NEW YORK – The gain or activation of the MYC gene at diagnosis correlates with poor prognoses and resistance to Merck's Temodar (temozolomide) in gliomas, according to a study published this week.
Scientists affiliated with the Hong Kong University of Science and Technology (HKUST) developed a machine learning algorithm that uses genomic data and other clinical factors to predict how a patient's glioma will evolve and respond to therapy.
The group used the algorithm, Cancer EvoLution for LOngitudinal data version 2 (CELLO2), to identify a particularly high-risk subset of glioma patients characterized by MYC amplifications and showed that this genetic feature drove resistance to Temodar in vitro.
Importantly, genetic data in the study came from 544 adult diffuse gliomas predominantly from patients of East Asian ancestry, in which the MYC amplifications appeared more common than in individuals of European genetic ancestry.
"The study provides some remarkable insights into the differences in genomic characteristics between patients from an East Asian, compared to a mostly Caucasian genetic background," Roel Verhaak, a professor of neurosurgery at Yale School of Medicine and who was not involved in the study, said by email.
The low-level MYC amplifications enriched in East Asian patients correlated with an increased risk of hypermutation in response to temozolomide therapy. Other significant differences included the lower penetrance of the rs55705857 glioma germline risk allele in East Asians compared to Caucasians.
Jiguang Wang, professor of life sciences at HKUST and the study's senior author, described the difference in germline rs55705857 penetrance between East Asian and European populations as "remarkable."
"This evidence implies that East Asian and Caucasian gliomas have an alternative way of activating the MYC pathway," he said. "A number of European gliomas are associated to germline risk allele rs55705857, while the East Asian gliomas tend to develop MYC copy number gain."
"These results are both important for our understanding of the disease, as well as relevant to patients and possibly patient clinical management," said Verhaak.
Sequencing data came from the Glioma Longitudinal AnalySiS (GLASS) consortium and over 100 newly sequenced glioma pairs from East Asian patients at multiple clinical sites in China and South Korea. All patients were diagnosed with diffuse glioma of three molecular subtypes known as IDHwt, IDHmut-noncodel, and IDHmut-codel.
The HKUST team used a computational tool called tumor evolutionary directed graphs (TEGDs) to infer the timewise order of genetic alterations in each patient and interrogated these "evolutionary trajectories" for early genetic and clinical features predictive of later events occurring after clinical interventions.
Several such features appeared to drive clonal evolution in response to therapies, including CDKN2A deletion, Ki-67 expression, MGMT methylation, and copy number gain of MYC.
Wang and his colleagues used these early predictors to train machine learning models capable of inferring a patient's cancer evolution. In the study, these models accurately predicted hypermutation and grade progression at tumor recurrence. They also pointed to MYC gain as a key genetic event promoting Temodar-associated hypermutation within the IDHmut-noncodel patient subgroup.
Importantly, the group not only recapitulated these findings in experiments using glioma cell lines, thereby validating the model's predictions to some degree, but showed the mechanism by which MYC likely drove Temodar resistance.
MYC appeared to accomplish this by binding to open chromatin and transcriptionally active genomic regions, making key mismatch repair genes more vulnerable to treatment-induced mutagenesis and therefore, hypermutation.
Notably however, not all Temodar-treated patients who later developed recurrent, treatment-resistant tumors displayed hypermutation.
"The non-hypermutated patients contained fewer MYC amplifications compared to the hypermutated ones," Wang said. "These patients might have other unknown mechanisms of drug resistance. Our lab is currently trying to identify such mechanisms to explain why the other patients are resistant to treatment."
Nonetheless, Wang said that his group's findings suggest that drug resistance in cancer is to some extent shaped by detectable genetic alterations in early cancer stages, which emphasizes the importance of tailoring therapies to individual cancers.
Wang is now planning a prospective study to further validate and refine CELLO2, while several lab members are also applying it to predicting tumor evolution in other cancer contexts.
"We are still improving the CELLO2 platform for better accuracy and a better user-friendly interface," he said.
While the algorithm is currently open access, Wang mentioned that "some key techniques might be patented in the future."
"[This] study is both timely and exciting," said Yale's Verhaak, adding that the new datasets complement existing resources from the GLASS Consortium. "Together, they will enable [scientists to] address many additional questions, for example, on the distribution of mutational signatures, or to understand the role that structural variations may play in glioma evolution."