Frequent Overexpression of HER3 in Brain Metastases from Breast and Lung Cancer

HER3/ErbB3 belongs to a family of receptor tyrosine kinases with oncogenic properties (1). In contrast to other known family members, such as HER2 (ErbB2), HER3 has an impaired kinase activity (2). Therefore, heterodimerization with other HER members, preferentially with HER2, are necessary for its own oncogenic potential (3). HER3 overexpression has been observed in several tumor types such as breast, ovarian, prostate, gastric, melanoma, colon, lung, and head and neck cancer (4–11). A meta-analysis performed in 2012 by Ocana and colleagues evaluated 12 studies with a median percentage of HER3 overexpression in cancers of 42.2% (17.0–69.7; ref. 12). Solid tumors with the highest frequency of HER3 overexpression included colorectal (69.0%), melanoma (65.0%), and gastric (59.0%) cancer. Moreover, increased HER3 expression was postulated as a potential cause for therapy resistance against other HER-targeted therapies (13, 14). To overcome therapeutic failure, approaches to directly target HER3 are emerging including abrogating the kinase activity of its dimerization partners by using small-molecule inhibitors or by directly targeting its extracellular domain with mAbs (15, 16).

HER3 is of particular interest in the context of brain metastases (BM) formation as HER3 signaling has been shown to be involved in metastatic spread and colonization to the brain (17, 18). BM are a major clinical challenge that affect approximately 20% of all patients with cancer. Higher incidences occur in certain tumor types like EGFR-mutated non–small cell lung cancer (NSCLC) or triple-negative (TNBC) or HER2-positive (HER2⁺) breast cancer (BCa), where BM tend to develop in up to 40% of cases (19). Treatment options are still limited, focusing on radiotherapy and, if applicable, surgical resection of the tumor mass (20). While initially effective, these interventions are associated with frequent neurologic impairment, hence significantly reducing quality of life, though rarely capable to improve prognosis beyond 2 years. Therefore, additional therapeutic options including systemic treatment are warranted to improve prognosis.

Molecular targets for personalized treatment approaches are of particular interest for BM treatment. Indeed, up to 50% of BM were shown to harbor potentially clinically relevant molecular alterations, which were not detected in their matched primary tumors (21). Recently, targeted treatments including antibody–drug conjugates (ADC) resulted in promising clinical activity in patients with BM (22–24). Furthermore, patritumab-deruxtecan (HER3-DXd), a novel ADC against HER3, demonstrated activity and safety in a phase I clinical trial including patients with metastatic NSCLC harboring an EGFR-activating mutation (25). In addition, early results of a phase I/II clinical trial in patients with breast cancer of any subtype showed promising efficacy of HER3-DXd as well (26).

In the current study, we aimed to investigate the frequency of HER3 overexpression in BM of patients with breast cancer and NSCLC as those entities most frequently developing BM. Considering the influence of immune cells on the efficacy of mAb-based therapies, we also evaluated the tumor microenvironment for its immune cell composition within the BM. We further used methylation profiles of HER3-positive (HER3⁺) and -negative (HER3⁻) specimens to identify relevant dysregulated pathways and gain further biological insight for future trial designs.

Patients diagnosed with BM from breast cancer or NSCLC between 1990 and 2020 at the Medical University of Vienna (Vienna, Austria) were identified. Formalin-fixed and paraffin-embedded (FFPE) material available for scientific purposes was restored from the pathology and neuropathology department, respectively. Clinical data were collected retrospectively using available medical records. This study was approved by the Ethics Committees of the Medical University of Vienna (approval no. 1975/2021) in accordance with the Declaration of Helsinki. The need for informed consent was waived by the competent regulatory authorities due to the retrospective nature of the study.

A total of 4-μm-thick shavings were prepared from FFPE tissue blocks of BM and primary tumor samples. IHC staining was performed on a Ventana Benchmark ULTRA machine (Roche Diagnostics) using the following antibodies: HER3 (Cell Signaling Technology, #12708, RRID:AB_2721919, 1:100), HER2 [PATHWAY HER2 (4B5), Roche catalog no. 790-4493, RRID:AB_2921204, ready-to-use (RTU)], Estrogen Receptor [CONFIRM ER (SP1), Roche, catalog no. 790-4324, RRID:AB_2857956, RTU], Progesteron Receptor [CONFIRM PR (1E2), Ventana Medical Systems catalog no. 790-4296, RRID:AB_2335976, RTU], CD3 (Lab Vision catalog no. RM-9107-S, RRID:AB_149922, 1:200), CD8 (Agilent catalog no. M7103, RRID:AB_2075537, 1:100), CD68 (Agilent catalog no. M0814, RRID:AB_2314148, 1:500). Heat-induced antigen retrieval was achieved by incubation with the Tris-EDTA–based Cell Conditioning 1 (CC1) buffer (Roche Diagnostics) for all antibodies. HER3 signal intensity was amplified by the application of the Ventana Amplification Kit (Roche). Protein expression was visualized using the UltraView DAB IHC Detection Kit (Ventana) and counterstained with Hematoxylin and Bluing Reagent (Ventana). A protocol including specific conditions for each antibody is provided in Supplementary Table S1. Specificity of the HER3 antibody was further validated by staining two breast cancer cell lines with known HER3 expression levels (Supplementary Fig. S1A).

Slides stained for immune cell markers were digitalized on a slide scanner for semiquantitative analysis using the Definiens Tissue Studio software v3.0 (Definiens Inc., RRID:SCR_023001) as described previously (27). Hormone receptors and HER2 expression were reevaluated in brain metastasis specimens by microscopic evaluation using established criteria (28, 29). Briefly, nuclear staining of estrogen receptor (ER) and progesteron receptor (PR) in at least 1% of tumor cells was considered hormone receptor–positive. Luminal tumors had to show positive expression for either ER and/or PR but they were not further grouped into A and B subtypes. HER2 and HER3 stainings were only considered if complete, strong membranous staining pattern of at least 10% tumor cells was present. HER2 expression was further graded into IHC 0, 1+, 2+, and 3+ while only IHC3+ and IHC2+/ISH+ tumors were termed HER2⁺, IHC1+ and IHC2+/ISH− as HER2-low, and IHC0 as HER2-negative (HER2⁻). We did not perform ISH analysis in IHC2+ BM samples instead adopted subtyping from primary samples. HER3 staining was graded according to complete, membranous staining intensity into negative, low and high based on staining intensity following aforementioned HER2 criteria (Fig. 1A). For analysis, HER3-low and -high tumors were combined into HER3⁺. Two independent observers evaluated the IHC-stained slides.

Genomic DNA was isolated from tumor cell–enriched (≥70%) FFPE blocks (n = 71) using the Maxwell FFPE Plus Kit (Promega) following the manufacture's instruction. Isolated DNA was quantified on a Qubit 2.0 Fluorometer (Thermo Fisher Scientific) using the associated assay kits and 500 ng were subsequently bisulfite converted using the EpiTect Fast Kit (Qiagen) according to the manufacturer's manual. Before each sample was applied to the Infinium EPIC Methylation BeadChip (Illumina) protocol, the Infinium HD FFPE Restore Kit (Illumina) was performed to improve sample quality. Processed microarray chips were scanned on an iScan device (Illumina, RRID:SCR_016388) to obtain raw fluorescence intensity (.idat) files.

Intensity data were imported into R using the RnBeads package (v2.12.2, RRID:SCR_010958) following the suggested pipeline which performs assay quality validation, probe filtering, and normalization steps as well as differential methylation analysis (30). Probes on sex chromosomes were excluded from subsequent analysis as well as SNP-enriched and cross-reactive probes. The Greedycut algorithm was used to remove probes and samples of highest impurity based on a detection P value above 0.05. The SWAN algorithm was chosen to normalize microarray data (31). Differentially methylated sites were selected on the basis of calculated rank cutoffs by RnBeads. P values on the site level were computed using the limma method. Dimensional reduction of the methylation data was calculated and visualized with the umap package for R. Heatmaps were generated using the ClustVis online tool and pathway enrichment analyses was performed utilizing the graphical web application ShinyGO v.0.76.3 (32, 33).

Differences between two groups were tested using a two-tailed Mann–Whitney test while differences between multiple (≥3) groups were tested using Kruskal–Wallis test with Dunn post hoc test for multiple comparison. The independence of categorical variables was tested by either χ² or Fisher exact test depending on sample size. An adjusted P value below 0.05 was considered statistically significant. Statistical analyses were performed using GraphPad Prism, Version 9.4.0 (RRID:SCR_002798), IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp, RRID:SCR_019096), and the R statistical environment v.4.1.1 (R Core Team; R Foundation for Statistical Computing, RRID:SCR_001905). Survival analyses were computed with the R package survminer v.0.4.9. Overall survival (OS) was defined as the time of BM diagnosis to all-cause death or last follow-up. A log-rank test was performed to compare survival of groups and a Cox regression model was performed to estimate HRs.

Raw data of DNA methylation analysis are publicly available at the Gene Expression Omnibus database respiratory under the accession number GSE220826. Further data will be shared upon reasonable request to the corresponding author.

We identified 696 patients with BM of breast cancer (n = 191) and NSCLC (n = 505) diagnosed at our institution in the last 30 years. From 180 patients, sufficient tissue material was available for scientific purposes and included into this study. The ethnicity of the overall cohort consisted exclusively of white Europeans. All patients with breast cancer were female, while 69.6% (32/48) of the patients with NSCLC were male. Of the 132 BCa-BM, 49 (37.1%) were of luminal subtype, 31 (23.5%) were HER2⁺, and 52 (39.4%) were TNBC. A total of 34/48 (70.8%) NSCLC-BM derived from lung adenocarcinomas, 12/48 (25.0%) had squamous, and 2/48 (4.2%) a mixed histology. While 87% of patients with NSCLC had only one BM at time of diagnosis, this only applied to 44.7% of patients with breast cancer. Matched extracranial samples were available from 46/48 patients with NSCLC which all derived from primary lung tumors except of one pleural carcinosis biopsy. Of the patients with NSCLC, 15/48 patients received adjuvant systemic treatment and 5/48 adjuvant radiotherapy of the primary, extracranial tumor (4/5 a combined radiochemotherapy of the primary tumor). In the breast cancer cohort, 87/132 patients received adjuvant systemic therapy and 92/132 adjuvant radiotherapy of the primary, extracranial tumor (69/132 a combined radiochemotherapy of the primary tumor). Further patients’ characteristics are summarized in Table 1. Patients’ characteristics according to HER3 expression levels are summarized separately in Table 2.

The expression pattern of HER3 varied between tumor samples from completely negative to low and high intensity (Table 1; Fig. 1A). Overall, 99/132 (75.0%) BCa-BM and 33/48 (68.8%) NSCLC-BM presented with HER3 positivity (Fig. 1B; Table 1). A total of 10/14 (71.4%) of NSCLC-BM with a squamous or mixed histology and 23/34 (67.6%) of the adenocarcinoma-based NSCLC-BM were HER3⁺. HER3 expression significantly differed between breast cancer subtypes: 27/31 (87.1%) HER2⁺ BCa-BM patients presented with HER3 expression, in contrast to 41/49 (83.7%) luminal breast cancer and 31/52 (59.6%) TNBC BM (P < 0.01; χ² test; Fig. 1C). A total of 37/61 (60.6%) patients with HER2⁻ breast cancer presented with HER3 expression and 31/37 (83.8%) of HER2-low patients were positive for HER3 (Fig. 1D; Supplementary Fig. S1B). Importantly, specimens with high levels of HER3 expression were also different according to breast cancer subtype: 19/31 (61.3%) in HER2⁺, 20/49 (40.8%) in luminal, and 16/52 (30.8%) in TNBC (P = 0.02; χ² test) BMs. While the amount of HER3 expression indeed decreased in association with HER2, 50% of HER2⁻ (IHC score 0) BCa-BM were still positive for HER3, with almost a third of them even expressing high levels of HER3 (Supplementary Fig. S1C).

From 46/48 patients (95.8%) with NSCLC, matched extracranial tumor material for the analysis of HER3 expression was available. HER3 expression was more commonly observed in BM than in extracranial tumors (34/47 vs. 19/47; P = 0.003; Fisher exact test; Fig. 1E); high HER3 expression was more frequently observed in BM than in extracranial tumors as well (48.9% vs. 8.5%; P < 0.001, Fisher exact test). A similar increase of expression was observed for HER2 in BM samples. Here, 22/48 (45.8%) of NSCLC-BM where HER2-low compared with 7/46 (15.2%) of NSCLC primary tumors (Supplementary Fig. S1D). We did not detect any HER2 IHC 3+ cases in the NSCLC cohort. Of the HER2-low NSCLC-BM, 19/22 (86.4%) were also positive for HER3 (Supplementary Fig. S1E). Given the long inclusion period of our study we investigated whether the year of sample acquisition and the results of HER3 IHC staining interact with each other. However, the decade of diagnosis did not differ significantly between HER3-high, HER3-low, and HER3⁻ samples (P = 0.3133, χ² test; Table 2).

Median OS from BM diagnosis was 10 months [interquartile range (IQR) = 3–21] in patients with breast cancer and 14 months (IQR = 6–45) in patients with NSCLC (Table 1). Regarding different breast cancer subtypes, patients with luminal breast cancer and HER3 expression presented with improved OS compared with patients without HER3 expression [P = 0.032, HR = 0.39, 95% confidence interval (CI) = 0.17–0.92], while no difference was observed in other breast cancer subtypes (Fig. 2A–C). By including the extent of expression into the analysis, only patients with high HER3 levels had longer OS (Table 2; Supplementary Fig. S2A–S2C). In patients with NSCLC-BM, no difference in survival was evident according to HER3 expression (P = 0.73, HR = proportional hazard assumption does not hold; Fig. 2D; Supplementary Fig. S2D).

Overall, there was no significant difference between CD3⁺ (P = 0.4774, Kruskal–Wallis test), CD8⁺ (P = 0.4464), and CD68⁺ (P = 0.3834) immune cell density in BM of patients with breast cancer with different HER3 expression (Fig. 3A). No statistically significant difference was observed in breast cancer subtypes either (Supplementary Fig. S3A–S3C). Similarly, no differences in immune cell infiltration based on HER3 status were observed in BM of patients with NSCLC, including CD3⁺ (P = 0.0821), CD8⁺ (P = 0.2848), and CD68⁺ (P = 0.9725) leukocytes (Fig. 3B). However, overall, the number of immune cells in the tumor microenvironment was higher in BM of patients with NSCLC compared with patients with breast cancer: CD3 (362 vs. 156 cells/mm², P < 0.001, estimated difference between means (est. MD) = 187.414, CI = 120.184–263.515, Mann–Whitney test), CD8 (131 vs. 67 cells/mm², P = 0.003, est. MD = 22.223, CI = 7.774–43.786), and CD68 (1,988 vs. 672 cells/mm², P < 0.001, est. MD = 1044.598, CI = 757.613–1277.654).

To gain further insight on potential molecular drivers of HER3 expression, we performed methylation analysis as an omics method providing a broad insight and directions for further in-depth analysis. To identify relevant pathways potentially affected by differing HER3 expression, we interrogate the DNA methylome of HER3⁺ and HER3⁻ BM. Overall, we analyzed 71 samples comprising of 16 patients with NSCLC (8 HER3⁺ and 8 HER3⁻ BM) and 55 patients with breast cancer (27 HER3⁺, 20 HER3⁻, and 8 HER3-low BM). A detailed description of the patient characteristics used for methylation analysis is shown in Supplementary Table S2. After data processing and clearing, 641,418 CpG sites remained for genome-wide methylation analysis. By performing a uniform manifold approximation and projection (UMAP) model for dimensional reduction using the top 10,000 CpG sites based on variance of beta values across the whole dataset, the BM samples separated into three major clusters: (i) a luminal/HER2⁺, (ii) a TNBC, and (iii) a NSCLC cluster (Fig. 4A). In addition, while HER3⁺ and HER3⁻ samples of TNBC and NSCLC-BM are not divisible within their clusters, only the HER3⁻ samples of HER2⁺ and luminal BM misclustered within TNBC and NSCLC-BM. Subsequently, we were focusing on a subtype-specific analysis to identify differentially methylated positions (DMP) between HER3⁺ and HER3⁻ BM. For luminal BM, we identified 66,999 (44,809 hypermethylated and 22,190 hypomethylated) DMPs, 56,673 (21,151 hypermethylated and 35,522 hypomethylated) for HER2⁺ BM, and 47,482 (21,901 hypermethylated and 25,581 hypomethylated) for TNBC-BM, but only 15 (six hypermethylated and nine hypomethylated) DMPs for NSCLC-BM indicating an inferior effect of HER3 expression in NSCLC (Fig. 4B). Next, we focused on the top 1,000 differentially methylated sites to define HER3-specific methylation signatures in each breast cancer subtype. A list of these sites for each subtype including relevant annotation is provided in Supplementary Tables S3–S5. Hierarchical clustering for each subtype separated HER3⁺ from HER3⁻ BM with only a single outlier in HER2⁺ and luminal breast cancer BM (Fig. 4C). Each detected CpG site was then annotated to its designated gene that were further used for pathway enrichment analyses to detect dysregulated pathways (Fig. 4D). Supplementary Table S6 summarizes these pathways giving additional relevant information. A recurrent hit, appearing in both, HER2⁺ and TNBC BM, was associated with Wnt signaling. Furthermore, EGF/EGFR signaling pathways including its known downstream PI3K/AKT signaling cascade appears deregulated in TNBC BM together with G protein–coupled receptor involved pathways. In Luminal BM, TGFβ signaling seems affected by differential methylation of selected target genes.

Furthermore, to determine whether HER3 expression may be regulated by DNA methylation, we specifically evaluated the methylation level of 33 selected CpG sites associated with the genomic locus of HER3 (ERBB3) on the Infinium Methylation EPIC microarray. However, we did not detect any differences between HER3⁺ and HER3⁻ BM including all analyzed samples indicating a DNA methylation independent regulation of HER3 expression in those samples (Supplementary Fig. S4A).

HER3 is of particular interest in the context of BM, as its expression is associated with BM development and serves as a promising therapeutic target for personalized therapy approaches. In the current study, we could detect widespread expression of HER3 in BM of patients with breast cancer and NSCLC. Indeed, more than two-thirds of BM samples presented with detectable HER3 expression of any intensity whereas almost half of them had high expression as defined by complete membranous staining of over 10% of tumor cells. BM from the HER2⁺ subtype presented most frequently with HER3 expression. However, also the majority of TNBC and NSCLC-BM samples showed HER3 expression. Given the lack of systemic treatment options particularly in TNBC BM so far, our data strongly support further development of HER3-directed therapies in the context of BM.

To gain more insight on functional aspects of HER3 expression in BM, we performed genome-wide DNA methylation profiling and pathway analysis. Here, we could verify subtype-specific methylation signatures as recently demonstrated by Capper and colleagues (34). This supports the diverse tissue origin of our BM samples and is an important quality measure of our data. In addition, methylation profiles of HER3⁺ and HER3⁻ samples did differ according to the underlying tumor subtype. The identified differentially methylation patterns are strongly associated with pathways known to facilitate and support brain metastatic outgrowth. For example, brain-derived neutrophic factor signaling including TRKB (NTRK2) appears dysregulated in HER3⁺/HER2⁺ BM and is associated with neurogenesis and synaptic development (35). The neutrophin receptor TRKB was shown to be highly activated in patient tissue and breast cancer cell lines and heterodimerization of HER2 and TRKB was postulated as a potential mediator for BM progression (36, 37). However, if TRKB could also serve as a potential binding partner of HER3 is yet unknown. Furthermore, we identified Wnt signaling to be dysregulated in BM of HER2⁺ and TNBC BM according to HER3 expression. Previously, the Wnt signaling pathway was identified as driver of BM in basal-like breast cancer (38). Indeed, HER3⁻ HER2⁺ BM samples clustered within the TNBC BM due to their similar methylation profile, supporting their potential biological relation. While we could not show direct pathways causing HER3 overexpression, we found no evidence that DNA methylation is involved in the regulation of HER3 expression in BM of patients with breast cancer and NSCLC.

HER3 was expressed more frequently in BM compared with the matched primary tumors of patients with NSCLC, supporting the importance of HER3 signaling for BM development. Interestingly, the rate of HER3 expression in our cohort of extracranial NSCLC samples appeared lower than that from other sources, ranging from 81% to 90% (25, 39). However, in their studies, Yonesaka and colleagues and Jänne and colleagues only included NSCLC with mutated EGFR indicating a correlation of the expression of different HER family members. Consistently, our observation that HER2⁺ breast tumors show the highest expression of HER3 coincide with other studies (40). Unfortunately, we were not able to include matched primary tumor samples from patients with breast cancer to compare HER3 expression in time and space of this tumor entity. Other reports investigating the expression of HER3 in primary breast tumors vary greatly, ranging from 15% to 67% (4, 41). However, preliminary data from other groups support our finding of increased levels of HER3 in BM (42). Da Silva and colleagues evaluated BM of patients with breast cancer in a smaller cohort in 2010 and found that 11/37 (29.7%) of the primary tumors, 22/37 (59%) of the matched metastases, and 13/21 (62%) of the unmatched BM (P = 0.019) showed positivity for HER3 by IHC (17). Taken together, these observations indicate HER3 as a critical mediator in brain metastatic spread, which could serve as a selective pressure to promote the outgrowth of HER3⁺ cells at this particular metastatic site. However, further mechanistic insight on the role of HER3 in the development of BM is warranted, particular in the context of new HER3-targeting ADCs. In line, preclinical data further support the importance of HER3 in the brain metastatic cascade, particularly in the context of coexpression with HER2. While HER2 represents an orphan receptor that lacks a ligand, neuregulin-1 (NRG1 or heregulin-1) binds to HER3 and leads to its dimerization and further activation (17). The heregulin-HER3-HER2 signaling was shown to promote the matrix metalloproteinase–dependent transendothelial migration of BM initiating cancer cells through the blood–brain barrier and thereby facilitate BM establishment (18). Indeed, levels of NRG-1 are particularly high in the brain and are secreted by both, microglial and neuronal cells (43–45). Therefore, further preclinical studies concentrating on the role of HER3 inhibition to prevent BM development would be of high interest.

A novel interesting therapeutic approach in cancer therapy involves conjugates of specific mAbs coupled to cytotoxic drugs (ADCs). HER3-directed ADCs like patritumab-deruxtecan have shown promising activity experimentally and in early clinical development (25, 46–48). Indeed, the correlation of HER3 expression and response to HER3-targeting ADCs is currently under investigation. The first published clinical study suggested a high response rate for the HER3 ADC patritumab-deruxtecan in patients with EGFR-mutated NSCLC (25). Although, this early clinical trial did not require HER3 expression as an inclusion criterion, membranous HER3 expression was detected retrospectively in all included patients. Therefore, the observed response rates toward the HER3-targeting ADC could be explained at least by an enrichment for HER3 expression in this particular study population. In the meantime, other current clinical trials define at least a low expression of HER3 as a mandatory biomarker for inclusion (26). Interestingly, response to the HER2-directed ADC trastuzumab-deruxtecan was observed for patients with very low HER2 expression as well, leading to its approval for patients with HER2-low breast cancer (49). Therefore, further investigation is required to define the predictive role of HER3 expression in the context of HER3-targeting ADCs. The next generation of ADCs has in comparison with earlier generation improved cleavable linker-payload systems, which enable stable drug delivery to cancer cells causing selective destruction of antigen-expressing cells (50). In addition, by passive diffusion of cytotoxic drug molecules into neighboring antigen-negative tumor cells a so-called bystander effect further increases the efficacy of novel ADCs (51). Given the recent reports of high intracranial activity of the HER2-directed ADC trastuzumab-deruxtecan in patients with HER2⁺ BCa-BM, ADCs might become a particular promising approach in this disease setting (22–24). The current data of high HER3 expression in BM and in particular of higher expression as compared with matched primary tumors, strongly support HER3 as a target for ADCs in systemic treatment of BM.

It has been shown by various clinical trials that the presence of tumor-infiltrating lymphocytes in HER2⁺ early and advanced breast cancer is associated with a favorable prognosis and predicts a beneficial anti-HER2 therapeutic response (52–54). We did not observe any HER3 expression–associated differences in the respective inflammatory profiles, instead a steady state across all cohorts was found. Despite this observation, the sufficient number of immune cells required for antibody-dependent immunogenicity remains unknown and the mechanism of ADCs may depend less upon antibody-dependent cellular cytotoxicity but on the activity of the cytotoxic agent.

HER3 is a kinase-defective receptor; therefore, its sole expression might have only limited oncogenic potential and therefore no impact on OS. More important than a prognostic impact is a potential predictive impact as a biomarker for response to HER3-targeting ADC. However, previous reports suggested a prognostic significance for HER3 in solid tumors, such as gastric cancer and prostate cancer (46, 55, 56). In contrast, we observed no consistent correlation of HER3 expression with survival prognosis. While certain reports indeed found HER3 levels do serve as prognostic factors in hormone receptor–negative breast cancer (57–59), we were not able to verify its role as a negative predictor for patients with BM of breast and lung cancer as only hormone receptor–positive (luminal) metastatic patients with breast cancer with high expression of HER3 were associated with better prognosis. An association of HER3 expression with OS was adjusted for established risk factors including age, gender, and graded prognostic assessment class. Here, the association of HER3 expression remained statistically significant in the subgroup of luminal breast cancer (HR = 0.39; 95% CI, 0.17–092; P = 0.032), while no association of the other investigated factors was observed. Interestingly, Lee and colleagues observed a similar favorable clinical outcome in patients with hormone receptor–positive breast cancer based on disease-free survival (60). Frankly, the long observational period covering three decades might have introduced confounder that limit definitive survival analyses. For example, standard lines of anticancer therapy have changed throughout the years for both, breast cancer and NSCLC. However, while we could not detect an influence of adjuvant chemotherapy given on OS in any of the investigated subgroups, we cannot deny that modern targeted therapy has generally prolonged survival of patients with cancer. Taken this heterogeneity into consideration including the rather small sample sizes per subgroup, any conclusion and generalization using a rather historical cohort mustbe taken with caution. Nevertheless, to the best of our knowledge, this study represents the most comprehensive work on HER3 expression in BM so far.

In conclusion, our data highlight the role of HER3 expression in BM and encourage further clinical development of HER3-targeting therapeutics in patients with BCa-BM and NSCLC-BM.

F. Eckert reports personal fees from Dr. Sennewald GmbH outside the submitted work. R. Bartsch reports personal fees from AstraZeneca, Eisai, Eli Lilly, Gilead, Grunenthal, MSD, Novartis, Pfizer, Pierre-Fabre, Roche, and Seagen and grants and personal fees from Daiichi outside the submitted work. G. Heller reports personal fees from X4 Pharmaceuticals outside the submitted work. M. Preusser reports personal fees and non-financial support from Bayer, Bristol-Myers Squibb, Novartis, Gerson Lehrman Group (GLG), CMC Contrast, GlaxoSmithKline, Mundipharma, Roche, BMJ Journals, MedMedia, AstraZeneca, AbbVie, Lilly, Medahead, Daiichi Sankyo, Sanofi, Merck Sharp & Dome, Tocagen, Adastra, Gan & Lee Pharmaceuticals, and Servier outside the submitted work. A.S. Berghoff reports grants from Daiichi Sankyo during the conduct of the study as well as grants, personal fees, non-financial support, and other support from Roche; personal fees and non-financial support from Merck, AstraZeneca, CeCaVa, Bristol Myers Squibb, and Daiichi Sankyo; and other support from Amgen and AbbVie outside the submitted work. No disclosures were reported by the other authors.

E. Tomasich: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. A. Steindl: Data curation, formal analysis, validation, investigation, writing–review and editing. C. Paiato: Data curation, software, formal analysis, investigation, methodology. T. Hatziioannou: Data curation, software, formal analysis, validation, visualization, methodology, writing–review and editing. M. Kleinberger: Software, formal analysis, investigation, methodology, writing–review and editing. L. Berchtold: Software, formal analysis, validation, writing–review and editing. R. Puhr: Software, formal analysis, validation, investigation. J.A. Hainfellner: Resources, formal analysis, validation. L. Müllauer: Resources, formal analysis, validation, methodology, writing–review and editing. G. Widhalm: Formal analysis, investigation. F. Eckert: Formal analysis, writing–review and editing. R. Bartsch: Conceptualization, formal analysis, investigation, writing–review and editing. G. Heller: Conceptualization, data curation, software, formal analysis, validation, visualization, methodology, writing–review and editing. M. Preusser: Conceptualization, resources, supervision, funding acquisition, writing–review and editing. A.S. Berghoff: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration.

The Austrian Federal Ministry for Digital and Economic Affairs, the National Foundation for Research, Technology and Development, and the Christian Doppler Research Association financially supported this study by research grants to M. Preusser. This study was further supported by a research grant of Daiichi Sankyo to A.S. Berghoff.

The financial support by the Austrian Federal Ministry for Digital and Economic Affairs, the National Foundation for Research, Technology and Development, and the Christian Doppler Research Association is gratefully acknowledged. This study was further supported by an unrestricted grant of Daiichi Sankyo.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.