A genome-wide search reveals a genetic survival pathway that is switched on in many of the most aggressive glioblastomas.
Glioblastoma is a notoriously deadly form of brain cancer. It’s difficult to get drugs past the protective blood-brain barrier to treat the disease, and even when drugs do reach their target, the cancer cells can be alarmingly resilient. Now, a genome-wide search has enabled Howard Hughes Medical Institute (HHMI) researchers to reveal a genetic survival pathway that is switched on in many of the most aggressive glioblastomas.
In the May 23, 2010, issue of the journal Nature Medicine, HHMI investigator Michael R. Green and colleagues report on a pathway that helps keep glioblastoma cells alive. Blocking various components of the pathway with existing chemical inhibitors—including one compound already approved for the treatment of certain cancers—inhibits the growth of tumor cells, they report.
In 2002, Green’s lab at the University of Massachusetts discovered that a protein called ATF5 is a cellular survival factor. Active in a variety of cells throughout the body, ATF5 opposes signaling molecules that trigger cell suicide. Since that discovery, researchers have found evidence that excessive ATF5 activity helps keep many kinds of cancerous cells alive. The most convincing evidence, Green says, is for the protein’s role in glioblastoma.
ATF5 is often abundant in brain tumors. It’s an attractive target for potential therapies, because the protein is not usually present in mature neurons. Furthermore, studies have shown that blocking ATF5 can kill glioma cells without affecting healthy cells. However, ATF5 belongs to a family of molecules that Green says have proven difficult to block with drugs. The protein is a transcription factor, meaning it does its work by binding to and switching on certain genes.
“It turns out that it’s been very difficult to develop typical small molecule drugs to transcription factors,” Green says. “The entire industry has had better success in targeting enzymes—they’re just more ‘druggable.’” Since ATF5 is likely to exert its cancer-promoting effects with the cooperation of several such enzymes, those abettors might be better targets for potential therapeutics.
When Green and postdoctoral researcher Zhi Sheng began their search for genes that help keep ATF5 on in glioma cells, Green says, they were most interested in understanding why the protein might be upregulated in glioblastomas and other cancers. What they found was that the protein is part of a “druggable pathway,” with components that can be targeted with several existing compounds.
To identify that pathway, Sheng used RNA interference to test the effects of turning off each of the mouse’s nearly 30,000 genes. RNA interference, a natural process that scientists have exploited as a research tool, involves introducing into cells a short segment of RNA that looks like one of the cell's own genes. This prompts the cell to destroy all identical copies of that RNA, thus shutting down production of the protein made from that gene. Rather than targeting each gene individually, the team used a collection of thousands of short hairpin RNAs (shRNAs) that have their own barcoding system. A unique tag is built into every shRNA so that researchers can test thousands of genes in a single lab dish, and then track the identities of those that had interesting effects.
Genome-wide screens conducted with this library of 62,400 shRNAs, created by HHMI investigators Greg Hannon and Steven Elledge, have enabled researchers to find genetic contributors to a variety of processes, including tumor growth, metastasis, and viral infection. The new study, however, was less straightforward than previous screens. Green and his colleagues were searching for genes that, when shut off, dramatically reduced the activity of ATF5—but they knew that without ATF5, glioma cells would die. Dead cells would make it impossible for them to find and identify the responsible shRNAs.
So Sheng designed a system in which shutting off genes that are needed for ATF5 activity would instead keep cells alive. To achieve this reverse effect, they created a hybrid gene in which the regulatory region of the mouse Atf5 gene controlled the production of a diphtheria toxin receptor, which is not normally present in mouse cells. Diphtheria toxin can enter and kill cells only via this receptor. They introduced the shRNA library into cells containing this hybrid gene, and then added diphtheria toxin. Only those cells that had received an shRNA that turned off ATF5—and therefore turned off the hybrid gene—would survive in the presence of the toxin. However, because an shRNA that shut off the hybrid gene would also shut off the endogenous Atf5 gene, and therefore kill the cell, they also expressed a copy of the Atf5 gene controlled by the regulatory region of a gene that is always on in the cell.
When Sheng introduced shRNAs into glioma cells carrying the hybrid gene and then treated the cells with diphtheria toxin, only a few of the cells survived. By reading the genetic “barcode” on the shRNAs in those cells, he identified 12 genes that keep Atf5 on in glioma cells. The team chose to further analyze three of those—fibroblast growth factor receptor substrate 2 (Frs2), p21 protein-activated kinase-1 (Pak1), and cAMP response element–binding protein-3–like-2 (Creb3l2). These three genes have known roles in signaling pathways that control gene expression and cell growth, and Green and his colleagues speculated that they might work together to turn on Atf5.
Through a series of experiments, the team confirmed that blocking FRS2, PAK1, or CREB3L2 in glioma cells, either genetically or with chemical inhibitors, reduced the activity of ATF5 and triggered cell death. Following the signaling pathway further, they found that ATF5 keeps cells alive by turning on a gene called Mcl1, and blocking that step of the pathway could also kill glioma cells.
Up to this point, the team had conducted their experiments in mouse cells. To investigate whether their findings had clinical relevance, they enlisted neurooncologist Richard Moser and other colleagues at the University of Massachusetts Medical School. Examining samples of brain tissue from patients, the group confirmed that the proteins in the ATF5 pathway were commonly expressed in human tumors. They found at least one of the proteins in more than half of their glioma samples, whereas they were undetectable in healthy brain tissue. Notably, patients without ATF5 in their tumors survived significantly longer than patients whose tumors had the protein.
With this evidence for clinical relevance, the team proceeded to experiments in a mouse model of glioblastoma. For these, they used the chemical inhibitor from their previous experiments that is most common clinically—a drug called sorafenib, which blocks CREB3L2 from turning on the Atf5gene. Sorafenib, marketed by Bayer under the name Nexavar, is approved for the treatment of certain kinds of kidney and liver cancers, and is currently being investigated in clinical trials for its ability to treat many other cancers, including glioblastoma.
Green’s team had encouraging results when they used sorafenib to treat tumors in a mouse model of malignant glioma: mice developed fewer tumors and those they did develop were significantly smaller than those in untreated animals. Combining sorafenib with temozolomide—a chemotherapeutic agent currently used to treat glioblastomas—impaired tumor growth even more dramatically.
Despite this success, Green says sorafenib is “not a great inhibitor. It’s a first-generation inhibitor, and it’s not the most specific or the most efficacious. It works, but [I think] there are better inhibitors out there now.” He and his team are eager to begin experiments with newer inhibitors that might target various components of the ATF5 signaling pathway and lead to cancer cell death even more effectively.