Comparative Genomics Fine-tunes Noisy Data

Study in Mice Implicates Gene in Spread of Melanoma

Genomics has invigorated the hunt for genes involved in cancer. It gives scientists an unprecedented capability: where they once slogged through swamps of experimental data to locate just one gene that might contribute to cancer, they now can scan the whole genome and catalog all of the genetic aberrations in tumors. Unfortunately, the view from up here is blurry and ever-changing.

Photo by Graham Ramsay

A genomic screen in a mouse led Lynda Chin (left) and Minjung Kim to a gene that is involved in spurring melanomas to metastasize.

DNA can be deleted, repeated, or mutated—any one of these able to shut down tumor suppressors and turn on oncogenes. These genetic changes can affect a single base pair of DNA or a large region of a chromosome that is home to multiple genes. In total, the list of candidate cancer genes can be hundreds long, so it is difficult to know which ones are truly significant and which are just along for the ride. In the end, the genomic view of cancer offers many paths to wander down, but no certainty about which will be dead ends. Being able to sort out the promising paths is important because, as Lynda Chin notes, it may be easy to identify candidate cancer genes with a genomewide screen, but it still takes two or three years to do the work to show a gene is truly relevant.

Chin, HMS associate professor of dermatology at the Dana–Farber Cancer Institute, led a study published in the June 30 Cell that suggests the effort of sorting out promising leads in human cancers can benefit from the mouse. Using genomic analyses of a mouse model of melanoma, her team uncovered a gene that helps tumors metastasize. In a companion paper in the same issue, Scott Lowe, a professor at Cold Spring Harbor Laboratory, employs a mouse model to uncover genes involved in some liver cancers. The authors believe that together their papers make a case for using comparative cancer genomics across species in the hunt for cancer genes.

Cross-species Mutations
Mouse models of cancer have been maligned because they often fail to predict the effects of treatments in humans. The standard model is a xenograft—cultured human tumor cells are injected under the skin of an immunodeficient mouse and allowed to grow. Instead, Chin’s and Lowe’s models involve engineering tumors with specific genetic mutations known to be important in human cancers. These models may not capture the full spectrum of variation in corresponding human disease, but they have been shown to better model subsets of cancers that arise from specific genetic changes.

Chin’s team, led by research fellow Minjung Kim, studied tumor cells that had acquired the potential to metastasize in a mouse model of nonmetastatic melanoma. The team thought these cells may have developed genetic aberrations that would shed light on metastasis in human melanomas. Even though primary melanoma is entirely curable, the metastatic disease is one of the deadliest cancers, with no existing effective therapeutic options. Finding genes that relate specifically to metastasis could lead to drugs that prevent melanoma’s deadly effects and provide markers to distinguish deadly melanomas from benign ones.

Image adapted by Rachel Eastwood from original courtesy of Lynda Chin

Filtering out the noise. A genomewide analysis of metastatic melanomas in the mouse and human shows regions that are repeated or amplified (red) and deleted (green). The blue arrows point to analogous regions that were amplified in both models. Compared to the relatively noisy human cancer genome, the mouse genome contains a much narrower set of candidate genes that might contribute to metastasis. The y axis shows copy number changes in tumor DNA, with 0 representing 2 copies, negative representing deletions, and positive representing amplifications.

Chin and her colleagues performed array-comparative genomic hybridization, a technique that allows researchers to detect regions of DNA that are amplified or deleted across the genome. The altered cell lines shared an amplified region that was not observed in their parental nonmetastatic cells. The region is analogous to a large chunk of human DNA on chromosome 6 that is frequently multiplied in metastatic melanomas in humans. Compared to this large territory, the region in the mouse was just a sliver, containing just eight genes. “In the mouse, since the genome is relatively stable and very simple, when you get the candidate you have high confidence” that it is a relevant one, Chin said.

Based on gene expression levels in the cells, the team narrowed this list to one candidate, Nedd9. In human melanoma samples, Nedd9 was overactive in more than half of the metastatic samples compared to benign samples. The gene seems to work by activating a molecule called focal adhesion kinase, which helps cells migrate and invade tissues. Chin believes this kinase and its related pathway are promising drug targets.

Cancer Genomics
Recent efforts to launch a federally funded Human Cancer Genome Project have focused attention on the problem of how best to apply genomic tools to cancer. Some scientists have questioned the usefulness of embarking on a billion-dollar project to sequence something as protean as cancer genomes, which by definition are unstable. Scott Lowe explained, “There’s a lot of noise in human tumors that may not be relevant to the genesis of the tumor.” A large-scale cancer-sequencing project in the U.K. has had one major success: the discovery that the majority of melanomas harbor mutations in the gene BRAF. However, other analyses have not yielded such clear candidates.

Instead, Lowe believes that “in the case of human cancers, we’ve probably found the changes that occur really frequently.” What is left are genes that are modified in a relatively small percentage of tumors—some of these may be important drug targets and others are not. But finding them involves diving into the noise, a potentially daunting task given the high number of possible targets.

Raju Kucherlapati, the Paul C. Cabot professor of genetics at the HMS–Partners Center for Genetics and Genomics, believes that methods to detect changes in gene copy number, like the kind used by Chin and Lowe, are a more cost-effective place to start. Amplification and deletion of genes is a major event in cancer genomes, sometimes in concert with mutations—but other times causing normal genes to become silent or overactive. “Even if we could efficiently sequence all the genes in the human cancer genome, you still wouldn’t be able to find these genes because they don’t have mutations in them,” he said. Chin noted that the key is to use comparative information wherever possible—across species, technologies, and cancer types—as filters that will help bring relevant information out of the noise.

—Courtney Humphries

top