DRUG METABOLISM AND PHARAMACOKINETICS

Characterization of covalent binding of tyrosine kinase inhibitors to plasma proteins Xiaoyun Liu, Dan Feng, Mingyue Zheng, Yongmei Cui, Dafang Zhong

To appear in: Drug Metabolism and Pharmacokinetics

Published by Elsevier Ltd on behalf of The Japanese Society for the Study of Xenobiotics.

Characterization of covalent binding of tyrosine kinase inhibitors to plasma proteins

Xiaoyun Liu a,b, 1, Dan Feng a,c, 1 , Mingyue Zheng a,b, Yongmei Cui c, Dafang Zhong a,b, *
a State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica,

Chinese Academy of Sciences, Shanghai 201203, China

b University of Chinese Academy of Sciences, Beijing 100049, China

c Department of Chemistry, Shanghai University, Shanghai 200444, China

Abstract:

Eight covalent tyrosine kinase inhibitors (TKIs) were investigated to determine the characteristics of their covalent binding to plasma proteins. The data revealed that their covalent binding to plasma proteins is of species difference. In addition to the reports on neratinib and pyrotinib, osimertinib, alflutinib, AST5902, and ibrutinib were confirmed to covalently bind to the Lys-190 of human serum albumin (HSA). Molecular docking was used to simulate the binding mode of TKIs to HSA. The results exhibited the non-covalent interactions between covalent TKIs and HSA, which stabilize the TKIs-HSA complex and explain the selectivity of covalent binding. The t1/2 values of TKIs that are covalently bound to HSA or human plasma proteins were studied in vitro, and the features highly correlated with the t1/2 were determined by quantitative calculations and linear modeling. Reversibility of the covalent binding and the factors affecting the process of reversibility were evaluated. In conclusion, acrylamide moiety of covalent TKIs can covalently bind to lysine residue of HSA, most of which were determined to be Lys-190. The covalent binding is of species difference, especially between animal and human. Except for osimertinib, covalent binding between TKIs and HSA are reversible.

Key words: Covalent tyrosine kinase inhibitors; Covalent binding; Human serum albumin; Molecular docking; Quantitative calculations; Linear modeling; Species difference

1. Introduction

The human genome encodes a total of 538 protein kinases, which play important roles in cell signal transduction, metabolism, and apoptosis [1]. The overexpression, imbalances or mutations in protein kinases can cause several diseases, especially developmental and metabolic disorders and cancer [2, 3]. To date, the FDA approved 52 small molecule protein kinase inhibitors and 46 of them are anti-tumor drugs [4]. Kinase inhibitors can be divided into non-covalent and covalent inhibitors according to their interactions with target proteins. Covalent inhibitors are designed to combine the backbone of the non-covalent inhibitor with an electrophilic Michael acceptor (warhead). The former binds the target pocket with non-covalent interactions and the latter forms covalent binding with the exposed cysteine residue [5, 6]. Compared with non-covalent tyrosine kinase inhibitors (TKIs), covalent TKIs have advantages of high potency, extended duration of action, increased therapeutic index, and low dosage [7-9].
Currently, the FDA approved 7 covalent TKIs for cancer therapy, including acalabrutinib, afatinib, dacomitinib, ibrutinib, neratinib, osimertinib, and zanubrutinib. South Korea approved olmutinib, and China approved pyrotinib and almonertinib. Except for acalabrutinib, that has propynamide warhead, other drugs have an acrylamide warhead.
Osimertinib is an epidermal growth factor receptor (EGFR) Thr790Met mutant inhibitor for the treatment of advanced non-small cell lung cancer (NSCLC) [10]. After the administration of a single oral dose of 20 mg [14C]-osimertinib to healthymale volunteers, plasma osimertinib and its two major metabolites, AZ5104 and AZ7550, only accounted for 0.8, 0.08 and 0.07% of the total plasma radioactivity, respectively, based on the AUC ratio [11]. The terminal half-life of the total plasma radioactivity was 474 h, which was consistent with the half-life of HSA (19-20 days) [12]. This may indicate that the osimertinib-related material covalently and irreversibly binds to HSA. Even though osimertinib accounted for less than 1% of the total plasma radioactivity, osimertinib demonstrated excellent clinical activity. After oral administration of [14C]-osimertinib (4 mg/kg) to male partially pigmented rats, male and female albino rats, the liver to blood ratios of radioactivity were 21.3, 25.2 and 37.3 at 1 h post-dosing, respectively [11]. Osimertinib could quickly distribute to target tissues, while the process of osimertinib covalent binding to plasma protein required time.

Based on the results described above, we put forward some questions. What is the binding site(s) of osimertinib to HSA? What is the binding mode of osimertinib with HSA. What is the t1/2 of osimertinib in the HSA solution and human plasma? Does the covalent binding exhibit species difference?
Neratinib is a pan-EGFR inhibitor that is active against human epidermal growth factor receptor 1 (HER1), HER2 and HER4 [13]. A previous study showed that neratinib was solely and covalently bound to the Lys-190 of HSA. Besides, the covalent binding showed species difference [14]. The covalent binding with HSA may affect the pharmacokinetics of the covalent TKIs, while the species difference of covalent binding may have impact on the species extrapolation of pharmacokinetics.

Pyrotinib is an oral TKI that targets the HER-1 and HER-2 receptors [15]. A previous study by our lab indicated that pyrotinib can also covalently bind to the Lys-190 of HSA [16]. Afatinib (EGFR inhibitor) and ibrutinib (Bruton tyrosine kinase inhibitor) can covalently bind to human plasma protein and HSA, respectively. However, the binding sites have not been determined yet [17, 18].
We speculated that covalent TKIs containing acrylamide warhead could covalently bind to the Lys-190 of HSA. The current work is aimed at: (1) studying the species difference in covalent binding to plasma proteins by covalent TKIs and identify their binding sites; (2) exploring the mode of covalent binding between covalent TKIs and HSA through molecular docking; (3) investigating the t1/2 of covalent TKIs in HSA solution and human plasma, and identifying the factors that affect this rate by quantitative calculations and linear modeling; (4) determining the reversibility of the covalent bonds formed by covalent TKIs and human plasma proteins. The covalent TKIs we studied are showed in Figure 1.
2. Materials and Methods

2.1. Materials

Afatinib, ibrutinib, neratinib, and olmutinib were purchased from Meilun Biotechnology Co., Ltd. (Dalian, China), and osimertinib was purchased from MedChemExpress Co., Ltd (Monmouth Junction, NJ, USA). Pyrotinib was provided by Shanghai Hengrui Pharmaceutical Co., Ltd. (Shanghai, China), and alflutinib and AST5902 were obtained from Shanghai Allist Pharmaceuticals Inc. (Shanghai, China). Human serum albumin was purchased from Sigma-Aldrich (St. Louis, Missouri,USA). The plasma of mouse (C57, mixed sex, n > 6), rat (Sprague Dawley, mixed sex, n > 6), rabbit (Oryctolagus cuniculus, mixed sex, n > 6), dog (Beagle, mixed sex, n > 6), and monkey (Cynomolgus, mixed sex, n > 6) were obtained from the Experimental Animal Center of Shanghai Institute of Materia Medica, Chinese Academy of Sciences (Shanghai, China). Human plasma (Asian, mixed sex, n > 30) was obtained from The Third Affiliated Hospital of Qiqihar Medical University (Qiqihar, China). EDTA was used as anticoagulant and the plasma was kept at −20°C. Pronase was purchased from Roche Inc. (Mannheim, Germany). Boc-lysine was purchased from GL Biochem (Shanghai) Ltd. (Shanghai, China). HPLC-grade ammonium acetate and the formic acid were purchased from Roe Scientific Inc. (Neward, NJ, USA) and Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan), respectively. The HPLC-grade acetonitrile and methanol were purchased from Merck KGaA (Darmstadt, Germany). The Milli-Q system was used to prepare deionized H2O (Molsheim, France).

2.2. Stability of TKIs in plasma

The stability of 8 covalent TKIs was investigated in plasma of different species. Frozen plasma from mouse, rat, rabbit, dog, monkey and human was used for stability evaluation. Stock solutions of TKIs were dissolved in dimethyl sulfoxide and then diluted into plasma to achieve a concentration of 1 μM containing 0.1% dimethyl sulfoxide. The samples were incubated at 37ºC for 6 h. Aliquots (50 μL) were added with 50 μL of internal standard (alflutinib-d3 /osimertinib-d9, 10/10 ng/mL), and the reaction was terminated by the addition of acetonitrile (150 μL), vortexed, and centrifuged, and the supernatant was analyzed by LC-MS/MS.

2.3. Identification of TKIs’ protein adducts

Human plasma was diluted 10-fold with PBS (0.1 M, pH 7.4) and incubated with 100 μM of TKIs (alflutinib, AST5902, olmutinib, osimertinib, and pyrotinib) for 36 h at 37ºC. The incubated samples were diluted 10-fold with deionized H2O and directly analyzed by UPLC/Q-TOF MS to identify the protein that forms adducts with covalent TKIs.
2.4. Pronase hydrolysis

Pronase is a non-specific protease that hydrolyzes proteins into single amino acids. For this experiment, 8 covalent TKIs (100 μM) were incubated with HSA solution (1%, w/v) at 37ºC for 24 h and then, 200 μL of the incubated sample was added with 10 μL of pronase (20 mg/mL) and incubated at 37ºC for 16 h. After hydrolysis, the samples were centrifuged at 10000 g for 10 min and analyzed by UPLC/Q-TOF MS.
2.5. HCl hydrolysis

Afatinib, alflutinib, AST5902, ibrutinib, olmutinib, and osimertinib were diluted with 50 μM of HSA for a final concentration of 100 μM. After incubation at 37ºC for 24 h, 200 μL of the incubated sample was added to 1 mL of HCl (2 M), followed by an incubation at 90ºC for 2 h. After incubation, 1 mL of NaOH (2 M) was added and the pH of the solution was adjusted to 3-4 using acetic acid and then UPLC/Q-TOF MS analysis was performed.
2.6. Covalent binding of TKIs with Boc-Lys

Eight covalent TKIs (100 μM) were incubated with 55.5 mM Boc-Lys (100 mM borate buffer, pH 10.2) at 37ºC for 24 h, respectively. The adducts were detected using UPLC/Q-TOF MS.
2.7. The binding mode of covalent TKIs and human serum albumin

The molecular docking program Induced Fit Docking and Covalent Docking were used to simulate the binding mode of TKIs with HSA (PDB ID: 418U). The protein was prepared by Protein Preparation Wizard of Schrödinger (LLC, New York, NY, 2019), with the protonation states of residues determined at pH 7.4. Lys-190 was set as the center for covalent docking of TKIs.
2.8. TKIs’ intrinsic reactivity with HSA and human plasma

Eight covalent TKIs (1 μM) were incubated with HSA (45 mg/mL, 0.1 M PBS) at 37ºC for 0 h, 24 h, 48 h and 72 h, respectively. In addition, neratinib was incubated with HSA (45 mg/mL, 0.1 M PBS) at 37ºC for 0 h, 4 h, 10 h, and 24 h. Another 1 μM of 8 covalent TKIs was incubated with human plasma at 37ºC for 0 h, 0.5 h, 1 h, 4 h, 6 h, 10 h, 24 h, 48 h, 72 h, 96 h and 120 h, respectively. The remaining amount of the parent was measured by LC-MS/MS. Pseudo-first-order rate constants were determined by plotting the natural log of TKIs relative concentration against time. The negative slope of the straight line is the pseudo-first-order rate constant. The pseudo-first-order rate constants were used to determine half-life with the following equation.
t1/2=ln2/pseudo-first order rate constant

2.9. Quantitative calculations and linear modeling

The protonation states and the lowest energy conformations of TKIs were prepared using LigPrep 4.6 and MacroModel 12.4 program in Schrödinger (LLC, New York, NY, 2019). We used electrophilic reagent CH3NH2 as a substitute of Lys-190 of HSA for calculation. The lowest energy conformer of TKIs and the reaction products of adduct reaction between TKIs and methylamine were further optimized using Gaussian 09 software (Gaussian Inc., Wallingford, Conneticut, USA). Full geometry optimizations and frequency analyses of DFT calculations were carried out at the B3LYP/6-311+G (d, p) level. Fukui function and local electrophilicity index were calculated according to the analysis of wave function of the optimized molecule using Multiwfn software 3.2 (Beijing Kein Research Center for Natural Sciences, Beijing, China). Transition-state optimizations were conducted for the additions of CH3NH2 to the electrophilic β-carbon atom of acrylamide to calculate the reaction activation energy (∆Ea).
There are 23 features, including docking score, activation energy and the Fukui function that were used as input parameters for model building. Lasso regression was used to describe constrained optimization problems. Under the constraint that the sum of the absolute values of the regression coefficients is less than or equal to the threshold, some coefficients became smaller and could be directly changed to 0, to have the ability to select features. When using lasso regression, the λ, based on the MAE and R2 with the reaction half-life of TKIs with HSA and human plasma protein was respectively changed. Under the circumstances, the coefficient of certain features was 0 and the more important features were selected. Then the selected features were

used as predictor variables for further model building with the Ridge regression for the half-life of TKIs in both HSA solution and human plasma. The leave-one-out (LOO) method was applied for the evaluation of the model and to determine the parameter λ.
2.10. Reversibility of TKIs covalently bound to HSA and plasma proteins

To examine the potential reversibility of the covalent binding formed by TKIs with HSA and human plasma proteins, HSA and human plasma were incubated with covalent TKIs (100 μM) at 37ºC for 48 h. After incubation, the samples were extracted with methyl-tert-butyl ether (MTBE) (with 4.5 volumes for 3 times). Aliquots (50 μL) of the aqueous layer (water residue) were suspended in triplicate in 950 μL PBS (0.1 M, pH 7.4) and sonicated for 10 min. Then, the mixtures were incubated with gentle shaking at 37ºC for 12 h, 24 h, and 36 h to release covalent TKIs that were covalently bound to HSA and human plasma proteins.
The effects of temperature and initial concentrations of covalent TKIs-protein adducts on the release rate were further studied. Samples of alflutinib, neratinib osimertinib, and pyrotinib with plasma concentrations of 10 μM and 100 μM were prepared and incubated at 37ºC for 48 h. Then, human plasma proteins containing covalently bound TKIs were suspended in 20-fold volume of phosphate buffer. The mixtures were incubated with gentle shaking at 4ºC, 37ºC, and 45ºC for 24 h and 48 h, respectively. The concentration of released covalent TKIs was determined by LC-MS/MS.
2.11. Analytical methods

After protein precipitation by acetonitrile, the concentration of TKIs in plasma and HSA were analyzed by LC-MS/MS. The analysis was performed using a Shimadzu LC-30 AD liquid chromatograph (Kyoto, Japan) coupled with a Triple Quad 6500+ mass spectrometer (AB Sciex, Canada). Data acquisition and processing were carried out using the Analyst software version 1.6.3 (AB Sciex, Canada). The detection of covalent TKIs-HSA adducts was performed following a reported method [18]. The samples of pronase hydrolysis and adducts of TKIs with Boc-Lys were detected at an ACQUITY UPLC system (Waters Corp., Milford, MA, USA) coupled with Waters Synapt G2 mass spectrometer (Milford, MA, USA). After HCl hydrolysis, samples were detected at an ACQUITY liquid chromatography system (Waters Corp., Milford, MA, USA) coupled with Q-TOF mass spectrometer 5600 (AB Sciex, Canada). The detailed LC-MS/MS and LC/MS methods are displayed in the Supplementary Material section.
3. Results

3.1. Plasma stability

The stability of the 8 covalent TKIs to plasma proteins is significantly different between species (Table 1). The stability of neratinib and osimertinib in human plasma is much lower than that of other species, which is consistent with previous studies [11, 14]. Afatinib, ibrutinib, and pyrotinib exhibited the least stability in human plasma, while olmutinib showed the least stability in dog plasma. Alflutinib and AST5902 had the least stability in rat plasma, followed by mouse plasma.
3.2. Identification of covalent TKIs’ protein adducts

Using pyrotinib as an example, two multi-charged molecular clusters were detected by UPLC/Q-TOF MS analysis and marked as * and **, respectively (Figure 2). The figure shows the minimum and maximum charge numbers of each ion, M + 45H + (HSA: m/z 1477.4, pyrotinib-HSA adduct: m/z 1490.3), M + 59H + (HSA: m/z 1127.1, pyrotinib-HSA adduct: m/z 1136.9). The deconvolution of the calculated molecular weight of the main ion cluster * was 66436 Da and for the less abundant ion cluster ** was 67020 Da. The molecular weight difference between the two was 584 Da, which is 2 Da higher than pyrotinib molecular weight. The small difference may result from the instrument deviation. Similarly, we have also detected the adducts of alflutinib, AST5902, olmutinib, and osimertinib with HSA (Supplemental Figure 1-4). Like ibrutinib, alflutinib and osimertinib could form two adducts with HSA, HSA-TKIs adduct and HSA-2TKIs adduct. The adducts of afatinib, ibrutinib and neratinib with HSA have been reported [14, 18].
3.3. Pronase hydrolysis

The covalent TKIs-HSA adducts were hydrolyzed by pronase. Except for lysine, adducts of covalent TKIs with other amino acids were not detected. The data showed that eight covalent TKIs can covalently bind to the lysine residue of HSA. The data of the TKIs-lysine adducts, including the calculated and observed m/z, fragment ions and retention time, are summarized in Supplemental Table 1. The fragment ions included a protonated molecular ion of TKIs which resulted from the neutral loss of lysine. In addition, other fragment ions are the product ions of the TKIs.
3.4. HCl hydrolysis

To confirm the binding sites of covalent TKIs (afatinib, alflutinib, AST5902, ibrutinib, olmutinib, osimertinib) with HSA, the covalent TKIs-HSA samples were hydrolyzed with HCl at 90ºC for 2 h. In addition to the parent drugs, we also detected amide hydrolysates of covalent TKIs and a series of covalent TKIs’ adducts with peptides, including lysine, glycine-lysine, lysine-alanine, glycine-lysine-alanine, and lysine-alanine-serine. The data of the TKIs-peptide adducts, including the calculated and observed m/z, fragment ions and retention time, are summarized in Supplemental Table 2. These results indicated that alflutinib, AST5902, ibrutinib, and osimertinib could covalently bind to the peptide glycine-lysine-alanine-serine. This peptide corresponded to HSA amino acid residues 189 to 192. Based on the detected peptide fragments, we confirmed that alflutinib, AST5902, ibrutinib, and osimertinib could bind to Lys-190 of HSA. However, afatinib and olmutinib may have low covalent binding degrees or insufficient sensitivities to detect their related covalent TKIs-peptide adducts. In summary, in addition to the reported results on neratinib and pyrotinib, we confirmed that alflutinib, AST5902, ibrutinib, and osimertinib can covalently bind to the Lys-190 of HSA.
3.5. Covalent binding with Boc-Lys

Eight covalent TKIs could form adducts with Boc-Lys under the condition of 100 mM borate buffer with pH 10.2. The data of the TKIs-Boc-lysine adducts, including the calculated and observed m/z, fragment ions and retention time, are summarized in Supplemental Table 3. The fragment ions included a protonated molecular ion of TKIs-Lys adducts and TKIs which resulted from the neutral loss of t-Butyloxy

carbonyl group and Boc-lysine, respectively. In addition, other fragment ions are the product ions of the TKIs.
3.6. The binding mode of TKIs and human serum albumin

HSA is composed of three similar domains (I, II, and III): domain I (1195), domain II (196383) and domain III (384585) (Supplemental Figure 5). A cylindrical structure is formed by subdomains A and B of each domain packing against each other and a hydrophobic pocket is formed by hydrophobic amino acids around the cylindrical groove [19, 20].
The Induced Fit Docking results show that subdomain B of domain I is the primary binding site of covalent TKIs to HSA. Pyrotinib may form hydrogen bonds with the hinge area of HSA in the hydrophobic pocket. The first one takes place between the carbonyl oxygen atom of acrylamide and Leu-115 that can reduce the electron cloud density of acrylamide and increase its electrophilicity. The other one forms between the hydrogen atom of pyrrole moiety and Arg-145 (Figure 3A). In addition, we also found other non-covalent interactions that were involved in the interaction of TKIs and HSA: π-π stackings were generated with the pyrrole moiety and its nearby amino acid residues (His-146, Phe-149 and Phe-157), and a π-cation effect was formed between the nitrogen atom of pyrrole moiety and His-146, quinoline and the Arg-114, respectively. Those non-covalent interactions can stabilize the pyrotinib-HSA complex and increase the hydrophobicity of pyrotinib in the hydrophobic pocket. The distance between the β-carbon atom of the pyrotinib acrylamide and the ε-amine of Lys-190 is 3.8 Å; therefore, Michael addition between

Lys-190 and pyrotinib acrylamide are likely to occur under the circumstances. From the covalent docking results, there is an additional hydrogen bond between the substituent cyanogen nitrogen atom of quinoline and Arg-185; however, we do not observe a hydrogen bond between the hydrogen atom of pyrrole moiety and Arg-145 (Figure 3B). Although the Covalent docking results of the interactions between covalent TKIs and HSA in hydrophobic pockets are different from the Induced Fit Docking results, the conformational changes of the two docking modes showed no difference (Figure 3C, 3D). The putative binding mode of other covalent TKIs with HSA by Induced Fit Docking and Covalent Docking were performed (Supplemental Figure 6-12).
3.7. Intrinsic reactivity of covalent TKIs with HSA and human plasma

The remaining amount of the 8 covalent TKIs decreased with increasing incubation time. Figure 4 and 5 plot the natural log of covalent TKIs’ relative concentration against time in HSA solutions and human plasma, respectively. The t1/2 of neratinib in HSA solution was 23.9 h, which showed that this compound has the fastest reaction rate with HSA among tested TKIs. For neratinib, the linear regression that included the 48 h and 72 h did not exhibit linearity. Thus, we incubated neratinib with HSA at 0 h, 4 h, 10 h, and 24 h to determine the t1/2 of neratinib in the HSA solution. The t1/2 of ibrutinib and olmutinib were found to be ≥ 72 h. In human plasma, osimertinib had the shortest t1/2 (1.1 h), followed by neratinib (4.9 h), ibrutinib (5.8 h), alflutinib (10.5 h), AST5902 (12.7 h) and pyrotinib (21.6). The t1/2 of olmutinib and afatinib were found to be ≥ 44 h. Covalent TKIs’ reactivity with HSA and human plasma proteins did not correlate well. Covalent TKIs showed less stability in human plasma than in the HSA solution. Although osimertinib was the most reactive compound with human plasma proteins, neratinib had the fastest covalent binding rate to HSA. Based on the results, we can speculate that osimertinib and ibrutinib could covalently bind to other plasma proteins in addition to HSA.

3.8. Quantitative calculations and linear modeling

Quantitative calculations and machine learning were used to screen for high correlative features of TKIs and HSA interactions such as β-carbon atom local electrophilicity index, electrophilicity index, activation energy (∆Ea) and docking score of covalent docking, which highly correlated with the half-life of covalent TKIs in HSA solution. These features were screened out using Lasso regression and labeled as X11, X12, X13, X14, respectively. These features were further used for building Ridge regression model. In the equation, the values of four features were used after normalization process (Equation 2, Figure 6A). The R2 value was 0.88 and the MAE value was 5.3 h. The value of t1/2 was calculated using the following equation:
t1 / 2 = 10.9 * X11 + 5.6 * X12 + 7.3 * X13 + 13.6 * X14 + 58.6 (2)Similarly, we found that the β-carbon atom Fukui Function, β-carbon atom local electrophilicity index, electrophilicity index and docking score of covalent docking highly correlated with half-life of covalent TKIs in human plasma by using the same screening method mentioned above (Figure 6B). The four features were labeled as X21, X22, X23, X24, respectively. The R2 value was 0.88 and the MAE value was 5.4

h. The value of t1/2 was calculated using the following equation:t1/2=14.3 * X21 + 12.9 * X22 + 7.4* X23 + 10.1* X24 + 19.8 (3)

3.9. Reversibility of covalent TKIs bound to plasma proteins and HSA

Table 2 shows the released amount of TKIs after incubation of TKIs-plasma proteins adducts and TKIs-HSA adducts with PBS (0.1 M, pH 7.4) at 37ºC for 12 h, 24 h, and 36 h. TKIs’ released amount increased with the increasing incubation time. This indicated that the covalent binding of TKIs to plasma proteins are reversible. The osimertinib-HSA adduct released a small amount of osimertinib after 12 h incubation and its amount decreased later during the incubation time. Olmutinib also showed similar phenomenon at the last point. The amount of released TKIs by other TKIs-HSA adducts increased with incubation time.
Temperature has a significant effect on the stability of covalent TKIs-plasma protein adducts (Supplemental Figure 13, Figure 14). When the temperature increased, the rate of TKI released from the adducts was faster. At 4°C, osimertinib-plasma protein adducts did not release osimertinib. After incubating 10 μM or 100 μM TKIs at 37°C for 48 h, the degree of covalent binding to plasma proteins were consistent between 10 μM and 100 μM. Therefore, the initial concentrations of covalent TKIs-plasma proteins adduct shown in Supplemental Figure 14 were approximately ten times of that observed in Supplemental Figure 13. The results showed that the higher the concentration of covalent TKIs-plasma proteins adduct, the faster the released rate of TKIs was.
4. Discussion

Plasma protein binding is a critical parameter of drugs because a free drug thought to elicit a pharmacological effect and be cleared from the body [21]. A variety of compounds tend to bind slightly more to human plasma compared to pre-clinical species [22]. The disposition of a drug can be affected by the formation of drug-plasma protein adducts. Human plasma proteins that bind to small-molecule drugs mainly include human serum albumin, α1-acid glycoprotein, globulin, and various lipoproteins [23, 24]. In this study, more covalent TKIs tend to covalently bind to human plasma proteins compared with the plasma of preclinical species. For instance, neratinib, which covalently bound to HSA, did not bind to the plasma proteins of mouse or rat. Relative to human, the amino acid sequences of albumin were 94%, 80%, 73% and 72% conserved in monkey, dog, rat and mouse, respectively[19]. Except for human and monkey, the amino acid residue 190 of the dog, rabbit, rat and mouse albumin was not a lysine. The species difference may be a result of differences in the amino acid sequence of albumin. In this circumstance, when scientists extrapolate the pharmacokinetics from mouse or rat to human, they should also consider the species difference in covalent binding.

With a concentration of 400‒700 μM, HSA is the most abundant protein in plasma [21]. Besides, the plasma protein binding measured in human plasma correlates well with that measured in HSA solution was reported [22]. Previous studies showed that some covalent TKIs can form adduct with HSA. Thus, the covalent binding of covalent TKIs to HSA was primarily studied. Adducts with HSA were detected for 8 covalent TKIs. The results showed that 8 covalent TKIs could bind to the lysine of HSA. Both neratinib and pyrotinib can covalently bind to the Lys-190 [14, 16]. In this study, we assumed that other 6 covalent TKIs can also bind to the Lys-190 of HSA. However, only alflutinib, AST5902, ibrutinib, and osimertinib were confirmed to bind to the Lys-190 of HSA. For afatinib and olmutinib, we did not detect their related covalent TKIs-peptide adducts. This may result from the low covalent binding degrees or insufficient sensitivities of LC-MS. Besides, alflutinib, ibrutinib and osimertinib can bind to another lysine residue of HSA, which need further study.

Nevertheless, we assumed that afatinib and olmutinib could covalently bind to the HSA Lys-190 during the simulation. The results of molecular docking indicated that the amino acids residues around Lys-190 can stabilize the conformation of covalent TKIs through non-covalent interactions. According to the Induced Fit Docking and Covalent Docking, it is known that the interaction mode of covalent TKIs and pockets centered on Lys-190 is reasonable.
The docking results showed that the microenvironment can increase Lys-190 intrinsic reactivity and that covalent TKIs are accessible to the Lys-190 of HSA. Meanwhile, the covalent TKIs-HSA complex was stabilized in a position that was conducive to Michael addition through a series of hydrogen bonds, π-π interactions and π-cation effect; thereby explaining the selectivity of modification.
The 8 covalent TKIs could form adducts with Boc-Lys (pH 10.2) at 37ºC when incubated with Boc-Lys. Our study showed that when the pH was decreased to 7.4, the Michael addition between covalent TKIs and Boc-Lys did not occur, which indicated that Michael addition is affected by pH. The pKa of the amino group of Boc-Lys was predicted to be 10.6 by the Advanced Chemistry Development (ACD/Labs) software (Percepta 2018.1). The pKa of the amino group of lysine residues in protein may be lower than the Boc-Lys, which was affected by microenvironment.

We first used HSA to conduct modeling, then the established method was applied to create the human plasma model. According to the linear modeling, the β-carbon atom local electrophilicity index of covalent TKIs and the covalent docking score of covalent TKIs and HSA have important effects on the half-life of covalent TKIs in human plasma. The higher the electrophilicity of the acrylamide β-carbon atom is, the stronger the atom’s ability to accept electrons; and the easier is the occurrence of an electrophilic reaction such as Michael addition. These data indicated that if drug discovery scientists wish to decrease the off-target reactivity, they could lower the electrophilicity of the acrylamide β-carbon atom. However, it will also decrease the affinity of drugs to the target TKIs. When optimizing the reactivity of covalent TKIs’ warheads, the balance between off-target nucleophiles and on-target protein needs to be considered.
The plasma radioactivity t1/2 of afatinib, pyrotinib, osimertinib, ibrutinib were 118 h, 47.9 h, 474 h, and 47.3 h, respectively [11, 16, 17, 25]. Alflutinib plasma radioactivity t1/2 is approximately 14 days (unpublished data). Osimertinib plasma radioactivity t1/2 was consistent with the t1/2 of HSA, indicating that its HSA covalent binding is irreversible. Our study showed that the covalent binding of osimertinib to HSA was irreversible in vitro. However, the adducts of osimertinib-human plasma proteins exhibited a partial reversibility, indicating that the covalent binding of osimertinib to other plasma proteins seems reversible. The plasma radioactivity t1/2 of the other drugs was less than the t1/2 of HSA, indicating that their covalent binding to HSA was reversible. The release rate of parent drugs from the adducts of covalent TKIs-plasma proteins depends on temperature and the initial concentration of the adducts of the TKIs-plasma proteins. In early drug development, in vitro experiments can be used to assess whether the covalent binding of covalent TKIs to human plasma proteins and HSA is reversible.

Our research has some limitations. The built model is only applicable to acrylamide warheads that are commonly used in the design of covalent TKIs. Except for Lys-190 of HSA, another lysine residue of HSA that alflutinib, ibrutinib and osimertinib bound to remains to be further studied.
5. Conclusion

In summary, our research showed that acrylamide of covalent TKIs can covalently bind to lysine residue of HSA, most of which were determined to be Lys-190. The covalent binding is of species difference, especially between animal and human. The result of molecule docking explained the binding mode of covalent TKIs and HSA. The electrophilicity of the acrylamide β-carbon atom is an important factor that affect the rate of covalent binding between covalent TKIs and plasma proteins. Except for osimertinib, covalent binding between TKIs and HSA are reversible.

Authorship Contributions

Participated in research design: Xiaoyun Liu, Dafang Zhong, Dan Feng, Mingyue Zheng, Yongmei Cui
Conducted experiments: Xiaoyun Liu, Dan Feng

Performed data analysis: Xiaoyun Liu, Dan Feng

Wrote or contributed to the writing of the manuscript: Xiaoyun Liu, Dafang Zhong, Dan Feng, Mingyue Zheng

Conflict of interest

The authors declare that there is no conflict of interest.

Funding

This work was partially supported by the National Natural Science Foundation of China (No 81521005) and the Strategic Priority Research Program of the Chinese Academy of Sciences (No XDA12050306).

References
[1] Bhullar KS, Lagaron NO, McGowan EM, Parmar I, Jha A, Hubbard BP, et al. Kinase-targeted cancer therapies: progress, challenges and future directions. Mol Cancer 2018;17(1):48. https://doi.org/10.1186/s12943-018-0804-2.
[2] Lahiry P, Torkamani A, Schork NJ, Hegele RA. Kinase mutations in human disease: interpreting genotype-phenotype relationships. Nat Rev Genet 2010;11(1):60-74. https://doi.org/10.1038/nrg2707.
[3] Brognard J, Hunter T. Protein kinase signaling networks in cancer. Curr Opin Genet Dev 2011;21(1):4-11. https://doi.org/10.1016/j.gde.2010.10.012.
[4] Roskoski R. Properties of FDA-approved small molecule protein kinase inhibitors: A 2020 update. Pharmacol Res 2020;152:104609. https://doi.org/10.1016/j.phrs.2019.104609.
[5] Singh J, Petter RC, Baillie TA, Whitty A. The resurgence of covalent drugs. Nat Rev Drug Discov 2011;10(4):307-17. https://doi.org/10.1038/nrd3410.
[6] Muller S, Chaikuad A, Gray NS, Knapp S. The ins and outs of selective kinase inhibitor development. Nat Chem Biol 2015;11(11):818-21. https://doi.org/10.1038/nchembio.1938.
[7] Awoonor-Williams E, Rowley CN. How reactive are druggable cysteines in protein kinases? J Chem Inf Model 2018;58(9):1935-46. https://doi.org/10.1021/acs.jcim.8b00454.
[8] Baillie TA. Targeted covalent inhibitors for drug design. Angew Chem Int Ed 2016;55(43):13408-21. https://doi.org/10.1002/anie.201601091.
[9] Johnson DS, Weerapana E, Cravatt BF. Strategies for discovering and derisking covalent, irreversible enzyme inhibitors. Future Med Chem 2010;2(6):949-64. https://doi.org/10.4155/fmc.10.21.
[10] Ramalingam SS, Yang JCH, Lee CK, Kurata T, Kim DW, John T, et al. Osimertinib As First-Line Treatment of EGFR Mutation–Positive Advanced Non–Small-Cell Lung Cancer. J Clin Oncol 2018.

https://doi.org/10.1200/JCO.2017.74.7576.

[11] Dickinson PA, Cantarini MV, Collier J, Frewer P, Martin S, Pickup K, et al. Metabolic disposition of osimertinib in rats, dogs, and humans: insights into a drug eesigned to bind covalently to a cysteine residue of epidermal growth factor receptor. Drug Metab Dispos 2016;44(8):1201-12. https://doi.org/10.1124/dmd.115.069203.
[12] Rehman MT, Khan AU. Understanding the interaction between human serum albumin and anti-bacterial/anti-cancer compounds. Curr Pharm Design 2015;21(14):1785-99. https://doi.org/10.2174/1381612821666150304161201.
[13] Deeks ED. Neratinib: First Global Approval. Drugs 2017;77(15):1695-704. https://doi.org/10.1007/s40265-017-0811-4.
[14] Wang J, Li-Chan XX, Atherton J, Deng L, Espina R, Yu L, et al. Characterization of HKI-272 covalent binding to human serum albumin. Drug Metab Dispos 2010;38(7):1083-93. https://doi.org/10.1124/dmd.110.032292.
[15] Blair HA. Pyrotinib: First Global Approval. Drugs 2018;78(16):1751-5.

https://doi.org/10.1007/s40265-018-0997-0.

[16] Meng J, Liu XY, Ma S, Zhang H, Yu SD, Zhang YF, et al. Metabolism and disposition of pyrotinib in healthy male volunteers: covalent binding with human plasma protein. Acta Pharmacol Sin 2019;40(7):980-8. https://doi.org/10.1038/s41401-018-0176-6.
[17] Stopfer P, Marzin K, Narjes H, Gansser D, Shahidi M, Uttereuther-Fischer M, et al. Afatinib pharmacokinetics and metabolism after oral administration to healthy male volunteers. Cancer Chemother Pharmacol 2012;69(4):1051-61. https://doi.org/10.1007/s00280-011-1803-9.
[18] Ballard TE, Dahal UP, Bessire AJ, Schneider RP, Geoghegan KF, Vaz AD. A tag-free collisionally induced fragmentation approach to detect drug-adducted proteins by mass spectrometry. Rapid Commun Mass Spectrom 2015;29(22):2175-83. https://doi.org/10.1002/rcm.7375.
[19] He XM, Carter DC. Atomic structure and chemistry of human serum albumin. Nature 1992;358(6383):209-15. https://doi.org/10.1038/358209a0.
[20] Carter DC, He XM. Structure of human serum albumin. Science 1990;249(4966):302-3. https://doi.org/10.1126/science.2374930.
[21] Trainor GL. The importance of plasma protein binding in drug discovery. Expert Opin Drug Dis 2007;2(1):51-64. https://doi.org/10.1517/17460441.2.1.51
[22] Colclough N, Ruston L, Wood JM, MacFaul PA. Species differences in drug plasma protein binding. Med Chem Commun 2014;5(7):963-7. https://doi.org/10.1039/c4md00148f.
[23] Tozer TN, Rowland M. Introduction to pharmacokinetics and pharmacodynamics: the quantitative basis of drug therapy. Lippincott Williams & Wilkins; 2006.
[24] Wasan KM, Cassidy SM. Role of plasma lipoproteins in modifying the biological activity of hydrophobic drugs. J Pharm Sci-US 1998;87(4):411-24. https://doi.org/10.1021/js970407a.
[25] Scheers E, Leclercq L, de Jong J, Bode N, Bockx M, Laenen A, et al. Absorption, metabolism and excretion of oral 14C radiolabeled ibrutinib: an open-label, phase I, single-dose study in healthy men. Drug Metab Dispos 2014;43:289-97. https://doi.org/10.1124/dmd.114.060061.

Figure Legends

Figure 1. The chemical structure of covalent tyrosine kinase inhibitors

Figure 2. Mass spectra of HSA and HSA-pyrotinib adduct, obtained by direct injections of 10X-diluted human plasma that was incubated with pyrotinib (A) and deconvoluted molecular weights of HSA and HSA-pyrotinib adduct (B).
Figure 3. Putative binding mode of pyrotinib with the human serum albumin by Induced Fit Docking (A) and Covalent Docking (B). The active site pocket of human serum albumin was depicted as the white surface by Induced Fit Docking (C), and Covalent Docking (D).
Figure 4. Incubation half-life of covalent TKIs with human serum albumin solution.

Figure 5. Incubation half-life of covalent TKIs with human plsama.

Figure 6. Linear model of half-life of covalent TKIs in HSA solution (A) and human plasma (B).

Table1. Mean (n = 3) covalent TKIs remaining after incubation at 37C for 6 h in mouse, rat, dog, monkey, and human plasma
Human Monkey Dog Rabbit Rat Mouse
Afatinib 83% 87% 93% 95% 90% 85%
Alflutinib 59% 85% 74% 77% 0.10% 16%
AST5902 83% 84% 70% 75% 24% 60%
Ibrutinib 61% 91% 81% 64% 88% 91%
Neratinib 44% 83% 87% 106% 106% 98%
Olmutinib 91% 99% 65% 88% 75% 70%
Osimertinib 0.60% 58% 68% 75% 13% 7%
Pyrotinib 82% 99% 92% 103% 95% 100%

Table 2. The amount of covalent TKIs released from covalent TKIs-plasma proteins adducts and covalent TKIs-HSA adducts Pyrotinib after incubation at 37C for differenttimes (n=3)