Researchers use gene-profiling to predict metastasis

Molecular oncologist Dr Kent Hunter, of the Laboratory of Population Genetics at the National Cancer Institute in Bethesda, Maryland, told the recent Lorne Genome Conference that it may soon be possible to predict a patient's risk of developing metastatic cancer based on their individual gene-expression profiles.

"Most primary tumours are not lethal," Hunter said. "Most deaths are due to metastatic cancers. Secondary tumours can occur decades after excision of primary tumours, and there's no good way to identify who will develop them.

"We don't know which breast cancers will metastasise, so we have to treat everyone with aggressive adjuvant therapy, even though a significant proportion of patients won't require it. Who can we spare the rigors of chemotherapy, and who should we treat more aggressively?

Despite its clinical importance, metastasis is an extremely inefficient process, Hunter said. A primary tumour may shed millions of cells into the circulatory system, but only a few will ever form secondary tumours.

The classical view contends that, when a primary tumour forms, only a tiny subpopulation of its cells acquire the "right" combination of mutations required to metastasise and colonise other organs. But gene-expression profiles don't support this model, suggesting instead that the propensity to metastasise is pre-programmed into the cell by the mutations that trigger the tumour.

So why don't all escapee cells metastasise?

Zeroing in on metastisis

Hunter's team sought to answer the question using laboratory mice that develop metastatic lung cancer by 100 days of age. They devised a simple out-breeding scheme to probe the genetic background against which secondary cancers occur.

The mice are heterozygous (two different alleles at many gene loci). Some of the combinations increase or reduce the risk of metastatic cancer, while others have no effect.

Assigning the mice to high- and low-risk lines, the NCI researchers performed QTL (quantitative trait loci) mapping to identify chromosomal segments shared by individuals within these lines.

They zeroed in on a gene-rich region on the proximal arm of chromosome 19, which QTL mapping suggested probably harboured a gene that reduces the risk of metastasis.

In these regions, 500 genes lie within a 10-megabase block. Kent's team performed a gene-for-gene comparison of the maternally and paternally derived alleles within this block, for the high and low-risk mice.

They narrowed their search for the anonymous, low-metastasis risk gene to just five genes, then sequenced all of them to identify any polymorphism that correlated with reduced metastasis risk.

One of the prime suspects was Sipa1, a member of the signalling pathway that regulates expression of Rac 1 GTPase in cells.

The low-metastasis risk allele of Sipa 1 has a single amino acid substitition that affects the strength with which it "docks" to another protein in the same signalling pathway.

Hunter's team use RNA interference to knock down expression of the Sipa 1 gene in the high-metastasis risk mice, using RNA interference - if they were right, the mice were likely to be carrying two high-risk alleles of Sipa 1.

The RNAi knockdown reduced Sipa 1 protein levels in the mice by 75 per cent. "It almost completely ablated the capacity of implanted tumours to metastasise," Kent said.

"It also turned out to have effects on the primary tumour - Sipa 1 is a tumour growth suppressor, and a metastasis enhancer. It's the first known gene with opposite effects on metastasitic and primary tumour - if you overexpress Sipa 1, the effects reverse."

Hunter's team searched international cancer databases for associations between Sipa 1 and cancer in humans.

They found an association in prostate cancer: Sipa 1 was upregulated in patients with metastatic tumours, in comparison to patients with non-metastatic primary tumours.

Predictive genetics

"If these genes really exist polymorphically [in human populations], the clinical implication is that we can use predictive genetics to identify those patients requiring more aggressive chemotherapy after surgery, and those who can safely go home."

Hunter said gene-expression microarrays were being developed to identify markers associated with poor prognosis in specific types of cancer, and were already being used in clinical tests in the Netherlands - but they were only 80 per cent accurate.

"It has to be a lot higher than that if you're making clinical judgments about treatment," he said.

Both models of oncogenesis were partly correct. It is oncogenes and mutant tumour suppressors that conspire to create a primary cancer that has the potential for metastasis. But only a tiny minority of cells among the millions that detach from the primary tumour acquire the extra genetic errors that activate their metastatic potential.

The problem with using microarrays to predict metastasis risk, and devise a personalised therapeutic regime, Hunter said, is that the signatures of the plethora of cancerous cells that have no metastatic potential will swamp the faint signatures the few metastatic cells.

Thus, the patterns seen on gene-expression-array profiles, instead of reflecting the ripple-down effects of primary, oncogenic mutations, may actually be driven by the patient's own genetic background.

Hunter's team took a set of 17 genes that are under- or overexpressed in metastatic tumours and compared their expression levels in high- and low-risk mouse genotypes.

Expression patterns differered between the high- and low risk lines, but within each line, were almost identical.

"It suggests that the expression patterns we are observing are due to the genetic backgrounds of each line, not to oncogenes."

Hunter said natural polymorphisms, rather than mutations, can explain differences in metastasis risk, since the polymorphisms are constitutionally present in a patient's cells, the same expression levels should be observed even before a primary cancer develops.

In other words, it may be possible to predict a patient's risk of developing metastatic cancer, based on their individual gene-expression profiles, even before they develop cancer.

Hunter and his colleagues tested their hypothesis by crossing females of the same genetic background with males from the high- and low-risk lineages. Female offspring in the F1 generation segregated into high and low-risk groups for developing metastatic cancer from primary breast-cancer implants.

The same 17 genes were differentially expressed in both lines, even though they would not be abnormally expressed in tumours, Hunter said.

"If it works in mammary tissues, it will work in other tissues," he said.

For reasons that are unclear, the 17 genes are differentially expressed to the greatest degree in epithelial cells in mouse saliva.

"If we can do it in F1 mice, can we do it in outbred populations, including human populations?" Hunter asked.

"Yes we can - we collected saliva at 21 and 60 says of age, and separated the mice into high- and low-metastasis risk groups, and expression levels were 80 per accurate in predicting high-risk animals."

"We believe that if we profile the same genes in serum, we could get to 95-100 per cent accuracy."

Hunter said it was now clear that oncologists need to consider the individual patient's genetic background when assessing the risk of metastisis, and prescribing a therapeutic regime.

"For a cell to metastasise, it must jump a series of hurdles," he said.

"It has to detach from primary tumour, invade the surrounding tissues, enter the vascular or lymphatic system and survive there, then exit the circulation, survive in a completely novel microenvironment, then proliferate into a clinically relevant mass.

"If it fails at any point, it fails the whole process."

The rigours of this obstacle course, Hunter believes, explain why so few of the millions of cells that detach from the primary tumour form secondary, metastatic tumours.

"The gene profiles are measures of the height of the hurdles, which differ from person to person," he said. "Different tissues also have hurdles of different heights.

"A basal program, or a few relatively simple programs drive metastasis, but the gene complement will not be the same identical in every tissue tested.

"And many of the genes in metastasis are not expressed in tumour cells. It's not clear yet what they do, but one possibility is that they're involved in immune surveillance."