Strategy for Determining the Free Fraction of Labile Covalent Modulators in Plasma Using Equilibrium Dialysis

Christian Leung, Jane R. Kenny, Cornelis E.C.A. Hop, Zhengyin Yan*

Abstract

Determination of free drug fraction (fu) in plasma can be challenging for labile covalent modulators due to the off-target reactivity of chemical warheads to matrix proteins. The resulting poor drug recovery yields low confidence in fu. Two approaches using diluted plasma and low temperature (4 & 20 C) for equilibrium dialysis (ED) have been investigated using covalent modulators including osimertinib, ibrutinib, rociletinib, afatinib, neratinib and acalabrutinib. Our data indicate that stability of covalent modulators in plasma varies in different species, and drug depletion may lead to overestimation of fu if true equilibrium is not reached. Additionally, although ED at low temperature improves the recovery of covalent modulators, the impact of low temperature may lead to underestimate of fu. Overall, ED using diluted plasma is a preferred method because of its faster equilibrium, improved recovery and free of temperature effect on fu. If low temperature ED must be used for extremely labile compounds, precaution must be taken to ensure no temperature dependence of fu in plasma. Nevertheless, an orthogonal ED approach is recommended for labile covalent modulators to confirm the true equilibrium and impact of temperature on fu. Additionally, this strategy can be used for determining fu of other liable compounds. © 2020 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.

Keywords:
Plasma protein binding
PPB
Temperature effect on protein binding Serum albumin a-acid glycoprotein Free fraction
Equilibrium dialysis
Covalent inhibitors
Covalent modulators
Drug-drug interactions
Labile compounds
Unstable drugs

Introduction

Covalent modulation of biochemical pathways has become an increasingly important strategy in drug discovery,1,2 and success of such strategy is evident by recent regulatory approvals of covalent modulators such as ibrutinib3 and osimertinib.4 Covalent modulators normally contain a binding “anchor” or reversible modulator and a chemically reactive moiety or “warhead” that is specifically designed to react with a corresponding amino acid side chain of an intended pharmacological target. Mechanistically, covalent modulation can be divided into two separate steps. Initially, the binding “anchor” reversibly associates with the pharmacological target to bring the chemical “warhead” within close proximity of a specific amino acid of the protein; subsequently, the chemical “warhead” forms a covalent bond with the side chain of the targeted amino acid to inactivate the biological function of the target.1,2
So far, several different types of electrophilic warheads including acrylamide, epoxide, aziridine, ester, ketone and nitrile have been deployed to specifically target various side chains of corresponding amino acids including cysteine, lysine, serine, tyrosine, glutamate and aspartate.2 Compared to traditional reversible modulators, covalent modulators are presumed to exert higher biochemical potency and/or selectivity due to the covalent bonding between the chemical “warhead” and its corresponding amino acid side chain of the intended target. Additionally, covalent modulators may render longer duration of action than traditional reversible modulators since the pharmacological function of the target remains inactivated or impaired until the covalently modified protein is degraded and native protein is newly synthesized in the body.
Although the mechanism of action is quite different for covalent modulators compared to traditional reversible modulators, the free drug hypothesis certainly remains true in which only free covalent modulators are available in the body to bind and subsequently react with the intended target to produce a desired pharmacological effect. Therefore, measurements of the fraction of unbound drug (fu) remain equally important for covalent modulators because plasma protein binding (PPB) is required as a very critical pharmacokinetic (PK) parameter for establishing pharmacokinetic/ pharmacodynamic (PK/PD) relationships, predicting clinical dosages, and assessing drug-drug interaction (DDI) potential as well as toxicity of drug candidates.5,6
There are different methodologies developed for measuring fu of traditional drugs with diverse physicochemical properties,7e9 and the most frequently used one is equilibrium dialysis (ED) because of its high-throughput and relatively low non-specific binding compared to others such as ultra-centrifugation (UC) and ultrafiltration (UF).10,11 However, regardless of PPB methodology used, measurements of fu can potentially be problematic for covalent modulators simply because these reactive warheads may also react with corresponding amino acid residues present in plasma proteins, leading to poor recovery and low confidence in fu value. For example, as shown Scheme 1, the most common acrylamide “warhead” can chemically react with the thiol group of free cysteine, glutathione (GSH) and plasma proteins such as serum albumin. Additionally, some acrylamide modulators can also form GSH conjugates catalyzed by glutathione transferases (GST) in plasma.12 Together these reactions can result in inaccurate fu if true equilibrium is not completely achieved in the assay. Additionally, these reactions can potentially cause drug depletion in plasma, leading to analytical challenges in PPB measurements for highly labile covalent modulators. For example, fu value of osimertinib, a highly labile covalent modulator for treating non-small-cell lung carcinomas, could not be determined experimentally; instead it was predicted using an in silico approach.12
Although chemical inhibitors, such as phenylmethylsulfonyl fluoride (PMSF) and 4-(2-aminoethyl) benzenesulfonyl fluoride (AEBSF), have been used to stabilize drug degradation mediated by hydrolytic enzymes in plasma,13 the same strategy does not work for covalent modulators, simply because depletion of covalent modulators in plasma can be mediated by both enzymatic and chemical reactions (Scheme 1). For instance, chemical inhibitors of GST may be able to block enzymatic reactions involved in drug depletion in plasma, but they cannot stop or slow down chemical reactions with the thiol group of cysteine, albumin and glutathione.
Also, chemical inhibitors are rarely effective against enzymes in all different species. Additionally, it is not certain that any given inhibitor will not compete with covalent modulators for the binding sites of serum albumin or alpha-1-acid glycoprotein. More importantly, it is not realistic to expect that all depletion mechanisms are well characterized and fully understood for each covalent modulator especially at the early stage of discovery. Therefore, alternative strategies are needed to improve recovery of labile covalent modulators in plasma and ensure the reliability of PPB data.
In the present study, we selected seven commercially available covalent modulators (Scheme 2) representing three different amide warheads such as unsubstituted acrylamides (osimertinib, ibrutinib, rociletinib and EGFX), substituted acrylamide (afatinib and neratinib) and butynamide (acalabrutinib) to systematically investigate the impact of different assay conditions on both drug recovery and PPB determinations. It is anticipated that the data generated can help strategize the assay design specifically for measuring fu values of labile covalent modulators (see Scheme 3).

Materials & Methods

Chemicals and Reagents

Pooled human, monkey (Cynomolgus), mouse (CD-1), and rat (Sprague Dawley) plasma were purchased from BioIVT (Westbury, NY); HPLC grade water was from J.T. Baker (Center Valley, PA); HPLC grade acetonitrile was from EMD Millipore (Billerica, MA). The Rapid Equilibrium Dialysis (RED) devices with inserts were obtained from Thermo Fisher Scientific Inc. (Rockford, IL). All covalent modulators were from the Genentech Compound Management System. All other chemicals and reagents were of analytical grade and were obtained from Sigma-Aldrich (St. Louis, MO) unless otherwise specified.

PPB Measurement Using Equilibrium Dialysis

All plasma protein binding experiments were performed in triplicate (n ¼ 3) using a Single-Use RED Device as previously described.14 Initially, stock solutions of test compounds were prepared in dimethyl sulfoxide (DMSO). The frozen plasma under study was first thawed, and then gradually adjusted to pH 7.4 with 0.5 M phosphoric acid. Fresh human plasma did not involve a thawing step. For the conventional ED method, individual drugs were spiked in plasma to achieve a final concentration of 5 mM unless specifically indicated. Initially, aliquots of 500 mL of phosphate buffer saline (133 mM PBS) were added to the receiver wells of the RED device. Subsequently 300 mL of the drug-plasma mixtures were transferred to the corresponding donor wells. The RED device was sealed with a gas permeable membrane, and then placed in a shaking incubator (450 rpm, VWR Symphony™) for a designated amount of time at 37 C with 5% CO2. At the end of incubation, aliquots of 30 mL samples were taken out of the RED device and matrix equalized with an equal volume of plasma or buffer. The resulting samples were then immediately quenched with 3-vol of ice-cold acetonitrile containing both propranolol and labetalol as internal standards (IS). After shaking for 15 min at 500 rpm on a Thermo Scientific Compact Digital MicroPlate Shaker, all samples were then centrifuged at 3700 rpm for 15 min (Beckman Coulter Allegra X 12R) to precipitate plasma protein. Supernatants of the centrifuged samples were collected then diluted with an equal volume of water prior to LC-MS/MS analysis.
For the dilution ED method, thawed plasma was first diluted with PBS (pH 7.4) to the desired concentration of plasma (10%), and then test compounds were added to a final concentration of 5 mM before being aliquoted to the RED device for equilibrium dialysis. For PPB measurements at lower temperature, the RED device was placed on a shaker housed in a refrigerator (4 C) or on a lab bench (room temperature, ~20 C) instead of an incubator at 37 C.

Recovery in Plasma

Recovery determinations were performed in triplicate (n ¼ 3). The plasma under study was thawed, and then gradually adjusted to pH 7.4 with 0.5 M phosphoric acid. Fresh human plasma did not require a thaw step. Individual drugs were spiked in plasma to achieve a final concentration of 5 mM. Aliquots of 30 mL of the drugplasma mixtures were taken and immediately stored at 80 C for drug recovery analysis at the end of PPB assay. The remainder of the drug-plasma mixtures were placed in the same conditions as the PPB assay being tested accounting for the temperature, shaking, CO2, and length of incubation. At the end of incubation, aliquots of 30 mL were taken from the previously incubated drug-plasma mixtures. Both aliquots of the drug-plasma mixtures from preincubation and post-incubation were matrix equalized with an equal volume of buffer, then immediately quenched with 3-vol ice cold acetonitrile containing both propranolol and labetalol as internal standards. After shaking for 15 min at 500 rpm on a Thermo Scientific Compact Digital MicroPlate Shaker, all samples were then centrifuged at 3700 rpm for 15 min (Beckman Coulter Allegra X 12R) to separate the plasma protein. Supernatants of the centrifuged samples were collected then diluted with an equal volume of water prior to LC-MS/MS analysis.

LC-MS/MS Analysis

LC-MS/MS analysis was performed using a 6500þ QTRAP® mass spectrometer coupled with a TurboIonSpray® ESI ion source (AB SCIEX, Redwood City, CA) and Shimadzu Nexera 2 UPLC (Kyoto, Japan). Chromatographic separation of all analytes was achieved using a Kinetex® C18 column (30 2.1 mm, 100 Å, 2.6 mm particle size) (Torrance, CA) along with mobile phase A consisting of 0.1% formic acid in HPLC grade water and mobile phase B consisting of 0.1% formic acid in acetonitrile. A generic LC gradient was used for all analytes where the flow rate was set to 1 mL/min, the run time at 2 min and the LC gradient as follows: 3% B for the first 0.1 min, ramped up to 95% B from 0.10 to 0.35 min, remained constant at 95% B for 0.50 min, and then decreased to 3% B within 0.65 min.

Calculation of fu

Fraction of unbound drug was calculated using Equation (1). The peak area ratios of test compound to IS in buffer and plasma wells were obtained respectively using LCeMS/MS as described above. For diluted plasma, the measured fraction of unbound drug (fu,d) was calculated using Equation (1) and then extrapolated to the undiluted fu using Equation (2), where D is the dilution factor of plasma.15
Recovery of drug (%) in plasma was calculated according to Equation (3). The peak area ratios of test compound in plasma aliquoted before equilibrium dialysis (T0) and plasma samples at the end of dialysis (T) were obtained using LCeMS/MS as described above and normalized to IS. Plasma samples at T0 were kept frozen at 80 C until dialysis ended. It should be noted that the reported recovery is different from mass balance, as drug loss could have occurred after mixing drug-plasma and before quenching with icecold acetonitrile for T0 samples and this loss was not accounted for in the recovery calculation herein.

Results

Impact of Assay Conditions on Recovery

To cover the worst-case scenario in the ED assay, all equilibrium incubations were performed for 24 h regardless of stability profile and fu range of the covalent modulator. To investigate drug loss resulting from plasma frozen and thawing during the assay, drugs were mixed with plasma, and resulting drug-plasma mixture was split into two identical portions; one portion was immediately crashed with ACN (with IS) and subsequently centrifuged to collect the supernatants; the second portion was frozen overnight and then thawed, and subsequently processed to collect the supernatants. Both sample sets were analyzed by LC-MS/MS. As shown in Table 1, either no or minor differences (<11%) were observed in human and mice between those two sample sets for all seven covalent modulators, suggesting that drug loss caused by frozen/thaw is not significant.
Drug recovery was initially determined at 37 C in plasma from human, monkey, rat and mouse respectively, and subsequent assessments mainly focused on mouse and human. As shown in Table 2, at the normal assay conditions (37 C & undiluted plasma), a majority of acrylamide modulators exhibited distinct species differences in plasma stability, with the exception of osimertinib. Osimertinib is highly labile in all species with recovery ranging from less than 0.1%e9%. The most striking species difference was observed for acalabrutinib, the only butynamide on our compound list, which is completely stable in both human and monkey plasma but extremely labile in both mouse and rat. It appears that substituted acrylamides are not consistently more stable than unsubstituted acrylamides, depending on species and individual modulators. For example, afatinib, one of two substituted acrylamides tested, is more stable in both human and monkey plasma compared to unsubstituted acrylamides including rociletinib, ibrutinib, osimertinib and EGFX; whereas, neratinib, another substituted acrylamide, is more stable than unsubstituted osimertinib and EGFX but less stable than rociletinib in human plasma and also less stable than ibrutinib in both human and monkey plasma. In rat plasma, substituted afatinib and neratinib are more stable than unsubstituted osimertinib and EGFX, but unsubstituted ibrutinib is the most stable one in mouse plasma. Clearly, more research is needed to further characterize the depletion mechanisms of various covalent modulators in plasma in order to help design and optimize ADME properties of covalent modulators, given the fact that the same mechanisms may play an important role in overall drug clearance in vivo.
As shown in Table 2, all covalent modulators are highly stable at 4 C in plasma from human and mouse despite complex depletion mechanisms. Even for the most labile covalent modulators such as osimertinib, EGFX and acalabrutinib, their recovery is greater than 65% in both species. Despite all enzymes in plasma are largely inactive, certain levels of instability were still observed at 4 C for osimertinib in human (34% depletion) and acalabrutinib in mouse (~30% depletion). Presumably, depletion of both acalabrutinib and osimertinib at 4 C is probably due to the remaining chemical reactivity in plasma although more mechanistic studies are warranted in the future.
In general, labile covalent modulators are more stable at room temperature (~20 C) compared to 37 C, regardless of plasma species (Table 2). For example, recovery of EGFX and osimertinib, the two most labile modulators in human plasma, increases from less than 1% to 48% and 14% respectively. Also, recovery of osimertinib and acalabrutinib in mouse plasma improves from less than 1% to 78% and 27% respectively. Similar increases in recovery were observed at room temperature for all labile covalent modulators in rat and monkey plasma (data not shown). Overall, recovery for all labile covalent modulators is generally lower in all species at room temperature compared to that at 4 C, likely due to slower rates of both chemical and enzymatic reactions at a lower temperature.
Diluted plasma has been commonly used in PPB assay to determine fu values of highly bound drugs.8,14,16 As shown in Table 2, recovery of all labile modulators such as ibrutinib, neratinib and osimertinib at 37 C increases from 37%, 22% and <0.1% in undiluted human plasma to 67%, 48% and 14% in diluted plasma (10%), respectively. In mouse plasma, recovery at 37 C of labile covalent modulators such as neratinib, EGFX, osimertinib, and acalabrutinib increases from 34%, 15%, 1% and <0.1% in undiluted to 74%, 47%, 5% and 12% in diluted plasma (10%), respectively. Obviously, for very highly labile compounds such as osimertinib and acalabrutinib, recovery at 37 C in diluted plasma is still relatively low compared to that at 4 C. Similar increases in recovery were observed in rat and monkey when diluted plasma was used (data not shown). More pronounced increase in recovery was found in diluted plasma (10%) from all species at room temperature (~20 C). As shown by the representative data in Table 2, all labile covalent modulators showed recovery greater than 65% in human and 70% in mouse when 10% plasma was used at room temperature.
Low recovery in PPB assays can be problematic analytically if drug levels in assay samples are below the lower limit of quantitation (LLOQ) in LC-MS/MS analysis. As shown in Fig. 1A, osimertinib in human plasma sample before dialysis showed peak intensity greater than 5 105 in LC-MS/MS detection, but it dropped to about 8 103, less than 2% of that in the original plasma sample after 24-hr incubation (Fig. 1B). Also, the level of free osimertinib in the buffer chamber is significantly lower (Fig. 1C). Obviously, low signal-to-noise ratios can lead to more significant data variability and inaccurate fu.

Determination of fu for Covalent Modulators Under Various Assay Conditions

The fu values of all seven covalent modulators were determined in plasma from different species under various conditions, and representative data from human and mouse are summarized in Table 3. In human plasma, highly comparable fu values were obtained at 37 C, 20 C and 4 C respectively for several covalent modulators including acalabrutinib, afatinib, neratinib, ibrutinib and rociletinib. The results seem to indicate that, for those covalent modulators, the effect of temperature on fu in human plasma is not significant. Consistently, for those same compounds, their fu values directly measured in undiluted plasma are also highly comparable to those extrapolated from 10% human plasma at both 37 C and 20 C. Negligible effect of temperature on fu was observed for osimertinib at 20 C (0.038) and 4 C (0.024), and both are consistent with the values extrapolated from 10% human plasma at 37 C (0.030) and 20 C (0.030), respectively.
In contrast, remarked temperature effect on fu was observed for EGFX at various assay conditions. As shown in Table 3, fu of EGFX directly measured in human plasma (0.02) at 37 C is in good agreement with the value (0.028) extrapolated from diluted plasma at 37 C. Similarly, identical fu values were obtained at 20 C using both un-diluted (0.003) and diluted human plasma (0.003). Both directly measured and extrapolated fu of EGFX at 37 C are significantly higher than their corresponding values obtained at 20 C; additionally, directly measured fu (0.0004) at 4 C is about 50 and 7fold lower compared to the values obtained at 37 C and 20 C, respectively. Consistently, a similar temperature effect on fu was observed for EGFX in mouse plasma. As shown in Table 3, fu of EGFX directly measured at 4 C (0.0003) is about 7 and 10-fold lower compared to the values obtained at 20 C (0.002) and 37 C (0.003), respectively. Since EGFX showed complete recovery in mouse plasma at both 4 C (101%) and 20 C (102%) (Table 2), it is reasonable to conclude that the temperature effect on fu observed is not likely caused by drug depletion at two different temperatures. This conclusion is substantiated by the similar results obtained in human plasma which show a more than 7-fold difference in fu for EGFX between 4 C and 20 C (Table 3).
It appears that increase of fu with temperature is more commonly seen in mouse than human plasma. As shown in Table 3, ibrutinib, osimertinib and rociletinib all showed significantly lower fu (4e20 folds) in mouse plasma at 4 C compared to their corresponding values obtained at either 20 C or 37 C, and increase of fu with temperature is consistent when comparing directly measured PPB data at 4 C to those values obtained at 37 C regardless of diluted or undiluted plasma. Also, the same trend in fu was observed for compounds with low recovery such as EGFX (15%) and good recovery in mouse plasma such as ibrutinib (94%). Additionally, the effect of temperature on fu was also observed for multiple covalent modulators across different species (Supplemental Material I), suggesting that the impact of temperature on fu is compound specific and species dependent.

Discussion

Covalent modulators are a class of new chemical modality aiming at those challenging drug targets. Despite their advantages such as higher potency and longer duration of action compared to reversible modulators, covalent modulators represent some unique challenges in assessing their DMPK properties due to chemical reactivity of the “warhead”. For plasma protein binding specifically, both chemical and enzymatic reactions in plasma may alter the equilibrium in the ED assay, depending on their overall reaction rate, the extent of non-specific binding, and diffusion rate of the covalent modulators across the semi-permeable membrane. Additionally, drug depletion caused by these reactions can potentially lead to analytical challenges for highly labile covalent modulators which can become more serious when PPB measurements at low drug concentrations are required, or a labile covalent modulator is either poorly ionized in the ion source or highly bound to plasma (fu < 0.01). Therefore, it is highly desired to stabilize labile covalent modulators in plasma to ensure high confidence in PPB data.
Our data clearly show that low temperature can stabilize labile covalent modulators in plasma (Table 2). However, the effect of temperature on plasma protein binding was observed in this study for some covalent modulators (Table 3) which is consistent to observations previously reported in literature.17e19 Kodama et al. have reported significant differences in free fraction for phenytoin in human plasma between 25 C and 37 C.17 Igari et al. also have observed that binding of thiopental to rat albumin and plasma is sensitive to temperature between 4 C and 37 C, whereas the impact of temperature is not noticeable in bovine albumin.18 The similar effect of temperature on fu was reported by Deltombe et al. for several uraemic toxins including hippuric acid, indole-3-acetic acid, indoxyl sulphate, and p-cresylsulphate.19 In contrast, using a larger set of compounds, Ryu et al. have recently shown that, all nineteen drugs, except for indomethacin, showed comparable fu values at 4 C and 37 C, and they have concluded that the size of compound sets, differences in PPB methodology (ED vs UC) and assay conditions are the confounding factors contributing to the discrepancies in the effect of temperature on drug plasma binding.20 However, our data clearly indicates that the effect of temperature on fu is not an assay related artifact, given the fact that it was observed for multiple compounds in a species-specific manner when the same PPB assay was used. Instead, it is possible that the effect of temperature on fu is compound specific and species related, and it happens to be more common for labile covalent modulators compared to traditional stable drugs previously tested by Ryu et al.20
Conceptually, a higher level of thermodynamic energy tends to weaken inter-molecular interactions and increase dissociation constant (Kd) between drug molecules and their binding partner such as albumin in plasma. Assuming a single binding site involved, increase of Kd translates to a higher fraction of free drug (fu) according to Equation (4) when the PPB assay is performed at an elevated temperature.
It is important to point out that, for a given drug, the effect of temperature on fu may vary significantly, depending on the magnitude of Gibbs free energy change [D(DG) ¼DGT2-DGT1)] at two different temperatures (37 C vs 4 C). For example, it is estimated that, for a 70% bound compound, a shift of fu from 0.30 to 0.20 requires Gibbs free energy change [D(DG)] of 690 cal/mol as temperature decreases from 37 C to 4 C (Supplemental Material II), which usually is considered within the range of normal experimental errors.8 Therefore, Gibbs free energy change must be higher as the assay temperature changes in order to incur a more marked shift in fu in plasma binding.
It is generally recognized that albumin has several distinct drug binding pockets, and many drugs can bind to more than one pocket primarily via hydrophobic interactions.21 Thus, it is possible that, for some drugs, hydrophobic interactions at multiple binding sites collectively are sufficiently strong to sustain the Gibbs free energy change (DGT2-DGT1) as long as the assay temperature maintains within a certain range (4e37 C), and as a result, the effect of temperature on fu is largely negligible as Ryu et al. have recently shown.20 Alternatively, this unique binding characteristic of albumin allows some drugs to be able to switch between different binding pockets without completely dissociating from the protein as the PPB assay temperature changes. In other words, these binding pockets of albumin may function as a “binding cushion” which affords some drugs to remain in bound as long as the assay temperature fluctuates within a certain range (4e37 C), and such “binding cushion” may help maintain the degree of disorder and thus minimize entropy change (DS) and Gibbs free energy as well (DG ¼ DH-T*DS). As a result, the effect of temperature on fu is insignificant.20 In contrast, if a particular drug just binds loosely to a single pocket without “binding cushion” or the outside of those binding pockets of albumin via weaker intermolecular interactions such as hydrogen bonding or ion pairing compared to hydrophobic interactions,22 the drug can easily dissociate from albumin (higher entropy) as temperature increases, which can lead to a more pronounced shift in fu as previously observed for phenytoin,17 thiopental,18 uraemic toxins19 and indomethacin.20
As stated by the “free drug theory”, only unbound drug molecules are pharmacologically active. When it applies to drug plasma binding assay specifically, only unbound drug molecules are available for either enzymatic or chemical reactions in the incubation mixture. For labile covalent modulators, since their instability is driven by chemical or enzymatic reactions of unbound molecules, it is reasonable to postulate that those compounds bind loosely to either a single binding pocket via weak hydrophobic interaction or the outside of these binding pockets. As a result, the temperature dependence of fu is more likely to occur for labile covalent modulators as observed in the current study compared to traditional stable drugs previously tested by others.20
It is generally accepted that, for stable drugs, recovery in the ED assay may not necessarily impact the reliability of fu as long as true equilibrium is reached,23 given the fact that low recovery is usually caused by high non-specific binding to the assay device, a reversible process that can potentially be saturated during equilibrium dialysis. Incontrast,depletionofcovalentmodulatorsinplasmaisacontinuous and irreversible process consisting of both chemical and enzymatic reactions.Dependingon the overall rateof reactions,degradation may eventuallyresultindrugdepletionintheplasmachamber,leadingtoa reversal of drug diffusion across the semipermeable membrane from the buffer to plasma. As a result, fu will be inflated if equilibrium is not re-established in a timely fashion. This hypothesis appears consistent with the time course of PPB measurements for osimertinib, a labile compound in monkey plasma. As shown in Fig. 2, two fu values of osimertinib measured from diluted (0.038 ± 0.002) and undiluted monkey plasma (0.040 ± 0.003) respectively converged after 6 h of dialysis, suggesting that true equilibrium was momentarily reached under both assay conditions. At later time points of the assay, fu of osimertinib remains steady up to 24 h in diluted plasma (0.047 ± 0.007) but increases steadily with incubation time in undiluted plasma (0.066 ± 0.004) as its recovery continuously dropped, suggesting that on-going drug depletion in plasma probably altered the equilibrium previously reached at the earlier time point (6 h). In contrast, true equilibrium in diluted plasma was maintained presumably because the diffusion rate across the semipermeable membrane is faster compared to that in undiluted plasma. The similar impactof low recovery on fu was also observed for acalabrutinib. Even though acalabrutinib is highlyliablein mouseplasmawith recoveryof less than 0.1% (Table 2), it was still possible to determine fu in undiluted plasma due to high sensitivity of LC-MS/MS method. However, as shown in Table 3, the free fraction (0.739 ± 0.097) measured in undiluted mouse plasma is significantly higher than all values measured under other assayconditions usingeither undiluted plasma at4C(fu ¼0.231±0.032;recovery:72%)and20C(fu ¼0.415±0.017; recovery: 27%), or 10% plasma at 37 C (fu ¼ 0.289 ± 0.026; recovery: 12%) and 20 C (fu ¼ 0.278 ± 0.018; recovery: 72%).
It is apparent that, for labile covalent modulators, accurate determinations of PPB require reasonable compound recovery to ensure both sufficient drug levels for analytical detection and true equilibrium achieved in the assay. Therefore, an orthogonal strategy using two different ED methods is proposed as described in Scheme 2. For a given covalent modulator, a regular PPB assay is first performed at 37 C using 24 h incubation to ensure true equilibrium. However, if the compound shows low recovery (<5%) in the regular ED assay or its peak intensity is too low for reliable LC-MS/MS quantitation, a bi-directional ED experiment using diluted plasma (10%) at 37 C for 24 h is subsequently conducted to simultaneously determine two fu values from plasma to buffer and from buffer to plasma, respectively.14 Plasma dilution not only facilitates drug diffusion across semipermeable membrane but also increases recovery as well as free drug concentration in assay samples to ensure sufficient sensitivity in LC-MS/MS analysis. For most covalent modulators, two fu values extrapolated from diluted plasma are expected to converge if true equilibrium is reached in the bi-directional ED assay.14 For a highly labile compound, it is still possible that recovery of the compound in diluted plasma is too low to reliably determine fu values in the bi-directional ED assay; consequently, low temperature ED assay can be conducted for 24 h to determine fu values at 4 C with undiluted plasma and 20 C with diluted plasma, respectively. It is important that two fu values measured at two different temperatures are comparable to confirm no significant temperature effect on fu. Otherwise, further plasma dilution (2e5%) may be considered in the ED assay unless binding saturation is a real concern.
The bi-directional ED method using diluted plasma can be easily adopted without any concerns about underestimating fu, but it may not work for extremely labile covalent modulators. It is also important to note that the diluted ED may be the best option for labile covalent modulators with high plasma binding (>99%) since fu values of highly bound compounds are too low to be reliably determined using the routine ED method.8 To ensure data quality, the bi-directional ED, an orthogonal ED approach, is preferred to confirm true equilibrium, given the possibility that drug depletion can potentially disrupt previously established equilibrium as demonstrated in the case of osimertinib (Fig. 2). However, one must realize that too much plasma dilution may raise the possibility of binding saturation especially for compounds with high AAG binding. In those cases, ultra-filtration should be assessed and utilized for PPB measurements if non-specific binding of the drug to apparatus is not significant. Additionally, in-silico approaches can be investigated as an alternative for highly labile covalent modulators24; alternatively, one may consider using the fu value measured at the initial phase to estimate the equilibrium fu by mathematic modeling as described previously by Wenlock et al.25
As demonstrated in the present study, low temperature ED can be very effective for those highly labile compounds, but it may potentially under-estimate fu for some covalent modulators such as EGFX (Table 3) since drug binding in vivo occurs at physiological temperature of 37 C. Additionally, drug diffusion rate across the semipermeable membrane is slower at 4 C, and therefore, ED at two different temperatures, an orthogonal ED approach, is proposed to ensure the data quality. However, accommodation of the assay device at low temperature is inconvenient, and compounds with very low solubility may still be problematic due to the potential effect of low temperature on aqueous solubility.

Conclusions

Covalent modulators represent a new and unique therapeutic class which primarily aims at highly valuable but challenging drug targets. Increasing interest in discovering and developing covalent modulations highlights the need to strategize our PPB methodology in order to reliably determine fu values of labile covalent modulators. To the best of our knowledge, the present study is the first to systematically evaluate ED assay conditions specifically for labile covalent modulators. Our results indicate that, for highly labile covalent modulators, low recovery can potentially lead to overestimation of fu if true equilibrium is not reached in the standard ED assay. Additionally, our data clearly demonstrate that, although ED at low temperatures (4 C & 20 C) improves recovery of labile covalent modulators, their fu values could potentially be underestimated significantly, an important observation which differs from that of traditional stable drugs previously published in literature by others.20 In conclusion, an orthogonal ED strategy is advised to determine fu for labile covalent modulators in order to ensure PPB data reliability, and such strategy can be applied to liable compounds other than covalent modulators.

References

1. Baillie TA. Targeted covalent inhibitors for drug design. Angew Chem. 2016;55: 13408-13421.
2. Ghosh AK, Samanta I, Mondal A, Liu WR. Covalent inhibition in drug discovery. ChemMedChem. 2018;14:889-906.
3. de Claro RA, McGinn KM, Verdun N, et al. FDA Approval: ibrutinib for patients with previously treated mantle cell lymphoma and previously treated chronic lymphocytic leukemia. Clin Cancer Res. 2015;21:3586-3590.
4. Greig SL. Osimertinib: first global approval. Drugs. 2016;76:263-273.
5. Schmidt S, Gonzalez D, Derendorf H. Significance of protein Neratinib binding in pharmacokinetics and pharmacodynamics. J Pharm Sci. 2010;99:1107-1122.
6. Liu X, Wright M, Hop CE. Rational use of plasma protein and tissue binding data in drug design. J Med Chem. 2014;57:8238-8248.
7. Banker MJ, Clark TH. Plasma/serum protein binding determinations. Curr Drug Metab. 2008;9:854-859.
8. Di L, Breen C, Chambers R, et al. Industry perspective on contemporary proteinbinding methodologies: considerations for regulatory drug-drug interaction and related guidelines on highly bound drugs. J Pharm Sci. 2017;106:34423452.
9. Kalvass JC, Phipps C, Jenkins GJ, et al. Mathematical and experimental validation of flux dialysis method: an improved approach to measure unbound fraction for compounds with high protein binding and other challenging properties. Drug Metab Dispos. 2018;46:458-469.
10. Zamek-Gliszczynski MJ, Ruterbories KJ, Ajamie RT, et al. Validation of 96-well equilibrium dialysis with non-radiolabeled drug for definitive measurement of protein binding and application to clinical development of highly-bound drugs. J Pharm Sci. 2011;100:2498-2507.
11. Ye Z, Zetterberg C, Gao H. Automation of plasma protein binding assay using rapid equilibrium dialysis device and Tecan workstation. J Pharm Biomed Anal. 2017;140:210-214.
12. Dickinson P, Cantarini M, Collier J, et al. Metabolic disposition of osimertinib in rat, dog, and man: insights into a drug designed to bind covalently to a cysteine residue of EGFR. Drug Metab Dispos. 2016;44:1201-1212.
13. Fung EN, Zheng N, Arnold ME, Zeng J. Effective screening approach to select esterase inhibitors used for stabilizing ester containing prodrugs analyzed by LC-MS/MS. Bioanalysis. 2010;2:733-743.
14. Chen Y, Kenny J, Wright M, Hop C, Yan Z. Improving confidence in the determination of free fraction for highly bound drugs using bidirectional equilibrium dialysis. J Pharm Sci. 2019;108:1296-1302.
15. Kalvass JC, Maurer TS. Influence of nonspecific brain and plasma binding on CNS exposure: implications for rational drug discovery. Biopharm Drug Dispos. 2002;23:327-338.
16. Riccardi K, Cawley S, Yates PD, et al. Plasma protein binding of challenging compounds. J Pharm Sci. 2015;104:2627-2636.
17. Kodama H, Kodama Y, Itokazu N, Shinozawa S, Kanemaru R, Sugimoto T. Effect of temperature on serum protein binding characteristics of phenytoin in monotherapy paediatric patients with epilepsy. J Clin Pharm Ther. 2001;26: 175-179.
18. Igari Y, Sugiyama Y, Awazu S, Hanano M. Interspecies difference in drug protein binding-temperature and protein concentration dependency: effect on calculation of effective protein fraction. J Pharm Sci. 1981;70:1049-1053.
19. Deltombe O, Dhondt A, Van Biesen W, Glorieux G, Eloot S. Effect of sample temperature, pH, and matrix on the percentage protein binding of proteinbound uraemic toxins. Analyt Methods. 2017;9:1935-1940.
20. Ryu S, Novak JJ, Patel R, Yates P, Di L. The impact of low temperature on fraction unbound for plasma and tissue. Biopharm Drug Dispos. 2018;39:437-442.
21. Peters T. All about Albumin: Biochemistry, Genetics and Medical Applications. San Diego: Academic Press; 1996.
22. Atkins P, de Paula J. Physical Chemistry for the Life Sciences. Oxford, UK: Oxford University Press; 2006.
23. Di L, Umland JP, Trapa PE, Maurer TS. Impact of recovery on fraction unbound using equilibrium dialysis. J Pharm Sci. 2012;101(3):1327-1335.
24. Watanabe R, Esaki T, Kawashima H, et al. Predicting fraction unbound in human plasma from chemical structure: improved accuracy in the low value ranges. Mol Pharmacol. 2018;15(11):5302-5311.
25. Wenlock MC, Barton P, Austin RP. A kinetic method for the determination of plasma protein binding of compounds unstable in plasma: specific application to enalapril. J Pharm Biomed Anal. 2011;55(3):385-390.