BACKGROUND AND METHODS

Despite the overwhelming clinical significance of metastases, the cellular and molecular mechanisms involved are largely unknown. In order to define significant differences between primary colon carcinomas and their metastases, we analyzed gene expression profiles of 12 sets of triple-paired tissues using 19 K human oligonucleotide microarrays. A total of 36 microarray experiments were analyzed by unsupervised two-way hierarchical clustering and multi-dimensional scaling (MDS).

RESULTS

Both methods completely distinguished normal mucosa from carcinoma, but failed to demonstrate a complete classification of primary and metastatic carcinomas. We found a separable tendency to be classified into the primary and metastatic colon carcinomas by MDS. In supervised hierarchical clustering, we identified 80 genes that were differentially expressed between paired primary and metastatic colon carcinomas. The 80 identified genes also successfully distinguished three validation sets of primary and lung-metastatic colon carcinomas. A specific set of genes was identified that distinguished the metastasis from the corresponding primary tumor in nearly half of the metastases analyzed.

CONCLUSIONS

We suggest that a more accurate model of the metastatic potential is based on a global tumor expression pattern along with the appearance of distinct metastatic variants. This molecular profiling may be useful for the future study of colon cancer metastasis.