The inflammatory milieu within the pancreatic cancer microenvironment correlates with clinicopathologic parameters, chemoresistance and survival

Pancreatic adenocarcinoma (PC) is the fourth leading cause of cancer deaths in the United States, due in part to nearly universal resistance to cytotoxic chemotherapy. Gemcitabine-based therapies achieve clinical benefit in approximately 24 % of patients with PC [1], but the overall survival advantages are sobering, ranging from a few weeks to months [1–3]. Complete surgical resection offers patients with PC the greatest survival benefit. However, this is achievable in fewer than 20 % of patients presenting with PC [4]. As a result, PC is projected to be the second leading cause of cancer deaths by 2030 [5]. There is a tremendous need to discover novel biomarker (s) or panels of biomarkers that can aid in detecting PC earlier, improving prognostic evaluation and predicting response to chemotherapy.

Inflammation within the PC microenvironment has been mechanistically linked to tumor progression and chemoresistance through NF-κB, IL-6, toll-like receptor and TGF-β signaling pathways [6–10]. However, the diagnostic and prognostic value of the inflammatory milieu within the PC microenvironment remains essentially undefined. While survival gains from immune cell infiltration into the tumor microenvironment have been conclusively demonstrated in colorectal and ovarian cancer [11–13], similar investigations have not yielded consistent results in PC [14, 15]. Patients with chronic pancreatitis are 5–15 times more likely to develop PC [16] and insights into the association between inflammation and PC stems from investigations of chronic pancreatitis. Potential environmental sequelae of pancreatitis such as hypoxia, the presence of reactive oxygen species, and acidosis may influence the development of PC [17]. Additionally, numerous soluble mediators, including TNF-α [18], TGF-α [19], TGF-β [20], IL-1β [21], IL-1α [22], IL-6 [23, 24], IL-8 [25], VEGF [26], and others have been implicated in PC carcinogenesis, tumor progression, and treatment resistance. However, the relationship between the inflammatory milieu and the spectrum of disease from normal pancreas to pancreatitis to pancreatic cancer has not yet been characterized. Therefore, the translational relevance of the microenvironmental inflammatory milieu to PC development and progression remains speculative.

We examined the inflammatory milieu present in the PC microenvironment from 36 freshly resected tumor specimens using a forty-one-item panel of cytokines, chemokines and growth factors to test the hypothesis that expression levels of these mediators harbor diagnostic and prognostic value. We first compared the inflammatory milieu of PC to that of pancreatitis (n = 9) and normal pancreas (n = 6). Inflammatory mediators were further evaluated in relation to prognostic clinicopathologic parameters, administration of neoadjuvant therapy, treatment resistance and patient survival. These data bring the field one step closer to the identification of biomarker panels that can aid in detecting disease earlier and classifying patients with respect to response to chemotherapy and most importantly, prognosis.

Due to the dismal clinical outcomes associated with pancreatic adenocarcinoma and the continued debate surrounding therapeutic interventions, there is a tremendous need for the development of tools that can supplement current diagnostic and prognostic efforts. The extent of genetic and phenotypic heterogeneity specific to PC represents a major obstacle to the clinical application of developing biomarkers in PC. Efforts to further understand clinical observations regarding pathologic signaling within the tumor microenvironment have provided a novel focus that have led to major breakthroughs. Examples include therapeutic successes following Nab-paclitaxel infusion that is dependent upon SPARC expression in desmoplastic tumor-associated stroma [32, 33], and consistent observations that partial responses achieved from CD40 agonists led to the infiltration of tumoricidal macrophages into the local microenvironment [34, 35]. In order to further understand clinically important paracrine signaling pathways within this local microenvironment, the work presented here has detected a unique inflammatory signature within pancreatic adenocarcinoma that is distinct from that of chronic, benign inflammation. Further, several members of this panel of markers were associated with specific clinicopathologic parameters, response to cytotoxic chemotherapy, and overall survival.

The almost universal development of treatment resistance and disease relapse following systemic cytotoxic or targeted therapies has made survival in PC achievable in only a small minority of patients. Mechanistically, chemoresistant phenotypes have been reproduced in vitro. However, the relevance of these findings to clinical practice remains unclear. For example, gemcitabine resistance has been linked to the expression of gemcitabine-metabolizing proteins and DNA repair enzymes as well as the downregulation of nucleoside transporters. However, the clinical value of identifying these markers in resected PC specimens has yielded conflicting results [36]. Here we demonstrate that not only is the exposure to gemcitabine-based therapy associated with a different inflammatory milieu within the tumor, but also that differences in the milieu associate with the degree of clinical response, whereby increased levels of intratumoral IL1-β and TNF-α are associated with poor histopathologic response to neoadjuvant therapy. These findings further support a wealth of investigations linking downstream NF-κB signaling to tumor progression and chemoresistance [37].

The observation that EGF levels correlated with the degree of lymph node metastasis is consistent with widespread evidence implicating EGF signaling in cancer progression and metastasis, culminating in a phase three trial employing EGFR inhibition in pancreatic cancer [3]. Conversely, high intratumoral concentrations of IL-4 displayed the opposite trend, correlating with reduced lymphatic metastasis, whereby patients with tumors high in IL-4 concentrations displayed roughly triple the survival of those with tumors expressing low levels of IL-4. In light of this finding it is important to note that direct stimulation of cancer cells with IL-4 generally results in augmented growth and proliferation [38–40]. However, this finding must be interpreted within the context of IL-4 signaling within the microenvironment. Indeed, constitutive IL-4 expressing cancers have demonstrated reduced growth in vivo due to the induction of a robust antitumor immune response [41]. Similarly, intratumoral levels of IL-8 and GM-CSF were predictive of tumor grade while IL-8 levels were also positively associated with tumor size. Again this is consistent with previous findings that suggest that IL-8 and GM-CSF produced in the tumor microenvironment promote immune evasion in PC [42–44]. Interestingly, the administration of cytotoxic chemotherapy was strongly associated with significantly lower intratumoral IL-8 concentrations. The investigation of the intratumoral inflammatory milieu has therefore revealed consistent correlations between IL-8 concentrations, histopathologic findings and the administration of chemotherapy.

As alluded to above, several of these mediators could also be used to predict survival. For instance, high intratumoral G-CSF levels correlated with reduced overall survival, which is supported in literature relating myeloid-derived suppressor cell infiltration to tumor progression and angiogenesis [45]. Of particular interest is the prolonged survival observed in patients with high intratumoral FGF-2, known to stimulate fibroblast migration, wound healing and generally thought to be a growth factor which supports growing tumors. However, it is generally accepted that FGF-2 is abundant in most tissues, concentrated in basement membranes and at cell surfaces in inactive forms. Tissue injury leads to FGF-2 activation and subsequent promotion of wound healing processes known to promote tumor growth, invasion and angiogenesis [46]. In this context, reduced FGF-2 concentrations in tissue homogenates may paradoxically reflect increased FGF-2 activation, which would lead to the expected findings of reduced survival.

The inability to follow the intratumoral inflammatory milieu over time represents a significant limitation to this type of analysis, as this information will be critical to elucidating the relationship between local inflammation and treatment strategies in PC. In addition, stratification of long-term survival into treatment-naïve and treatment-exposed tumors will be essential in validating these relationships. However, this analysis currently lacks the power to dichotomize in this fashion. While grouping mediators into functionally relevant categories may address our current lack of power, the pleiotropic nature of these soluble mediators may lead to improper interpretations in the absence of functional analyses. Further, it has not escaped our notice that VEGF demonstrated no correlation with survival in this analysis. The extensive body of work associating VEGF signaling with angiogenesis and tumor progression has led many groups to investigate potential correlations between VEGF expression and survival in PC. Subsequent analyses have yielded conflicting results [47–49]. Importantly, this is not the first clinical cohort to demonstrate a nonsignificant correlation between intratumoral VEGF levels and overall survival in PC.

In summary, pancreatic adenocarcinoma is a devastating malignancy with an extremely poor prognosis. High ratios of tumor stroma to cancer cells plague the sensitivity of cytologic diagnosis of PC; In fact, even direct pathologic analysis of PC biopsies can yield inconsistent results with high interobserver variability [50]. Highly specific, reliable biochemical signatures obtained from these small samples that improve diagnostic sensitivity could dramatically improve the clinical care of PC patients. Further, treatment algorithms for PC in the absence of metastasis are currently anatomic based and lack attention to variations in biology. Thus, there is further need to develop accurate biomarkers capable of predicting response to systemic therapies or the futility of surgical or radiation therapy. Unfortunately, the current literature is characterized by marked variability between individual studies as to the relative prognostic impact of several biomarkers in PC. Here, in contrast to studies that evaluated these properties using a single or a couple biomarkers, we have identified novel relationships between tissue examination and clinical outcome by quantitatively evaluating the milieu of the tumor microenvironment utilizing fresh pancreatic surgical specimens. It is important to note that even though some of these associations are counterintuitive based on the currently understood biology, the greater context and complexity of the tumor microenvironment in PC is not currently appreciated. Thus, these data combined with a better understanding of the context-dependence of inflammatory signaling, may eventually offer the opportunity to identify patterns that improve interpretations in cancer and emphasize the importance of investigating the tumor microenvironment as a whole. Nonetheless, results presented here exhibit a high degree of reproducibility and provide rationale to prospectively evaluate these markers as diagnostic and prognostic tools.