Integrative whole-genome and transcriptome analysis of HER2-amplified metastatic breast cancer

Amplification of the ERBB2 gene, encoding human epidermal growth factor receptor 2 (HER2), occurs in approximately 15% of primary breast cancers (PBC) and causes overexpression of the protein kinase receptor HER2. Among patients with a HER2-amplified (HER2-positive/HER2+) tumor, targeted monoclonal antibodies binding and blocking this receptor of which trastuzumab is the most commonly administered proved to be very effective [1], resulting in a major increase in survival in this inherently aggressive breast cancer subtype [2]. As a result, HER2+ tumors constitute a distinct subgroup from a clinical perspective. In contrast, whole-genome sequencing (WGS) and whole-exome sequencing (WES) data from PBC revealed that HER2+ tumors are heterogeneous with regard to their molecular characteristics and often cluster with all other subtypes [3].

At the protein level, overexpression of HER2 in metastatic breast cancer (MBC) compared PBC to is rather stable [4]. Despite very pronounced and durable responses to regimens containing HER2-targeted agents, resistance is not uncommon. For example, around 40% of HER2 + MBC patients treated with dual HER2 blockade combined with docetaxel in the first line of therapy according to current clinical guidelines experience progression within the first year of treatment [5, 6]. The stable overexpression of HER2 despite disease progression suggests that mechanisms other than those involving downregulation or inactivation (e.g. via mutation) of the receptor are responsible for resistance. Several alternative trastuzumab-related mechanisms have been proposed in the literature, including activating ERBB2 mutations, alternate receptor cleavage and epitope masking, but also alterations in the downstream pathways and activation of alternative signaling pathways [7, 8]. Furthermore, therapy resistance might be caused by resistance against the cytotoxic backbone, given the observation from clinical data that admission of trastuzumab beyond progression has clinical benefit [9, 10].

Results from targeted sequencing of a panel of genes in both primary and metastatic HER2+ tumors revealed that there was a significant enrichment of mutations in the MAPK pathway in metastatic tumors indicating that in response to treatment with targeted therapy, a proportion of HER2+ cancers switch from PI3K/AKT signaling to MEK/ERK signaling [11]. As opposed to targeted data, genome wide data might yield even more insight into resistance mechanisms, but to date there is no report on a large and clinically annotated cohort of metastatic HER2+ MBC.

The present study therefore uses WGS data from metastatic breast cancer and aims to compare the observed alterations and mutational processes with publicly available WGS data of unpaired primary breast cancer. Additionally, we investigate the associations between observed alterations in metastatic cases and response to HER2-targeted therapy to gain more insight into resistance mechanisms. Finally, we perform integrative genomic and transcriptomic analyses to evaluate downstream HER2-driven signaling in HER2-amplified and non-amplified tumors.

In the present work, we performed an in-depth investigation into the genomic and transcriptomic characteristics of HER2-amplified MBC biopsies incorporating available clinical data on (prior) therapy and treatment response. To our knowledge, this is the first large cohort that reports specifically on WGS and RNAseq of HER2+ metastases, also including a comparison to publicly available WGS data of primary BC tumors.

In general, there were few genomic differences observed comparing (unpaired) metastatic vs primary HER2+ cases. The differences that were observed, namely a higher rate of TP53 mutations and a higher contribution of signature DBS2, are either not subtype-specific (TP53) or does not appear highly affected in HER2+ MBC (DBS2) [25]. Similarly, no prominent changes were observed in the genome of patients that had been treated prior to biopsy with anti-HER2 therapy compared to treatment-naïve patients, implicating that resistance mechanisms are either diverse or driven by the therapeutic backbone of chemotherapy or endocrine therapy.

Analysis of the genomic characteristics in relationship to post-biopsy anti-HER2-targeted treatment revealed a strong predictive model, which could be partly validated in independent public data [11]. Predictors for fast progressive disease (< 9 months) were the presence of PIK3CA and CDK12 mutations, an amplified region on chromosome 8 (8p11.23 with ZNF703 as closest gene) and a high contribution of mutational signature DBS3. The impaired outcome on HER2-targeted therapy in the presence of PIK3CA mutations has been shown in preclinical and clinical literature many times before and has prompted clinical trials, targeting the PI3K pathway and HER2 simultaneously [11, 29‐31]. A phase II trial with the pan PI3K inhibitor buparlisib showed a limited clinical benefit rate of 14%, accompanied by grade 3–4 toxicity in 70% of patients, and did not meet its primary target [32]. The less toxic alpha subunit-specific PI3K inhibitor alpelisib, however, has shown efficacy and tolerability in ER+- and HER2-negative MBC in the SOLAR-1 trial, especially for PIK3CA-mutated cases [33, 34]. The efficacy of alpelisib combined with anti-HER2-targeted therapy is now being investigated in PIK3CA-mutated patients in the ALPHABET trial (NCT05063786) [35]. CDK12 amplification has been linked to anti-HER2 therapy resistance before [36]. Amplification of (the region surrounding) ZNF703 was not described for HER2+ breast cancer, but has been linked to luminal B breast cancer before, which is known to be inherently aggressive [37]. Interestingly, the DBS3 signature has not been linked to breast cancer before [15]. Although in our set only 8 tumors (12%) harbored a contribution of > 10%, this was the strongest predictive factor in our multivariable model. Not much is known on the role of DBS3 in breast cancer. In gastrointestinal cancers, this signature has been linked to a hypermutator phenotype mediated by POLE mutations, which were not significantly enriched in our MBC cohort [38].

Supervised clustering using expression levels of a set of HER2-associated genes did result in a HER2-specific sample cluster. Remarkably, this cluster did not only contain virtually all HER2+ samples, but also included the majority of ERBB2-mutated samples and a substantial number of cases with an ERBB2 CN below our threshold for HER2 positivity. In 36% of the ERBB2 low CN metastatic samples, we observed a low expression of ESR1, even though their primary tumor was ER+ and as such, endocrine therapy was received. Further analyses of the gene expression data showed that samples from this HER2-specific cluster had significantly higher expression of genes that are related to the MAPK pathway, which was not observed in PBC. It can therefore be hypothesized that both resistance to endocrine therapy and expression or upregulation of the ERBB kinase pathways is correlated with MAPK upregulation. This observation is in concordance with the work of Razavi et al., who showed enrichment of MAPK alterations in a large cohort of endocrine-resistant HER2-negative tumors [28]. Together, our results provide a rationale for targeting the MAPK pathway in a subset of patients, resistant to endocrine therapy or to anti-HER2 therapy. Wagle et al. [27] showed that cell lines with an upregulated MAPK pathway on RNA level are sensitive to MAPK inhibition, independent of mutational status. Since targeting multiple pathways in patients is often limited by toxicity, gene expression analysis may provide a better snapshot of the most dominant tumor driving pathway, complementary to using specific mutations for matching patients with a specific treatment. Consequently, RNA sequencing may be indispensable to predict response on therapy and to help prioritize the best strategy to pursue, which is also substantiated by a recent large pan-cancer study [39, 40]. With the knowledge that the treatment landscape for breast cancer is changing rapidly [41], it will be of important clinical significance to use such integrative analyses to distinguish between resistant tumors primarily driven by ERBB2 aberrations and those driven by downstream pathways.

One of the main players in the rapidly changing treatment landscape is the highly efficacious anti-HER2-targeted antibody–drug conjugate trastuzumab–deruxtecan (T-DXd). On the basis of the DESTINY trials, this agent additionally gained approval for the HER2-low-expressing subgroup [42, 43]. However, the current clinical workflow defining HER2 status, involving IHC and/or FISH does not seem discriminative enough the define this subgroup [44, 45]. Moreover, the recently published DAISY trial even showed that although HER2 expression is the most determinant of T-DXd efficacy, objective responses could also be achieved in the HER2 ‘ultra-low’ subgroup [46]. In line with the foregoing results, we observed comparable ERBB2 expression across the defined clusters. We did not have formal IHC results from the metastatic lesions to make a comparison, but from our exploratory analysis we have no reason to assume that the clinical HER2 low subgroup can be distinguished on genomic level.

Although this dataset is unique with respect to its size and the availability of paired WGS and RNAseq data, the pathological assessment of the HER2 status of the metastatic tissue was lacking. Therefore, the HER2+ amplification status was inferred from the genomic data. To prevent false positives, we set a stringent copy number threshold for calling a sample HER2+, potentially resulting in a number of false negative cases. We estimated the false negative rate at 5%, derived by the number of patients who were HER2− according to our copy number threshold, but who were treated with anti-HER2 therapy prior to the metastatic biopsy, suggesting a HER2+ status for the primary tumor. However, using this stringent cutoff, WGS data from 68 truly HER2+ MBC could be compared to publicly available WGS data from PBC. Another limitation was the heterogeneity in pretreatment, especially in the drugs that accompanied anti-HER2 treatment, although we did correct for this in our analyses. Next, for clinical practice, repeated biopsies are most likely not always feasible. Lastly, our cohort is large, but some of our analysis are underpowered to draw conclusions on the absence of HER2-specific aberrations at the genomic level.