Imagine a dandelion flower that's 'gone to seed' — the smallest gust of wind disperses the seeds in every direction and you never know where this weed will grow next. Metastasis of tumour cells is similar — cells in the primary tumour undergo a sequence of genetic changes that cause the altered cells to migrate and form secondary tumours. But why do some patients with cancer develop metastases years after the primary tumour has been resected? As Schmidt-Kittler and colleagues report in Proceedings of the National Academy of Sciences, the answer challenges current views that metastasis is a late event in tumour formation.

The biological age of a tumour can be estimated from the number of chromosomal aberrations it contains, so a similar pattern of aberrations would be expected in metastases if cells migrate late in tumorigenesis. As bone is the preferred site of metastasis in patients with breast cancer, the authors began by isolating single disseminated cancer cells from the bone marrow of patients presenting with breast cancer with (M1) or without (M0) metastases. The cells were analysed by comparative genomic hybridization (CGH) — an in situ technique that is used to identify regions of amplification and deletion in the genome — and their chromosomal profile was compared with that of their matched primary tumour. The number of chromosomal aberrations was increased in M1 cells. But, to the authors' surprise, M0 cells had fewer chromosomal aberrations than either M1 cells or the primary tumours, indicating that the cancer cells migrate early in tumour development. But, how early does this migration occur?

The authors looked for chromosomal rearrangements that occurred during crisis — due to telomere shortening — and found that, unlike M1 cells, M0 cells had not undergone this process. As the passage through crisis is often completed early in tumour development, the M0 cells must migrate far earlier than previously thought. But are M0 cells really related to the primary tumour? The authors used a microsatellite polymerase chain reaction technique to identify loss of heterozygosity and showed that at least some M0 cells share sub-chromosomal deletions with the primary tumour. Cancer cells must therefore leave the breast tissue before the tumour is removed and accumulate the changes that generate the secondary tumour over time, which explains the long latency periods. So, targets for adjuvent therapies should be defined directly on disseminated cancer cells.