For several decades, the molecular changes within tumor cells were studied in order to understand factors responsible for promoting tumor progression and metastasis, while little attention was paid to the possible contributory role of tumor microenvironment. Recent evidence suggests that the tumor microenvironment, which is composed of a very complex network of extracellular matrix (ECM) proteins and many cell types, such as endothelial cells, stromal (mesenchymal) stem cells, pericytes, fibroblasts and immune cells, plays a critical role in tumor progression and metastasis
[36, 37]. Among these components, MSCs have been the focus of intensive investigation
[9, 17, 38–45].
In the present report, we examined the crosstalk between tumor cells and MSCs and we investigated the effect(s) of tumor secreted factors on MSCs at the cellular and molecular levels. As surrogates for malignant tumors, we employed a number of well characterized cancer cell lines. We reported that secreted factors from FaDu cells led to significant morphological and genetic changes in MSCs with enhanced expression of pro-inflammatory cytokines, and similar responses were also observed when additional tumor cell lines were evaluated. However, these effects were not universal for all malignant cell lines. For example, MCF7 and HT-29 did not exert these effects. Our findings corroborate recent findings of the presence of morphological and functional changes in mouse MSCs in response to cancer cell lines (MDA-MB-231, PANC-1 and U87) CM
, which exhibit a carcinoma-associated fibroblast (CAF)-like myofibroblastic phenotype.
Interestingly, several of the pro-inflammatory molecules identified in the current study have been linked to cancer progression. For instance, cancer cells that overexpress CXCL1 and 2 were found to be more primed for survival at metastatic sites, and are capable of attracting CD11b(+)Gr1(+) myeloid cells into the tumor that enhance cancer cell survival and enhance their chemoresistance and metastatic ability
. In addition to that, CXCL2 was also found to be involved in cancer-associated bone destruction
. A recent study has reported differentiation of human MSCs into pericyte–like cells upon exposure to glioblastoma tumor CM
. In our current study, we observed no evidence of differentiation of MSCs into pericytes or endothelial-like cells using an in vitro angiogenesis assay (Figure
6a). In fact, MSCs exposed to FaDu or MDA-MB-231 CM failed to form any vascular-like tubular networks compared to control MSCs, suggesting MSCs have lost their ability to support angiogenesis
. Nonetheless, MSCs exposed to tumor CM also exhibited poor adipocytic and osteoblastic differentiation potential (Figure
6b), probably as a result of differentiation into pro-inflammatory cells. Glioblastoma are known for their high angiogenic capability and the secretion of high levels of VEGF
, which might account for the variable effects of CM from breast, lung, prostate, and head and neck cancer models investigated in the current study compared to published glioblastoma data
; hence, the response of MSCs to tumor secreted factors can vary depending on the tumor type.
Our gene expression data revealed significant correlation between the expression of a panel of genes involved in inflammation and the metalloprotease pathway (CCL8, CCL5, CXCL6, CXCL5, SAA1, MMP12, EHF, CCL3, CSF2, CXCL3, IL6, IGF2, CXCL2 and IL1b) in MSCs exposed to FaDu and to those exposed to MDA-MB-231, PC-3 and NCI-522 CM, while the expression of these genes was almost unchanged in MSCs exposed to MCF7 CM (Figure
3). These data support our hypothesis of the ability of tumor cells to recruit MSCs to their stroma and which in turn induce inflammation, either directly or through recruiting circulating immune cells (Figure
10b). It seems that this model does not apply to all cancer models since in the MCF7 model, MSCs seemed to promote tumorigenicity via direct interaction with tumor cells (Al-toub et al., in preparation).
Bioinformatics and pathway analysis of gene expression data from tumor cell lines revealed that the phenotypic changes were mostly observed in MSCs exposed to CM from cell lines with a pro-inflammatory nature (such as, FaDu and PC-3, Figure
7c). Indeed our investigation has identified tumor-derived IL1β to be the primary driver of the pro-inflammatory phenotype observed in MSCs exposed to tumor CM, whereas treating MSCs with recombinant IL1β mimicked the effects of tumor CM at the cellular and molecular level (Figures
Nonetheless, we also identified signaling via FAK and, to lesser extent, MAPK to be critical for the induction of the observed phenotype (Figure
4). In contrast, pharmacological inhibition of TGFβ signaling in MSCs led to substantial enhancement in the observed changes in phenotype and gene expression in MSCs exposed to MDA-MB-231 CM (Figure
5a and b), which was also associated with a slight increase in cell proliferation [see Additional file
5: Figure S3]. Treating MSCs with recombinant TGFβ1 and TGFβ3 in the presence of FaDu CM led to significant inhibition of the observed phenotype at the cellular and molecular level (Figure
5c and d), which further implicated TGFβ signaling in negatively regulating MSC differentiation in response to tumor CM. Thus, our findings corroborate previous studies suggesting a role for the TGFβ signaling pathway in regulating mesenchymal stem cell differentiation