NEW YORK (GenomeWeb) – A team led by researchers at the University of San Diego Moores Cancer Center has developed a new computational tool that analyzes cancer patients' gene copy number losses and gains in a way that can predict affected molecular pathways and identify therapies that may be effective in targeting them.
The team published a study in Nature Communications last week describing their initial work with the tool, called Haploinsufficient/Triplosensitive Gene (HAPTRIG), in serous ovarian cancer. According to the authors, by identifying pathways that should be significantly disrupted by somatic copy number alterations (SCNAs), HAPTRIG was able to suggest novel potential therapies with strong anti-tumor effects in pre-clinical testing.
The study's lead authors, UCSD pathology professor Dwayne Stupack, and Joe Ryan Delaney, currently a postdoc in reproductive medicine at the center, said in an interview this week that their investigation was based on a desire to glean information from monoallelic copy number changes, which can make up the bulk of a tumor's genomic changes, to help fight cancer.
Monoallelic changes have been largely unappreciated as a contributor to cancer, the authors wrote. But across cancer types, more genes are affected by single gene-dose changes than by allelic deletions, larger amplifications, and point mutations combined.
"People have looked at copy number changes before but they only focused on copy number changes that involved total loss of a gene or on massive amplifications," Stupack said, "not on these plus-one and minus-one changes."
"I think a lot of people acknowledge that cancer cells are not very good caretakers of their genome, and these changes are just a symptom of this lack of care. That may be the underlying truth, but in the end, there is still selection and tumor cells are being selected for [aggressiveness] through one pathway or another pathway," he explained.
With HAPTRIG, the team created a tool that could cut through the "noise" of SCNAs that have limited or no effect to find instances where single gene copy number losses do appear to combine in a way that significantly affects a targetable cancer pathway.
According to Stupack and Delaney, the tool relies on two important factors lacking in other algorithmic methods for identifying pathways, such as Gene Set Enrichment Analysis (GSEA).
First, HAPTRIG looks specifically for proteins present across many organisms "with the idea that these are going to be the ones that are really going to be central cogs in the machinery," Stupack said. Secondly, it takes into account how many other proteins each cog interacts with.
"For example, the hub of a wheel would be rated more importantly than something around the outside of the wheel," he explained.
Although some sequencing efforts have identified potentially targetable genes in serous ovarian cancer, this has only been in a small minority of individuals. Meanwhile, patients frequently show widespread SCNAs.
"With [GSEA], when we ran that across the genomes from serous ovarian cancer [patients], there was only one pathway that was detected out of the 180 that we tried. We were very skeptical of that because there are so many changes present in the genome, so we decided to add these additional markers of computation, like conservation across different organisms and [connections between proteins], to develop HAPTRIG," Delaney said.
When the team used HAPTRIG on samples from serous ovarian cancer patients, they were able to pick out several expected pathways — for example, suppression of the p53 pathway, enhancement of the focal adhesion pathway, and disruption of homologous recombination repair pathways. But apart from these, their other top hits were not canonical drivers.
One unexpected pathway that stood out in particular was a network of genes involved in cell autophagy.
Narrowing in on this, the team first showed that autophagy was indeed suppressed in serous ovarian cancer cells. They then picked a combination of existing US Food and Drug Administration-approved compounds that target this pathway by disrupting proteostasis, which they dubbed COAST (Combination of Autophagy Selective Therapeutics).
To validate their computational findings, the team tested the COAST drug combo, which includes chloroquine/nelfinavir with rapamycin or dasatinib, or both, in several cellular and mouse models.
Not only did the drugs appear to work strongly, but they also showed activity in a mouse xenograft derived from a recurrent chemotherapy-resistant patient. The chemo drugs cisplatin and docetaxel did not alter the growth of the model at all, while COAST showed a "striking complete ablation of tumor growth," the authors wrote.
Importantly, the COAST combination appears to be less toxic than standard chemotherapy and is relatively inexpensive, which Delaney and Stupack said presents a rationale for immediately moving from these early experiments into clinical evaluation for human use.
Stupack is now working on an application for funding for a Phase I trial of COAST.
"We had such good results that we thought it should go quickly into the clinic to look at safety," he said. "Because all of these drugs are approved already, it was pointed out to us that people may start using them if they get desperate, so it's all the more important to do safety experiments right away."
Beyond this investigation of COAST in serous ovarian cancer, Delaney and Stupack said they see potential for HAPTRIG in finding similar patterns in other cancers that might reveal other novel therapy combinations to bring through clinical trials.
They haven't solidified what their next steps will be, but they have made the complete HAPTRIG code available online to allow the community to easily perform their own analyses. The resource allows the investigation of 187 pathways in 21 cancer types.
According to Stupack, while this is far off and will require much more research, it's possible that HAPTRIG could also become an engine for therapeutic prediction on a patient-by-patient basis, rather than only for identifying targets common to a group of patients.
For example, he said, as institutions and companies move toward broader use of exome sequencing to look for driver mutations, the same data could be analyzed for SCNA patterns that speak to other targetable pathways as well.
"When you have sequencing becoming cheap enough, you could analyze a person's exome, look for genes that were lost, map the pathways, and you could use this as a sort of an adjunct to just looking for driver mutations to come up with a more complete treatment strategy for a patient," Stupack said.