Noncovalent Wild-type–Sparing Inhibitors of EGFR T790M

“Oncogene-addicted” cancers define a clinical context in which rationally targeted drug therapies have been somewhat successful. In many cases, these cancers are defined by the presence of mutationally activated tyrosine kinases, such as BCR–ABL, c-KIT, HER2, and EGF receptor (EGFR; ref. 1). Consequently, intense efforts have been focused on tyrosine kinase inhibitors (TKI) as promising molecularly targeted therapies for genotype-defined subsets of cancers (1–4). In approximately 10% to 20% of non–small cell lung cancers (NSCLC), somatic-activating alleles affecting the catalytic kinase domain of EGFR have been well correlated with the clinical response to the small-molecule EGFR kinase inhibitors gefitinib and erlotinib (5, 6). In the adenocarcinoma subtype of NSCLC, nearly 90% of these mutations are the missense mutation L858R or in-frame small deletions affecting exon 19. Both of these mutations have been shown to promote the activation of EGFR signaling and a state of EGFR dependency (7, 8).

Despite the dramatic clinical responses to gefitinib and erlotinib that have been observed in some advanced NSCLCs, treated patients invariably develop acquired resistance to these drugs, typically around 1 year after the initiation of treatment (9). Approximately 50% of patients, who initially responded to EGFR TKI therapy and subsequently develop drug resistance have acquired, within their tumors, a secondary mutation within the EGFR kinase domain, a substitution of methionine for threonine at position 790 (T790M; refs. 10, 11). In vitro studies have shown that this mutation renders EGFR TKI refractory while preserving catalytic function in the presence of gefitinib or erlotinib.

Two potential mechanisms by which the EGFR T790M mutation confers drug resistance have been proposed. Several groups have focused on the “gatekeeper model,” which was originally described in the context of the analogous T315I mutation of the BCR–ABL fusion kinase associated with acquired drug resistance in chronic myelogenous leukemia patients treated with the ABL TKIs imatinib and dasatinib (12). Similarly, substitution with the bulkier methionine in EGFR T790M mutants causes a steric hindrance, thus preventing drug binding by EGFR inhibitors (10, 11, 13). A more recent report proposed another mechanism in which the T790M substitution increases the binding affinity of EGFR for ATP, resulting in reduced cellular potency of reversible EGFR TKIs (14). Although the specific resistance mechanisms associated with the T790M substitution remain controversial, relapsed NSCLCs with acquired T790M mutations seem to remain dependent on EGFR signaling for their growth, prompting substantial efforts to discover second-generation EGFR inhibitors that can overcome the effects of the T790M substitution.

Several second-generation EGFR kinase inhibitors that covalently bind to a cysteine residue within the EGFR catalytic domain (Cys 797) have shown preclinical therapeutic potential for overcoming EGFR T790M through increased occupancy of the ATP binding site (13, 15, 16). However, all of these irreversible inhibitors currently undergoing clinical testing, such as BIBW2992, PF00299804, and HKI-272, have thus far shown limited clinical efficacy, possibly because of their potency against wild-type EGFR, leading to skin rash and gastrointestinal toxicity, which has limited their maximal dosing to levels less than those that may be required to achieve drug exposure sufficient to overcome the EGFR T790M mutation (17, 18). An encouraging recent study, however, showed a preclinical irreversible pyrimidine-based mutant-selective EGFR inhibitor with greater potency against EGFR T790M than current clinical quinazoline-based irreversible inhibitors (19).

Using a high-throughput cancer cell line screening platform to profile 705 tumor-derived cancer cell lines for sensitivity to a variety of validated and investigational anticancer small compounds (20), we unexpectedly identified a bis-indole–based tool compound that inhibits EGFR T790M resistance–associated mutants and was largely inactive against wild-type EGFR. A structurally related reversible kinase inhibitor, PKC412, that is currently undergoing phase III clinical testing as a FLT3 kinase inhibitor was found to exhibit potent inhibition of EGFR T790M, while completely sparing wild-type EGFR. These findings indicate that it should be possible to develop reversible EGFR T790M inhibitors for which dosing is not limited by on-target toxicities, and may therefore be advantageous relative to currently available irreversible EGFR inhibitors.

Among a variety of kinase inhibitors profiled for growth-inhibitory activity against a panel of 705 human cancer cell lines derived from various solid tumor types, we tested Gö6976, a widely used staurosporine-related inhibitor of “classical” PKCs (protein kinase C-α, -β, and -γ), which have been implicated in oncogenesis (21). Less than 4% of tested cell lines exhibited strong sensitivity to this compound, as defined by more than 70% growth suppression at 1 micromolar (Fig. 1A; Supplementary Dataset S1). Notably, among the identified Gö6976-sensitive cell lines, 2 EGFR-mutant NSCLC cell lines, PC-9 and HCC827, were unexpectedly strongly growth inhibited by Gö6976.

We initially hypothesized that PKC might positively regulate the EGFR pathway and that disruption of this regulation by Gö6976 would lead to EGFR inhibition and growth suppression in these cells. To determine whether the classical PKC pathway drives proliferation and survival in PC-9 and HCC827 cells, we treated cells with bis-indolylmaleimide I (Bis) and sotrastaurin, 2 other classical PKC inhibitors. However, these inhibitors failed to detectably affect growth of either PC-9 or HCC827 cells, even at high concentrations (Supplementary Fig. S1A). Although all 3 inhibitors potently suppressed the PKC pathway, only Gö6976 induced apoptosis, as shown by PARP cleavage (Supplementary Fig. S1B). Moreover, inhibition of PKC-α and/or -β by RNA interference (RNAi) caused no measurable effects on EGFR signaling or cell viability, (Supplementary Fig. S1C and S1D), whereas RNAi-mediated knockdown of EGFR, as a positive control, induced apoptosis in both cell lines (Supplementary Fig. S1D). Together, these results suggest that Gö6976 promotes apoptosis in EGFR-mutant PC-9 and HCC827 cells by targeting a PKC-independent cell survival pathway.

To explore the molecular basis for Gö6976 sensitivity in PC-9 and HCC827 cells, we profiled this compound against a panel of 442 human kinases using the Ambit kinome profiling platform (Supplementary Dataset S2). Significantly, Gö6976 (at 500 nanomolar) exhibited substantial binding ­affinity for EGFR delE746_A750, T790M, or L858R/T790M mutants, whereas it showed significantly less affinity for wild-type EGFR (Fig. 1B).

By sorting the original cell line screening data on the basis of tissue origin, we found that, among 107 tested NSCLC-derived cell lines, 3 of 9 of the most sensitive cell lines harbor activating EGFR mutations, consistent with the observation that Gö6976 may be especially efficacious in EGFR-mutant cell lines (Fig. 1C). In light of the kinome profiling data, we selected 3 NSCLC lines harboring the EGFR delE746_A750 mutation and 1 NSCLC line with the L858R/T790M mutation for further evaluation. Gö6976 effectively reduced cell viability in each of these lines, with IC₅₀s of approximately 100 to 200 nanomolar for EGFR delE746_A750-mutant lines and 800 nanomolar for the L858R/T790M-mutant line (Fig. 1D). It has been previously reported that the EGFR-mutant cell lines PC-9 and HCC827 are very sensitive to erlotinib, whereas cells harboring an EGFR T790M mutant (NCI-H1975) are relatively erlotinib resistant (11). We compared the ability of Gö6976 and erlotinib to suppress EGFR signaling in these cells. Gö6976 suppressed EGFR signaling as efficiently as erlotinib in both PC-9 and HCC827 cells. Significantly, EGFR pathway signaling was effectively suppressed by Gö6976 in the erlotinib-resistant NCI-H1975 cells, with an IC₅₀ of approximately 100 nanomolar (Fig. 1E and F). Taken together with the Ambit profiling results, these findings suggested that Gö6976 can directly inhibit the activity of the EGFR T790M mutant.

To further explore the ability of Gö6976 to directly inhibit wild-type and mutant forms of EGFR, we used an in vitro autophosphorylation assay with the recombinant cytoplasmic domain of EGFR. We initially compared the ability of erlotinib and Gö6976 to inhibit ATP-dependent autophosphorylation of wild-type, T790M, and L858R/T790M forms of EGFR. As expected, erlotinib effectively inhibited autophosphorylation of wild-type EGFR and was ineffective against T790M or L858R/T790M-mutant forms. In contrast, Gö6976 was ineffective against wild-type EGFR, even at 10 micromolar, exhibited weak inhibitory activity against the isolated T790M-mutant EGFR, and very potently inhibited the autophosphorylation of the L858R/T790M double mutant EGFR at an IC₅₀ <100 nanomolar (Fig. 2A). It is worth noting that the purified L858R/T790M-mutant EGFR protein is associated with a baseline level of phosphorylation that can be detected even in the absence of added ATP; consequently, even complete inhibition of ATP-dependent autophosphorylating activity would be expected to yield a protein product associated with low-level phosphorylation in this assay. These findings show that Gö6976 is a wild-type–sparing potent inhibitor of the erlotinib-resistant L858R/T790M mutant.

Gö6976 is derived from staurosporine, an indolocarbazole alkaloid natural product, and is a relatively nonselective kinase inhibitor. Thus, although Gö6976 was originally developed to target classical PKCs, recent studies have revealed inhibitory activities for other cancer-associated kinases, including checkpoint kinase 1 (CHK1), fms-like tyrosine kinase 3 (FLT3), and Janus-activated kinase 2 (JAK2; refs. 22, 23). Notably, some indolocarbazole derivatives have shown greater kinase selectivity and have been developed for clinical use. Therefore we extended these findings to examine potential EGFR-inhibitory activity for other structurally related indolocarbazole derivatives that are currently undergoing clinical development. Two such compounds, PKC412 and CEP-701, are small-molecule inhibitors that were developed to target the FLT3 kinase, which is frequently activated in acute myeloid leukemia (24). Gö6976, CEP-701, and PKC412 share the same indolocarbazole backbone (Fig. 2B).

To examine the potential activity of these investigational compounds as inhibitors of NSCLC-associated EGFR mutants, we again used the in vitro EGFR autophosphorylation assay, as well as cell survival assays in the context of NSCLC cell lines expressing either wild-type or the T790M-mutant EGFR. In the in vitro biochemical assays, both PKC412 and CEP-701 were found to potently inhibit EGFR T790M and EGFR L858R/T790M autophosphorylating activity. Although CEP-701 displayed some weak activity against wild-type EGFR, PKC412 did not cause detectable inhibition of wild-type EGFR at a concentration as high as 10 micromolar. Moreover, PKC412 exhibited the greatest potency among the 3 indolocarbazole inhibitors tested against T790M-mutant EGFR and was more than 100-fold more potent than Gö6976, despite their structural similarity (Fig. 2C).

This activity was similarly reflected in cell viability assays. Thus, NCI-H1975 NSCLC cells, which harbor EGFR L858R in cis with the T790M substitution, are refractory to erlotinib but exhibit striking sensitivity to all 3 of the tested indolocarbazole compounds. In contrast, erlotinib displayed some growth-inhibitory activity on treated NCI-H322 NSCLC cells, which express wild-type EGFR (IC₅₀ ∼100 nanomolar), whereas these cells were largely refractory to the indolocarbazole compounds (Fig. 2D). Collectively, these data indicate that some indolocarbazole analogues, including 2 compounds currently undergoing clinical investigation, display potent inhibition of the EGFR T790M mutant in vitro and in NSCLC cell line models.

The dosing limitation associated with the irreversible EGFR inhibitors currently undergoing clinical evaluation, such as BIBW2992, PF00299804, and HKI-272, is likely to reflect, at least in part, their lack of selective inhibitory activity for T790M EGFR mutants versus wild-type EGFR (17, 18). Because we observed potent activity of PKC412 on EGFR T790M mutants, we directly compared the potency and selectivity of PKC412, BIBW2992, and HKI-272 against wild-type and mutant forms of EGFR. As reported previously, both of the irreversible inhibitors suppressed the autophosphorylating activity of wild-type EGFR at least as effectively as they inhibited T790M EGFR or L858R/T790M EGFR, indicating that these agents would be unable to reach a sufficient plasma concentration in patients to cause EGFR T790M inhibition without the adverse effects caused by inhibition of wild-type EGFR. Of these 2 inhibitors, HKI-272 displayed significantly less potency on EGFR T790M than wild-type EGFR (Fig. 3A). This finding is consistent with previous preclinical results, suggesting that a relatively high concentration of HKI-272 would be required to overcome EGFR T790M–medated erlotinib resistance in NSCLC cell lines (25). In contrast, EGFR T790M mutants were inhibited at concentrations of PKC412 as low as 3 to 30 nanomolar, whereas wild-type EGFR was not detectably affected by PKC412 concentrations as high as 10 micromolar (Fig. 3A).

It has been reported that pyrimidine-based irreversible EGFR inhibitors exhibit greater selectivity against EGFR T790M than clinical quinazoline-related inhibitors in tested cell lines and mouse models (19). To compare the properties of one of these irreversible inhibitors, WZ4002, with the reversible PKC412 inhibitor, we conducted an in vitro EGFR autophosphorylation assay. As previously reported, WZ4002 exhibited potent suppression of ATP-dependent autophosphorylation of EGFR T790M or EGFR L858R/T790M. At most tested concentrations, WZ4002 was ineffective against wild-type EGFR; however, unlike PKC412, at concentrations above 1 micromolar, WZ4002 showed detectable inhibition of wild-type EGFR (Fig. 3A).

To extend these findings, we next conducted enzyme inhibition studies with purified recombinant wild-type and mutant EGFR proteins to compare the activities of PKC412, BIBW2992, and WZ4002. All 3 compounds showed similar K_i values (<10 nanomolar) for the L858R/T790M and del E746_A750/T790M-mutant proteins, within 2-fold of each other, indicating that they are all very potent T790M inhibitors. Notably, only PKC412 and WZ4002 were more potent against the T790M mutants than wild-type EGFR (Supplementary Fig. S2A). Interestingly, when PKC412 was preincubated with enzyme before ATP addition, K_i values of PKC412 for both EGFR L858R/T790M and del E746_A750/T790M significantly shifted the selectivity ratio for wild-type EGFR to 92- and 58-fold, respectively (Fig. 3B). Moreover, when we compared off-rates of PKC412 for L858R/T790M and wild-type EGFR in this experimental setting, PKC412 showed a very slow off-rate (residence time of 133 minutes; data not shown) for the mutant only. This finding suggests that one mechanism contributing to the selectivity and potency of PKC412 against EGFR T790M may be the longer occupancy of the ATP binding site once it binds. Overall, PKC412 exhibited the greatest selectivity for EGFR T790M versus wild-type EGFR (Fig. 3B and Supplementary Fig. S2A and S2B). These results provide further evidence that PKC412 is a very potent inhibitor of T790M-mutant forms of EGFR and does not significantly inhibit wild-type EGFR.

To extend the comparative in vitro enzyme inhibition studies to cell line models, we next assessed the ability of the various EGFR T790M inhibitors to block ligand-induced (TGF-α) EGFR phosphorylation in NR6 cells (immortalized murine fibroblasts) engineered to stably express wild-type, L858R/T790M, or del E746_A750/T790M-mutant forms of EGFR. As shown, PKC412, BIBW2992, and WZ4002 displayed similarly potent ability to inhibit L858R/T790M, and WZ4002 was the most potent inhibitor of ligand-induced phosphorylation of del E746_A750/T790M (Fig. 4A and B). The immunoblot results were confirmed with the use of a quantitative immunocytochemical approach to detect phosphorylated EGFR (pEGFR) and total EGFR (Fig. 4A and B). In contrast, at submicromolar concentrations, HKI-272 only inhibited the L858R/T790M mutant. CEP-701 and Gö6976 effectively inhibited both of the T790M-containing mutants (Fig. 4C and D). Phosphorylation of wild-type EGFR in ligand-stimulated NR6 cells was largely suppressed by BIBW2992 and was modestly affected by WZ4002, only at high concentrations. Notably, activation of wild-type EGFR was not significantly inhibited by PKC412, consistent with the findings that PKC412 selectively inhibits T790M-containing forms of EGFR (Supplementary Fig. S3). Moreover, we confirmed the reversible nature of PKC412′s activity in PC-9 cells that had been selected in vitro for erlotinib resistance and were found to have acquired a del E746_A750/T790M EGFR mutation (Supplementary Fig. S4) by examining the recovery of pEGFR after treatment in a drug washout assay (Supplementary Fig. S5A and S5B).

We next examined the activity of the 3 most potent EGFR T790M inhibitors in NSCLC-derived cell lines, including NCI-H1975, which harbors an EGFR L858R/T790M mutation, and NCI-H820, which harbors an EGFR del E746_A750/T790M mutation. Significantly, PKC412 and BIBW2992 exhibited approximately 2-fold less potency, whereas WZ4002 showed approximately 40-fold less potency in NCI-H1975 cells (EGFR L858R/T790M) than in NR6 cells expressing L858R/T790M EGFR (Fig. 5A). Unexpectedly, the potency of PKC412 against del E746_A750/T790M EGFR in NCI-H820 was increased by approximately 100-fold relative to NR6 cells expressing the same EGFR mutant. However, WZ4002 potency against del E746_A750/T790M remained the same in both cell lines (Fig. 5B). A similar observation was made in PC-9 cells harboring a del E746_A750/T790M EGFR mutation (Fig. 5C). This cell line also displayed significant sensitivity to PKC412, approximately 30-fold more than that seen in NR6 cells expressing del E746_A750/T790M EGFR.

We then directly compared the efficacy of these various inhibitors in assays of proliferation and apoptosis. Notably, the growth inhibition assay revealed smaller IC₅₀ variations between inhibitors than the ligand-induced EGFR activation assay (Figs. 5A–C and 6A). EGFR signaling was invariably suppressed at early time points and apoptosis was induced at later time points by all of the tested inhibitors (Fig. 6B). These findings suggest that the mechanism of ligand-induced EGFR activation may differ from that of ligand-independent activation, and that PKC412 has greater potency against ­ligand-induced EGFR activation than other tested compounds in EGFR T790M-mutant NSCLC cells.

To confirm the observed PKC412 efficacy in an in vivo tumor model, we conducted xenograft studies using NCI-H1975 NSCLC and PC-9/ER lines, which harbor EGFR L858R/T790M and del E746_A750/T790M, respectively. We first determined whether PKC412 could suppress EGFR signaling in vivo. Established NCI-H1975 tumor xenografts were treated daily for 3 days with 100 mg/kg, 200 mg/kg of PKC412, vehicle, or 100 mg/kg of gefitinib. Significant suppression of pEGFR and pAKT was observed following PKC412 treatment (Fig. 7A), whereas vehicle ­control or gefitinib (100 mg/kg) failed to inhibit EGFR signaling, as previously reported (11). We then assessed the ability of these treatments to retard tumor growth. Mice bearing established NCI-H1975–derived tumors were treated daily with orally administered PKC412, gefitinib, or vehicle control for the duration of the study (Fig. 7B and C). PKC412 significantly suppressed tumor growth by 50% compared with control-treated animals at day 12.

In the PC-9/ER xenograft study, we included a WZ4002 treatment group for comparison (Fig. 7D and E). Notably, PKC412 did not suppress tumor growth until 12 days on treatment, after which time tumors began to regress. These kinetics of response were similarly observed in the NCI-H1975 xenograft study (Fig. 7B). In both studies, no significant body weight loss was detected (Supplementary Fig. S6A and S6B), suggesting that the observed tumor regression is not due to accumulated toxicity of PKC412. In addition, immunoblotting of PKC412-treated xenografts (4 hours after the last dosing), confirmed effective suppression of EGFR signaling in tumors, indicating that the inhibitory effect of PKC412 against EGFR T790M was maintained during the course of treatment (Fig. 7C and E). Tumor growth in mice treated with WZ4002 was efficiently inhibited for 21 days of the study. Pharmacokinetic analysis was conducted from plasma and tumor samples immediately after the final dosing and revealed a C_max of 2.91 micromolar 30 minutes after treatment and a plasma concentration of 0.659 micromolar 24 hours after treatment (Supplementary Fig. S7A–S7C).

We extended the in vivo analysis to examine the efficacy of PKC412 in a genetically engineered NSCLC model that develops lung adenocarcinomas driven by transgenic expression of EGFR L858R/T790M. In this study, PKC412-treated tumors were growth inhibited for the first 2 weeks of treatment and then began to increase in size at 3 and 4 weeks of treatment in 3 mice, albeit at a slower rate than vehicle-treated tumors (numbers 4212, 3208, and 4220), whereas steady tumor growth suppression during the overall time period of treatment was observed in the other 3 treated mice (Supplementary Fig. S8A). No significant weight loss was observed during 4 weeks of PKC412 treatment (Supplementary Fig. S8B). Biochemical analysis showed that the antitumor activity of PKC412 is well correlated with a significant decrease in phosphorylation of EGFR L858R/T790M of tumors treated for 5 days, indicating that PKC412 can effectively suppress EGFR T790M activity in vivo in this model (Supplementary Fig. S8C).

The successful development of gefitinib and erlotinib for the treatment of EGFR-mutant NSCLCs has been a very significant advancement in the clinical management of metastatic cancer. Moreover, the observed association between the clinical activity of these agents and the presence of activated alleles of EGFR within a subset of tumors has helped to advance the paradigm of “personalized medicine” for patients with cancer. Although these drugs were initially developed as inhibitors of wild-type EGFR, the fact that they exhibit approximately 10-fold increased potency against the mutationally activated forms of EGFR, which has been attributed to the reduced affinity of these mutants for ATP (26, 27), has probably also contributed to their clinical efficacy. Despite their impressive clinical activity in a subset of treated patients, the inevitable acquisition of drug resistance has prompted significant efforts to develop ­second generation inhibitors, especially those that can overcome the frequently observed EGFR T790M gatekeeper mutation. Although the first-generation EGFR inhibitors effectively compete for ATP in the context of the clinically observed activating mutants, the T790M substitution at the gatekeeper position restores ATP affinity, consequently imposing a more formidable challenge for competitive inhibition (14).

The need to overcome EGFR T790M-mediated resistance to gefitinib or erlotinib has prompted substantial efforts to discover inhibitors that exhibit increased potency against EGFR T790M or limit access of ATP to the binding pocket. Thus far, all of the reported investigational agents that have been developed to target EGFR T790M are irreversible EGFR inhibitors which covalently occupy the ATP binding site as a means of reducing ATP binding and thereby inhibiting catalysis. Indeed, these second generation irreversible inhibitors, including BIBW2992, PF00299804, CI-1033, and HKI-272 exhibit substantially greater potency against EGFR T790M than gefitinib or erlotinib in in vitro studies. However, like the first generation inhibitors, the irreversible inhibitors are also very active against wild-type EGFR, leading to an on-target dose-limiting toxicity associated with severe skin rash in some patients. Our studies have confirmed that BIBW2992 and HKI-272 are in fact more potent against wild-type EGFR than the T790M-containing mutants, and due to dose-limiting toxicity, such inhibitors may not reach sufficient plasma concentrations to effectively inhibit EGFR T790M in patients (17).

WZ4002 is a recently reported irreversible inhibitor that displays much improved selectivity against EGFR T790M over wild-type EGFR in preclinical studies (19). Indeed, at concentrations as low as 10 nanomolar, WZ4002 inhibits ATP-dependent autophosphorylation of EGFR T790M without any observed effects on wild-type EGFR. However, at higher concentrations, we found that this compound also inhibits wild-type EGFR in vitro and in cell line studies. Because of the irreversible nature of its inhibitory mechanism, prolonged administration of this agent could potentially affect wild-type EGFR, particularly in tissues where it accumulates.

The noncovalent indolocarbazole compounds described here, including Gö6976 and PKC412, seem to function as more selective inhibitors of EGFR T790M, which is consistent with a recent study implicating Gö6976 as an EGFR T790M inhibitor (28). Importantly, we showed that PKC412 is at least 100-fold more potent against EGFR T790M than Gö6976 and is comparably potent with the irreversible EGFR inhibitors, but with 500-fold less activity against wild-type EGFR in drug-treated cells. Consistent with these findings, the in vitro binding specificity of indolocarbazole compounds for EGFR T790M has been previously reported (29). PKC412 is currently undergoing phase III clinical testing in patients with acute myelogenous leukemia (AML) with activating FLT3 mutations, and its safety has been shown in a phase I trial of patients with solid tumors. Although PKC412 is a potent inhibitor of PKC, FLT3, KIT, KDR, PDGFR-α, and -β in vivo (30–33), inhibition of these signaling kinases may not be detrimental to normal cell physiology in most tissues. Similarly, it has been somewhat unexpectedly observed that potent multikinase inhibitors, including dasatanib and sunitinib, can be safely administered to patients with cancer.

Recently, a concern regarding the potential clinical efficacy of PKC412 was raised with respect to its relatively low free concentration in plasma due to its rapid metabolization rate, and substantial binding affinity for the serum protein alpha-1-acid glycoprotein (AAG; ref. 34). However, despite these pharmacokinetic limitations, the clinical findings with FLT3-mutant AML patients treated with PKC412 have been encouraging. Moreover, clinical responses were well correlated with FLT3 inhibition, and despite the rapid metabolization of PKC412, concentrations of its major metabolite, CGP52421, remain high (20–25 micromolar) in plasma and may be active against the target (24, 34–36). Furthermore, the direct measurement of PKC412 and its metabolites in solid tumor tissues of metastatic melanoma patients treated with PKC412 at 225 mg/day for 28 days showed that the tissue concentration of PKC412 was relatively high (median 593 nanomolar)—8 to 200 times the IC₅₀ required for inhibition of EGFR T790M mutants in NCI-H1975 and PC-9/ER cells. Considering that the in vitro binding affinity of CGP52421 (median tissue concentration, 1.5 micromolar) for EGFR T790M mutants is similar to that of PKC412, the cumulative plasma concentration of CGP52421 and PKC412 would be expected to be sufficient to inhibit EGFR T790M (37, 38). Notably, although our xenograft studies did not show tumor regression in the relatively short treatment window, the tumor inhibition curve showed a trend to regression at the time when animals needed to be taken down due to vehicle-associated ulceration in the NCI-H1975 model, and significant tumor regression was observed in the PC-9/ER model, raising the possibility that the slow accumulation of the CGP52421 metabolite could yield increased efficacy after a period of time during which an active drug metabolite accumulates.

In summary, these findings show the use of broad cancer cell line sensitivity profiling to identify unanticipated and potentially useful applications for small-molecule kinase inhibitors. The follow-up studies have provided a proof-of-principle demonstration that some selective indolocarbazole derivatives can function as potent inhibitors of EGFR T790M in a reversible manner in vitro and in vivo, largely sparing wild-type EGFR. This suggests that such inhibitors may be effective without the adverse effects associated with the irreversible EGFR T790M inhibitors currently undergoing clinical evaluation, which are also potent inhibitors of wild-type EGFR. It is also possible that these reversible inhibitors could be used in combination with first-generation EGFR TKIs, as the first-generation inhibitors seem to be somewhat more active against the classical EGFR-activating mutations, and might therefore be most effective once a secondary T790M mutation has been acquired. Because some NSCLC tumors harbor multiple EGFR alleles, with or without the presence of T790M mutations (8), a combination strategy might be required to optimally suppress signaling from more than one form of mutant EGFR within a single tumor. Finally, it will be of interest to determine whether structurally related alternative inhibitors can be generated that maintain these properties while showing even greater selectivity for EGFR over other kinases.

Human cancer cell lines were tested to assess treatment effects on viability using an automated platform as previously described (20). Cells were treated with 1 micromolar Gö6976 for 72 hours and then assayed for cell viability. Cell lines were either maintained in RPMI-1640 or in DMEM/F12 (GIBCO) supplemented with 10% FBS (GIBCO), 50 U/mL penicillin, 50 U/mL streptomycin, and 2 mmol/L l-glutamine (GIBCO). PC-9 cells were kindly provided by Dr. Kazuto Nishio (National Cancer Center Hospital, Tokyo, Japan). HCC827, NCI-1975, and NCI-H820 cells were obtained from the American Type Culture Collection. Cells were tested and authenticated by single-nucleotide polymorphism genotyping. NR6 EGFR lines and PC-9/ER cells were authenticated by immunoblotting for EGFR and sequencing EGFR from PCR-amplified genomic DNA.

Cell viability was assessed using the fluorescent DNA-staining dye SYTO60 (Invitrogen). A total of 5 × 10³ cells were plated in 96-well plates in triplicate, and the following day cells were treated with a variety of drug concentrations. After 72 hours, cells were fixed in 4% formaldehyde and stained with SYTO60 followed by fluorescent measurement using the Odyssey Imaging system (absorption at 700 nm; LI-COR). Drug sensitivity was calculated as the fraction of drug-treated cells relative to untreated cells. Data were subjected to a nonlinear regression model, and drug–response curves were obtained using GraphPad Prism version 5.3 (GraphPad Software, Inc.).

Gö6976 was obtained from EMD Chemicals Inc. Bisindolylmaleimide I was purchased from Tocris Bioscience, and sotrastaurin was from Axon Medchem. Cep-701 and PKC412 were obtained from LC Laboratories. BIBW-2992 and HKI-272 were purchased from Selleckchem. Gefitinib was obtained from Astrazeneca. WZ4002 and erlotinib were synthesized at Genentech.

All EGFR recombinant proteins were purchased from Millipore. A total of 100 ng of protein was used for autophosphorylation reactions. Reactions were carried out in 8 mmol/L MOPS-NaOH pH 7.0, 1 mmol/L EDTA, 10 mmol/L MnCl₂, 10 mmol/L MgCl₂, 0.8 mol/L (NH4)₂SO₄, and 1 mmol/L ATP with or without various concentrations of inhibitors. Reactions were at 34°C for 15 minutes and were stopped by adding SDS-sample buffer. Samples were electrophoresed on SDS-PAGE, and phosphorylated EGFR was detected by an anti-phospho tyrosine antibody.

EGFR proteins were purchased from Invitrogen or Carna Biosciences. The protein constructs used in enzymatic assays were EGFR wild-type (catalytic domain aa668-1210), T790M (catalytic domain aa668-1210); L858R (catalytic domain aa668-1210), L858R/T790M (catalytic domain aa668-1210), delE746_A750 (catalytic domain aa669-745, 751-1210), delE746_A750/T790M (catalytic domain aa669-745, 751-1210). PKC412 was preincubated with EGFR kinase (wild-type 0.006 micromolar; T790M 0.017 micromolar; L858R 0.016 micromolar; L858R/T790M 0.0092 micromolar; delE746_A750 0.004 micromolar; delE746_A750/T790M 0.0046 micromolar) in 50 mmol/L HEPES, pH 7.5, 10 mmol/L MgCl₂, 4 mmol/L MnCl₂, 0.01% Brij-35, 1 mmol/L dithiothrietol (DTT) for 30 minutes followed by the addition of 5 micromolar ATP (Sigma) and 1 micromolar Fl-EEPLYWSFPAKKK-CONH₂ peptide substrate (Caliper Life Sciences). After an additional 30 minutes (60 minutes for EGFR T790M), reactions were terminated by addition of EDTA to a final concentration of 80 mmol/L. Concentrations of substrate and product were quantified using a mobility shift chip (Caliper Life Sciences) on a LabChip 3000 instrument. Dose response curves were fit to the percentage of inhibition data using the Morrison Equation. K_i values are subsequently calculated using the equation K_i = Morrison K_i/(1 + [ATP]/K_m) for competitive inhibitors (39). ATP K_m values for enzymes were as follows: wild-type, 1.7 micromolar; T790M, 2.0 micromolar; L858R, 5.3 micromolar; L858R/T790M, 1.3 micromolar; delE746_A750, 2.8 micromolar; and delE746_A750/T790M, 2.1 micromolar.

Cells were lysed in NP-40 lysis buffer containing protease inhibitor cocktail (Roche). Lysates were prepared by taking supernatants from centrifugation for 15 minutes at 12,000 × g. Equivalent amounts of proteins were loaded and separated by SDS-PAGE followed by transfer to membranes. Antibodies used for immune detection of proteins were pEGFR (Y845; #6963), EGFR (#2232), AKT (#9272), pERK (#9101), ERK (#9102), pS6 (#2211), S6 (#2217), PKCα (#2056), PKC substrates (#2261), cleaved PARP (#9541), pTyr (#9411), and GAPDH (#2118; Cell Signaling), pEGFR (Y1068; Abcam, ab40815), and pAKT (Invitrogen, 44621G).

Erlotinib-resistant PC-9 clones were established by exposing parental cells to gradually increasing concentrations of erlotinib for 3 months. Clones capable of proliferating in the presence of drug were isolated and confirmed to be erlotinib resistant. A clone exhibiting the highest level of EGFR T790M expression and no antiproliferative response to 10 micromolar of erlotinib was selected and designated as PC-9/ER.

All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee at Genentech and carried out in a facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care. NCI-H1975 cells were cultured in RPMI-1640 medium containing 10% serum with 2 mmol/L l-glutamine, and 5 × 10⁶ cells were implanted subcutaneously into right flanks of Balb/c nude mice. When tumor sizes reached approximately 200 mm³, mice were randomized into 3 groups of 10 mice each. One group of mice was treated with PKC412 100 mg/kg, as described previously (32), another group was treated with gefitinib 100 mg/kg, and a third group of mice was treated with vehicle alone. All treatments were stopped at day 16 because of severe ulceration across the 3 treatment cohorts due to the unusual required formulation for PKC412 (32). The pharmacodynamic studies for each treatment were conducted with an additional 3 mice after 2 days of treatment. For the PC-9/ER xenograft study, PC-9/ER cells were cultured in RPMI-1640 medium with 10% serum and 2 mmol/L l-glutamine. Harlan athymic nude mice were inoculated subcutaneously into the right flank area with 5 × 10⁶ cells suspended in Hank's balanced salt solution (HBSS)/Matrigel. When tumor sizes reached approximately 200 to 300 mm³, mice were randomized into 4 groups of 7 mice each. Each group of mice was dosed via daily oral gavage with PKC412 100 mg/kg, erlotinib 50 mg/kg, WZ4002 25 mg/kg, or vehicle alone for 21 days. Tumor volumes were determined using digital calipers (Fred V. Fowler Company, Inc.) using the formula (L × W × W)/2. The pharmacodynamics studies were conducted 4 hours after the final treatment.

M. Merchant serves as a consulting scientist for Genentech and has an ownership interest (including patents) in Roche. K. Politi has ownership interest (including patents) as an inventor on a patent owned by Molecular MD. All Genentech authors, with the exception of H.-J. Lee, are employees of Genentech and may be shareholders of Roche Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H.-J. Lee, S. Malek, J. Settleman

Development of methodology: H.-J. Lee, L. Shao, S. Sideris, S. Ubhayakar, J. Settleman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-J. Lee, T.P. Heffron, X. Ye, S. Sideris, S. Malek, E. Chan, M. Merchant, H. La, R.L. Yauch, V. Pirazzoli, K. Politi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-J. Lee, G. Schaefer, S. Sideris, S. Malek, M. Merchant, V. Pirazzoli, K. Politi, J. Settleman

Writing, review, and/or revision of the manuscript: H.-J. Lee, G. Schaefer, T.P. Heffron, S. Malek, M. Merchant, V. Pirazzoli, K. Politi, J. Settleman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Pirazzoli

Study supervision: G. Schaefer, J. Settleman

The authors thank members of the Settleman laboratory for helpful discussions; members of the MGH Cancer Center's Center for Molecular Therapeutics for sensitivity profiling of cancer cell lines; Jose Imperio and the In Vivo Study Group at Genentech for conducting the pharmacokinetic study, and to Erik Shapiro and Dorit Granot from the Yale Magnetic Resonance Resource Center for assistance with mouse imaging.

This study was supported by grants R00CA131488 and R01CA120247 from the National Institutes of Health (to K. Politi). Additional support was received from Uniting Against Lung Cancer (to K. Politi), the Labrecque Foundation (to K. Politi), and the American Italian Cancer Foundation (to V. Pirazzoli).