RON is not a prognostic marker for resectable pancreatic cancer

Pancreatic cancer (PC) still remains one of the most aggressive and lethal of human cancers, with a lack of prognostic biomarkers and effective treatments contributing to its poor prognosis and high mortality. Around 15% of PC patients are eligible to undergo potentially curative pancreatic surgery at the time of presentation. However, those with operable PC still only have an 18–23% 5-year survival [1, 2], with high incidences of local recurrence and hepatic metastases occurring within 1–2 years after surgery [3]. This emphasizes the need for adjuvant intervention. Gemcitabine is the current standard for post-operative chemotherapy and delays the development of recurrent disease in some PC patients [4]. One explanation for why only a subset of PC patients respond favourably to adjuvant treatment is the molecular heterogeneity of PC [5]. The concept of stratifying patients based on their molecular signatures has been successful in breast cancer and is the basis of modern clinical oncology [6]. Implementation of this treatment strategy for PC may also prove beneficial. The literature is replete with biomarkers of prognosis and therapeutic responsiveness identified through small and/or single pancreatic cancer patient cohorts. However, a recent systematic review has shown that very few of these have been independently validated [7]. The calcium binding protein S100A2 is one of the few prognostic biomarkers where this is the case [7]. In PC, tumors that are negative for S100A2 have a significant survival benefit with pancreatectomy compared to tumors with moderately-high to high expression [8].

Recepteur d’origine nantais (RON), also referred to as macrophage stimulating 1 receptor (MST1R), is a receptor tyrosine kinase (RTK) and member of the mesenchymal epithelial transition factor (MET)-proto-oncogene family. Similar to other cell surface receptors, RON is activated upon binding of its ligand, the hepatocyte growth factor-like (HGFL) protein, also referred to as macrophage stimulating protein (MSP) and macrophage stimulating 1 (MST1). HGFL is synthesized from hepatocytes and secreted as an endocrine mediator in an inactive precursor form called pro-MSP. Subsequently, the type II transmembrane proteinase MT-SP1, also known as suppression of tumorigenicity 14 (ST14), processes pro-MSP to its active form near the cell surface, where it activates RON, leading to the initiation of multiple signaling cascades that impact upon cellular motility and survival [9].

RON is overexpressed in PC relative to the ductal epithelium of non-malignant pancreas, and its signaling enhances migration, invasion, and survival of PC cells and promotes resistance to gemcitabine in experimental models, thus making it a potential therapeutic target and a possible marker of prognosis [10–12]. Although RON is associated with poor survival in gastroesophageal cancer [13] and a three-gene signature involving RON, MSP, MT-SP1 is a strong indicator for metastasis and poor prognosis in breast cancer [14], the prognostic significance of RON in PC remains unknown. The goal of this study was to use a comprehensive cohort of PC patients to assess RON as a biomarker of prognosis and therapeutic responsiveness.

Pre-clinical studies identified RON as a potential predictive marker for gemcitabine response in pancreatic cancer [11, 12]. This study examined RON as a prognostic and predictive biomarker in three large well-annotated cohorts of patients (total of 492) with resectable PC using immunohistochemistry for the training and validation set and gene expression arrays for the ICGC cohort. Apart from the training cohort, RON expression was not associated with survival. The validation cohort consisted of 316 patients and had a 99% power to detect a hazard ratio of 1.90 (assuming it has a HR and an expression pattern similar to the training cohort with a Type I error of 0.05). However, RON overexpression was only observed in 5.4% of patients in the validation set as compared to 22.1% in the training set, but the reason for the much lower proportion of RON over-expressers in the validation set is not known. The unexpected lower proportion of the over-expressers in the validation cohort has reduced the power to detect a difference and therefore, the inability to detect a difference may be due to a Type 2 error rather than there being truly no difference between the two groups. A post-hoc analysis of the validation set using different cut-off points for RON expression could not demonstrate any differential survival (data not shown). It is therefore likely that due to the smaller number of patients, the training cohort returned a false-positive result, which highlights the importance of independent validation in biomarker discovery and development.

Both the antibody used for IHC and the probe used for mRNA analysis recognized the same RON isoforms: the full-length inactive precursor pro-RON, the active RON β-chain and short form sf-RON [15], thus enabling comparisons to be made between the cohorts.

RON mRNA was expressed in a continuous manner in the third, prospectively acquired cohort, so two cutoffs were used: the mean (Figure 5A), and an 80%:20% low:high cutoff to match the frequency utilized in the restrospective cohorts (Figure 5B). Using both approaches, RON was not associated with prognosis. Additional to RON expression alone, a combination gene signature of RON, MSP and MT-SP1, which is prognostic for breast cancer, was also evaluated. However, in PC this three-gene signature, and even two-gene signature combinations, did not correlate with patient survival. We acknowledge that there is reduced power in the 2-gene and 3-gene signatures due to low sample numbers in each strata, but it should be noted that this is the largest prospectively accrued cohort of PC patients with mRNA data to date. Collectively, this suggests that the role of RON is cancer-type specific and that RON is unlikely to be a major metastatic driver in PC. However, the ICGC cohort is still relatively young, currently with a median follow up of 14.1 months, and exhibits relatively small numbers. Therefore, the negative result can also possibly be due to the study being under-powered.

Data from this study do not support RON as a prognostic or predictive biomarker in resectable PC. This study, however, did demonstrate that RON is expressed in a large proportion of PC (training set: 61 of 64, 95.3%; validation set: 253 of 281, 90.0%), which is consistent with previous published data [10]. Due to increased expression during PC progression RON may still prove to be an effective therapeutic target.

To improve the efficiency of gemcitabine in PC, several studies have evaluated the co-administration of gemcitabine with other cytotoxic and biological agents [22, 23]. However, the epidermal growth factor receptor (EGFR) inhibitor, erlotinib, was the only agent to show efficacy in a phase III setting [24]. The EGFR is commonly overexpressed in PC and is associated with poor prognosis and disease progression [25]. The frequent overexpression of RON in PC suggests that it could also be a target for therapy. Supporting this concept, silencing RON expression reduced growth of pancreatic cancer xenografts and increased gemcitabine-induced apoptosis [11]. In addition, function-blocking monoclonal antibodies and small molecular inhibitors directed against RON have demonstrated promising results in pre-clinical models [10, 26, 27].

RON is not associated with prognosis or therapeutic responsiveness in resectable PC, in the current cohorts of patients. However it may still represent a therapeutic target in this disease.