Neratinib Efficacy and Circulating Tumor DNA Detection of HER2 Mutations in HER2 Nonamplified Metastatic Breast Cancer

This article is featured in Highlights of This Issue, p. 5659

HER2 (ERBB2) is a well-established therapeutic target in breast cancer (1–13), and HER2-negative (nonamplified) breast cancers overall do not benefit from HER2-directed drugs (14). However, the recent identification of recurrent HER2 mutations (HER2^mut) in a subset of HER2 gene nonamplified breast cancer suggested an additional HER2 targeting opportunity (3, 15–25). HER2 mutations cluster in the tyrosine kinase and extracellular dimerization domains of HER2, leading to enhanced kinase activity and tumorigenesis in preclinical models (16, 26, 27). Importantly HER2^mut render tumor cells sensitive to HER2-targeted agents, especially neratinib (16, 26, 27), a potent irreversible pan-HER inhibitor (28–34). We therefore conducted a phase II trial of neratinib in patients with HER2^mut, nonamplified metastatic breast cancer (MBC; Mutant HER2 trial: MutHER). The primary endpoint was clinical benefit rate (CBR). Secondary endpoints included progression-free survival (PFS), toxicity profile, HER2^mut frequency in MBC, and analysis of ctDNA for HER2^mut detection and response monitoring. The Clinicaltrials.gov# is NCT01670877.

Patients with HER2-negative (0 or 1+ by immunohistochemistry or nonamplified by FISH) MBC, at least 18 years old, Eastern Cooperative Oncology Group (ECOG) Performance Status (PS) ≤2, measurable or evaluable disease by RECIST 1.1, and adequate organ function were preregistered for HER2^mut screening by DNA sequencing of archival primary or metastatic tumor performed centrally at the Clinical Laboratory Improvement Amendment (CLIA)-certified Washington University Genomic and Pathology Service (GPS). Patients were eligible for screening while receiving other treatment. The initial requirement of at least one prior systemic therapy was subsequently removed to improve accrual. Local HER2 mutation testing in a local CLIA laboratory was also allowed.

Eligibility criteria for registration included tumor positive for somatic HER2^mut identified by CLIA-certified laboratories, recent disease progression, adequate organ function, QTc interval ≤450 msec (men) or ≤470 msec (women), and left ventricular ejection fraction ≥ institutional lower limit of normal, ≥1 week wash-out from radiotherapy or systemic therapy. Patients receiving other cancer therapy, strong CYP3A4 inducers or inhibitors, uncontrolled concurrent illness, ≥grade 2 diarrhea, being pregnant, or breastfeeding were not eligible. Treated brain metastases stable for ≥3 months without steroids were allowed. This study was conducted in accordance with the Declaration of Helsinki and the principles of Good Clinical Practice, and was approved by each center's Regulatory and Ethics Committees. All participants provided written informed consent.

The primary objective was rate of CB defined as complete/partial response (CR/PR) or stable disease (SD) ≥24 weeks. Secondary objectives included HER2^mut frequency and clinicopathologic characteristics of these patients, PFS on neratinib, toxicity profile, and analysis of ctDNA for HER2^mut detection and response monitoring.

Sample size was calculated based on Simon's Optimal two-stage design to enroll 10 patients in the first stage and 19 patients in the second stage to allow 80% power and a one-sided 0.05 significance level to detect an anticipated 20% CBR against the null hypothesis of 5%. At least one CB was required to proceed to the second stage. The primary endpoint is met if at least four of 29 achieved CB. After observing that almost all HER2^mut tumors were ER⁺, the protocol was amended to add fulvestrant to the regimen (if ER⁺). This pragmatic decision led to an early stop of enrollment to neratinib monotherapy. Here, we report the results from the 16 patients who received neratinib monotherapy prior to the activation of this amendment.

Patients were started with neratinib orally at 240 mg daily in a 28-day cycle. Diarrhea prophylaxis with loperamide was mandatory during the first cycle of therapy. Loperamide was administered at an initial dose of 4 mg with the first dose of neratinib on cycle 1 day 1, followed by 2 mg every 4 hours for 3 days, then 2 mg every 6 to 8 hours during the first cycle of therapy, and as needed. Subjects were allowed to escalate neratinib dose to 320 mg daily if no intolerable grade 2 or higher treatment-related adverse events (AE) were experienced during a complete cycle of treatment. A maximum of three dose reductions in 40-mg decrements were permitted for toxicities. AEs were accessed by NCI CTCAE 4.0 weekly during the first cycle and day 1 of subsequent cycles. Patients underwent tumor evaluation by RECIST 1.1 every two cycles and echocardiograms every four cycles.

Central tumor DNA sequencing for HER2 mutation was performed at the GPS laboratory. Sanger technology was initially used to sequence exons 8 and 18 to 24 of HER2, which applied to approximately 20% of samples by October 2014. Subsequently, a PCR based next-generation sequencing (NGS) assay was used to sequence all HER2 coding exons. PCR was performed using the 48.48 high-throughput access array system (Fluidigm Corp.). Cluster generation and sequencing were performed using Illumina's HiSeq2500 Reagent Kit (200 cycles), and 2 × 101 paired-end sequence reads were generated. Each patient's DNA was processed in three independent technical replicates to generate and sequence three independent amplicon libraries. Variant call was performed on all BAM files together by a combination of commercially available and custom-developed scripts to generate a multi-sample variant call file. Variants were called if at least three of the four BAM files had evidence for the variant, and the average variant frequency was ≥10%.

Digital sequencing of cell-free DNA was performed by Guardant Health, Inc. (Guardant360, www.guardanthealth.com/guardant360/), a CLIA-certified and College of American Pathologists (CAP)-accredited clinical laboratory. Note that 5 to 30 ng of ctDNA was isolated from plasma (two 10-mL Streck tubes drawn for each patient), and sequencing libraries were prepared with custom in-line barcode molecular tagging and complete sequencing at 15,000× read depth. The panel utilizes hybrid capture followed by NGS of all exons in 30 genes, including HER2, and critical exons (those reported as having a somatic mutation in COSMIC) of 40 additional genes to detect and report single-nucleotide variants and small indels in 70 genes, copy-number amplifications in 18 genes, and select fusions (Supplementary Fig. S1). Postsequencing bioinformatics matches the complementary strands of each barcoded DNA fragment to remove false-positive results (35). The variant allele fraction (VAF) was computed as the number of mutated DNA molecules divided by the total number (mutated plus wild type) of DNA fragments at that allele; VAF was reported as a percentage.

The CBR was calculated as the proportion of CR, PR, or SD≥24 weeks with 90% exact binomial confidence intervals (CIs) by RECIST 1.1. Response duration was defined as the duration between the first scan demonstrating disease response and that at progression. PFS was defined as weeks from treatment initiation to progression or death. The associations of HER2 mutations with histology subtype (lobular vs. ductal), hormone receptor status (ER⁺/PR⁺ vs. ER⁻/PR⁻), and the source of sample (primary vs. metastatic) were assessed using Fisher exact tests. The diagnostic ability of ctDNA sequencing for HER2 mutations was summarized by sensitivity and specificity, with corresponding 90% CI. The association between ctDNA-mutant VAF and clinical outcomes (PFS and tumor size change) was also assessed using Spearman correlation coefficients. The data were analyzed using the standard package of SAS (Version 9.3, SAS Institute).

Between December 1, 2013, and August 15, 2015, 636 women with HER2 nonamplified MBC, median age 56 (range, 23–87) years, were consented for central HER2 sequencing at Washington University CLIA-certified GPS laboratory (Supplementary Fig. S2). Among the 579 patients eligible for preregistration, 517 had tissue available for testing (Table 1). Adequate quantity and quality of tumor DNA were extracted and successfully sequenced for 381 of 517 (74%) samples (primary n = 256, metastasis n = 125) with nine of 381 (2.4%) samples positive for at least one HER2 mutation. Notably, all nine cases with HER2 mutations were also positive for ER and/or PR. The incidence of HER2 mutation was nine of 277 (3.2%) in hormone receptor–positive cases in contrast with the 0 of 82 (0%) in triple-negative breast cancers. However, the difference in HER2 mutation incidence by hormone receptor status did not reach statistical significance (P = 0.219), likely because of the small sample size. Lobular cancers had a significantly higher incidence of HER2 mutations, 7.8% (4/51), compared with ductal histology, 1.6% (5/309), (P = 0.026; Table 1). There was no difference in the detection rate of HER2 mutations in tumors from primary (2.3%) or metastatic sites (2.4%).

An additional 13 patients with HER2^mut were identified by an outside CLIA-lab (Supplementary Fig. S2). Figure 1 shows the distribution of the HER2 mutations identified in all 22 cases. L755S was the most common mutation (n = 7, 32%). The median patient age was 58 (range, 31–72) years, 15 of 22 (68%) were ductal, seven of 22 (32%) were lobular cancers, and 21 of 22 (95%) were ER⁺. Retrospectively, 21 of 22 patients presented initially with early-stage disease. The median recurrence-free survival from surgery to recurrence was 35.9 (90% CI, 15.4–93.8) months.

Sixteen patients received neratinib monotherapy, including 14 with known-activating mutations and two of unknown significance (no preclinical data; Fig. 1; Supplementary Fig. S2). Table 2 details the patient characteristics. As of data cutoff, all patients had stopped study therapy due to progression (n = 15, 94%) or AE (n = 1, 6%).

All 16 patients were evaluable for AE (Table 3). Treatment was well tolerated, and most AEs were grades 1 and 2, with one patient who discontinued therapy due to AE (grade 3 fatigue and dehydration). The most common grade 2 and above AEs included diarrhea (69%), anorexia (44%), and fatigue (31%), and there were no grade 4 AEs. Diarrhea (n = 4, 25%) was the only treatment-related grade 3 AE that occurred in more than one patient, but the duration was short, lasting a median of 1.5 days (range, 1–3 days). Neratinib was reduced to 200 mg daily in four (25%) patients due to grade 3 diarrhea (n = 2), grade 2 nausea/anorexia (n = 1), and grade 3 fatigue/dehydration leading to neratinib discontinuation (n = 1). Neratinib was escalated to 320 mg daily in five patients (three PD, two SD ≥24 weeks) in cycles two (n = 4) and four (n = 1) and continued until disease progression, except in one patient who subsequently reduced to 240 mg 10 days after due to grade 2 nausea/anorexia and diarrhea.

One of the 10 patients enrolled in the first stage achieved CB; therefore, the study continued to the second stage until the activation of the protocol amendment for subsequent patients to receive the combination of fulvestrant and neratinib if ER⁺. In total, 16 patients received neratinib monotherapy. The median number of metastatic regimens received prior to study entry was three (range, 2–10). Five patients experienced CB including one CR (6%), one PR (6%), and three SD ≥24 weeks (19%). The CBR was 31% (90% CI, 13%–55%), which met the primary endpoint. The median response duration in patients who achieved CB was 24 (range, 24–66) weeks. PFS and changes in tumor size were shown in Figure 2A and B, respectively. An example of response is illustrated in Figure 2C. Among the 15 patients with ER⁺ breast cancer, only one (pt 16) received a CDK4/6 inhibitor (ribociclib) for metastatic disease prior to study enrollment. This patient experienced prolonged disease stabilization on neratinib (PFS 37 weeks). Treatment benefit was observed across different HER2 mutations including L755S, V777_G778 insGSP, P780_Y781 insGSP, V842I, S310F, and L869R. Multi-gene panel next-generation tumor sequencing results were available for 13 patients: common co-occurring mutations included CDH1 (5/13), PIK3CA (4/13), and TP53 (6/13).

ctDNA sequencing for HER2^mut detection and response monitoring was performed using a 70-gene digital sequencing assay (Guardant Health; Supplementary Fig. S1). Plasma collected from patients with HER2-activating mutations was analyzed at baseline as positive controls (n = 14) and additionally tested following 4 weeks on neratinib (n = 13) and upon progression (n = 9). Because a sample size of at least 30 negative controls was required to ensure 90% confidence of >90% testing specificity, plasma collected at preregistration for 40 patients negative for HER2^mut by tumor sequencing at GPS was analyzed.

Among the 14 positive control cases, two were negative for any HER2^mut by ctDNA sequencing despite the detection of other genetic alterations in the same sample, and 11 were positive for the same HER2^mut as by tumor testing (sensitivity: 11/14; 79%, 90% CI, 53%–94%; Supplementary Table S1). One other patient (Pt 14), SD ≥ 24 weeks on neratinib, had discrepant HER2^mut alleles by tumor and ctDNA sequencing. ctDNA sequencing identified HER2 L869R and D769Y at baseline, in contrast to the S310F and V842I detected in the breast cancer specimen collected from several years prior to trial enrollment. Interestingly, the ctDNA VAFs for both HER2 L869R and D769Y were reduced at week 4, and increased at progression, accompanied by the emergence of several other HER2 mutations, including the T798I mutation in the kinase domain of HER2 analogous to the EGFR T790M “gate-keeper” drug resistance mutation (Fig. 3A; refs. 36–38).

Among the 40 negative control cases, 32 were informative for ctDNA sequencing interpretation (eight had no detectable ctDNA mutations), and all 32 were negative for HER2^mut despite the detection of other somatic mutations in each case. Therefore, the specificity of ctDNA for HER2^mut detection was 32 of 32 (100%; 90% CI, 91%–100%; Supplementary Table S1). The overall concordance rate between plasma and tumor HER2 sequencing results was 43 of 46 (93.5%; 90% CI, 87%–100%; Supplementary Tables S1 and S2).

We further queried the ctDNA sequencing data from 1,834 advanced breast cancer patients clinically tested at Guardant Health between October 2015 and August 2016 using the same assay. HER2-activating mutations in the absence of HER2 amplification were identified in 48 of 1,584 (3.0%) evaluable patients. The incidence and distribution (Supplementary Fig. S3) were similar to prior tumor-based analyses (16).

Among the 11 patients (four PD and seven non-PD as best response by RECIST) with baseline HER2 mutation detected by ctDNA who also had subsequent blood collections, HER2^mut VAF decreased at week 4 in nine patients, including all seven patients with non-PD as their best response. The HER2^mut VAF at week 4 was reduced to nondetectable in the patient with CR (Pt 15). Although two patients with PD as the best response (Pts 1 and 12) demonstrated a decrease in HER2^mut VAFs at week 4, their HER2^mut VAFs subsequently increased at week 8, when progression was radiographically detected. In contrast, the other two of four patients with PD as the best response (Pts 7 and 11) had a rise in HER2^mut VAFs at week 4, consistent with early progression (Fig. 2D and 3; Supplementary Table S2). The absolute levels of HER2^mut VAFs at week 4 were significantly associated with PFS (Spearman correlation coefficient rho = –0.69, P = 0.017, n = 11) and tumor size change (rho = 0.67, P = 0.05, n = 9). All patients had a rise in HER2^mut VAFs upon disease progression.

A number of genetic alterations were also identified in ctDNA sequencing, which co-occur with HER2 mutations (Supplementary Table S3), with the VAFs of HER2 mutations among the highest in each case. The VAFs of these co-occurring genetic alterations trended similarly to that of HER2 mutations (Fig. 3). These data further support HER2 mutations being a driver event in HER2^mut breast cancers which can be treated with neratinib.

HER2-targeted agents have markedly improved outcomes for patients with HER2-positive breast cancer (2, 5–9). In this phase II study, single-agent neratinib, an irreversible pan-HER inhibitor, demonstrated a CBR of 31% (90% CI, 13%–55%), in a heavily pretreated patient population with HER2-mutated nonamplified MBC. Neratinib was well tolerated with limited duration of grade 3 diarrhea (median = 1.5 days) and only one patient discontinuing therapy due to AE. The incidence of HER2-activating mutations in primary breast cancer approximates 2% to 3% overall, with an approximately threefold higher incidence in lobular cancers (3, 15–17). Therefore, approximately 5,000 to 7,500 (2%–3% of 250,000; ref. 39) new breast cancers each year in the United States likely have a HER2 mutation at diagnosis. The overall prevalence of HER2 mutation in MBC can only be estimated, but in the United States, at least 200,000 patients are likely to be living with advanced disease today based on a relapse risk of 30% and median survival of 3 years (40). Based on the incidence of 2% to 3%, 4,000 to 6,000 patients have a HER2-mutant metastatic tumor. This prevalence is sufficient to conduct randomized trials to establish a standard of care if HER2 mutation screening were widely available.

This study demonstrates the feasibility of targeting a rare breast cancer patient population with a multi-institutional, collaborative effort. The logistics of large-scale screening efforts were the major challenge, hampered by insufficient tumor tissue in approximately one-third of patients. ctDNA-based HER2 sequencing provides a noninvasive, highly specific and sufficiently sensitive alternative to invasive approaches to be recommended in a clinical trial setting to assist in patient identification.

This study was proof of concept and had a small sample size. However, the investigation met the predefined primary endpoint and provided strong evidence that HER2 mutation is a valid therapeutic target. Although there have been several single case reports of patients with HER2 mutated, nonamplified MBC responding to HER2-targeted agents, this is the first report of a formal study addressing this question (22, 25). Although the CBR of 31% (90% CI, 13%–55%) is modest, it is significant in a heavily pretreated patient population that received a median of 3 (range, 2–10) prior treatment regimens in the metastatic setting. As an example, single-agent palbociclib showed a CBR of 19% overall, and 29% in the HR⁺ HER2⁻ subset, in patients with Rb+ advanced breast cancer who had progressed through at least two prior lines of hormonal therapy (41).

In summary, our data indicate efficacy of neratinib for HER2-mutated nonamplified breast cancer and provide a rationale for a large-scale ctDNA-based screening program to identify a sufficient number of patients with HER2-mutant breast cancer to establish a new standard of care. The universal, and in many cases, quite rapid development of acquired resistance despite a clear initial molecular response by HER2 mutation VAF in ctDNA indicates that rational combination therapy is the next step. The strong association between ER positivity and HER2-activating mutations demands dual targeting of ER and HER2 pathways based on the success of combining ER and HER2-targeted agents in HER2-amplified disease (42–44). This phase II trial has therefore been amended to investigate the combination of fulvestrant and neratinib in subsequent patients with ER-positive HER2-mutated nonamplified breast cancer and to allow ctDNA determination of HER2 mutation status as an eligibility criterion (NCT01670877). Other ongoing studies that target HER2-mutated advanced solid tumors include the SUMMIT neratinib basket study (NCT01953926), the NCI-MATCH trial with afatinib (NCT02465060), and My Pathway trial with trastuzumab plus pertuzumab (NCT02091141).

C.X. Ma reports receiving commercial research grants from and is a consultant/advisory board member for Novartis, Pfizer, and Puma Biotechnology. R. Bose is a consultant/advisory board member for Genetech. G. Kimmick reports receiving commercial research grants from BioNovo, Bristol-Myers Squibb, GlaxoSmithKlein, Novartis, and Puma Biotechnology. E. Winer is a consultant/advisory board member for Genetech, LEAP, and Tesaro. M. Naughton reports receiving speakers bureau honoraria from Genetech and Novartis. D. Tripathy is a consultant/advisory board member for Puma Biotechnology. M. Cobleigh is a consultant/advisory board member for Puma Biotechnology. C. Anders is a consultant/advisory board member for Angiochem, Eli Lilly, Genentech, Geron, Kadmon, Merrimack, Nektar, Novartis, Sanofi, and to BBB. K.C. Banks holds ownership interest (including patents) in Guardant Health. R.B. Lanman holds ownership interest (including patents) in Guardant Health. R. Bryce and A. Lalani are employees of PUMA Biotechnology. D.F. Hayes reports receiving commercial research grants from Puma Biotechnology, reports receiving other commercial research support from AstraZeneca, Janssen, and Pfizer, and holds ownership interest (including patents) in OncImmune. K. Blackwell is a consultant/advisory board member for Puma Biotechnology. M.J.C. Ellis is an employee of and holds ownership interest (including patents) in Bioclassifier LLC, and is a consultant/advisory board member of AstraZeneca, Novartis, Pfizer, and Puma Technology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.X. Ma, R. Bose, E. Winer, M. Naughton, R. Bryce, K. Blackwell, M.J.C. Ellis

Development of methodology: C.X. Ma, J. Pfeifer, M. Pegram, H. Al-Kateb, M.J.C. Ellis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.X. Ma, R.A. Freedman, M.L. Telli, G. Kimmick, M. Naughton, M. Goetz, C. Russell, D. Tripathy, M. Cobleigh, A. Forero, T.J. Pluard, C. Anders, P.A. Niravath, C. Bumb, R.B. Lanman, J. Pfeifer, D.F. Hayes, M. Pegram, K. Blackwell, P.L. Bedard, M.J.C. Ellis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.X. Ma, R. Bose, F. Gao, R.A. Freedman, G. Kimmick, M. Cobleigh, A. Forero, K.C. Banks, R.B. Lanman, M. Pegram, K. Blackwell, P.L. Bedard, H. Al-Kateb, M.J.C. Ellis

Writing, review, and/or revision of the manuscript: C.X. Ma, R. Bose, F. Gao, R.A. Freedman, M.L. Telli, G. Kimmick, E. Winer, M. Naughton, M. Goetz, C. Russell, D. Tripathy, M. Cobleigh, A. Forero, T.J. Pluard, C. Anders, P.A. Niravath, S. Thomas, J. Anderson, K.C. Banks, R.B. Lanman, R. Bryce, A.S. Lalani, J. Pfeifer, D.F. Hayes, M. Pegram, K. Blackwell, P.L. Bedard, H. Al-Kateb, M.J.C. Ellis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.X. Ma, S. Thomas, J. Anderson, K.C. Banks, M.J.C. Ellis

Study supervision: C.X. Ma, D. Tripathy, A. Forero, K. Blackwell, M.J.C. Ellis

We thank patients and their families for participation in this study. We thank physicians, nurses, research, and regulatory coordinators for their work, PUMA Biotechnology for trial support, and Army of Women for patient referral. We acknowledge Stephanie Myles for assisting protocol development and Zach Skidmore for graphing Supplementary Fig. S3.

This work is funded in part by Siteman Cancer Center, The Foundation for Barnes-Jewish Hospital Cancer Frontier Team Science Award (C.X. Ma, R. Bose, M.J.C. Ellis, and J. Pfeifer), NCI Cancer Clinical Investigator Team Leadership Award (C.X. Ma), and Puma Biotechnology Inc. M.J.C. Ellis is a McNair Medical Foundation Scholar and a Cancer Prevention Institute of Texas established investigator.

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.