EGFR mutant lung adenocarcinomas treated first-line with the novel EGFR inhibitor, XL647, can subsequently retain moderate sensitivity to erlotinib

Abstract

Introduction

EGFR mutant lung cancers are sensitive to EGFR tyrosine kinase inhibitors (TKIs). Unfortunately, they develop resistance, often due to acquisition of a second-site mutation (T790M). Current EGFR TKIs select for T790M in preclinical models of acquired resistance. We explored whether all EGFR TKIs similarly select for the T790M mutation using data from early clinical trials and established in vitro models of acquired resistance.

Methods

We analyzed the clinical characteristics of 8 patients with metastatic EGFR mutant lung adenocarcinoma who were treated first-line with XL647 and then progressed. XL647 is an ATP-competitive inhibitor of EGFR, HER2, KDR, and EPHB4. Additional molecular preclinical studies were performed to characterize resistance.

Results

Four patients displayed confirmed partial responses (PRs), three patients had unconfirmed PRs, and one patient displayed stable disease. Only one of five patients’ tumor samples available for analysis after disease progression harbored the T790M mutation. Eight patients subsequently received erlotinib, with (n=3) or without (n=5) chemotherapy. Three of five patients treated with single agent erlotinib derived additional benefit, staying on drug up to 9 months. EGFR mutant PC-9 cells with acquired resistance to XL647 did not harbor the T790M mutation, displayed a distinct mRNA profile from PC-9 cells with T790M-mediated resistance, and were moderately sensitive to erlotinib in growth inhibition assays. Crystal structure analyses of XL647/EGFR T790M did not reveal a different binding mode from that of erlotinib.

Conclusions

The findings of this exploratory study suggest different EGFR TKIs may select for distinct mechanisms of resistance. These results raise the possibility that different EGFR TKIs could be sequentially used to improve outcomes in patients with EGFR mutant lung cancer. Further work investigating this hypothesis is warranted.

INTRODUCTION

Mutations within the gene encoding the epidermal growth factor receptor (EGFR) are found in a subset of non-small cell lung cancer (NSCLC) patients. Particular mutations, such as deletions in exon 19 (19 del) and point mutations in exon 21 (L858R) are associated with high sensitivity to the tyrosine kinase inhibitors (TKIs) gefitinib and erlotinib.^1-3 Unfortunately, acquired resistance to these agents occurs after approximately one year.^{4, 5} The most common mechanism of resistance involves a second site mutation within EGFR, i.e. T790M. This change restores receptor affinity for its natural substrate, ATP, versus drug.^6-8 The T790M mutation is rarely found in TKI-naïve samples but is observed in ~50% of tumors with acquired resistance to erlotinib or gefitinib.^{9, 10}

Multiple “second-generation” inhibitors have been developed to overcome T790M-mediated resistance.¹⁰ Irreversible EGFR inhibitors, such as neratinib (HKI-272), dacomitinib (PF 00299804), and afatinib (BIBW-2992), bind covalently to the receptor and are more potent in vitro against T790M than gefitinib or erlotinib.^11-13 However, continuous exposure of EGFR mutant lung adenocarcinomas cells in vitro to these agents still leads to the resistance mediated by T790M.^{9, 14, 15} In patients, the clinical activity of neratinib and afatinib against T790M-harboring tumors has been disappointing, possibly because of an inability to achieve high enough plasma concentrations to inhibit the double mutant receptor without significant toxicity. ^{16, 17} The activity of dacominitib in this patient population remains to be established.

XL647 is a distinct “second-generation” small molecule inhibitor of EGFR (IC₅₀ 0.3 nM). The compound is a reversible ATP-competitive TKI that also potently inhibits ERBB2 (16 nM), KDR (1.5 nM), and EPHB4 (1.4 nM).¹⁸ Preclinical studies showed efficacy against EGFR-driven tumors, including potentially those harboring T790M.¹⁸ However, in a phase II trial for NSCLC patients that displayed acquired resistance to gefitinib or erlotinib, XL647 induced only a 3% response rate, and patients with confirmed T790M-harboring tumors had worse progression-free survival.¹⁹ Thus, the efficacy of this agent in the setting of acquired resistance was also discouraging.

Concurrently, a separate phase II trial of first-line treatment with XL647 in patients who met defined clinical criteria (i.e. lung adenocarcinoma, never-smokers, females, documented EGFR mutation) was performed. While the radiographic response rate in the overall population was 19.6%, 8 of 14 (57%) patients with EGFR mutant tumors displayed significant tumor shrinkage (confirmed partial (PR) or complete response) (Pietanza et al, in preparation). These clinical data demonstrate that EGFR TKI-naïve tumors with drug-sensitive (19 del and L858R) EGFR mutations are highly sensitive to treatment with XL647. In this exploratory study, we investigated mechanisms of disease progression on XL647 in patients with EGFR mutant tumors and in relevant preclinical models.

MATERIALS AND METHODS

Patients

Trial enrollment was limited to patients with histologically confirmed diagnosis of stage IIIB or IV adenocarcinoma and either a documented EGFR mutation or at least one of the following clinical criteria: Asian, female, minimal smoking history (<15 pack-year smoking history and no smoking within the previous 25 years) or no smoking history. Tumor responses were assessed as per the Response Evaluation Criteria in Solid Tumors (RECIST).²⁰ For complete details of the trial see Pietanza et al., (in preparation). Patients at MSKCC were re-biopsied with consent on an IRB-approved protocol (#04-103). Among the eight total patients in the phase II study with EGFR mutant tumors who displayed confirmed PRs, two were at an outside institution, one remains on study, and one patient came off XL647 after response but before disease progression because of toxicity. These patients were therefore not studied further. Four additional patients on the study with EGFR mutant tumors at MSKCC were studied, including three with unconfirmed PRs and one with stable disease.

Derivation and characterization of TKI-resistance cell lines

Resistant cells were derived using established protocols.⁹ Cells were deemed resistant when they could grow in ~1 μM XL647.

Immunoblotting

Immunoblotting and phospho-receptor tyrosine kinase arrays were performed as per standard protocols.⁹

Molecular characterization of tumor cell DNA and resistant cells

Direct dideoxynucleotide-based sequencing and PCR-RFLPs were performed as per standard protocols.^{7, 9} Additional multiplex mutational profiling was performed as published.²¹ MET FISH was performed as per standard protocols.²² Array comparative genomic hybridization (aCGH) was performed as previously described.²³

For microarray experiments, RNA was extracted from sensitive and resistant cells in the presence and absence of TKI, as previously described.²⁴ Microarray profiling was performed using U133 Plus chips (Affymetrix). Microarray data were normalized using the Robust MultiChip Averaging (RMA) algorithm as implemented in the Bioconductor package Affy.²⁵ The gene expression values were transformed to log2 scale. The moderated t-statistic implemented in the Bioconductor LIMMA package was used to detect differentially expressed probe sets between PC-9/XL-R and PC-9/ER +PC-9/BR. This statistic has the same interpretation as the standard t-statistic, however the standard errors were calculated to shrink towards a common value by the empirical Bayes model to borrow information across all genes.²⁶ To control for false discovery rates, p-values for moderated t-tests were adjusted using the method of Benjamini and Hochberg.²⁷ For heatmap generation, 28,088 probes with absolute beta coefficients greater than 1.5 were selected (in the LIMMA model comparing PC-9/XL-R and PC-9/ER + PC-9/BR groups). For hierarchical clustering, complete linkage and Euclidean distance were used.

Crystal structure analysis

The crystal structures of both the Apo form of the gatekeeper mutant of EGFR (T790M) at 2.7Å resolution and also its complex with XL647 at 2.3Å resolution were solved. Crystals of the Apo form of EGFR-T790M were prepared using protein purified from over 90L of expression media, into which XL647 was soaked.

RESULTS

Characterization of EGFR mutant tumors from patients treated with XL647

We studied in detail the clinical course of eight patients with EGFR mutant lung cancer who were treated at MSKCC with first-line XL647 but then progressed on treatment (Table 1). All patients had tumors harboring either 19 dels or L858R. Five patients were treated on a 5 day on, 9 day off schedule (350 mg po qd), while three were treated with daily dosing (300 mg po qd). Four patients displayed confirmed PRs, three patients had unconfirmed PRs, and one had stable disease. The median time on drug was 7 months (range 4-19 mos). Only one of five (20%) patients’ tumor samples available for analysis after disease progression harbored the T790M mutation; this patient underwent dose reduction from 300 mg po qd to 200 mg po qd due to toxicity and had only an unconfirmed PR. All re-biopsied tumors showed adenocarcinoma histology. Upon pathologic review there was no mention of any spindle like mesenchymal morphologic appearance that would represent an epithelial to mesenchymal transition (EMT; data not shown), or features of small cell carcinoma, ruling out histological changes as a mechanism of resistance.²⁸ One of these samples underwent additional interrogation using massively-parallel (454) sequencing (limit of detection 0.2%⁹); the T790M mutant allele was still not detected (data not shown). Tissue from only one patient was amenable to FISH analysis for amplification of MET, a rarer mechanism of acquired resistance;^{23, 24, 28, 29} the sample was negative.

Table 1. Characteristics of 8 patients at MSKCC with drug-sensitive EGFR mutant lung adenocarcinomas treated with XL647.

d- days; PR – partial response; uPR – unconfirmed PR; SD – stable disease; mos – months; L – left; R – right; 1ry – primary; 2ry – secondary; mutn – mutation; amp – amplification; Y – yes; N – no; chemo – chemotherapy. Patient 4 lacked a confirmatory scan due to progression of disease. Patients 7 and 8 had PR at their first post-XL647 scans; however, they were dose reduced secondary to toxicity to 200mg daily due to grade 3 rash and diarrhea, respectively. Their subsequent scans failed to confirm PR. Post-XL647 biopsy sites (*) were from lesions that were involved at baseline, improved or stabilized with treatment, and then progressed; only patient 7 had a new liver metastasis (**) that was biopsied on progression, although had had liver involvement prior to starting XL647.

ID	Treatment with XL647			Biopsies			Genotype			Treatment with erlotinib
	Dosing schedule	Best response	Time on drug (mos)		Site*	Histology	1ry mutn	2ry mutn	MET amp	Clinical benefit from erlotinib	Single agent or with chemotherapy	Time on drug (mos)
1	5d on, 9d off	PR	19	Pre- Post-	R. pleura n/a	Adeno n/a	19 del	n/a	n/a	N	Single agent	2
2	5d on, 9d off	PR	9	Pre- Post-	L. lung mass L. pleura	Adeno Adeno	19 del	None	No	Y	Single agent	9
3	5d on, 9d off	PR	6	Pre- Post-	L2 vertebra L. 6th rib	Adeno Adeno	19 del	None	n/a	Y	With chemo	7
4	5d on, 9d off	uPR	5	Pre- Post-	R. pleural fluid n/a	Adeno n/a	19 del	n/a	n/a	Y	Single agent	5
5	5d on, 9d off	SD	4	Pre- Post-	L. lung mass L. pleural fluid	Adeno Adeno	L858R	None	n/a	Y	With chemo	11
6	Daily	PR	14	Pre- Post-	L. pleura n/a	Adeno n/a	L858R	n/a	n/a	Y	Single agent	4
7	Daily	uPR	6	Pre- Post-	R. lung mass Liver lesion**	Adeno Carcinoma	19 del	T790M	n/a	N	Single agent	2
8	Daily	uPR	8	Pre- Post	R. lung mass R. lung mass	Adeno Adeno	L858R	None	n/a	Y	With chemo	22

At the discretion of the treating physicians, the eight patients who came off study were subsequently immediately treated with erlotinib, either as a single-agent (n=5) or in combination with chemotherapy (n=3). Interestingly, three of five patients treated with single-agent erlotinib derived additional long-term benefit from the EGFR TKI, staying on drug from 4 to 9 months (Table 1; Figure 1A, 1B). One patient underwent tumor biopsy at the time of progression after XL647, and the T790M mutation was not detected. Two of these patients did not have tumor available for analysis after disease progression, but their clinical courses suggested that their disease lacked T790M. The patient whose disease harbored the T790M mutation progressed after 2 months on subsequent erlotinib treatment (Figure 1C). The three patients who received chemotherapy with erlotinib stayed on erlotinib for 7, 11, and 22 months, respectively.

Figure 1. Representative case histories.

A. Patient 2 was a 63 year old white male former light smoker whose disease harbored an EGFR 19 del and displayed a partial response (PR) to XL647 (computed tomography (CT) scans of the chest shown in top left and middle panels). After 9 months, the patient came off study following growth of a left lower lobe lung mass (arrowed) and development of a left pleural effusion (top right panel). The patient underwent a video-assisted thorascopic surgery (VATS) pleurodesis and was switched to single agent erlotinib (bottom panels). The ground-glass opacities are likely due to the pleurodesis. There was an 18% decrease in all lesions (by RECIST criteria) at the patient’s nadir; the left lower lobe itself decreased by 24%. B. Patient 4 was a 68 white female former light smoker whose disease harbored a 19 del and displayed an unconfirmed PR to XL647 (left panels, CT scans). Therapy was discontinued after 5 months when the patient developed leptomeningeal disease (right upper panels; axial T1 post magnetic resonance images). Tumor tissue from this patient was not available at the time of developing resistance. However, following a switch to erlotinib single agent therapy, the leptomeningeal disease resolved (right upper panels). Concurrently, the patient’s right pleural effusion decreased (right lower panels; chest radiographs). C. Patient 7 was a 62 year old Filipino female never smoker whose tumor harbored a 19 del. She exhibited an unconfirmed PR on XL647 (left and middle panels; CT scans) but disease progressed in the lung and liver after ~5 months (right panels). Biopsy of the growing liver lesion revealed presence of the EGFR T790M secondary mutation. Consistent with the presence of this mutation, the patient did not derive benefit from subsequent single-agent erlotinib therapy.

Derivation of EGFR mutant cells with acquired resistance to XL647

These intriguing clinical histories prompted us to examine properties of XL647 in preclinical models. Compared to erlotinib, XL647 (Figure 2A) displayed similar growth inhibitory effects against TKI-sensitive PC-9 lung adenocarcinoma cells, which harbor an exon 19 del, and H1975 cells, which harbor an EGFR L858R TKI-sensitive mutation in addition to a T790M TKI-resistant mutation. XL647 was more potent than erlotinib at higher concentrations of drug against H1975 cells (Figure 2B).

Figure 2. Activity of XL647 against EGFR mutant lung cancer cell lines.

A.Chemical structures of erlotinib and XL647. B. Growth inhibition assays comparing the activity of XL647 with erlotinib against PC-9 (19 del) and H1975 (L858R/T790M) cells.

We next used a well-established protocol to develop PC-9 cells with acquired resistance to XL647. We previously reported the development of PC-9 cell lines with acquired resistance to erlotinib or afatinib that were developed concurrently; both were found to harbor the T790M mutation.⁹ After about 160 days of selection, PC-9 cells with acquired resistance to XL647 (PC-9/XL-R cells) were able to proliferate in concentrations 20x the IC₅₀ of the parental cells (Figure 3A). Analysis of protein lysates from PC-9/XL-R cells demonstrated that phosphorylation of EGFR was unaffected by addition of low or high concentrations of XL647 (Figure 3B). Direct sequencing of all the coding exons (2 to 28) of EGFR revealed only the primary 19 del in DNA from PC-9/XL-R cells (Figure 3C). More sensitive SNaPShot analysis also did not detect the T790M mutation (limit of detection ~5%²¹; data not shown). Array comparative genomic hybridization (aCGH) of DNA from PC-9/XL-R cells did not show amplification of the MET kinase (Figure 3D). Assessment of the phosphorylation state of 42 receptor tyrosine kinases (RTKs) in the PC-9/XL-R cells in the presence of inhibitor using phospho-RTK arrays revealed no obvious changes compared to parental cells (Figure 3E). The exact mechanism of resistance of PC-9/XL-R cells remains to be elucidated.

Figure 3. Derivation of PC-9 cells with acquired resistance to XL647.

A. PC-9/XL-R cells grew in 20x the IC₅₀ of parental PC-9 cells in growth inhibition assays. B. Immunoblotting studies show decreased phosphorylation of EGFR in the presence of XL647 in parental but not resistant cells. C. Direct dideoxynucleotide sequencing chromatograms from EGFR exons 19 (ex 19) and 20 (ex20) show presence of the 19 del but no T790M mutation in DNA from PC-9/XL-R cells. D. Array comparative genomic hybridization shows no amplification of MET on chromosome 7 in PC-9/XL-R cell line DNA. Y-axis: copy number gain/loss. E. Phospho-receptor tyrosine kinase (RTK) arrays on lysates from PC-9 and PC-9/XL-R cells in the absence or presence of drug reveal no obvious differences in phospho-RTK status (1: pEGFR; 2: pHER3; 3: pFGFR3; 4: pIR; 5: pIGF1-R; 6: pDTK (TYRO3)).

PC-9/XL-R cells remain moderately sensitive to erlotinib

To investigate the characteristics of the PC-9/XL-R cells, we compared the mRNA expression profiles of PC-9/XL-R versus corresponding erlotinib-(PC-9/ER) and afatinib (BIBW-2992)-resistant (PC-9/BR) cells. Using supervised clustering algorithms, PC-9/XL-R cells clustered completely separately from both sets of T790M-containing cells, regardless of the absence or presence of drug (Figure 4A). A list of differentially expressed genes is available as Supplemental Digital Content. These data further support that PC-9/XL-R cells are distinct from T790M-harboring PC-9/ER and PC-9/BR cells.

Figure 4. PC-9/XL-R cells are distinct from T790M-harboring PC-9 cells.

A. Heatmap generated via supervised clustering shows that mRNA expression profiles from PC-9 cells with acquired resistance to XL647 (PC-9/XL-R) are distinct from T790M-harboring PC-9 cells developed via long-term culture in erlotinib (PC-9/ER) or afatinib (BIBW-2992) (PC-9/BR). mRNA profiles were obtained from cells grown in the absence or presence of inhibitor. B. PC-9/XL-R cells retain moderate sensitivity to erlotinib, compared to PC-9 cells harboring the T790M mutation. C. Immunoblotting studies show decreased phosphorylation of EGFR and ERK at major signaling residues (tyrosine (Y) 1068 in EGFR, and threonine (T) 202/tyrosine (Y) 204 in ERK) in the presence of erlotinib in parental and PC-9/XL-R cells but not in PC-9 cells harboring the T790M mutation.

We next examined the sensitivity of PC-9/XL-R cells to erlotinib. Similar to the observations in patients, these cells remained intermediately sensitive to erlotinib (Figure 4B). By contrast, PC-9 cells harboring the T790M mutation were resistant to erlotinib (Figure 4B). Biochemical analysis of PC-9/XL-R cells corroborated the growth inhibitory data and showed decreased phosphorylation of EGFR and its downstream target, ERK, following treatment with high and low doses of erlotinib (Figure 4C).

Selection for T790M-harboring clones in mixed cell populations may differ between XL647 and erlotinib

To determine if XL647 could select for cells harboring T790M mutations in mixed cell populations, we mixed sensitive PC-9 cells (without T790M) with PC/BR clonal cells harboring the T790M mutation at known.^{9, 14} Each population was then treated with DMSO vehicle, erlotinib, or XL647 for 7 days (Figure 5). As previously reported, vehicle-treated cells harbored almost non-detectable levels of the T790M allele at the end of the treatment cycle, likely due to growth of the faster growing non-T790M-harboring sensitive cells.^{9, 10} Treatment with erlotinib selected for high levels of the T790M allele. By contrast, the T790M allele peak height was not as high in cells treated with XL647. These data are consistent with the notion that XL647 may not be able to overcome T790M-mediated resistance but that the selective pressure exerted by the drug may be different than that of erlotinib.

Figure 5. Effect of TKI treatment on the number of T790M-harboring cells in mixed cell populations.

(A) Using a T790M-harboring clone (PC-9/BR, c1⁹), cell populations were mixed to have 0, 25 or 100% resistant cells. The baseline panel shows forward sequence tracings from exon 20 of EGFR (the underlined codon encodes T790). Cell populations were then treated with DMSO, XL647, or erlotinib at the indicated doses for 7 days. (B) Chromatograms display exon 20 forward sequences of EGFR from PC-9 cells treated with different drugs (*ACG→ATG).

Crystal structure of XL647 within EGFR

To gain additional insights into XL647, the crystal structures of both the Apo form of EGFR T790M at 2.7Å resolution and also its complex with XL647 at 2.3Å resolution were solved. The binding mode of XL647 to EGFR-T790M was similar to other agents with quinazoline cores (Figure 6A). Comparison of the binding mode of XL647 in EGFR-T790M versus XL647 in EGFR-wildtype showed that the substitution of T790 for the larger methionine residue causes a slight shift of the aniline ring “forward” and away from M790 (Figure 6B).

Figure 6. Crystal structure modeling of XL647 in the EGFR ATP binding pocket.

A. The binding mode of XL647 to EGFR-T790M is largely as anticipated: the quinazoline core forms a single hydrogen bond attachment to the protein kinase “linker”, the solubilizing group extends toward solvent, and the tri-halo aniline sits in the “aniline pocket” between the gatekeeper mutant methionine (T790M) and the salt bridge. B. Comparison of the binding mode of XL647 in EGFR-T790M (cyan carbons) versus XL647 in EGFR-wildtype (green carbons). Note that the substitution of T790 for the larger methionine residue causes a slight shift of the aniline ring “forward” and away from M790. The rotation of the solubilizing group by ~180° appears to be due to the presence of a new crystal lattice packing contact not seen in the wild-type structure.

DISCUSSION

T790M-mediated resistance remains a major clinical problem in NSCLC patients whose tumors harbor activating mutations within EGFR. More potent “second-generation” irreversible EGFR TKIs such as neratinib and afatinib have been tested in early phase trials, but results have been disappointing. Consistent with these clinical results, preclinical models have shown that neratinib and afatinib still select for T790M-mediated resistance in vitro, ^{9, 14} perhaps because they still inhibit the drug-sensitive EGFR mutants at lower doses than the T790M mutant. Analogously, the second-generation EGFR TKI, XL647, showed minimal activity in a trial for patients who progressed on gefitinib or erlotinib.¹⁹ These studies collectively indicate that once patients’ tumors have acquired the T790M mutation, single agent EGFR TKIs with greater sensitivity against drug-sensitive versus -resistant mutants will not be clinically effective. In this setting, a more promising approach may be dual targeting of EGFR with an irreversible inhibitor plus an anti-EGFR antibody, a combination which recently showed a >40% response rate in an phase IB/II trial.³⁰ Alternatively, small molecule inhibitors with selectivity against the T790M mutant appear promising in preclinical models.³¹

Here, we examined resistance that develops to XL647 in EGFR TKI-naïve EGFR mutant lung cancer. From analogous studies involving gefitinib or erlotinib, we have shown that >50% of patients with acquired resistance harbor T790M.²² Surprisingly, for XL647, only one of five patients’ tumor samples available for analysis harbored the T790M mutation. Two patients who did not have tumor available for analysis also benefited from erlotinib, suggesting that their disease also lacked T790M-mediated resistance. Consistent with these clinical findings, PC-9 cells with acquired resistance to XL647 did not harbor the T790M mutation, displayed a distinct mRNA profile from PC-9 cells with T790M-mediated resistance, and were moderately sensitive to erlotinib in growth inhibition assays. These studies suggest that it may be possible to find EGFR TKIs that do not necessarily select for T790M-mediated resistance, even if they have greater potency against drug-sensitive mutants.

This study has some limitations. First, only eight patients were studied, only 5 had tumor tissue acquired after resistance developed, and some patients had only unconfirmed PRs. Although only one of these five samples harbored the T790M mutation, the numbers were too small to make meaningful comparisons against those who developed acquired resistance to gefitinib or erlotinib. In two cases without available tissue at the time of progression (patients 4 and 6, Table 1), we cannot rule out that these tumors harbored T790M and just progressed slowly while on erlotinib therapy for 4 and 5 months, respectively.²² Second, patients were treated with XL647 on different dosing schedules. We cannot exclude that patients did not acquire the T790M mutation because of related pharmacokinetic parameters; for example, we have shown that different dosing schedules can clearly influence whether cells acquire T790M in vitro or not.⁹ However, in modeling acquired resistance, we treated PC-9 cells with erlotinib, afatinib, and XL647 in parallel, and only XL647 treated cells did not develop T790M-mediated resistance. This result suggests that dosing is not a critical issue. Third, three of the patients studied received erlotinib plus chemotherapy after progression on XL647. Thus, we were unable to assess whether patients subsequently responded to single agent erlotinib. Finally, we do not yet know how XL647 induces resistance. We have excluded known mechanisms, including second-site mutations, MET amplification, or a change in tumor morphology (data not shown). Crystal structure analyses also were unrevealing, as XL647 seems to bind in the EGFR ATP-binding pocket in a manner similar to erlotinib. Whether concurrent inhibition of other kinases like ERBB2, KDR, and EPHB4 affects the development of T790M-mediated resistance remains to be determined. However, afatinib, which also inhibits ERBB2, and vandetanib, which inhibits EGFR and VEGFR, both select for T790M-mediated resistance in vitro,^{9, 32} suggesting that inhibition of these kinases may not be relevant.

Despite these caveats, we present these clinical and pre-clinical data because they raise the notion that different EGFR TKIs may select for distinct mechanisms of resistance. Thus, it may be possible to use non-cross-resistant EGFR TKIs sequentially to improve outcomes in patients with EGFR mutant lung cancer.