SIAM News Blog

Modeling MGMT Methylation to Better Understand Brain Cancer

By Lina Sorg

Glioblastoma multiforme (GBM) is a malignant, intrinsic strain of brain cancer. It is both extremely fast-growing and remarkably aggressive. With standard treatment methods, patients have a median survival rate of 15 months; chance of two-year survival hovers around 30 percent. Treatment regimens begin with surgery to remove most of the tumor, followed by radiation and chemotherapy with a cytotoxic drug called temozolomide (TMZ), which attaches a damaging methyl group to DNA and triggers cell death. Methyltransferase (MGMT), a DNA repair gene, codes for a protein that can mend the lesions caused by TMZ.

The region of DNA that initiates gene transcription is called the gene promoter. Methylation—a form of alkylation in which a methyl group replaces a hydrogen atom in DNA—of a gene promoter can silence a gene. Cells with a methylated MGMT promoter region are sensitive to TMZ, while cells with an unmethylated MGMT promoter region are resistant. Nevertheless, researchers do not yet fully understand the evolutionary processes behind the emergence and development of TMZ-resistant tumor subpopulations. 

Several recent clinical studies have compared MGMT methylation status in matched GBM samples at diagnosis and recurrence of a tumor. These studies revealed frequent reduction in MGMT methylation between primary and recurrent GMB, meaning that most methylated primary tumors become unmethylated recurrent tumors after TMZ treatment. This shift begged the following question: can evolutionary selection alone explain the reduction in MGMT methylation between tumor diagnosis and recurrence, or does TMZ actively influence MGMT’s methylation status? During a contributed presentation at the 2018 SIAM Conference on the Life Sciences, which took place this week in Minneapolis, Minn., Katie Storey of the University of Minnesota presented a stochastic evolutionary model of GBM response to standard treatment to investigate the role of MGMT methylation in TMZ resistance. 

Katie Storey's stochastic continuous-time branching model has three distinct phases.

Storey’s multi-type, continuous-time branching model incorporates mechanisms of MGMT methylation and demethylation. It also accounts for three cellular subtypes. Type 1 comprises GBM cells with fully-methylated MGMT promoters, while type 2 contains GBM cells with one methylated and one unmethylated DNA strand; these are called hemi-methylated MGMT promoters. And type 3 encompasses GBM cells with completely unmethylated MGMT promoters. Type-1 cells are drug-sensitive, but type-2 and type-3 cells are drug-resistant and capable of mending TMZ-inflicted lesions. “Each cell type is equipped with intrinsic birth and death rates, which vary during treatment,” Storey said.

She used both clinical and experimental data from GBM patients to calibrate the parameters of her model, which has three distinct phases. Phase 1 includes tumor growth prior to detection, as well as the subsequent surgery and a three-week recovery period. Phase 2 consists of concurrent radiotherapy and chemotherapy with another three-week recovery. In phase 3, radiation ceases and adjuvant chemotherapy with TMZ repeats in 28-day cycles until the tumor recurs. 

The division of type-1, drug-sensitive cells is fairly straightforward, but division of type 2 and type 3 cells is decidedly more complicated.
Immediately after cell division, methylation and demethylation at representative CpG sites in the DNA drive conversions between cell types. Three major enzymes are responsible for DNA methylation: DNMT1 conducts maintenance methylation, while DNMT3a and DNMT3b handle de novo (new) methylation. During DNA replication, new cells pair with parent DNA strands. The division of type-1, drug-sensitive cells is relatively straightforward, while type 2 and type 3 cell division is more complicated. After replication, DNMT1 moves through the DNA and methylates any unmethylated sites that were missed during replication. DNMT3a and DNMT3b follow suit, methylating sites regardless of their current methylation status. Storey incorporated these methylation dynamics into her model. Variables include the probability of maintaining methylation at a previously-methylated CpG site and the probability of de novo methylation at an unmethylated CpG site, both following cell division. 

Storey used a Monte Carlo simulation to examine the percentage distribution of methylation at the time of primary GBM diagnosis. First she tested whether evolutionary selection alone—without the influence of TMZ—can explain methylation patterns; the outcome was not consistent with clinical results. Next, Storey wondered whether TMZ impacts de novo methylation. “When there’s no de novo methylation in the presence of TMZ, we still see a pretty small change between detection and recurrence,” she said. “It’s not sizable enough to account for the clinically-observed drop.”

Finally, Storey hypothesized that the presence of TMZ influences maintenance methylation. The corresponding simulation averaged a 30 percent drop in methylation levels. “There was a substantial decrease in MGMT methylation percentage between direction and recurrence, which was consistent with clinical results,” she said. This suggests that TMZ actively inhibits maintenance methylation, accounting for the clinically-observed downward shift in MGMT methylation between GBM tumor diagnosis and recurrence. It also confirms that evolutionary selection alone cannot sufficiently explain this phenomenon. In the future, Storey hopes to determine an optimal adjuvant TMZ dosing schedule—contingent upon the level of MGMT methylation at tumor diagnosis—to more effectively treat GBM.

Lina Sorg is the associate editor of SIAM News


blog comments powered by Disqus