For patients with solid tumours, the biggest threat to survival is metastasis — the spread of tumour cells from the original growth to other sites in the body. For biologists studying cancer, a major challenge is to identify the underlying molecular changes that switch cells to a metastatic state, with the ultimate aim being to devise treatments that inhibit metastasis. Previous research has concentrated on the contribution of individual genes to metastasis. Now, gene-expression profiling, using high-density DNA microarrays, is revolutionizing our approach to studying cancer.
On page 532 of this issue, Clark et al.1 describe how they used this approach to identify several genes that are selectively upregulated in metastatic mouse and human melanoma cells compared with their non-metastatic counterparts. They find that, remarkably, overexpression of one of these genes — RhoC — can stimulate metastasis all by itself. Meanwhile, Bittner et al.2 (page 536 of this issue) have used microarrays to compare different subgroups of human melanoma, and also find a distinct pattern of gene expression in highly invasive melanoma cells.
In DNA microarrays, probes for the messenger RNA products of up to 10,000 different genes are present on a single 'chip', usually a glass slide. The chips are used to determine which of these genes are expressed (that is, which are transcribed into mRNA) in a selected cell type3. This technology has already been used to classify cancers, such as leukaemia, according to their gene-expression profile3, 4.
As their starting material, Clark et al.3 used two poorly metastatic cell lines derived from a human and a mouse melanoma. Using an in vivo selection procedure in mice (see Fig. 1 on page 532), they isolated highly metastatic variants of each cell line, and compared the expression profiles of 7,070 human genes or 6,347 mouse genes in three different human or mouse metastases with their non-metastatic counterparts. Of course, these genes represent just a fraction of those in the respective genomes. But only 16 of the human genes studied — and a similar number of the mouse genes — were expressed at a significantly higher level in the metastases than in the primary tumours. This small number indicates that the products of these genes are likely to be important in metastasis, rather than being upregulated as an indirect consequence of the altered cell phenotype.
Most of these upregulated genes were not found in both the human and the
mouse metastases. Some of the genes, however, were not actually present on
both arrays. It is perhaps not surprising that there is no unique gene-expression
'fingerprint' for metastatic melanomas: Bittner et al.2 show that the gene-expression profiles of different subgroups of
human melanoma vary greatly. What is significant is that three of the genes
identified by Clark et al. — those encoding fibronectin, thymosin 
4
and RhoC — showed increased expression in all of the human and mouse
melanoma-derived metastases1.
What do these three proteins do, and why might they be upregulated in metastases
(Fig. 1)? Fibronectin is a component of the extracellular
matrix and promotes the migration of several cell types, including melanoma
cells5. The production of fibronectin by melanoma cells might
allow them to lay down their own promigratory matrix. Interestingly, increased
fibronectin expression also correlated with higher invasive capacity in the
study by Bittner et al.2 (which did not, however, include
RhoC or thymosin 4).
Figure 1: The results of Clark et al.1 and Bittner et
al.2 indicate that the proteins RhoC1, fibronectin1, 2 and thymosin 
4 (ref. 1) are
required for the metastasis, or spread, of melanoma cells.
![Figure 1 : The results of Clark et al. and Bittner et
al. indicate that the proteins RhoC, fibronectin and thymosin |[beta]|4 (ref. 1) are
required for the metastasis, or spread, of melanoma cells. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nature/journal/v406/n6795/images/406466aa.0.jpg)
Cell migration involves protrusion of the plasma membrane (lamellipodium
extension) at the leading edge of the cell; the formation of new sites of
adhesion to the extracellular matrix at the front; the release of old adhesion
sites at the back; and contraction of actomyosin-based cytoskeletal filaments
in the cell body to move the bulk of the cell forward. Rho proteins regulate
actomyosin-based contractility and adhesion turnover, and RhoC might be involved
in enhancing these steps of cell migration. Thymosin 4 buffers monomeric
actin in cells, and could act to provide actin monomers for rapid polymerization
into actomyosin filaments in lamellipodia. Fibronectin is an extracellular
matrix protein, and deposition of fibronectin by the cell could promote migration
by signalling through specific receptors on the cell surface.
Thymosin 4 binds monomeric actin, a component of the cytoskeleton,
and may act as an actin buffer, preventing spontaneous polymerization of actin
monomers into filaments but supplying a pool of actin monomer when the cell
needs filaments6. It is not clear how increased expression of
thymosin 4 might promote metastasis, but it is likely to relate to the
need for cells to migrate.
Clark et al.1 chose RhoC for further study. RhoC is a member of the Rho family of small GTP-hydrolysing proteins, several of which are known to regulate cell migration7, 8; RhoC is highly related to RhoA and RhoB. Unlike the related Ras genes, the three Rho genes have not yet been found to be mutated in human cancers, but they can contribute to cell transformation to a cancerous state, motility and invasion, at least under experimental conditions8, 9. However, the expression of RhoA and RhoB is not altered in metastatic melanoma cells1, and it will be interesting to find out how RhoC expression is specifically increased.
Clark et al. also find that simply overexpressing RhoC in vitro
in the poorly metastatic human melanoma cell line can induce it to become
highly metastatic. This is surprising because several other genes are also
upregulated consistently in the metastases, and one would expect these genes
to contribute to metastasis. Perhaps the higher level of overexpression of
RhoC in the transfected cells compared with the metastases allows RhoC
to induce metastasis in the absence of other changes. The poorly metastatic
cells are on the verge of metastasis, and it probably takes little to push
them further. It would be interesting to know the gene-expression profile
of the transfected cells. Is overexpression of RhoC sufficient to
induce the increased expression of the other metastasis-associated genes,
such as those encoding fibronectin or thymosin 4? Or is the pattern
of expression now different?
It is perhaps unexpected that increased expression of RhoC enhances the in vitro migration and invasion of melanoma cells, given that constitutively active RhoA inhibits the migration or invasion of several cell types10, 11, 12. However, RhoA is not as potent at inducing melanoma metastasis or migration as RhoC1, and the effect of RhoC on cell migration in vitro has not previously been studied. These results point to a subtle difference in the properties of RhoA and RhoC. Also, it is wild-type, not constitutively active, RhoC that is overexpressed in the transfected melanoma cells. Wild-type Rho proteins cycle between a GTP-bound and a GDP-bound conformation, whereas constitutively active mutants are locked in the GTP-bound form. Cycling may be important for Rho proteins to promote cell migration. Indeed, a 'fast-cycling' mutant of RhoA has considerably greater cell-transforming potential in fibroblasts than does constitutively active RhoA13.
One drawback to microarray analysis is that it cannot identify post-transcriptional changes in protein expression or activity. Mutant Ras, for example, can contribute to cancer but would not be identified by microarray analysis of human tumours, as only its activity — not its expression level — is altered by mutation. But the approach taken by Clark et al. and by Bittner et al. does provide an unbiased method by which to pinpoint important, and potentially new, contributors to cancer. It should now be possible to work back from these transcriptional changes to identify the one or more key genetic or epigenetic events that induce metastasis. The challenge will be to extend this work from a mouse model of metastasis to human patients.