An Acquired HER2 ^T798I Gatekeeper Mutation Induces Resistance to Neratinib in a Patient with HER2 Mutant–Driven Breast Cancer

DNA-sequencing efforts have revealed that ERBB2, the gene encoding the HER2 receptor tyrosine kinase, is mutated in a wide variety of cancer types, including 2% to 3% of primary breast cancers (1–3), with a higher incidence in lobular breast cancers (4). More than 70% of HER2 mutations in breast cancer are found in the absence of HER2 (ERBB2) gene amplification (2). Some of the common HER2 mutations promote HER2 kinase activity and transform breast epithelial cells and other cell types (5–9). Given that irreversible EGFR/HER2 tyrosine kinase inhibitors (TKI), such as neratinib and afatinib, have shown preclinical activity against several HER2 mutants (5, 7–9), clinical trials with neratinib (SUMMIT trial; NCT01953926) and afatinib (NCI-MATCH; NCT02465060) focused on patients with HER2-mutant cancers are in progress. However, sustained clinical activity of ATP mimetics in patients with advanced cancer has generally been limited by the acquisition of drug resistance. Mutation of the “gatekeeper” residue within the kinase's ATP-binding pocket, such as ABL^T315I, KIT^T670I, and EGFR^T790M, is a common mechanism of acquired resistance. Here, we report for the first time a case of a HER2 gatekeeper mutation in a patient with nonamplified HER2-mutant breast cancer with acquired resistance to neratinib.

Targeted capture next-generation sequencing (NGS; ref. 10) of DNA from a skin metastasis in a 54-year-old female with estrogen receptor (ER)/progesterone receptor (PR)–positive, HER2 nonamplified lobular breast carcinoma identified an ERBB2^L869R (HER2^L869R) somatic mutation (Supplementary Table S1). Prior therapies included chemotherapy, tamoxifen, aromatase inhibitors, everolimus, and trastuzumab. The tumor also harbored a truncation mutation in CDH1, ERBB3^E928G, and amplification of CCND1 and FGF3/4/19. Interrogation of the cBioPortal (n > 21,000), Project GENIE (n > 18,000), the Catalogue of Somatic Mutations in Cancer (COSMIC; n > 50,000), Foundation Medicine (n > 40,000), and Guardant Health (n > 17,000) databases found 16 additional cancers harboring ERBB2^L869R and one L869Q mutation, including 12 breast cancers (Supplementary Table S2). In addition, a recent study reported four instances of ERBB2^L869R among 413 invasive lobular breast cancers (4).

The L869R mutation is located within the activation loop of the HER2 kinase domain. Sequence alignment of the HER2, EGFR, and BRAF kinase domains showed that HER2^L869R is homologous to BRAF^V600E, a gain-of-function mutation found in >50% of melanomas (11), and EGFR^L861R/Q, an activating mutation in non–small cell lung cancer (NSCLC; Fig. 1A; ref. 12). We performed structural modeling of the L869R mutation using Rosetta (13) and examined the residue pair energies involving L869. The mutation resulted in the addition of a strong attractive interaction between R869 and D769 (Fig. 1B and C). This interaction potentially stabilizes the active conformation of the C helix. We also predict that mutating L869 to a polar residue (Arg) disrupts the autoinhibitory contacts between the C helix and the activation loop helix, resulting in destabilization of the inactive conformation of the kinase, similar to EGFR^L858R (14).

On the basis of these structural data, we hypothesized that HER2^L869R would display increased signaling and transforming capacity. To test this, we stably transduced MCF10A breast epithelial cells with lentiviral vectors encoding HER2 wild-type (WT) or HER2^L869R. Cells expressing HER2^L869R exhibited increased phosphorylation of AKT, ERK, and S6, which were blocked by neratinib (Fig. 1D). Phosphorylation of HER2^WT, but not HER2^L869R, was blocked by the reversible HER2/EGFR TKI lapatinib, whereas neratinib inhibited phosphorylation of both WT and mutant receptors. Expression of HER2^L869R enhanced MCF10A cell proliferation in growth factor–depleted media (Fig. 1E) and colony formation in three-dimensional (3-D) Matrigel in the absence of EGF and insulin (Fig. 1F) compared with MCF10A/HER2^WT cells. Growth of MCF10A/HER2^L869R cells was inhibited by neratinib but not by lapatinib, whereas the HER2^WT cells were sensitive to both TKIs. With these supporting data, the patient was enrolled in a clinical trial with single-agent neratinib (NCT01953926). Upon treatment, the patient exhibited an excellent clinical response, showing near disappearance of multiple skin metastases after 20 days (Fig. 1G), and a 77% reduction in marker lesions by RECIST criteria after 8 weeks.

Because the patient harbored a co-occurring ERBB3^E928G mutation, a known activating mutation in HER3 (15), we next asked whether HER2^L869R and HER3^E928G might co-operate to drive HER2 signaling. Mutations in ERBB2 and ERBB3 often co-occur in cancer. In the American Association for Cancer Research (AACR) Project GENIE dataset (>18,000 sequenced tumors), 8.3% of ERBB2-mutated cancers also harbor mutations in ERBB3, whereas only 2.3% of ERBB2 WT cancers have ERBB3 mutations (q value = 1.3 × 10⁻¹⁰; www.cbioportal.org/genie). ERBB2^L869R and ERBB3^E928G were found to co-occur in another breast cancer case in the METABRIC dataset (16). Structural modeling of the HER2^L869R/HER3^E928G double mutant predicted that the HER3 mutation, located at the dimer interface, may enhance heterodimerization of the kinase domains through decreased bulk and electrostatic repulsion (Supplementary Fig. S1A). Calculating the change in free energy of WT heterodimers compared with mutant heterodimers demonstrated a significant difference in the capacity of the latter to bind to one another (Supplementary Fig. S1B). Furthermore, coexpression of the HER2^L869R and HER3^E928G intracellular domains resulted in enhanced transphosphorylation of HER3 and ERK as substrates compared with that induced by expression of either mutant alone (Supplementary Fig. S1C). Phosphorylation of mutant HER2 and HER3, as well as the elevated downstream signaling induced by the expression of both mutants, was blocked by treatment with neratinib (Supplementary Fig. S1D). These data suggest that these co-occurring mutations in ERBB2 and ERBB3 enhance ERBB signaling output, which, in turn, can be blocked by neratinib.

After 5 months on therapy, the patient developed a painful metastasis in the sternum. The addition of the ER antagonist fulvestrant to neratinib induced a prompt symptomatic and clinical response. After 10 additional months on the combination, the patient progressed with new skin metastases. Targeted tissue-based NGS analysis of DNA from a new skin metastasis and plasma tumor, cell-free DNA (cfDNA; Guardant360) revealed that ERBB2^L869R remained (44% allele frequency and 8.7% cfDNA, respectively; Supplementary Table S3). ERBB3^E928G remained in the post-treatment biopsy as well. ERBB2^T798I was found in plasma (1.3% cfDNA), but not in DNA from the synchronous skin metastasis. Additional single-gene deep sequencing of plasma ERBB2 using two rounds of targeted capture (average >4,000 reads per sample) in an independent plasma sample from that used for the Guardant360 test failed to identify ERBB2^T798I in any of the plasma samples obtained at study enrollment or during the first 9 cycles of neratinib, but increased to 1.0% of reads at the time of clinical progression (Fig. 2A; Supplementary Table S4). In contrast, ERBB2^L869R was detected in 6.8% of reads in the pretreatment sample, decreasing considerably during therapy, and rebounding up to 15.2% at progression. These data suggest that ERBB2^T798I was acquired during neratinib therapy.

HER2^T798I is homologous to EGFR^T790M and imatinib-resistant KIT^T670I in gastrointestinal stromal tumors (Fig. 2B). EGFR^T790M drives resistance to first- and second-generation EGFR TKIs in NSCLC by two mechanisms: first, by mediating steric hindrance of ATP-competitive drugs, and second, by increasing the affinity of ATP, resulting in enhanced phosphotransfer and kinase activity (17). To determine whether HER2^T798I functions in a similar manner, we constructed computational models of HER2^WT and HER2^T798I bound to neratinib. We found that the increased bulk of the isoleucine at position 798 would result in steric hindrance when neratinib binds (Fig. 2C). The closest approach between nonhydrogen atoms from residue T798 to neratinib is 4.1 Å in HER2^WT, whereas this distance is reduced to 3.6 Å in HER2^T798I, resulting in a reduced size of the binding pocket. Therefore, the isoleucine substitution at position 798 is expected to reduce neratinib binding.

Next, we asked whether the T798I mutation would block neratinib action. HEK293 cells transfected with HER2^WT, HER2^L869R, HER2^T798I, or HER2^L869R/T798I (both mutations in cis) were treated with increasing doses of neratinib for 4 hours. Low doses of neratinib (20 nmol/L) blocked pHER2, pAKT, and pERK in cells expressing HER2^WT or HER2^L869R, but not in cells expressing HER2^T798I or HER2^L869R/T798I treated with up to 180 nmol/L neratinib (Fig. 2D). To confirm these findings, we stably transduced MCF10A cells with WT and mutant HER2. We noted that HER2^T798I and HER2^T798I/L869R were poorly expressed in HEK293 and MCF10A cells (Fig. 2D and E). Treatment with the proteasome inhibitor MG132 for 24 hours restored expression of the T798I mutants (Supplementary Fig. S2), suggesting that this mutation decreases protein stability. Cells expressing HER2^L869R or HER2^L869R/T798I, but not HER2^T798I alone, displayed enhanced pAKT, pERK, and pS6 (Fig. 2E). Furthermore, although HER2^L869R and HER2^L869R/T798I induced EGF-independent MCF10A cell proliferation, HER2^T798I did not (Fig. 2F). Although untreated MCF10A/HER2^L869R/T798I cells did not proliferate as fast as cells expressing HER2^L869R, they were the only cells that grew in the presence of neratinib. A similar slow growth rate has been reported in EGFR TKI–resistant cell lines and patients' tumors harboring EGFR^T790M (18). MCF10A cells expressing both mutations displayed reduced sensitivity to neratinib (IC₅₀ = 154 nmol/L) compared with cells expressing HER2^L869R (IC₅₀ = 23.9 nmol/L; Fig. 2G). These results suggest that, like EGFR^T790M for gefitinib and erlotinib, HER2^T798I confers a growth advantage in the presence of neratinib.

The lack of transforming capacity of HER2^T798I alone suggests that it is not a driver oncogene, but an acquired alteration as a result of therapeutic pressure. Consistent with this speculation, HER2^T798I is exceedingly rare in tumors from patients not treated with HER2 TKIs (Supplementary Table S2). Of all of the tumors sequenced in the cBioPortal, COSMIC, Foundation Medicine, and Guardant Health databases (more than 100,000 samples sequenced in all), HER2^T798I was found in only one colorectal cancer cell line (Foundation Medicine) and one endometrial cancer cell line (Cancer Cell Line Encyclopedia), strongly suggesting that in the patient reported herein, T798I was acquired due to selective pressure of neratinib treatment.

We next examined a panel of other irreversible EGFR/HER2 TKIs for their ability to block HER2^L869R/T798I. These included afatinib, a covalent EGFR/HER2 inhibitor, the EGFR inhibitor osimertinib (AZD9291), which exhibits selectivity against mutant EGFR (including T790M) but does not block WT HER2 (19), and AZ5104, an osimertinib metabolite that inhibits WT HER2 and EGFR (20). We performed computational modeling of HER2^L869R/T798I bound to neratinib, afatinib, osimertinib, and AZ5104 (Fig. 3A–D). These small molecules are expected to bind HER2 using the same mechanism and position by which they bind EGFR. By analogy with EGFR, the HER2 kinase is predicted to adopt distinct conformations when bound by each inhibitor. Afatinib and neratinib have covalent binding modes that project deeply into the substrate binding pocket of the HER2 kinase (Fig. 3A and B). The sterically larger side chain of HER2^T798I decreases the available space and decreases the polar character of the binding pocket. This is predicted to affect neratinib binding, which, by being the largest of these small molecules, extends the deepest into the pocket. Although afatinib is predicted to make slight contact with T798I, it does not insert as far into the tunnel as neratinib does. Osimertinib and AZ5104 are predicted to bind much less deeply on the lip of the pocket (Fig. 3C and D). On the basis of these studies, HER2^T798I is predicted to disrupt neratinib binding, but is not expected to significantly affect the binding of afatinib, osimertinib, or AZ5104.

We next tested the ability of the panel of inhibitors to block mutant HER2 in stably transduced NR6 mouse fibroblasts, which lack endogenous EGFR (21), and MCF10A cells. In both cell types, neratinib more efficiently blocked HER2 phosphorylation in cells expressing HER2^WT or HER2^L869R compared with cells expressing HER2^L869R/T798I (Fig. 3E and F). Treatment with afatinib and AZ5104 blocked phosphorylation of HER2^WT as well as both HER2 mutants. In contrast, osimertinib failed to inhibit HER2^WT, HER2^L869R, or HER2^T798I. Inhibition of pAKT, pERK, and pS6 with all small molecules mirrored that of pHER2 in MCF10A cells (Fig. 3F).

MCF10A/HER2^L869R and MCF10A/HER2^L869R/T798I were highly sensitive to afatinib and AZ5104 in growth factor–depleted media, whereas higher doses of osimertinib were required to block the growth of both cell types (Fig. 4A). Neratinib and AZ5104 showed similar IC₅₀ values in HER2^L869R-expressing cells, whereas neratinib was less effective against HER2^L869R/T798I-expressing cells. In 3-D Matrigel, 100 nmol/L of neratinib, afatinib, or AZ5104 completely blocked acini formation by MCF10A/HER2^L869R cells, whereas 100 nmol/L of osimertinib only slightly suppressed acini growth (Fig. 4B). Both neratinib and osimertinib failed to suppress growth of MCF10A/HER2^L869R/T798I cells in 3-D Matrigel, whereas this was completely blocked by afatinib and AZ5104, suggesting that the latter two inhibitors are able to overcome HER2^T798I-mediated drug resistance.

Recent reports have proposed the acquisition of HER2 mutations in patients with HER2^WT amplification treated with anti-HER2 therapies (22). In addition, neratinib has shown clinical activity and is being used in patients with HER2^WT amplification (23). Thus, we tested whether a HER2 gatekeeper mutation would confer resistance to neratinib when present in a background of HER2^WT amplification. We used HER2-amplified BT474 cells stably expressing HER2^T798M, which we previously reported to be lapatinib resistant (24). BT474^GFP control cells and BT474/HER2^T798M cells were treated with vehicle (DMSO), lapatinib, neratinib, afatinib, osimertinib, or AZ5104. Lapatinib failed to suppress pHER2, pAKT, pERK, and pS6 in HER2^T798M-expressing cells (Fig. 4C). Treatment with neratinib inhibited pHER2, pAKT, and pS6 in BT474^GFP cells but not in BT474/HER2^T798M cells. Consistent with the findings in MCF10A cells, afatinib and AZ5104, but not osimertinib, blocked pAKT, pERK, and pS6 in both BT474^GFP and BT474/HER2^T798M cells. As only approximately 3% of the ERBB2 alleles in the BT474/HER2^T798M cells harbor the mutation (24), these data suggest that just a few HER2^T798M alleles can confer resistance to neratinib, but not afatinib, in cells with HER2^WT gene amplification.

We report herein the identification of a HER2^T798I gatekeeper mutation in a patient with HER2-mutant, nonamplified breast cancer with acquired resistance to neratinib. Structural modeling showed that the T798I mutation results in a steric clash with neratinib, which would reduce drug binding. HER2^T798I directly promoted resistance to neratinib in lentivirally transduced cell lines. In contrast to neratinib, afatinib and the metabolite of osimertinib, AZ5104, blocked HER2^T798I-induced signaling and cell growth.

Although the initial neratinib-sensitizing HER2^L869R mutation induced constitutive phosphorylation of AKT, ERK, and S6 and displayed gain-of-function activity when expressed in breast epithelial cells (Fig. 1), we failed to observe increased phosphorylation of this mutant compared with HER2^WT (Figs. 1D and 2D and E). We speculate that the L869R mutation likely removes autoinhibitory interactions, thus placing the kinase in a better position to interact with other ERBB receptors and adaptor proteins/downstream substrates (25, 26). Notably, the HER2-mutant cancer also harbored a known activating HER3^E928G mutation (15). We speculate these comutations resulted in increased dependence on the ERBB pathway and contributed to the tumor's initial sensitivity to neratinib. Consistent with this speculation, preliminary results from the SUMMIT trial show that among 17 patients who exhibited clinical benefit from neratinib, 2 patients harbored ERBB3 missense mutations, whereas none of the 25 patients who did not benefit harbored ERBB3 alterations in their cancer (27).

Our findings parallel the identification of the EGFR^T790M gatekeeper mutation in NSCLC resistant to EGFR inhibitors. We note that for EGFR, two nucleotides would need to be mutated to change the threonine codon at position 790 to an isoleucine [ACG (Thr) > ATA, ATC, or ATT (Ile)], whereas only one nucleotide change is needed for the T790M mutation (ACG > ATG). The opposite is true for ERBB2 [ACA (Thr) > ATA (Ile) vs. ACA > ATG (Met)]. Thus, it is easier for the tumor to mutate ERBB2 codon 798 to an isoleucine rather than a methionine.

EGFR^T790M is reported to promote resistance by simultaneously increasing ATP affinity and decreasing drug binding (28). Although our data suggest that the HER2^T798I mutation could affect neratinib binding through steric interactions, it could similarly affect ATP binding and kinase activity. Although the change in distance (0.5 Å) from residue 798 to neratinib could theoretically be accommodated by conformational changes, the structural evidence suggests that replacing a polar amino acid (Thr) with a hydrophobic residue (Ile) would decrease ATP affinity. The WT Thr side chain contains an -OH group that faces the ATP-binding site. In the AMP-PNP–bound crystal structure of EGFR (2GS7.pdb), that -OH group is within 3.4 Å of the N6 of AMP-PNP. Replacing the Thr with Ile would remove that favorable interaction and is expected to decrease ATP affinity. These structural assessments are consistent with our cell-based findings that T798I-expressing cells do not show increased HER2 phosphorylation, even when corrected for expression levels (Fig. 2D and E).

HER2^T798I and EGFR^T790M also differ in that the former is exceedingly rare in untreated tumors (Supplementary Table S2), whereas EGFR^T790M also occurs in germline DNA and can promote lung cancer formation (29), suggesting that EGFR^T790M itself is oncogenic. This is also consistent with the notion that HER2^T798I alone is not oncogenic, but requires another activating mutation in cis (e.g., L869R) to promote HER2 signaling and oncogenic growth (Fig. 2).

We previously reported that a HER2^T798M gatekeeper mutation increased HER2 autophosphorylation and association of HER3 with the p85-regulatory subunit of PI3K (24). In the current study, HER2^T798I alone did not appear to enhance HER2 signaling or HER2-induced proliferation more than HER2^WT (Fig. 2E and F). This discrepancy may be due to differences in experimental conditions (i.e., serum starvation), differences between the Met and Ile residues, or lower expression of the mutant receptor compared with WT, which we observed in multiple cell lines expressing T798I alone or in cis with L869R (Figs. 2E and 3E). We speculate that the decreased expression of the mutant may be due to decreased protein stability (Supplementary Fig. S2). Despite this decreased expression, MCF10A cells expressing HER2^T798I/L869R displayed increased phosphorylation of HER2 signaling targets and EGF-independent proliferation compared with MCF10A/HER2^WT cells, as well as robust growth in the presence of neratinib (Figs. 2E–G and 4B), altogether suggesting that even low levels of HER2^T798I can promote neratinib resistance.

We are unable to determine whether ERBB2^L869R and ERBB2^T798I occur in cis in the patient's plasma, as these two mutations are 213 bp apart, longer than the length of cfDNA fragments shed from tumor cells. In NSCLC, EGFR^T790M is usually found on the same allele as the initial TKI-sensitizing EGFR mutation (30), suggesting that the two ERBB2 mutations may also occur in cis. In addition, the allele frequency of ERBB2^T798I in plasma tumor cfDNA in the patient progressing on neratinib was lower than the frequency of ERBB2^L869R (Fig. 2A; Supplementary Table S3) consistent with HER2^L869R being the initial driver mutation, and HER2^T798I representing an acquired subclonal drug-resistant mutation. A similar relationship is typically seen with somatic EGFR^T790M in the plasma of patients progressing on EGFR inhibitors compared with the level of the original drug-sensitive EGFR mutation (31). We also note that HER2^T798I was not found in a new skin metastasis synchronous with the progression on neratinib, suggesting spatially heterogeneous mechanisms of drug resistance. This finding is consistent with other reports where plasma may serve as a repository of different acquired drug-resistant mutations found in some but not all metastatic sites, whereas a tissue biopsy of a single lesion may produce a less complete picture, as suggested by studies with drug-resistant NSCLC expressing EGFR^T790M. For example, a subset of patients with EGFR TKI–resistant NSCLC with EGFR^T790M detected in plasma but not in a tumor biopsy still responded to osimertinib (32).

PIK3CA^M1043I, an activating mutation in the p110α catalytic subunit of PI3K (33), was found at 0.1% frequency in the same plasma sample where HER2^T798I was first detected (Supplementary Table S3). PIK3CA mutations are associated with resistance to anti-HER2 therapy in HER2-overexpressing breast cancers (34). Whether PIK3CA^M1043I contributes to a multifactorial resistance to neratinib is also possible but beyond the scope of this report. Although afatinib and neratinib are both irreversible covalent EGFR/HER2 TKIs, we found that afatinib, but not neratinib, was able to block HER2^L869R/T798I activity. We speculate that because neratinib is larger than afatinib, the former is more likely to be affected by a steric clash with the bulkier isoleucine residue in HER2^T798I (Fig. 3A and B). Treatment with low doses of afatinib (10 nmol/L), easily achievable in patients (35), completely blocked growth of MCF10A/HER2^L869R/T798I cells, whereas treatment with neratinib at clinically achievable concentrations (36) failed to do so (Fig. 4A and B). We also observed moderate activity of the osimertinib metabolite AZ5104 (Fig. 4). However, this drug is not being developed independently of osimertinib, and only approximately 10% of osimertinib is metabolized into AZ5104 in humans (20).

Immediately following progression on neratinib, the patient was treated with capecitabine chemotherapy. The patient responded well and remains in a partial response approximately 1 year later. We repeated NGS of her plasma tumor DNA after approximately 6 months on capecitabine; ERBB2^L869R cfDNA dropped to 0.4% and ERBB2^T798I and CCND1 amplification were no longer detectable, consistent with the decrease in tumor burden and the patient's clinical response. If the patient progresses on capecitabine and the ERBB2 mutations are once again detectable, there will be strong consideration for treatment with afatinib at that time. As more patients with HER2-mutant cancers are treated with HER2 TKIs such as neratinib, we expect that acquired HER2^T798I may be observed more frequently. We propose that afatinib is active against HER2^T798I and is an alternative worthy of clinical investigation in cancers harboring the HER2 gatekeeper mutation. Finally, this report supports the development of HER2^T798I-selective inhibitors that would spare the toxicity associated with therapeutic inhibition of WT ERBB receptors.

Extraction of cfDNA from plasma was performed using a fully automated QIAGEN platform, QIAsymphony SP, and QIAsymphony DSP Virus/Pathogen Midi Kit following centrifugation. Sequence libraries were prepared according to the KAPA Hyper protocol (Kapa Biosystems) with the ligation of Illumina sequence adaptors, followed by PCR amplification and clean-up. Barcoded libraries were hybridized with DNA probes targeting all coding exons of ERBB2 (Integrated DNA Technologies) in two successive captures, using a protocol modified from the NimbleGen SeqCap Target Enrichment system. The first capture was incubated at 55°C for 16 hours, followed by postcapture washes and 16 cycles of PCR amplification. The second capture was incubated at 65°C for 4 hours, followed by postcapture washes and 3 to 5 cycles of PCR amplification. Captured libraries were sequenced on an Illumina HiSeq as paired-end 100-bp reads.

Structural modeling of inhibitor-bound HER2^WT, HER2^L869R/HER3^E928G, and HER2^L869R/T798I was performed using Rosetta. Detailed procedures are available in Supplementary Methods.

The MCF10A breast epithelial cells (ATCC CRL-10317; purchased in 2012) and HEK293 human embryonic kidney cells (ATCC CRL-1573; purchased in 2006) were from ATCC. Cell lines were authenticated by ATCC prior to purchase by the short tandem repeat method. The 293FT cells were purchased from Invitrogen (cat. no. R70007). The NR6 cells have been described previously (21), as have BT474^GFP and BT474/HER2^T798M (24). ERBB2^T798M was verified by sequencing cDNA using primers for ERBB2. Other than routinely checking cell morphology for consistency with published images, no other authentication was performed.

The 293FT, HEK293, and NR6 cells were maintained in DMEM supplemented with 10% FBS and 1× antibiotic–antimycotic (Gibco). BT474 cells were maintained in Improved Minimum Essential Media supplemented with 10% FBS, 1× antibiotic–antimycotic, and 100 μg/mL G418. MCF10A cells were maintained in MCF10A complete media (DMEM/F12 supplemented with 5% horse serum, 20 ng/mL EGF, 10 μg/mL insulin, 0.5 μg/mL hydrocortisone, 0.1 μg/mL cholera toxin, and 1× antibiotic–antimycotic). For experiments under growth factor–depleted conditions, MCF10A cells were grown in DMEM/F12 supplemented with 1% charcoal/dextran-stripped serum, 10 μg/mL insulin, 0.5 μg/mL hydrocortisone, 0.1 μg/mL cholera toxin, and 1× antibiotic–antimycotic. Cell lines were routinely evaluated for Mycoplasma contamination. All experiments were completed less than 2 months after thawing early-passage cells.

The following inhibitors were used: MG132 (Selleck Chemicals), lapatinib (LC Laboratories), neratinib (PUMA Biotechnology), afatinib (Selleck Chemicals), and osimertinib and AZ5104 (AstraZeneca Pharmaceuticals).

Cells were washed with PBS and lysed on ice in RIPA lysis buffer plus protease and phosphatase inhibitors. Protein concentration was measured using the BCA protein assay reagent (Pierce). Lysates were subjected to SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad). Immunoreactive bands were detected by enhanced chemiluminescence following incubation with horseradish peroxidase–conjugated secondary antibodies (Promega). Detailed information on antibodies is available in Supplementary Methods. Immunoblot bands were quantified from inverted images using ImageJ software.

MCF10A cells were seeded in black clear-bottom 96-well plates (Greiner Bio-One) at a density of 1,000 cells per well in growth factor–depleted media. The next day, media were replaced with 100 μL media containing increasing amounts of inhibitor (0.17 nmol/L–10 μmol/L in 3-fold dilutions). After 5 to 6 days, nuclei were stained with 10 μg/mL Hoechst 33342 (Thermo Fisher Scientific) at 37°C for 20 minutes. Fluorescent nuclei were counted using the ImageXpress Micro XL automated microscope imager (Molecular Devices).

For 3-D growth assays, cells were seeded on growth factor–reduced Matrigel (BD Biosciences) in 48-well plates following published protocols (37). Inhibitors were added to the medium at the time of cell seeding. Fresh media and inhibitors were replenished every 3 days. Following 9 to 14 days, colonies were stained with 5 mg/mL MTT for 20 minutes. Plates were scanned and colonies measuring ≥100 μm were counted using GelCount software (Oxford Optronix). Colonies were photographed using an Olympus DP10 camera mounted in an inverted microscope.

Informed consent was obtained from the patient described in this study. The clinical trial (NCT01953926) was conducted in accordance with the Declaration of Helsinki and approved by an Institutional Review Board.

All experiments were performed using at least three technical replicates and at least two independent times. P values were generated by ANOVA followed by Tukey multiple comparisons test unless otherwise indicated. Data are presented as mean ± SD. IC₅₀ values were generated through GraphPad Prism (version 6.0).

Detailed descriptions of NGS, multiple sequence alignment, determination of mutation frequencies, transient transfections, and generation of stable cell lines are available in Supplementary Methods.

R. Lanman has ownership interest (including patents) in Guardant Health, Inc. D.M. Hyman reports receiving commercial research grants from AstraZeneca and Puma Biotechnology and is a consultant/advisory board member for Atara Biotherapeutics, Chugai, and CytomX. A.S. Lalani has ownership interest (including patents) in Puma Biotechnology. C.M. Lovly is a consultant/advisory board member for Ariad, Clovis, Genoptix, Novartis, Pfizer, and Sequenom. C.L. Arteaga is a consultant for Puma Biotechnology, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.B. Hanker, M.R. Brewer, D.B. Solit, D. Cross, J. Meiler, C.L. Arteaga

Development of methodology: A.B. Hanker, M.R. Brewer, D.B. Solit, J. He, J. Meiler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.B. Hanker, M.R. Brewer, J.P. Koch, R. Lanman, M.F. Berger, D.M. Hyman, D.B. Solit, J. He, R.E. Cutler Jr, C.L. Arteaga

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.B. Hanker, M.R. Brewer, J.H. Sheehan, G.R. Sliwoski, R. Nagy, R. Lanman, M.F. Berger, D.M. Hyman, C.M. Lovly, J. Meiler, C.L. Arteaga

Writing, review, and/or revision of the manuscript: A.B. Hanker, M.R. Brewer, J.H. Sheehan, R. Nagy, R. Lanman, D.M. Hyman, D.B. Solit, V. Miller, R.E. Cutler Jr, A.S. Lalani, D. Cross, C.M. Lovly, J. Meiler, C.L. Arteaga

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.R. Brewer, D.M. Hyman, C.L. Arteaga

Study supervision: M.R. Brewer, D.M. Hyman, D.B. Solit, J. Meiler, C.L. Arteaga

We would like to acknowledge the American Association for Cancer Research and its support in the development of the AACR Project GENIE registry, as well as members of the consortium for their commitment to data sharing. Interpretations are the responsibility of study authors. In addition, we offer our sincere gratitude to the patient for her contribution to this study.

This work was supported by NIH/NCI K12 award CA090625 (to A.B. Hanker), NCI R01 grant CA080195 (to C.L. Arteaga), ACS Clinical Research Professorship Award CRP-07-234-06-COUN (to C.L. Arteaga), a research grant from Puma Biotechnology, NIH Breast Cancer Specialized Program of Research Excellence (SPORE) grant P50 CA098131, and Vanderbilt-Ingram Cancer Center Support Grant P30 CA68485.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.