There are some cell lines in which epithelial–mesenchymal transition (EMT) occurs as the result of a joint effort between H-ras and transforming-growth-factor-β (TGF-β). How relevant this is to the multistage nature of in vivo tumour progression, though, is a burning question.

So Allan Balmain's group studied whether changes in the levels of H-ras and TGF-β are significant during tumour progression. They did this using a series of well-characterized tumour cell lines that arise from initiated cells that carry activating mutations in the Hras1 gene. And, as they now report in Nature Cell Biology, Smad2 (a downstream target of TGF-β signalling) and H-ras surpass discrete thresholds during progression from early-stage papillomas, through squamous carcinomas, to late-stage undifferentiated spindle-cell tumours.

First, the authors studied the molecular changes that occur when squamous carcinomas are converted into spindle-cell tumours. TGF-β-mediated transcriptional activity was very high in the spindle cells, and phosphorylated Smad2 accumulated in the nucleus, which indicated that the TGF-β pathway was activated in these cells. Furthermore, in primary material from spindle-cell tumours, but not from differentiated tumours or squamous carcinomas, Smad2 was phosphorylated and predominantly localized in the cytoplasm.

Although Smad2 alone induced changes in the migration of squamous carcinoma cells, only in the presence of increased levels of mutated H-ras did changes in cell shape and the expression of genes such as α-smooth-muscle actin (a mesenchymal marker) occur, resulting in EMT.

The authors then investigated whether, once this stage has been reached, TGF-β signalling through Smad2 is still necessary for tumour progression. Expression of a dominant-negative form of Smad2 showed that this is indeed the case; spindle cells that expressed this construct reverted to a more epithelial phenotype and took on many features of epithelial gene expression. Notably, surface expression of αvβ3 integrin was lost, and this coincided with the loss of collagen-matrix invasion. In vivo, this correlated with an inability to form tumours. By contrast, parental spindle cells or spindle cells that express a dominant-active form of Smad2 formed tumours, and those that were formed by dominant-active Smad2 were particularly invasive. Expression of dominant-active Smad2 also promoted extravasation into the target tissue, and a subsequent increase in lung metastases.

As the ability of a tumour to metastasize is the main determinant of whether or not cancer patients die, these findings that different thresholds of H-ras and TGF-β activity — intermediate levels of Smad2 that cooperate with H-ras to induce EMT and invasiveness, and even higher levels of Smad2 that are required for metastasis — are crucial for metastasis offer the opportunity for the design of small-molecule inhibitors to prevent the spread of tumours.