Researchers from Ohio State University have developed a score combining methylation and gene expression that proved a stronger predictor for the prognosis of acute myeloid leukemia patients than previously identified genetic markers.
In a paper published at the end of December in the Journal of Clinical Oncology, the researchers reported the development and validation of the combined predictor, which integrates gene expression and methylation patterns associated with seven genes.
Guido Marcucci, the study's first author, told PGx Reporter this week that AML, and particularly its cytogenetically normal subset, has been well investigated in terms of genetic markers like mutations and gene expression.
"However," he said, "when we try to separate patients in terms of outcome based on what we have learned using current mutations and gene expression markers, it is clear that we are missing some part of the story."
"Patients with the best molecular profile, not all of them achieve long-term survival, and some will die despite having a favorable profile and vice versa for those with an [unfavorable] profile," He added.
According to the study authors, theirs is not the first effort looking to methylation as a potentially more powerful prognostic marker in AML, but previous efforts have not attempted to combine epigenetic and genetic information in a single predictive score.
In the study, Marcucci and his team first used next-generation sequencing to identify differentially methylated regions associated with already-known prognostic mutations in a discovery cohort of 134 cytogenetically normal AML patients over the age of 60. Using a set of the most common prognostic mutations in AML, including IDH1, NPM1, and FLT2-ITD, the team looked for methylation patterns associated with mutated versus wild-type patients, establishing a set of methylated sites present in the mutated patients.
The researchers then took the resulting list of methylated regions, and limited themselves to genes for which both methylation patterns and gene expression were significantly associated with patient outcomes.
Overall the group identified 80 individual genes for which higher promoter methylation was associated with longer overall patient survival. Seven of these genes were also linked to outcome through their expression levels, and the group collected these into a weighted score.
Patients in the discovery cohort with the lowest scores had more than an 80 percent increase in their odds of achieving remission and a three-fold lower risk of disease relapse or death compared with patients with the highest scores.
Interestingly, for each of the seven genes in the score, the pattern of gene expression and methylation predicting outcomes was similar— higher methylation and lower gene expression predicted better outcomes.
According to Marcucci, this was unexpected, but exciting for the group, because it represented additional biological plausibility.
"It was exciting, because it goes along with the idea that when you have a hypermethylated gene, expression is low, and it fits with the functional knowledge we have on how methylation may impact genes," he said.
Compared to other previously reported prognostic markers for AML, the team found that the combined epigenetic/genetic score was more predictive of outcome based on statistical measures of the different markers' Akaike information criterion, or AIC.
The team then tested the seven-gene score in four independent patient sets, representing a total of 355 patients.
In each, the prognostic performance of the score from the discovery cohort was mirrored closely, Marcucci said. Additionally, in multivariate models, the score remained an independent predictor of remission and overall survival after adjusting for other clinical and molecular variables.
Though the group's initial score was weighted, Marcucci said that the team found that the relative contribution of each of the seven genes in the score was very close, meaning a non-weighted scoring method, which would be easier to implement in a clinical commercial test, was also highly predictive of outcome.
Patients in the discovery set with either zero or one highly expressed gene — approximately 15 percent of the cytogenetically normal AML population according to the study authors — had a 96 percent remission rate, a 32 percent three-year disease free survival rate, and a 39 percent three-year overall survival rate, compared to 25 percent, 0 percent, and 4 percent respectively for patients with six or seven highly expressed genes.
Like the weighted scoring method, this unweighted scoring also showed higher predictive value than other prognostic markers based on AIC calculations in both the discovery and validation cohorts.
"When we compared [our] predictor, with respect to the other molecular markers, essentially with 99 percent probability the model that used our predictor is better than the other molecular markers, so we feel that we gain a lot of information over other markers," Marcucci said.
According to the study authors, the results suggest that the unweighted scoring method could be used to identify a small but clinically significant subset of patients more likely to respond to chemotherapy with complete remission, and longer overall survival.
For older patients especially, the authors wrote, such a tool could be valuable to spare those patients who are unlikely to respond to the toxicity of chemotherapy.
Marcucci said he and his colleagues are working on developing a rapid assay based on the results that can be used in a planned prospective clinical study to support future commercialization of the score as a laboratory-developed test.