The results of our study add to some previous observations that fibronectin and periostin may have prognostic significance in ovarian cancer. The exact mechanism of this phenomenon remains to be elucidated. However, some explanation comes from the fact that both of these proteins are involved in shaping the structure and regulating function of the extracellular matrix (ECM). ECM is a major component of the tumor microenvironment, together with the vascular network, fibroblasts, immune and stem cells, etc. This complex environment is deeply involved in the control of tumor progression.
4.1. The Role of Fibronectin in Ovarian Cancer
Cellular fibronectin has broad biological activity, serving as a structural scaffold and playing an important role in the regulation of cell adhesion and motility. The main fibronectin-specific integrin, which is present on most cells, consists of two subunits: α5 and β1 (reviewed in: [33
]). Importantly, fibronectin–integrin interactions proved not only able to supply the mechanical force necessary for cell migration, but also to be engaged in signal transduction. Thus, although the role of fibronectin in cancer is only partially understood, there are multiple potential mechanisms emerging, through which this role may be exerted. It seems that the main mechanism is based on the activation of the focal adhesion kinase (FAK). Downstream FAK/Src signaling can have very diverse cellular effects: from the induction of epithelial to mesenchymal transition (EMT) and acquisition of invasive phenotype, to providing proliferative signaling and even inducing antiapoptotic activity (reviewed in: [6
]). It was confirmed by Mitra et al. that such signaling occurs in ovarian cancer cells: fibronectin binding to α5β1-integrin leads to a direct association of α5-integrin with the c-Met kinase, activating it in a ligand-independent manner. Subsequently, c-Met associates with Src, and activates Src and FAK. This mechanism may promote the invasion and metastasis of ovarian cancer cells [34
Moreover, fibronectin seems to take part in the cross-talk between ovarian cancer cells and other cells of the tumor microenvironment. In our work, we observed that tumor samples derived from omental metastatic disease demonstrated significantly stronger expression of fibronectin than samples of primary tumors. Similar results are returned when querying FN1
using the CSIOVDB transcriptomic microarray database which integrates gene expression data from 48 published datasets (comprising 3,431 epithelial ovarian cancer samples) [35
]: it could be found that FN1
mRNA is significantly increased in metastatic peritoneal tumors compared with primary ovarian tumors (p
= 2.22 × 10−6
for comparison between classes: “Tumor” vs. “Peritoneal”). Further, Kenny et al. showed that fibronectin expression was higher in the omental metastases than in the primary tumors [36
]. These authors performed a series of very elegant in vitro and in vivo experiments, showing that TGF-β produced by ovarian cancer cells can activate a TGF-β receptor/RAC1/SMAD-dependent signaling pathway in the mesothelial cells, what induces EMT and results in overexpression of fibronectin. Fibronectin, in turn, promotes the adhesion and invasion of cancer cells. This mechanism may be responsible for the preferential implantation of ovarian cancer cells on the surface of omentum, through interactions with mesothelial cells lining its surface. Such a hypothesis is supported by the observation that blocking fibronectin production in primary mesothelial cells in vitro decreases the adhesion, invasion, and proliferation of ovarian cancer cells. A pro-tumourigenic role for fibronectin is also confirmed by the fact that SKOV3 ovarian cancer cells have reduced invasiveness and metastatic potential in fibronectin knockout mice [36
Another key component of the cancer microenvironment are cancer associated fibroblasts (CAFs). Yeung et al. examined the gene expression profile of microdissected ovarian cancer cells and CAFs derived from HGSOC tumors [37
]. They observed that cancer cells overexpress TGF-β, while CAFs can respond through activated Smad signaling. Thus, it can be speculated that similar relationship to those described between ovarian cancer and mesothelial cells, also pertain to CAFs.
Increased fibronectin expression is generally regarded as a marker of mesenchymal phenotype. Gene expression data from CSIOVDB indicate that higher FN1
mRNA significantly correlates with an EMT score. Accordingly, the highest FN1
expression is observed in the mesenchymal subtype of ovarian cancer which is characterized by the worst prognosis [38
In this work, we have demonstrated that stronger stromal fibronectin expression in the tumor is associated with the shorter OS of patients with advanced ovarian cancer. Prognostic significance of fibronectin was first noticed by Franke et al. [22
] who analyzed 211 tumors of different histological type to observe that higher fibronectin expression correlated significantly with worse OS and was an independent prognostic factor (p
= 0.009). Also in the CSIOVDB, FN1
mRNA levels higher than the median are significantly associated with both, reduced OS (p
< 0.0001) and reduced DFS (p
< 0.0001). In the multivariate analysis FN1
expression retains as an independent predictor of DFS.
CSIOVDB also indicates that higher FN1
expression is related with resistance to the first line chemotherapy. It was also observed by others that the level of fibronectin (as well as matrix metalloproteinase 9) was significantly higher in tumor samples and in the ascites fluid of the recurrent ovarian cancer patients and in the group of patients who died from the disease, as compared to the non-recurrent cases [23
]. These data further support the connection between fibronectin expression and worse prognosis in ovarian cancer.
CSIOVDB provides also the information that higher FN1
expression is related to serous histology, with higher FIGO stage, and higher grade. These results were not confirmed in our tissue microarray analysis (see Figure 4
). Discrepancies between CSIOVDB and our results may be partially due to the fact that we analyzed protein expression, while CSIOVDB supplies the data on mRNA quantity; these are two distinct levels of gene expression, not always coherent.
There have also been attempts to use fibronectin for diagnosis and treatment. It was shown that a nine-biomarker diagnostic panel, including fibronectin, could better discriminate ovarian cancers from benign ovarian masses than the OVA1 test (at a threshold sensitivity of 90%, the nine-marker panel gave 88.9% specificity, compared to 63.4% for the OVA1) [39
Making use of fibronectin in therapy is based on the assumption that, in adults, alternatively-spliced fibronectin variants containing extra domains A and B (EDA/EDB) are primarily limited to sites of malignancy [9
]. One strategy which has been tested, utilizes recombinant human antibody L19 (specific to the EDB) fused with cytokines such as interleukin 2 (IL2) or tumor necrosis factor α (TNFα). As EDB is strongly expressed in stromal and neo-vascular structures during cancer progression, this strategy could deliver therapeutic agents directly to the tumor and reduce side effects of the systemic activity of these cytokines [41
]. A phase I/II clinical trial with L19-IL2 in patients with solid tumors allowed to evaluate safety, tolerability, recommended phase II dose and showed early signs of activity [42
]. However, another phase I/II clinical trial with L19-TNF in patients with solid tumors did not showed objective tumor responses. Transient stable disease occurred in 19 of 31 evaluable patients [43
4.2. Periostin Significance in Ovarian Cancer
Periostin, also called osteoblast-specific factor 2 (OSF-2), is a multifunctional ECM glycoprotein. Periostin is capable of binding to multiple integrin receptors (αvβ3, αvβ5, α6β4), thus affecting the regulation of the intracellular signaling pathways associated with protein kinases PI3K/AKT and FAK. This protein plays a role in the adhesion and migration as well as in EMT and remodeling of the extracellular matrix. Periostin is also implicated in the metastases of cancer cells and lymph- and angiogenesis (reviewed in: [16
]. Some data indicate that periostin may exert an important role in ovarian cancer and may be potentially useful as a prognostic and predictive biomarker.
In our previous microarray study, we found the 96-gene prognostic signature including POSTN
which had the highest change in mRNA level (fold change 21.69) between two groups of patients with significantly different OS [5
]. Further, the output from CSIOVDB indicates that POSTN
mRNA level is significantly associated with OS and DFS. However, in the multivariate analysis periostin is an independent prognostic factor only with regard to DFS. Further, Karlan et al. (2014) identified a prognostic gene signature containing POSTN
]. They found that genes from this signature were related with TGFβ- and integrin-signaling. Patients with an upregulated POSTN/TGFβ
signature had significantly shorter OS than patients with lower expression of these genes.
Another microarray study performed by Ryner et al. (2015) revealed that POSTN was overexpressed in primary resistant ovarian cancers, while downregulated in chemosensitive ones [46
]. In their study, POSTN
was found to be regulated coordinately with FAP, LOX, TIMP3
, similarly like in our previous study [5
]. We also share a common observation that a higher degree of desmoplasia is correlated with the higher stromal expression of periostin. In addition, Ryner et al. observed that periostin is more highly expressed in recurrent tumors than in non-recurrent ones. Using in situ hybridization and IHC, they revealed that periostin was predominantly produced by CAFs.
Sung et al. (2015) evaluated periostin expression by IHC, using a tissue microarray with 308 samples derived from ovarian tumors of different histology and stages [25
]. They observed a correlation of stromal periostin expression with more advanced FIGO stage, suboptimal cytoreduction, tumor recurrence, and survival: patients with high periostin expression in tumor stroma had shorter OS and DFS. Similarly, as in our study, there was no correlation between periostin expression in cancer cells and survival.
4.3. FN1&POSTN Score
Both analyzed proteins, fibronectin and periostin, are engaged in similar cellular processes and are indicated as related to the survival of ovarian cancer patients. In addition, we have shown that both proteins have higher expression in omental metastases than in primary tumors, which supports their role in metastatic process. There are not many data concerning their mutual expression, however it was postulated that periostin can be a modulator of fibronectin production during gingival healing [47
]. Thus, it is possible that both proteins can be expressed coordinately or within short time interval. For these reasons, we decided to check whether the combined score based on their immunohistochemical staining in the tumor stroma (FN1&POSTN score) will perform better than individual scores for each protein. Indeed, in the multivariate analysis, this score was significant (p
= 0.044), with a p
-value slightly worse than that for fibronectin only (p
= 0.037), but better than periostin only, which was insignificant. When analyzing all five survival curves for patients with scores of 2, 3, 4, 5, and 6 separately, it could be noticed that this result was spoiled by the occurrence of two patients who behaved unexpectedly. One patient with a score 2 tumor (indicating favorable prognosis) had the shortest OS (4.8 months) out of the entire cohort of 108 patients. Likewise, one patient with score 6 (indicating worst prognosis) had the second longest OS (131.17 month) out of the whole cohort. These exceptions cause the FN1&POSTN score to behave similarly to classical prognostic factors, for which it is also frequently observed that some patients with good prognosis progress quickly and die early, while some patients with bad prognosis have unexpectedly long OS. This indicates that further studies are necessary to find molecular markers (or their panels) which will enable more precise prognosis.