Relative Selectivity of Covalent Inhibitors Requires Assessment of Inactivation Kinetics and Cellular Occupancy: A Case Study of Ibrutinib and Acalabrutinib

Kinases form an attractive class of targets for small molecule inhibitors, but similarity among their adenosine triphosphate binding sites presents difficulties for developing selective drugs. Standard methods of evaluating selectivity of most reversibly bound drugs account for binding affinity but not the two-step process, affinity and inactivation, occurring during covalent inhibition. To illustrate this concept, we assessed the selectivity of Bruton’s tyrosine kinase (BTK) over TEC kinases by two novel therapeutics: ibrutinib and acalabrutinib. The two-step process and time-dependent inhibition unique to covalent inhibitors were evaluated with two biochemical assays measuring enzymatic function and inhibition kinetics. The selectivity for BTK over TEC found in these biochemical analyses was 1–1.5 for ibrutinib and 3.0–4.2 for acalabrutinib. To further assess drug selectivity in a more physiologically relevant context, we developed cell-based occupancy assays that quantify the percentage of drug-inactivated kinases. Cellular selectivity of BTK over TEC was determined after MWCL-1 cells, and samples from patients with chronic lymphocytic leukemia (CLL) were treated for durations and concentrations based on human pharmacokinetics of each drug. In MWCL-1 cells, BTK/TEC selectivities measured at 0.5, 1, and 3 hours were 2.53, 1.05, and 1.51 for ibrutinib and 0.97, 1.13, and 2.56 for acalabrutinib, respectively. The equivalent selectivity measured in samples from patients with CLL were 1.31 ± 0.27 and 1.09 ± 0.11 for ibrutinib and acalabrutinib, respectively. Collectively, our data show that when properly accounting for time-dependent factors and relevant cellular context, ibrutinib and acalabrutinib demonstrate similar selectivity for BTK over TEC. SIGNIFICANCE STATEMENT This study shows relative selectivity of covalent inhibitors toward different kinase targets should be assessed with both affinity and inactivation kinetics to accurately account for time-dependent effects of covalent binding and assessed in a cellular matrix to reproduce the physiologic context of target inhibition. This is illustrated with a case study of ibrutinib and acalabrutinib for which selectivity assessment with appropriate assays, as opposed to measuring binding affinity with KINOMEscan alone, corroborate emerging clinical data demonstrating similar safety profiles between the therapies.


Introduction
Bruton's tyrosine kinase (BTK), a member of the TEC family of kinases, mediates B-cell receptor signaling and has emerged over the past decade as an effective clinical target for first-line therapy and for treating relapsed/refractory chronic lymphocytic leukemia (CLL) (Stevenson et al., 2011;Woyach et al., 2012). Ibrutinib is the only once-daily BTK inhibitor that is approved in the United States, Europe, and other countries for the treatment of several B-cell malignancies, including CLL and mantle cell lymphoma (MCL) (). Acalabrutinib, another BTK inhibitor taken twice daily, has received conditional approval in the United States only for the treatment of relapsed/refractory MCL and is in development for the treatment of other B-cell malignancies (AstraZeneca Pharmaceuticals LP, 2017). Although it has been demonstrated that both ibrutinib and acalabrutinib potently inhibit BTK via covalent bond formation, selectivity for this target over other TEC family kinases has not been rigorously assessed.
In 2017, Barf and colleagues (2017) reported the pharmacology of acalabrutinib and highlighted its biochemical and cellular selectivity for BTK over other kinases in the TEC family. The authors used two different biochemical assays to assess the inhibition selectivity of BTK over TEC for several covalent BTK inhibitors, including ibrutinib and acalabrutinib. In their study, the potency for BTK was measured using the immobilized metal ion affinity-based fluorescence polarization (IMAP) assay, and the potency for TEC was measured using a time-resolved fluorescence resonance energy transfer Financial support for this work was provided by Pharmacyclics LLC, an AbbVie Company.
This work was previously presented by M. Hopper et al. at the 60th ASH Annual Meeting and Exposition on December 1-4, 2018 (poster 3498).
Role of the Funder/Sponsor: Pharmacyclics LLC, an AbbVie Company, sponsored and designed the study. Study investigators and their research teams collected the data. The sponsor confirmed data accuracy and performed analysis of the data. Medical writing support was funded by the sponsor.
(TR-FRET) assay. The two assays differed in that the IMAP assay informed enzymatic activity of the kinases, whereas the TR-FRET assay measured inhibition via competitive binding and thus only informed affinity. Neither of these assays was optimal for quantifying drug selectivity, because they assessed only part of the inactivation mechanism of action. Accordingly, although each assay reported higher potency for both BTK and TEC by ibrutinib relative to acalabrutinib, selectivity between the two kinases could not be inferred from the IC 50 derived from the two different assay platforms. The authors further profiled inhibition via competitive binding of ibrutinib and acalabrutinib against a panel of 395 nonmutant human kinases using the KINOMEscan Assay Platform (Eurofins DiscoverX Corporation, Fremont, CA). This experiment showed that when both drugs were tested at the same concentration, 1 mM, ibrutinib was more potent against BTK and TEC than acalabrutinib. Notably, physiologically relevant exposures for acalabrutinib are 5-fold higher than for ibrutinib (Byrd et al., 2013(Byrd et al., , 2016; although these results were consistent with earlier studies (Honigberg et al., 2010;Byrd et al., 2016) in finding ibrutinib as the more potent inhibitor, the arbitrary use of 1 mM resulted in data that were not informative toward understanding target inhibition of the two drugs at physiologically relevant exposures (Byrd et al., 2013(Byrd et al., , 2016. Kinase inhibition via covalent bond formation is a two-step process that begins with the compound interacting with the kinase driven by affinity (represented by K i ) and ends with an inactivation step (K inact ). Accordingly, potency of covalent inhibitors cannot be determined using only traditional IC 50 measurements because this does not account for the entirety of the two-step process of covalent bond formation (Bauer, 2015). Here, we used the ratio of K inact /K i , which is the preferred metric to rank potency of different covalent inhibitors against a target (Bauer, 2015). In fact, Barf and colleagues (2017) discussed the relevance of these two parameters in their article, highlighting the faster inactivation rate for an acrylamide-substituted acalabrutinib and higher target affinity by ibrutinib as the primary factors contributing to the difference in potency of each compound. Furthermore, both the IMAP and the TR-FRET platforms are biochemical assays that measure the enzymatic reactions in relatively artificial systems that do not account for the complexity of the cellular environment, which has high (millimolar) ATP concentrations and hundreds of protein interactions. Collectively, it is clear that selectivity can be accurately assessed only in a comprehensive manner, including analysis of binding kinetics, enzymatic activity, and cellular occupancy at physiologically relevant drug concentrations. In the present study, we assessed these three parameters to rigorously evaluate the selectivity of ibrutinib and acalabrutinib for BTK and TEC.

Materials and Methods
Biochemical Assay-Enzymatic Function. Biochemical enzymatic IC 50 data were generated by Nanosyn (Santa Clara, CA). Determination of inhibitor potency against BTK enzyme was carried out using the microfluidic-based LabChip 3000 Drug Discovery System (Caliper Life Sciences, Hopkinton, MA), which uses capillary electrophoresis to separate phosphorylated and nonphosphorylated peptides. Briefly, the enzyme reaction was started by preincubating  , in which K i * is the stable complex forming constant and K i is the overall final inhibitory constant (for covalent irreversible inhibitors, k2 5 k inact and k22 5 0). Upon plotting concentration of inhibitor [I] vs. k obs , goodness of fit parameters for the two drugs were determined using either hyperbolic or linear fit from the time-dependent inhibition equation: , 1 mM dithiothreitol, and 5 mM magnesium chloride at 25°C. After enzyme-inhibitor preincubation, the kinase reaction was initiated by adding peptide substrate and ATP (50 mM) prepared in the same assay buffer, followed by further incubation of reaction mixture for 3 hours. At the end of the incubation, the reaction was quenched by a buffer containing 50 mM of EDTA. Appropriate control samples (0% inhibition in the absence of inhibitor, DMSO only, and 100% inhibition in the absence of enzyme) were assembled in replicates of four and used to calculate the percentage of inhibition in the presence of compounds. IC 50 values were determined by fitting the inhibition curves using a fourparameter sigmoidal dose-response model using XLfit 4 software (IDBS, Boston, MA). A similar protocol was used to determine the IC 50 values for TEC kinase, with minor modifications to optimize for enzyme, peptide, and ATP concentrations. Biochemical Assay-Kinetics of Target Inhibition. For kinetic studies, similar enzyme assay conditions were employed as described in the Biochemical Assay-Enzymatic Function section. The generation of progress curves for BTK/TEC peptide phosphorylation in the presence of test inhibitor was performed at 11 drug concentrations (10, 8.3, 6.9, 5.8, 4.8, 4.0, 3.4, 2.8, 2.3, 1.9, and 1.6 nM). After initiating the enzyme reaction, the real-time curves were obtained for a total period of 5 hours using a climate-controlled LabChip 3000 Drug Discovery System. Progress curves were then fitted by XLfit 4 software using the time-dependent inhibition equation , in which V i is the initial velocity, V s is the steady-state velocity, and K obs represents the rate of inactivation. For time-dependent inhibitors, the obtained K obs values were plotted against the compound concentration using either a hyperbolic or linear fit. The model with the better goodness of fit was selected. These plots enabled determination of K inact and K i values. The acquisition of all progress curve data points and the calculation of kinetic parameters were performed by Nanosyn.
Cell Treatments. The MWCL-1 cell line, which was established from a patient with Waldenström macroglobulinemia (Hodge et al., 2011), was obtained from Mayo Foundation for Medical Education and Research (Rochester, MN). MWCL-1 treatments were performed in six-well tissue culture plates using 6 Â 10 6 cells per well at a concentration of 2 Â 10 6 cells/ml in Gibco RPMI 1640 (ThermoFisher Scientific, Waltham, MA) supplemented with Gibco 1Â Penicillin-Streptomycin (ThermoFisher Scientific), 10% heat-inactivated fetal bovine serum, and 1 mM Gibco Sodium Pyruvate (ThermoFisher Scientific). Cells were treated with inhibitors for 0.5, 1, or 3 hours at 37°C and 5% CO 2 , with a final DMSO concentration of 0.1% (v/v). Following treatment, cells were washed two times with 1Â PBS to remove excess compound. Cell pellets were lysed in 1Â PBS with 0.1% NP-40 detergent and 1Â Protease Inhibitor Cocktail (Sigma-Aldrich). Cell lysate protein concentrations were quantified by Pierce BCA Protein Assay (ThermoFisher Scientific) per manufacturer's instruction using a FlexStation 3 Multi-Mode Microplate Reader (Molecular Devices, San Jose, CA). Treatment-free cryopreserved CLL peripheral blood mononuclear cells (PBMCs) isolated from four donors were obtained from Discovery Life Sciences (Huntsville, AL). Samples were thawed, washed once with Gibco RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum, and resuspended in fresh medium. Cells were plated in a six-well tissue culture plate using 9 Â 10 6 cells per well at a concentration of 3 Â 10 6 cells/ml and incubated at 37°C with 5% CO 2 for 2 hours prior to compound treatment. Treatments and cell lysis were performed using the same procedure as was used for MWCL-1 cells.
Enzyme Occupancy Assays. Cell-based BTK and TEC occupancy assays were developed to evaluate drug selectivity between the two kinases in a more physiologically relevant system. An MSD Small Spot High Bind plate (Meso Scale Diagnostics, Rockville, MD) was coated with 35 ml of 10 mg/ml anti-TEC antibody (Sigma-Aldrich) or 35 ml of 1 mg/ml anti-BTK antibody and incubated overnight at 4°C. The following day, the plate was warmed to room temperature, washed three times with 200 ml of 1Â MSD Tris Wash Buffer (Meso Scale Diagnostics), and blocked with 200 ml of 3% MSD Blocker A (Meso Scale Diagnostics) for 1 hour. In a 96-well plate, 25 ml of 2 mg/ml steady-state velocity (for covalent irreversible inhibitors, V s 5 0). From the goodness of fit and statistical analysis data it is clear that the two drugs exhibited a hyperbolic fit (curve-fitting software XLfit4 yielded a goodness of fit with lower x 2 values and F-Test greater than 0.95 in all the cases), indicating that the binding of the two drugs follows a two-step mechanism. The first step includes reversible binding of the inhibitor to the enzyme, which is followed by the second step of covalent bond formation for both BTK and TEC enzymes. Conc., concentration.
cell lysate was incubated with 25 ml of 50 nM proprietary biotinylated probe for 1 hour to allow the probe to bind TEC or BTK kinase not occupied by drug. The MSD plate was then washed three times, and 45 ml of probe-labeled lysate was added to each well and incubated for 2 hours. The plate was then washed three times prior to the addition of 45 ml of MSD SULFO-TAG Streptavidin (Meso Scale Diagnostics) diluted 1:500 in 1Â PBS. After 1 hour of incubation, 150 ml of 1Â MSD Read Buffer (Meso Scale Diagnostics) was added to each well and the plate was read immediately on an MSD Sector S 600 plate reader (Meso Scale Diagnostics). Unless otherwise specified, incubation steps were performed at room temperature with 300 rpm shaking.

Inhibition Kinetics of BTK and TEC by Ibrutinib and
Acalabrutinib. Inhibition of kinases by covalent binding is a two-step process that is driven by affinity with the target followed by inactivation by covalent bond formation. Therefore, the assessment of relative selectivity toward different targets for covalent inhibitors should be evaluated using measurement of both binding affinity and time-dependent inactivation (Strelow, 2017). In the current study, biochemical analyses of enzymatic function were performed and kinetics measuring both of these parameters were derived using the LabChip 3000 Drug Discovery System. Biochemical IC 50 values for kinase inhibition demonstrated BTK and TEC selectivity ratios of 1.0-fold for ibrutinib and 4.2-fold for acalabrutinib (Table 1). These data demonstrated similar binding affinity of ibrutinib for both BTK and TEC kinase, as indicated by their similar K i values: 0.95 6 0.009 nM (BTK) and 1.8 6 0.05 nM (TEC) ( Table 1). Examples of the enzymatic progression curves and hyperbolic fit curves that were used to calculate the kinetic parameters in Table 1 (K i and K inact ) as well as the equations used for the calculations are shown in Fig. 1. The complete set of kinetic data for all determinations is included as Supplemental Material (Supplemental Figs. 1-8; Supplemental Tables 1-8). The rates of enzymatic inactivation (K inact ) of ibrutinib against both BTK and TEC were also similar: 0.011 6 0.0004 (BTK) and 0.013 6 0.0005 (TEC). In contrast, acalabrutinib demonstrated 20-fold higher binding affinity for BTK (8.7 6 0.5) compared with TEC (160 6 37) but a 5-fold faster rate of TEC inactivation (0.012 6 0.001) compared with BTK inactivation (0.0021 6 0.0004). Therefore, despite these large differences in binding and inactivation, selectivity for BTK over TEC, as measured by the K inact /K i ratio with this method, were within 2-fold for the two drugs: a selectivity ratio of 1.5 for ibrutinib and 3 for acalabrutinib (Table 1). This difference in kinetic properties of the two compounds highlights why different approaches are necessary to rank selectivity between reversible and covalent inhibitors. For reversible binding, relative selectivity can be inferred from the IC 50 of affinity analysis alone because potency is driven by affinity in these cases. For covalent inhibitors, the IC 50 does not reflect the combined affinity plus inactivation steps. The apparent IC 50 for covalent inhibitors decreases as a function of time because of the depletion of active enzyme caused by covalent bond formation. Accordingly, this analysis demonstrated how binding affinity measured alone inadequately reflects selectivity among different targets of covalent inhibitors.
Acalabrutinib and Ibrutinib Exhibit Similar Selectivity for BTK Over TEC in MWCL-1 Cells and Human CLL PBMCs. MWCL-1 cells were treated with eight different Fig. 3. Dose-response curves of BTK and TEC occupancy in human CLL PBMCs. Occupancy of BTK (blue lines) and TEC (red lines) is shown for human CLL PMBCs treated with ibrutinib (left) and acalabrutinib (right). Occupancy was assessed at 3 hours of exposure to ibrutinib or acalabrutinib. BTK Versus TEC Selectivity and Relative Kinase Inhibition concentrations of either ibrutinib or acalabrutinib for 0.5, 1, and 3 hours (Fig. 2). The time points were selected to encompass the T max (i.e., the time for drug to reach maximal plasma concentration in patients) of both compounds (Advani et al., 2013;Barf et al., 2017). With very short exposure (0.5 hours), ibrutinib showed slightly higher selectivity for BTK over TEC: when 50% of BTK was occupied, only 25% of TEC was occupied (Fig. 2, left). Acalabrutinib selectivity was not observed after 0.5-hour exposure (i.e., there was no separation for BTK and TEC dose-response curves). BTK over TEC selectivity was minimal for both drugs with 1-hour exposure (Fig. 2, center) across eight concentrations. With 3 hours of treatment, there was modest selectivity indicated by the small separation between the BTK and TEC dose-response curves for both drugs (Fig. 2, right), and this selectivity was nearly equivalent for both drugs. In effect, these cellular results are consistent with the biochemical results; ibrutinib exhibited slightly higher potency on BTK over TEC at 0.5 hours, but similar potency against the two kinases became apparent with longer treatment time ( Table 2). The time-dependent effect of kinase inhibition observed here is not unlike the physiologic activities of these drugs in vivo. Because the drug-binding reaction is irreversible, the amount of drug-bound targets increases with longer duration of exposure. The current data with MWCL-1 cells show that the two drugs have nearly equivalent selectivity between BTK and TEC, and the ranges of selectivity ratios (BTK/TEC) were similarly small (1.05-2.53 and 0.97-2.56 for ibrutinib and acalabrutinib, respectively) throughout the time window of 0.5-3 hours ( Table 2) that is required to achieve maximum occupancy resulting in target inhibition (Barf et al., 2017). Protein content and cellular compositions of PBMCs from patients with CLL may differ from those of the MWCL-1 cells, so we further evaluated the differential selectivity of the two drugs in cryopreserved PBMCs isolated from patients with CLL who were previously untreated. Cellular occupancy assays, which measure the percentage of target kinases bound by drugs, were conducted after each patient sample was treated with either ibrutinib or acalabrutinib for 3 hours (Fig. 3). Despite the potential biological variation among different donors, IC 50 values against BTK and TEC were highly similar for both drugs across four individual patients (Table 3). Ibrutinib was more potent toward both kinases, with IC 50 for BTK ranging from 1.3 to 1.5 nM and TEC ranging from 1.3 to 2.1 nM.
Acalabrutinib was approximately 10-fold less potent against both kinases, with IC 50 for BTK ranging from 7.9 to 11.9 nM and TEC ranging from 10.2 to 11.5 nM. Results of this occupancy assay are consistent with findings from biochemical analyses from the present study and others' work (Bye et al., 2015); ibrutinib is generally a more potent compound than acalabrutinib toward members of the TEC kinase family. However, as observed in MWCL-1 cells, the two drugs' BTK/ TEC selectivity ratios observed in samples from patients with CLL were remarkably similar: 1.32 6 0.27 for ibrutinib and 1.09 6 0.11 for acalabrutinib (Table 3). These data indicate that the two compounds exerted equivalent selectivity for BTK over TEC in a cellular environment despite having the higher potency of ibrutinib against BTK and TEC kinases based on IC 50 values ( Fig. 4; Table 3). In general, the implication of higher potency is that a drug is more reactive toward its targets and less drug is needed to achieve pharmacologic activity, whereas selectivity is a measure of the ratio of on-target/off-target effects, such as BTK over TEC.
Here, the collective data indicate that ibrutinib is a more Fig. 4. Apparent selectivity of ibrutinib and acalabrutinib. Binding affinity data are reported by Barf et al. (2017). Enzyme function assay data are the IC 50 ratios of TEC/BTK, and inactivation kinetics data are the K inact /K i ratios for ibrutinib (blue) and acalabrutinib (red). Cellular occupancy data are mean TEC/BTK selectivity measured by IC 50 ratio for human CLL PMBCs. potent drug than acalabrutinib toward its targets, but acalabrutinib does not offer improved on-target/off-target selectivity.

Discussion
Kinases form an attractive target class for small molecule inhibitors. Specifically, targeting BTK has resulted in therapies for numerous hematologic malignancies that previously had substantial unmet medical needs. However, the high similarity of the ATP binding site on protein kinases, a typical target of inhibition, mounts a continuous challenge for developing selective small molecule inhibitors for this target class. Our understanding of the selectivity of covalent and noncovalent inhibitors is highly dependent on the assessment methods, which have different implications and interpretations for irreversible and reversible binding modes. Crucial to this is the understanding that covalent kinase inhibition is a two-step process involving initial binding (driven by affinity) and time-dependent inactivation (driven by covalent bond formation). Thus, to evaluate biochemical potency and understand target selectivity of covalent inhibitors, both steps must be evaluated.
Published pharmacokinetics for ibrutinib and acalabrutinib report that the two compounds reach peak plasma concentrations between 1 and 2 hours and between 0.6 and 1.1 hours after dosing, respectively (Advani et al., 2013;Byrd et al., 2016). In a recent pharmacokinetic/pharmacodynamic modeling study of a covalent BTK inhibitor, it was shown that target occupancy in B cells and drug plasma exposure reach peak levels at the same time points (Daryaee et al., 2017). C max for ibrutinib at 420 mg once daily (310 nM) was significantly lower than that of acalabrutinib at 100 mg twice daily (1.78 mM) (Byrd et al., 2016;Chen et al., 2018). These data show that direct comparison between the two drugs using in vitro systems should be conducted at concentrations consistent with this 5-fold or greater difference in human exposure levels. Without such considerations, the physiologic effects of these agents cannot be reasonably discerned. In one example of such oversight, Bye and colleagues (2015) argued that acalabrutinib is a less potent inhibitor of TEC phosphorylation based on in vitro analysis using equal concentration of either drug but subsequently discovered that therapeutic doses of acalabrutinib did inhibit TEC based on their own findings from patient samples (Bye et al., 2015)(). These authors concluded that the apparent potency of acalabrutinib toward both BTK and TEC in their in vitro experiments was lower than that observed in patients, because blood samples derived from patients treated with ibrutinib and acalabrutinib exhibited very similar platelet aggregation responses. These inconsistencies highlight the importance of adjusting in vitro exposure according to a drug's therapeutic concentration to obtain experimental results that are informative toward understanding the drug's effects in humans.
The recent high interest in understanding kinase selectivity profiles of various BTK inhibitors stems from a desire to optimize patient outcomes with treatment, minimize adverse events, and understand how off-target effects may be related to such events. For example, bleeding is an adverse event observed in patients treated with BTK inhibitors, with most cases being grade 1 or 2 in severity (Mock et al., 2018). Although clinical data were still being accumulated, an earlier hypothesis based on kinase affinity data proposed that acalabrutinib might lead to lower bleeding rates because of its higher selectivity for BTK relative to TEC (Byrd et al., 2016;Barf et al., 2017;Awan et al., 2019). However, when properly adjusted for clinical exposure, covalent binding mechanism, and enzymatic activity, our data and those of other groups (Bye et al., 2017) now indicate that ibrutinib and acalabrutinib have similar selectivity between BTK and TEC. This similar selectivity is consistent with a recent study suggesting that platelet inhibition and potential hemorrhagic risk predicted by in vitro closure time are likelyclass effects across five BTK inhibitors: ibrutinib, zanubrutinib, acalabrutinib, tirabrutinib, and evobrutinib (Denzinger et al., 2019). Furthermore, recent mechanistic investigations of the effects of ibrutinib and acalabrutinib on platelet functions suggested that polymorphism of drug efflux pumps might sensitize some patients toward platelet aggregation impairment more than others (Series et al., 2019). Because polymorphism in drug efflux pumps is an intrinsic factor, platelet aggregation did not improve when the authors retreated the same patient samples with acalabrutinib. Emerging clinical data now increasingly indicate that the bleeding rates of ibrutinib and acalabrutinib are similar, with more bleeding events reported for patients treated with acalabrutinib; 60% of 99 patients treated with first-line acalabrutinib reported bleeding events of all grades, and two of these patients showed grade 3 events after longer follow-up (median time on study of 33 months) (Byrd et al., 2018).
In summary, our study applies fundamental biochemistry principles to demonstrate how standard methods used to evaluate target selectivity for reversible inhibitors fail to fully characterize irreversible inhibitor selectivity. The current case study of the two covalent inhibitors ibrutinib and acalabrutinib highlights how these results, when performed at physiologic concentrations and in a more relevant cellular context, better inform the understanding of possible off-target clinical observations of bleeding events with these two drugs.