Skip to main content
Premium Trial:

Request an Annual Quote

DNA Sequencing Illuminates 11 Genetic Subtypes of AML


NEW YORK (GenomeWeb) – Patients diagnosed with acute myeloid leukemia may be better served if they are first subtyped into one of 11 different genomic categories, according to researchers from the Wellcome Trust Sanger Institute and elsewhere.

An international group of scientists led by the Sanger Institute have sequenced cancer genes from over 1,500 patients with acute myeloid leukemia, and defined 11 different genomic subgroups that had distinct clinical outcomes.

While the work, which was published today in the New England Journal of Medicine, needs to be confirmed, it could influence how AML patients are treated and monitored throughout the course of their disease. In addition, if validated, the results could be incorporated into the World Health Organization's system for classifying AML.

The researchers enrolled 1,540 patients who had taken part in three prospective clinical trials as part of the German-Austrian AML Study Group and sequenced 111 cancer genes from patient samples collected prior to their treatment, using that data as well as cytogenetic and clinical data to define AML genomic subgroups.

The sequencing identified 5,234 driver mutations in 76 genes. Around 86 percent of patients had two or more driver genes, while in 62 individuals, or 4 percent, there were no driver mutations.

Next, they applied statistical information to look for patterns of mutations and gene-gene interactions, which subdivided the patients into 11 different classes. All but 166 individuals, about 11 percent, could be assigned to one of the genomic subtypes, and 56 patients, or 4 percent, met the criteria for two or more subgroups.

Currently, the World Health Organization classifies AML into eight different genetic subtypes defined by large-scale genomic events that include chromosomal translocations and inversions as well as gene fusions.  The researchers identified those same eight classifications, but also found three addition subgroups. One group, comprising 18 percent of the patients, had mutations to genes encoding chromatin, RNA splicing regulators, or both. Thirteen percent of patients had TP53 mutations, chromosomal aneuploidies, or both. And 1 percent of patients fell into a third group defined by IDH2 mutations.

"We recaptured what was already known," senior author Peter Campbell told GenomeWeb. "But we also fleshed out those groups and added a few more," he said. In addition, because AML has been genomically characterized so well by many different groups, "we think that this picture, this subtyping, will be pretty stable" going forward. Although more genes will be discovered, he said, those will likely be rare genes, while the bigger picture will remain the same.

Another implication of the study is that it showed that certain genomic features and patterns do predict clinical outcome. After the group identified the 11 genomic subtypes from 1,500 patient samples and saw that the different subtypes did correlate with clinical outcome, they applied the same classification system to samples sequenced as part of the Cancer Genome Atlas and found that the subtyping scheme as well as clinical predictions largely held true.

"A large component of the variation in survival can be explained by the underlying genomic basis of the disease," which can lead to better "predictions about which patients will be cured and which will die," Campbell said. "It's by no means perfect, but much better than what's in the clinic at the moment."

Of the eight previously identified genomic subtypes of AML, there were one or two that had been controversial, Campbell said, with researchers debating the merit of those subtypes and whether they were in fact subtypes. So, it was both somewhat surprising as well as reassuring that the Sanger team confirmed those eight types.

In addition, currently, about half of AML patients do not fall into any of the eight standard subtypes. For those patients, physicians have a hard time providing accurate predictions about the course of their disease, which can impact how aggressively they are treated.

The discovery of three additional subtypes went a long way toward reducing the number of patients that could not be classified. About one-third of the individuals in the NEJM study fell into one of the three new groups and would not have been subtyped under the current WHO guidelines. Those patients, without being subtyped are "quite diverse," first author Elli Papaemmanuil told GenomeWeb. "It would be really hard clinically to predict whether those patients would have a good outcome," she said.

But the new subtypes go a long way toward reducing that group of patients that cannot be typed. In fact, two of the three newly identified subgroups comprised the second and third most populous subgroups, behind only the subgroup of patients defined by having NPM1 mutations. That group consisted of 418 patients, or 27 percent, of the whole group.

Despite the researchers' ability to whittle down 1,540 patients into 11 different subtypes with distinct clinical outcomes, they noted that there was still a large amount of variability within each subtype.

"Even within a particular disease category that has more homogenous biology, even then the variable genetics within that subgroup has an impact on outcomes," Campbell said.

Researchers have known for quite some time based on other genomic studies that there is a so-called long tail of cancer genes that are mutated in just a few percent of patients, but this study adds to the idea that those genes are impacting outcomes of the specific patients. Interestingly, Papaemmanuil noted that when the overall group is considered, those genes may make up 5 percent, but looking at the subgroups, it is often the case that patients with one of those rarely mutated genes will cluster in one group, comprising upwards of 50 percent of that group, she said. Such a finding adds to the idea that it's time to "stop considering AML as one group and consider the distinct subtypes," she said.

Campbell and Papaemmanuil said that their next steps are to validate their findings in larger groups of patients as well as to apply the same methods to patients with other types of blood cancers.

For instance, Campbell said, patients with myeloid neoplasms are usually diagnosed as either having an acute myeloid disease, a myelodysplastic syndrome, or having a myeloproliferative disorder. While this study looked at the most aggressive form, acute myeloid leukemia, there are patients who fall into the other two categories who are also high risk and may have similar mutations. Campbell said he would like to study these other categories to see if in fact individuals actually fall into one of the 11 subtypes identified in this study.

Individuals who are initially diagnosed with having a myeloproliferative disorder, myelodysplastic syndrome, or AML, "go down completely different treatment paths," Campbell said, "but it can be challenging to diagnose based on morphological grounds," and it could be that genomic classifications could ensure that patients go down the most appropriate treatment path.

In addition, Papaemmanuil said that following validation studies, researchers will work with the WHO to discuss including the new genomic classifications in the updated guidelines. The WHO revised its classifications in 2015, she said, and the next revision is anticipated to occur in 2018. She added that one of the collaborators on the study is part of the panel that makes recommendations about myeloid classifications.

"Our data is being discussed and evaluated for the next revision," she said. "At the same time, we're validating our results and expanding the studies."