Variety Is the Spice of Life, but Maybe Not in Gastroesophageal Adenocarcinomas

It was not long ago that the anti-HER2 antibody trastuzumab demonstrated both the therapeutic potential and current limitations of targeted therapies in advanced gastroesophageal adenocarcinomas. Unlike the significant benefits seen with matched therapies in oncogene-driven non–small cell lung cancer and HER2-positive breast cancer, trastuzumab plus chemotherapy for patients with first-line advanced HER2-positive gastroesophageal adenocarcinomas resulted in only a 47% response rate versus 35% for chemotherapy alone. Countless front-line phase III trials adding targeted agents to chemotherapy, including anti-EGFR inhibitors (cetuximab, panitumumab) without biomarker selection, and MET ligand–blocking antibodies (rilotumumab, onartuzumab) with suboptimal biomarker selection, all failed to improve outcomes. Later-line clinical trials attempting to evaluate targeted agents based on biomarkers detected at initial diagnosis were even more disappointing. Why have we failed to replicate in gastroesophageal adenocarcinomas the advancements of precision medicine realized in other tumor types? Why have targeted agents, including trastuzumab, either failed or only modestly improved outcomes in most patients with gastroesophageal adenocarcinomas, even when harboring the target?

In this issue of Cancer Discovery, Sanchez-Vega and colleagues present a single-institution prospective phase II trial of the irreversible pan-HER inhibitor afatinib, alone or in combination with trastuzumab in patients with trastuzumab-refractory HER2-positive gastroesophageal adenocarcinomas (1). Pretreatment and postprogression biopsies were analyzed with MSK-IMPACT next-generation sequencing (NGS), and a subset of patients underwent molecular imaging with ⁸⁹Zr-trastuzumab PET imaging.

Second-line studies evaluating continued HER2-directed therapy after failure of first-line trastuzumab-containing therapies in gastroesophageal adenocarcinomas have been disappointing. To date, attempts utilizing small-molecule tyrosine kinase inhibitors (“TyTAN”; lapatinib), antibody–drug conjugates (“GATSBY”; TDM1), and changes in chemotherapy backbone (“T-ACT”; trastuzumab) failed to improve outcomes for HER2-positive disease, as determined by HER2 status prior to first-line therapy. The authors astutely noted “HER2 conversion” during their study and amended the protocol to include only those patients with persistently HER2-positive tumors just prior to starting therapy on study, a concept reported previously (2, 3). However, despite this extra (and necessary) selection, the authors here report only modest activity of single-agent afatinib with a response rate of 10% (2/20) and median progression-free survival of 2 months. In the cohort receiving afatinib in combination with trastuzumab (n = 12), the response rate was no better. Most patients had progressed on two or more prior trastuzumab-containing combinations and therefore were heavily pretreated, potentially explaining the poor activity observed. More importantly, though, it is becoming well-recognized that treating with a monotherapy-targeted agent, or even dual targeted therapy toward one or two receptors, will not be enough for this profoundly molecularly heterogeneous disease. The results from this study yet again point to this problem. To date, no reported studies have addressed all of these concepts simultaneously, continuing anti-HER2 therapy beyond first progression, ensuring persistent HER2 positivity in such patients, and combining with other active “generalized” therapy, namely second-line chemotherapy, to both address HER2-negative clones and synergize toward HER2-positive clones. This remains a hypothesis that merits specific prospective testing and is ongoing (3).

To explore the specific genomic determinants of afatinib response and innate resistance, the authors obtained tissue samples at trastuzumab resistance, but prior to afatinib. Among the 24 of 32 (75%) patients with evaluable biopsies for genomic analyses, all retained HER2 positivity (those without persistent HER2 positivity were excluded, as discussed above). Four patients harbored concurrent EGFR amplification from post-trastuzumab samples (n = 3) or pre-trastuzumab samples (n = 1). The authors note that EGFR coamplification was enriched in patients with the greatest response and hypothesize that coamplification (confirmed by dual-probe FISH) sensitizes to afatinib. The biological underpinnings as to why coamplification would increase afatinib sensitivity versus either target alone is not explored. The authors were unable to obtain paired samples from initial diagnosis; therefore, it is unknown whether the EGFR coamplification was acquired post-trastuzumab or was present initially. Tumors acquiring EGFR coamplification at trastuzumab resistance could suggest dependence on HER pathway signaling and would therefore be more likely to respond to the pan-HER inhibitor afatinib. However, recent reports observed HER2/EGFR coamplifications usually present at initial diagnosis, in some instances in separate cell populations, or in other patients as in this report within the same cells (4, 5). It may be as simple as coamplification of EGFR and HER2 within cells renders higher sensitivity if simultaneously targeted. Further studies including single-cell analyses are likely to be informative. Consistent with prior studies, trastuzumab resistance alterations in MYC, KRAS, PIK3CA, MET, and NF1 were associated with lower probability of response to afatinib (6).

Molecular imaging has the potential to complement traditional disease assessment methods, assessing heterogeneity and early response prediction. The authors provide some preliminary proof of concept using ⁸⁹Zr-labeled trastuzumab PET (Zr-T PET) to highlight previously reported intertumoral HER2 heterogeneity. In a subset of patients (n = 8) undergoing baseline and repeat Zr-T PET and CT, worse outcomes were noted in CT-detectable metastatic sites with no baseline Zr-T PET uptake, presumably related to low/absent HER2 expression in some tumor sites. Not surprisingly, those patients with homogenous (agreement between CT-visible and Zr-T PET avid) metastatic sites were enriched for patients with the largest tumor regression on afatinib. Noninvasive imaging of tumoral heterogeneity could be an attractive tool to incorporate into targeted therapy trials, although barriers including standardization, cost, and independent predictive/prognostic significance remain to be determined. Furthermore, because it cannot be performed in patients having received trastuzumab recently, the applicability of Zr-T PET is more limited to baseline assessment or those receiving anti-HER2 agents other than trastuzumab. Moreover, given the numerous putative predictive biomarkers in gastroesophageal adenocarcinomas, including amplification of the receptor tyrosine kinases (RTK) MET, EGFR, and FGFR2, the concept of multiplex PET to assess each simultaneously sounds fancy, but would potentially accentuate the barriers discussed. Rather, cell-free DNA (cfDNA) collection at serial time points may be a more cost-effective multiplexed assessment to characterize heterogeneity and identify optimal therapeutic targets, as described previously (5). Interestingly, orthogonal support for Zr-labeled imaging was recently shown using pretreatment Zr-atezolizumab PET to predict clinical response to an anti–PD-L1 agent across tumor types in a phase I trial (7).

Utilizing a rapid autopsy program, the authors obtained multiple tissue samples in 3 patients with initial tumor response then subsequent progression and death. Multiple studies in advanced gastroesophageal adenocarcinomas have now repeatedly confirmed both baseline and adaptive intra- and intertumor molecular heterogeneity among various biomarkers, including RTK amplifications, presenting a major barrier to targeted therapies (Fig. 1; refs. 5, 8). Rapid autopsy programs provide a unique tissue source to examine lesion-specific changes and genomic evolution, well exemplified here. Among the 3 patients analyzed, variable putative resistance mechanisms included intralesional variations in EGFR copy-number status (presumably reflecting EGFR nonamplified subclones) and MET amplification. Although not described, rapid autopsy in primary progressors may also have been informative. Support for the functional importance of acquired or innate coamplification was demonstrated by transient response to afatinib plus cabozantinib in a patient and patient-derived xenograft (PDX). PDX models, however, are limited in that they represent only the tumor from which they were derived, not reflecting the overall heterogeneity within the patient. Although further functional work is needed to optimize combination approaches, the authors and patients should be recognized for understanding the power of autopsy programs and comprehensive evaluation of pre- and post-treatment samples to enhance understanding of resistance mechanisms to direct future therapeutic strategies.

Therapeutic resistance often comes at a fitness cost to the cancer cell. Inhibitor cycling, adaptive dosing strategies, and up-front combinations may maximize benefit from targeted agents in context-specific scenarios (Fig. 1; ref. 9). Preexisting and adaptive RTK coamplifications in gastroesophageal adenocarcinomas appear to be rather unique. Comparative genomic analysis of gastroesophageal adenocarcinomas versus colorectal adenocarcinomas has confirmed higher frequency cooccurring genomic alterations in HER2-amplified gastroesophageal adenocarcinomas versus colorectal adenocarcinomas among the chromosome-instable (CIN) subtype (10). The etiologic factors driving the somewhat uniquely higher rates of AA > AC transversions in gastroesophageal adenocarcinoma CIN tumors remains unknown, although oxidative damage from reflux may be a driver of this mutational signature. Importantly, the rates of co-occurring RTK amplifications are lower in tumor types associated with greater magnitude of benefit from RTK-directed mABs (HER2-positive colorectal adenocarcinomas and breast cancer). A theoretical model in which tissue-specific risk factors (e.g., reflux) driving a mutational signature enriched in gastroesophageal adenocarcinomas (AA > AC signature), which predisposes to CIN subtype gastroesophageal adenocarcinomas with focal RTK copy-number amplifications and co-occurring genomic alterations, has clear therapeutic implications and may underlie the intrapatient molecular heterogeneity observations made here and by others (Fig. 1). The clinical utility of NGS assays, as used in this study, continues to increase as we gain a greater appreciation of the impact of the genomic context of a given biomarker, both at diagnosis and over time.

Overall, the phase II trial and correlative work presented by Sanchez-Vega and colleagues add to the cumulative body of work highlighting the need to better longitudinally understand heterogeneity and to design specific strategies to systematically address it if we are ever to achieve targeted therapy success in gastroesophageal adenocarcinomas (3). Development of standardized comprehensive heterogeneity assessments to stratify and inform treatment is critical to achieving the promise of precision medicine in gastroesophageal adenocarcinomas. Although we would never advocate for taking variety out of life, we certainly think understanding the spicing is critical in gastroesophageal adenocarcinomas.

S.J. Klempner has received speakers bureau honoraria from Foundation Medicine and is a consultant/advisory board member for Lilly Oncology, Astellas, and Foundation Medicine. D.V.T. Catenacci is a consultant/advisory board member for Merck, BMS, Lilly, Genentech, Taiho, Astellas, Amgen, Guardant Health, and Foundation One.