Interactions between anti-EGFR therapies and cytotoxic chemotherapy in oesophageal squamous cell carcinoma: why clinical trials might have failed and how they could succeed

Oesophageal cancer is the sixth most common cause of death from cancer globally, and squamous cell carcinomas of the oesophagus (ESCC), is the dominant histological subtype of oesophageal cancer worldwide [1]. Patients frequently present with advanced disease and, as a result of late stage diagnosis and limited treatment options, 5-year survival rates remain low at around 15%. Current treatments depend on the tumour stage, co-morbidities and patient performance status; surgery is curative in fewer than half of patients and the majority of patients receive palliative treatment, including chemotherapy [2]. Currently, cytotoxic chemotherapy provides a systemic therapy option for palliative treatment of ESCC, but there are no licenced targeted therapies or predictive biomarkers and, therefore, an unmet need for more effective approaches [3].

First-line palliative chemotherapy usually involves a fluoropyrimidine/platinum combination but, eventually, all patients develop progressive disease with some receiving second-line treatment with a taxane or irinotecan monotherapy [4]. Recently, a study in patients progressing after fluoropyrimidine/platinum chemotherapy demonstrated that the PD-1 inhibitor nivolumab improved overall survival compared to taxane monotherapy (ATTRACTION-3 trial [5]). Although the progression free survival and the proportion of patients responding were similar in both groups, the responses to nivolumab were more durable but took longer to occur than responses to taxanes. These findings highlight the importance of identifying the minority subgroup of patients who would benefit long-term from nivolumab, but, in the short term, taxanes are superior. However, the low objective response rate (20%) and poor long-term survival with taxanes in this setting, indicates that treatment resistance is a major clinical challenge that needs to be addressed.

One approach to develop novel therapies is to identify and target oncogenic drivers and efforts to characterise genome alterations within tumour tissue is now enabling the selection of biomarkers for precision medicine targeted therapies. Potential drivers of oesophageal tumourigenesis include the epidermal growth factor receptor (EGFR). Copy number gain is detected in around 20% of tumours [6, 7], while EGFR is overexpressed in around 50% of ESCC tumours and correlates significantly with tumour invasion [6]. Targeting EGFR with EGFR tyrosine kinase inhibitors (TKi), such as gefitinib, erlotinib or afatinib, inhibits the proliferation of oesophageal cancer cell lines in vitro [8, 9], but, clinical trials of EGFR inhibitors in oesophageal cancer, including ESCC, have shown mixed results. Monotherapy trials in unselected patients with EGFR inhibitors indicate that there is an EGFR-driven minority ESCC subgroup who gain survival, symptomatic control and health-related quality of life benefits from EGFR inhibitors [10, 11]. In ESCC, EGFR CNG assessed by FISH, and/or EGFR protein over-expression have shown promise as predictive biomarkers to identify this benefiting subgroup, but needs prospective validation [3, 7, 12, 13]. Cell line models and patient-derived xenografts also demonstrate an EGFR-driven subgroup of ESCC sensitive to EGFR inhibitors and characterised by EGFR CNG and/or EGFR protein over-expression. However, even in these biomarker-selected groups, intrinsic and acquired resistance to EGFR inhibitors remains significant [8, 14, 15]. The considerable heterogeneity of EGFR CNG and protein over-expression observed in ESCC may be a key determinant of resistance [16], with rapid selection and outgrowth occurring of EGFR CNG and protein overexpression-negative tumour cell sub-clones that are unresponsive to EGFR inhibitors. This emphasises the importance of combining EGFR inhibitors with a therapy that is effective against EGFR ‘negative’ sub-clones, and ideally one that would also synergise with EGFR inhibitors towards the EGFR ‘positive’ driven sub-clones. To address this, a number of clinical trials have investigated the combination of EGFR inhibitors and cytotoxic chemotherapy. Clinical trials combining EGFR inhibitors and platinum/fluoropyrimidine chemotherapy in the advanced stage setting or with platinum fluoropyrimidine-based concurrent chemoradiotherapy in the curative treatment setting, have not shown an incremental benefit [17]. In the largest randomised trial in ESCC, in molecularly unselected patients with advanced stage disease, the addition of the anti-EGFR monoclonal antibody pantitumumab to cisplatin and 5FU chemotherapy did not improve overall survival [18]. Similarly, in unselected advanced stage gastroesophageal adenocarcinoma (GOA) patients, a negative impact on overall survival was observed with addition of panitumumab to epirubicin, oxaliplatin and capecitabine [19, 20]. Conflicting results have also been reported in trials of platinum-based chemotherapy in combination with EGFR TKi in non-small cell lung cancer patients (NSCLC) [21‐23]. In contrast, the addition of the EGFR TKi erlotinib, to definitive chemoradiotherapy for ESCC, which included a taxane (paclitaxel and cisplatin) was beneficial [24].

Overall, there is evidence of an EGFR-driven and EGFR inhibitor-responsive subgroup of ESCC and, thus, the potential to combine current standard of care cytotoxic chemotherapies with EGFR inhibitors to improve outcomes. The development and clinical use of EGFR inhibitors for oesophageal cancer, however, has been hampered by conflicting trial results, which could be a consequence of biomarker-unselected patient cohorts, and also because questions regarding potential antagonistic effects of co-administration of EGFR inhibitors with cytotoxic chemotherapy have been raised [25‐28]. Furthermore, the clinical relevance of this is increased by findings which suggest that, consistent with observation in other tumour types, EGFR signalling is a key determinant of resistance to immune checkpoint inhibitors in ESCC [29], and accordingly EGFR-driven ESCC are likely to be a subgroup that derive less, or no, benefit from immune checkpoint inhibitors.

To address this, and to identify the drug combinations most likely to benefit ESCC patients, in this study, we have investigated the outcomes from platinum/fluoropyrimidine chemotherapy in EGFR CNG positive and negative gastroesophageal cancer patients and then the combinatorial activity of EGFR inhibitors with cytotoxic drugs in ESCC cell lines with EGFR CNG and varying intrinsic sensitivity to gefitinib. Cytotoxic drugs included platinum-based chemotherapies (cisplatin and oxaliplatin), a taxane (docetaxel) and a topoisomerase inhibitor (irinotecan) which were tested in concurrent and sequential administration settings.

Oesophageal squamous cell carcinoma (ESCC) patients whose tumours have EGFR CNG and/or EGFR protein overexpression may represent a subgroup that benefits from EGFR inhibitor monotherapy [7, 12, 13]. However, even in this biomarker-selected subgroup of ESCC, significant proportions of patients do not respond to EGFR inhibitors and durable responses are uncommon, indicating that primary and acquired clinical resistance is a major clinical challenge. Heterogeneity is a predominant feature of ESCC, including for EGFR CNG and protein over-expression. Tumours with a higher number of genomic clonal subpopulations that are not EGFR-driven are less likely to respond significantly to monotherapy with an EGFR inhibitor. Therefore, EGFR combination treatments are likely to be important to optimise treatment effectiveness in EGFR CNG positive ESCC.

Several clinical trials have investigated the combination of EGFR inhibitors with cytotoxic chemotherapy or concurrent chemoradiotherapy [17‐20, 24]. In the palliative setting, in both ESCC and GOA, the addition of EGFR inhibitors to platinum plus fluoropyrimidine chemotherapy has not improved overall survival [18‐20]. Similarly, in the radical treatment setting, the addition of EGFR inhibitors to concurrent chemoradiotherapy, with a platinum and fluoropyrimidine chemotherapy backbone has not improved overall survival [17]. However, addition of EGFR inhibition to chemoradiotherapy in ESCC with a chemotherapy backbone incorporating paclitaxel did improve overall survival, even in biomarker-unselected patients [24]. Consistent with these clinical trial results we observed that EGFR CNG positive gastroesophageal cancers in both the palliative and neoadjuvant setting appear to be more sensitive to platinum fluoropyrimidine-based cytotoxic chemotherapy. This suggests that the use of an EGFR inhibitor could reduce, or negate, the benefit of platinum-based cytotoxic chemotherapy in patients with EGFR-driven tumours and would thus be antagonistic. We confirmed drug antagonism in EGFR CNG cell lines. Because no GOA cell lines with EGFR CNG are available, our cell line experiments were restricted to ESCC. This is a limitation of our work, and deriving EGFR CNG GOA cell lines to confirm our findings would be advantageous. The finding of synergy between docetaxel and gefitinib observed in 2-D cell culture was supported by our findings in longer-term 2-D colony formation assays and in 3-D spheroid culture models. However, not all cell lines used in this study were able to grow in 3-D culture and likewise, to overcome this limitation of the current study, testing of additional EGFR CNG positive cell lines that are able to survive anchorage-independent seeding would be advantageous. Investigation of drug synergy in an in vivo xenograft model, evaluation of the impact of EGFR CNG on survival in a cohort of ESCC patients treated with docetaxel monotherapy, as well as extension of investigation to other EGFR inhibitors would also be beneficial. In addition, elucidation of the mechanism of synergy between docetaxel and gefitinib, or other EGFR inhibitors, in EGFR CNG positive ESCC would also provide additional understanding and may suggest predictive biomarkers to identify those patients that might benefit most from this combination.

The majority of earlier reports have suggested that EGFR protein expression and/or EGFR gene copy number gain are associated with shorter survival of gastroesophageal cancers including adenocarcinomas and ESCCs [38‐40]. However, longer survival, or no impact, has been observed in other studies [38‐40]. In contrast to our investigation, the influence of treatment modality in general and cytotoxic chemotherapy in particular on how EGFR impacts the clinical outcomes has not been investigated in the majority of these studies. Our data shows an interaction between neo-adjuvant PBC and the impact of EGFR and, consistent with the findings of previous studies, indicates that the treatment patients receive should be taken into account when the impact of EGFR is being considered. Furthermore, Smyth et al., reported that higher tumour EGFR RNA expression is associated with shorter overall survival in gastroesophageal cancer patients treated with peri-operative PBC [41]. These findings contrast with ours, and suggest that EGFR copy number measured by FISH and EGFR RNA expression measured by nanostring may be measuring different aspects of gastroesophageal cancer biology. For example EGFR FISH-positive tumours almost invariably overexpress EGFR RNA (and EGFR protein), but up to 50% of EGFR FISH-negative tumours also have high EGFR RNA and protein [42, 43]. The lack of benefit from gefitinib in EGFR CNG-negative patients (which might have still high expression EGFR RNA and or protein) suggests that EGFR protein or RNA expression may be a less reliable marker of EGFR-driven gastroesophageal cancers and a less reliable marker of the impact of EGFR signalling and EGFR inhibitors on PBC sensitivity, resistance or clinical outcomes. Nevertheless, these findings also indicate the importance of considering the type of EGFR assay used in gastroesophageal predictive and prognostic studies. To our knowledge the impact of EGFR CNG on overall survival in advanced stage patients treated with palliative platinum fluoropyrimidine chemotherapy (PBC) has not been previously reported.

Our data indicating that the combination of gefitinib with cisplatin is antagonistic (in TE8 and OE21 cells) are at odds with reports that treatment of TE8 xenograft tumours with cisplatin in combination with EGFR inhibition by cetuximab significantly reduces their size [44]. Such discrepancies may arise due to the nature of the mechanism of inhibition of EGFR (small molecule EGFR TKi versus blocking antibody with potential antibody-dependent cellular cytotoxicity effects). However, clinical results of EGFR monoclonal antibodies both as monotherapy and in combination with PBC in ESCC have been similar to those demonstrated with EGFR TKis [10, 18, 20‐23, 27, 28, 35]. In addition, in line with our conclusions, the POWER phase III RCT in advanced ESCC did not demonstrate any benefit of the addition of the humanised monoclonal anti-EGFR antibody panitumumab to cisplatin plus fluoropyrimidine chemotherapy [18]. POWER enrolled molecularly unselected ESCC patients, but, a retrospective analysis demonstrated that EGFR IHC did not correlate significantly with overall survival, and EGFR copy number was not investigated. The antagonism between cisplatin and oxaliplatin and EGFR inhibition provides a key explanation for this, and other, negative clinical trials and suggests that, even if these trials had been undertaken in biomarker-selected patients, benefit from the addition of EGFR inhibitors may not have been observed.

Previous studies in KYSE30 cells have suggested that the sequence of administration of gefitinib in combination with cytotoxic agents determines efficacy. Synergy was noted with cisplatin, carboplatin, oxaliplatin, docetaxel and paclitaxel followed by gefitinib [45]. However, our studies in a wider cell panel suggest that the effect of concurrent gefitinib with platinum-based cytotoxic drugs is cell line-dependent and thus the effects of this combination may be unpredictable in the clinical setting. In contrast, we demonstrated that gefitinib and docetaxel was consistently synergistic. This observation is consistent with the demonstration that addition of erlotinib to chemoradiotherapy including paclitaxel was beneficial in ESCC [24]. In this study there was no biomarker selection, and we hypothesise that in this trial the greatest benefit from addition of erlotinib will be seen in those patients who are EGFR CNG and/or have EGFR protein overexpression. Our dose reduction analysis indicates that equivalent reductions in tumour cell growth could be achieved with lower doses of gefitinib and docetaxel if used in combination, which could have significant advantages in terms of reduced financial costs of the treatment and, more critically, reduced toxicity to the patient. The cell line studies also suggest that the sequence of administration of taxane and EGFR inhibitor is critical. The administration of gefitinib prior to docetaxel invariably resulted in antagonism which is consistent with studies in both an NSCLC cell line [46] and in KYSE30 ESCC cells [45]. This data suggests that careful scheduling, or drug holidays may be required to avoid possible antagonistic drug interactions. In support of this, when used in combination with paclitaxel, pulsatile administration of gefitinib has proved more effective than continuous dosing in murine models of breast cancer, [47].

The molecular mechanism of drug combination antagonism or synergy in this setting is unclear and is under investigation. In vitro studies analysing potential combination therapies on both head and neck SCC and NSCLC cell lines suggest that there is also antagonism between gefitinib and cisplatin in other tumour types [25‐27]; the effects have been variously attributed to cisplatin cytotoxicity being dependent on EGFR phosphorylation and degradation [28], induction of epithelial to mesenchymal transition (EMT) which is associated with increased resistance to gefitinib [26], reduced cisplatin entry into the cell and increased DNA repair or cell cycle arrest in G₁. However, antagonism can be overcome by concurrent use of autophagy inhibitors (without any apparent effect on the cell cycle) [27] suggesting that factors other than the phase of the cell cycle may be involved. Previous reports in colorectal cancer cells have suggested that treatment of cells with cytotoxic agents increases phosphorylation of EGFR rendering cells more sensitive to the effects of EGFR TKis, whereas antagonistic interactions result from a cytotoxic drug induced a decrease in EGFR phosphorylation [37, 48]. In NSCLC, the mechanism of synergistic interaction was also suggested to be due to increased docetaxel-induced phosphorylation of EGFR and its subsequent inhibition following gefitinib addition [6]; however, similar analysis of our ESCC cell lines over several repeat experiments was inconclusive. The antagonistic effect of sequential administration of gefitinib prior to docetaxel could be due to cell cycle effects, where gefitinib induces a G1 cell cycle arrest thus rendering taxanes (which are primarily mitotic spindle inhibitors acting in G2/M) ineffective [45, 46]. Since concurrent administration of gefitinib or administration after docetaxel is synergistic, in the clinic, this would suggest that an interrupted schedule of an oral TKi like gefitinib, might be needed in combination with a taxane. Alternatively, and in contrast to common clinical practice at present, an anti-EGFR monoclonal antibody should potentially be administered after docetaxel and not before, when combined with a taxane.

Recently, the ATTRACTION-3 study demonstrated that the PD-1 inhibitor nivolumab provided improved overall survival compared to taxane monotherapy suggesting that nivolumab is a new standard of care for ESCC after progression with fluoropyrimidine/platinum chemotherapy [5]. However, given that only a minority subgroup of patients respond to nivolumab and responses took a median of 2.6 months to occur, its use presents clinical challenges in this setting, where patients often have high tumour burdens and are very symptomatic. In addition, in squamous cell carcinomas including ESCC, EGFR activation is associated with depleted tumour infiltrating lymphocytes and resistance to immune checkpoint inhibition (ICI) [29]. EGFR activation leads to increased anaerobic glycolysis in tumour cells, glucose depletion and accumulation of lactate in squamous cell carcinomas, meaning that tumour-infiltrating T cells, may have to compete for metabolic fuels. ICI appear to be less effective in EGFR mutant positive NSCLC [29, 49], and EGFR activation has been associated with hyper-progression following ICI therapy [50]. Early phase trials in NSCLC have revealed problematic toxicity combining EGFR inhibitors and immune checkpoint inhibitors [51]. Together these data suggest that EGFR-driven ESCC, identified by EGFR CNG and /or EGFR protein overexpression are likely to be less sensitive to nivolumab which is unlikely as a monotherapy to provide an effective treatment for this group of patients. As such, taxanes will likely remain one standard of care for ESCC after progression with fluoropyrimidine/platinum chemotherapy, either before or after nivolumab. Overall, our results contribute additional evidence to support investigation of EGFR inhibitors in EGFR CNG positive ESCC and suggest that a combination strategy with taxanes has the potential for synergism thereby optimising clinical impact and effectiveness. Since taxanes are a standard of care for ESCC after progression with fluoropyrimidine/platinum chemotherapy evaluating the benefit of the addition of an EGFR inhibitor to docetaxel or paclitaxel in tumours that are EGFR CNG and/or EGFR protein overexpressed by IHC would be the most appropriate initial area of clinical investigation.

In summary, drug combination studies indicate that targeting EGFR in ESCC cells carrying EGFR copy number gain may negate or reduce anticancer effects of platinum-based chemotherapy; however, EGFR inhibitors are efficacious and synergistic in combination with docetaxel when scheduled correctly. We recommend further preclinical studies and consideration of clinical investigation of scheduled anti-EGFR therapies combined with taxanes for ESCC patients with tumours expressing high EGFR by IHC and/or EGFR CNG.