Identifying the challenges of biomarkers that predict response to HER2-directed therapies

Abstract

The identification of the human epidermal growth factor receptor 2 (HER2) as a targetable marker of adverse prognosis along with the development of HER2-directed therapies, have significantly improved outcomes for women with HER2-overexpressing breast cancers. However, both resistance to treatment and treatment-related toxicities have contributed to a less-than-ideal effect even as we have access to an increasing number of HER2-directed therapies making treatment selection potentially complex. Therefore, identifying biomarkers that predict response and/or resistance to specific HER2-directed therapies is an attractive clinical aim. In doing so, it is hoped that patients may be spared ineffective treatments and unnecessary toxicity, while maximizing benefit from available therapies. However, distinguishing and quantifying relevant markers have been significant obstacles. Both genomic and proteomic strategies have been employed, yet discrepancies have arisen and no clearly predictive marker has been identified beyond HER2 itself. Recent data may change this situation and possibly provide guidance for future research.

Keywords::

• capecitabine
• EGFR
• gene copy number
• HER2
• lapatinib

1. Background

Overexpression of the human epidermal growth factor receptor 2 (HER2), a member of the epidermal growth factor receptor (EGFR) family, occurs in 20 – 30% of breast cancers and in the absence of targeted therapy is associated with clinically aggressive disease including shortened disease-free and overall survival [1]. This adverse prognostic effect has been largely abrogated by the development of specific HER2-directed therapies, including initially the monoclonal antibody trastuzumab [2]. Subsequently, a number of other drugs have emerged that are at varying stages of development, including pertuzumab, ado-trastuzumab emtansine, other antibody-conjugates, inhibitors of the phosphoinositide 3-kinase (PI3K) pathway, small molecule tyrosine-kinase inhibitors (TKIs) and inhibitors of heat shock protein 90 [3]. Despite the success of HER2-directed therapies, many patients have disease that will ultimately relapse or progress due to de novo or acquired resistance leading to premature death. Accordingly, the ability to identify patients who are more or less likely to respond to specific HER2-directed therapies would carry significant clinical import. However, this task remains challenging given the innate biological heterogeneity of breast cancer, multiple mechanisms of action of HER2-directed therapies and, presumably, numerous mechanisms of drug resistance. Indeed, several studies have aimed to identify biomarkers of resistance and/or response to anti-HER2 therapies and one major area of interest is EGFR, which is known to heterodimerize with HER2 and may therefore influence the efficacy of HER2-directed therapies [1,2]. Another promising marker thus far is the presence of mutations in the PTEN/PI3K/AKT pathway [4-9]. Overall, results from studies to date have been mixed and overexpression of HER2 itself remains the only proven predictor of response. Therefore, simply stated the optimal biomarker of benefit from anti-HER2 therapy (other than HER2 itself) remains elusive.

In this issue of Expert Opinion: Pharmacotherapy, Fabi and colleagues report their analysis of EGFR gene copy number (GCN) as a predictor of sensitivity to the small molecule, orally bioavailable, TKI lapatinib in patients with HER2 overexpressing breast cancers. Notably, this TKI also possesses activity against EGFR (also called HER1) [10]. There is a strong biological rationale for the use of this agent and research approach given that the presence of EGFR gene mutations and increased EGFR GCN are associated with clinical benefit from anti-EGFR therapy in other solid tumors. In this study, 50 patients with metastatic HER2-positive breast cancer and disease progression following one trastuzumab-based therapy were treated with lapatinib (1250 mg/day continuously) and capecitabine (2000 mg/m²/day for 14 days every 3 weeks) with an overall response rate of 38% (95% confidence interval 24 – 51%). Using archived paraffin-embedded tissue from primary and metastatic tumors (where available), EGFR gene status was then evaluated by fluorescent in situ hybridization (FISH). Although the authors acknowledge this was an exploratory study, a sample size of 50 patients was planned based on the hypothesis that increased EGFR GCN would be associated with an odds ratio (OR) of > 2 for tumor response (with 80% power and significance level of 5%). In total, EGFR GCN was successfully performed in 40/50 samples. The authors then generated a receiver operating characteristic (ROC) curve and determined that a cut-off of > 3.36 EGFR copies/nucleus could distinguish responders from non-responders. Importantly, EGFR GCN > 3.36 was significantly associated with response in univariate (OR 6.0, p = 0.01) and multivariate analyses (OR 7.5, p = 0.013). Response rates were highest (73%) in tumors with the highest EGFR GCN (> 3.36) and strongest HER2 expression (3+ by immunohistochemistry). As might be expected, these patients had longer median overall survival compared to those with low EGFR GCN. The authors conclude that, pending validation, EGFR GCN might be a useful tool to select patients for combination therapy with lapatinib and capecitabine.

2. Conclusion

As novel HER2-directed therapies become implemented in clinical practice, it will be increasingly important to consider the specific mechanisms of HER2 signaling disruption when assessing the utility of potential biomarkers. Many markers including c-MYC amplification, insulin-like growth factor-1 receptor (IGF-1R), and PTEN loss are under investigation as prognostic and/or predictive markers [11]. Still more markers are being explored in correlative and preclinical studies including the MET oncogene, hepatocyte growth factor, polymorphisms in cell cycle checkpoint genes and others [12-15]. As described above, PI3K mutational status has emerged as a promising prognostic marker. While the PIK3CA mutation of PI3K was not predictive of treatment response in the CLEOPATRA study (discussed below), wild type PIK3CA was associated with better outcomes [4]. Furthermore, in a series of prospectively collected tumor tissue from women with metastatic HER2-positive breast cancer, Chandarlapaty and colleagues reported a higher rate of PTEN loss and PI3K mutations in the group treated with trastuzumab versus the control group without trastuzumab exposure [7]. We applaud the authors for adding a novel observation to the developing field of resistance markers for anti-HER2 therapy. Their work clearly demonstrates the potential utility of this line of inquiry even as it illuminates some of the many challenges ahead.

3. Expert opinion

The development of validated biomarkers of therapeutic benefit to individual anti-HER2 therapies is of major clinical importance. As such, the study by Fabi et al. is of interest and has a number of strengths. Blinded analyses were conducted by three separate investigators, suggesting reproducibility of results. The ability to perform EGFR GCN on archived paraffin samples means this technique might be generalizable to clinical practice. However, contrasting reports have quantified EGFR by immunofluorescence and/or mRNA levels [5,6,8,9]. In the Alliance (formerly the North Central Cancer Treatment Group, NCCTG) N9831 Trial, EGFR expression was determined by immunofluorescence. A high level of EGFR expression, using a threshold determined in an independent retrospective cohort, was associated with worse outcomes and diminished benefit from trastuzumab given concurrently with chemotherapy in the adjuvant setting [5]. In contrast, EGFR expression, determined by immunohistochemistry or mRNA level by reverse transcription PCR, was not associated with lapatinib benefit in a retrospective subset analysis of tumor specimens from nearly 1000 patients with metastatic breast cancer who received treatment with this TKI plus chemotherapy versus chemotherapy alone [6]. Similarly, Baselga and colleagues reported the results of an extensive biomarker analysis in the Clinical Evaluation Of Pertuzumab And TRAstuzumab (CLEOPATRA) study in which no marker, including EGFR mRNA level, was predictive of benefit from HER2-directed therapy [4]. However, it is possible that the lack of a HER2 treatment-naïve control group may have abrogated a potential signal in this study. Collectively, these studies were all considerably larger than the study by Fabi et al., in which tissue was available in only 40/50 planned patients. Furthermore, the relationship between EGFR GCN and EGFR expression, whether determined immunohistochemically or by mRNA level, is not well understood. Indeed, EGFR has been historically difficult to measure. It is conceivable, then, that GCN could represent a more useful method of EGFR assessment.

A weakness in this study is the lack of separate training and validation cohorts. Furthermore, the cut-off point, at which a greater EGFR GCN (3.36) was predictive of lapatinib sensitivity, was determined within the same study population and, therefore, may not be confirmed when tested in larger or alternate populations. Finally, since all patients received both chemotherapy and anti-HER2 therapy, the relative benefit of each therapy and the resultant performance of the biomarker cannot be clearly elucidated. This leads us to question whether GCN can be used clinically to select patients for lapatinib therapy today. Although response rates were higher in patients with GCN > 3.36 (64%) than in those with GCN < 3.36 (23%), the test failed to adequately predict responders from non-responders. Assuming these results are reproducible, a clinician could tell an individual patient that their disease is more or less likely to respond based on this test, but that some patients with low EGFR GCN will still have disease response and some with high EGFR GCN will have disease progression. Therefore, even if it is validated, the current data suggest that this test cannot be used clinically without additional refinement. A key issue is that treatment in the setting of incurable metastatic breast cancer is aimed at both symptom palliation and quality of life as well as overall survival rather than tumor response. As a consequence, it may, or may not, help to more accurately predict response rates with a validated test. Finally, as noted, the design and sample size of this study do not allow it to differentiate between the prognostic and predictive implications of high EGFR GCN. It is possible that the presence of high EGFR GCN would have been associated with high response rates and relatively long survival irrespective of treatment, including second-line trastuzumab-based therapy. In this setting it is unclear from this study whether the apparent lapatinib sensitivity is largely related to HER2, rather than EGFR expression, as would be corroborated by the biomarker analysis in CLEOPATRA discussed above.

Declaration of interest

PG Morris has received honoraria from OncologyStat. NM Iyengar has nothing to declare.