Combined Blockade of Activating ERBB2 Mutations and ER Results in Synthetic Lethality of ER+/HER2 Mutant Breast Cancer

Estrogen receptor positive (ER+) breast cancer is the most common breast cancer subtype, representing about 80% of all patients. Endocrine therapies, such as selective ER modulators (SERMs; i.e., tamoxifen), selective ER downregulators (SERDs; i.e., fulvestrant), and estrogen suppression with aromatase inhibitors (AI), are standard of care treatment for patients with advanced ER+ breast cancer. Many large randomized clinical trials have proven the effectiveness of these therapies in preventing metastatic recurrence (1, 2). However, about 20% of patients diagnosed with operable ER+ tumors will recur during or after adjuvant endocrine therapy. Despite the overall efficacy of these therapies, mortality from antiestrogen-resistant ER+ tumors still accounts for the majority of breast cancer deaths each year. Up to now, the only mechanisms of antiestrogen resistance that are supported in the clinic are HER2 amplification (3, 4), mutations in the ligand-binding domain (LBD) of ESR1 (5, 6), and dysregulation of the CDK4/6 pathway (7).

Human epidermal growth factor receptor 2 (ERBB2/HER2) is frequently altered via gene amplification in breast cancer. Some ER+ breast cancers may gain HER2 amplification and/or overexpression as a result of treatment pressures and/or tumor evolution (4). This may allow amplified HER2 signaling to drive tumor progression, thus reducing the dependence of a tumor on ER while also inducing resistance to antiestrogen therapy (3, 8, 9). The ERBB family of transmembrane receptor tyrosine kinases (RTK) consists of EGFR (ERBB1), HER2 (ERBB2), HER3 (ERBB3), and HER4 (ERBB4). Binding of ERBB ligands to the extracellular domain of EGFR, HER3, and HER4 induces the formation of kinase active hetero-oligomers (10). HER2 does not bind ERBB ligands directly, but it is in a conformation that resembles a ligand-activated state and favors dimerization (11, 12). Activation of HER2 induces transphosphorylation of the ERBB dimer partner and stimulates several intracellular pathways such as RAS/RAF/MEK/ERK, PI3K/AKT/TOR, Src kinases, and STAT transcription factors (reviewed in ref. 13). Activation of these pathways allows for extensive crosstalk with ER transcriptional activity, potentially leading to endocrine resistance.

ERBB2 missense mutations have been found in approximately 2% to 4% of breast cancers (14, 15), occurring with higher frequency in metastatic tumors (cBioPortal). Activating ERBB2 mutations occur most commonly in the kinase and extracellular domains in the absence of HER2 amplification (14–18). Among all ERBB2 missense mutations in breast cancer, approximately 70% occur in ER+ breast cancers and close to 80% of these occur in the kinase domain (KD; cBioPortal). Herein, we report that ERBB2 missense mutations occur at a significantly higher frequency in metastatic disease, thus suggesting these variants are acquired as a mechanism of resistance to endocrine therapy. Our studies also demonstrate that ERBB2 mutations hyperactivate HER3 and the PI3K/AKT/mTOR signaling pathway. In turn, this results in estrogen-independent growth and resistance to endocrine therapy. Based on these data, we posit that dual targeting of activating ERBB2 mutations or their downstream effectors, HER3/PI3K, and the ER pathway is necessary for optimal growth inhibition of ER+/HER2 mutant breast cancers.

The ER+ breast cancer cell line MCF-7 (ATCC HTB-22) was previously isogenically modified using AAV-mediated gene targeting to include 3 ERBB2 missense mutations (G309A, L755S, V777L) and maintained as previously described (18). Cell lines were not carried past 20 passages.

ER status of ERBB2 mutants was determined using databases from cBioPortal (cohorts include TCGA, METABRIC, GENIE, MBC Project, France 2016) and databases from Foundation Medicine. The frequency of ERBB2 mutations in primary and metastatic tumors was determined using cBioPortal (cohorts listed above, tissue-based) and databases from Guardant Health (plasma-based). Several selected patients from the Guardant Health database were analyzed for allelic frequency as a function of treatment course.

Clonogenic growth assays to determine resistance to estrogen deprivation or to fulvestrant were performed in 10-cm dishes. Drug sensitivity assays were carried out in 12-well dishes at a seeding density of 2,000 cells/well. Plates were stained using crystal violet. Cell growth was quantified using both a LI-COR Odyssey Infrared plate reader of stained monolayers and cell numbers measured with a Coulter Counter. For experiments with siRNAs, cells were reverse-transfected with HER3 siRNAs or scrambled control (Cell Signaling Technology) according to the Lipofectamine RNAiMAX protocol (Invitrogen). The next day, 2,000 cells/well were reseeded in full medium ± 1 μmol/L fulvestrant or estrogen-free medium ± 1 nmol/L estradiol. Cells were counted on days 0, 3, and 6 after plating. For assessment of HER3 levels, cells were harvested 3 days posttransfection and cell lysates were subjected to immunoblot analysis.

Cells were seeded at a density of 3,000 cells/well into 96-well plates in triplicate and treated with a dose range (0, 31.25, 62.5, 125, 250, 500, and 1,000 nmol/L) of fulvestrant and neratinib each. Media was replenished every other day. At the end of 5 days, cells were fixed and stained with crystal violet. Stained monolayers were lysed in 1% SDS and read at an absorbance of 560 nm using a GloMax Plate reader. Combination Index values were calculated for the dose combination 1 μmol/L fulvestrant and 250 nmol/L neratinib using the computer program CompuSyn. Values less than 1 indicate a synergistic interaction.

Three-dimensional (3D) morphogenesis assays were carried out as previously described (18). Briefly, cells were seeded in growth factor-reduced Matrigel (BD Biosciences) in the absence of estradiol in 8-well chamber slides at a density of 2.5 × 10⁴ cells/mL. Chambers were supplemented with 1 nmol/L estradiol, 1 μmol/L fulvestrant, and/or 200 nmol/L neratinib. Chambers were imaged with an Olympus DP20 microscope on day 14.

Cells were washed with PBS and lysed on ice in NP-40 lysis buffer plus protease and phosphatase inhibitors. Protein concentration in cell lysates was measured using BCA protein assay reagent (Pierce). Lysates were subjected to SDS-PAGE followed by transfer to nitrocellulose membranes (Invitrogen) as previously described (19). Primary antibodies included: p-HER2 Y1248 (Millipore 06-229), HER2 (CS2242), p-HER3 Y1197 (CS4561), HER3 (CS12708), PathScan Multiplex Western Cocktail I (for p-AKT S473, p-p90RSK, p-ERK, and p-S6), p-S6K (CS9205), AKT (CS9272), ERK (CS9102), S6K (CS9202), p90RSK (CS9333), S6 (CS2217), p-ERα S167 (SC-101676), ERα (SC-73479), and GAPDH (CSXP5174). Immunoreactive bands were detected by enhanced chemiluminescence following incubation with horseradish peroxidase-conjugated secondary antibodies. The Invitrogen NuPage system was used. Nitrocellulose membranes were cut horizontally to probe with multiple antibodies.

Cells were plated in 96-well plates at a density of 4,000 cells/well in Improved Minimum Essential Medium, Gibco (IMEM) media with 10% charcoal-stripped serum; 24 hours later, cells were transfected with pERE (estrogen responsive elements)-luciferase and pCMV-Renilla plasmids. Twenty-four hours posttransfection, cells were treated with vehicle (control), 1 nmol/L estradiol, 1 μmol/L fulvestrant, and/or 200 nmol/L neratinib. Luciferase activity was measured 24 to 48 hours later using the Dual Luciferase Kit (Promega) according to the manufacturer's instructions utilizing the Promega GloMax plate reader.

Gene expression analysis was performed as previously described (20). Briefly, cells were plated in estrogen-free media in triplicate. Cells were harvested and RNA was purified using the RNeasy Kit (Qiagen). cDNA was generated using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems) and analyzed using the Estrogen Receptor PCR Array (Qiagen; PAHS-005Z).

Mouse experiments were approved by the Vanderbilt Institutional Animal Care and Use Committee. Female ovariectomized athymic mice were implanted with a 14-day release 17β-estradiol (E2) pellet (0.17 mg). Twenty-four hours later, 5 × 10⁶ MCF-7 cells harboring a ERBB2 L755S or V777L mutation, as indicated, suspended in PBS and Matrigel (BD Biosciences) at a 1:1 ratio were injected subcutaneously into the dorsum of each mouse. Approximately 4 weeks later, mice bearing tumors of ≥150 mm³ were randomized to treatment with vehicle (control), neratinib [40 mg/kg/day via orogastric gavage (o.g.)], fulvestrant (5 mg/wk, s.c.), everolimus (5 mg/kg/day, o.g.), and/or alpelisib (30 mg/kg/day, o.g.), as indicated. Tumor volume in mm³ was measured 2 times a week by using the formula: volume = width² × length/2. Portions of tumors were snap frozen or fixed in 10% neutral-buffered formalin and embedded in paraffin for subsequent analyses. Five millimeters of paraffinized sections were used for IHC using a phospho-S6 antibody (Cell Signaling 2211). Sections were scored by an expert pathologist (PLG-E) blinded to treatment arm. Fulvestrant was kindly provided by AstraZeneca Pharmaceuticals.

Cell lysates were harvested using ice cold lysis buffer [1% NP-40, 20 mmol/L Tris pH 7.4, 10% Glycerol, 150 mmol/L NaCL, 1 mmol/L EDTA, 1 mmol/L EGTA, 5 mmol/L sodium pyrophosphate, 50 mmol/L NaF, 10 mmol/L β-glycerophosphate, protease, and phosphatase inhibitor cocktails (Sigma-Aldrich)] and rotated at 4°C for 1 hour. Lysates were then clarified by spinning at 10,000 × g at 4°C for 15 minutes. Protein concentrations were measured using BCA standard curves (Pierce). Lysates were precleared with 50 μL of Protein G agarose beads (Life Technologies 10003D) at 4°C for 1 hour. Immunoprecipitation was carried out using the Invitrogen Immunoprecipitation Kit (10004D) as directed. Lysates were next subjected to SDS-PAGE and immunoblot analysis. Immunoblots were quantified using ImageJ software.

Known EGFR-like structures were identified using either HHPRED search of the PDB with the HER2 KD query sequence (SwissProt sequence P04626.1, residue range 685–1,255) or BLAST sequence search (21) of all protein kinase-like domain sequences defined in Evolutionary Classification Of structure Domains (ECOD; ref. 22). Kinase structure domains outlined in ECOD were defined manually as active or inactive based on the position of the regulatory C helix and the proximity of the K753/E770 sidechains (numbered according to HER2), whose ion pair interaction are a hallmark of protein kinase activation. Manually defined active and inactive conformations were double checked using Dali superpositions. The flexibility of residue positions corresponding to HER2 L755 were assessed using the structure B-factor of the Cα atom for all identified EGFR-like kinase structures containing coordinates for that residue. B-factors were normalized by the Z-score of the PBD, where the mean and standard deviation are calculated from all Cα atoms in the same chain. Accessible surface (Gerstein) of wild-type (WT) EGFR (pdb 4riw, chain B) and mutant (pdb 4riw, chain B, L723 replaced with Ser) structures were calculated with default parameters (23). Changes in calculated surface exposure between WT and mutant were compared for all residues, with those surrounding the mutation becoming increasingly exposed. The stability scores resulting from mutating L755S in the background of WT EGFR and HER2 crystal structures, as well as HER2 structure models based on EGFR templates, were assessed with the site directed mutator (SDM) server (24), which calculates a potential energy function based on the environment-specific amino acid substitution frequencies within homologous protein families.

All experiments were performed using 3 technical replicates and at least 2 independent times. P values were calculated using GraphPad Prism (version 6.0) by ANOVA followed by Tukey multiple comparisons test.

Interrogation of cBioPortal and Foundation Medicine databases found that approximately 70% of ERBB2 mutations are detectable in ER+/HER2 nonamplified breast cancers, with the majority of them occurring in the KD (Supplementary Table S1). Further interrogation of several genomic databases available through the cBioPortal interface (TCGA, METABRIC, GENIE, MBC Project, France 2016), as well as cancer patients' plasma specimens in the Guardant Health database, showed a significantly higher occurrence of ERBB2 mutations in patients with metastatic disease versus primary tumors (Fig. 1A). Within the GENIE database, the percentage of ERBB2 mutations in metastatic biopsies (4.3%) almost doubled that in primary cancers (2.5%), altogether suggesting they are an acquired mechanism of antiestrogen therapy resistance. This was further supported by plasma ctDNA analysis from a small cohort of patients with ER+/HER2-negative (HER2 nonamplified) breast cancer in the Guardant Health database (Fig. 1B). Comparison between pretreatment and postprogression on endocrine therapy showed the emergence and/or an increase in the allelic frequency of activating ERBB2 mutations. An in-depth mutational analysis of tumors from patient 1, harboring the L755S mutation, and patient 2, harboring the V777L mutation, showed the emergence of the ERBB2 mutations as a dominant alteration in later samples (Fig. 1C and D). Interrogation of online databases demonstrated that ERBB2 mutations were mutually exclusive with ESR1 mutations, a mechanism of escape from hormone dependence (Supplementary Fig. S1). These suggest that ERBB2 mutations operate independently of the ER pathway in the development of antiestrogen resistance.

In a case study of a postmenopausal woman with advanced ER+ breast cancer who had progressed after multiple endocrine therapies and was considered resistant to antiestrogens, next gene sequencing (NGS) of DNA from a skin metastasis revealed a ERBB2 L869R activating mutation (Fig. 1E). The patient was placed on neratinib monotherapy and, after exhibiting an excellent clinical response, progressed with bone and lymph node metastases (16). Addition of the ER antagonist fulvestrant to neratinib resulted to in significant and prolonged tumor regression suggesting that ER+ breast cancers with ERBB2-activating mutations may require this therapeutic combination.

To determine if ERBB2 missense mutations are causally associated with endocrine therapy resistance, we used ER+ MCF7 breast cancer cells that had been isogenically modified to contain 3 relatively common ERBB2 missense mutations in the cell's genome; G309A in the extracellular domain (ECD), and L755S and V777L, both in the KD (18). Isogenic cell lines were previously developed using AAV-mediated gene targeting, which incorporates the desired point mutation into the host cell's genome through homologous recombination. Mutations were incorporated into a single allele and expressed under the cell's endogenous promoter. This provides a biologically relevant model that best mimics mutations in primary tumors. MCF7s require estrogen in order to propagate exponentially. MCF7^L755S and MCF7^V777L cells but not cells with the ECD mutation, MCF7^G309A, or cells with WT HER2 exhibited robust growth in the absence of estradiol (Fig. 2A), a scenario akin to that of a postmenopausal patient treated with aromatase inhibitors. Establishment of long-term estrogen-deprived (LTED) stable cell lines showed similar trends with MCF7 cells harboring both KD mutations propagating early in the absence of estradiol. Conversely, MCF7^WT and MCF7^G309A cells exhibited near-complete growth arrest and required 20+ days before achieving confluence (Fig. 2B and C).

Similarly, MCF7^L755S and MCF7^V777L cells were resistant to the selective estrogen receptor degrader (SERD) fulvestrant, whereas MCF7^WT and MCF7^G309A cells were growth inhibited >90% compared with untreated controls (Fig. 2D). Development of fulvestrant-resistant lines exhibited trends nearly identical to the development of the LTED lines (Supplementary Fig. S2). Fulvestrant was still able to downregulate ER protein levels as well as inhibit ligand-independent and estradiol-induced ER reporter activity in all 4 cell types (Fig. 2E and F), suggesting the resistance associated with the ERBB2 KD mutations was not explained by loss of fulvestrant's action on ER.

We next investigated the effect of the pan-HER irreversible tyrosine kinase inhibitor (TKI) neratinib alone or in combination with fulvestrant on cells with WT or mutant HER2. MCF7 cells with WT HER2 and with the ECD mutation showed a near complete response to fulvestrant alone, whereas cells harboring the 2 KD mutations responded only partially. Treatment with neratinib partially inhibited proliferation of all 3 cells harboring knock-in ERBB2 mutations. Because low levels of WT HER2 play no oncogenic role in MCF7 cells, neratinib was inactive against MCF7^WT cells (Fig. 3A). Complete growth inhibition of the HER2 mutant cells was only observed upon treatment with both neratinib and fulvestrant, further suggesting that in ER+ tumors harboring activating ERBB2 mutations, both ER and HER2 signaling drive cell viability. To determine if neratinib and fulvestrant were synergistic, a dose range (0, 31.25, 62.5, 125, 250, 500, and 1,000 nmol/L) of each drug alone and in combination was analyzed to calculate the combination index. For the combination of 250 nmol/L of neratinib and 1 μmol/L of fulvestrant, MCF7^V777L cells (Fig. 3B) and MCF7^L755S cells (Supplementary Fig. S3) exhibited a combination index of 0.22 and 0.49, respectively, both of which are considered synergistic. Similar results to the 2D in vitro assays were observed in 3D-matrigel. MCF7^L755S and MCF7^V777L cells formed large invasive irregular acini in the absence of estrogen, which were ablated by the addition of neratinib, consistent with a transformed phenotype driven by aberrant HER2 signaling (Fig. 3C, top 2 rows). Picomolar concentrations of estradiol rescued MCF7^L755S and MCF7^V777L from the effect of the HER2 TKI; addition of fulvestrant to estradiol and neratinib restored acinar growth inhibition, further supporting dual ER and HER2 blockade are required to ablate their invasive phenotype (Fig. 3C, last 2 rows).

In ER+ breast cancer cells with amplification of WT HER2, there is clear evidence of crosstalk between the HER2 and ER signaling pathways (8, 25). Thus, we next examined if nonamplified mutant HER2 was stimulating estradiol-independent ER transcriptional activity. Using an ERE luciferase reporter, baseline and estradiol-stimulated ER transcriptional reporter activity was overall no different in HER2 mutant versus WT cells. Moreover, treatment with fulvestrant but not with neratinib abrogated basal and/or estradiol-induced transcriptional activity in all HER2 mutant cells (Fig. 3D), further suggesting no interdependence of mutant HER2 and ER signaling in MCF7^L755S and MCF7^V777L cells. We next examined a real-time (RT) Profiler array of 84 ER-regulated genes in MCF7^L755S and MCF7^V777L cells under estrogen-deprived conditions. When compared with MCF7^WT cells, cells harboring the ERBB2 KD mutations showed modest changes in expression (≥2-fold) in only a small number of ER regulated genes. MCF7^V777L exhibited modest upregulation in L1CAM, JUNB, and BCAR1, and downregulation in PGR1 (Fig. 3E), whereas MCF7^L755S exhibited upregulation in WISP2 and modest downregulation in FOXA1, CTSD1, and GPER1 (Supplementary Fig. S4). The absence of common genes between the 2 KD mutants suggests ER transcriptional activity does not play a role in the estrogen independent growth observed in both.

HER2-dependent transformation and metastatic progression of breast cancer cells with WT HER2 gene amplification are mainly attributed to hyperactivation of the PI3K/AKT/mTOR survival pathway, with HER2/HER3 heterodimers being the most transforming of this receptor network (26, 27). HER3, which lacks intrinsic kinase activity, is able to potently activate PI3K via its 6 docking sites for the p85 adaptor subunit of the PI3K dimer (28). Moreover, several approved HER2 antagonists exert their antitumor effect, at least in part, by inhibiting phosphorylation of HER3 and disabling PI3K/AKT/mTOR signaling (9, 29). Further, phosphorylated HER3 was reported to be elevated in cells endogenously expressing the ERBB2^V777L mutation compared with cells with WT HER2 (18). Thus, we next examined the role of HER3 in transformation induced by nonamplified mutant HER2. Immunoblot analysis showed detectable Y1197 p-HER3 in MCF7^V777L cells; treatment with fulvestrant resulted in a clear increase in p-HER3 in MCF7^L755S and MCF7^V777L cells (Fig. 4A). This increase was abrogated by neratinib, suggesting it is due to HER2-mediated trans-activation of HER3. To determine whether HER3 function is causally associated with endocrine therapy resistance, we knocked down HER3 with 2 independent siRNAs. siRNA-mediated knockdown of HER3 was confirmed by immunoblot analysis of lysates from transfected MCF7^V777L cells (Fig. 4B). Monolayer growth analysis showed that transfection of HER3 siRNA resensitized MCF7^V777L cells to fulvestrant (Fig. 4C). HER3 activation requires heterodimerization with other ERBB family receptors. Immunoprecipitation of HER3 followed by HER2 immunoblot of HER3 antibody pulldowns revealed that both HER2 KD mutants exhibit higher levels of HER3:HER2 complexes (Fig. 4D and E), further implying that ER+/HER2 mutant cells gain ER independent growth through enhanced activation of HER3.

Structural analysis of the HER2^V777L by Bose and colleagues previously determined that the V777L mutation mirrored alterations observed in the EGFR regulatory DFG motif with altered kinase activity. These authors also proposed that HER2^L755S adopts a confirmation that may sterically inhibit small molecule binding and in return could produce resistance to lapatinib (14). Thus, we next inquired a structural explanation for the apparent enhanced association of HER2^L755S with HER3 using known HER2 and EGFR structures. While numerous crystal structures are available for EGFR, both in active and inactive states as well as bound to HER3, relatively few are available for HER2. Given the homologies in amino acid sequence (Supplementary Fig. S5), the proposed mechanism of activation of the HER2 KD (30), and previously described structural analysis methods by Bose and colleagues (14), we compared both EGFR and HER2 structures (Fig. 4F). HER2 residue L755 resides at the C-terminus of the β-strand prior to the regulatory αC helix in the kinase N-lobe. L755 is adjacent to the characteristic ion pair (K753/E770) whose formation of a salt bridge marks the active kinase conformation (31). HER3 binds the active EGFR kinase, forming an asymmetric heterodimer similar to activating homodimers of either EGFR or HER2 (30, 32). Superposition of inactive EGFR and HER2 kinase structures with active EGFR structures alone or bound to HER3 highlighted the relative positions of the ion pair, αC helix and L755. Notably, L755 forms a stable hydrophobic core in the inactive conformations (Fig. 4F, orange sidechains), whereas it appears more flexible in the active conformations (Fig. 4F, magenta sidechains). This flexibility extends to the following loop, which helps form the HER3 binding site in the activating heterodimer. To examine if this observed flexibility is a general feature of active EGFR conformations, we compared the B-factors of sidechains corresponding to L755 in 122 available EGFR-like kinase structures (including mainly EGFR, 2 HER2, and 4 HER4) in either active or inactive conformations (Fig. 4G). Indeed, the distribution of normalized B-factors for L755 shifted from average flexibility in inactive conformations to increased flexibility in active conformations, suggesting the HER2^L755S conformation promotes increased binding with HER3.

To identify differential intracellular signaling between cells with mutant versus WT HER2, we used an array of 43 phosphorylated human kinases in estrogen-deprived MCF7 cells. The main difference between both cell types was persistent phosphorylation of p70S6K (S6K) in MCF7^L755S and MCF7^V777L cells versus MCF7^G309A and MCF7^WT cells (Fig. 5A). Immunoblot analysis of S6, the downstream target of S6K, was also clearly higher in cells with ERBB2 KD mutations in the presence or absence of estrogen (Fig. 5B). Further interrogation of signal transducing molecules by immunoblot analysis revealed elevated levels of T308 p-AKT, S473 p-AKT, and p-S6 in cells with ERBB2 mutations, particularly in cells with ERBB2 KD mutations (Fig. 5C). In MCF7^G309A and MCF7^WT cells but not in MCF7^L755S and MCF7^V777L cells, treatment with fulvestrant completely inhibited p-S6 levels. In MCF7^V777L cells, treatment with neratinib did not affect p-S6 levels except in combination with fulvestrant (Fig. 5C).

We next examined whether inhibition of signaling molecules altered in cells harboring ERBB2 KD mutations would restore sensitivity to fulvestrant. For this purpose we used the PI3Kα inhibitor alpelisib, the TORC1 inhibitor everolimus, and the MEK1/2 inhibitor selumetinib. Clonogenic growth assays showed treatment with alpelisib and everolimus, but not with selumetinib, restored the growth inhibitory action of fulvestrant (Figs. 5D; Supplementary Fig. S6). Quantification revealed there was no statistically significant difference between alpelisib, everolimus, or neratinib when combined with fulvestrant (Supplementary Fig. S7). Further, immunoblot analysis of cell lysates showed that alpelisib and everolimus but not selumetinib, each in combination with fulvestrant completely abrogated p-S6 levels (Supplementary Fig. S8), supporting a role for PI3K/AKT/mTOR but not RAS/RAF/MEK/ERK signaling on mutant ERBB2-induced endocrine resistance. In each of these studies, we used doses of small molecules that effectively inhibited their molecular target, S473 p-AKT for alpelisib, p-S6 for everolimus, and p-ERK for selumetinib (Supplementary Figs. S9–S11).

Finally, we extended these findings to studies in mice bearing HER2 mutant xenografts. Athymic ovariectomized mice with established MCF7^V777L tumors were first treated with fulvestrant, which arrested tumor growth but did not induce tumor regressions (Fig. 6A). In a second study, neratinib and fulvestrant, each alone, prevented tumor growth but only the combination induced regression of both MCF7^V777L and MCF7^L755S xenografts (Fig. 6B; Supplementary Fig. S12). In a third experiment, mice with established MCF7^V777L tumors were treated with fulvestrant alone or in combination with neratinib, alpelisib, or everolimus. As observed in the in vitro growth experiments, there was no significant difference in tumor growth among any of the combinations with fulvestrant (Fig. 6C). Finally, established MCF7^V777L tumors were treated with everolimus, neratinib/everolimus, fulvestrant/everolimus, or the triple combination of neratinib/fulvestrant/everolimus to determine the role of p-S6 in tumor progression. Similar to neratinib and fulvestrant, everolimus was modestly effective as a single agent. The combinations of neratinib/everolimus and neratinib/fulvestrant equally suppressed tumor progression and to a higher degree than single agent everolimus. Interestingly, treatment with the triple combination of neratinib, fulvestrant, and everolimus induced close to complete tumor regression (Fig. 6D). We speculate the incomplete inhibition of p-S6 by everolimus may be due to mTOR independent activation of S6K through mutant HER2-induced activation of RSK (33, 34). IHC analysis of tumor sections revealed that treatment with the triple combination induced a significant decrease in p-S6 levels when compared with single-agent or any double combination, thus correlating with the near complete tumor responses observed (Supplementary Fig. S13).

Although some mechanisms of endocrine therapy resistance in ER+ breast cancer have been well studied, several of these remain to be discovered and characterized. Here we report that ERBB2 mutations are predominantly acquired in late-stage ER+ breast cancer and may play an important role in the progression of estrogen independent ER+ tumors. It is generally accepted that in ER+ breast cancers with overexpression of WT HER2, there is molecular crosstalk between HER2 and ER signaling pathways, resulting in endocrine therapy resistance (35–38). Whether similar transactivation between these 2 pathways is operative in ER+/HER2 nonamplified mutant breast cancer cells is unclear. Our data suggested that ERBB2 mutations are acquired in metastatic ER+ disease after exposure to endocrine therapy (Fig. 1). Hence, we examined herein the role of ERBB2 mutations in endocrine resistance. Analysis of the cBioPortal and Genie databases revealed the ERBB2 mutations have nearly complete mutual exclusivity from activating ESR1 mutations (Supplementary Fig. S1; refs. 5, 6), suggesting that HER2 variants use may use an ERα-independent mechanism to mediate escape from estrogen suppression. Our studies with MCF7 cells with isogenically incorporated ERBB2 mutations revealed that variants L755S and V777L in the KD, but not G309A in the extracellular domain, generate estrogen independent growth and resistance to fulvestrant (Fig. 2). Combined with the clinical data, this result strongly supports ERBB2-activating KD mutations are an acquired mechanism of endocrine therapy resistance.

Previous studies have demonstrated that patients with HER2-nonamplified breast cancer do not benefit from HER2-directed therapies (39). However, the identification of ERBB2 mutations in breast cancers without HER2 gene amplification has led to clinical trials showing activity of HER2-directed therapies against these “low HER2” tumors (40, 41). The ongoing SUMMIT ERBB2 Mutation Trial (ClinicalTrial identifier NCT01953926) identified neratinib as an effective drug against HER2 mutant breast and is currently assessing the efficacy of neratinib in combination with fulvestrant and trastuzumab (40, 42). In this study, using in MCF7 cells harboring activating ERBB2 KD mutations, we determined dual blockade of ER and HER2 signaling was required to achieve complete growth arrest and cell death (Fig. 3A and 5D). Very low levels of estrogen are observed in postmenopausal women. To recapitulate these conditions, we used picomolar levels of estradiol. These low levels of estradiol rescued cells with ERBB2 KD mutations from neratinib action (Fig. 3C), further suggesting the need of simultaneously disabling ER and mutant HER2 function to achieve an optimal antitumor effect. Crosstalk between ER and HER2 signaling and the ability of HER2 to downregulate ER have been well characterized in ER+/HER2 amplified cells and tumors (43). The heterozygous, low copy incorporation of ERBB2 mutations into the MCF7 genome used herein is biologically relevant to HER2 mutant tumors, as the overwhelming majority of these variants are not amplified in primary human tumors. In our studies, we found no evidence to suggest HER2 mutant signaling activates ER function and/or limits the effect of fulvestrant or estrogen deprivation on ER transcriptional activity (Figs. 2E, 2F, 3C, and 3D; Supplementary Fig. S4). These suggest that unlike endocrine resistance drive by HER2 amplification (44), HER2 mutant tumors develop anti-ER therapy resistance independently of ER activity.

Induction of HER3 was previously identified as a mechanism for promoting re-expression of genes mediating endocrine-resistant breast cancer cell growth (45). In line with this report, we found constitutively activated HER3 levels in MCF7^L755S and MCF7^V777L cells, particularly upon treatment with fulvestrant (Fig. 4A). HER3 is catalytically inactive and requires hetero-dimerization with other ERBB (HER) receptors. Upon dimerization, ERBB co-receptors phosphorylate HER3 which, in turn, can couple to signal transducers mainly the PI3K/AKT/mTOR pathway (44). Structural analysis of L755S HER2 suggested an increase in flexibility in the active state, thus allowing for an increased propensity for heterodimerization and activation of HER3. Mutation of L755 to serine replaces a hydrophobic sidechain that helps stabilize the N-lobe core in inactive EGFR-like kinases with a smaller hydrophilic side chain (Fig. 4F). The serine mutation increases solvent exposure of the sidechain in both EGFR and HER2 structures according to SDM (24) and allows the surrounding residues to become more solvent exposed, thus increasing flexibility (46). We propose this increased flexibility mimics that observed in activated states of EGFR-like kinase structures, thus shifting HER2 towards the activated state and enhancing its binding to HER3 (Fig. 4G). This would be consistent with the observed hyperactivity of the HER3/PI3K/AKT/mTOR axis in in MCF7^L755S and MCF7^V777L cells and its potential central role on the transformed phenotype induced by ERBB2-activating mutations.

Inhibition of HER2 (with neratinib), TORC1 (with everolimus), or PI3Kα (with alpelisib) in combination with fulvestrant were equipotent in vitro (Fig. 5D) and in vivo (Fig. 6C); however, complete tumor regressions in vivo were only achieved with the triple combination of fulvestrant/neratinib/everolimus (Fig. 6D). IHC analysis revealed that inhibition of both TORC1 and mutant HER2 were required in order to achieve an optimal reduction in p-S6 levels (Supplementary Fig. S13). Previous studies have shown that MCF7^V777L cells exhibit increased PI3K signaling and cell migration (18). Further, online databases (cBioPortal, Genie) revealed that ERBB2 mutations have a tendency to co-occur (log odds ratio = 0.323) with PIK3CA mutations (Supplementary Fig. S1), suggesting a natural selection geared towards amplification of PI3K signaling. Combined with our data, these data also suggest that combinations of endocrine therapy with PI3K pathway inhibitors would be a rational strategy worthy of clinical investigation in cancers with activating ERBB2 mutations.

In conclusion, we report herein that ERBB2-activating mutations predominate in late stages of ER+ breast cancer after long-term ER suppression. These mutations led to estrogen independent growth and fulvestrant resistance with no significant changes in ER transcriptional activity. Dual blockade of HER2 and the ER pathways was necessary to effectively inhibit growth of ER+/HER2 mutant tumors. In addition, HER2 mutant signaling hyperactivated HER3/PI3K/AKT/mTOR, thus suggesting a causal association of aberrant activity of this pathway to endocrine resistance in patients with ER+/HER2 mutant breast cancer. Taken together, these studies suggest that patients with ER+/HER2 nonamplified breast cancer-harboring ERBB2 mutations would benefit from HER2 targeted therapies in combination with hormonal therapy.

A. Mathew is a consultant/advisory board member for AstraZeneca. E.H. Bernicker is a consultant/advisory board member for Abbvie, AstraZeneca, and Guardant Health. M. Cristofanilli reports receiving commercial research grants from Merus, Novartis, and Pfizer. V.A. Miller is an employee of and holds ownership interest (including patents) in Foundation Medicine, and is a consultant/advisory board member for Revolution Medicines. R.B. Lanman holds ownership interest (including patents) in BIolase, Inc., Forward Medical, Inc., and Guardant Health, Inc., and is a consultant/advisory board member for Forward Medical, Inc. C. Arteaga reports receiving commercial research grants from Lilly, Pfizer, and Radius, holds stock options in Provista and Y-Trap, and is a consultant/advisory board member for Daiichi Sankyo, H3Biomedicine, Lilly, Merck, Novartis, OrigiMed, PUMA Biotechnology, Radius, Sanofi, Symphogen, and TAIHO Oncology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Croessmann, L. Formisano, C.L. Arteaga

Development of methodology: S. Croessmann, L. Formisano, L.N. Kinch, A. Mathew, J. He, R.B. Lanman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Croessmann, L. Formisano, P.I. Gonzalez-Ericsson, D.R. Sudhan, R.J. Nagy, A. Mathew, E.H. Bernicker, M. Cristofanilli, R.B. Lanman, C.L. Arteaga

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Croessmann, L. Formisano, L.N. Kinch, P.I.Gonzalez-Ericsson, R.J. Nagy, A. Mathew, M. Cristofanilli, J. He, R.E. Cutler, Jr., V.A. Miller, R.B. Lanman, N.V. Grishin, C.L. Arteaga

Writing, review, and/or revision of the manuscript: S. Croessmann, L.N. Kinch, P.I. Gonzalez-Ericsson, D.R. Sudhan, R.J. Nagy, A. Mathew, E.H. Bernicker, M. Cristofanilli, R.E. Cutler, Jr., A.S. Lalani, V.A. Miller, R.B. Lanman, N.V. Grishin, C.L. Arteaga

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Croessmann, R.J. Nagy, A.S. Lalani, C.L. Arteaga

Study supervision: S. Croessmann, N.V. Grishin, C.L. Arteaga

Other (funding): C.L. Arteaga

We thank Teresa Dugger for general administration and technical assistance and Ariella Hanker for helpful discussions. This study was funded by NIH Breast SPORE grant P50 CA098131, Vanderbilt-Ingram Cancer Center Support grant P30 CA68485, UT Southwestern Simmons Cancer Center Support Grant P30 CA142543 CPRIT RR170061 grant, Susan G. Komen for the Cure Foundation grant SAC100013 (to C.L. Arteaga), grants from the Breast Cancer Research Foundation (to C.L. Arteaga), National Institutes of Health (GM094575 and GM127390 to N.V. Grishin) and the Welch Foundation (I-1505 to N.V. Grishin).

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.