HER2-Mediated Internalization of Cytotoxic Agents in ERBB2 Amplified or Mutant Lung Cancers Free

Each year, thousands of patients with cancers bearing amplifications or mutations in the human epidermal growth factor receptor 2 (ERBB2, encoding HER2) are without effective treatments. HER2 is a transmembrane receptor tyrosine kinase (RTK), and ERBB2 is amplified/overexpressed in 15% to 20% of breast and gastroesophageal cancers, for which targeted anti-HER2 therapy has transformed the standard of care (1–3). In addition to gene amplification, HER2 is also hyperactivated by specific mutations, most of which are located either in its extracellular domain or in its kinase domain. The common denominator of both overexpression and mutation is receptor hyperactivation via an increased homo- or heterodimerization and activation of downstream signaling cascades such as the PI3K and MAPK pathways (4). Mutations in or amplification of ERBB2 are also found in approximately 4% of lung tumors (5, 6), for which there is currently no approved targeted therapy. HER2-targeted agents investigated in patients with lung cancers over the past two decades include the mAbs trastuzumab and pertuzumab, and tyrosine kinase inhibitors such as afatinib, dacomitinib, and neratinib (7, 8). Although clinical activity was seen, neither class of drugs has produced transformative results in clinical trials to date (9–13). Antibody–drug conjugates (ADC) are emerging antitumoral agents that combine the unique binding capabilities of mAbs with the cytotoxic activity of chemotherapy to specifically target and harm cancer cells (14). The mechanism of action of ADCs involves the recognition and binding of the mAb backbone to the extracellular domain of a cancer-specific transmembrane protein, the internalization of the ADC–protein complex via the endocytic pathway, and the release of the cytotoxic payload that ultimately kills or arrests the proliferation of the malignant cell (14). To date, there are five FDA-approved ADCs; three are recommended for hematologic malignancies, and two, ado-trastuzumab emtansine (T-DM1, Genentech/Roche) and trastuzumab deruxtecan (T-DXd, formerly DS-8201a, Daiichi-Sankyo/AstraZeneca; ref. 15), are indicated for HER2-positive breast cancer (16), with the latter also showing promising results in patients with HER2-positive gastric cancer (17, 18).

In this work, we demonstrate that activated HER2, regardless of its oncogenic potential or the addiction status of the cancer cell to its downstream signaling pathways, can serve as a vehicle to funnel potent chemotherapeutic agents into lung tumors. Integrating parallel laboratory and clinical data on the mechanisms of response, we also propose two different strategies to improve the efficacy of anti-HER2 ADCs: cotreatment with irreversible pan-HER inhibitors that enhance receptor ubiquitination and internalization of ADCs or switching to an ADC bearing a different cytotoxic payload.

We hypothesized that ERBB2-amplified or -mutated tumors have exquisite susceptibility to HER2 ADCs on the basis of high-rate receptor internalization and trafficking, regardless of their intrinsic dependence on HER2 signaling for cell growth and/or survival. To test whether the presence of an activating mutation influences HER2 internalization rate, affecting in turn the internalization rate of ADC–HER2 complexes, we established isogenic models using a nontransformed breast cell line (MCF10A) and a lung cancer cell line (NCI-H2030) expressing either V5-tagged wild-type (WT) or mutant (S310F or L755S) HER2 or empty vector (EV) as control. We used these models to quantitate the internalization rate of T-DM1 linked to a pH-sensitive dye (pHrodo-T-DM1) that becomes fluorescent only at low pH, providing a positive indication of ADC–HER2 endocytosis. Albeit with some variations in the kinetics, we consistently observed that cells expressing either S310F or L755S mutants internalize more T-DM1 than WT cells in both cell models (Fig. 1A–D). In particular, MCF10A cells expressing the HER2^L755S mutant show the highest pHrodo-T-DM1 signal at earlier time points and morphologic changes at later time points (Fig. 1A and B), compatible with the onset of cell death, as confirmed by increased PARP cleavage (Fig. 1E). In both cell lines, expressing comparable WT or mutant HER2, receptor levels were reduced upon T-DM1 treatment, likely due to internalization and subsequent degradation (Fig. 1E and F).

On the basis of these data, we posited that T-DM1 can accumulate in both mutant and amplified lung tumors. To evaluate this in the clinical setting, we performed ⁸⁹Zr-trastuzumab PET/CT functional imaging in patients with lung cancer bearing HER2 alterations. We found that ⁸⁹Zr-trastuzumab can accumulate in both ERBB2-amplified and ERBB2-mutant tumors (Fig. 1G, H).

We then tested the antitumoral activity of T-DM1 in different patient-derived xenografts (PDX) established from patients with lung cancer with tumors bearing ERBB2 mutations or amplification. Treatment with T-DM1 resulted in durable responses in both HER2-mutant (YVMA insertion) PDXs and in xenografts with co-occurrent amplification (2.8-fold) and mutation (S310F) of ERBB2 (Fig. 1I and J). This treatment was well tolerated with no significant change in animal body weight (Supplementary Fig. S1A and S1B).

The above data provided the rationale for a dedicated analysis of the clinical activity of T-DM1 in patients with ERBB2-amplified or -mutant lung cancers within the context of a histology-agnostic phase II basket trial (NCT02675829; Supplementary Fig. S2). In total, 49 patients (18/49 patients were part of the previously reported ERBB2-mutant cohort of this clinical trial; ref. 19) with ERBB2-mutant and/or -amplified lung cancer (Table 1) were treated with T-DM1, and responses were measured by RECIST version 1.1 and/or modified PET Response Criteria in Solid Tumors (PERCIST). Modified PERCIST, recently used to assess antitumor activity of new cancer therapies in clinical trials (13, 20, 21), was employed to capture RECIST nonmeasurable disease. The best overall response rate (ORR) by either RECIST or modified PERCIST for ERBB2-amplified/mutant patients was 51% [25/49; 95% confidence interval (CI) 36–66; Fig. 2A and B], with a median progression-free survival (PFS) of 5 months (95% CI, 3.5–5.9; Fig. 2C). In the subset of ERBB2-mutant and/or -amplified patients with RECIST evaluable disease, the ORR by RECIST was 31% (13/42; 95% CI, 18–47; Supplementary Fig. S3A and S3B), whereas in the subset of ERBB2-mutant and/or -amplified patients with PERCIST evaluable disease, the ORR by modified PERCIST was 64% (14/22; 95% CI, 41–83; Supplementary Fig. S3C and S3D). Except for 1 case (2%) of grade 3 febrile neutropenia, treatment was well tolerated (Table 2).

In accordance with our preclinical data (Fig. 1), we did not observe appreciable differences in response when we stratified patients according to ERBB2 genetic status: The best ORR by either RECIST or modified PERCIST was 55% (6/11; 95% CI, 23%–83%) for ERBB2-amplified patients, 50% (14/28; 95% CI, 31–69) for ERBB2-mutant patients and 50% (5/10; 95% CI, 19–81) for concurrently ERBB2-mutant and -amplified patients (Fig. 2A). These data strongly suggest that a response to T-DM1 can be achieved to the same extent in lung cancers with either amplified or mutant ERBB2.

ERBB2 copy number obtained by next-generation sequencing (NGS) using the Fraction and Allele-Specific Copy Number Estimates from Tumor Sequencing (FACETS) algorithm (22) correlated with HER2 status as assessed by FISH and IHC (Supplementary Fig. S3E–S3G; Supplementary Table S1). A trend of association between the degree of ERBB2 amplification by FACETS with patient response to T-DM1 therapy was observed, albeit not reaching statistical significance (Supplementary Fig. S3H). There were 6 patients with ERBB2-amplified lung cancer and concurrent EGFR mutation who had progressed on a prior EGFR inhibitor, and 2 of these patients responded to T-DM1 (Supplementary Table S1).

Despite the promising clinical activity observed with T-DM1 in ERBB2-mutant/amplified lung cancers, some patients' tumors were intrinsically refractory to T-DM1 and some responses were short-lived. We then sought strategies to enhance or prolong the duration of the responses to T-DM1 in lung cancers. The irreversible pan-HER kinase inhibitors neratinib and afatinib are approved agents for HER2-overexpressing breast cancer and EGFR-mutant lung cancer, respectively. In addition to their strong activity in inhibiting HER2 phosphorylation, they are thought to induce receptor polyubiquitination and internalization, likely due to hindrance of HSP90 binding (23–25). We thus tested whether these findings could be translated to ERBB2-amplified or -mutant lung cancers. Cotreatment of Calu-3 (ERBB2 amplified) and LUAD-10 (patient-derived ERBB2^L755P mutant) lung cancer cells with the irreversible inhibitor neratinib markedly increased T-DM1 internalization, as quantitated by pHrodo-T-DM1 live imaging (Fig. 3A and B). In contrast, cotreatment with the reversible HER2 inhibitor lapatinib, known to promote receptor stabilization and accumulation at the plasma membrane (26), decreased the internalization of pHrodo-T-DM1 (Fig, 3A and B). Consistently, ubiquitination of immunoprecipitated HER2 was stronger than controls upon treatment with the irreversible inhibitors neratinib or afatinib but not with the reversible inhibitors lapatinib or tucatinib in both cell lines (Fig. 3C and D). This increase in HER2 ubiquitination and internalization was recapitulated by HSP90 inhibition (Fig. 3C and D; Supplementary Fig. S4A and S4B). The same pattern of HER2 ubiquitination was reproduced also in the NCI-H2030 isogenic models (Fig. 3E). Collectively, these data suggest that irreversible inhibitors of HER2 increase poly-ubiquitination and subsequent internalization of the receptor via HSP90 dissociation in ERBB2-amplified or -mutant lung cancers. Importantly, this phenomenon seems to be prevented by reversible kinase inhibitors.

To determine whether the enhanced receptor internalization induced by cotreatment with neratinib and T-DM1 was sufficient to enhance antitumor efficacy, we treated the same lung PDXs bearing ERBB2 amplification and mutation (S310F) shown in Fig. 1J with either neratinib, T-DM1, or the combination. Although both T-DM1 and the combination of T-DM1 and neratinib induced marked tumor regression, the effect was more durable with the combination (despite negligible activity observed with neratinib monotherapy; Fig. 3F). All treatments were well tolerated by the animals (Supplementary Fig. S5).

Because the combination of T-DM1 and neratinib is not given as standard of care at progression on T-DM1, we evaluated the clinical activity of this combination in a 41-year-old patient with ERBB2-amplified breast cancer enrolled in an ongoing clinical trial (NCT01494662). At the time of study entry, the patient had relapsed after multiple lines of anti-HER2 therapy, including T-DM1 (see details in Methods). Immediately upon T-DM1 progression, neratinib was added to T-DM1 and the patient experienced a brisk partial response (-38%) 6 weeks into therapy (Fig. 3G) and continued with this treatment until intracranial progression at week 18 (at that time the patient remained in confirmed partial response extracranially). This prompt response is consistent with our preclinical data that cotreatment with neratinib improved internalization and efficacy of T-DM1, and with a recently published trial showing encouraging efficacy and tolerability of T-DM1 and neratinib in metastatic breast cancer, albeit in a T-DM1–naïve setting (27).

Given that T-DXd has demonstrated significant efficacy in T-DM1–resistant ERBB2-amplified breast cancers (18), we next evaluated its antitumor activity in the lung PDXs bearing ERBB2 amplification and mutation (S310F) shown in Figs. 1J and 3F, that developed resistance to T-DM1 over time. Treatment with T-DXd induced striking responses resulting in complete tumor regression in all the mice (Fig. 4A), without any sign of toxicity (Supplementary Fig. S6A), similar to the treatment with the combination of T-DM1 and neratinib (Fig. 3F). As a preliminary proof-of-concept, we treated the corresponding patient with lung cancer bearing both ERBB2 amplification and mutation (S310F) with T-DXd on a phase I trial (NCT02564900) upon progression on T-DM1 after an initial response on 4 months of treatment. Strikingly, the patient achieved a partial response to T-DXd (70% tumor shrinkage) that lasted for one year (Fig. 4B). Moreover, we observed that although lung PDXs harboring a YVMA ERBB2 mutation (different from the PDX shown in Fig. 1I) were initially sensitive to all ADC-containing therapeutic regimens, mice receiving T-DM1 or the combination of T-DM1 and neratinib relapsed shortly after treatment discontinuation. In contrast, animals treated with T-DXd monotherapy did not show any sign of tumor regrowth up to 1 month after treatment discontinuation (Fig. 4C; Supplementary Fig. S6B). To rule out the possibility that the observed superior activity of T-DXd compared with T-DM1 monotherapy was due to faster internalization kinetics, we performed a parallel live-imaging experiment with pHrodo-T-DM1 and pHrodo-T-DXd in Calu-3 cells. As expected, we did not observe any significant difference in the internalization rate of the two ADCs (Supplementary Fig. S6C).

Collectively, these data suggest that both the combination of T-DM1 with pan-HER irreversible kinase inhibitors or switching to T-DXd result in superior activity compared with T-DM1 monotherapy (Fig. 4D).

The activity of T-DM1 shown here represents the first clinical trial evidence that ERBB2 amplification in lung cancers may be therapeutically targeted, which stands in contrast to two decades of disappointing efforts targeting HER2 protein expression in lung cancers (7). This study also confirmed the clinical activity of anti-HER2 ADCs in lung cancers with HER2-activating mutations, regardless of the quantity of protein expression (Supplementary Table S1). These results validate our hypothesis that receptor hyperactivation through gene amplification or mutation facilitates its ubiquitination and internalization, which may be therapeutically exploited through ADCs. Although two independently conducted clinical trials of T-DM1 targeting HER2 protein expression (IHC 2+/3+) in lung cancers did not reach expected outcomes, a signal of encouraging responses was seen in tumors with ERBB2 amplification or mutation, which is consistent with our findings (28, 29). We also observed clinical activity of T-DM1 across all activating HER2 mutation subtypes (Supplementary Table S1), in contrast to the differential activity of tyrosine kinase inhibitors across mutation subtypes seen in previous studies (11, 30, 31). This finding may be explained by the different mechanism of action of ADCs compared with tyrosine kinase inhibitors, that is, the internalization of receptor–ADC complex to deliver the cytotoxic payload rather than inhibiting oncogenic signaling. This also explains the observed responses to T-DM1 in EGFR-mutant and ERBB2-amplified lung cancer. Therefore, ADCs represent a fundamentally different targeted therapy for lung cancer, as their efficacy is independent of the drug's ability to inhibit oncogenic driver signaling.

Our laboratory and clinical data are consistent with the idea that surface-localized and cycling HER2 can serve as a carrier for HER2-specific ADCs. Therefore, strategies that augment the dynamics of internalization may increase the efficacy of these drugs. We observed that cotreatment with a pan-HER irreversible inhibitor such as neratinib enhanced internalization of HER2–ADC complexes in vitro, resulting in potent antitumor activity in vivo. Importantly, although treatment with reversible inhibitors such as lapatinib or tucatinib decreased HER2 ubiquitination, treatment with irreversible inhibitors such as neratinib or afatinib increased such ubiquitination, similar to HSP90 inhibition. A superior activity of the combination of the irreversible HER inhibitor poziotinib and T-DM1 compared with single agents has been recently shown in a lung cancer PDX model (32). However, the proposed mechanism of action in that report is HER2 stabilization at the membrane, which is in contrast with our findings. Moreover, it has been shown that receptor stabilization at the plasma membrane increases antibody-dependent cytotoxicity exerted by natural killer cells binding to antibodies such as trastuzumab or cetuximab (26, 33).

It has been shown that HSP90 binding regulates the stability of mature membrane-bound form of HER2. Indeed, HSP90 inhibition triggers receptor ubiquitination, destinating HER2 to proteasomal degradation (23, 34). Here, we propose that FDA-approved irreversible pan-HER inhibitors, by increasing HER2 ubiquitination, may act as HER2-specific HSP90 inhibitors. Thus, HER2 targeting by irreversible inhibitors would be used not to obtain a stronger inhibition of the kinase and the downstream signaling, but to further enhance internalization of the receptor–ADC, improving the antitumor activity of the payload. As a matter of fact, we also speculate that subtherapeutic doses (or perhaps pulsatile treatment schedules) of neratinib or afatinib may be sufficient to maximize ADC-dependent cell death/tumor shrinkage, thus minimizing the adverse effects associated with daily treatment with these agents (35).

Although combined neratinib and T-DM1 elicited an ORR of 63% with a manageable toxicity profile in a phase I trial of 19 patients with HER2-positive breast cancer previously treated with trastuzumab- and pertuzumab-based therapies (27), these responses were not compared to T-DM1 monotherapy and no data are available on the efficacy of this combination in patients with T-DM1–refractory breast cancer. Furthermore, there are no current clinical trials evaluating the clinical benefit of the combination of T-DM1 and pan-HER irreversible inhibitors in comparison with T-DM1 monotherapy in ERBB2-mutant or -amplified lung cancers. Similarly, it is currently unknown whether the antitumor effects of T-DXd, recently FDA-approved for metastatic breast cancer (36), would also be enhanced by the up-front combination with neratinib. The question of potential overlapping gastrointestinal toxicity of the two agents may only be addressed through clinical trial investigation.

The superior efficacy of T-DXd compared with T-DM1 observed in our models is unlikely to be the consequence of differential ADC internalization. Rather, we posit that genetic, epigenetic, or transcriptional mechanisms that render tumor cells more refractory to microtubule-directed chemotherapy (37, 38) may be at play in this context. The antitumor activity of T-DXd in tumors that are or become resistant to T-DM1 may be the result of a combination of variables, including the higher drug-to-antibody ratio compared with T-DM1, diversity between stability and cleavability of the respective linkers, a cytotoxic payload with a different mechanism of action, and the higher cell permeability of the payload resulting in bystander cytotoxicity in the neighboring cells (39). The identification of biomarkers of response to the available ADCs (or their payloads) will be critical for a better patient stratification in future trials.

Despite the earlier setbacks, several clinical development strategies targeting HER2 in lung cancers are now under way. T-DXd is currently being evaluated in phase I/II clinical trials as a single agent (NCT03505710) or in combination with immune checkpoint inhibitor pembrolizumab (NCT04042701). Trastuzumab duocarmazine is another anti-HER2 ADC which has demonstrated activity in patients with T-DM1–resistant or low HER2–expressing breast cancers (40). Other anti-HER2 ADCs in early-phase development include ZW-49 (NCT03821233), DHES0815A (NCT03451162), RC48-ADC (NCT03500380), and A166 (NCT03602079). Novel irreversible pan-HER tyrosine kinase inhibitors such as pyrotinib and poziotinib have also shown promising activity in early-phase clinical trials targeting HER2-mutant lung cancers (32, 41). As the development of brain metastases is common in HER2-mutant lung cancers, affecting 47% of patients, central nervous system activity should be further evaluated in future trials as it is in breast cancer trials (42, 43). As these trials of HER2-targeted agents bring renewed hope, our findings provide clinical evidence and mechanistic insights to guide the development of ADCs and their combinations for patients with lung cancers and other solid tumors.

In summary, ADC-based therapies are a promising new treatment for patients with ERBB2-amplified or -mutant lung cancers. Concurrent treatment with irreversible pan-HER kinase inhibitors or ADC therapy switching can improve the efficacy of these agents. Our findings can potentially be extended to other HER2-activated tumor types for which pharmacologic HER2 signaling blockade is insufficient to elicit strong and durable responses, and may help guide future clinical trials and expand the field of ADCs as cancer therapy.

We declare compliance with all relevant ethical regulations.

A cohort of 49 patients with lung adenocarcinomas carrying either amplification of or mutations in ERBB2 was enrolled in a phase II basket trial at MSK #15-335 (NCT02675829) to assess the clinical activity of ado-trastuzumab emtansine (T-DM1). The primary objective was the determination of ORR according to RECIST v1.1 or modified PERCIST criteria as assessed by investigator. Secondary objectives included assessment of PFS and toxicity according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.1 (NCI CTCAE v4.1). The data cutoff for this analysis was August 1, 2019. All patients received T-DM1 at 3.6 mg/kg by intravenous infusion every 21 days until disease progression or unacceptable toxicity. Physical examination and safety assessments were performed every 3 weeks. Tumor assessments using contrast enhanced CT chest abdomen pelvis and/or PET were performed at baseline, week 6, week 12, then every 12 weeks thereafter until disease progression. One patient had no RECIST or PERCIST measurable disease but was evaluable for progression-free survival. Therefore, the waterfall plot included 48 patients. The 95% exact CI for ORR was calculated using the Clopper–Pearson method. PFS time was estimated by the Kaplan–Meier method. ERBB2 mutation and amplification status was evaluated using MSK-IMPACT (see targeted tumor sequencing section below), another NGS assay in a CLIA-certified laboratory, and/or FISH. FISH was performed using FDA-approved probe sets (PathVysion, Abbott, and ERBB2 IQFISH pharmDx, Dako). HER2 protein was assessed by IHC using the 4B5 antibody (Ventana).

All patients provided written informed consent for genomic sequencing of tumor DNA, and review of medical records for detailed demographic, pathologic, and clinical data and for publication of this information as part of an institutional review board (IRB)–approved investigator sponsored trial (NCT02675829). Research protocols for tumor collection and analysis were approved by the IRB of Memorial Sloan Kettering Cancer Center (MSK).

The patient shown in Fig. 3F has provided written informed consent to DF/HCC protocol #09-204, which is an IRB-approved, prospective cohort study enrolling patients with metastatic breast cancer seen for at least one visit to Dana-Farber Cancer Institute (Boston, MA). The consent includes permission to publish deidentified data. In addition, the patient consented to DF/HCC protocol #11-344, a prospective clinical trial testing neratinib-based therapies for patients with breast cancer brain metastases (NCT01494662), and received the combination of neratinib + T-DM1 under this protocol. The investigator-sponsored trial's principal investigator (Dr. Rachel Freedman) has provided permission to report on the extracranial response only for this single trial patient ahead of reporting of the primary trial endpoint (central nervous system objective response rate). Before being enrolled in this clinical trial, the patient was treated with doxorubicin–cyclophosphamide followed by paclitaxel–trastuzumab–pertuzumab, achieving a pathologic complete response at surgery. The patient then completed adjuvant trastuzumab treatment and started docetaxel–trastuzumab–pertuzumab, when metastatic relapse (bilateral pulmonary nodules and mediastinal/hilar/supraclavicular adenopathy) occurred 27 months after completion of adjuvant trastuzumab treatment. The patient achieved a clinical response and was transitioned to trastuzumab–pertuzumab maintenance. Seven months after the initial diagnosis of metastatic disease, symptomatic brain metastases were diagnosed, and treated with surgery followed by whole-brain radiotherapy. Trastuzumab and pertuzumab were continued for an additional 2 months; however, the patient progressed with pericardial effusion and pulmonary metastases, and T-DM1 monotherapy was started. Tumor burden stabilized for 6 months, but resistance to T-DM1 emerged with enlargement of the lung lesions. At this stage, neratinib was added to T-DM1.

The patient shown in Fig. 4D is a 73-year-old woman with ERBB2^S310F-mutant and -amplified metastatic lung adenocarcinoma, who has progressed on prior cisplatin–pemetrexed chemotherapy and ipilimumab–nivolumab immunotherapy. She achieved an initial partial response to T-DM1 followed by progression of disease after 4 months of treatment. She was then treated with T-DXd on a phase I clinical trial (NCT02564900) and achieved a partial response with 70% tumor shrinkage on RECIST v1.1 with clinical benefit that lasted 1 year.

DNA from formalin-fixed paraffin-embedded tissue and matched germline DNA underwent targeted NGS assay using MSK-IMPACT. In brief, this assay uses a hybridization-based exon capture design to interrogate all protein-coding exons and select introns of oncogenes, tumor-suppressor genes, and key members of pathways that may be actionable using targeted therapies. In this study, 468 key cancer-associated genes were analyzed. Sequencing data were analyzed as previously described to identify somatic single-nucleotide variants, small insertions and deletions, copy-number alterations, and structural rearrangements (15).

ERBB2 copy-number analysis was performed using a computational algorithm termed FACETS to adjust for tumor purity, ploidy, and variant allele frequency, as described previously (22). The degree of ERBB2 amplification was correlated with clinical response to anti-HER2 therapy.

For preclinical studies, T-DM1 was obtained from MSK pharmacy and T-DXd was obtained from Daichii Sankyo Co., Ltd. The HER2 inhibitors neratinib (provided by Puma Biotechnology), afatinib (Selleckchem, #S1011), lapatinib (Selleckchem, # S1028), and tucatinib (Selleckchem, #S8362) and the HSP90 inhibitor ganetespib (Selleckchem, #S1159) were dissolved in DMSO (10 mmol/L stock), aliquoted and stored at 20°C. The endocytosis inhibitor dynasore (Abcam, #ab120192) was dissolved in DMSO (30 mmol/L stock), aliquoted, and stored at 20°C.

MCF10A isogenic cell lines were purchased from ATCC (CRL-10317) and were grown in DMEM/F12 (Invitrogen #11330-032) supplemented with horse serum 5% (Invitrogen #16050-122), EGF 20 ng/mL (PeproTech #AF-100-15), hydrocortisone 0.5 mg/mL (Sigma #H-0888), cholera toxin 100 ng/mL (Sigma #C-8052), insulin 10 μg/mL (Sigma #I-1882), penicillin/streptomycin 1% under standard conditions.

NCI-H2030 were purchased from ATCC (CRL-5914) and were grown in RPMI 1640 (Thermo Fisher Scientific, #11875093) supplemented with 10% heat-inactivated FBS (Thermo Fisher Scientific, #16140071), penicillin/streptomycin 1% under standard conditions.

Calu-3 cell line was purchased from ATCC (HTB-55) and the cells were grown in DMEM/F12 supplemented with 10% FBS, penicillin/streptomycin 1% under standard conditions.

Patient-derived LUAD-10 cells were established in our laboratory from a PDX engrafted with the pellet isolated from the pleural effusion of a patient with HER2^L755P-mutant lung cancer from a phase II basket trial at MSK #15-335 (NCT02675829), after T-DM1 progression. Cells were plated on collagen-coated Petri dishes and cultured in DMEM/F12 supplemented with 10% FBS, penicillin/streptomycin 1% under standard conditions.

MCF10A and NCI-H2030 isogenic cell lines stably expressing EV, ERBB2 WT, or ERBB2 mutants were obtained by lentiviral infection as follows. Briefly, 7 × 10⁶ 293T cells were seeded into 10-cm dishes. Cells were transfected with 1.2 μg pMD2.G, 2.4 μg pCMV-dR8.2, and 3.6 μg pLX302-EV, pLX302-ERBB2 WT, pLX302-ERBB2 S310F, or pLX302- ERBB2 L755S using jetPRIME (Polyplus-transfection) following the manufacturer's instructions. Medium was refreshed 6 hours post-transfection. The supernatant was collected 72 hours and filtered with 0.45-μm filters.

MCF10A and NCI-H2030 cells (3 × 10⁵) were seeded into 6-well plates and transduced with different dilutions of freshly collected lentiviruses with 8 μg/mL Polybrene for 24 hours. Cells were then trasnfered into 10-cm dishes and selected with 2 μg/mL (MCF10A) or 5 μg/mL (NCI-H2030) of puromycin for 2 days. Cells stably expressing comparable levels of WT or mutant HER2 were chosen for subsequent studies.

pHrodo-TDM1 and pHrodo-T-DXd were generated using pHrodo iFL Red Microscale Protein Labelling Kit (Invitrogen #P36014) according to manufacturer's instructions. Briefly, a reaction solution containing 100 μL of the ADC (1 mg/mL), 10 μL of sodium bicarbonate (1 mol/L), and 3.3 μL of pHrodo iFL Red STP esteramine-reactive dye (2 mmol/L in DMSO) was incubated for 30 minutes at room temperature. pHrodo-ADCs were purified by centrifugation using a purification resin, aliquoted, and stored at 20°C.

For internalization assays, 4 × 10⁴ cells were seeded on 8-well chamber slides (Nunc Lab-Tek II Chamber Slide, Thermo Fisher Scientific #154534). The day after, cells were incubated with serum-free medium containing Hoechst (Thermo Fisher Scientific #62249) for live-cell fluorescent staining of DNA, and pHrodo-ADC (1 μg/mL, unless stated differently in the figure legend) for 30 minutes at 4°C and then released at 37°C in the heated live chamber of the Zeiss LSM880 confocal microscope. Positions were decided using brightfield (at least 4 positions per sample) and 2 μm–thick z-stack images of brightfield and red fluorescence were taken every hour over 16 hours. Images were then analyzed using Fiji software and results were displayed as average number of cytoplasmic red fluorescent dots per cell (trafficking index).

Where indicated, cells were incubated with serum-free medium containing Hoechst stain, pHrodo-ADC, and neratinib (100 nmol/L, unless stated differently in the figure legend) or lapatinib (100 nmol/L, unless stated differently in the figure legend) or HSP90 inhibitor (200 nmol/L) or DMSO as control.

Total protein lysates (20 μg) were extracted using RIPA buffer and separated on SDS-PAGE gels (NuPAGE 4–12% Bis-Tris Protein Gels, Invitrogen) according to standard methods.

Membranes were probed using the following antibodies: anti-total HER2 Rabbit mAb (29D8, Cell Signaling Technology #2165), anti-total HER2 Mouse mAb (CB11, Thermo Fisher Scientific #MA1-35720), anti-phospho-HER2 (Tyr1248, Cell Signaling Technology #2247), anti-Cleaved PARP Asp214 human specific (Cell Signaling Technology #9541), anti-phospho-tyrosine (Cell Signaling Technology #8954), anti-β-Actin 13E5 (Cell Signaling Technology #4970), and ubiquitin (P4D1, Cell Signaling Technology #3936S).

For immunoprecipitation (IP) assays, cells were incubated in 5% FBS medium in the presence of the proteasome inhibitor MG-132 (10 μmol/L), together with 10 μg/mL of T-DM1 (or IgG control, Cell Signaling Technology #2729) and 100 nmol/L of tyrosine kinase inhibitors or HSP90 inhibitor at 37°C for 6 hours (except for LUAD-10, which were incubated for 3 hours). Cells were then washed in cold PBS and lysed using NP-40 buffer (150 mmol/L NaCl, 10 mmol/L Tris pH 8, 1% NP-40, 10% glycerol). Twenty micrograms of proteins were used as total lysates. Protein lysates (350 μg for Calu-3 and NCI-H2030, 1500 μg for LUAD-10) were incubated rotating at 4°C for 4 hours in the presence of 10 μg/mL trastuzumab, which was added only to samples lacking T-DM1 or IgG control. Agarose Protein A/G beads (Santa Cruz Biotechnology #sc-2003) were then added and samples were incubated rotating at 4°C for 1 hour. Beads were centrifuged at 3,000 rpm for 1 minute and supernatant was removed. Similarly, beads were washed twice using NP-40 buffer and once using nuclease-free sterile water. Finally, immunoprecipitates were eluted using 1× sample buffer (NuPAGE LDS Sample Buffer, Thermo Fisher Scientific #NP0008 and NuPAGE Sample Reducing Agent, Thermo Fisher Scientific #NP0009) at 95°C for 5 minutes.

PDXs were generated as follows: 6-week-old NOD scid gamma female mice were implanted subcutaneously with specimens freshly collected from patients at MSK under an MSK-approved IRB biospecimen protocol #06-107. Tumors developed within 2 to 4 months and were expanded into additional mice by serial transplantation. At this point, PDXs were subjected to high-coverage NGS with the MSK-IMPACT assay. For efficacy studies, treatment was started when tumor volumes reached approximately 100 mm³. Xenografts were randomized and dosed with T-DM1 (15 mg/kg, intravenously once a week), neratinib (20 mg/kg, orally 5 days a week), a combination of the two agents, T-DXd (10 mg/kg, intravenously once every 3 weeks), or vehicle as control (saline, orally 5 days a week). Mice were observed daily throughout the treatment period for signs of toxicity. Tumors were measured twice weekly using calipers, and tumor volume was calculated using the formula length × width² × 0.52. Body weight was also assessed twice weekly. At the end of each treatment, animals were sacrificed, and tumors were collected for biochemistry and histology analysis. Mice were cared for in accordance with guidelines approved by the MSK Institutional Animal Care and Use Committee and Research Animal Resource Center. Six to 7 mice per group were included in each experiment.

Statistical analyses were conducted using GraphPad Prism 8 (GraphPad Software Inc.). Two-way ANOVA test was used to evaluate significant differences in tumor volumes in in vivo efficacy studies. Data are presented as mean ± SEM. *, P ≤ 0.05; **, P < 0.01; ***, P ≤ 0.001; ****, P < 0.0001. Median PFS and 95% CI were calculated using the Kaplan–Meier method. Association of ERBB2 copy number with response status was evaluated using the Wilcoxon rank-sum test. Spearman correlation was used for correlating copy-number assessments by FISH, IHC, and NGS.

All genomic results and associated clinical data for all patients in this study are publicly available in the cBioPortal for Cancer Genomics at the following URL: http://cbioportal.org/msk-impact. All other relevant data are included in the article and/or Extended Data files.

B.T. Li is a consultant/advisory board member at Roche/Genentech, Biosceptre International, Thermo Fisher Scientific, Mersana Therapeutics, Hengrui Therapeutics, and Guardant Health, reports receiving commercial research grants from Roche/Genentech, Daiichi Sankyo, Lilly, Hengrui Therapeutics, Illumina, Guardant Health, BioMed Valley Discoveries, AstraZeneca, GRAIL, MORE Health, and Amgen, and has ownership interest in two pending institutional patents at Memorial Sloan Kettering Cancer Center (US62/685,057, US62/514,661). S. Misale is a consultant at Boehringer-Ingelheim. A.S. Lalani is SVP, Translational Medicine, at Puma Biotechnology, Inc. and has ownership interest (including patents) in the same. I. Diala is a Manager, Clinical Science, at Puma Biotechnology. N.U. Lin is an advisory board member at Puma, Seattle Genetics, and Daichii Sankyo and reports receiving commercial research grants from Genentech, Merck, Pfizer, and Seattle Genetics. D.B. Solit is an advisory board member for Pfizer, Loxo Oncology, Lilly Oncology, Vivideon Therapeutics, Illumina, and QED Therapeutics. M.F. Berger is a consultant at Roche. T. Ng reports receiving commercial research grants from Daiichi Sankyo and AstraZeneca. M. Offin has received speakers bureau honoraria from PharmaMar, Novartis, and Targeted Oncology. J.M. Isbell has ownership interest (including patents) in LumaCyte, LLC and is a consultant/advisory board member for Roche Genentech. D.R. Jones is senior medical advisor at Diffusion Pharmaceuticals, and a consultant at Merck and AstraZeneca. H.A. Yu is a consultant at AstraZeneca and Daiichi and reports receiving commercial research grants from AstraZeneca, Pfizer, Daiichi, Novartis, and Lilly. S. Thyparambil is a senior director, R&D, at mProbe. A. Bhalkikar is an associate scientist at mProbe, Inc. F. Cecchi is an associate director, proteomics, translational medicine, at AstraZeneca. D.M. Hyman is a chief medical officer at Loxo Oncology and Eli Lilly, is a consultant at Chugai Pharma, Boehringer Ingelheim, AstraZeneca, Pfizer, Bayer, Genentech, and Fount/Kinnate, and reports receiving commercial research grants from AstraZeneca, Puma Biotechnology, Loxo Oncology, and Bayer. J.S. Lewis reports receiving commercial research support from Genentech. A.L. Ho is an advisory board member at Eisai, Sanofi Genzyme, Ayala Pharmaceuticals, Regeneron, CureVac, Klus Pharm, Prelude Therapeutics, Novartis, AstraZeneca, Merck, Bristol-Myers Squibb, Genentech/Roche, and Sun Pharmaceuticals, has received conference and travel fees from Kura Oncology and Ignyta, reports receiving commercial research grants from Eisai, Novartis, Ayala, Astellas, Daiichi Sankyo, Pfizer, Allos Pharm, Kura Oncology, AstraZeneca, Merck, Bristol-Myers Squibb, Genentech, Celldex, Bayer, and Lilly, and has received speakers bureau honoraria from Medscape. V. Makker is a principal investigator at Eisai, Merck, and Karyopharm, is a consultant at IBM Watson, reports receiving commercial research grants from Eisai, Merck, Karyopharm, Lilly, AstraZeneca, Takeda, and Genentech, and is a consultant/advisory board member for Eisai, Merck, and Karyopharm. J.S. Reis-Filho is a consultant at Goldman Sachs, is a consultant and SAB member at Paige.AI, is an SAB member at Volition RX, and is an ad hoc SAB member at Roche Tissue Diagnostics, Ventana Medical Systems, Genentech, InVicro, and Novartis. P. Razavi is a consultant/advisory board member at Novartis, AstraZeneca, and Foundation Medicine and reports receiving institutional research funding from Grail Inc. M.G. Kris is a consultant at AstraZeneca, Pfizer, and Daiichi, and has immediate family members who are consultants at Regeneron. J. Tsurutani reports receiving commercial research grants from Eisai, Daiichi Sankyo, Kyowa Kirin, Chugai, Eli Lilly, Nihon Kayaku, Pfizer, and MSD, has received speakers bureau honoraria from Daiichi Sankyo, Eisai, Kyowa Kirin, Novartis, Chugai, Eli Lilly, Nihon Kayaku, Taiho, Pfizer, and AstraZeneca, and is a consultant/advisory board member at Daiichi Sankyo, Eli Lilly, Asahikasei, and MSD. N. Rosen is an SAB member at AstraZeneca, Chugai, Beigene, Zai Laboratories, MapKure, and Ribon, is a consultant for Tarveda, Boehringer Ingelheim, Concello, and Jubilant, reports receiving commercial research grants from Boehringer Ingelheim and Chugai, and has ownership interest (including patents) in Beigene, Kura, and Zai Laboratories. C.M. Rudin is a consultant at AbbVie, Amgen, Ascentage, Vavotek, Bicycle, Celgene, Daiichi Sankyo, Genentech Roche, Ipsen, Loxo/Lilly, and Pharmamar, is an SAB member at Harpoon and Bridge Medicines, and reports receiving a commercial research grant from Daiichi Sankyo. M. Scaltriti reports receiving commercial research grants from Menarini Ricerche, Daiichi-Sankyo, AstraZeneca, Immunomedics, Puma Biotechnology, and Targimmune, has ownership interest in Medendi.org and two pending institutional patents at Memorial Sloan Kettering Cancer Center (US62/685,057, US62/514,661), and is a consultant/advisory board member for Daiichi-Sankyo, Menarini Ricerche, and AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.T. Li, F. Michelini, S. Misale, A.S. Lalani, M. Scaltriti

Development of methodology: B.T. Li, F. Michelini, S. Misale, Y. Cai, D.B. Solit, P.R. Barber, M. Offin, M.E. Arcila, G.A. Ulaner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.T. Li, F. Michelini, S. Misale, E. Cocco, L. Baldino, S. Shifman, M.L. Myers, H.-Y. Tu, C. Xu, M. Mattar, M. Little, B. Qeriqi, G. Weitsman, A.S. Lalani, R.A. Freedman, N.U. Lin, D.B. Solit, M. Offin, J.M. Isbell, D.R. Jones, S. Thyparambil, W.-L. Liao, A. Bhalkikar, D.J. Buonocore, M.G. Kris, J.T. Poirier, J. Tsurutani, G.A. Ulaner, E. de Stanchina, C.M. Rudin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.T. Li, F. Michelini, S. Misale, E. Cocco, H.-Y. Tu, M.L. Myers, G. Weitsman, C.J. Wilhelm, A.S. Lalani, R.A. Freedman, N.U. Lin, D.B. Solit, M.F. Berger, M. Offin, H.A. Yu, S. Thyparambil, A. Bhalkikar, F. Cecchi, J.S. Lewis, V. Makker, J.S. Reis-Filho, P. Razavi, M.G. Kris, R. Shen, G.A. Ulaner, M. Scaltriti

Writing, review, and/or revision of the manuscript: B.T. Li, F. Michelini, S. Misale, C. Xu, H.-Y. Tu, M.L. Myers, C.J. Wilhelm, A.S. Lalani, I. Diala, R.A. Freedman, N.U. Lin, D.B. Solit, T. Ng, M. Offin, J.M. Isbell, D.R. Jones, H.A. Yu, A. Bhalkikar, F. Cecchi, D.M. Hyman, J.S. Lewis, A.L. Ho, V. Makker, J.S. Reis-Filho, P. Razavi, M.E. Arcila, M.G. Kris, J.T. Poirier, R. Shen, J. Tsurutani, G.A. Ulaner, N. Rosen, C.M. Rudin, M. Scaltriti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.T. Li, L. Baldino, H.-Y. Tu, M.L. Myers, C. Xu, I. Khodos, A.S. Lalani, D.B. Solit, M. Offin, D.M. Hyman, M.G. Kris, M. Scaltriti

Study supervision: B.T. Li, S. Misale, M.G. Kris, M. Scaltriti

We thank the MSK imaging unit and the MSKCC pharmacy for their help and all the members of the Scaltriti group for useful discussion. This study was funded by the National Cancer Institute under the MSK Cancer Center Support Grant/Core Grant (P30 CA008748), NIH Grants U54 OD020355-01 (to E. de Stanchina) and P01 CA-129243, the Conquer Cancer Foundation of ASCO-AACR Young Investigator Translational Cancer Research Award, the Carol Lowenstein Fund, and a kind gift from Mr. and Mrs. Peter Pritchard. This work was also partially funded by an unrestricted grant from MORE Health, Genentech, Puma Biotechnology, and Daiichi Sankyo. E. Cocco is a recipient of an MSK society scholar prize. P.R. Barber and T. Ng were funded by the CRUK UCL Centre (C416/A25145), CRUK City of London Centre (C7893/A26233), and CRUK KCL-UCL Comprehensive Cancer Imaging Centre (CRUK and EPSRC) in association with the MRC and DoH (C1519/A16463 and C1519/A10331).