WZ4002（周文军）

LETTERS Novel mutant-selective EGFR kinase inhibitors against EGFR T790M Wenjun Zhou1,2*, Dalia Ercan3,4*, Liang Chen3,4*, Cai-Hong Yun1,2*, Danan Li3,4, Marzia Capelletti3,4, Alexis B. Cortot3,4, Lucian Chirieac5, Roxana E. Iacob6,7, Robert Padera5, John R. Engen6,7, Kwok-Kin Wong3,4,8,9, Michael J. Eck1,2, Nathanael S. Gray1,2 & Pasi A. Ja¨nne3,4,8 The clinical efficacy of epidermal growth factor receptor (EGFR) kinase inhibitors in EGFR-mutant non-small-cell lung cancer (NSCLC) is limited by the development of drug-resistance muta- tions, including the gatekeeper T790M mutation1–3. Strategies targeting EGFR T790M with irreversible inhibitors have had limited success and are associated with toxicity due to concurrent inhibition of wild-type EGFR4,5. All current EGFR inhibitors possess a structurally related quinazoline-based core scaffold and were identified as ATP-competitive inhibitors of wild-type EGFR. Here we identify a covalent pyrimidine EGFR inhibitor by screening an irreversible kinase inhibitor library specifically against EGFRT790M. These agents are 30- to 100-foldmore potent against EGFR T790M, and up to 100-fold less potent against wild- type EGFR, than quinazoline-based EGFR inhibitors in vitro. They are also effective in murine models of lung cancer driven by EGFR T790M. Co-crystallization studies reveal a structural basis for the increased potency and mutant selectivity of these agents. These mutant-selective irreversible EGFR kinase inhibitors may be clini- cally more effective and better tolerated than quinazoline-based inhibitors. Our findings demonstrate that functional pharmaco- logical screens against clinically importantmutant kinases represent a powerful strategy to identify new classes ofmutant-selective kinase inhibitors. EGFR kinase inhibitors, gefitinib and erlotinib, are effective clinical therapies for NSCLCs that harbour activating mutations in the EGFR kinase domain1,6. The most common EGFR mutations, L858R and delE746_A750, impart both an increased affinity for gefitinib or erlo- tinib and a decreased affinity for ATP relative to wild-type (WT) EGFR7,8. The clinical efficacy of gefitinib or erlotinib is, however, ultimately limited by the development of acquired drug resistance such as by mutation of the gatekeeper T790 residue (T790M), which is detected in 50% of clinically resistant patients2,3. Unlike the analo- gous T315I mutation in ABL, which introduces a steric impediment for imatinib binding, EGFR T790M only modestly affects gefitinib binding. However, more importantly, it restores the affinity for ATP, similar to that of WT EGFR9. Most EGFR inhibitors are based on a 4-anilinoquinazoline core scaffold andwere initially identified as ATP-competitive inhibitors of WT EGFR. They include irreversible inhibitors that, unlike gefitinib, contain an electrophilic functionality that undergoes aMichael addi- tion reaction with a conserved cysteine residue present in EGFR (Cys 797). The covalent nature of these compounds allows them to achieve greater occupancy of the ATP site relative to reversible inhibitors, thus providing the ability to inhibit EGFR T790M in pre- clinical models, despite the increased ATP affinity conferred by this secondary mutation4,10,11. However, all current irreversible inhibitors are less potent in cell-linemodels harbouring EGFRT790M than those with an EGFR-activating mutation alone (Supplementary Fig. 1) and, at clinically achievable concentrations, these agents do not inhibit EGFR T790M in vitro10–13. Because the ATP affinity of EGFR T790M is similar toWT EGFR, the concentration of quinazoline-based EGFR inhibitors required to inhibit EGFR T790Mwill also effectively inhibit WT EGFR. In patients, this concurrent inhibition ofWT EGFR results in skin rash and diarrhoea, and limits the ability to achieve plasma concentrations sufficient to inhibit EGFR T790M. Consequently, the clinical efficacy of the irreversible EGFR inhibitors CI-1033, HKI-272 and PF00299804 has been limited, especially in patients with gefitinib- or erlotinib-resistant NSCLC, and the dose-limiting toxicity has been diarrhoea and skin rash5,14,15. We hypothesized that the anilinoquinazoline scaffoldmay not be the most potent or specific for inhibiting EGFR T790M because it relies on the small size and hydrogen bonding interactions with the gatekeeper threonine of WT EGFR. We prepared a focused library of common kinase inhibitor core scaffolds where one of the side chains was modi- fied with an acrylamide group at a position that molecular modelling predicted to react with Cys 797. This library was screened for com- pounds that could inhibit the growth of both gefitinib-resistant (PC9GR; delE746_A750/T790M) and -sensitive (PC9; delE746_A750) cell lines but were not toxic up to 10mM against A549 (KRASmutant) or H3122 (EML4-ALK) cells. We compared our findings with both reversible (gefitinib) and irreversible EGFR inhibitors (CL-387,785 and HKI-272). Three closely related pyrimidines, WZ3146, WZ4002 and WZ8040, were identified from the screen that possessed up to a 300-fold lower half-maximum inhibitory concentration (IC50) against the PC9GR cells compared with clinical-stage inhibitors such as HKI-272 (Fig. 1a, b and Supplementary Table 1).Weobserved a similar increased potency of the WZ compounds in the H1975 (L858R/ T790M) cell line and in Ba/F3 cells harbouring EGFR T790M (Fig. 1b and Supplementary Tables 1 and 2). The increased cellular potency correlated with inhibition of EGFR, AKT and ERK1/2 phosphorylation in NSCLC cell lines (Fig. 1c and Supplementary Fig. 2) and with the more potent inhibition of EGFR phosphorylation by WZ4002 in NIH-3T3 cells expressing different EGFR T790Mmutant alleles (Fig. 1d and Supplementary Fig. 3). The profile against ERBB2 was markedly different: the WZ compounds were less potent than CL-387,785 or HKI-272 (Supplementary Tables 1 and 2) and did not inhibit ERBB2 1Department of Cancer Biology, 2Department of Biological Chemistry and Molecular Pharmacology, 3Lowe Center for Thoracic Oncology, 4Department of Medical Oncology, Dana- Farber Cancer Institute, 44 Binney Street, Boston, Massachusetts 02115, USA. 5Department of Pathology, Brigham andWomen’s Hospital, Boston, Massachusetts 02115, USA. 6The Barnett Institute of Chemical & Biological Analysis, 7Department of Chemistry and Chemical Biology, Northeastern University, Boston, Massachusetts 02115, USA. 8Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts 02115, USA. 9Ludwig Center at Dana-Farber/Harvard Cancer Center, Boston, Massachusetts 02115, USA. *These authors contributed equally to this work. Vol 462 |24/31 December 2009 |doi:10.1038/nature08622 1070 Macmillan Publishers Limited. All rights reserved©2009 CALMAN Typewritten Text CALMAN Typewritten Text CALMAN Typewritten Text 用心生活馆 CALMAN Typewritten Text CALMAN Typewritten Text CALMAN Typewritten Text CALMAN Typewritten Text 肺癌QQ群2970957 CALMAN Typewritten Text CALMAN Typewritten Text CALMAN Typewritten Text CALMAN Typewritten Text 肺癌QQ群2970957 phosphorylation in 3T3 cells expressing the ERBB2 gatekeeper (T798I) mutation (data not shown). Analysis of recombinant EGFR T790M kinase incubated with WZ3146 by electrospray mass spectrometry revealed stoichiometric addition of one inhibitor molecule to the protein. Analysis of a pepsin digest of the modified protein by tandem mass spectrometry identified Cys 797 as the site of modification, thus verifying covalent bond formation between WZ3146 and EGFR (Sup- plementary Fig. 4). 100 80 60 40 20 0 100 4,000 3,000 2,000 1,000 100 80 60 40 20 0 4,000 3,000 2,000 1,000 100 80 60 40 20 0 80 60 40 20 0 Gefitinib CL-387,785 HKI-272 WZ3146 WZ4002 WZ8040 4,000 3,000 2,000 1,000 100 80 60 40 20 0 4,000 3,000 2,000 1,000 100 80 60 40 20 0 100 100 100 80 60 40 20 0 120 140 160 180 200 80 60 40 20 0 IC 50 (n M ) IC 50 (n M ) N Cl O Y N N X N N H N H WZ3146 X=O, Y=H WZ4002 X=O, Y=OMe WZ8040 X=S, Y=H a c d b WZ3146 WZ4002 WZ4002 Gefitinib HKI-272 CL-387,785 CL-387,785Concentration (µM) pEGFR EGFR pAKT AKT pERK 1/2 ERK 1/2 Tubulin Concentration (nM) EGF (10 ng ml–1) pEGFR EGFR Tubulin 0 0. 01 0. 1 1. 0 0 0. 01 0. 1 1. 0 0 0. 01 0. 1 1. 0 H3255 PC9 H1975 PC9 GR L858R E746_A750 L858R/T790M E746_A750/T790M 0 0 10 10 0 1, 00 0 0 0 10 10 0 1, 00 0 0 0 10 10 0 1, 00 0 0 0 10 10 0 1, 00 0 – + + + + – + + + + – + + + + – + + + + Figure 1 | WZ4002, WZ3146 and WZ8040 are novel EGFR inhibitors, suppress the growth of EGFR-T790M-containing cell lines and inhibit EGFR phosphorylation. a, Chemical structures of the WZ compounds. b, IC50 values for NSCLC cell lines (top) and Ba/F3 cells (bottom), with genotypes corresponding to the NSCLC cell lines, treated with indicated drugs. Growth was assessed using the MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3- carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) survival assay. c, Comparison of WZ3146, WZ4002 and CL-387,785 on EGFR signalling in PC9 GR cells. The cells were treated with the indicated concentrations of each drug for 16 h. Cell extracts were immunoblotted to detect the indicated proteins. d, Comparison of EGFR inhibitors on EGFR phosphorylation in 3T3 cells expressing del E746_A750/T90M. The cells were treated with indicated concentrations of each drug for 16 h and stimulated with EGF (10 ng ml21) 15 min before lysis. Cell extracts were immunoblotted to detect the indicated proteins. 125 100 75 50 25 0 0 P er ce nt ag e co nt ro l 10–3 10–2 10–1 100 101 Drug concentration (µM) CL-387,785 HKI-272 WZ3146 WZ4002 WZ8040 WZ4002 CL-387,785 Gefitinib HKI-272 Concentration (nM) EGF (10 ng ml–1) – + + + + – + + + + – + + + + – + + + + 0 0 10 10 0 1, 00 0 0 0 10 10 0 1, 00 0 0 0 10 10 0 1, 00 0 0 0 10 10 0 1, 00 0 pEGFR EGFR Tubulin Vehicle E G FR p E G FR Erlotinib WZ4002 100 80 60 40 20 0 Erlotinib WZ4002 Vehicle P < 0.001 P > 0.05 p E G FR p os iti ve (% ) a c b d Figure 2 | WZ4002 is less potent than quinazoline EGFR inhibitors against WT EGFR in vitro and in vivo. a, EGFR vIII Ba/F3 cells treated with WZ or quinazoline EGFR inhibitors. The mean (n5 6) and standard deviation is plotted for each drug and concentration. b, Comparison of EGFR inhibitors on EGFR phosphorylation in 3T3 cells expressing WT EGFR. The cells were treated with indicated concentrations of each drug for 16 h and stimulated with EGF (10 ng ml21) 15 min before lysis. Cell extracts were immunoblotted to detect the indicated proteins. c, Immunohistochemical analysis of hair bulb from erlotinib- or WZ4002-treated mice using EGFR and pY1173 EGFR. Only treatment with erlotinib results in significant inhibition of EGFR phosphorylation. Scale bar, 50 mm. d, Quantification of frequency of phospho-EGFR staining from vehicle- (n5 3), erlotinib- (n5 3) and WZ4002- (n5 2) treated mice. The means and standard deviations are plotted for drug treatment. NATURE |Vol 462 |24/31 December 2009 LETTERS 1071 Macmillan Publishers Limited. All rights reserved©2009 CALMAN Typewritten Text 肺癌QQ群2970957 CALMAN Typewritten Text 肺癌QQ群2970957 We profiled WZ3146 and WZ4002 against a panel of 400 kinases using the Ambit kinome screening platform (Supplementary Table 3 and Supplementary Fig. 5). For WZ4002, kinases that exhibited greater than 95% inhibition relative to the dimethylsulphoxide control (Ambit score ,5) at 10mM were selected for measurement of their dissociation constants (Supplementary Table 3). In addition to EGFR, we observed potent inhibition of several of the ten kinases that possess a cysteine at the same position as EGFR, including a subset of the TEC- family kinases (Supplementary Fig. 6). Cross-reactivity with BMX has been reported for irreversible quinazoline-derived EGFR inhibitors16. To confirm whether the observed binding activity translated into cel- lular inhibition,WZ4002 andWZ3146 were profiled against Ba/F3 cells transformedwith TEL fusions of BMX, BLK, JAK2 and JAK3.WZ4002, which possesses an ortho-methoxy group at the C2-aniline substituent, is more selective for EGFR than WZ3146 (Supplementary Table 4). We next determined whether the increased potency of the WZ compounds against mutant EGFR also applied to WT EGFR. We used EGFR WT HN11 cells17 and Ba/F3 cells harbouring the EGFR vIII mutation, which contains aWT kinase domain (Fig. 2a and Sup- plementaryTable 2). These compoundswere 3- to 100-fold less potent, with WZ4002 being least potent, than CL-387,785 and HKI-272 at inhibiting the growth of the EGFR WT cells. Furthermore, WZ4002 was 100-fold less effective at inhibiting phosphorylation of WT EGFR than the quinazoline inhibitors (Fig. 2b). Similarly, WZ4002 inhibited EGFR kinase activity of recombinant L858R/T790M protein more potently than that of WT EGFR, whereas the opposite was observed with HKI-272 and gefitinib (Supplementary Fig. 7a, b). To understand better the potency and relative selectivity for EGFR T790M over WT EGFR, we determined the crystal structure of WZ4002 in complex with EGFR T790M (Fig. 3a, b, Supplementary Fig. 8 and Supplementary Table 5). The compound binds within the ATP-binding cleft of the enzyme, forming the expected covalent bond with Cys 797. As expected based upon co-structures of related pyrimidine-derived inhibitors with CDK2 (ref. 18), JNK1 (ref. 19) and FAK20, the anilinopyrimidine core of WZ4002 forms a bidentate hydrogen bonding interaction with the ‘hinge’ residue Met 793 (Fig. 3b). The chlorine substituent on the pyrimidine ring contacts the mutant gatekeeper residue, Met 790. The hydrophobicity con- ferred by this mutation likely contributes to the potency of these compounds against the T790M mutant. The aniline ring forms a hydrophobic interaction with the a-carbon of Gly 796 and its methoxy substituent extends towards Leu 792 and Pro 794 in the hinge region. The greater selectivity of theWZ4002 compound likely derives from the fact that both JAK3 and TEC-family kinases have a bulkier residue (tyrosine in JAK3, phenylalanine in TEC-family kinases) in the position of Leu 792, which would be expected to interfere sterically with the methoxy group in WZ4002. The reactive acrylamide moeity and the linking phenyl ring comprise the other arm of the inhibitor. The ‘linker’ phenyl ring lies roughly perpendic- ular to the pyrimidine core; this orientation juxtaposes the acryla- mide with the thiol of Cys 797 for covalent bond formation (Fig. 3b). We further determined whether WZ4002 is effective in vivo by using mouse lung cancer models harbouring either EGFR L858R/T790M or Del E746_A750/T790M. We chose WZ4002 for the in vivo studies because in vitro it was least potent against WT EGFR and other cysteine-containing kinases (Supplementary Tables 2 and 4) but was effective againstEGFRT790M.Apharmacokinetic studywasperformed to determine the achievable plasma concentration (429ngml21), half- life (2.5 h) and the oral bioavailability (24%) of WZ4002 (Supplemen- tary Tables 6–8). In a pharmacodynamic study, WZ4002 effectively inhibited EGFR, AKT and ERK1/2 phosphorylation (Fig. 4a), which was associated with a significant increase in TdT-mediated dUTP nick-end labelling (TUNEL)-positive and a significant decrease in Ki67-positive cells compared with mice treated by vehicle alone (Fig. 4b, c). To evaluate whether WZ4002 imparted a differential effect on WT EGFR in vivo, we evaluated EGFR phosphorylation in the hair bulb from mouse skin after treatment with either erlotinib or WZ4002 (Fig. 2c, d).Only erlotinib significantly inhibitedEGFRphosphorylation in the hair bulb. In a 2-week efficacy study, WZ4002 treatment resulted in significant tumour regressions compared with vehicle alone in both T790M-containing murine models (Fig. 4d, e and Supplemen- tary Fig. 9). Histological evaluation of the lungs after treatment con- firmed significant resolution of the tumour nodules, with only a few small residual nodules and nodule remnants that had evidence of treat- ment effect with decreased cellularity and increased fibrosis consistent with remodelling/scarring (Fig. 4f). Therewere no signs of overt toxicity comparedwithmice treatedbyvehicle aloneduring the studyas assessed by changes in weight (data not shown), serum creatine and total white blood cell count (Supplementary Fig. 10). Our studies identify a novel structural class of EGFR kinase inhi- bitors that are effective in vitro and in in vivomodels harbouring the EGFR T790M mutation. Given the marked activity in models with established EGFR T790M, we determined whether WZ4002 treat- ment could also prevent the development of EGFR T790M using in vitro models harbouring EGFR-activating mutations. Unlike with gefitinib or HKI-272, which when used at their achievable plasma concentrations lead to development of EGFR T790M in vitro13,21,22, a b Met 790 Leu 792 Met 793 Pro 794 Val 726 WZ4002 Leu 718 Cys 797Gly 796 WZ8040 WZ4002 Aniline Methyl- piperazine Pyrimidine Methoxy Acryl- amide Figure 3 | Crystal structure of WZ4002 bound to EGFR T790M. a, Chemical structures of WZ8040 and WZ4002 are shown schematically in a manner resembling the conformation adopted in complex with the kinase. b, Crystal structure of WZ4002 in complex with EGFR T790M mutant (PDB ID 3IKA). WZ4002 binds the active conformation of the kinase, with both the regulatory C-helix and the ‘DFG’ segment of the activation loop in their inward, active positions. The EGFR kinase is shown in a ribbon representation (blue) with the bound inhibitor in yellow. Side-chain and main-chain atoms are shown for selected residues that contact the compound. Expected hydrogen bonds to the backbone amide and carbonyl atoms of Met 793 are indicated by dashed lines. Note also the covalent bond with Cys 797. The structure was refined to a crystallographic R value of 21.3% (Rfree5 25.4%) with data extending to 2.9-A˚ resolution (see Methods for further crystallographic details). LETTERS NATURE |Vol 462 |24/31 December 2009 1072 Macmillan Publishers Limited. All rights reserved©2009 we were unable to isolate any EGFR-T790M-containing clones from WZ4002-treated Ba/F3 or PC9 NSCLC cells (Supplementary Table 9). These findings suggest that WZ4002 could also be used as initial therapy for patients with EGFR-mutant NSCLC and may ultimately lead to a longer time to disease progression than currently achieved with gefitinib1. Our crystallographic studies provide insight into why WZ4002 is so much more effective against L858R/T790M than HKI-272. Although both share the irreversible component, the anilinopyrimi- dine scaffold of WZ4002 is an intrinsically better fit for the mutant gatekeeper methionine (Supplementary Fig. 11). To test this hypo- thesis further, we prepared WZ4003, a reversible analogue of WZ4002 that is non-reactive towards Cys 797. WZ4003 binds to the L858R/T790M mutant 100-fold more tightly than it does to the WT EGFR (Supplementary Table 10), confirming that the scaffold per se is indeed specific for the mutant kinase. Importantly, the WZ compounds rely on covalent bond formation for potent cellular inhibition, as evidenced by the 100-fold increase in the IC50 of WZ4002 against the EGFR C797S mutants and by the significantly reduced cellular IC50 of WZ4003 against T790M containing Ba/F3 cells (Supplementary Table 2). These observations highlight the importance of using a library of irreversible kinase inhibitors, as WZ4003 would not have been identified in the initial cellular screen. Mutations, including those at the gatekeeper residue, are a common mechanism of drug resistance to kinase inhibitors. The current approach, using a cellular screen expressing the mutant kinase of interest, can be applied to identify novel agents specifically against drug resistance or oncogenicmutations implicated in human cancers. Such agents