Seventy percent of breast cancers express the estrogen receptor (ER), and agents that target the ER are the mainstay of treatment. However, virtually all people with ER+ breast cancer develop resistance to ER-directed agents in the metastatic setting. Beyond mutations in the ER itself, which occur in 25–30% of people treated with aromatase inhibitors1–4, knowledge about clinical resistance mechanisms remains incomplete. We identified activating HER2 mutations in metastatic biopsies from eight patients with ER+ metastatic breast cancer who had developed resistance to aromatase inhibitors, tamoxifen or fulvestrant. Examination of treatment-naive primary tumors in five patients showed no evidence of pre-existing mutations in four of five patients, suggesting that these mutations were acquired under the selective pressure of ER-directed therapy. The HER2 mutations and ER mutations were mutually exclusive, suggesting a distinct mechanism of acquired resistance to ER-directed therapies. In vitro analysis confirmed that the HER2 mutations conferred estrogen independence as well as—in contrast to ER mutations—resistance to tamoxifen, fulvestrant and the CDK4 and CDK6 inhibitor palbociclib. Resistance was overcome by combining ER-directed therapy with the irreversible HER2 kinase inhibitor neratinib.
The epidermal growth factor receptor 2 (ERBB2, which encodes HER2) is frequently altered in cancer; in breast cancer this is manifested primarily via gene amplification5 or HER2 overexpression. There is considerable in vitro evidence that HER2 signaling may have a complementary role to the estrogen pathway in estrogen receptor–positive (ER+) breast cancer, through ER cross-talk or downstream signaling to provide survival signals in the context of estrogen deprivation6,7,8,9,10,11,12,13 (see Supplementary Note). Mutations in HER2 are comparatively rare in breast cancer, accounting for 1.6% of primary breast cancers in The Cancer Genome Atlas study14. Several activating hotspot mutations in the kinase and extracellular domains have been characterized14,15,16,17. However, the role of such mutations in metastatic breast cancer (MBC) is less well understood.
As part of an ongoing sequencing study of ER+ MBC18, we performed whole-exome sequencing (WES) of metastatic tumor biopsies from 168 patients, nearly all of whom had received prior endocrine therapy18. In 12 patients, we identified mutations in ERBB2, including hotspot mutations in the kinase domain, as well as in the extracellular, transmembrane, and cytoplasmic domains (Fig. 1a). The kinase domain mutants HER2 p.Leu755Ser, p.Val777Leu, and p.Leu869Arg have previously been identified and characterized as activating in breast cancer14,15. The transmembrane domain alteration p.Ser653Cys (homologous to the activating epidermal growth factor receptor (EGFR) alteration p.Ser645Cys19,20) has not been described in breast cancer, although it was previously observed in one patient with urothelial bladder carcinoma, where it was characterized as activating and sensitive to lapatinib21. One metastatic biopsy had both p.Gly727Ala and p.Val777Leu alterations; this combination has been reported in breast cancer22, and homologous alterations to p.Gly727Ala are activating in EGFR (Gly719)23 and found in combination with other EGFR mutations in non-small cell lung cancer24. The extracellular domain and cytoplasmic domain mutants have not been previously reported. Overall, the increased prevalence of HER2 alterations in the metastatic setting seen here (compared to primary ER+ breast cancer5,25,26) is consistent with recent sequencing studies that included MBC27,28,29,30,31 (Supplementary Fig. 1 and Supplementary Table 1), though those studies did not specifically discuss HER2 alterations in MBC.
To determine whether these 12 HER2 alterations were acquired in the metastatic setting, we obtained and performed WES on corresponding primary tumor biopsies collected before exposure to any endocrine therapy in eight patients18. In six of the eight primary tumors (75%), the HER2 alterations identified in the metastatic biopsies were not observed, suggesting that these alterations were acquired over the course of therapy (Fig. 1a, red triangles).
Among eight patients with activating HER2 alterations in their metastatic biopsies (Fig. 1b), the activating HER2 alterations were found to be acquired in four patients, shared in one, and indeterminate in three. All patients were treated with ER-directed therapy before their metastatic biopsies, including tamoxifen (five patients), aromatase inhibitors (six patients) and fulvestrant (two patients). For the four patients with uncharacterized HER2 alterations, two were acquired, one shared, and one indeterminate (Supplementary Fig. 2). Additional alterations were observed in 593 cancer genes in metastatic biopsies and corresponding primary tumors (Fig. 2a). Detailed clinicopathological features, therapies, and genomic data for all patients and samples are in Supplementary Tables 2–6.
To further explore the role of acquired HER2 alterations in these tumors, we performed an evolutionary analysis to evaluate clonal structure and dynamics. We evaluated the change in the estimated fraction of tumor cells harboring each genomic alteration (the cancer cell fraction, CCF) from the pretreatment primary biopsy to the resistant metastatic biopsy (Fig. 2b). In all four patients, HER2 alterations were not detected in the primary tumor, despite power to detect alterations at this locus (Supplementary Table 5). In three of the metastatic biopsies, the activating HER2 alterations were clonally acquired. In the fourth biopsy, the HER2 alteration was subclonal, although evaluation of the clonality was confounded by a concurrent HER2 amplification. Other alterations, including known driver alterations, were clonal in both primary and metastasis in these patients. Evolutionary analysis for patients with uncharacterized HER2 alterations is in Supplementary Fig. 3.
Although ESR1 mutations are an established mechanism of resistance to aromatase inhibitors in ER+ MBC1,2,3,4, none of the patients with acquired HER2 alterations had ESR1 mutations in their metastatic biopsies, suggesting that these might be mutually exclusive events. Examination of publicly available sequencing data from the American Association for Cancer Research (AACR) Project GENIE database V1.0.1 (ref. 30) showed that ERBB2 and ESR1 mutations were indeed mutually exclusive, consistent with our results. Of 1,019 MBC samples (ER+ and ER–), 48 samples had an ERBB2 mutation only, 109 had an ESR1 mutation only, and one had both mutations (odds ratio = 0.17, P = 0.0269, one-sided Fisher’s exact test). Based on these findings, along with the higher overall incidence of acquired activating HER2 alterations in our cohort, we hypothesized that HER2 alterations are a mechanism of acquired resistance to ER-directed therapy in ER+ MBC.
To investigate whether acquired HER2 alterations directly confer resistance to ER-directed therapy, we acutely expressed all HER2 mutant proteins observed in the 12 patients in the ER+ HER2 breast cancer cell lines T47D and MCF7 through lentiviral transduction and examined the impact of acute infection on susceptibility to endocrine agents. These included estrogen deprivation (using charcoal-dextran-stripped serum medium), which recapitulates aromatase inhibitor treatment in vitro, tamoxifen, and the selective estrogen receptor degraders fulvestrant and GDC-0810 (ref. 32). As controls, we transduced cells with wild-type HER2 and the kinase-dead HER2 mutant p.Asp845Ala16.
T47D cells expressing alterations in the kinase domain (p.Leu755Ser, p.Val777Leu and p.Leu869Arg) or transmembrane domain (p.Ser653Cys) were strongly resistant to estrogen deprivation in vitro, conferring a level of resistance equivalent to the previously described resistance-associated ESR1 ligand-binding domain alteration p.Tyr537Ser (Fig. 3a). Wild-type HER2 conferred only modest resistance, and the kinase-dead p.Asp845Ala alteration16 did not alter sensitivity as compared to green fluorescent protein (GFP)-expressing cells. We obtained similar results in MCF7 cells (Supplementary Fig. 4a).
Although ESR1 ligand-binding domain mutations confer robust resistance to estrogen deprivation, they confer only partial resistance to tamoxifen and fulvestrant1,2,3,4. In contrast, HER2 kinase domain and transmembrane domain mutants were completely resistant to tamoxifen (Fig. 3b), fulvestrant (Fig. 3c) and GDC-0810 (Fig. 3d). Similar results were observed in MCF7 cells (Supplementary Fig. 4). Wild-type HER2 as well as extracellular domain and cytoplasmic domain mutants conferred intermediate resistance between that of ESR1 p.Tyr537Ser and the kinase domain and transmembrane domain mutants (Fig. 3b–d and Supplementary Figs. 4–6). Expression levels of HER2 were similar for all HER2 constructs (Fig. 3e,f). Expressing the p.[Gly727Ala]+[Val777Leu] kinase domain alteration (observed in patient 0300252) conferred resistance to all ER-directed therapies (Supplementary Fig. 7). All in vitro drug sensitivities were consistent with clinical resistance phenotypes in patients bearing the respective HER2 alterations (Fig. 1b). Similar findings were seen at low levels of mutant HER2 expression, by using HER2 mutants expressed under a tetracycline-responsive promoter in cells grown in low doses of doxycycline (Supplementary Fig. 8).
HER2 activates pro-survival signaling pathways in cells, including RAS/RAF/MAPK and PI3K/AKT. HER2 mutants were associated with hyperphosphorylation of both ERK and AKT under conditions of estrogen deprivation or inhibition (Fig. 3e,f). ESR1 p.Tyr537Ser did not hyperactivate MAPK or AKT signaling, whereas extracellular domain and cytoplasmic domain mutants had effects similar to wild-type HER2 (Supplementary Fig. 5e,f). HER2 mutant cells also had lower levels of ER than controls (Fig. 3e), consistent with prior studies in tamoxifen-resistant cells that also showed lower levels of ER33. Treatment with fulvestrant led to an additional decrease of ER in HER2 mutant cells, but did not affect AKT or ERK phosphorylation (Fig. 3f).
To investigate whether ER signaling is suppressed in HER2 mutant cells, we examined transcript levels of ESR1 and the ER targets progesterone receptor (PGR), GREB1 and TFF1 by quantitative PCR with reverse transcription (qRT-PCR). ESR1 transcript levels were significantly downregulated in HER2 mutant cells along with PGR and GREB1 (Fig. 3g), although TFF1 was upregulated. The downregulation of PGR was consistent with the phenotype observed in the metastatic biopsies from patients, where acquisition of activating HER2 alterations coincided with loss of progesterone receptor expression by immunohistochemistry (Figs. 1b and 2). The extracellular domain and cytoplasmic domain HER2 mutants suppressed ER, progesterone receptor and GREB1 only slightly (Supplementary Fig. 5g).
To further examine transcriptional changes associated with HER2 mutants, we performed RNA sequencing on cells expressing the four activating HER2 mutants as well as GFP, wild-type HER2, kinase-dead HER2, and ESR1 p.Tyr537Ser, which were all treated with either dimethylsulfoxide (DMSO) or fulvestrant (Fig. 4a). Principal component analysis of the transcriptomes showed that the four HER2 activating mutants clustered together and remained separate from a cluster containing cells expressing ESR1 p.Tyr573Ser, as well as from a cluster containing cells expressing GFP and kinase-dead HER2, and one containing wild-type HER2 (Fig. 4b). Under treatment with fulvestrant, 5,293 genes were significantly differentially expressed (q value = 0.01) between cells expressing the activating HER2 mutants and GFP (Fig. 4c and Supplementary Tables 7 and 8). We next defined a shared HER2-MUT expression signature across the four activating HER2 mutants. Gene set enrichment analysis34 showed that the shared signature is enriched for ERBB1 and ERBB2 signaling and RAS/MAPK signaling compared to GFP and wild-type HER2 cells treated with fulvestrant or DMSO (Fig. 4d, panels 1 and 2; Supplementary Tables 9–12).
Canonical ER targets were among both the upregulated and downregulated HER2-MUT genes (Fig. 4d, panel 3), consistent with our quantitative RT-PCR of PGR, GREB1 and TFF1 (Fig. 3g). We hypothesized that HER2 mutants may lead to reprogrammed ER signaling, similar to the previously described growth factor–induced ER cistrome35. Indeed, the HER2-MUT signature was significantly enriched for induction of growth factor–induced ER targets in HER2 mutant cells (Fig. 4d, panel 4), along with a concomitant suppression of targets induced by estradiol, as previously observed in tamoxifen-resistant MCF7 cells33. Individually, all four HER2 mutants demonstrated the elevated RAS/MAPK transcriptional signature (Fig. 4e) as well as the ERBB1 and ERBB2 signatures and growth factor–driven ER signatures (Supplementary Fig. 9).
HER2 kinase domain alterations activate HER2 signaling through conformational changes to the catalytic domain that destabilize the inactive conformation of the kinase14 (Fig. 5a). The p.Gly727Ala and p.Val777Leu co-alterations are in apposition as indicated, and probably modify kinase domain conformation and sensitivity to reversible kinase inhibitors, as shown for EGFR previously36. In contrast, the mechanism of activation of the p.Ser653Cys transmembrane domain alteration has not been reported (Fig. 5b). We hypothesized that the cysteine residues form disulfide bridges, leading to constitutive HER2 homodimerization. To test this, we performed immunoblotting under nonreducing conditions to look for intact HER2 dimers. HER2 p.Ser653Cys but none of the other mutants we tested showed higher-molecular-mass species, indicating reduction-sensitive dimers (Fig. 5c). Thus, p.Ser653Cys probably functions by constitutive dimerization, as has been observed for HER2 extracellular domain mutants p.Gly309Glu and p.Glu321Gly16 as well as other transmembrane domain mutants37,38 (Fig. 5d).
Neratinib, an irreversible kinase inhibitor with anti-pan-HER activity, inhibits HER2 p.Leu755Ser, p.Val777Leu, and p.Leu869Arg in vitro14,15 and has shown promise in monotherapy clinical trials in breast cancer patients with activating HER2 alterations39,40. Low doses of neratinib resensitized HER2 mutant cells to fulvestrant (Fig. 6a). These cells were also sensitive to neratinib monotherapy at a higher dose (Supplementary Fig. 10c). In all cases, combination with fulvestrant inhibited viability better than neratinib alone, suggesting that inhibition of HER2 mutants restored sensitivity to fulvestrant. In contrast, neratinib only partially resensitized ESR1 p.Tyr537Ser cells to fulvestrant (Fig. 6a,b), and high-dose neratinib monotherapy was also ineffective in these cells (Supplementary Fig. 10a,c and Supplementary Note).
Neratinib treatment of HER2 mutant cells decreased both ERK and AKT phosphorylation (Fig. 6b and Supplementary Fig. 10b). In contrast, pharmacologic inhibition of individual downstream effectors MEK, ERK, PI3K, AKT and MTOR, alone or in combination with fulvestrant, did not restore sensitivity to fulvestrant in HER2 mutant cells (Supplementary Fig. 10e,f).
Combination treatment with endocrine therapy and CDK4 and CDK6 inhibitors is currently a standard-of-care treatment for ER+ MBC. HER2 mutant cells were cross-resistant to the CDK4 and CDK6 inhibitor palbociclib, both alone and in combination with fulvestrant (Fig. 6c and Supplementary Fig. 10d). Consistent with this, RNA sequencing (RNA-seq) analysis showed that HER2 mutant cells treated with palbociclib, either as single agent or with fulvestrant, grouped with DMSO- and fulvestrant-treated cells; cells treated with neratinib, with or without fulvestrant, grouped separately (Fig. 6d). RAS/MAPK transcriptional activity remained robust in the presence of palbociclib monotherapy or combination therapy (Fig. 6e). However, neratinib treatment, both as a single agent and with fulvestrant, led to repression of RAS/MAPK transcriptional activity (Fig. 6e) and other transcriptional effects of HER2 mutants (Supplementary Fig. 11). This was borne out clinically in one patient in the cohort with an acquired HER2 p.Val777Leu alteration after developing disease progression while on treatment with tamoxifen, who had intrinsic resistance to subsequent letrozole and palbociclib therapy (Fig. 6f). The discovery of the acquired HER2 alteration in the metastatic biopsy prompted enrollment of the patient into a phase 2 trial of fulvestrant plus neratinib41, resulting in a partial response lasting 6 months, consistent with our in vitro findings (Fig. 6a).
Taken together, our results suggest that activating HER2 alterations are a distinct mechanism of acquired resistance to multiple forms of ER-directed therapy in MBC that can be overcome by an irreversible HER2 inhibitor. Our observations also suggest that, with the increasing clinical use of selective estrogen receptor degraders, which can overcome ESR1 mutations, the prevalence of HER2 alterations might increase. The acquisition of targetable activating alterations in the metastatic setting highlights the importance of serial profiling of metastatic tumor biopsies or cell-free DNA from blood at the time of resistance in people with ER+ MBC. Identification of HER2 alterations in real time may help identify patients who will not benefit from ER-directed therapies or should be directed to clinical trials testing strategies to overcome this mechanism of resistance. One such strategy, the combination of fulvestrant and neratinib, is being tested in phase 2 clinical trials42 with promising preliminary results reported in individual patients15 including within our cohort (Fig. 6f). Ultimately, the use of upfront combinations to preempt the emergence of HER2-mutant-resistant clones may lead to more durable responses in people with ER+ MBC.
Gene Expression Omnibus (GEO), https://www.ncbi.nlm.nih.gov/geo/; Database of Genotypes and Phenotypes (dbGAP), https://www.ncbi.nlm.nih.gov/gap; Picard, http://picard.sourceforge.net/; Firehose, http://www.broadinstitute.org/cancer/cga/Firehose; Indelocator, http://www.broadinstitute.org/cancer/cga/indelocator; MuTect2, https://software.broadinstitute.org/gatk/documentation/tooldocs/current/org_broadinstitute_gatk_tools_walkers_cancer_m2_MuTect2; Novoalign, www.novocraft.com/products/novoalign/; ReCapSeg, http://gatkforums.broadinstitute.org/categories/recapseg-documentation; Oncotator, http://www.broadinstitute.org/cancer/cga/oncotator; CCF and evolutionary analysis, https://software.broadinstitute.org/cancer/cga/absolute.
Patient and tumor samples
Before any study procedures, all patients provided written informed consent for research biopsies and whole-exome sequencing of tumor and normal DNA, as approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board (DF/HCC Protocol 05-246). Metastatic core biopsies were obtained from patients and samples were immediately snap frozen in optimal cutting temperature compound (OCT) and stored at −80 °C. Archival formalin-fixed, paraffin-embedded (FFPE) blocks of primary tumor samples were also obtained. A blood sample was obtained during the course of treatment, and whole blood was stored at −80 °C until DNA extraction was performed.
DNA was extracted from primary tumors, metastatic tumors and normal samples for all patients and whole-exome sequencing was performed as detailed below.
DNA extraction was performed as described51. For whole blood, DNA was extracted using magnetic bead–based chemistry in conjunction with the Chemagic MSM I instrument manufactured by Perkin Elmer. After red blood cell lysis, magnetic beads bound to the DNA and were removed from solution using electromagnetized rods. Several wash steps followed to eliminate cell debris and protein residue from DNA bound to the magnetic beads. DNA was then eluted in Tris-EDTA buffer. For frozen tumor tissue, DNA and RNA were extracted simultaneously from a single frozen tissue or cell pellet sample using the AllPrep DNA/RNA kit (Qiagen). For FFPE tumor tissues, DNA and RNA were extracted simultaneously using Qiagen’s AllPrep DNA/RNA FFPE kit. All DNA was quantified using Picogreen.
DNA libraries for massively parallel sequencing were generated as described51 with the following modifications: the initial genomic DNA input into the shearing step was reduced from 3 µg to 10–100 ng in 50 µl of solution. For adapter ligation, Illumina paired-end adapters were replaced with palindromic forked adapters (purchased from Integrated DNA Technologies) with unique dual indexed 8-base index molecular barcode sequences included in the adapter sequence to facilitate downstream pooling. With the exception of the palindromic forked adapters, all reagents used for end repair, A-base addition, adapter ligation and library enrichment PCR were purchased from KAPA Biosciences in 96-reaction kits. In addition, during the post-enrichment solid-phase reversible immobilization (SPRI) bead cleanup, elution volume was reduced to 30 µl to maximize library concentration, and a vortexing step was added to maximize the amount of template eluted.
Solution-phase hybrid selection
After library construction, hybridization and capture were performed using the relevant components of Illumina’s Rapid Capture Exome Kit and according to the manufacturer’s suggested protocol, with the following exceptions: first, all libraries within a library construction plate were pooled before hybridization. Second, the Midi plate from Illumina’s Rapid Capture Exome kit was replaced with a skirted PCR plate to facilitate automation. All hybridization and capture steps were automated on the Agilent Bravo liquid handling system.
Preparation of libraries for cluster amplification and sequencing
After post-capture enrichment, library pools were quantified using quantitative PCR (qPCR, KAPA Biosystems) with probes specific to the ends of the adapters; this assay was automated using Agilent’s Bravo liquid handling platform. Based on qPCR quantification, libraries were normalized and denatured using 0.1 N NaOH on the Hamilton Starlet.
Cluster amplification and sequencing
Cluster amplification of denatured templates was performed according to the manufacturer’s protocol (Illumina) using HiSeq 2500 Rapid Run v1/v2, HiSeq 2500 High Output v4 or HiSeq 4000 v1 cluster chemistry and HiSeq 2500 (Rapid or High Output) or HiSeq 4000 flowcells. Flowcells were sequenced on HiSeq 2500 using v1 (Rapid Run flowcells) or v4 (High Output flowcells) Sequencing-by-Synthesis chemistry or v1 Sequencing-by-Synthesis chemistry for HiSeq 4000 flowcells. The flowcells were then analyzed using RTA v.1.18.64 or later. Each pool of whole-exome libraries was run on paired 76 base pair runs, with two 8-base index sequencing reads to identify molecular indices, across the number of lanes needed to meet coverage for all libraries in the pool.
Sequence data processing
Exome sequence data processing was performed using established analytical pipelines at the Broad Institute. A BAM file was produced with the Picard pipeline (see URLs), which aligns the tumor and normal sequences to the hg19 human genome build using Illumina sequencing reads. The BAM was uploaded into the Firehose pipeline (see URLs), which manages input and output files to be executed by GenePattern52.
Sequencing quality control
Quality control modules within Firehose were applied to all sequencing data for comparison of the origin for tumor and normal genotypes and to assess fingerprinting concordance. Cross-contamination of samples was estimated using ContEst53.
Somatic alteration assessment
MuTect54 was applied to identify somatic single-nucleotide variants. Indelocator (see URLs), Strelka55, and MuTect2 (see URLs) were applied to identify small insertions or deletions. A voting scheme was used for inferred indels that required scoring by at least two of three algorithms.
Artifacts introduced by DNA oxidation (so-called OxoG) during sequencing were computationally removed using a filter-based method56. In the analysis of primary tumors that are FFPE samples we further applied a filter to remove FFPE-related artifacts57.
Reads around mutated sites were realigned with Novoalign (see URLs) to filter out false-positive results that are due to regions of low reliability in the reads alignment. At the last step, we filtered mutations that are present in a comprehensive WES panel of 8,334 normal samples (using the Agilent technology for WES capture), aiming to filter either germline sites or recurrent artifactual sites. We further used a smaller WES panel of 355 normal samples that are based on Illumina technology for WES capture, and another panel of 140 normal samples sequenced without our cohort18 to further capture possible batch-specific artifacts. Annotation of identified variants was done using Oncotator58 (see URLs).
Copy number and copy ratio analysis
To infer somatic copy number from WES, we used ReCapSeg (see URLs), calculating proportional coverage for each target region (that is, number of reads in the target divided by total number of reads) followed by segment normalization using the median coverage in a panel of normal samples. The resulting copy ratios were segmented using the circular binary segmentation algorithm59.
To infer allele-specific copy ratios, we mapped all germline heterozygous sites in the germline normal sample using GATK Haplotype Caller60 and then evaluated the read counts at the germline heterozygous sites to assess the copy profile of each homologous chromosome. The allele-specific copy profiles were segmented to produce allele-specific copy ratios.
Cancer cell fraction and evolutionary analysis
Analysis using ABSOLUTE
To properly compare single-nucleotide variants and indels in paired metastatic and primary samples, we considered the union of all mutations called in either of the two samples. We evaluated the reference and alternate reads in each patient’s primary and metastatic tumors, including mutations that were not initially called in one of the samples. These mutations in matched samples were used as input for ABSOLUTE61. The ABSOLUTE algorithm uses mutation-specific variant allele fractions together with the computed purity, ploidy and segment-specific allelic copy ratio to compute CCFs.
Analysis of clonal dynamics using PHYLOGIC
To evaluate the mutation clonality in the patient-matched primary and metastatic samples, we used PHYLOGIC clustering of the mutation-specific CCFs as described62,63. The CCF of indels may be systematically underestimated owing to counting of only a subset of the reads supporting the indel event in cases of longer indels. For example, CCF was underestimated in the CDH1, c.437_447delCTCCTGGCCTC indel in patient 300504 in both metastatic and primary samples, but upon re-evaluation of the BAM files and supporting reads we assigned this mutation to the shared truncal mutations (clonal in both primary and metastatic samples). Further details regarding CCF and evolutionary analysis are online (see URLs).
Selected cancer genes (mutational landscape of potential drivers)
The list of mutated genes (Fig. 2) include: (1) significantly mutated genes in breast cancer64 or significantly mutated genes in our metastatic cohort18, including ERBB2, PIK3CA, CDH1, PTEN, ATM, CDC42BPA, CTCF, ESR1, FOXP1, NF1 and SETDB1; (2) genes with plausible clinical impact57 including ETV5, CDK12, MITF, PDGFRA, RARA and TSC1; and (3) genes from Cancer Gene Census, COSMIC database v82 (ref. 65), including MYH9, CHD4, FAT4, POT1, RMI2, ARID2, CASC5, CHCHD7, EBF1, ERCC5, HIP1, IL6ST, KAT6A, KMT2A, LZTR1, REL, RHOA, SALL4 and WRN.
ER+ HER2 T47D (American Type Culture Collection (ATCC), HTB-133) and MCF7 cells (ATCC, HTB-22) were cultured in phenol red–free RPMI-1640 (Gibco, 11835-030) or phenol red–free MEMα (Gibco, 41061-029) supplemented with 10% FBS (Gemini, 100-106), respectively. HEK293T/17 (ATCC, CRL-11268) cells were cultured in DMEM (Gibco, 11995-065) supplemented with 10% FBS. For serum starvation conditions, the appropriate medium was supplemented with 10% charcoal-dextran-stripped serum (Gemini, 100-119).
Generation of HER2 mutant plasmids and cells
HER2 mutants were generated by site-directed mutagenesis of the ERBB2 open reading frame in a pDONR223 vector backbone, using the QuikChange II site-directed mutagenesis kit (Agilent Technologies, 200523). After sequencing, the mutant constructs were cloned into the pLX307 vector backbone or the pDEST40 vector using the gateway LR clonase kit (Life Technologies, 11791019) to facilitate generation of mutant-encoding lentivirus, as described below.
To generate HER2 mutant or control lentivirus, 293T cells were transfected to produce viral particles using FuGENE HD (Promega, E2311) with VSV-G and Δ8.91 envelope and packaging plasmids. After 72 h of incubation the supernatant was filtered through a 0.45-μm filter (Corning, 431225), tested for the presence of lentiviral particles using Lenti-x GoStix (Takara, 631244) and stored at –80 °C until transduction. Lentiviral transduction of T47D cells plated in 6-well dishes was accomplished by viral spinoculation of cells in medium containing 4 µg ml–1 polybrene (Santa Cruz, sc-134220). After overnight incubation, medium was replaced. The cells were selected in 1.5 µg ml–1 puromycin (Gibco, A11138-03) 48 h after infection for 2 d, after which medium was replaced with fresh medium containing puromycin, and selected for a further 2 d. After confirmation of complete selection in uninfected control wells, cells were trypsinized, counted and plated for viability, western blotting, qPCR or HER2 dimerization assays as described below.
In cases in which estrogen deprivation was required, cells were cultured in phenol red–free RPMI containing 10% charcoal-dextran-stripped FBS (Gemini, 100-119) after selection.
HER2 inducible cell lines
To generate HER2 inducible cell lines, pDONR223 constructs encoding various HER2 mutants or controls, as well as pDONR223-GFP and pDONR223-HER2 wild type, were cloned into the Gateway-compatible pLX403 vector, which drives transgene expression using a Tet-On system.
The pLX317-GFP and pLX317-HER2 controls and pDONR223-HER2 plasmid were obtained from the Broad Institute The RNAi Consortium (TRC) portal. The VSV-G envelope and Δ8.91 packaging plasmids, pLX307 Gateway cloning destination vector and pLX403 vector were gifts from Levi Garraway (Dana-Farber Cancer Institute). Lentivirus encoding all mutants and controls was generated as described above, followed by transfection into T47D cells and bulk selection with puromycin to generate a stable cell line in each case. The cells were treated with various concentrations of doxycycline (Clontech, 631311) ranging from 25 to 500 ng ml–1, for various time periods ranging from 3 h to 1 week, to determine the minimal dose and induction time of doxycycline required to stimulate HER2 mutant expression over endogenous levels. Transgene expression was seen as early as 3 h after induction at 25 ng ml–1 doxycycline.
T47D HER2 mutant cells were plated in RPMI with 10% charcoal dextran-stripped serum (CSS) at 1,000 cells per well of 96-well Viewplates (Perkin-Elmer, 6005181). After 2 d of serum starvation, cells were switched over to complete medium or treated with estradiol (Sigma-Aldrich, E2758) as appropriate, and treated with a range of doses of the corresponding drug. Cells were re-treated after 3 d. One week after the start of treatment, viability was determined using CellTiter-Glo as described below. The normalized data were plotted in GraphPad Prism, and analyzed by nonlinear regression curve fit using the log (inhibitor) versus response three-parameter model.
CellTiter-Glo viability assays
One week after the start of treatment, the medium in 96-well plates was replaced with 100 μl fresh medium per well and brought to room temperature. CellTiter-Glo (Promega, G7572) was added to each well (20 μl), lysed for 2 min at ~250 r.p.m. and then equilibrated at room temperature for 20 min. A plate reader was programmed to integrate luminescence for 500 ms per well. Background luminescence from medium-only wells was subtracted from all values, and raw values were normalized against untreated wells for each cell line.
Chemicals and antibodies
Chemicals used included iodoacetamide (Sigma-Aldrich, A3221-10V), fulvestrant (Sigma-Aldrich, I4409-25mg), neratinib (Selleck Chemicals, S2150), (Z)-4-hydroxyTamoxifen (Sigma-Aldrich, H7904-25mg), GDC-0810 (Medkoo, 206041), AZD5363 (Thermo Fisher Scientific, NC0488926), VX-11e (Selleck Chemicals, S7709), trametinib (GSK1120212, Selleck Chemicals, S2673), BYL719 (Selleck Chemicals, S2814), RAD001/everolimus (Selleck Chemicals, S1120) and palbociclib (Selleck Chemicals, S1116). Primary antibodies used included antibodies to p-HER2 (EMD Millipore, 06-229), HER2 (Cell Signaling Technology, clone 44E7, 2248S), ERα (Santa Cruz Biotechnology, clone HC-20, SC-543), p-ERK1 and p-ERK2 (CST, clone E10, 9106S), ERK1 and ERK2 (CST, clone 137F5, 4695S), p-AKT (CST, clone D9W9U, 12694S), AKT (CST, 9272S), CCND1 (CST, clone 92G2, 2978), GAPDH (SCB, clone V-18, SC-20357) and β-actin (Santa Cruz, 47778), as well as goat anti-rabbit (Pierce, 32260), rabbit anti-goat (Invitrogen, 81-1620) and goat anti-mouse (Novex, A16090) antibodies.
About 5.0 × 105 HER2 mutant or control cells were serum-starved for 2 d, then treated with 1 μM fulvestrant or 1 μM neratinib for 24 h. Whole-cell protein extracts were made by passive lysis with 50 μl of lysis buffer, comprised of RIPA buffer (Sigma-Aldrich, R0278-50ml), 1 mM DL-dithiothreitol (DTT, Sigma-Aldrich, D0632-5G), 1 mM phenylmethylsulfonyl fluoride (Sigma-Aldrich, P7626-5G), protease inhibitor (Sigma-Aldrich, P8340-5ml), and Halt phosphatase inhibitor (Thermo Fisher Scientific, 78428). Cell pellets were lysed on a rotator at 15 r.p.m. for 15 min at 4 °C; lysates were then cleared by centrifugation at 14,000g for 15 min at 4 °C.
Protein was quantified by bicinchoninic acid assay (BCA, Pierce BCA Protein Assay kit, 23225) according to the manufacturer’s microplate procedure. Samples were prepared using 30 μg of protein, Bolt LDS Sample Buffer (Life Technologies, B0007), and DTT, then heated to 70 °C for 10 min according to Novex’s NuPAGE Bis-Tris Gel sample preparation recommendations. Samples were run on a Bolt 4–12% Bis-Tris Plus gel (Thermo Fisher Scientific, NW04125BOX) at 200 V for 35 min in 1× Bolt MOPS SDS running buffer (Life Technologies, B0001). Protein was transferred to nitrocellulose membranes using the BioRAD transblot turbo according to the fast-transfer protocol in the Trans-Blot Turbo RTA Transfer kit. Membranes were blocked at room temperature for 1 h in 5% milk in Tris-buffered saline (TBS, Bio-Rad, 170-6435) with 0.1% Tween-20 (Sigma-Aldrich, P9416-100 ml). The membranes were incubated overnight at 4 °C with primary antibody at a 1:1,000 dilution in 1% milk in TBS-Tween (TBS-T). The following day the membranes were washed 3× for 5 min with TBS-T and incubated with secondary antibody at a 1:2,000 dilution in 1% milk in TBS-T for 1 h at room temperature. After washing 4 × 10 min with TBS-T, membranes were treated with Pierce ECL Plus Western Blotting Substrate (Life Technologies, 32132) for 5 min and exposed to autoradiography film (Fisher Scientific, NC9648989). Full scans of all western blots in main and supplementary figures are in Supplementary Data.
Quantitative PCR with reverse transcription
About 8.0 × 105 T47D HER2 mutant or control cells were serum-starved in RPMI-1640 supplemented with 10% CSS for 48 h. Cells were washed with PBS and trypsinized, and cell pellets were collected and washed once with PBS. RNA was harvested using a RNeasy Mini kit (Qiagen 74104) according to the supplied protocol. RNA (1 μg) was reverse transcribed to cDNA on a ProFlex PCR System thermocycler, according to the RNase inhibitor–free protocol in the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, 4368814) user manual.
The cDNA was diluted eight-fold in double-distilled H2O and mixed with PowerUp SYBR Green Master Mix (Applied Biosystems, A25742) and appropriate forward and reverse primer pairs specific for ESR1 and selected targets in a 20-μl reaction according to the manufacturers’ specifications. The quantification was performed on an Applied Biosystems 7300 Real-Time PCR System. Each sample was run in triplicate and the raw data were analyzed by the 2–ΔΔCt method and normalized to GFP.
Primers were ordered from Integrated DNA Technologies at 25-nm scale and standard purification. Primer sequences are in Supplementary Table 13.
RNA sequencing on HER2 mutant cell lines and controls
RNA sequencing experimental setup and treatment of cells
T47D cells were infected de novo with lentivirus encoding HER2 mutants and controls (GFP, wild-type HER2, kinase-dead HER2 and ESR1 p.Tyr537Ser) in the pLX307 plasmid as described above. Upon completion of selection with puromycin, the cells were plated in 96-well plates in RPMI supplemented with 10% CSS for 48 h. Cells were then treated with DMSO, 1 μM fulvestrant, 1 μM neratinib, 10 μM palbociclib, 1 μM fulvestrant + 1 μM neratinib, or 1 μM fulvestrant + 10 μM palbociclib for 241 μg h–1. The cells were washed 3× with ice-cold PBS and lysed with TCL buffer (Qiagen 157013305) containing 1% β-mercaptoethanol (Sigma, M3148), transferred to PCR plates (Qiagen, 951020401), sealed, spun down for 1 min at 3,000 r.p.m. and immediately frozen at −80 °C until RNA extraction. For each specific construct and drug condition we performed six biological replicates, for a total of 288 transcriptomes (8 constructs × 6 drugs × 6 replicates).
RNA sequencing protocol
Plates with cell lysates were thawed and purified with 2.2× RNAClean SPRI beads (Beckman Coulter Genomics). The RNA-captured beads were air dried and processed immediately for RNA secondary structure denaturation (72 °C for 3 min) and cDNA synthesis. We performed SMART-Seq2 as described66 with minor modifications in the reverse transcription step. We made a 15-µl reaction mix for each PCR and performed ten cycles for cDNA amplification. We used 0.2 ng cDNA of each population and one-eighth of the standard Illumina NexteraXT (Illumina FC-131-1096) reaction volume in both the tagmentation and PCR amplification steps. Uniquely indexed libraries were pooled and sequenced with NextSeq 500 high-output V2 75 cycle kits (Illumina FC-404-2005) and 38 × 38 paired-end reads on an Illumina NextSeq 500 instrument, aggregating three NextSeq runs.
Reads alignments, expression quantification and quality control
Reads were mapped to the human genome (hg19) with STAR aligner67 (version 2.5.2b) with default parameters against hg19 of the human genome. Transcriptome quality and expression quantification was conducted using RNA-SeQC68. Samples with <12,000 unique genes were removed from subsequent analysis, excluding 26 samples and retaining 262 profiles, with a minimum of three replicates for each activating HER2 mutant (median of six and lower quantile of five replicates). Samples passing quality control had a mean of 17,363 detected genes by at least three reads (s.d. = 755) and a mean of 7,060,935 uniquely mapped pairs of reads (s.d. = 1,191,466).
Principal component analysis
The transcripts per million measured expression per gene was log normalized. To correct for batch effects among the three 96-well plates, we used the ComBat algorithm with default parameters69. Principal component analysis was performed with the R implementations of the function prcomp as part of the stats package using the default parameters.
The clustering of transcriptional profiles across various drug conditions used genes with high dispersion among samples, as calculated by LogVMR (variance-to-mean ratio, VMR) as implemented in FindVariableGenes function, Seurat R package.
Differential expression analysis
Definition of HER2-MUT signature
The HER2-MUT signature was inferred from the list of differentially expressed genes (DEGs) by comparing the four activating HER2 alterations (p.Val777Leu, p.Ser653Cys, p.Leu869Arg and p.Leu755Ser) with a given control open reading frame under a given drug condition. Supplementary Tables 9 and 10 depict the HER2-MUT signature with DEG between the mutants and GFP under fulvestrant and DMSO, respectively. Supplementary Tables 11 and 12 depict the HER2-MUT signature with DEGs between the mutants and wild-type HER2 under fulvestrant and DMSO, respectively.
Transcriptional signatures analysis
We used fast gene set enrichment analysis34 with 50,000 permutations and gene set size limited to a minimum of 3 and a maximum of 1,000 genes, respectively. A total set of 5,095 gene sets was analyzed including the c2, c6 and hallmark collections from MSigDB46 augmented with a breast cancer gene set collection35,71,72. Significant gene sets, controlling for a false discovery rate (FDR) of 5%, are in Supplementary Tables 9 and 10 for enrichments of HER2 mutants versus GFP DEGs under treatment with fulvestrant and DMSO, respectively. Significant gene sets, controlling for a FDR of 5%, are in Supplementary Tables 11 and 12 for enrichments of HER2 mutants versus HER2 wild-type DEGs under treatment with fulvestrant and DMSO, respectively. The evaluation of signature strength for single samples was done using gene set variation analysis73 with default parameters using the log2-normalized, batch-corrected expression for each sample.
HER2 dimerization assay
The HER2 dimerization assay was performed as described16. Briefly, T47D HER2 mutant or control cells were cultured in phenol red–free RPMI-1640 with 10% FBS in six-well plates until nearly confluent (~3–4 d). Cells were washed twice with cold PBS containing 10 mM iodoacetamide (Sigma-Aldrich, A3221) while on ice. Cells were actively lysed with 200 μl of TGP buffer, comprising 50 mM Tris base (Sigma-Aldrich, T1503), 10% glycerol (Sigma-Aldrich, G5516), 1% Triton X-100 (Sigma-Aldrich, T8787), 10 mM iodoacetamide, 1 mM phosphatase inhibitor and 1 mM protease inhibitor, for 20 min at 4 °C. Lysates were cleared by centrifugation and quantified by BCA. The samples were prepared in LDS sample buffer without DTT, and 25 μg was loaded on gels as described in the “Western blotting” section. After transfer, membranes were probed with anti-HER2 and anti-GAPDH for loading control as described above.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Tumor and germline whole-exome sequencing data generated and analyzed for this study have been deposited in the access-controlled public repository dbGaP with accession code phs001285. RNA-seq data are available through GEO under accession GSE121411. Additional data generated in this study, including tumor exome analysis, are available within the paper and in the supplementary information.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank Q. Quartey and P. Ram for technical assistance, F. Luo, R. Jeselsohn, C. Strathdee, I. Leshchiner, D. Rosebrock, D. Livitz and G. Getz for technical advice, and B. Kaplan, H. Greulich, C. Strathdee, F. Luo, E. Goetz and L. Garraway for providing reagents. We thank C. Johannessen, M. Brown, M. Meyerson and R. Bose for helpful discussions and comments on the manuscript. We are grateful to all the patients who volunteered for our tumor biopsy protocol and generously provided the tissue analyzed in this study. This work was supported by the Department of Defense W81XWH-13-1-0032 (N.W.), AACR Landon Foundation 13-60-27-WAGL (N.W.), National Cancer Institute Breast Cancer SPORE at DF/HCC P50CA168504 (N.W.), Susan G. Komen CCR15333343 (N.W.), the V Foundation (N.W.), the Breast Cancer Alliance (N.W.), the Cancer Couch Foundation (N.W.), the MBC Collective (N.W.), Breast Cancer Research Foundation (N.U.L. and E.P.W.), ACT NOW (to Dana-Farber Cancer Institute Breast Oncology Program), Fashion Footwear Association of New York (to Dana-Farber Cancer Institute Breast Oncology Program), Friends of Dana-Farber Cancer Institute (to N.U.L.), the Klarman Family Foundation and HHMI (to A.R.) and Dana-Farber/Harvard Cancer Center SPORE grant P50CA168504.