The stromally-derived paracrine factors, HGF and TGF-β, have enjoyed the limelight in recent literature on epithelial-stromal crosstalk. Most cell types have the capacity to both express and respond to TGF-β25
. in contrast, HGF is primarily expressed by fibroblasts, while the cognate receptor, c-Met, is primarily expressed by epithelia45
. There are multiple reports to support the transforming ability of HGF46
. As also discussed above, the role of TGF-beta is more complex and involves both tumour-suppressive and promoting roles.
The above mentioned report on tissue recombination of irradiated pancreatic fibroblasts with pancreatic cancer cells describe an elevated metastatic potential of the developing tumours36
. This phenomenon was associated with increased c-Met activation in the carcinoma cells and TGF-β1 expression by the irradiated pancreatic fibroblasts. The fact that the authors did not observe a concomitant increase in HGF secretion by the fibroblasts suggests the possibility that the expression of an HGF-related ligand may be involved, such as MSP (macrophage stimulating protein) which can activate a heterodimeric receptor consisting of c-Met and the c-Met related molecule RON. Alternatively, the overexpression of c-Met, a common finding in many cancers, can be associated with ligand independent activation48
or increased sensitivity to physiological levels of HGF (note new reference 48A
Apparent contradictions in our current understanding of the role of TGF-β in epithelial-stromal interactions have also emerged. How can overexpression of TGF-β with enhanced signalling and loss of TGF-β signalling through knockout of Tgfbr2 both result in enhanced tumourigenesis? The overexpression of TGF-β by fibroblasts in mammary and pancreas transplantation models of cancer formation35,36,44
can affect both epithelial cells and stromal fibroblasts. In contrast, the loss of TGF-β signalling in the Tgfbr2fspKO
model affects TGF-β signalling only in fibroblasts. However, the enhanced HGF signalling in the two models developing mammary tumours is likely the result of epithelial responses to paracrine factors (since c-Met receptors are primarily found in epithelial cells48
). It is of interest that conditional deletion of Tgfbr2
in the epithelia of the mammary gland and prostate have no detectable phenotypic alterations (unpublished observations).
poses other important questions. If TGF-β is stimulatory of an activated stromal environment, how does the loss of TGF-β signalling in stromal fibroblasts still allow for an otherwise activated stromal phenotype? The fact that tumours in the Tgfbr2fspko
mice are associated with an activated stroma suggests important roles for factors other than TGF-β in the development of an activated tumour microenvironment. As already discussed, a dual role for TGF-β as a tumour suppressor and promoter when acting on epithelial cells has emerged. However, in light of the above data42,44
, the paradigm of the early and late roles of TGF-β singalling on tumourigenesis needs to be adapted to include the influences of both fibroblasts and epithelia in tumour susceptible tissues. In particular, both the direct effects of TGF-β on epithelia (direct via TGF-β receptors) and secondary effects (through the regulation of other growth factors) need to be considered.
While the role of TGF-β in epithelia from various tissues has long been accepted as growth inhibitory and morphogenic through multiple downstream singalling proteins39
, the proliferative role of TGF-β in fibroblasts has been less clear as a result of the heterogeneity of fibroblasts (Box 1
). Notably NIH3T3 and dermal fibroblasts in culture are growth stimulated when treated with TGF-β59
. However, in the Tgfbr2fspko
mice, the loss of TGF-β signalling in fibroblasts of the entire mouse 42
, for the most part, had little effect on fibroblast abundance in most tissues examined (i.e. no evidence of stromal hyperplasia in the skin, lung, kidney, mammary gland, esophagus, liver, small intestine, or colon). This might indicate a tissue-selective role for TGF-β singalling in maintaining fibroblast homeostasis or the presence of redundant growth inhibitory signals from other factors that do not require the TGF-β type II receptor. There was however significant stromal hyperplasia in the prostate and forestomach of Tgfbr2fspko
mice, the same organs that undergo epithelial transformation42
. It is therefore important to keep in mind that multiple changes in several growth factor pathways as well as tissue-specific responses will ultimately determine the outcome of the complex epithelial-fibroblast interactions in tumours compared to that in the non-disease state (Box 1
). In addition, our current knowledge on the role of many growth factors is primarily derived from studies of epithelial cells in culture. Undoubtedly, the ability to conditionally knock out growth factor signaling specifically in fibroblasts or epithelial cells will continue to advance our understanding of these networks of paracrine and autocrine signalling on epithelial proliferation and transformation that operate in vivo.
Box 1. Not all fibroblasts are created equal
The figure illustrates stromal-epithelial interactions in normal and tumour tissues. Panel a
illustrates wild-type stomach fibroblasts interacting with both the squamous epithelia (SE) of the forestomach and columnar glandular epithelia (GE) of the stomach body (below, green line indicates the area of epithelial transition). Panels b
depicts the progression of squamous cell carcinoma (SC) in the forestomach as it infiltrates the glandular stomach epithelia (GE). The progression of the carcinoma is likely maintained through reciprocal signals to and from the fibroblasts (asterisk, panel b
). Fibroblasts have the capacity to proliferate and/or take on an activated form that can be supportive and even initiate epithelial hyperplasia and eventual tumourigenesis. In turn, carcinoma-derived paracrine proliferative factors signal to the stroma. The cartoon in panel d
depicts the non-carcinogenic balance epithelia and fibroblast maintain through dynamic interactions between the two compartments. A curious aspect of the Tgfbr2fspko
is that, apart from the prostate and forestomach, the other tissues in the mouse had no apparent signs of carcinogenic transformation. Functional differences between fibroblast from different organs my explain this.
In a reductionist approach to understanding epithelial-stromal interactions in normal tissues and in cancer, the fibroblast component is often generalized and obscure differences in fibroblasts from different tissues or differences within the same tissue. However, it is becoming clear that different fibroblast can have distinct functions. For example, during lung development, epithelial induction capacities are known to differ based on whether the mesenchyme is derived from the trachea or the tips of the growing lungs69
. The origin of the mesenchymal fibroblasts apparently determines their sensitivity to sonic hedgehog (Shh) derived from the developing lung epithelia. Shh signals through the patched receptor homologue (Ptc) found in the adjacent mesenchyme to stimulate proliferation70
. In turn, the specific mesenchyme supports the normal development and branching morphogenesis of the lung epithelia.
In mice, the fibroblasts associated with the squamous epithelium of the forestomach and esophagus appear phenotypically similar to the fibroblasts associated with the adjacent glandular epithelia of the stomach body. Although a similar proportion of fibroblasts in the esophagus, forestomach, and glandular stomach compartments were deficient for TGF-β singalling in Tgfbr2fspko mice, only the squamous epithelia of the forestomach responded to the oncogenic paracrine signals. It is possible that the loss of TGF-β signalling in stromal fibroblasts results in proliferative autocrine and paracrine signals to which the fibroblasts and squamous epithelia of the forestomach, uniquely responded with the formation of invasive carcinoma, but not other other epithelia of e.g. the esophagus which showed no hyperplastic or neoplastic response in the Tgfbr2fspko mice. Thus in this model, the paracrine signals required for malignancy of the glandular epithelia likely differ from that of the adjacent squamous epithelia.