Skip to main content

UCSF Glioblastoma Projects Using Molecular Data to Inform Subtypes, Match Patients to Treatments


NEW YORK − Researchers involved in the Glioblastoma Precision Medicine Program, an effort launched at the University of California, San Francisco two years ago, are making headway in cataloging the molecular aberrations that drive glioblastoma and assessing whether treating patients based on these genomic markers can improve their outcomes. 

Although it is more common for patients to have brain metastases from aggressive tumors in the lung, breast, or colon, glioblastoma is the most common malignant tumor that originates in the brain. There are between 12,000 and 14,000 glioblastoma cases diagnosed each year, but these tumors are difficult to treat.

Glioblastomas tend to be resistant to many different treatments because of their location in the brain and because of the nature of these tumors, according to Susan Chang, director of the UCSF Division of Neuro-oncology. "It's becoming increasingly recognized that glioblastoma, like other cancers, is not a single type of tumor," Chang said. "There are many different subgroups because of the molecular features … They have different mutations and different signaling pathway abnormalities that differentiate them." 

These molecular features can be associated with glioblastoma patients' prognosis and influence their ability to respond to treatments.

Chang noted that in glioblastoma, there have been some incremental advances in broadening treatment options, though standard of care still involves radiation, chemotherapy, or surgery.

"Patients are not cured. And in fact, by 12 months or so, more than 50 percent of the patients have already progressed, despite initial treatment with radiation and chemotherapy and maximal safe resection," she said. "It's a very refractory cancer."

However, there have been efforts to better understand the genomic underpinnings of the disease. The National Cancer Institute's Cancer Genome Atlas broke ground into mapping the genetic alterations in glioblastoma in 2008, and further expanded upon this effort in 2013

The NCI researchers identified four molecular subtypes of glioblastoma that responded differently to aggressive treatment: proneural, neural, classical, and mesenchymal. Additionally, in younger adults, a methylation signature that reflects MGMT gene silencing correlated with improved survival. The Cancer Genome Atlas also homed in on five genetic mutations that appear to play a role in glioblastoma including NF1, ERBB2, TP53, PIK3R1, and TERT.

Even though the TCGA identified different molecular features of glioblastoma, "there was no real clinical annotation to that information," said Chang. Researchers could group cases with specific molecular features and look at survival on a population level, but they could not integrate the molecular data with other prognostic features such as age, extent of the surgery patients had, how much tumor was removed, or the impact of any neoadjuvant or follow-up treatments.

This is the type of analysis researchers at UCSF, led by Anette Molinaro, wanted to do when they decided to build the Brain Tumor Center Database. Within the database, researchers collect molecular and genomic data, alongside clinical data, so they can construct a more complete picture of how patients with certain types of glioblastomas fare on different therapies. The searchable database is one of the major projects underway in the Glioblastoma Precision Medicine Program.

"That was really to comprehensively annotate any patient operated [on] at UCSF with a glioblastoma, whether at the time of diagnosis or at the time of recurrence," said Chang. Since December 2017, UCSF started sending portions of resected tumors from grade 4 glioma or glioblastoma patients to be sequenced on the UCSF500 Cancer Panel, which scours approximately 500 cancer-associated genes for mutations, copy number alterations, fusions, and other structural rearrangements. UCSF also performs additional analysis on the tumor samples such as methylation profiling and RNA sequencing. The resulting information is then uploaded into the database, which has been operational since 2018.

Currently, the database contains the molecular and clinical profiles of approximately 500 patients, including 90 patients who were sequenced retrospectively.

"The goal is to try to be able to group patients based on their genetic abnormalities and treatment history and then be able to match that potentially with not only the outcomes that are happening with those patients, but specific treatments that we could then target for those patients," said Chang.

The group published the first set of analysis on 165 adult patients in Neuro-Oncology in November 2018. 

Out of these glioblastoma patients, 10 had a H3 K27M mutation and 19 had an IDH mutation. Among the IDH-mutated glioblastoma patients, the most common additional mutations were in TP53, ATRX, CDKN2A, and PDGFRA.

These findings went beyond the earlier TCGA research and improved researchers' understanding of the biology of glioblastomas. "The IDH mutation is one that's been seen in lower grade tumors," explained Chang. Low grade gliomas can over time progress to higher grade gliomas, or glioblastomas, and "patients whose tumors arise from a lower grade and become grade 4, they tend to have the IDH mutation."

On the other hand, patients with newly diagnosed or primary glioblastomas don't tend to have the IDH mutation. Older patients also tend not to have the IDH mutation, according to Chang.

Research like this suggests the possibility that IDH inhibitors might be a good option in this setting. To date, IDH inhibitors have only been approved in leukemia, but there is growing interest in exploring the efficacy of these drugs in brain tumors.

Forma Therapeutics and Agios have programs targeted to IDH-mutated gliomas, which have shown early signs of efficacy. Oncoceutics is also advancing a targeted drug for the treatment of H3 K27M-mutated gliomas through Phase II trials, which has garnered support from the National Brain Tumor Society.  

In the Neuro-Oncology paper, researchers further reported that in patients who did not have an IDH or H3 K27M mutation, the most common genetic alterations were in TERT, EGFR, CDKN2A, PTEN, NF1, TP53, PIK3R1, PDGFRA, CDK4, MDM2, LZTR1, and STAG2.

Chang highlighted that the TERT promoter mutation is one of the more common mutations in primary glioblastomas, but there is no standard treatment for this subset of patients. TERT mutations affect the way telomerase behaves in cancer cells and can enable cancer cells to divide continuously. "It gives the cells immortality," Chang said, adding that there have been attempts "to attack telomerase" but these have been unsuccessful because our normal cells need telomerase to work. However, if there was a way to target the TERT mutation in glioblastoma patients without affecting normal telomerase function, she said that could be an interesting targeted treatment avenue to explore for this subgroup of patients.

Within UCSF's database project, researchers are also testing patients at the time of recurrence and charting the genomic evolution of glioblastoma. "We have now what we call paired samples from the same patient across the trajectory of illness," said Chang. "And what we can do is see the mutations that arise at the time of recurrence that the tumors may not have at diagnosis."

Feasibility of genomic medicine

While the database project is aimed at identifying the genomic features that can inform treatment, in a smaller project within the UCSF Glioblastoma Precision Medicine Program, researchers are applying these learnings and testing out how recurrent glioblastoma patients fare when they receive tailored treatments based on actionable mutations identified through the UCSF500 panel. The 15-patient feasibility study, led by Jennifer Clarke, is fully enrolled. 

Researchers are primarily interested in seeing if in the span of a month after patients have been diagnosed with recurrence, they can conduct molecular profiling, capture the treatment-relevant data, and treat patients with standard-of-care or US Food and Drug Administration-approved drugs or combinations. The study will also track patients' overall survival and progression-free survival on the treatments they receive.

Patients in this study have all undergone surgical resection of their tumors and had a sample sent for sequencing. The results for each patient are discussed within a molecular tumor board, which includes a genomic pathologist, a pharmacologic researcher, and other experts. They recommend drugs based on the molecular data, factoring in available information such as the ability of agents to penetrate the blood-brain barrier and side effects.

When considering combinations, for example, the board considers the toxicities that might result from pairing different agents and references the literature for information on potential dose adjustments and other lab tests that may be necessary.

Researchers may also develop additional cell lines, organoids, or avatars from patients' tumors to evaluate if these ex vivo models can predict how they will respond to the treatment.

"Every patient becomes a little trial, because you're [evaluating] drugs based on each patient's specific characteristics," said Chang.