Abstract
Mounting evidence has revealed that despite the high degree of sequence homology between cytochrome P450 3A isoforms (i.e., CYP3A4 and CYP3A5), they have the propensities to exhibit vastly different irreversible and reversible interactions with a single substrate. We have previously established that benzbromarone (BBR), a potent uricosuric agent used in the management of gout, irreversibly inhibits CYP3A4 via mechanism-based inactivation (MBI). However, it remains unelucidated if CYP3A5—its highly homologous counterpart—is susceptible to inactivation by BBR. Using three structurally distinct probe substrates, we consistently demonstrated that MBI was not elicited in CYP3A5 by BBR. Our in silico covalent docking models and molecular dynamics simulations suggested that disparities in the susceptibilities toward MBI could be attributed to the specific effects of BBR covalent adducts on the F-F′ loop. Serendipitously, we also discovered that BBR reversibly activated CYP3A5-mediated rivaroxaban hydroxylation wherein apparent Vmax increased and Km decreased with increasing BBR concentration. Fitting data to the two-site model yielded interaction factors α and β of 0.44 and 5.88, respectively, thereby confirming heterotropic activation of CYP3A5 by BBR. Furthermore, heteroactivation was suppressed by the CYP3A inhibitor ketoconazole in a concentration-dependent manner and decreased with increasing preincubation time, implying that activation was incited via binding of parent BBR molecule within the enzymatic active site. Finally, noncovalent docking revealed that CYP3A5 can more favorably accommodate both BBR and rivaroxaban in concert as compared with CYP3A4, which further substantiated our experimental observations.
SIGNIFICANCE STATEMENT Although it has been previously demonstrated that benzbromarone (BBR) inactivates CYP3A4, it remains uninterrogated whether it also elicits mechanism-based inactivation in CYP3A5, which shares ∼85% sequence similarity with CYP3A4. This study reported that BBR exhibited differential irreversible and reversible interactions with both CYP3A isoforms and further unraveled the molecular determinants underpinning their diverging interactions. These data offer important insight into differential kinetic behavior of CYP3A4 and CYP3A5, which potentially contributes to interindividual variabilities in drug disposition.
Introduction
CYP3A4 is the most abundant cytochrome P450 (P450) isoform expressed in the liver. Coupled with its wide substrate and catalytic promiscuity, it is known to be involved in the oxidative metabolism of >50% of all marketed drugs (Wilkinson, 1996; Wrighton et al., 1996). Consequently, it has been the focal point of intense pharmacological research. Recently, CYP3A5—its highly homologous counterpart—has also gained prominence in xenobiotic metabolism. Although both CYP3A4 and CYP3A5 possess overlapping substrate specificities, they are known to have different relative contributions to its clearance (Tseng et al., 2014). Furthermore, expression levels of CYP3A5 can vary considerably between ethnic populations because of the occurrence of the defective *3 allele that encodes for a truncated and nonfunctional protein. Conversely, in wild-type *1 carriers, the hepatic protein expression of CYP3A5 may be on par with or even surpass that of CYP3A4 in some individuals (Lamba et al., 2002; Lin et al., 2002). As such, CYP3A5 may further augment CYP3A-mediated drug metabolism and necessitate dose adjustments (Kamdem et al., 2005). Importantly, apart from their differing metabolic capacities and expression levels, mounting evidence has revealed that despite the high degree of sequence homology between both CYP3A isoforms, they have the propensities to exhibit vastly different reversible and irreversible interactions with a common substrate (Okada et al., 2009; Sugiyama et al., 2011; Guo et al., 2019).
Mechanism-based inactivation (MBI) is one such irreversible interaction that is of clinical significance because of an irrevocable loss of enzymatic activity that can only be restored upon biosynthesis of new enzymes. Hence, the extent of pharmacokinetic drug-drug interaction tends to be more profound than with a direct reversible inhibitor (Bjornsson et al., 2003). A key molecular initiating event underscoring MBI is the prior bioactivation of the parent drug into an electrophilic reactive intermediate that subsequently primes it for covalent adduction to nucleophilic amino acid residues within the active site (Kamel and Harriman, 2013). Notably, several research groups have characterized multiple aberrant CYP3A4-mediated bioactivation pathways stemming from the benzofuran and dibrominated phenolic moiety of the uricosuric agent benzbromarone (BBR) (Fig. 1A) (Kitagawara et al., 2015; Wang et al., 2017, 2019). These findings led us to posit that BBR can potentially undergo metabolic activation to reactive electrophilic species that irreversibly inhibit CYP3A4 via MBI. Indeed, these postulations were confirmed by our recent findings, which established BBR as an irreversible MBI of CYP3A4 (Tang et al., 2021). However, it remains unelucidated as to whether CYP3A5 could also be susceptible to MBI by BBR. At this outset, previous studies interrogating the MBI in CYP3A4 and CYP3A5 have revealed vastly different susceptibilities and potencies to MBI between both isoforms (Pearson et al., 2007; Hong et al., 2016). Moreover, because of the comparatively larger active site in CYP3A isoforms, probe substrate–specific inactivation has also been observed (Chan et al., 2012). Taken together, it is therefore imperative to interrogate the MBI susceptibilities of both CYP3A isoforms to BBR individually using an array of structurally distinct probe substrates to investigate whether it elicits any isoform- or probe substrate–specific inactivation.
Moreover, CYP3A isoforms may also be implicated in reversible atypical kinetic interactions. Atypical P450 kinetics deviate from the conventional Michaelis-Menten model due to nonfulfillment of assumptions crucial to the derivation of its kinetic parameters, hence confounding estimation of in vitro intrinsic clearances, which hinders accurate in vitro–in vivo extrapolations. Examples of such atypical kinetic phenomena include substrate inhibition and homotropic/heterotropic cooperativity among many others (Tracy, 2006). Although atypical kinetics are uncommon, they are more frequently reported with CYP3A isoforms. Although the exact mechanistic basis underpinning such observation remains nebulous, it is thought to be ascribed to the presence of a highly plastic active site that is capable of accommodating multiple substrates simultaneously with different binding modes (Sevrioukova and Poulos, 2013). In that regard, the crystal structure of CYP3A5 has been recently solved (Hsu et al., 2018; Hsu and Johnson, 2019). Remarkably, analysis of its active site topology revealed salient plasticity differences between CYP3A4 and CYP3A5. To date, it remains unknown whether the CYP3A isoforms exhibit atypical kinetics with BBR. As such, assessment of potential atypical kinetics needs to be discerned between CYP3A4 and CYP3A5.
In this study, we investigated the specific nature of interactions between BBR and the CYP3A isoforms. Our in vitro results revealed that BBR irreversibly inhibits CYP3A4 via MBI, whereas it reversibly activates CYP3A5-mediated rivaroxaban hydroxylation through heterotropic cooperative interactions. Our in silico findings further provided plausible mechanistic insights on the structural and molecular determinants underpinning their diverging interaction profiles.
Materials and Methods
Chemicals and Reagents
BBR and testosterone were obtained from Tokyo Chemical Industries (Tokyo, Japan). Dexamethasone, ketoconazole, prednisolone, and carbamazepine were acquired from Sigma-Aldrich (St. Louis, MO). Rivaroxaban was purchased from Carbosynth (Berkshire, UK). Midazolam was procured from Tocris Bioscience (Bristol, UK). Human recombinant cytochrome P450 3A4 and 3A5 Supersomes (1000 pmol/ml) (rCYP) coexpressing cytochrome b5 and NADPH-P450 reductase and the NADPH-regenerating system comprising NADP+ and glucose-6-phosphate (G6P) (NADPH A) and glucose-6-phosphate dehydrogenase (G6PDH) (NADPH B) were purchased from BD Gentest (Woburn, MA). High-performance liquid chromatography–grade acetonitrile was acquired from Tedia Company Inc. (Fairfield, OH). Ultrapure water (type I) was obtained using a Milli-Q water purification system (Millipore Corporation, Bedford, MA). All other commercially available chemicals were of analytical or high-performance liquid chromatography grade.
Screening of MBI of CYP3A4 and CYP3A5 by BBR
The potential MBI of CYP3A4 and CYP3A5 by BBR was first evaluated by preparing primary incubation mixtures (n = 3) consisting of 20 pmol/ml rCYP3A4 or rCYP3A5, BBR (at either 0, 5, or 25 µM, with the latter two corresponding to low and high concentrations, respectively), G6PDH, and 100 mM potassium phosphate buffer (pH 7.4). The enzymatic reaction was initiated via the addition of NADP+/G6P after prewarming at 37°C for 5 minutes. The final primary incubation mixture (100 µl) contained <1% v/v organic solvent. At various preincubation intervals (0, 3, 8, 15, 22, and 30 minutes), a 5-µl aliquot of each primary incubation mixture was withdrawn and transferred to 95 µl of prewarmed secondary incubation mixture consisting of a CYP3A-specific probe substrate, an NADPH-regenerating system, and 100 mM potassium phosphate buffer (pH 7.4). This yielded a 20-fold dilution. Specifically, three structurally disparate substrates of CYP3A at concentrations more than ∼4× their respective Km were included in our screening assay (i.e., 200 µM testosterone, 25 µM midazolam, and 50 µM rivaroxaban). The secondary incubation mixtures were incubated at 37°C for an additional 10 minutes (for assays involving testosterone or midazolam) or 2 hours (for experiments involving rivaroxaban). After this, an 80-µl aliquot was immediately removed and quenched with equal volumes of ice-cold acetonitrile spiked with either 1 µM prednisolone (internal standard for quantification of 6β-hydroxytestosterone and 1′-hydroxymidazolam) or 4 µM dexamethasone (internal standard for quantification of hydroxylated rivaroxaban). The quenched samples were centrifuged at 4000g at 4°C for 30 minutes to obtain the supernatant for liquid chromatography tandem mass spectrometry (LC/MS/MS) analysis.
Substrate Depletion of BBR in CYP3A4 and CYP3A5
Incubation mixtures comprising 20 pmol/ml rCYP3A4 or rCYP3A5, 1 µM BBR, G6PDH, and 100 mM potassium phosphate buffer (pH 7.4) were prepared in triplicates. After preincubating at 37°C for 5 minutes, the reaction was initiated via the addition of NADP+/G6P. The final primary incubation mixture (100 µl) contained <1% v/v organic solvent. Subsequently, at various intervals (0, 5, 10, 15, 30, 60, and 120 minutes), an 80-µl aliquot of each incubation mixture was withdrawn and quenched with equal volumes of ice-cold acetonitrile spiked with 10 nM carbamazepine (internal standard). The quenched samples were then centrifuged at 4000g at 4°C for 30 minutes, after which aliquots of the supernatant were sampled to quantify the amount of BBR remaining using LC/MS/MS.
CYP3A5 Molecular Dynamics Simulations
The apo- [Protein Data Bank (PDB): 6MJM] and holo- (PDB: 5VEU) CYP3A5 crystal structures were first downloaded from RCSB PDB and processed using the protein preparation wizard in Maestro (Schrödinger, NY) (Rose et al., 2015). Briefly, for each protein structure retrieved, only chain A was considered. All heteroatoms except for the protein, heme prosthetic group, and cocrystallized ligand (if available) were removed from the structure. Thereafter, the missing residues and side chains were added using Prime (Schrödinger, NY), and the ionization states of titratable groups were determined using Epik (Schrödinger, NY). After this, the simulation systems for each of the conditions (apo and holo) were built using CHARMM-GUI (Jo et al., 2008). In the holo condition, the GROMACS-compatible parameters for ritonavir (bound ligand for PDB: 5VEU) were generated using Antechamber (Wang et al., 2006). The physiologic concentration of 150 mM NaCl was maintained by adding an appropriate amount of Na+ and Cl− ions. In total, each solvated system consisted of approximately 100,000 atoms. Molecular dynamics (MD) simulations were carried out using GROMACS 2018.2 (Abraham et al., 2015) with the CHARMM36 forcefield for proteins (Best et al., 2012) and the CHARMM TIP3P water model (Jorgensen et al., 1983). Energy minimization and equilibration of the system were conducted using standard CHARMM-GUI steps. Briefly, the system was first minimized using the steepest descent after a 125-ps equilibration step. During the equilibration, positional restrains of 400 and 40 kJ/mol/nm2 were imposed on the protein backbone and side-chain heavy atoms, respectively. To improve sampling, we generated 20 equilibration trajectories with different random number seeds in replicates. After this, each equilibrated system was subjected to a production run of 20 ns without any positional restraint. All production runs were performed in a constant particle, pressure, and temperature ensemble. Specifically, the temperature was kept at a constant 310 K using Nosé-Hoover thermostat (Hoover, 1985) whereas the pressure of 1 bar was maintained using Parrinello-Rahman pressure coupling (Nosé and Klein, 1983). Long-range interactions were calculated using the particle-mesh Ewald method (Darden et al., 1993) and van der Waals’ interactions were described using a cutoff of 12 Å for all simulations. Periodic boundary conditions were employed in all three directions. CYP3A4 and CYP3A5 sequence alignment revealed considerable differences in the F-F′ loop and C-terminal region of these proteins. We created an in silico apo-CYP3A5 double mutant (G214D/Q479G) (CYP3A5MUT) in which Gly214 and Gln479 on the F-F′ loop were replaced by Asp and Gly as observed in CYP3A4 to ascertain the biochemical implications of these variations. An additional set of 20 trajectories (20 ns each) were collected for the apo-CYP3A5MUT. In total, we collected 1.2 µs long trajectory data for all three conditions (apo-CYP3A5WT, holo-CYP3A5WT, apo-CYP3A5MUT).
Root-Mean-Square Deviation–Based Hierarchical Clustering
Root-mean-square deviation (RMSD)-based clustering was performed on apo- and holo-CYP3A5 trajectories. Briefly, we first concatenated all apo- and holo-CYP3A5 trajectories and computed the pairwise RMSD matrix. In particular, the heavy atoms of the heme prosthetic group were used for superimposition, and the pairwise RMSD was calculated for residues 209-243 and 465-490, which comprise the F-F′ loop (residue 211-220) and C-terminal loop. The computed RMSD matrix was then subjected to clustering using the hierarchical-clustering algorithm implemented in SciPy with an RMSD cutoff of 3 Å. Finally, MD frames representing the centroid of the clusters were extracted for subsequent covalent docking.
In Silico Ligand Preparation
Ligands for the covalent docking were prepared as previously described (Tang et al., 2021). Briefly, we obtained 3D coordinates of BBR from the PubChem Database (Kim et al., 2019), from which two epoxide metabolites, namely, benzbromarone-6,7-benzofuran-epoxide (BBR-6,7-BF-epoxide) (Fig. 1B) and BBR-5,6-BF-epoxide (Fig. 1C) were derived using the sketcher module in the Schrödinger suite (Schrödinger, NY). Thereafter, the lowest energy conformer was generated for both epoxide metabolites using the LigPrep (Schrödinger, NY).
In Silico Binding Site Prediction and Covalent Docking
Covalent docking was performed as previously described (Tang et al., 2021). Briefly, potential ligand-binding pockets for each of the included MD frames and CYP3A5 crystal structures were first identified using the site recognition software Sitemap (Schrödinger, NY). As our previous work suggested that BBR epoxides could covalently alkylate cysteine residues vicinal to the F-F′ loop and adversely modulate its flexibility, we first directed our focus to the cysteine residues in CYP3A5 that are in close proximity to the F-F′ loop. Interestingly, unlike CYP3A4, there was an absence of cysteine residues near the F-F′ loop of CYP3A5. Consequently, we computed a list of accessible serine residues in binding pockets that lie near the F-F′ loop using an in-house analytical algorithm. The docking of BBR-6,7-BF-epoxide and BBR-5,6-BF-epoxide was performed in two sequential stages. We first noncovalently docked the metabolites in the binding pockets containing the accessible serine using Glide (Friesner et al., 2004). If noncovalent docking was successful, we then performed covalent docking for accessible serine residue within the pocket using CovDock (Zhu et al., 2014) in the Schrödinger suite. In particular, we docked each BBR epoxide metabolite to Ser206 and Ser239. The covalent docking scores obtained were then combined, sorted, and ranked.
Mixed-Type Inhibition of CYP3A4 and CYP3A5 by BBR
The mixed-type inhibition of BBR against CYP3A4 and CYP3A5 was investigated using three structurally distinct probe substrates (i.e., testosterone, midazolam, and rivaroxaban). All experiments were conducted in triplicates in 96-well plates. Reaction mixtures for IC50 experiments consisted of either 20 pmol/ml rCYP3A4 or rCYP3A5, BBR (0, 0.1, 0.3, 1, 3, 10, 30, and 100 µM), probe substrate (50 µM testosterone, 3 µM midazolam, or 10 µM rivaroxaban), G6PDH, and 100 mM potassium phosphate buffer (pH 7.4). After prewarming at 37°C for 5 minutes, the reaction was initiated via the addition of NADP+/G6P, yielding a final primary incubation mixture (100 µl) with <1% v/v organic solvent. The reaction mixture was incubated at 37°C for 10 minutes (for assays involving testosterone or midazolam) or 2 hours (for experiments involving rivaroxaban). After this, aliquots of the samples were quenched, centrifuged, and subjected to LC/MS/MS analysis for the quantitation of either 6β-hydroxytestosterone, 1′-hydroxymidazolam, or hydroxylated rivaroxaban.
Calculation of Mixed-Type Inhibition Parameters (IC50)
The IC50 value was determined using the log(inhibitor) versus normalized response—variable slope model based on eq. 1 in GraphPad 8.0.2 (San Diego, CA): wherein IC50 represents the half-maximal inhibitory concentration, [I] is the in vitro concentration of the mixed-type inhibitor, Hill slope is the Hill coefficient, and Y is the % enzyme activity compared with control.
Heterotropic Activation of CYP3A5-Mediated Rivaroxaban Hydroxylation by BBR
The heteroactivation kinetics of CYP3A5-mediated rivaroxaban hydroxylation was investigated under time-linear conditions (i.e., 2 hours) (unpublished data) in 96-well plates by preparing incubation mixtures containing 20 pmol/ml rCYP3A5, BBR (0, 1, 3, and 10 µM), rivaroxaban (2.5, 5, 15, 30, and 50 μM), G6PDH, and 100 mM potassium phosphate buffer (pH 7.4) in triplicates. After preincubating at 37°C for 5 minutes, the reaction was initiated via the addition of NADP+/G6P and further incubated at 37°C for 2 hours. The final primary incubation mixture (100 µl) contained <1% v/v organic solvent. Subsequently, aliquots of the samples were quenched, centrifuged, and subjected to LC/MS/MS analysis for the quantitation of hydroxylated rivaroxaban.
Calculation of Heteroactivation Kinetic Parameters
Kinetic constants (i.e., Vmax and Km) were derived from nonlinear least-square regression analysis by fitting hydroxylated rivaroxaban formation data collected at various rivaroxaban and BBR concentration to the Michaelis-Menten equation (eq. 2) in GraphPad 8.0.2 (San Diego, CA). wherein v is the rate of enzyme activity, Vmax (maximum rate of reaction) and Km (Michaelis constant) are kinetic constants for substrate metabolism, and [S] is the in vitro concentration of substrate.
Additionally, the data were also fitted to nonlinear least-square regression based on the two-site model equation (eq. 3) (Korzekwa et al., 1998) using Sigma Plot 14.0 (San Jose, CA) for three-dimensional surface plot analysis of rivaroxaban hydroxylation in the presence of BBR. wherein Vmax (maximum rate of reaction) and Km (Michaelis constant) are kinetic constants for substrate metabolism, [S] is the in vitro substrate concentration, [B] is the in vitro activator concentration, KB is the activator binding constant, and α and β are interaction factors indicating changes in Km and Vmax resulting from activator binding, respectively, in which, α and β < 1 is indicative of a decrease in Km or Vmax, respectively, whereas α and β > 1 represents an increase in Km or Vmax, respectively. For activation, α < 1 and β > 1. Equation 3 only takes into consideration metabolism of the probe substrate (i.e., rivaroxaban) and assumes that product release is fast relative to oxidation rates, rapid equilibrium, and single occupancy in that only a single molecule of the probe substrate (i.e., rivaroxaban) and activator (i.e., BBR) are bound to the enzymatic active site simultaneously.
Effect of CYP3A Inhibitor on Heteroactivation
To investigate whether heteroactivation of CYP3A5-mediated rivaroxaban hydroxylation can be abolished by CYP3A5 inhibition, the potent CYP3A competitive inhibitor ketoconazole at a concentration of 0.1 and 1 µM [roughly 1× and 10× the Ki for CYP3A5 (Greenblatt et al., 2011)] was introduced in triplicates to the primary incubation mixture comprising 20 pmol/ml CYP3A5, BBR (0, 0.1, 0.3, 1, 3, 10, and 30 µM), rivaroxaban (10 µM), G6PDH, and 100 mM potassium phosphate buffer (pH 7.4). The reaction was initiated by the addition of NADP+/G6P after preincubation at 37°C for 5 minutes. After incubation at 37°C for 2 hours, aliquots of the samples were quenched, centrifuged, and subjected to LC/MS/MS analysis for the quantitation of hydroxylated rivaroxaban. Incubation mixtures that excluded the addition of either ketoconazole or both BBR and ketoconazole served as the negative controls.
Effect of Preincubation on Heteroactivation
Incubation mixtures consisting of 20 pmol/ml CYP3A5, BBR (0 or 10 µM), G6PDH, and 100 mM potassium phosphate buffer (pH 7.4) were prepared in triplicates. After prewarming at 37°C for 5 minutes, the reaction was initiated via the introduction of NADP+/G6P, yielding a final primary incubation mixture (99 µl) with <1% v/v organic solvent. At various time points (0, 10, and 20 minutes), 10 µM rivaroxaban (1 µl) was added into the incubation mixture, and the reaction was allowed to continue for another 2 hours at 37°C. After this, aliquots of the samples were quenched, centrifuged, and subjected to LC/MS/MS analysis for the quantitation of hydroxylated rivaroxaban.
Noncovalent Docking of BBR and Rivaroxaban in CYP3A4 and CYP3A5
To elucidate possible structural insights underpinning the differential susceptibilities of both CYP3A isoforms to heteroactivation by BBR, we independently docked BBR and rivaroxaban to CYP3A4 and CYP3A5. Briefly, we first arbitrarily divided the CYP3A orthosteric binding site (OBS) into two separate cavities—namely the proximal binding pocket (PBP), which is the spatial region of the OBS located within 15 Å from the heme iron, and the distal binding pocket (DBP), which comprises the rest of the binding pocket not subsumed under the PBP. A second series of RMSD-based hierarchical clustering was conducted on CYP3A4 and CYP3A5 crystal structures and MD frames using OBS residues, which yielded a total of four crystal structures from CYP3A4 and one crystal structure along with three MD frames from CYP3A5 for noncovalent docking. Here, we employed two docking schemes that consisted of two individual noncovalent docking stages. In the first docking scheme, we first noncovalently docked BBR to the PBP and generated 100 docking poses. A top-scoring ligand docking pose in which the hydroxyl oxygen atom in the 2,6-dibromophenol group of BBR oriented within 5 Å of the heme iron was selected and used to perform a second round of docking in which rivaroxaban was docked in the DBP. Conversely, in the second docking scheme, we first docked rivaroxaban to the PBP followed by BBR to the DBP. As per the previous scheme, we also generated 100 docking poses for rivaroxaban in the PBP, which yielded a top-scoring ligand docking pose in which the oxygen atom in the morpholinone moiety of rivaroxaban oriented within 5 Å of the heme iron was selected for a second round of docking. To evaluate and compare the extent of binding of BBR and rivaroxaban in the OBS of CYP3A4 and CYP3A5, we computed the overall OBS volume and solvent-accessible surface area (SASA) of the ligand bound in the DBP. SASA values obtained were then further converted into percentage ligand exposure using eq. 4 wherein SASAbound and SASAfree represent the SASA of the ligand in bound and free form, respectively. Higher SASA and ligand exposure values are indicative of weaker protein-ligand interactions.
Measurement of Residual CYP3A Activity and BBR via LC/MS/MS
All samples were analyzed using the LC/MS/MS system consisting of an Agilent 1290 Infinity ultra-high pressure liquid chromatography (Agilent Technologies Inc., Santa Clara, CA) interfaced with AB SCIEX Triple Quad 3500 tandem mass spectrometry system (AB SCIEX, Framingham, MA). Chromatographic separation was achieved with an ACQUITY UPLC BEH C18, 1.7 µm, 2.1 × 50 mm column (Waters, Milford, MA). The aqueous mobile phase (A) was 0.1% formic acid in water, whereas the organic mobile phase (B) was 0.1% formic acid in acetonitrile. Mobile phases were delivered at a flow rate of 0.5 ml/min. The column and sample temperature were set at 45°C and 4°C, respectively. The gradient elution conditions were as follows: linear gradient from 20% to 80% B (0–1.20 minutes), isocratic at 100% B (1.21–2.00 minutes), and isocratic at 20% B (2.01–2.50 minutes). All analytes were detected in positive electrospray ionization mode. The source-dependent MS parameters were as follows: ion spray voltage = 5500 V; source temperature = 500°C; curtain gas = 25 psi; ion source gas 1 (sheath gas) = 30 psi; and ion source gas 2 (drying gas) = 30 psi. The multiple reaction monitoring transitions and compound-dependent MS parameters of the analytes are summarized in Supplemental Table 1. Chromatographic peak integration was performed using SCIEX OS version 1.5.0.23389 (AB SCIEX, Framingham, MA). For all LC/MS/MS analyses, the peak area of the analyte was expressed as a ratio to the peak area of the internal standard.
Results
Absence of Time-Dependent Inactivation of CYP3A5 by BBR
Our screening experiments revealed that BBR elicited a time- and concentration-dependent reduction in CYP3A4 activity in all three probe substrates used (Fig. 2, A–C), which corroborates our previous findings on the irreversible covalent MBI of CYP3A4 by BBR (Tang et al., 2021). Additionally, covalent modification of the CYP3A4 apoprotein and destruction of the prosthetic heme (i.e., via alkylation or oxidative fragmentation) by BBR is further substantiated by our electrospray ionization–MS analyses, which demonstrated that the intensity of the native CYP3A4 spectrum peak decreased by approximately 30.3% when coincubated with 50 µM BBR for 30 minutes as compared with vehicle controls (refer to Supplemental Material; Supplemental Fig. 1, A–D for more details). In contrast, time-dependent reduction in CYP3A5 activity by BBR was not observed in all three probe substrates used (Fig. 2, D–F). As MBI hinges on the prior bioactivation of the substrate to a reactive intermediate, we performed substrate depletion studies to ascertain whether BBRs are substrates of CYP3A4 and CYP3A5. Measuring the amount of parent drug remaining over 2 hours revealed that both CYP3A isoforms were capable of metabolizing BBR but with different efficiencies. Specifically, the percentage of BBR remaining after 2 hours (expressed as mean ± S.D) was 52.62 ± 1.82% in CYP3A4 incubations and 81.51 ± 0.29% in CYP3A5 incubations (Supplemental Fig. 2). These results confirm that BBR is a substrate of both CYP3A isoforms, with BBR being metabolized to a lower extent by CYP3A5.
CYP3A5 Crystal Structure Analysis
To gain mechanistic insights on the structural determinants underpinning the differential MBI susceptibilities to BBR in CYP3A4 and CYP3A5, we closely compared the crystal structures of both these CYP3A isoforms. Notably, there was an absence of cysteine residues near the F-F′ loop of CYP3A5. Rather, the two cysteine residues (i.e., Cys58 and Cys239) that we previously determined to lie close to the F-F′ loop in CYP3A4 were substituted to Trp58 and Ser239 in CYP3A5 (Supplemental Fig. 3). Consequently, we directed our search toward nucleophilic serine residues in CYP3A5. Our structural analysis identified two accessible serine residues, namely Ser206 and Ser239, near the F-F′ loop in CYP3A5 that were predicted to be amenable to covalent binding by BBR epoxide metabolites.
CYP3A5 MD Simulations
Because of the availability of only two reported crystal structures (i.e., PDB: 6MJM and 5VEU) (Hsu et al., 2018; Hsu and Johnson, 2019), multiple short MD simulations were performed for holo- and apo-CYP3A5 to generate an ensemble of CYP3A5-associated MD frames. This allows for a meaningful comparison of covalent docking results to be made with CYP3A4, which possesses multiple holo and apo crystal structures. Here, our simulations yielded a total of 40 MD trajectories (20 for each condition) with a cumulative length of 800 ns (400 ns for each condition). Subsequently, after concatenating and superimposing all trajectories, RMSD-based clustering segregated frames into 28 distinct clusters. MD frames representing the centroid of each cluster were extracted for subsequent covalent docking.
CYP3A5 Covalent Docking
BBR-6,7-BF-epoxide and BBR-5,6-BF-epoxide were covalently docked to 28 representative frames obtained from our MD trajectory clustering analysis along with the previously reported apo (PDB: 6MJM) and holo (ritonavir-bound) (PDB: 5VEU) CYP3A5 crystal structures. Results from our covalent docking demonstrated that BBR adducts at Ser239 could similarly adopt a binding pose adjacent to (minimum distance <3.5 Å) the F-F′ loop (Fig. 3) (Supplemental Table 2). Moreover, the top covalent docking scores obtained for BBR-6,7-BF-epoxide and BBR-5,6-BF-epoxide adducts in CYP3A5 (−4.4, −3) (Table 1) were comparable to scores previously reported for CYP3A4 (−3.1, −3.5) (Table 1) (Tang et al., 2021). Taken together, our findings suggest that both CYP3A4 and CYP3A5 have similar propensities to form adducts with BBR epoxide metabolites.
Our previous covalent docking models with CYP3A4 demonstrated that BBR-Cys239 adducts could stabilize a single F-F′ loop conformation in which the F-F′ loop residues are positioned in close proximity to the C-terminal loop (termed as the “closed” conformation) (Fig. 4C). Interestingly, this “closed” conformation was also observed in apo (PDB: 1TQN) and holo (ritonavir-bound) (PDB: 3NXU) crystal structures of CYP3A4 (Fig. 4, A and B) (Yano et al., 2004; Sevrioukova and Poulos, 2010). Comparative analysis of covalent docking poses in CYP3A5 revealed that BBR-Ser239 adducts could stabilize two structurally distinct F-F′ loop conformations. On one hand, as evident in cluster-13, the resulting stabilized F-F′ loop may adopt the “closed” conformation that is also seen in the apo (PDB: 6MJM) and holo (ritonavir-bound) (PDB: 5VEU) CYP3A5 crystal structures (Fig. 4, D–F). On the other hand, as illustrated in cluster-18, the F-F′ loop may be oriented away from C-terminal loop, resulting in a wider entrance (5.3 Å wider) for the OBS (termed as the “open” conformation) (Fig. 4F). To rationalize observed differences in F-F′ loop conformation between the two CYP3A isoforms, we performed sequence alignment to identify key residue differences. Interestingly, Asp214 in CYP3A4 was found to be substituted to a more flexible Gly214 in CYP3A5 (Supplemental Fig. 3). Apo and holo crystal structures of CYP3A4 suggest that Asp214 plays an important role in stabilizing the “closed” conformation of the F-F′ loop by forming hydrogen bonds with Gly481 in the C-terminal loop (Fig. 4, A and B). Although a complementary G479Q substitution is evident in the C-terminal loop of CYP3A5 (Supplemental Fig. 3), apo and holo crystal structures reveal that Gln479 does not form any hydrogen bonds with residues on the opposing F-F′ loop (Fig. 4, D and E). To further ascertain the impact of hydrogen bonding between F-F′ and C-terminal loop in CYP3A5 on F-F′ loop conformations and its ramifications in modulating the channel entrance, we computed the minimum distances between the backbone atoms of Gly214 from Gln479 side-chain polar atoms and Gly480 (structurally equivalent to Gly481 in CYP3A4) backbone atoms (see Supplemental Table 3 for more details) using apo-CYP3A5WT MD trajectories. Using a distance cutoff of 3.5 Å, we determined that Gly214 interacts with Gln479 and Gly480 in only 2.4% (Fig. 5A) and 6.5% (Fig. 5B) of the frames, respectively. A comparative interaction analysis in apo-CYP3A5MUT MD trajectories revealed that although G214D form negligible interactions (0.3%) with Q479G (Fig. 5C), G214D interestingly exhibited a substantially higher interaction frequency (23.3%) (Fig. 5D) with Gly480. Moreover, the interaction between G214D and Gly480 is very dynamic (Supplemental Fig. 4). Taken together, our findings suggest that the D214G substitution in CYP3A5 renders its F-F′ loop more flexible than CYP3A4. Consequently, the CYP3A5 F-F′ loop could move dynamically closer to the C-terminal loop in cluster-13 in one instance and further away from the C-terminal loop in cluster-18 in another instance (Fig. 4F). When framed in the context of our in vitro findings, we posited that the dynamic flexibility of the F-F′ loop in cluster-18 could have preserved substrate accessibility to the OBS, explaining the lack of MBI despite covalent adduction by BBR epoxides at Ser239. Conversely, BBR adducts at Ser206 were found to be oriented further away from the F-F′ loop and entrance to the OBS (Supplemental Fig. 5), hence we are inclined to believe that covalent adduction at Ser206 is likely to be less pertinent to mechanistically explain the lack of MBI in CYP3A5.
Mixed-Type Inhibition of CYP3A4 and CYP3A5 by BBR
Our earlier MBI screening assays (Fig. 2) also hinted that BBR could directly inhibit CYP3A4 and CYP3A5. Consequently, we proceeded to probe their mixed-type inhibitions to better characterize the nature of interaction between BBR and both CYP3A isoforms. Our findings demonstrated that BBR reversibly inhibited CYP3A4-mediated testosterone 6β-hydroxylation, midazolam 1′-hydroxylation, and rivaroxaban hydroxylation with IC50 values (expressed as mean ± S.D) of 9.11 ± 1.1 μM, 21.47 ± 1.1 μM, and 42.35 ± 6.34 μM, respectively (Fig. 6, A–C). Similarly, when CYP3A5 was incubated with BBR, a characteristic sigmoidal dose-dependent decrease in testosterone 6β-hydroxylation and midazolam 1′-hydroxylation activities was evident with IC50 values (expressed as mean ± S.D) of 11.76 ± 2.0 M and 43.68 ± 2.31 μM, respectively (Fig. 6, D and E). Intriguingly, this trend was not recapitulated in incubation mixtures comprising CYP3A5 and rivaroxaban. Rather, CYP3A5-mediated rivaroxaban hydroxylation was enhanced in a concentration-dependent manner by BBR (Fig. 6F). Notably, this effect peaked at 10 μM, beyond which a decline in enzyme activity was observed instead. This apparent activation of CYP3A5 metabolism was only present with rivaroxaban and absent when the United States Food and Drug Administration–recommended prototypical probe substrates (i.e., testosterone and midazolam) were used (Fig. 7A).
Heterotropic Activation of CYP3A5-Mediated Rivaroxaban Hydroxylation by BBR
To further expound on the observed atypical kinetic phenomenon, we characterized its kinetics to quantitatively evaluate the impact of BBR on rivaroxaban affinity (Km) and CYP3A5 rivaroxaban hydroxylation activity (Vmax). Fitting the data to the Michaelis-Menten model demonstrated that apparent Vmax was augmented, whereas Km (expressed as mean ± S.D.) was reduced from 34.51 ± 12.67 μM to 23.91 ± 4.60 μM (Fig. 7B). This yielded an estimated ∼5.69-fold increase in apparent Vmax/Km ratio at the highest concentration of BBR (10 μM) and hints that the apparent intrinsic clearance of rivaroxaban by CYP3A5 increases in the presence of BBR (Table 2). Taken together, our findings further substantiated heteroactivation of CYP3A5-mediated rivaroxaban hydroxylation by BBR.
Refinements in the heteroactivation kinetics were made using the two-site model proposed by Korzekwa et al. (1998), which produced good fits (R2 = 0.99) (Table 3). The Km (expressed as mean ± S.D) of CYP3A5-mediated rivaroxaban hydroxylation in the absence of BBR was determined to be 36.72 ± 19.15 μM using the two-site model, which is comparable to the value derived from the Michaelis-Menten model (34.51 ± 12.67 μM). Interaction factors α and β were determined to be 0.44 and 5.88, respectively (Table 3). Since α < 1 and β > 1, our results suggest that BBR activates the metabolism of rivaroxaban via a decrease in Km and an increase in Vmax. At this outset, this complex kinetic relationship is also observable from the three-dimensional surface plot (Fig. 8). Finally, the KB was determined to be 14.55 μM (Table 3).
Effect of CYP3A Inhibitor on Heteroactivation
Activation of CYP3A5-mediated rivaroxaban hydroxylation by BBR was suppressed in the presence of 0.1 µM ketoconazole (Fig. 7C). However, the concentration-dependent enhancement of CYP3A5 metabolism was still apparent and similarly peaked at 10 µM of BBR. In contrast, coincubation with 1 µM ketoconazole (∼10× Ki) completely abolished heteroactivation elicited by BBR.
Effect of Preincubation on Heteroactivation
To delineate whether heteroactivation is incited by parent BBR or its metabolites, the effects of preincubation were investigated. Our results reveal that enhancement of CYP3A5-mediated rivaroxaban hydroxylation decreased moderately with increasing preincubation time (Fig. 7D).
Noncovalent Docking of BBR and Rivaroxaban in CYP3A4 and CYP3A5
Our in silico structural analyses revealed that volumes of the OBS did not vary considerably between CYP3A4 (720.6–1360.0 Å3) and CYP3A5 (889.6–1376.0 Å3) (Tables 4 and 5). Additionally, noncovalent docking models revealed that in both docking schemes, the first ligand docked to the PBP (i.e., BBR in scheme 1 and rivaroxaban in scheme 2) in the absence of the second ligand in the DPB produced more favorable docking scores in CYP3A4 (avg. docking scores of −5.4 and −6.4 in schemes 1 and 2, respectively) as compared with that in CYP3A5 (avg. docking scores of −4.5 and −5.2 in schemes 1 and 2, respectively), thereby implying a higher binding affinity of the first ligand bound to CYP3A4 (Tables 4 and 5). On the contrary, after the PBP is occupied with the first docked ligand, the second ligand docked to the DBP (i.e., rivaroxaban in scheme 1 and BBR in scheme 2) exhibited an opposing trend in that CYP3A5 yielded more favorable binding scores (avg. docking scores of −7.7 and −6.3 in schemes 1 and 2, respectively) than CYP3A4 (avg. docking scores of −6.9 and −4.7 in schemes 1 and 2, respectively). This suggested a greater likelihood for CYP3A5 to accommodate the second docked ligand in the presence of the first docked ligand. Concordant with these findings, we also observed smaller SASA values and ligand exposure from the second ligand docked to CYP3A5 (ligand exposure of 1.7%–5.4% and 0.9%–12.7% in schemes 1 and 2, respectively) than to CYP3A4 (ligand exposure of 8.0%–26.7% and 8.8%–27.0% in schemes 1 and 2, respectively) in all but one case (Tables 4 and 5).
Discussion
In this study, we delineated the unique dichotomous interactions of BBR with CYP3A4 and CYP3A5 and characterized the kinetics and structural determinants underpinning their differing inactivation or activation profiles for the first time. As the CYP3A isoforms share ∼85% sequence homology, we postulated that the observed differences in their interactions with BBR may be ascribed to the subtle yet important nuances in their active sites architecture, which consequently facilitate differential irreversible or reversible interactions.
Using three structurally distinct probe substrates of CYP3A, we confirmed the presence and absence of MBI by BBR against CYP3A4 and CYP3A5, respectively. Although it has been previously demonstrated experimentally that the reactive electrophilic epoxide metabolites of BBR were primarily generated by the CYP3A subfamily (Wang et al., 2016), because of the low turnover rate of BBR by CYP3A5, it is plausible that the lack of observed MBI elicited by CYP3A5 in our assays could stem from its low intrinsic clearance, which could also reduce the likelihood of generating the epoxide metabolite implicated in its MBI. To account for this, we have included a high concentration of BBR (25 µM) in our MBI screening experiments and allowed sufficient time for CYP3A5-mediated metabolism/bioactivation of BBR by preincubating BBR with CYP3A5 for up to 30 minutes before the subsequent introduction of the probe substrate. In spite of these, our findings consistently demonstrated a lack of MBI in CYP3A5 in all three CYP3A probe substrates used. As such, we posited that slight alterations in active site architecture of CYP3A4 and CYP3A5 could be responsible instead for its diverging MBI susceptibilities despite the bioactivation of BBR to its epoxides in both isoforms. Sequence alignments further revealed that these two reactive cysteines that we previously identified in CYP3A4 are substituted to Trp58 and Ser239 in CYP3A5. Consequently, we initially proposed that discrepancies in the observed MBI susceptibilities between the two CYP3A isoforms may be attributed to the different nucleophilicities of the side chains in relation to their ability to sequester reactive metabolites generated via bioactivation. However, covalent docking models have since suggested that both Cys239 in CYP3A4 and Ser239 in CYP3A5 bear similar propensities to form adducts with BBR epoxide metabolites. Although unexpected, it has been previously reported that covalent modification of the enzymatic active site need not always precipitate in a loss of enzyme activity (Barr et al., 2020). At this outset, we are inclined to believe that there may be a more nuanced explanation underscoring the different MBI susceptibilities in both CYP3A isoforms. In that regard, additional structural insights gleaned from our docking models have proposed that BBR-Cys239 adducts in CYP3A4 stabilized a “closed” conformation of the F-F′ loop, which could restrict the approach of an incoming substrate into the OBS and result in enzyme inactivation. In contrast, MD simulations and docking models in CYP3A5 revealed that its F-F′ loop possesses a greater degree of structural flexibility, which allowed for the subsequent stabilization of both a “closed” and “open” F-F′ loop conformation by BBR-Ser239 adducts. This is further substantiated by our apo-CYP3A5MUT MD trajectories, which proposed that its structural flexibility results from D214G substitution in F-F′ loop of CYP3A5. As the F-F′ loop is oriented directly above the OBS and plays an instrumental role in substrate-enzyme binding by functioning as a gating mechanism involved in substrate access and egress, its resulting conformation is anticipated to have important implications to substrate binding. Consequently, we postulated that the greater inherent flexibility coupled with the maintenance of an “open” F-F′ loop conformation may have preserved substrate accessibility to the OBS and translated to a lack of MBI elicited by BBR in CYP3A5. However, it should be noted that our assertions here were based on covalent docking results of BBR epoxide metabolites to centroid frames extracted from MD simulations of CYP3A5 due to the inaccessibility of Ser239 in the two published crystal structures. Nonetheless, such an observation elegantly illustrates how the ensemble-based approach that we adopted for docking analysis could circumvent limitations associated with the lack of reported CYP3A5 crystal structures.
While interrogating the MBI susceptibilities of CYP3A4 and CYP3A5 by BBR, we observed a reduction in residual enzyme activity with increasing BBR concentrations in the absence of preincubation, implying possible reversible inhibition by BBR. Our experiments subsequently confirmed mixed-type inhibition of CYP3A4 with all three probe substrates used. Atypically, although we demonstrated that mixed-type inhibition of CYP3A5 could be recapitulated with testosterone and midazolam, there was an apparent enhancement of CYP3A5 activity when rivaroxaban was used as the probe substrate in the presence of BBR, hinting at a possible heterotropic activation of CYP3A5-mediated rivaroxaban hydroxylation by BBR. These serendipitous findings prompted us to characterize the observed heteroactivation using a more representative kinetic model that could simultaneously account for the kinetics of both the probe substrate and activator. Korzekwa et al. (1998) previously described a two-site model that was harnessed to fit heteroactivation of CYP3A-mediated metabolism of midazolam by erlotinib (Dong et al., 2011). Here, we demonstrated that the two-site model could also provide a good fit for our data (R2 = 0.99). Interaction factors of α (0.44) and β (5.88) were in accordance with the criteria for activation (i.e., α < 1 and β > 1). Coincubation with ketoconazole suppressed the enhancement of CYP3A5-mediated rivaroxaban hydroxylation in a concentration-dependent manner, further lending support to the notion that heteroactivation by BBR occurred within the active site (Houston and Galetin, 2005). Finally, considering that the amount of BBR decreased moderately with increasing preincubation time due to its metabolism by CYP3A5 and increasing preincubation time resulted in a reduction in activation of CYP3A5-mediated rivaroxaban hydroxylation to a similar extent, heteroactivation of CYP3A5 by BBR is possibly elicited by parent BBR and not its metabolites.
Current understanding on the structural and molecular determinants of heterotropic activation is tenuous but X-ray crystallographic evidence points toward the existence of a large active site with remarkable plasticity that can simultaneously accommodate two or more substrate molecules within different binding pockets. Although both CYP3A4 and CYP3A5 are known to have considerably large active sites, a recently study reported that the active site cavity volume for CYP3A5 (1866 Å) was considerably larger than that in CYP3A4 (1386 Å) (Hsu et al., 2018), which may explain the disparity in heteroactivation susceptibilities between the two CYP3A isoforms. However, the active site of CYP3A4 has also been previously found to be large enough to bind two or more bulky molecules (i.e., ketoconazole) (Sevrioukova and Poulos, 2013). Consequently, rather than differences in cavity size, it is plausible that active site architectural differences could instead favor different binding conformations for BBR and rivaroxaban such that the two may only coexist in CYP3A5 but not in CYP3A4 due to spatial and steric restrictions. In alignment with this postulation, our in silico analyses have shown that when the PBP is occupied by the first docked ligand (i.e., rivaroxaban or BBR), there is a greater binding affinity for the second docked ligand in the DBP of CYP3A5 than in CYP3A4 regardless of the order of docking. This observation is substantiated by the lower SASA and ligand exposure values for the second ligand docked to the DBP in CYP3A5 than in CYP3A4, thereby implying that in comparison with CYP3A4 the OBS of CYP3A5 can more favorably accommodate both BBR and rivaroxaban in concert, which could potentially explain our in vitro observations pertaining to the heteroactivation of rivaroxaban metabolism by BBR in CYP3A5 but not in CYP3A4. At its core, modeling atypical kinetics is most rigorous when supported by mechanistic insights describing such enzyme-substrate interactions at a molecular level. As such, we hypothesize that elucidating the precise binding poses of BBR and rivaroxaban in CYP3A5 is key to understanding the mechanism of heteroactivation of rivaroxaban metabolism by BBR. Although unelucidated, a plausible mechanism could arise from initial binding of BBR to the effector binding pocket near the heme moiety in CYP3A5, which in turns elicits a conformational change in the substrate binding pocket that increases binding affinity toward rivaroxaban. Finally, it should be noted that although the in silico assays described in this study do not yield direct experimental evidence, they provide plausible and key mechanistic insights on the molecular determinants underpinning their diverging reversible and irreversible interaction profiles. Consequently, additional work involving site-directed mutagenesis and MS-based proteomics needs to be performed to validate our conjectures.
In conclusion, our findings demonstrated that BBR exhibits differential interactions with CYP3A4 and CYP3A5. Specifically, BBR irreversibly inactivates CYP3A4 via MBI, whereas BBR reversibly activates CYP3A5-mediated rivaroxaban hydroxylation through heterotropic cooperative interactions. Disparities in the susceptibility toward MBI were suggested to be attributed to the effects of covalent adducts on the F-F′ loop. Our findings here illuminate the atypical dichotomous inactivation and activation kinetics between BBR and the two highly homologous CYP3A isoforms and pave the way for future investigations into their clinical implications. Finally, these results reinforce the importance of discerning between the kinetic behavior of CYP3A4 and CYP3A5 due to their propensities for diverging interaction profiles with a common substrate.
Authorship Contributions
Participated in research design: Tang, Verma, Fan, Chan.
Conducted experiments: Tang, Verma, Yong.
Contributed new reagents or analytic tools: Lin, Fan.
Performed data analysis: Tang, Verma, Yong, Li, Wang.
Wrote or contributed to the writing of the manuscript: Tang, Verma, Li, Wang, Fan, Chan.
Footnotes
- Received February 4, 2021.
- Accepted June 21, 2021.
↵1 L.W.T.T. and R.K.V contributed equally to the work.
This work is supported by the Agency for Science, Technology and Research (A*STAR) Industry Alignment Fund–Pre-Positioning (IAF-PP) Funding [Grant H18/01/a0/C14] provided to H.F and E.C.Y.C and the National University Heart Centre Singapore (NUHCS) Cardiovascular Research Institute (CVRI)–Core Fund [Grant NUHSRO/2019/082/Core] and SCEPTRE CG Seed Grant [Grant NMRC/CG/M008/2017] provided to E.C.Y.C. L.W.T.T is supported by the National University of Singapore (NUS) President’s Graduate Fellowship (PGF).
The authors declare that they have no conflicts of interest with the contents of this article.
↵This article has supplemental material available at molpharm.aspetjournals.org.
Abbreviations
- BBR
- benzbromarone
- BF
- benzofuran
- DBP
- distal binding pocket
- G6P
- glucose-6-phosphate
- G6PDH
- glucose-6-phosphate dehydrogenase
- KB
- activator binding constant
- Km
- Michaelis constant
- LC/MS/MS
- liquid chromatography tandem mass spectrometry
- MBI
- mechanism-based inactivation
- MD
- molecular dynamics
- MS
- mass spectrometry
- MUT
- mutant
- OBS
- orthosteric binding site
- P450
- cytochrome P450
- PBP
- proximal binding pocket
- PDB
- Protein Data Bank
- rCYP3A4
- recombinant cytochrome 3A4
- rCYP3A5
- recombinant cytochrome 3A5
- RMSD
- root-mean-square deviation
- SASA
- solvent-accessible surface area
- WT
- wild type
- Copyright © 2021 by The American Society for Pharmacology and Experimental Therapeutics