Aberrant gene expression in mucosa adjacent to tumor reveals a molecular crosstalk in colon cancer

Colorectal cancer (CRC) is a complex disease in which many genes, proteins, and molecular processes are implicated. Proteins do not work independently in a tumor cell, but are organized into co-regulated units or pathways that perform a common biological function [1]. Relevant molecular mechanisms involved in cancer are gene regulation, signaling, cell metabolism, and the connections between them, among others [2]. In addition to the tumor cell intrinsic complexity, increasing data support the main role of tumor microenvironment in the mechanisms of CRC progression [3‐5]. Tumor microenvironment is composed by a heterogeneous population of stromal cells such as fibroblasts and immune cells, extracellular matrix components and secreted factors. All these components work orchestrated by molecular transducers like integrins engaging cell-cell and cell-matrix signaling that in turn enhance tumor growth [6].

Besides, a colorectal tumor is not an isolated entity growing in a restricted location of the body. An active communication exists not only between different cell communities within the tumor bulk but also between the tumor and the non-tumor distant mucosa. Hence, the patient’s gut environment constitutes the framework where the tumor evolves and this relationship promotes and includes a complex and tight correlation with inflammation, blood vessels formation, nutrition and gut microbiome composition [7]. Consequently, studying the micro-ecology context of a tumor is central to understand colorectal carcinogenesis. The tumor influence on environment could both promote an anti-tumor and a pro-tumor response. Some microenvironments, particularly those associated with tissue injury, are favorable for progression of mutant cells, whereas others restrict it. Cancer cells can also instruct surrounding tissues to undergo changes that promote malignancy [8].

Field cancerization or the field-effect is a theory first described by Slaughter et al. in oral carcinoma [9]. In the initial phase of the multistep carcinogenesis, a stem cell acquires genetic alterations and forms a “patch”, a clonal unit of altered daughter cells. Further alterations convert the “patch” into a field of pre-neoplastic cells. Although only one cell becomes tumoral, the remaining field (adjacent mucosa) continues in a “pre-neoplastic-state” composed of morphologically normal, but biologically altered epithelial cells. Since this field is a pre-tumor site predisposed towards development of cancer, this hypothesis could explain local recurrences after surgery [10].

Understanding the complex ways in which cancer cells interact with their surroundings, both locally in the tumor organ and systemically in the body as a whole has implications for effective cancer prevention and therapy. In contrast to the gene-centric view, a systems biology approach (defined as the analysis of the molecular relationship between genes and proteins as a whole) can be useful to depict a global view of the cancer disease not only as a tumor cell but as an intricate systemic disease [11].

In this study, mRNA expression from paired tumor (T) and adjacent mucosa from CRC patients (A) and mRNA from mucosa healthy donors (H) were measured using microarrays. The inclusion of samples from healthy subjects has allowed us assessing whether adjacent mucosa from colon cancer patients differs from healthy donors’ mucosa possibly due to the tumor presence. Indeed, a number of differentially expressed genes (DEG) were found between these two entities (A vs. H). Considering their level of expression in tumor tissues, these DEGs were classified as “Tumor-like”, “Trend” or “Adjacent-specific” (A vs. T) patterns. To explain the mechanisms that regulate these patterns of differential expression, networks mimicking transcription regulation were used to search for those transcription factors directly influencing DEG. Then, a systems biology approach using PPIN was applied to describe a crosstalk between cytokines and other proteins secreted by the tumor and receptors activated in the adjacent colon tissue; and vice versa, providing new insights into the micro-ecology of colorectal tumorigenesis. Finally, relevant cytokines and receptors up-regulated in tumor tissue were identified comparing T vs. H expression (Figure 1). Further elucidation of these interactions could be helpful in the development of novel therapeutic strategies oriented to disrupt this molecular crosstalk.

There is clear evidence of the relevance of the tumor-microenvironment crosstalk for carcinogenesis [12‐15]. Here we describe altered patterns of expression of the adjacent mucosa from colon cancer patients that could be a direct response against the tumor or induced by the tumor. The analysis of transcriptional profiles and the regulatory networks derived from them allowed us identifying the pathways involved in tumor-microenvironment crosstalk.

We can not discard that at least part of the differences found between adjacent and healthy mucosa were explained by the existence of a pre-neoplastic field in the gut. Studies of adjacent mucosa of the head and neck tumors indicate that such fields can expand more than 7 cm in diameter [10]. Nevertheless, a study in CRC by Jothy S. et al. reported a gradient of carcinoembryonic antigen (CEA) expression expanding only 5 cm. from the peritumor area [16]. In our study, adjacent tissue from patients was dissected from the proximal tumor resection margin, with a minimum distance of 10 cm. However, a recent paper by Hawthorn et Mojica suggests that the field effect cancerization could be evident up to 10 cm. from the tumor [17].

Previous studies usually have compared paired tumor and adjacent mucosa tissues, which can result in misleading interpretations. We have used a large sample of healthy mucosa as reference for gene expression comparisons and have identified a large number of DEG that can be grouped into three altered patterns: “tumor-like”, “trend”, and “adjacent-specific”. Our conclusion is that adjacent normal mucosa is not so normal. In fact, studies that only compare tumor and adjacent mucosa may miss good cancer biomarkers candidates, because many genes are deregulated in adjacent mucosa mimicking the tumor expression.

The predominant functions of DEG are mainly related to response to stimulus, extracellular matrix (ECM) remodeling, organ morphogenesis, and cell adhesion. Remodeling of the ECM network though controlled proteolysis regulates tissue tension, generate pathways for migration, and release ECM protein fragments to direct normal developmental processes such as branching morphogenesis [8]. Collagens are major components of the ECM of which basement membrane type IV and interstitial matrix type I are the most prevalent. Abnormal expression, proteolysis and structure of these collagens influence cellular functions to elicit multiple effects on tumors, including proliferation, initiation, invasion, metastasis, and therapy response [18]. It has been described that integrins that connect various cell types play a vital role in the survival of a growing tumor mass by orchestrating signaling pathways activated through cell-cell and cell-matrix interactions [6]. In our system, integrins ITGA5 and ITGA9 emerged as active signal transducers, occupying central positions in the cellular networks. This result suggests that integrins are not only vital proteins in tumor cells but also in normal-adjacent cells. Moreover, our results indicate that proteins implicated in the described crosstalk are predominantly over-expressed by the tumor stroma. This result underscores the important role of this compartment in CRC carcinogenesis.

One important finding is that DEG are enriched in transcription factors. This indicates the existence of a transcriptional program driving the altered expression pattern observed in adjacent mucosa. A loop including members of the AP-1 family of transcription factors emerged as the most significant one in the analysis. Interestingly, these TF are over-expressed in both adjacent mucosa and tumor tissue. AP-1 members homo or hetero dimerize to assemble the activator protein 1 (AP-1). AP-1 transcription factor acts synergistically with SMAD3/SMAD4 component and is implicated in the regulation of a variety of cellular processes including proliferation and survival, differentiation, growth, apoptosis, cell migration, and inflammation [19, 20]. Topologically, these nodes have a low centrality but a high eccentricity in the transcriptional network. This result can be a little controversial since it is widely accepted that the more centered a node is the more important their functional role in the studied system [21]. However, a recent publication postulates that nodes with high eccentricity could be quickly activated by external factors [22]. This observation could explain the radial position of AP-1 members Jun, Fos, FosB and JunB into the transcriptional network as important fast effectors mediating response against the tumor.

We hypothesized that cytokines and other signaling proteins secreted by the tumor activate membrane receptors of adjacent mucosa cells that initiate this transcription factor activity. Tumor-secreted growth factors act as paracrine agents distorting the normal tissue homeostasis. In turn, tumors are both maintained or attacked by signals from the surrounding microenvironment inducing stromal reaction, angiogenesis and inflammatory responses. To gain insight into the molecular mechanisms underlying this phenomenon, a bi-tissue PPIN analysis strategy was performed to extract patterns of receptor activation in both directions from adjacent mucosa to tumor and vice-versa. Robo genes appeared as the most recurrently receptors activated in tumor membrane by Slit family of proteins. Slits have been implicated in regulating a variety of life activities, such as axon guidance, neuronal migration, neuronal morphological differentiation, tumor metastasis, angiogenesis and heart morphogenesis [23]. Several studies have demonstrated dual roles for Slit and Robo in cancer, acting as both oncogenes and tumor suppressors [24]. This bi-functionality is also observed in their roles as axon guidance cues in the developing nervous system, where they both attract and repel neuronal migration [25]. In CRC, Slit2 up-regulation has been reported as beneficial for the overall survival of patients [26]. Slit is under-expressed in patients with metastatic colorectal cancer and their over-expression in cells resulted in an inhibition of cell migration through AKT-GSK3β signaling pathway. In our data, no significant association between Slit2 or Slit3 level of expression and prognosis was found.

CLU-RELN-LRP8 was other afferent axis to consider for further analysis. CLU codifies the protein Clusterin that has been described as both tumor suppressor and pro-survival factor in colon cancer depending on the intra- and extracellular microenvironment crosstalk [27]. In fact, it has been reported that Clusterin is a protein that shares the intracellular information with the microenvironment and it also experiences a systemic diffusion, acting as a factor that synergistically interacts with their surrounding microenvironment [28]. Moreover, it has been proposed as a diagnostic biomarker in colon cancer [29]. The other CRC-mucosa-secreted protein activating LRP8 receptor in tumor is Reelin (RELN), a glycoprotein that plays an important role in neuronal migration through the activation of lipoproteins receptors such as LRP8 [30]. Also, Reelin has been proposed as a pro-metastatic factor due to their role in cancer cell migration through TGF-β pathway activation [31].

Efferent pathways were also of interest. LIF is a member of the IL6 family of cytokines that displays pleiotropic effects on various cell types and organs [32, 33]. In our system, its receptor LIFR was expressed in the colonic epithelium. It has been reported that LIF stimulates the Jak/STAT pathway to produce nitric oxide (NO) [34, 35]. Based on this, we hypothesize that, in our model, tumor LIF activates Jak-STAT pathway in normal epithelial cells through LIFR receptor leading to NO release and the subsequent creation of a pro-inflammatory environment. Moreover, in our model, Angiotensinogen (AGT) was produced by the tumor and their receptor (AGTR1) was located in membrane from adjacent tissue. Since Angiotensinogen is the precursor form of the active peptide Angiotensin, the pair AGT-AGTR1 makes up the renin-angiotensin system (RAS), usually associated with cardiovascular homeostasis but recently associated with tumor growth [36]. RAS could play a synergistic effect with LIF inducing NO production, leading to inflammation, macrophage infiltration and tumor-induced fibrosis. In addition to their pro-inflammatory role, it has been reported that NO can activate notch-signaling pathway leading to the induction of tumors [37].

Conceptually, an active sub-network includes differentially expressed and connected proteins in a given phenotype. Here we have described a sub-network including membrane receptors over-expressed in normal adjacent tissue acting together in cell-adhesion and with functions on cell surface signal transduction that finally activate the AP-1 transcription factor. ROR2 has emerged as an important link in the crossroad between cell surface entering signal and Fos/Jun transcriptional role as previously described [38]. ROR2 is tyrosine-kinase receptor that plays an important role in developmental morphogenesis [39] and in our network it was activated by the tumor-secreted WNT5A, a WNT pathway signaling mediator.

We do not exclude the possibility that genes having a pivotal role in crosstalk between adjacent and tumor tissue also have a direct relationship with prognosis. In our data, expression of Fos and Jun were found to be protective when over-expressed in adjacent but not in tumor tissue (log-rank p-value = 0.042). Further studies are needed to experimentally corroborate this hypothesis and to test the utility of these transcription factors as prognosis biomarkers. Nevertheless, a complex equilibrium between positively pro-survival and pro-apoptotic signals given by the microenvironment ultimately influences the tumor growth and their plasticity. This could be one of the reasons why prognosis signatures that only take into account tumor but not adjacent tissue expression fail to accurate predict patients’ outcome [40].

The study has some methodological and technical limitations. Though we obtained adjacent mucosa from the farthest resection margin and usually required at least 10 cm, it is possible that some of the variability observed among adjacent mucosa might be related to the distance to the tumor that we cannot analyze. Also, despite a careful dissection of tumor blocks before RNA extraction was done, a normal adjacent tissue infiltration can exist in some tumor samples. Regarding analytical methods, the network analysis only considered well-annotated genes. Some TFs were excluded from the transcriptional network analysis due to their low variability in our data. For these reasons, some genes with a putative role in colon tissue remodeling could have been missed. In fact, we did not find TGF-β, proposed as an important microenvironment modifier [4] because its probeset had very low expression level in our microarray. Finally, our study only included colon specimens, which could raise a concern about generalizability of the results. However, we have previously analyzed that the expression levels are very similar in colon and rectal tumors [41] and this has been confirmed in the TCGA study [42].

In conclusion, gene expression in cells comprising normal adjacent tissue in CRC patients is not so normal and this could have important implications in colorectal cancer prognosis and progression. A systems-level approach has been useful to gain insight into the molecular mechanisms by which adjacent mucosa activates a transcriptomic program in response to cytokines and other signaling proteins secreted by the tumor. We hypothesize that a crosstalk exists, not only between different cell communities within the tumor bulk, but also between colorectal tumor cells and adjacent mucosa, which reacts against the tumor like against a wound. Tumor-secreted growth factors act as paracrine agents distorting the normal tissue homeostasis. In turn, tumors are both maintained and/or attacked by signals from the surrounding microenvironment inducing stromal reaction, angiogenesis and inflammatory responses. Disrupting this intricate molecular network of cell-cell communication and signal transduction could be a therapeutic target in CRC patients.