Article pubs.acs.org/jmc
Structure-Guided Development of Covalent and Mutant-Selective Pyrazolopyrimidines to Target T790M Drug Resistance in Epidermal Growth Factor Receptor Julian Engel,†,∥,# Steven Smith,†,# Jonas Lategahn,† Hannah L. Tumbrink,†,¶ Lisa Goebel,† Christian Becker,† Elisabeth Hennes,†,■ Marina Keul,† Anke Unger,‡ Heiko Müller,‡,⊥ Matthias Baumann,‡ Carsten Schultz-Fademrecht,‡ Georgia Günther,§ Jan G. Hengstler,§ and Daniel Rauh*,† †
Faculty of Chemistry and Chemical Biology, TU Dortmund University, Otto-Hahn-Straße 4a, D-44227 Dortmund, Germany Lead Discovery Center GmbH, Otto-Hahn-Straße 15, D-44227 Dortmund, Germany § Leibniz Research Centre for Working Environment and Human Factors (IfADo), TU Dortmund University, D-44139 Dortmund, Germany ‡
S Supporting Information *
ABSTRACT: Reversible epidermal growth factor receptor (EGFR) inhibitors prompt a beneficial clinical response in non-small cell lung cancer patients who harbor activating mutations in EGFR. However, resistance mutations, particularly the gatekeeper mutation T790M, limit this efficacy. Here, we describe a structure-guided development of a series of covalent and mutant-selective EGFR inhibitors that effectively target the T790M mutant. The pyrazolopyrimidine-based core differs structurally from that of aminopyrimidine-based third-generation EGFR inhibitors and therefore constitutes a new set of inhibitors that target this mechanism of drug resistance. These inhibitors exhibited strong inhibitory effects toward EGFR kinase activity and excellent inhibition of cell growth in the drugresistant cell line H1975, without significantly affecting EGFR wild-type cell lines. Additionally, we present the in vitro ADME/DMPK parameters for a subset of the inhibitors as well as in vivo pharmacokinetics in mice for a candidate with promising activity profile.
■
INTRODUCTION
rendering ATP-competitive inhibitors less effective in the presence of cellular ATP concentrations.14,17 Second-generation EGFR inhibitors that also contain a 4-aminoquinazoline core (afatinib18 and dacomitinib19) were developed and thought to overcome T790M drug resistance as they incorporate a Michael acceptor that covalently modified a unique cysteine (Cys797) at the lip of the EGFR ATP-binding site (Figure 1).20−22 These covalent inhibitors, characterized by marginal off-rates as well as maximal drug−target residence time, were expected to inhibit T790M drug-resistant EGFR efficiently, but at clinically achievable concentrations they displayed insufficient efficacy in patients due to severe side effects. The clinical data revealed that 4-aminoquinazoline inhibitors are inappropriate to target the gatekeeper mutant variant of EGFR, as they likewise inhibit wild-type EGFR, accounting for the observed side effects and represent a dose-limiting factor.20,21,23−25 Hence, a third-generation of EGFR inhibitors (CO-1686 (rociletinib26) and AZD9291 (osimertinib27)), structurally based on an aminopyrimidine, were developed and seemed to avoid a steric clash with the mutant methionine gatekeeper (Figure 1). Both covalent
First-generation epidermal growth factor receptor (EGFR) inhibitors gefitinib1 and erlotinib2 provided clinical benefit in patients harboring somatic activating mutations in the gene encoding EGFR such as the L858R point mutation and exon 19 deletions that drive oncogenesis in non-small cell lung cancer (NSCLC) (Figure 1). Treatment with tyrosine kinase inhibitors (TKIs) achieved impressive objective response rates of 50−80% and a doubled median progression-free survival, therefore representing a prime example of targeted cancer therapy that opened up a new era in cancer treatment.2−7 However, patients acquired resistances and suffered a relapse within 18 months of treatment. Biopsies of resistant tumors revealed a single point mutation of the gatekeeper residue (T790M) as the major resistance mechanism.8−10 Replacement of the threonine with methionine leads to steric repulsion of the 4-aminoquinazoline-based first-generation EGFR inhibitors and results in a slightly different binding mode that accounts for a dramatic loss of inhibitory activity.11−13 In addition to the steric repulsion, the change in ATP affinity is well documented.14−17 While the L858R point mutation leads to a distinct loss of ATP affinity to EGFR, the T790M mutation restores the ATP affinity to almost the same level as in wild-type EGFR, © 2017 American Chemical Society
Received: April 3, 2017 Published: August 30, 2017 7725
DOI: 10.1021/acs.jmedchem.7b00515 J. Med. Chem. 2017, 60, 7725−7744
Journal of Medicinal Chemistry
Article
Figure 1. Chemical structures of representative examples of the three generations of EGFR inhibitors. The electrophiles are highlighted in green.
we sought to identify a new set of covalent inhibitors that specifically target the gatekeeper mutant. Considering the knowledge we gained from first- and second-generation inhibitors as well as from our previous work, our structureguided design was focused on the following three main criteria, which we expected to be crucial for the development of an effective inhibitor that targets the drug-resistant gatekeeper mutant form of EGFR: (i) a distinct scaffold to specifically target the Met-gatekeeper mutant avoiding steric interference, (ii) mutant-selective inhibition while sparing the wild-type form as an essential requirement to reduce on-target toxicity that is associated with severe side effects, and (iii) an electrophile optimized with respect to geometry and orientation for the efficient alkylation of Cys797 as a critical element to overcome T790M drug resistance.35−37 We focused our design efforts on the well-studied 4-amino pyrazolopyrimidine as a core element as this moiety supplies an appropriate hinge-binding element and allows for the rapid and straightforward synthesis of a focused collection of small molecules for exploring the efficiency of potent inhibition of the kinase activity of EGFRT790M, the selectivity profiles, and covalent binding characteristics.35,38−40 We identified N1 of the pyrazolopyrimidine as a suitable position for the attachment of a Michael acceptor linked to a 3-piperidine as the electrophilic system that has been shown to be appropriate for alkylation of an analogous cysteine to EGFR (Cys481 in Bruton’s tyrosine kinase, BTK) (Figure 2).41,42 Furthermore, we have learned from previous studies on EGFR inhibitors that a restricted rotational flexibility of the electrophile’s linker, which enables a preferred orientation toward Cys797, has a major effect on the kinetics of the covalent bond formation that may drive the inhibitory potency.37 As 4-amino pyrazolopyrimidines, with a PP143 derived substitution pattern, provide this potential, we focused our design on the SARs of various substitutions at the 3-position to
inhibitors showed excellent in vivo potency as well as reduced on-target toxicity and could progress into human clinical trials.26,27 Rociletinib has recently been updated to an objective response rate of 45% to T790M-positive patients. Furthermore, elevated levels of glucose in the blood were observed which could be explained by a noncovalent metabolite that inhibits the insulin-like growth factor 1 and the insulin receptor and, thus, its clinical development was stopped.28−32 In fact, only one third-generation covalent inhibitor has yet achieved FDA approval. Osimertinib showed excellent results in clinical studies as well as reduced on-target toxicity and was approved by the FDA as Tagrisso in 2015.33,34 Here, we present the expanded structure-guided development and the biochemical as well as cellular evaluation of a pyrazolopyrimidine-based series of covalent inhibitors that target the drug-resistant EGFR mutant T790M in a highly mutant-selective manner. In a previous publication, we presented complex crystal structures of mutant EGFR, which revealed the binding mode of this class of pyrazolopyrimidines and gave a first explanation for the potency and selectivity toward the gatekeeper mutated variant.35 On the basis of these preliminary results, we now further investigated the structure− activity relationship (SAR) and examined the in vitro ADME/ DMPK parameters for a representative set of the pyrazolopyrimidines as well as in vivo pharmacokinetics for one candidate that displayed the most promising overall profile.
■
RESULTS Structure-Guided Development of Pyrazolopyrimidine-Based Inhibitors. Prompted by the failure of secondgeneration EGFR inhibitors in clinical studies and further stimulated by the significant impact of the first covalent T790M inhibitors beyond 4-aminoquinazolines osimertinib and rociletinib, 7726
DOI: 10.1021/acs.jmedchem.7b00515 J. Med. Chem. 2017, 60, 7725−7744
Journal of Medicinal Chemistry
Article
Figure 2. Schematic representation of a structure-based design approach to develop a new series of EGFR inhibitors to target the T790M gatekeeper mutant of EGFR. The design strategy is based on a 4-amino pyrazolopyrimidine scaffold and is focused on the particular derivatization of the 3-position of the pyrazolopyrimidine and the N3-position of the piperidine moiety (each highlighted in pink). The 4-amino pyrazolopyrimidine binding mode is adopted (PDB 4hct) and was modeled into the active site of EGFR-T790M (PDB 3ika).
Figure 3. Rational design of the pyrazolopyrimidines to covalently address the methionine gatekeeper mutant of EGFR. Schematic illustration of the pyrazolopyrimidine core structure derived from the nonselective kinase inhibitor PP1 and proposed substitutions to investigate the interplay with Met790 (R1) and to gain valuable insights into the covalent bond formation with Cys797 (R2).
investigate in what way a change in size, orientation, and electronic properties will affect the activity and selectivity on the different EGFR mutant variants (Figures 2 and 3). We initially designed derivatives that contained rather simple moieties at 3-position that differed in their electronic properties (e.g., containing a halogen, ethyne, or cyclopropyl). To further investigate the tolerable size of C-3 substitutions, we included benzyl and phenyl moieties with different types of halogen substitutions in the meta and para positions. Ultimately, we designed derivatives with sterically advanced bicyclic moieties at C-3 to exploit the maximum tolerable size of lipophilic extensions into the pocket adjacent to the gatekeeper residue Met790. Synthesis of a Focused Library of Covalent Pyrazolopyrimidines. We synthesized a small, focused library of covalent pyrazolopyrimidines in which we paid particular attention to the derivatization of the 3-position of the
pyrazolopyrimidine core to elucidate on the maximal tolerable size for the pocket near the methionine gatekeeper (Met790) and the potential for beneficial interactions while retaining the crucial direct hinge contact. Further efforts were directed to exploring covalent bonding to Cys797 by altering the type of Michael acceptors and to producing reversible counterparts for comparative evaluation. We employed a general synthetic route for synthesizing pyrazolopyrimidines 1d−z (Scheme 1).35 Iodination of the commercially available 1H-pyrazolo[3,4-d]pyrimidin4-amine (2) at the 3-position resulted in the formation of compound 3. Subsequent Mitsunobu coupling was performed with N-Boc-protected (S)-3-hydroxypiperidine to generate chiral compound 4, which served as a generic intermediate for rapid library generation. Suzuki coupling with various aryl, benzyl, and heteroaryl boronic acids or esters as well as Sonogashira coupling with cyclopentane-equipped alkynes led to intermediate 5. Subsequent Boc deprotection with TFA 7727
DOI: 10.1021/acs.jmedchem.7b00515 J. Med. Chem. 2017, 60, 7725−7744
Journal of Medicinal Chemistry
Article
Scheme 1. Synthesis of a Subset of Pyrazolopyripimidine Derivatives (1d−z)a
Reagents and conditions: (i) NIS, DMF, 80 °C; (ii) N-Boc-(S)-3-hydroxypiperidine, DIAD, PPh3, THF, rt; (iii) boronic acid/ester, Pd(PPh3)4, satd Na2CO3, DME/EtOH (3:1, v/v), 90 °C; (iv) CuI, Pd(PPh3)2Cl2, TEA, rt; (v) TFA in DCM, rt; (vi) acryloyl or propionyl chloride, DIPEA, THF, 0 °C; (vii) (E)-4-(dimethylamino)but-2-enoic acid, EDC, HOBt, DIPEA, DCM, rt.
a
the inhibition of L858R (IC50 = 0.003 μM) and L858R/T790M (IC50 = 0.04 μM) but led to a complete loss of T790M selectivity (ratio WT:T790M = 0.5). The phenol moiety seemed to have a major influence on the activity, most probably due to a contact with an adjacent glutamate residue (Glu762) of the helix C (Figure 4B). To further investigate the potential benefits of the penetration in the subpocket near the gatekeeper residue, we attached ethynyl- and prop-1-ynyl bridged cyclopentane moieties (1s and 1t, respectively) to restrict the substituents in an orientation facing the gatekeeper residue. In contrast to compound 1c (terminal alkyne), compounds 1s (IC50 = 1.4 μM) and 1t (IC50 = 0.9 μM) displayed less inhibitory activity against L858R/T790M and more strongly inhibited activating mutant L858R (IC50 = 0.06 and 0.1 μM, respectively), sustainably indicating that the extended substitutions underwent steric repulsion with Met790. Notably, the benzylequipped compound 1r demonstrated potent kinase inhibition of the relevant mutant variants of EGFR displaying a 24-fold increased inhibition against EGFRL858R/T790M compared to the phenyl derivative (1d). The more flexible benzyl group probably adopts an orientation toward Met790 and particularly Lys745 that allows a more beneficial interaction (Figure 4C). Triggered by this observation, we sought to increase the size of the substituents and attached bicyclic aromatic moieties (compounds 1u−z). Therefore, we incorporated 1-naphthyl (compound 1x) and 2-naphthyl (compound 1u) moieties at the 3-position of the pyrazolopyrimidine scaffold. These alterations increased the inhibition of the activating mutation (3-fold increase on L858R compared to 1r) and particularly of the drug-resistant (L858R/T790M; 1x IC50 = 0.002 μM; 1u IC50 < 0.001 μM) mutants of EGFR, leading to a superior selectivity over EGFR-WT (ratio WT:T790M; 1x = 30; 1u > 20). Conversion of naphthyl into an azaindole (compound 1v) or a methylindole (compound 1w) yielded comparable trends with IC50 values in the single-digit nanomolar range. As these inhibitors showed preferential inhibition of the gatekeeper mutated form of EGFR, the highly lipophilic and sterically demanding moieties seemed to induce crucial interactions while avoiding adverse interference with the gatekeeper residue (Supporting Information, Figure S1). In addition to their high potency, the bicyclic aromatic substitutions reduced the
resulted in secondary amine 6, which was decorated with a Michael acceptor using acryloyl chloride, resulting in structures 1d, 1g, 1i, 1k, 1n, 1p−t, 1v, and 1y−z or with 4-(dimethylamino)-2-butenoic acid to generate amides 1f and 1m via standard amide coupling. Reversible counterparts (1e, 1h, 1j, 1l, and 1o) were synthesized with propionyl chloride. Compounds 1a−c were synthesized through a slightly altered synthetic approach presented in the Supporting Information (Schemes S1−S3). The entire chemical data of compounds 1u, 1x, and 1w have been previously described.35 Covalent Pyrazolopyrimidines Effectively Inhibit the Kinase Activity of the Drug-Resistant EGFR Mutant. We used an activity-based assay to evaluate the inhibitory activity of pyrazolopyrimidines 1a−z on wild-type, activating (L858R), and drug-resistant (L858R/T790M) EGFR (Table 1). Notably, structure 1a inhibited the wild-type, L858R, and L858R/ T790M variants of EGFR (IC50 = 6.9, 2.5, and 1.6 μM, respectively) and shows a reasonable selectivity toward the drug-resistant mutant of EGFR (ratio WT:T790M = 4.3). Because compound 1a possesses a hydrogen atom at C-3 that is not suitable for beneficial interaction with Met790, the observed activity is directly linked to the inhibitory effect of the acrylamide piperidine-decorated 4-amino pyrazolopyrimidine core, rendering it potentially useful for covalently targeting EGFR-T790M. The introduction of small lipophilic moieties showed an interesting effect on all variants of EGFR. While an ethyne at the 3-position (compound 1c) led to an increased activity on all EGFR variants but a loss of selectivity (ratio WT:T790M = 2), the introduction of iodine (compound 1b) resulted in an even lower kinase activity but a very good selectivity toward the double mutant of EGFR (EGFR-WT IC50 = 0.3 μM, EGFR-L858R/T790M IC50 = 0.02 μM, and ratio WT:T790M = 14, respectively). For structure 1d, which is equipped with a phenyl group at the 3-position, we could observe a loss in activity and selectivity compared to 1b, indicating an insufficient interaction within the lipophilic subpocket that is in line with the X-ray-based modeling studies (Figure 4A). Phenyl derivatives with an additional fluorine or trifluoromethyl group in the meta or para position (1g, 1i, 1l, and 1n, respectively) or the introduction of the bioisosteric cyclopropyl moiety (1p) tended to follow the same inhibitory profile. Notably, introduction of a 4-phenol moiety increased 7728
DOI: 10.1021/acs.jmedchem.7b00515 J. Med. Chem. 2017, 60, 7725−7744
Journal of Medicinal Chemistry
Article
Figure 4. Model of 1d (A), 1q (B), 1r (C), and 1y (D) in EGFR-T790M (PDB 5j9y). The models provide a reasonable explanation for the observed SAR. 1y with its potential of steric repulsion between Met766 and Glu762 are in line with the rather moderate activity in the biochemical assay. In contrast, 1d with a comparable activity lacks sufficient hydrophobic interactions in the back pocket of EGFR-L858R/T790M. With its intrinsic flexibility, the benzyl moiety of 1r is capable of adapting a position stacked in between Met790 and Lys745, comparable to the more potent bicyclic derivatives. The phenol derivate (1q) seems to be able to form an additional hydrogen bond with Glu762 and therefore gain selectivity compared to the other monocyclic derivatives.
compounds 1u, 1x, or 1w confirmed covalent modification consistent with a complete labeling of the targeted protein. Nano-LC-MS/MS analysis precisely verified a single alkylation of Cys797.35 Mutant-Selective Inhibitory Effects on Drug-Resistant NSCLC Cell Line H1975. We evaluated the inhibitory efficacy of pyrazolopyrimidine-based compounds 1a−z on a cell line with an EGFR activating mutation (HCC827, EGFRdelE746_A750) and on the drug-resistant NSCLC cell line H1975, which harbors EGFR-L858R/T790M. To investigate mutant-selective characteristics, we also tested the compounds’ effect on a cancer cell line harboring wild-type EGFR (A431, Table 2). Additionally, we tested on KRAS-mutated cell lines (A549, and H358; Table 2, and Supporting Information, Table S1, respectively) to exclude the possibility of off-target toxicity. Cells were treated with compounds 1a−z (0−30 μM) for 96 h. Gefitinib, afatinib, N-(3-((5-chloro-2-((2-methoxy-4-(4methylpiperazin-1-yl)phenyl)amino)pyrimidin-4-yl)oxy)phenyl)acrylamide (7, WZ400244), osimertinib, rociletinib, and ibrutinib served as references. Compounds 1u−x demonstrated excellent inhibitory effects on HCC827 (EC50 = 86%). With respect to plasma protein binding, the 2-naphthyl containing derivative was the only compound that showed a somewhat lower free fraction (f u100% = 0.02) compared to the other derivatives of the series that revealed satisfactory f u100% values ranging from 0.09 to 0.37, which are significantly higher than the fraction unbound levels determined for osimertinib, rociletinib, and ibrutinib (Table 3). As expected, the intrinsic clearance (Clint) of the compounds was very high in both mouse and human liver microsomes, thus being similarly unstable as
was achieved with compounds 1w and 1x (30-fold and 60-fold, respectively). Compound 1v evoked an inhibitory effect in EGFR wild-type cells (EC50 = 1.4, 3.5, and 2.3 μM in A431, A549, and H358, respectively), consistent with its strong in vitro inhibition of wild-type EGFR. Compound 1y, which contains a dimethylamine at the 4-position of the naphthyl moiety, exhibited less activity in H1975 cells (EC50 = 2.7 μM), in line with our activity-based data. Also consistent with our activity-based assay, compounds with less-lipophilic extensions at the pyrazolopyrimidine C-3 (compounds 1c, 1d, 1g, 1i, 1k, and 1p) had less effect on H1975 cells (EC50 = 1.8−16 μM), while the unsubstituted derivative 1a did not display inhibitory activity. Notably, compounds 1q (4-phenol substitution at C-3) and 1b (iodine substitution at C-3) strongly inhibited the T790M mutant in the biochemical setting (IC50 = 0.04 and 0.02 μM, respectively, Table 1) but were less potent against H1975 cells (EC50 = 1.84 and 3.9 μM, respectively). In accordance with our activity-based data, reversible inhibitors 1e, 1h, 1j, 1l, and 1o did not affect any of the tested cancer cell lines, in contrast to their covalent counterparts (compounds 1d, 1g, 1i, 1k, and 1n), highlighting how crucial a maximized drug−target residence time (characteristics of covalent inhibition) is to target T790M drug resistance. Further, the dimethylamino butenamide Michael acceptor equipped inhibitors 1f and 1m demonstrated significantly less inhibitory potency against H1975 cells (EC50 = 30 μM for compound 1f and EC50 = 21 μM for compound 1m) than their respective acrylamide analogues (compound 1d, EC50 = 3.7 μM; compound 1k, EC50 = 4.2 μM), also consistent with our activity-based assay data. ADME/DMPK Parameters of a Subset of Pyrazolopyrimidines. For a representative subset of pyrazolopyrimidines, the physicochemical and in vitro ADME parameters such as kinetic solubility and permeability as well as plasma and metabolic stability were examined (Table 3). Despite the 7731
DOI: 10.1021/acs.jmedchem.7b00515 J. Med. Chem. 2017, 60, 7725−7744
Journal of Medicinal Chemistry
Article
Table 2. Cellular Growth Inhibition (EC50) of Reference Compounds and Pyrazolopyrimidines against the Cancer Cell Lines A431, HCC827, H1975, and A549 cancer cell lines EC50 [μM]
a
compd
A431 (EGFR-WT)
HCC827 (EGFR-delE746_A750)
H1975 (EGFR-L858R/T790M)
A549 (KRAS-G12S)
ratio A431:H1975
1a 1b 1c 1d 1e 1f 1g 1h 1i 1j 1k 1l 1m 1n 1o 1p 1q 1r 1s 1t 1u 1v 1w 1x 1y 1z ibrutinib gefitinib afatinib 7 osimertinib rociletinib
nia 18.2 ± 7.06 29.6 ± 0.69 12.7 ± 4.37 ni 29.8 ± 0.39 11.7 ± 3.81 ni 17.0 ± 5.92 ni 16.0 ± 2.81 ni 13.7 ± 1.57 8.87 ± 4.38 ni 28.1 ± 4.06 4.17 ± 2.20 19.9 ± 9.59 10.9 ± 2.32 8.77 ± 2.88 3.60 ± 1.52 1.39 ± 0.59 6.78 ± 4.45 29.4 ± 1.67 24.5 ± 5.76 12.8 ± 4.63 2.38 ± 1.66 1.71 ± 0.79 0.2 ± 0.07 2.21 ± 0.44 0.67 ± 0.23 1.66 ± 0.12
9.90 ± 1.02 0.589 ± 0.13 2.05 ± 0.375 0.07 ± 0.02 ni 1.19 ± 0.48 0.07 ± 0.35 22.3 ± 4.85 0.08 ± 0.03 25.9 ± 3.58 0.65 ± 0.13 ni 6.39 ± 1.62 0.16 ± 0.08 23.6 ± 6.96 0.31 ± 0.12 0.07 ± 0.02 0.15 ± 0.02 1.15 ± 0.14 3.15 ± 0.55
