Identification of Infigratinib as a Potent Reversible Inhibitor and Mechanism-Based Inactivator of CYP2J2: Nascent Evidence for a Potential In Vivo Metabolic Drug-Drug Interaction with Rivaroxaban

Abstract

Infigratinib (INF) is a fibroblast growth factor receptor inhibitor that was recently United States Food and Drug Administration-approved for the treatment of advanced or metastatic cholangiocarcinoma. We previously established that INF inhibited and inactivated cytochrome P450 3A4 (CYP3A4). Here, in a follow up to our previous study, we identified for the first time that INF also elicited potent competitive inhibition and mechanism-based inactivation of CYP2J2 with kinetic parameters K_i, K_I, k_inact, and a partition ratio of 1.94 µM, 0.10 µM, 0.026 minute⁻¹, and ∼3, respectively, when rivaroxaban was harnessed as the probe substrate. Inactivation was revealed to exhibit cofactor-dependency and was attenuated by an alternative substrate (astemizole) and direct inhibitor (nilotinib) of CYP2J2. Additionally, the nature of inactivation was unlikely to be pseudo-irreversible and instead arose from covalent modification due to the lack of substantial enzyme activity recovery after dialysis and chemical oxidation, as well as the lack of a resolvable Soret band in spectral scans. Glutathione trapping confirmed that the identity of the putative reactive intermediate implicated in the covalent inactivation of both CYP2J2 and CYP3A4 was identical and likely attributable to an electrophilic p-benzoquinonediimine intermediate of INF. Finally, mechanistic static modeling revealed that by integrating the previously arcane inhibition and inactivation kinetic parameters of CYP2J2-mediated rivaroxaban hydroxylation by INF illuminated in this work, together with those previously documented for CYP3A4, a 49% increase in the systemic exposure of rivaroxaban was projected. Our modeling results predicted a potential risk of metabolic drug-drug interactions between the clinically relevant combination of rivaroxaban and INF in the setting of cancer.

SIGNIFICANCE STATEMENT

This study reported that INF elicits potent reversible inhibition and mechanism-based inactivation of CYP2J2. Furthermore, static modelling predicted that its coadministration with the direct oral anticoagulant rivaroxaban may potentially culminate in a metabolic drug-drug interaction (DDI) leading to an increased risk of major bleeding. As rivaroxaban is steadily gaining prominence as the anticoagulant of choice in the treatment of cancer-associated venous thromboembolism, the DDI projections reported here are clinically relevant and warrant further investigation via physiologically based pharmacokinetic modelling and simulation.

Introduction

The fibroblast growth factor receptors (FGFRs) are a family of four ligand-dependent transmembrane receptor tyrosine kinases that mediate signal transduction of many important cellular processes ranging from differentiation to survival (Powers et al., 2000). Consequently, genetic mutations in FGFRs could result in their deranged constitutive activation and confer unbridled oncogenic properties to malignant cells (Touat et al., 2015). This explains why pharmacological repression of the FGFR signaling axis is steadily gaining interest as a nascent therapeutic paradigm in cancer (Porta et al., 2017). Currently, there are a total of three FGFR inhibitors that have been approved by the United States Food and Drug Administration (FDA) (Chakrabarti et al., 2022) (Supplemental Fig. 1, A–C). Infigratinib (INF) represents the newest of the three and was approved in 2021 for the treatment of locally advanced or metastatic cholangiocarcinoma harboring FGFR2-activating mutations. Despite being a relatively new class of drugs, the use of INF and other FGFR inhibitors is expected to soar in the near future as more refractory cancers have been identified to be closely intertwined with FGFR genomic aberrations (Weaver and Bossaer, 2021).

Apart from the constellation of debilitating symptoms intrinsic to malignancy, a highly prevalent complication afflicting cancer patients is that of cancer-associated venous thromboembolism (CA-VTE), which encompasses both deep vein thrombosis and pulmonary embolism (Timp et al., 2013). Unsurprisingly, CA-VTE is reported to be one of the most common causes of death in cancer, ranking only second to disease progression (Khorana et al., 2007). As such, it is imperative to manage this comorbidity to optimize treatment outcomes. At present, the factor Xa direct oral anticoagulant rivaroxaban is increasingly being used in the primary prevention or treatment of CA-VTE over conventional parenteral anticoagulants and vitamin K antagonists due to its linear pharmacokinetics, predictable dose-response relationship, and the absence of major dietary effects, thereby obviating the need for routine laboratory monitoring (Mueck et al., 2014). Unfortunately, one major limitation hindering its widespread clinical adoption is the well founded concern pertaining to its potential for pharmacokinetic drug-drug interactions (DDI) (Grillo et al., 2012). In broad strokes, such an interaction typically arises from the interference of its systemic clearance pathways by another drug that is concomitantly administered, thereby resulting in its accumulation throughout the body, which may precipitate dose-dependent toxicities. Given that rivaroxaban therapy is associated with the presence of narrow therapeutic indices with steep dose-exposure relationship toward major hemorrhagic events (Ismail et al., 2018), it is therefore germane to prospectively identify clinical DDI scenarios that could detrimentally augment rivaroxaban exposure and pose a bleeding risk.

In that regard, previous studies have revealed that approximately two-thirds of the administered rivaroxaban dose undergo hepatic metabolism with major contributions by cytochrome P450 (P450) 3A4 (CYP3A4) and 2J2 (CYP2J2) enzymes (Weinz et al., 2009). Collectively, the P450s constitute a ubiquitous class of hemeproteins that serve as one of the major drivers of oxidative metabolism in the human body (Guengerich, 2001). However, the catalytic activities of P450 may be directly and/or indirectly suppressed by a perpetrator drug molecule. In the former, the perpetrator noncovalently binds to the active and/or allosteric site of the P450 isoform and reversibly inhibits its ability to metabolize another substrate. Whereas, in the latter, the perpetrator drug is first metabolically activated by the enzyme into a chemically reactive intermediate that covalently alkylates the P450 apoprotein and/or its prosthetic heme or forms a coordination complex with the heme catalytic ferrous (Fe²⁺)_, thereby engendering a distinctive time-dependent loss of its activity via a phenomenon known as mechanism-based inactivation (MBI) (Orr et al., 2012). In contrast to reversible inhibition, the loss of activity evoked by MBI is irreversible, as the enzyme is irrevocably destroyed by covalent modification and can only be restored by de novo protein synthesis (Ho et al., 2015). Consequently, the likelihood and magnitude of DDI elicited by an MBI is usually more profound than with a direct reversible inhibitor (Bjornsson et al., 2003).

In our previous study, we established for the first time that INF reversibly inhibited and irreversibly inactivated CYP3A4-mediated rivaroxaban hydroxylation in a manner consistent with MBI (Tang et al., 2021b). However, it remained nebulous as to whether INF could also evoke inhibition and/or inactivation of CYP2J2-mediated metabolism of rivaroxaban. Here, we unraveled that INF is a novel and potent competitive inhibitor and covalent MBI of CYP2J2. The subsequent integration of CYP2J2 and CYP3A4 reversible inhibition and time-dependent inhibition (TDI) kinetic parameters into a mechanistic static model predicted a clinically significant in vivo metabolic DDI with rivaroxaban that was projected to culminate in an increased risk of major bleeding.

Materials and Methods

Chemicals and Reagents

Erdafitinib (purity 99.66%), INF (purity 99.70%), and pemigatinib (purity 99.88%) were acquired from MedChem Express (Monmouth Junction, NJ). Rivaroxaban (purity ≥ 98.0%) was obtained from Carbosynth (Berkshire, UK). Astemizole, catalase, dexamethasone, glutathione (GSH), and nilotinib were procured from Sigma-Aldrich (St. Louis, MO). Potassium ferricyanide was bought from VWR International (Leuven, Belgium). Human recombinant P450 2J2 Supersomes (rhCYP2J2) coexpressing both NADPH P450 oxidoreductase and cytochrome b₅ and the NADPH regenerating system comprising NADP+ and glucose-6-phosphate (G6P) (NADPH A) and glucose-6-phosphate dehydrogenase (G6PDH) (NADPH B) were commercially purchased from Corning Gentest (Woburn, MA). Ultrapure water (type I) was prepared in-house using a Milli-Q water purification system (Millipore Corporation, Bedford, MA). All available chemicals were of analytical or high-performance liquid chromatography grade.

CYP2J2 TDI Assay

TDI assays were conducted as described in our previous works with minor modifications (Tang et al., 2021a, 2021b, 2022). The factor X_a direct anticoagulant rivaroxaban was used as a clinically relevant probe of CYP2J2. Foremost, a CYP2J2 screening assay involving the three FDA-approved FGFR inhibitors was conducted. Briefly, primary incubation mixtures (n = 3) comprising 20 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and either erdafitinib, pemigatinib, or INF (each at 10 µM) were prepared and equilibrated at 37°C for 5 minutes with constant gentle stirring. After that, the reaction was initiated via the addition of NADP+/G6P (final NADPH concentration 1 mM). Subsequently, at various preincubation intervals (i.e., 0, 3, 8, 15, 22, and 30 minutes), aliquots of each primary incubation mixture were diluted 20-fold into a standard TDI secondary incubation mixture that consisted of the CYP2J2 probe substrate rivaroxaban (50 µM; representing the concentration at ∼4× K_m), an NADPH regenerating system (1 mM), and potassium phosphate buffer (0.1 M; pH 7.4). The secondary incubation mixtures were then further incubated for 30 minutes at 37°C. After that, aliquots of the secondary incubation mixture were withdrawn and added to an equal volume of ice-cold acetonitrile spiked with internal standard (4 µM dexamethasone) to stop the reaction and induce protein precipitation. The quenched samples were then centrifuged at 4000g for 30 minutes at 4°C and the resultant supernatants were transferred to a 96-well microplate for ultra-high performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) quantification of hydroxylated rivaroxaban, which serves as a surrogate for the residual CYP2J2 activity (Supplemental Methods). Vehicles controls were performed by the addition of DMSO in place of the FGFR inhibitor. Subsequently, as only INF was determined to elicit TDI of CYP2J2, the TDI assay was repeated using multiple concentrations of INF (0, 0.1, 0.25, 1, 2.5, and 5 µM). Finally, to further interrogate the role of NADPH cofactor in enzymatic inactivation, parallel experiments were performed by substituting NADP+/G6P with potassium phosphate buffer (0.1 M; pH 7.4).

Calculation of TDI Kinetic Parameters (K_I and k_inact)

The TDI kinetic parameters that consisted of the inactivator concentration at half-maximum inactivation rate constant (K_I) and the maximum inactivation rate constant (k_inact) were calculated by first normalizing the averages of triplicate peak area ratios to vehicle at each respective preincubation timepoint to yield the percentage of CYP2J2 activity remaining after TDI by INF. Thereafter, the natural logarithm of this value was plotted against preincubation time for each concentration of INF. The resulting data points were fitted to linear regression and the observed first-order inactivation rate constant (k_obs) of INF against CYP2J2-mediated rivaroxaban hydroxylation was obtained from the slope of the initial linear decline in activity. Here, data points from the preincubation timepoints 0 to 22 minutes of each INF concentration were harnessed in the calculation of its k_obs due to a deviation from linearity occurring at the 30-minute mark. After that, k_obs was plotted against INF concentrations [I] for the determination of its TDI kinetic parameters (K_I and k_inact) using nonlinear least square regression based on eq. 1 in GraphPad 8.0.2 (San Diego, CA) (Kitz and Wilson, 1962).

Eq. 1 assumes that the change of [I] during the preincubation period is negligible and that the loss of enzymatic activity purely stems from inactivation by INF. The k_inact/K_I ratio was determined by dividing the mean values of k_inact by K_I.

Substrate Depletion of INF by CYP2J2

Incubation mixtures (n = 3) comprising 10 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 µM) were prepared, equilibrated, and initiated with NADP+/G6P in the same manner described earlier. The reaction was allowed to proceed at 37°C. At several predefined intervals (0, 2.5, 5, 7.5, 10, 15, 30, 45, 60, and 90 minutes), aliquots of each incubation mixture were withdrawn and added to an equal volume of ice-cold acetonitrile spiked with internal standard (0.1 µM erdafitinib) to stop the reaction. The samples were then centrifuged and assayed for the amount of unchanged INF remaining by UPLC-MS/MS as described in detail in our previous study (Tang et al., 2021b).

Partition Ratio of the Inactivation of CYP2J2 by INF

Incubation mixtures (n = 3) comprising 50 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (0–5 µM) were prepared, equilibrated, and initiated with NADP+/G6P in the same manner described earlier. The reaction mixtures were incubated for a protracted duration of 45 minutes to allow CYP2J2 inactivation to go into completion. Thereafter, aliquots of each primary incubation mixture were withdrawn and diluted 20-fold into the TDI secondary incubation mixture and incubated at 37°C for a further 30 minutes. Samples were then quenched, centrifuged, and assayed for residual CYP2J2 enzyme activity by UPLC-MS/MS (Supplemental Methods). The turnover number and partition ratio were computed as outlined in our previous study (Tang et al., 2021c).

Effect of an Alternative Substrate and a Direct Inhibitor on Inactivation

To investigate whether the presence of a competing substrate could protect CYP2J2 from inactivation by INF, excessive amounts of the alternative CYP2J2 substrate astemizole (Matsumoto et al., 2002) (at 2 and 8 µM; equivalent to 1:2 and 1:8 molar ratio of INF to astemizole) and the potent direct inhibitor nilotinib [at 0.1 and 1 µM; corresponding to ∼1× and ∼10× its K_i value (Cheong et al., 2022)] were individually incorporated into the primary incubation mixtures (n = 3) that comprised 20 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 µM). After that, aliquots of each primary incubation mixture were diluted 20-fold into the TDI secondary incubation mixture at several predefined preincubation intervals (i.e., 0, 3, 8, and 15 minutes). After an additional 30 minutes of incubation at 37°C, the samples were quenched, centrifuged, and assayed for residual CYP2J2 enzyme activity by UPLC-MS/MS (Supplemental Methods). Primary incubation mixtures that omitted the inclusion of either astemizole/nilotinib or both INF and astemizole/nilotinib served as the negative and vehicle controls, respectively.

Effect of an Exogenous Nucleophile and a Scavenger of Reactive Oxygen Species on Inactivation

Primary incubation mixtures (n = 3) that comprised 20 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 µM) were enriched with either GSH (2 mM) or catalase (800 U/ml). After prewarming at 37°C for 5 minutes, the enzymatic reaction was initiated via the addition of NADP+/G6P. After that, aliquots of each primary incubation mixture were diluted 20-fold into the TDI secondary incubation mixture at several predefined preincubation intervals (i.e., 0, 3, 8, and 15 minutes). After an additional 30 minutes of incubation at 37°C, the samples were quenched, centrifuged, and assayed for residual CYP2J2 enzyme activity by UPLC-MS/MS (Supplemental Methods). Primary incubation mixtures that did not contain either GSH/catalase or both INF and GSH/catalase served as the negative and vehicle controls.

Effect of Dialysis on the Reversibility of Inactivation

To determine if the inactivation of CYP2J2 elicited by INF was reversible in vitro, we conducted an array of three complementary assays as described in many of our previous works (Tang et al., 2021a, 2021b, 2022). In the first series of experiments involving dialysis, primary incubation mixtures that comprised 20 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 µM) were prepared, equilibrated, and initiated with NADP+/G6P. The enzymatic reaction was allowed to proceed for 30 minutes. After that, a 5 μl aliquot was transferred to the TDI secondary incubation mixture yielding a 20-fold dilution. Concurrently, the remaining primary incubation mixture was transferred to a Slide-A-Lyzer MINI dialysis device (molecular weight cutoff of 10,000; Pierce Chemical Co., Rockford, IL) and floated in a glass beaker containing 0.5 l of ice-cold potassium phosphate buffer (0.1 M; pH 7.4). The buffer system was maintained on ice with gentle stirring and was refreshed at the second hour. After 4 hours, a 5 μl aliquot of the dialyzed mixture was withdrawn from the dialysis device and transferred to the respective TDI secondary incubation mixture as described above. All TDI secondary incubation mixtures were further incubated at 37°C for 30 minutes and subsequently quenched, centrifuged, and assayed for residual CYP2J2 enzymatic activity by UPLC-MS/MS (Supplemental Methods). Vehicle control was performed by substituting INF with DMSO in the primary incubations. Three independent experiments were conducted, each in triplicates.

Effect of Chemical Oxidation on the Reversibility of Inactivation

In the second series of experiments involving chemical oxidation with the strong oxidizing agent potassium ferricyanide, three sequential incubations were performed. Foremost, primary incubation mixtures (n = 3) that consisted of 20 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 μM) were prepared, equilibrated, and initiated with NADP+/G6P. After the incubation at 37°C for either 0 or 30 minutes, 20 µl of the primary incubation mixture were aliquoted into an equal volume of secondary incubation mixture containing potassium phosphate buffer (0.1 M; pH 7.4) formulated with or without potassium ferricyanide (2 mM). The secondary mixtures were then allowed to incubate at 37°C for another 10 minutes. Thereafter, 10 µl of the mixture was withdrawn and diluted 10-fold into a tertiary incubation mixture containing rivaroxaban (50 µM), an NADPH regenerating system (1 mM), and potassium phosphate buffer (0.1 M; pH 7.4). The reaction mixture was further incubated at 37°C for another 30 minutes and subsequently assayed for residual CYP2J2 activity by UPLC-MS/MS (Supplemental Methods). Vehicle control was performed by substituting INF with DMSO in the primary incubations. Three independent experiments were conducted, each in triplicates. The percentage of CYP2J2 metabolic activity remaining after 0 or 30 minutes of incubation with INF compared with the corresponding vehicle controls was calculated using eq. 2 and 3, respectively.

Spectral Difference Scanning

In the third series of experiments involving spectral difference scanning, incubation mixtures containing 200 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 µM) were prepared, prewarmed at 37°C for 5 minutes, and initiated via the addition of NADP+/G6P. After that, it was immediately scanned from 400 to 500 nm at 5-minute intervals over a 1-hour duration using a Hidex Sense microplate reader (Hidex, Turku, Finland) maintained at a constant 37°C. The spectral differences were obtained by comparing the UV absorbances between the sample and reference wells that consisted of vehicle in place of INF. Additionally, the degree of metabolite-intermediate complex (MIC) formation was also semiquantitatively assessed by measuring the absorbance difference between 454 nm and the isosbestic point at 490 nm with time.

GSH Trapping Assay

GSH trapping experiments of INF were performed as previously described with some minor modifications (Tang et al., 2021b). Briefly, incubation mixtures (500 µl) containing 50 pmol/ml rhCYP2J2, G6PDH, potassium phosphate buffer (0.1 M; pH 7.4), and INF (1 µM) were fortified with saturating concentrations of GSH (50 mM) and equilibrated at 37°C for 5 minutes. The reaction was then initiated via the addition of NADP+/G6P and incubated at 37°C for 1 hour. Thereafter, an equal volume of ice-cold acetonitrile was added to quench the reaction. The resulting mixture was centrifuged at 14,000g for 15 minutes at 4°C. After that, the supernatant was decanted to a new microcentrifuge tube and dried under a gentle stream of nitrogen gas (TurboVap LV; Caliper Life Science, Hopkinton, MA). The residue was subsequently reconstituted with 60 µl of acetonitrile-water mixture (3:7), vortexed, and centrifuged at 14,000g for another 15 minutes at 4°C. Finally, the supernatant was carefully removed and transferred to a vial for UPLC-MS/MS analysis. Samples that omitted the inclusion of INF in the incubation mixture served as the vehicle controls. Putative INF-derived GSH adducts formed in situ were detected via UPLC-MS/MS as previously described (Tang et al., 2021b) and is expounded in greater detail in the Supplemental Methods.

CYP2J2 Reversible Inhibition Assay

Incubation mixtures (n = 3) comprising 10 pmol/ml rhCYP2J2, G6PDH, rivaroxaban (10 µM), potassium phosphate buffer (0.1 M; pH 7.4), and either erdafitinib, pemigatinib, or INF (each at 10 µM) were prepared, equilibrated, and initiated with NADP+/G6P. The reaction was allowed to proceed at 37°C for 30 minutes. After that, aliquots of the samples were quenched, centrifuged, and assayed for residual CYP2J2 enzyme activity by UPLC-MS/MS. Vehicle controls was performed by substituting the respective FGFR inhibitors with DMSO. After that, as only INF was determined to elicit reversible inhibition of CYP2J2, dose-response and steady-state kinetic experiments were performed in a similar manner described above with minor modifications. Briefly, the incubation mixtures (n = 3) for dose-response experiments comprised 10 pmol/ml rhCYP2J2, G6PDH, rivaroxaban (10 µM), potassium phosphate buffer (0.1 M; pH 7.4), and INF (0–50 µM), whereas incubation mixtures (n = 3) for steady-state kinetic experiments consisted of 10 pmol/ml rCYP2J2, G6PDH, rivaroxaban (2.5, 5, 15, and 30 µM – concentrations spanning its K_m), potassium phosphate buffer (0.1 M; pH 7.4), and INF (0, 1, 3, and 10 µM).

Calculation of Reversible Inhibition Kinetic Parameters (IC₅₀ and K_i) and Determination of Modes of Inhibition

The half-maximal inhibitory concentration (IC₅₀) was determined from dose-response experiments using the log(inhibitor) versus response-variable slope (four parameters) model based on eq. 4 in GraphPad 8.0.2 (San Diego, CA).

where E₀ is the minimum effect, E_max is the maximum effect, [I] is the in vitro concentration of the reversible inhibitor, Hill slope is the Hill coefficient, and Y is the % enzyme activity compared to control.

The equilibrium dissociation constant for the enzyme-inhibitor complex (K_i) was derived by nonlinear least-square regression analysis of hydroxylated rivaroxaban formation data (expressed in peak area ratio) collected at various rivaroxaban and INF concentrations using eq. 5 for competitive inhibition in GraphPad 8.0.2 (San Diego, CA). The mode of reversible inhibition was determined based on statistical evaluation of the Michaelis-Menten plots by corrected Akaike’s information criterion and further verified via visual inspection of the transformed Lineweaver-Burk and Dixon plots.

where v is the rate of enzyme activity, V_max (maximum rate of reaction) and K_m (Michaelis constant) are kinetic constants for substrate metabolism, [S] is the in vitro concentration of substrate, [I] is the in vitro concentration of inhibitor, and K_i is the equilibrium dissociation constant for the enzyme-inhibitor complex.

Prediction of the Extent of Metabolic DDI Using a Mechanistic Static Model

Reversible inhibition (i.e., K_i) and TDI (i.e., K_I and k_inact) kinetic parameters of INF against CYP2J2-mediated rivaroxaban hydroxylation determined in this work and the corresponding values for CYP3A4 that were derived in our previous study (Tang et al., 2021b) were incorporated into a mechanistic static model developed previously by Fahmi et al. (2008) and later further refined by Isoherranen et al. (2012), which takes into consideration the net effects of reversible and TDI of P450 in both the liver and gut to predict the likelihood and severity of metabolic DDI. Here, the area under the curve ratio (AUCR) refers to the ratio of the systemic exposure of rivaroxaban in the presence of INF to that in the absence of INF and is described by eq. 6.

where A describes TDI of either CYP3A4 or CYP2J2 observed in the liver,

B describes reversible inhibition of either CYP3A4 or CYP2J2 observed in the liver,

X describes TDI of CYP3A4 observed in the gut, and

Y describes reversible inhibition of CYP3A4 observed in the gut.

[I]_H represents the in vivo concentrations of INF available to the P450 enzymes in the liver and was designated as the maximum unbound steady-state plasma concentration of the inactivator ([I]_max,u) and the maximum unbound concentration of inhibitor at the inlet to the liver ([I]_inlet,max,u) to predict metabolic DDIs precipitated by TDI (term A) and reversible inhibition (term B), respectively (Cheong et al., 2017). The in vivo concentrations for [I]_H in term A and B were chosen based on the respective estimates that were previously, retrospectively validated to yield the best correlation between predicted and observed DDIs reported in literature (Fahmi et al., 2009) and were calculated to be 0.016 and 0.517 µM, respectively (Supplemental Table 1). [I]_G represents the in vivo concentrations of INF available to CYP3A4 in the gut and was estimated to be 2.358 µM (Supplemental Table 1). The degradation rate constants for CYP3A4 in the liver (k_deg,H) and intestine (k_deg,G) were 0.00032 and 0.00048 minute⁻¹ based on a half-life of 36 and 24 hours, respectively (Fahmi et al., 2008). Conversely, due to the paucity of clinical pharmacokinetic data available for CYP2J2 to perform such similar calculations, the means of the k_deg estimates for the various hepatic P450 isoforms (0.00026 minute⁻¹) was adopted instead (Yang et al., 2008). The fraction of rivaroxaban metabolized by CYP3A4 (f_m,CYP3A4) and CYP2J2 (f_m,CYP2J2) was reported to be 0.18 and 0.14, respectively (Mueck et al., 2014), whereas the fraction of rivaroxaban escaping intestinal metabolism (F_G) was previously calculated by our group to be 0.89 (Cheong et al., 2017).

After that, the calculated AUCR values were incorporated into eq. 7 (Ismail et al., 2018) to predict the risk of major bleeding resulting from the augmented systemic exposure of rivaroxaban arising due to the inhibition and inactivation of its hepatic and gut metabolism by INF.

where AUC_ss (3.146 mg.h/l) is the mean steady-state plasma exposure of rivaroban determined from population pharmacokinetic modeling of nonvalvular atrial fibrillation patients with normal renal function (i.e., creatinine clearance > 50 ml/min) and receiving 20 mg once daily of rivaroxaban (Mueck et al., 2014). This particular population was adopted as a surrogate for patients receiving rivaroxaban for the prophylaxis or treatment of CA-VTE due to the same maintenance dose utilized (i.e., 20 mg once daily) and also because, like cancer, atrial fibrillation tends to be more common in older patients (Mantha et al., 2017). Notably, the baseline risk of major bleeding in this group of patients – assuming an AUCR of 1 – was computed to be 4.31%.

Results

Time-, Concentration-, and Cofactor-Dependent Inactivation of CYP2J2 by INF

Although we previously reported that all three FDA-approved FGFR inhibitors (i.e., erdafitinib, pemigatinib, and INF) elicited TDI of CYP3A4 and culminated in their MBI (Tang et al., 2021a, 2021b, 2022), findings from our preliminary screen revealed that only INF exhibited appreciable TDI of CYP2J2 (Fig. 1A). This is evident from the more profound loss of residual CYP2J2 activity occurring at longer preincubation durations. Armed with these preliminary findings, we proceeded with a more comprehensive investigation of its TDI kinetics using six different INF concentrations that subsequently confirmed that INF inactivated CYP2J2-mediated rivaroxaban hydroxylation in both a time- and concentration-dependent manner (Fig. 1B). Furthermore, as the k_obs values yielded seemed to tend toward a maximum rate, it implied that the inactivation elicited by INF against CYP2J2 is saturable and exhibited pseudo-first-order kinetics (Fig. 1C). As a result, a Kitz-Wilson plot was constructed that determined that the k_inact and K_I values corresponded to 0.026 ± 0.001 minute⁻¹ and 0.10 ± 0.02 µM, respectively, which translated to a k_inact/K_I ratio of 260.0 minute⁻¹mM⁻¹ (Table 1). Lastly, the omission of the P450 cofactor NADPH was also found to completely nullify the loss of rivaroxaban hydroxylase activity (Fig. 1D).

Fig. 1.

(A) Preliminary inhibition screen revealed only INF elicited time-dependent inhibition of CYP2J2. (B) Time- and concentration-dependent inactivation of CYP2J2 by INF using rivaroxaban as probe substrate. (C) Nonlinear least square regression of observed first-order inactivation rate constants (k_obs) versus INF concentration yielded k_inact and K_I values of 0.026 ± 0.001 minute⁻¹ and 0.10 ± 0.02 µM. (D) Cofactor NADPH-dependent inactivation of CYP2J2 by INF. Each point in (A–D) represents the mean and S.D. of triplicate experiments.

TABLE 1.

CYP2J2 and CYP3A4 TDI and reversible kinetic parameters for INF derived using morpholinone hydroxylation of rivaroxaban as a surrogate marker of residual P450 activity

P450 Isoform	KI (µM)	kinact (min−1)	kinact/KI (min−1 mM−1)	Partition Ratio	Ki (µM)	Reference
CYP2J2	0.10 ± 0.02	0.026 ± 0.001	260.0	3	1.94 ± 0.31
CYP3A4	4.17 ± 0.93	0.068 ± 0.005	16.4	41	0.97 ± 0.06	(Tang et al., 2021b)

Data are presented as mean ± S.D.

Substrate Depletion of INF by CYP2J2

After that, substrate depletion studies were performed to ascertain if INF is a substrate of CYP2J2. Monitoring the depletion of 1 µM INF in rhCYP2J2 revealed that after a 90-minute incubation there was a reduction of INF to approximately half its initial levels (i.e., 50.0% ± 4.0%), hence confirming that INF was indeed metabolized by CYP2J2 (Fig. 2). This was further confirmed in incubation mixtures that lacked rhCYP2J2, in which levels of INF did not differ significantly after 90 minutes (data not shown). Interestingly, the substrate depletion plot also possessed a rather salient biphasic depletion profile that appeared to corroborate our earlier findings on the TDI of CYP2J2 (data not shown). This was made more apparent by applying a natural log transformation to the substrate depletion plot, which produced two distinct linear phases (i.e., spanning from 0–7.5 minutes and 10–90 minutes), from which we could derive two distinct elimination rate constants (termed k_fast and k_slow, respectively) from the gradient of the straight lines (Fig. 2). The values obtained for k_fast and k_slow were calculated to be 0.0360 ± 0.0055 minute⁻¹ and 0.0050 ± 0.0003 minute⁻¹, respectively. Taken together, it further substantiated that TDI (or inactivation) of the enzyme had occurred due to the ∼7.2-fold reduction in its elimination rate constant.

Fig. 2.

Substrate depletion of INF by CYP2J2. Percentage of INF remaining against time in the presence of CYP2J2 plotted on a seminatural logarithmic scale. Each point represents the mean and S.D. of triplicate experiments.

Partition Ratio of the Inactivation of CYP2J2 by INF

The partition ratio is an estimate of the number of inactivator molecules that are metabolized and liberated from the enzymatic site without causing inactivation relative to each molecule of the enzyme that gets inactivated (Orr et al., 2012). Adopting a previously described titration method (Silverman, 1995), the turnover number for the inactivation of CYP2J2 by INF was found to be ∼4, which in turn corresponded to a partition ratio of ∼3 (Fig. 3A; Table 1).

Fig. 3.

(A) Partition ratio for the inactivation of CYP2J2 by INF, determined by extrapolating the intercept of the linear regression line at lower ratios and the straight line for the high ratios to the x-axis, was estimated to be 3. Inactivation of CYP2J2 was attenuated in the presence of (B), an alternative CYP2J2 substrate (astemizole), and (C) a direct CYP2J2 inhibitor (nilotinib). (D) Conversely, the presence of either an exogenous nucleophile (GSH) or a scavenger of reactive oxygen species (catalase) did not protect against enzymatic inactivation. Each point in (A–D) represents the mean and S.D. of triplicate experiments.

Effect of an Alternative Substrate and a Direct Inhibitor on Inactivation

The inactivation of CYP2J2 by INF was markedly attenuated in a dose-dependent fashion by the incorporation of both astemizole and nilotinib (Fig. 3, B and C). Notably, the parallel downward reduction in enzyme activity observed across all preincubation timepoints in the presence of 1 µM nilotinib may be ascribed to the potent reversible inhibition elicited against CYP2J2 by nilotinib.

Effect of an Exogenous Nucleophile and a Scavenger of Reactive Oxygen Species on Inactivation

On the contrary, the inclusion of GSH and catalase did not offer any appreciable protection from inactivation, as evident from the similar reduction in enzymatic activity in all incubations, regardless of whether GSH or catalase was present (Fig. 3D).

Effect of Dialysis on the Reversibility of Inactivation

Dialysis, chemical oxidation, and spectral difference scanning were conducted to mechanistically elucidate the underpinning nature and modality of inactivation of CYP2J2 by INF. The results of the first series of experiments revealed that dialysis for 4 hours at 4°C did not restore enzyme activity (Fig. 4A). Rather, the percentage of CYP2J2 activity remaining paradoxically declined after dialysis. This anomaly may be rationalized by enzymatic degradation that occurred during the course of dialysis, as corroborated by the similar post dialysis decrease observed in the vehicle control arm.

Fig. 4.

(A) Percentage activity of CYP2J2 remaining did not increase after extensive dialysis at 4°C for 4 hours. (B) Similarly, potassium ferricyanide (KFC) did not significantly (P value = 0.947) restore the metabolic activity of CYP2J2 after a 30-minute incubation with 1 µM INF. (C) Spectral difference measured over 60 minutes failed to elicit a Soret band in the absorbance ranges of 448–458 nm for CYP2J2 incubated with 1 µM INF. (D) Similarly, a comparison of the absorbance at the reference of 454 nm against the isosbestic point at 490 nm failed to demonstrate an increase in the extent of MIC formation over time. Results from (A and B) depict the mean and S.D. of three independent experiments conducted in triplicates.

Effect of Chemical Oxidation on the Reversibility of Inactivation

In the second series of experiments involving chemical oxidation, we determined that the mean CYP2J2 activity in the 30-minute preincubation samples that comprised potassium ferricyanide was not significantly different (P = 0.947) from its counterpart, which did not contain potassium ferricyanide (Fig. 4B), using an unpaired t test.

Spectral Difference Scanning

Lastly, in the third series of experiments, the incubation of INF with CYP2J2 failed to produce an absorbance maxima in the Soret region (448–458 nm) (Fig. 4C). The lack of this spectrally detectable peak (also termed as a Soret band) is further ascertained by tracking the increase in absorbance between 454 nm and the isosbestic point at 490 nm against time, which failed to exhibit a parabolic increase (Fig. 4D). The results of these three assays were in concordance and collectively asserted that the inactivation of CYP2J2 by INF was not reversible in vitro and was unlikely to have arisen via the formation of pseudo-irreversible MIC.

GSH Trapping

Consequently, a GSH trapping assay was conducted to uncover reactive intermediates of INF that could have engendered covalent inactivation of CYP2J2. This was achieved by enriching incubation mixtures with the nucleophilic scavenger GSH, which traps the labile electrophilic species generated via metabolic activation as stable adducts and facilitates their detection and eventual structure elucidation via the application of two UPLC-MS/MS survey scans. As anticipated, negative precursor ion scan at m/z 272 and positive ion constant neutral loss monitoring of 129 Da evinced one peak suggestive of an INF-derived GSH adduct termed INF-G1 (retention time: 6.32 minutes) with [M+H]⁺ ion at m/z of 865.3 that was absent in vehicle controls (data not shown) and only detected in incubation mixtures that contained INF (Fig. 5, A and B). The resultant MS/MS spectrum of INF-G1 acquired by the product ion scan of m/z 865.3 in positive ion mode confirmed that it belonged to a GSH conjugate due to its characteristic collision-induced dissociation fragmentation pattern yielding neutral mass loss of 129 Da (corresponding to the loss of a pyroglutamate moiety) (Fig. 5C).

Fig. 5.

(A) Total ion chromatogram for precursor ion scan of m/z 272 in negative electrospray mode and (B) constant neutral loss scan of 129 Da in positive electrospray mode of INF incubations in the GSH trapping assay. Notably, one peak (retention time: 6.32 minutes) corresponding to the INF-derived GSH adduct (INF-G1) was detected in both survey scans. (C) Representative product ion MS/MS spectra and proposed mass fragmentation pattern of the INF-derived GSH adduct (INF-G1) (m/z 865.3) formed in situ in the GSH trapping assay.

Reversible Inhibition of CYP2J2 by INF

Apart from evoking TDI, we were also curious to interrogate whether INF elicited any reversible inhibition of CYP2J2. Using one-way ANOVA followed by Dunnett’s post hoc test, findings from our preliminary screen revealed that only INF exhibited significant (P < 0.05) reversible inhibition of CYP2J2 (Fig. 6A). After that, we demonstrated that INF inhibited CYP2J2-mediated rivaroxaban hydroxylation in a concentration-dependent manner and yielded a sigmoidal-shaped dose-response curve with an experimentally derived IC₅₀ value of 2.38 ± 0.31 µM (Fig. 6B). Finally, nonlinear least-square regression analysis and visual inspection of the transformed Lineweaver-Burk plot allowed us to delineate that INF competitively inhibited CYP2J2-mediated rivaroxaban hydroxylation with a K_i value of 1.94 ± 0.31 µM (Fig. 6, C and D; Table 1).

Fig. 6.

(A) Preliminary inhibition screen revealed that only INF elicited significant (P < 0.05) direct inhibition of CYP2J2. (B) Dose-response curves depicting direct inhibition elicited by INF against CYP2J2-mediated rivaroxaban hydroxylation with an IC₅₀ value of 2.38 ± 0.31 µM. (C) The formation rate of hydroxylated rivaroxaban (expressed in peak area ratio) was plotted against various INF concentrations and fitted in the Michaelis-Menten kinetic model, which determined K_i to be 1.94 ± 0.31 µM. (D) Lineweaver-Burk plot revealed that the mode of direct inhibition elicited by INF against CYP2J2 was competitive.

Prediction of the Extent of Metabolic DDI Using a Mechanistic Static Model

The in vitro reversible inhibition (i.e., K_i) and TDI (i.e., K_I and k_inact) kinetic parameters of INF against CYP2J2- and CYP3A4-mediated rivaroxaban hydroxylation were integrated into a mechanistic static model that quantitatively predicts the combined impact of P450 inhibition and inactivation (TDI) elicited by INF on the systemic exposures of rivaroxaban, which can serve to inform the potential likelihood and severity of metabolic DDI. Our static modeling projections revealed that by taking into account only the inhibition of hepatic and gut CYP3A4, it resulted in an elevated AUCR of 1.27, which corresponded to a 5.46% increase in the risk of major bleeding associated with rivaroxaban (Table 2). However, when hepatic CYP2J2 inhibition and inactivation kinetic parameters illuminated in this present study were woven into the static model, it predicted that the AUCR and risk of major bleeding were further augmented to 1.49% and 6.63%, respectively (Table 2).

TABLE 2.

Prediction of the AUCR of rivaroxaban and the corresponding risk of major bleeding arising from combined impact of reversible inhibition and TDI elicited by INF

	No Inhibition	Inhibition of Hepatic and Gut CYP3A4	Inhibition of Hepatic and Gut CYP3A4 and Hepatic CYP2J2
Predicted AUCR	1.00	1.27	1.49
Risk of Major Bleeding	4.31%(baseline)	5.46%	6.63%

Discussion

INF is a potent and selective FGFR inhibitor that was recently FDA-approved for the treatment of advanced or metastatic cholangiocarcinoma harboring aberrant FGFR2 mutations. We previously elucidated that INF inhibited and inactivated CYP3A4. Here, in a follow up to our previous study, we identified for the first time that INF also elicits potent reversible inhibition and MBI of CYP2J2 and provided prospective pharmacokinetic modeling evidence for a potential in vivo metabolic DDI with the direct oral anticoagulant rivaroxaban.

In vitro reversible inhibition and TDI assays conducted for a particular perpetrator drug typically rely on a P450 isoform-selective probe substrate as a surrogate of enzyme activity. The corresponding kinetic parameters (i.e., K_i, K_I, and k_inact), which are reflective of the enzyme-drug binding affinities at the molecular level, can be quantitatively derived. Although these kinetic constants were traditionally regarded as intrinsic features of the perpetrator and were assumed to be independent of the probe substrate employed, such reductionistic assumptions are no longer tenable and only hold true when both the probe substrate and perpetrator are competing for the same binding site on the enzyme. Furthermore, as mechanistic investigations into the structural biology of CYP3A enzymes have revealed the existence of multiple ligand-binding sites that may facilitate different dynamic orientations of the perpetrator within its active site topology in a probe-substrate dependent manner (Ekroos and Sjögren, 2006), the biochemical interactions observed with one CYP3A probe substrate may not mirror those obtained with another (Galetin et al., 2005). This was corroborated in our recent study, where we demonstrated that INF elicited substantially more potent eversible inhibition and TDI of CYP3A4 when rivaroxaban was used as a probe substrate in place of its two FDA-recommended probes (i.e., testosterone and midazolam) (Tang et al., 2021b), thereby giving credence to the prevailing theory that inhibition and inactivation kinetic parameters exhibit probe substrate-dependency. Consequently, all reversible inhibition and TDI enzymatic assays involving CYP2J2 reported in this work were similarly conducted using rivaroxaban in place of its prototypical substrate astemizole to ensure the accuracy of the kinetic parameters obtained for subsequent prediction of its in vivo metabolic DDI risk via static modeling (Supplemental Methods; Supplemental Table 2).

Juxtaposing the reversible inhibition and TDI kinetic parameters obtained for CYP2J2 against those previously reported for CYP3A4 allowed us to discern that, whereas INF reversibly inhibited both P450 isoforms equipotently with submicromolar K_i values (Table 1), the k_inact/K_I ratio derived from CYP2J2 was ∼15.9-fold higher compared with CYP3A4 (Table 1). Furthermore, the partition ratio, which functions as a quantitative estimate of its inactivation efficiency, was also found to be vastly lower (∼13.7-fold) in CYP2J2 in comparison with CYP3A4 (Table 1). Consequently, our findings here suggested that the TDI of CYP2J2 by INF was immensely more potent and efficient than that of CYP3A4. We next sought to investigate if the underlying TDI observed in CYP2J2 could be alluded to MBI – a unique subset of TDI that fundamentally arises from the metabolic activation (or bioactivation) of the parent drug to a chemically reactive intermediate. Due to its unique biochemistry, an archetypal MBI is known to exhibit the following cardinal features in addition to its time-dependent hallmark: occurrence only in a catalytically competent system (i.e., presence of cofactors/coenzymes), saturable kinetics of inactivation, protection against inactivation by a competing substrate or inhibitor but not by nucleophilic scavengers, irreversibility of inactivation, and a 1:1 binding stoichiometry (Silverman, 1995). Our findings pertaining to the lack of inactivation in incubations mixtures deprived of the NADPH cofactor hinted that the metabolic activation of INF by CYP2J2 was a key molecular-initiating event that preceded the time-dependent loss of CYP2J2 activity. Moreover, whereas the ncorporation of GSH and catalase rendered no protection against inactivation, enriching the primary incubation mixtures with either astemizole or nilotinib greatly ameliorated the time-dependent loss of CYP2J2 evoked by INF. Taken together, these results precluded artifactual causes of inactivation and instead lent further support to the notion that enzymatic inactivation likely occurred within its active site and hence could be protected by a competing substrate and inhibitor of CYP2J2.

Next, an array of three complementary assays were conducted to delineate wheter the time-dependent loss of CYP2J2 activity arose pseudo-irreversibly via the formation of MIC or irreversibly by means of covalent modification. With regards to the former, MIC refers to the chelation of the heme catalytic ferrous by the chemically reactive intermediate through a strong coordinate covalent (dative) bond. Although these coordination complexes are extremely stable in vivo, they may be dissipated under certain in vitro conditions, which allows them to be experimentally differentiated from covalent modification in the latter. This is routinely accomplished via dialysis or chemical oxidation, which reverts the heme iron back to its reduced ferric ground state, dissociates it from the tight-binding complex with the inactivator, and restores its catalytic activity. Additionally, MIC is also associated with an absorbance maxima in the Soret region (448–458 nm); therefore, the detection of this spectral peak can serve to further corroborate the formation of MIC. On the contrary, covalent modification of the P450 apoprotein and/or heme moiety cannot be reversed by both aforementioned methodologies. Here, our findings asserted that the nature of inactivation of CYP2J2 by INF was unlikely to be pseudo-irreversible and instead stemmed from covalent modification by a putative reactive intermediate. This proposed nature of inactivation also aligns closely with those previously obtained for CYP3A4. Finally, we managed to evince one peak reminiscent of a prospective INF-derived adduct (termed INF-G1) in our GSH trapping assays. Intriguingly, a closer analysis of its product ion MS/MS spectra revealed that the fragmentation pattern of INF-G1 generated in rhCYP2J2 incubations was completely identical to those corresponding to the GSH adduct that we previously generated in rhCYP3A4 incubations (Tang et al., 2021b). This allowed us to deduce that the identity of the putative reactive intermediate of INF implicated in the covalent inactivation of both CYP2J2 and CYP3A4 were the same and likely arose via metabolic activation of INF to an electrophilic p-benzoquinonediimine intermediate (Fig. 7). At this juncture, previous homology modeling studies of CYP2J2 have revealed that although there were some slight geometrical differences, its active site cavity volume was similar to that of CYP3A4 (i.e., 1420 ų versus 1585 ų) (Lee et al., 2010). This explained why there were broad overlaps in substrate recognition and metabolism between CYP2J2 and CYP3A4. Consequently, it is unsurprising that INF could be metabolically activated to the p-benzoquinonediimine intermediate by both P450 isoforms at the same bioactivation “hot-spot”. Along this tangent, it may also denote that a more nuanced explanation underpinning the salient discrepancies in their inactivation potencies and efficiencies could be attributed to their underlying active site architectural difference, which may facilitate the formation of the reactive species and/or the eventual covalent adduction to its enzymatic substructural target to a greater extent in CYP2J2 than in CYP3A4. However, these postulations have to be further validated in future in silico covalent docking or molecular dynamics simulations (Tang et al., 2021d).

Fig. 7.

Proposed bioactivation pathway of INF by CYP2J2.

Our static modeling projections revealed that by taking into account the previously arcane reversible inhibition and TDI of CYP2J2-mediated rivaroxaban hydroxylation by INF illuminated in this work, the predicted AUCR and risk of major bleeding was exacerbated by a respective 17% and 21% as compared with the inhibition of hepatic and gut CYP3A4 alone. These nascent, albeit untested, modeling results highlight a potential risk of metabolic DDI between the clinically relevant combination of rivaroxaban and INF in the disease setting of cancer. However, one caveat of the DDI predictions gleaned from our modeling was that they were based on static estimates of plasma concentration (Supplemental Fig. 2; Supplemental Table 3). This arises due to an intrinsic limitation of the mechanistic static model, which does not take into consideration dynamic fluxes and interindividual variabilities (i.e., age, disease progression, organ dysfunction) in both INF and rivaroxaban plasma concentration. Furthermore, as INF is a relatively new drug on the market, there is a dearth of in vivo clinical data at present to verify our static metabolic DDI predictions. Notwith standing both these limitations, our findings hinted that the unique dichotomous inhibition modality of CYP2J2 and CYP3A4 by INF could potentially culminate in a metabolic DDI with the direct oral anticoagulant rivaroxaban and warrants further investigation via physiologically based pharmacokinetic modeling and simulation, which is currently underway.

Finally, unlike CYP3A, CYP2J2 is a relatively understudied P450 isoform that is also predominantly expressed in the heart and physiologically functions to catalyze the epoxidation of arachidonic acid to epoxyeicosatrienoic acids that mediate a myriad of cardioprotective effects ranging from anti-inflammation to cardiac ion channel modulation (Das et al., 2020). As such, perturbations of cardiac CYP2J2 metabolism of arachidonic acid may potentiate cardiotoxicities by increasing susceptibilities to cardiac insults and arrhythmias (Sisignano et al., 2020). Intuitively, many clinical drugs that potently inhibit or inactivate CYP2J2 have also been correlated with deleterious cardiac electrophysiological and cardiotoxic effects (Karkhanis et al., 2017). For instance, the second-generation antihistamine agents astemizole and terfenadine were withdrawn from the market due to their significant proarrhythmic effects (i.e., QTc prolongation and torsades de pointes tachyarrhythmia). Interestingly, both these drugs were later found to elicit potent inhibition of CYP2J2 in primary human cardiomyocytes in addition to hERG cardiac potassium channels (Evangelista et al., 2013). Therefore, our laboratory speculated that the inhibition of CYP2J2 could further exacerbate their cardiotoxic (i.e., proarrhythmic) propensities via the inhibition of epoxyeicosatrienoic acids biosynthesis. Indeed, we demonstrated that the antiarrhythmic agent dronedarone that potently inactivated the CYP2J2-mediated arachidonic acid-epoxyeicosatrienoic acids axis leading to cardiac mitochondrial toxicity was rescuable via epoxyeicosatrienoic acids enrichment (Karkhanis et al., 2018). Consequently, it is tempting to speculate that in addition to constituting a plausible pharmacokinetic DDI risk with CYP2J2 substrates (i.e., rivaroxaban), INF could also inhibit and inactivate cardiac CYP2J2 and dysregulate the endogenous arachidonic acid-epoxyeicosatrienoic acids metabolic pathway and precipitate cardiac-related toxicological adverse events. Although there are no current reports of cardiotoxicities or arrhythmias with INF, our findings here underscored the impetus to interrogate its inhibition of arachidonic acid metabolism in future studies.

In conclusion, we established that INF elicits potent reversible inhibition and MBI of CYP2J2 and that its coadministration with rivaroxaban may potentially result in a clinically significant in vivo metabolic DDI. As rivaroxaban is gaining prominence as the anticoagulant of choice in CA-VTE, the static DDI predictions reported in this study bear clinical significance and will be further extended to dynamic in silico physiologically based pharmacokinetic modeling and simulations.

Abbreviations

AUCR

area under the curve ratio

CA-VTE

cancer-associated venous thromboembolism

DDI

drug-drug interaction

FDA

United States Food and Drug Administration

FGFR

fibroblast growth factor receptor

G6P

glucose-6-phosphate

G6PDH

glucose-6-phosphate dehydrogenase

GSH

glutathione

INF

infigratinib

K _I

inactivator concentration at half-maximum inactivation rate constant

K _i

reversible inhibition constant

k _inact

maximum inactivation rate constant

K _m

Michaelis constant

k _obs

observed first-order rate constant of inactivation

MBI

mechanism-based inactivation

MIC

metabolite-intermediate complex

P450

Cytochrome P450

TDI

time-dependent inhibition

UPLC-MS/MS

ultra-high performance liquid chromatography tandem mass spectrometry

Authorship Contributions

Participated in research design: Tang, Chan.

Conducted experiments: Tang, Wu.

Performed data analysis: Tang, Wu.

Wrote or contributed to the writing of the manuscript: Tang, Chan.