—In reply

We agree with Fidler and Kripke that primary tumor cells are genetically heterogeneous. The relationship of this heterogeneity to the metastatic phenotype, however, is unclear. Although poorly metastatic cell lines can give rise to highly metastatic progeny after in vitro selection and transplantation, such experiments do not prove the existence of metastatic heterogeneity in vivo. Our microarray-based studies in human tumors support a model in which the propensity to metastasize reflects the predominant genetic state of a primary tumor rather than the emergence of rare cells with the metastatic phenotype. Our microarray experiments are only able to detect subsets of cells comprising a significant portion of the tumor.

It is formally possible, as Fidler and Kripke suggest, that the metastasis gene-expression signature that we observed is actually a composite signature resulting from distinct subsets in the tumor, with each subset expressing a portion of the metastasis signature. This interpretation, however, would require independent clonal expansion of multiple cells in the tumor, with each cell expressing a portion of the metastasis signature. The selective advantage of such various partial metastatic phenotypes is not obvious. We therefore favor the more parsimonious explanation that the mechanism of transformation of primary tumors dictates metastatic behavior.

Our results could also be explained by the mechanism proposed by Hunter, Welch & Liu, whereby host genetic background could be important in cancer metastasis. In this regard, they refer to an analysis of our recently reported 17-gene metastasis expression signature in transgene-induced mouse tumors arising in varying genetic backgrounds. Their findings, though difficult to evaluate without full review of the data, are of substantial interest, particularly the finding that primary tumors associated with metastases largely express the conserved 17-gene expression program due to these background differences. These findings raise the possibility that genetic modifiers in humans may influence the expression of metastasis programs.

The potential presence of such modifiers has important implications but does not demand another model of metastasis. Indeed, their results are consistent with our view that metastatic behavior is established early in the pathogenesis of tumors and is reflected in the bulk of primary tumors. The extent to which the metastasis program in human tumors is governed by tumor cell intrinsic factors as opposed to host cell factors must now be determined.

See Genomic analysis of primary tumors does not address the prevalence of metastatic cells in the population by Fidler and Kripke

See Genetic background is an important determinant of metastatic potential by Hunter et al.