EGFR Inhibition Enhances the Cellular Uptake and Antitumor-Activity of the HER3 Antibody–Drug Conjugate HER3–DXd

Epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKI) are the standard of care for advanced EGFR-mutant non–small cell lung cancer (NSCLC; refs. 1, 2). For patients with the advanced EGFR-mutant cancer harboring the common EGFR exon 19 deletion and L858R mutations, osimertinib is the preferred first-line EGFR inhibitor of choice given superior progression-free and overall survivals compared with prior generation EGFR TKIs (3). However, resistance to osimertinib invariably develops and the mechanisms of resistance are quite diverse (4, 5). Although targeting specific osimertinib resistance mechanisms is feasible, it may not be practical, and frequently no targetable resistance mechanism is identified (4). HER3 is overexpressed in various cancer types, including lung cancer (6, 7), making it an attractive therapeutic target. Both EGFR and HER3 are ubiquitously expressed in primary NSCLC tumors, and EGFR/HER3 expression was previously detected in 67% of circulating tumor cells of NSCLC patients (7). Furthermore, alterations in HER3 have not been described as resistance mechanisms to EGFR TKIs, and as such targeting HER3 maybe a novel approach to treat drug-resistant forms of EGFR-mutant cancers

Patritumab deruxtecan (HER3–DXd; U3-1402) is an antibody–drug conjugate (ADC) composed of a human HER3-targeting antibody (patritumab) linked to a topoisomerase I inhibitor (DX-8951 derivative, or DXd; ref. 8). Patritumab (also known as U3-1287) has been tested as a single agent in 57 patients (20 were NSCLC patients and the majority had been treated with prior EGFR inhibitors) and demonstrated no single-agent activity (9). HER3–DXd is currently being evaluated as a single agent in EGFR inhibitor–resistant EGFR-mutant NSCLC, HER3-positive metastatic breast cancer, and metastatic colorectal cancer (NCT03260491, NCT02980341, and NCT04479436). The determinants of the HER3–DXd efficacy are presently not well understood.

In the current study, we queried the single-agent efficacy of HER3–DXd in preclinical models of EGFR TKI-resistant NSCLCs harboring different drug-resistance mechanisms. Although the efficacy of HER3–DXd as a single agent was variable across the models, we sought to develop a strategy to enhance the efficacy of HER3–DXd through pretreatment with osimertinib.

Cells were seeded into 96-well plate the previous day of the assay start (6,000 cells/well) to obtain 30%–40% confluency. Next day, the cells were treated with either DMSO or osimertinib and preimaged for 6 to 8 hours before the addition of the conjugated antibodies. HER3–DXd (or control IgG) was conjugated with fab-pHrodo fragments (Essen Biosciences) using 1:3 molar ratio. Antibodies were conjugated 20 minutes in the dark, +37 °C, after which, the conjugated ADCs were administrated to the cells, and the imaging was continued immediately. Imaging and analysis were performed using IncuCyte Zoom/S3 live-cell imagers (Essen Biosciences) and quantified using the IncuCyte software.

Cell lines (Table 1) were obtained from ATCC and cultured in RPMI medium (Gibco) with 10% FBS (Sigma) and 1% penicillin/streptomycin (Life Technologies). DFCI-284 and DFCI-243 were established by using previously published protocol (10). Cells were tested for Mycoplasma routinely by PCR testing. HER3–DXd was a kind gift of Daiichi Sankyo. Osimertinib and gefitinib were purchased from Selleck (#S7297 and #S1025). Antibodies used in the study are described in Supplementary Table S1.

1,000 cells/well were seeded on the first day in 96-well plates. Next day, cells were treated, and CellEvent Caspase-3/7 green apoptosis detection reagent (Thermo Fisher Scientific) was added to the cells. Cells were imaged every 2 hours, and medium and drugs were changed every week.

CellTiter-Glo (Promega) was conducted according to the manufacturer's instructions.

Cells were pretreated for 24 hours with gefitinib or osimertinib and then continuously treated together with 30 μg/mL cycloheximide. Protein lysates were harvested at the indicated time points.

When two groups were compared, a two-tailed unpaired t test was used to calculate significance. One-way ANOVA was used when comparing three or more groups, and Tukey multiple comparisons test was used for post hoc analysis. Fisher exact test was used to analyze the regrowth of the patient-derived xenograft (PDX) tumors after treatment (Supplementary Fig. S5G and S5J). The used statistical tests are indicated in the figure legends. Data are presented as mean ± SD or SEM, indicated in the figure legends. P ≤ 0.05 was considered significant, and all assays were repeated using three to six biological replicates. GraphPad Prism 7.04 software was used for statistical analyses. Graphical abstract was created with BioRender.com (Publication agreement number: QA22SQP9UO).

Cells were detached with Accutase (Thermo Fisher Scientific), and both cells and culture medium were collected. Cells were counted with Countess cell counter (Invitrogen), suspension confluency was adjusted to 1 × 10⁶ cells/mL, after which, the cells were stained with 1:100 zombie viability dye for 20 minutes in the dark at room temperature. Cells were spin down and washed with PBS, followed by 30 minutes staining with conjugated antibodies in flow buffer (10% FBS-PBS), on ice and in the dark. Samples were spin washed three times, resuspended into flow buffer and acquired using Fortessa analyzer (BD). The data were analyzed with FlowJo version 10.6 (TreeStar).

PDX tumors were collected and minced with a scalpel, after which, they were shook 140 RPM in 0.2% collagenase A and cell culture medium for 3 to 4 hours, +37°C. Cell suspension was treated with TrypLe Express solution (Thermo Fisher Scientific) for 20 minutes +37°C to obtain a single-cell suspension. Cell suspension was adjusted to 1 × 10⁶ cells/mL and stained with zombie viability dye for 20 minutes at room temperature in the dark. Cells were spin down and washed with PBS, followed by 30 minutes staining with conjugated antibodies in flow buffer (10% FBS-PBS), on ice and in the dark. Samples were spin washed three times, resuspended into flow buffer and acquired using Fortessa analyzer (BD). The data were analyzed with FlowJo version 10.6 (TreeStar).

Cells grown and treated on coverslips were fixed with 4% PFA and permeabilized with 0.1% Triton X-100. Two-dimensional (2D) immunofluorescence stainings were performed using standard protocols. Three-dimensional (3D) cultured samples were fixed with 2% PFA, permeabilized with 0.25% Triton X-100, and blocked with 10% goat serum in PBS. Primary antibodies were diluted in blocking buffer, and 3D cultures were stained overnight. After three washes with IF buffer (0.1% BSA, 0.2% Triton X-100, 0.05% Tween-20 in PBS), the cells were incubated with secondary antibodies in blocking buffer and washed again. Nuclei were counterstained with 1 μg/mL DAPI (Cell Signaling Technology) and mounted with Immu-Mount reagent (Fisher Scientific). Imaging was performed using Nikon Eclipse 80i and Leica SP5 confocal microscopes (DFCI Confocal and Light Microscopy Core Facility). Quantifications for nuclear staining, Ki67, and pH2aX were performed using ImageJ software.

Four-micron-thick formalin-fixed, paraffin-embedded (FFPE) tissue sections were stained at the Brigham and Women's Hospital Pathology Core. After deparaffinization, sections were treated with the EDTA-based (pH 9.0) Epitope Retrieval Solution at 90°C. Endogenous peroxidase was blocked for 15 minutes. Antibodies were diluted 1:100 and incubated for 30 minutes. Envision + polymer system (DAKO, cat. #4011) was used to detect rabbit antibody for 30 minutes. Sections were detected with DAB and counterstained with hematoxylin. HER3 was visually scored by a pathologist for the generation of H score according to the formula: [3 × (% cells 3+) + 2 × (% cells 2+) +1 × (% cells 1+)].

Cells were seeded in a 96-well plate, 1,000 cells/well, and the following days the cells were treated using HP D300e digital dispenser. Imaging and analysis were performed using IncuCyte Zoom/S3 imagers (Essen Biosciences).

Cells were transiently transfected with 1 μg of the Firefly luciferase-expressing Her3 promoter reporter plasmid (ErbB-3-pGL3, gift from Frederick Domann, Addgene #60899) and 100 ng of the Renilla luciferase-expressing pRL-CMV control plasmid (Promega, cat. #E2261) for 24 hours using Fugene HD transfection reagent (Promega, cat. #E2311). Cells were then drug treated, and luciferase activities were measured using the Dual-Glo Luciferase Assay system (Promega cat# E2920).

Patient-derived xenografts were generated from tumor biopsies or pleural effusions from EGFR-mutant patients undergoing clinical biopsies and propagated in mice. All patients provided written informed consent. The study was conducted in accordance with the Declaration of Helsinki and was approved by the Dana-Farber Cancer Institute. All animal studies were conducted at Dana-Farber Cancer Institute with the approval of the Institutional Animal Care and Use Committee in an Association for Assessment and Accreditation of Laboratory Animal Care International–accredited vivarium. PDX tumors for DFCI-161, DFCI-284, DFCI-295, DFCI-306, DFCI-359, and DFCI-429 were derived from pleural effusions collected from patients as part of routine clinical care. Effusions were immune depleted and enriched for cancer cells, and these cancer cells were cultured on plastic for three days in RPMI-1640 media supplemented with 10% FBS and 1% antibiotic prior to subcutaneous implantation. The PDX tumor for DFCI-315 was derived from a surgical specimen and implanted subcutaneously. The PDX tumor for DFCI-243 was derived from a surgical biopsy and implanted into subrenal capsule for expansion. Following initial implantation, all PDX models were expanded and passaged continually in mice as subcutaneous tumors. All tumors used in efficacy studies were implanted subcutaneously. For the HCC4006 xenograft model, cells were grown in RPMI-1640 media supplemented with 10% FBS and mice were implanted subcutaneously with 5×10⁶ cells/mouse in 50% Matrigel (Corning, 356231). All PDX tumors and HCC4006 cells were implanted in 8- to 10-week-old female NSG (NOD.Cg-Prkdc^scid Il2rg^tm1WjI/SzJ) mice purchased from Jackson Labs (005557-NSG; RRID: IMSR_JAX:005557). Following implantation, tumor establishment and growth were monitored by caliper measurements twice per week. Typically, when tumors reached 150–250 mm³, mice were randomized by tumor volume into various treatment groups. Mice harboring tumors were treated with either human IgG control (Bethyl Laboratories) diluted with PBS, HER3–DXd diluted with 10 mmol/L acetate buffer with 5% sorbitol, pH 5.5 or osimertinib formulated as a fine suspension in 0.5% hydroxypropyl methylcellulose (HPMC) in water. IgG control or HER3–DXd were dosed intravenously either once weekly or once every three weeks, whereas osimertinib was administered orally once daily. During treatment, tumor measurements were taken using calipers and body weights monitored twice a week. Animals were euthanized if the tumor volume exceeded 2,000 mm³ or if the tumors became ulcerated/necrotic, and tumor samples were harvested and snap frozen for subsequent analysis.

Cas9 targeting crRNAs to target EGFR exon 20 were designed using the MIT CRISPR design tool. Homologous repair template encoding T790M/C797S EGFR was supplied to parental PC9 cells as a short single-stranded linear DNA. Silent mutations were incorporated into the donor strand where possible to protect the template from recleavage by Cas9 following integration, and to assess editing efficiency through sequencing. To minimize the effects of codon bias on protein translation, silent mutations incorporated within the donor template were either natural variants or corresponded to codons encoding tRNAs found at comparable frequency in the cell as those encoded by the endogenous codon. Targeting constructs and donor constructs were electroporated into cells alongside nuclear-targeted recombinant active Cas9 protein. Following gene editing, successfully recombined cells were selected with osimertinib, beginning 72 hours after nucleofection. Colonies, including the PC9 T790M/C797S clone used in the studies, were picked by hand under a light microscope and expanded in separate wells under continual osimertinib selection. Allele frequencies were determined through deep amplicon sequencing through the MGH CCIB DNA core CRISPR Sequencing platform. Osimertinib was washed out from PC9 T790M/C797S cells for at least one week prior to initiation of in vitro and in vivo studies.

Osimertinib-resistant DFCI-243 cells were created by treating the cells with increasing doses of osimertinib for multiple weeks, ending up with cells resistant to up to 100 nmol/L osimertinib.

For in vivo experiment, the PC9 T790M C797S cells were harvested, and 5×10⁶ cells with 50% Matrigel (Fisher Scientific) were implanted subcutaneously in the right flank of the 9-week-old female NCr nude mice. Tumors were allowed to establish with an average tumor size of 164 mm³ before randomization into various treatment groups. Mice were treated as above.

The 3D culture sample was carefully minced with a blade and incubated 3 hours with gentle shaking (140 rpm) +37 °C in RPMI medium containing 0.2% of Collagenase A (Sigma), 10% serum, and 1% penicillin/streptomycin (Life Technologies). The samples were centrifuged at 1,400 rpm for 5 minutes, and the pellet was washed with 5 mL PBS. Fragments were recentrifuged, resuspended in Matrigel (BD), and seeded to 8-well chamber slides (Nunc) for 3D culture or 96-well plates for CellTiter-Glo assay. Samples were cultured in RPMI medium with 10% serum and 1% penicillin/streptomycin.

Cells grown and treated on glass coverslips were fixed with 4% PFA and permeabilized with 0.1% Triton X-100. Proximity ligation assay was performed using the manufacturer's protocol (Sigma-Aldrich). For formalin-fixed paraffin-embedded tissues, paraffin was removed from the slides with xylene and ethanol series and the samples were permeabilized using 0.25% Triton X-100 for 25 minutes. Antigens were recovered in microwave for 20 minutes using citrate buffer (Abcam), and the slides were mouse-on-mouse blocked 30 minutes RT using Background Buster (Innovex). After this, proximity ligation assay was performed using the manufacturer's protocol. Nuclei were stained with 1 μg/mL DAPI (Cell Signaling Technology) for 10 minutes and the slides were mounted using Immu-Mount (Fisher Scientific). Proximity ligation assay (PLA) quantification was performed with ImageJ software.

Cell lines were lysed with RIPA lysis buffer. SDS-PAGE and Western blot were done using standard protocols.

To study the relevance of HER2 and HER3 in EGFR inhibitor–resistant cancers, we evaluated the protein expression levels of these two receptors using IHC in EGFR-mutant PDX models resistant to either first- or third-generation EGFR inhibitors, harboring a broad range of acquired resistance mechanisms (Fig. 1A; Supplementary Table S2). Although most models expressed both HER2 and HER3, high HER3 expression was more prevalent, and unlike HER2, HER3 expression was detected in all of the models. We also noted HER3 expression in HER2 mutant and amplified models, although in those models HER2 expression was more prevalent. The widespread HER3 expression across all tumor models suggested, that HER3 could serve as a therapeutic target independent of the specific mechanism of resistance to EGFR inhibitors.

Prior preclinical studies suggest that the efficacy of HER3–DXd is linked to the baseline HER3 expression in breast cancer (8). To evaluate the effects of HER3–DXd in NSCLC, we determined the efficacy of single-agent HER3–DXd in eight PDX models covering a wide range of HER3 expression (Fig 1B–E; Supplementary Fig. S1A and S1B; Supplementary Table S2). Single treatment of HER3–DXd (10 mg/kg, weekly), compared with IgG control, led to a significant tumor growth delay in the majority of the models, except in DFCI-306, which has negligible HER3 expression. The drug was well tolerated (Supplementary Fig. S1C–S1E). HER3–DXd efficacy was dose-dependent (Supplementary Fig. S1F). In two models, DFCI-161 (EGFR L858R/MET amplified) and DFCI-315 (HER2 exon20 V777_G778insGSP) HER3–DXd treatment was associated with substantial tumor regressions, while in the other models it led to tumor growth delay (Fig. 1F). Control IgG antibody did not affect the growth of the tumors (Supplementary Fig. S1G).

We further evaluated the efficacy of HER3–DXd in a model containing the EGFR C797S (11). As we were unable to generate a PDX model from an osimertinib resistant patient with this mutation, we engineered a model using CRISPR-Cas9. Using the EGFR-mutant PC9 cell line, we introduced EGFR T790M and C797S in cis as commonly detected in NSCLC patients. The resulting cells were resistant to EGFR inhibitors compared with the parental population (Supplementary Fig. S1H and S1I; Supplementary Table S3). We next evaluated the efficacy of both osimertinib and HER3–DXd in this model. The PC9 T790M/C797S tumors also expressed HER3 and were resistant to single-agent Osimertinib, whereas single-agent HER3–DXd treatment led to a significant tumor growth delay (Fig. 1G; Supplementary Fig. S1J and S1K).

Our initial studies demonstrated variable efficacy of HER3–DXd as a single agent in EGFR-mutant PDX models and no efficacy in the PDX model that lacked HER3 expression. We previously demonstrated that EGFR inhibitor treatment leads to an increase in HER3 levels in vitro (12), and as such we aimed to determine if EGFR inhibition could potentially enhance the efficacy of HER3–DXd.

To systematically evaluate the impact of EGFR inhibition on HER2 and HER3 expression, six EGFR-mutant NSCLC cell lines were treated with gefitinib or osimertinib, after which, HER2 and HER3 membrane expression were evaluated using flow cytometry (Fig. 2A). To make the analysis more comprehensive, the membrane expression of HER2 and HER3 in viable single cells was further divided into negative, low, and high expression groups (Supplementary Fig. S2A). Twenty-four-hour pretreatment with EGFR inhibitors increased the HER3 membrane levels in all six cell lines in a dose-dependent manner, and the increase was observed in both the amount of HER3-positive cells and the intensity of HER3 expression (Fig. 2B; Supplementary Fig. S2B). In half of the models HER2 membrane expression levels also increased, but in the remaining models HER2 was already highly expressed at baseline, and further increase could not be detected (Supplementary Fig. S2C). In the cell lines carrying the T790M resistance mutation making the cells resistant to gefitinib, osimertinib treatment more effectively increased HER3 membrane expression, suggesting that the effect was dependent on on-target EGFR inhibition. In a time-course assay we observed that the increase in HER3 expression peaked between 8 and 24 hours following the start of EGFR inhibitor treatment (Fig. 2C and D). There was no further increase in HER3 expression when treatment was extended beyond 24 hours (Supplementary Fig. S2D). Treatment with the anti-EGFR monoclonal antibody cetuximab did not lead to an increase in HER3 surface expression (Supplementary Fig. S2E).

PLA is a useful tool for measuring protein–protein interactions in situ and has been used to study and quantify the dimerization interactions of EGFR family (13, 14). A positive PLA signal is detected when the distance between two proteins is less than 40 nm. We used PLA to test if the EGFR TKI-induced increase in membrane HER3 could also lead to formation of HER3:EGFR dimers, suggesting active HER3 signaling. In all three studied cell lines, we noticed a significant increase in membrane HER3:EGFR proximity following EGFR inhibitor treatment (Fig. 2E and F), suggesting that the increase in HER3 expression was also associated with the formation of EGFR:HER3 heterodimers.

Next, we examined whether the increase in membrane expression of HER3 was regulated at the level of mRNA transcription or through protein stability. In a cycloheximide chase assay, HER3 protein levels increased after EGFR inhibitor treatment, suggesting that HER3 increase was regulated at the level of protein stability (Fig. 2G and H; Supplementary Fig. S2F and S2G). However, we also observed minor increases in HER3 promoter activity measured using a luciferase promoter reporter assay (Supplementary Fig. S2H and S2I), and mRNA expression (Supplementary Fig. S2J), suggesting that the upregulation was partly also occurring at the level of transcription. Inhibition of EGFR downstream signaling by either an AKT inhibitor (MK-2206) or a MEK inhibitor (trametinib) partially phenocopied the effect of EGFR inhibition in PC9 cells (Supplementary Fig. S2K), suggesting that inhibition of both signaling pathways contributed to the observed HER3 feedback upregulation.

To study if EGFR inhibition alters HER3 expression also in vivo, four PDX models with variable baseline HER3 expression levels were treated with vehicle or osimertinib for 24 hours followed by analysis of HER3 expression (Fig. 2I). We observed that in all four models osimertinib pretreatment increased HER3 total protein expression (Fig. 2J). In addition, the number of HER3-positive viable epithelial cells (marked by EpCam) also increased following osimertinib treatment (Fig. 2K). We noted a substantial increase in the protein levels of HER3 even in two additional models that were originally low or negative for HER3, including the DFCI-306 model (Supplementary Fig. S2L and S2M). HER3 expression did not increase further by extending the treatment time to 48 hours (Fig. 2L). PLA analysis from the osimertinib-treated PDX tumors also revealed an increase in HER3:EGFR proximity analogous to the in vitro observations (Fig. 2M and N).

Because membrane expression of the target is a key factor for the binding and internalization of ADC (15), we further evaluated the biological significance of the increased HER3 membrane expression on HER3–DXd. We tracked the internalization of HER3–DXd over time, using a pH-sensitive pHrodo Red conjugate, which is nonfluorescent outside the cell in neutral pH, but fluoresces once it is internalized and processed in the endosomes and lysosomes that have a more acidic environment (16). The cells were pretreated for 8 hours either with vehicle, gefitinib, or osimertinib, after which, the pHrodo-labeled control IgG-ADC or HER3–DXd was added to the cells, and the internalization was measured every 2 hours using live-cell imaging (Fig. 3A). The 8-hour time point was selected to avoid possible nonspecific internalization caused by EGFR inhibitor–induced cell death. Cell lines with different baseline HER3 levels were tested (Fig. 3B).

In all three cell lines, HER3–DXd was internalized to a greater extent than the control IgG-ADC, suggesting that the internalization was dependent on antibody binding (Supplementary Fig. S3A). In both DFCI-243 and H1975 cell lines, there was a 3- to 5-fold higher intake of HER3–DXd over time in cells pretreated with osimertinib, compared with vehicle (Fig. 3D and E; Supplementary Fig. S3A). In the HCC4006 cell line, with very high baseline HER3 expression, there was an increase in the ADC intake, but the fold increase in the internalization was less compared with the two lower HER3-expressing lines (Fig. 3C–E; Supplementary Fig. S3B–C). In the DFCI-243 cell line expressing low baseline HER3, osimertinib pretreatment significantly increased the uptake of HER3–DXd over time (Fig. 3F and G). DFCI-243 harbors EGFR T790M gefitinib resistance mutation, and gefitinib treatment did not lead to enhanced HER3–DXd internalization, suggesting that the increased uptake was dependent on target engagement with EGFR. Importantly, EGFR TKI pretreatment did not increase IgG-ADC internalization in any of the cell lines (Supplementary Fig. S3D), suggesting that the increase in internalization was specific to the EGFR inhibition–induced HER3 and not due an increase in nonspecific internalization.

HER3–DXd is linked to a topoisomerase I inhibitor (exatecan derivative (DXd), which induces DNA damage following internalization and lysosomal cleavage. To further study the potential therapeutic effect of combining osimertinib with HER3–DXd combination treatment, we evaluated the impact of the single agents and the combination on DNA damage. The combination treatment increased the levels of activating phosphorylation in ATM kinase in all three cell lines, followed by increased phosphorylation of its downstream target H2aX, indicating higher amount of DNA double-strand breaks (DSB) more compared with the single agents (Fig. 3G and H; Supplementary Fig. S3E). In addition, the combination led to a greater increased in both cleaved caspase-3 and PARP. We further measured the amount of DSB (pH2aX) and monitored cell proliferation (Ki-67 positivity) using immunofluorescence. Indeed, we observed that the osimertinib/HER3–DXd combination induced more DSB damage than the single agents alone and led to decrease in Ki-67–positive cells (Fig. 3I–M). Collectively, these findings demonstrate that EGFR inhibitor pretreatment enhances the cellular uptake of HER3–DXd, resulting in a greater degree of DNA damage.

To study the therapeutic potential of the osimertinib/HER3–DXd combination, we tested the efficacy of HER3–DXd in EGFR-mutant cell lines with and without osimertinib pretreatment. The cells were first pretreated with osimertinib for 24 hours, followed by continuous treatment with both osimertinib and HER3–DXd. In both the HCC406 and H1975 cells, the combination was more effective than each of the single agents (Fig. 4A and B). The DFCI-243 was quite sensitive to single-agent osimertinib and as such we compared the regrowth of the cells following drug withdrawal and noted a significant difference in the regrowth for the combination compared with osimertinib alone (Fig. 4C). Consistent with the effects on growth, we observed an increase in apoptosis in the combination groups compared with the single agents (Fig. 4D–F).

Our studies suggest that on-target EGFR inhibition is necessary for HER3 upregulation and as such for the efficacy of the osimertinib/HER3–DXd combination. We tested PC9 cells resistant to gefitinib (PC9 GR) or osimertinib (PC9 T790M/C797S) for their HER3 expression and for the efficacy of the osimertinib/HER3–DXd combination. We noted that HER3 membrane and total expression was dependent on EGFR target engagement, because gefitinib did not lead to increased membrane/total HER3 expression in PC9 GR cells, whereas osimertinib did not increase HER3 in PC9 T790M/C797S cells (Fig. 4G and H). Same pattern was observed in the DFCI-243 cells that were made resistant to osimertinib by treating the cells with increasing doses of osimertinib (Supplementary Fig. S4A). PC9 cells were very sensitive to single-agent osimertinib and HER3–DXd and, as such, there was only a slight increase in efficacy following treatment with the combination of the two drugs in both the PC9 and PC9 GR cells. In contrast, there was no increase in efficacy of the combination in the PC9 T790M/C797S cells, suggesting that on-target EGFR inhibition is important for the combination efficacy (Supplementary Fig. S4B–S4D).

We next studied the efficacy of the osimertinib/HER3–DXd combination using ex vivo 3D-cultured PDX tumor fragments. PDX tumors were isolated from mice, and tumor-derived fragments were treated with either single agents or the combination analogous to the cell lines. Viability was measured on days 7 and 12. We observed significant effect of the combination at day 12 in both the DFCI-282 and DFCI-243 models compared with the single agent (Figs. 4I–K), In contrast, in the DFCI-249 model, HER3–DXd was effective as a single agent, and there was no significant added benefit from osimertinib pretreatment.

The in vitro findings suggested that the osimertinib/HER3–DXd combination should be further evaluated in an in vivo setting. To first evaluate the potential toxicities of the combination, nontumor-bearing mice were treated with osimertinib (25 mg/kg, daily) and HER3–DXd (10 mg/kg, weekly) for 21 days. We observed no significant toxicities of the combination as measured by mouse body weights and by examining complete blood counts (Supplementary Fig. S5A–S5C). To evaluate the combination efficacy, we chose DFCI-243 PDX model, since in the DFCI-243–derived cell line we detected low HER3 levels (Fig. 3B), modest single-agent HER3–DXd activity (Fig. 4C) and added benefit from the osimertinib combination in both 2D cell line and 3D-cultured PDX-derived tumor fragments (Fig. 4C and Fig. 4G–I). Like in the in vitro models, the mice were first treated with osimertinib and 24 hours following which they were treated with continuous daily osimertinib (10 mg/kg) in combination with weekly HER3–DXd (3 mg/kg or 10 mg/kg) for 28 days (Fig. 5A). We saw tumor growth delays with both single-agent osimertinib and HER3–DXd (Fig. 5B–D); however, all the osimertinib-treated tumors regrew (Fig. 5C) after the drug withdrawal, as did the majority of the HER3–DXd-treated tumors (Fig. 5D). The combination was associated with a greater maximal tumor reduction (Supplementary Fig. S5D) and the fewest number of mice exhibiting tumor regrowth following drug withdrawal (Fig. 5E; Supplementary Fig. S4E–S4G). Interestingly, in the combination group using the lower HER3–DXd dose (3 mg/kg) 2/8 tumors were able to regrow, whereas in the higher dose (10 mg/kg HER3–DXd) group only 1/8 tumor regrew. The difference in regrowth between single-agent HER3–DXd and osimertinib + HER3–DXd groups was statistically significant.

To test the efficacy of the combination in a second model, we chose HCC4006, which has moderate single-agent HER3–DXd sensitivity in vitro (Fig. 4A). The HCC4006 model was more sensitive to single-agent osimertinib than the DFCI-243 model (Fig. 5F and G) but had variable responses to the single-agent HER3–DXd (Fig. 5H). Greatest maximal tumor reduction and slowest tumor regrowth after withdrawal were again associated with the combination group (Fig. 5I; Supplementary Fig. S5H). The tumor growth difference between the single-agent treatment groups and the combination groups was statistically significant (Supplementary Fig. S54I). The difference in regrowth between single-agent HER3–DXd and osimertinib + HER3–DXd groups was statistically significant (Supplementary Fig. S5J).

Targeting HER3 represents an attractive strategy to treat EGFR inhibitor–resistant EGFR-mutant NSCLC. HER3 is expressed in the majority of EGFR-mutant cancers, and HER3 is not a known resistance mechanism to EGFR kinase inhibitors. In addition, as resistance mechanisms can be diverse, or sometimes no known or targetable mechanism is identified, targeting HER3 represents a practical and uniform approach to treat drug-resistant cancers (4, 17). HER3–DXd is undergoing clinical development as a single agent in EGFR inhibitor–resistant EGFR-mutant cancers (NCT03260491), and efficacy (response rate: 39% and progression-free survival 8.2 months) has thus far been observed (18). However, as efficacy of HER3–DXd is not uniform, we aimed to develop a strategy that could potentially enhance the efficacy of single-agent therapy.

Our studies demonstrate that EGFR inhibition not only leads to increased HER3 expression but that this is associated with an increase in HER3–DXd internalization, DNA damaging effects, and efficacy in vitro and in vivo. Previous studies of HER2 TKI lapatinib and EGFR/HER2 targeting TKI poziotinib have demonstrated that TKIs can induce increased cell-surface accumulation of HER2 by reduced receptor ubiquitination, and poziotinib-induced HER2 was found to sensitize cells to HER2 ADC T-DM1 in various cancer types (19, 20). Similarly, cotreatment with irreversible pan-HER inhibitors neratinib or afatinib enhanced HER2 ubiquitination and receptor internalization, resulting in increased T-DM1 efficacy in preclinical models of ERBB2-amplified lung cancer (21), suggesting that TKI-induced receptor internalization could be a generalized strategy to enhance efficacy of ADCs. However, the mechanistic basis, following TKI treatment, differs for HER2 and HER3.

There are several potential clinical scenarios where a combination of osimertinib and HER3–DXd could be evaluated including in patients who have developed osimertinib resistance. Our findings however suggest that EGFR inhibition is necessary for both an increase in HER3 membrane expression and for the combination efficacy of the two drugs. As such, patients who have developed an on-target EGFR mutation as their resistance mechanism to osimertinib, such as EGFR C797S, would not be expected to benefit from this combination. The optimal clinical scenario for the combination is where EGFR-mutant cancers are sensitive to the single agents.

Although both osimertinib and HER3–DXd can independently, and through different mechanisms, be effective in EGFR-mutant cancers, our findings suggest that the combination maybe even more effective given this unique interaction. At the moment, single-agent osimertinib is the standard of care for advanced EGFR-mutant NSCLC. However, the efficacy of osimertinib is limited by the development of acquired drug resistance. The combination of osimertinib and HER3–DXd may thus be one way to limit or delay resistance. These preclinical studies have led to the initiation of a phase I clinical trial combining osimertinib with HER3–DXd (NCT04676477). The study will include evaluation of the combination of osimertinib and HER3–DXd in patients previously treated with osimertinib and those who are EGFR inhibitor naïve.

Y. Zhao reports personal fees from Dana-Farber Cancer Institute during the conduct of the study. C. Yu reports other support from Daiichi Sankyo, Inc. during the conduct of the study and other support from Daiichi Sankyo, Inc. outside the submitted work. Y. Kamai reports personal fees from Daiichi Sankyo Co., Ltd. outside the submitted work. C.P. Paweletz reports other support from Daiichi Sankyo during the conduct of the study. P.C. Gokhale reports grants from Marengo Therapeutics, Epizyme Inc, Daiichi Sankyo, and Foghorn Therapeutics outside the submitted work. P.A. Jänne reports grants and personal fees from Daiichi Sankyo during the conduct of the study; grants and personal fees from AstraZeneca, Boehringer-Ingelheim, and Eli Lilly, personal fees from Pfizer, Roche/Genentech, Chugai, LOXO Oncology, SFJ Pharmaceuticals, Voronoi, Biocartis, Novartis, Sanofi, Takeda Oncology, Mirati Therapeutics, Transcenta, Silicon Therapeutics, Syndax, Nuvalent, Bayer, Esai, Allorion Therapeutics, Accutar Biotech, and AbbVie, and grants from PUMA, Revolution Medicine, and Astellas outside the submitted work; in addition, P.A. Jänne has a patent for EGFR mutations issued and licensed to LabCorp. No disclosures were reported by the other authors.

H.M. Haikala: Conceptualization, formal analysis, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. T. Lopez: Formal analysis, investigation, and visualization. J. Köhler: Formal analysis, investigation, and visualization. P.O. Eser: Formal analysis, investigation, and visualization. M. Xu: Investigation. Q. Zeng: Investigation. T.J. Teceno: Investigation. K. Ngo: Investigation. Y. Zhao: Investigation. E.V. Ivanova: Investigation. A.A. Bertram: Resources and visualization. B.A. Leeper: Resources and investigation. E.S. Chambers: Resources and investigation. A.E. Adeni: Resources and investigation. L.J. Taus: Investigation. M. Kuraguchi: Investigation. P.T. Kirschmeier: Resources, supervision, and funding acquisition. C. Yu: Conceptualization and supervision. Y. Shiose: Supervision and investigation. Y. Kamai: Supervision and investigation. Y. Qiu: Conceptualization and supervision. C.P. Paweletz: Resources, supervision, and funding acquisition. P.C. Gokhale: Conceptualization, supervision, funding acquisition, writing–review and editing. P.A. Jänne: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This study was supported by the American Cancer Society (CRP-17–111–01-CDD to P.A. Jänne), the National Cancer Institute (R35 CA220497 to P.A. Jänne), and Daiichi Sankyo (to P.A. Jänne). H.M. Haikala was supported by Sigrid Jusélius Foundation, The Finnish Cultural Foundation, and Orion Research Foundation. C.P. Paweletz had additional funding from the Expect Miracles Foundation and the Robert A. and Renée E. Belfer Family Foundation.

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.