Using deep sequencing techniques, investigators of a study published in Nature determined they were able to tease out the predictors that are associated with the development of acute myeloid leukemia (AML).1

By looking at the genes that are commonly mutated in AML, and analyzing preleukemic hematopoietic stem and progenitor cells (HSPCs), the researchers could provide proof-of-concept for the genetic predictors of the transformation of pre-AML cells into malignant cells — and they could differentiate these aberrations from those that accumulated as a result of age-related clonal hematopoiesis (ARCH).

The investigators compared blood cells from 95 individuals collected 6.3 years prior to an AML diagnosis with 414 people in a control group that were considered healthy individuals. Through a comparison of gene panels, they found driver mutations (ARCH-PD) in 73.4% of the pre-AML cases at a median of 7.6 years before a positive AML diagnosis occurred.

Continue Reading

The median number of ARCH-PD mutations per participant was significantly higher in the pre-AML group compared with the control group (P < .0005) — and in that pre-AML group, 39% of the participants older than 50 years had a driver mutation with a variant frequency of more than 10% compared with 4% in controls.

Related Articles

The most commonly mutated genes across pre-AML cases and controls who had ARCH-PD mutations were DNMT3A and TET2, though these genes are thought to confer less of a risk of malignant transformation. Mutations to TP53 and U2AF1, however, were found to confer a high risk of subsequent AML.

Most importantly, the researchers determined that the “mutational landscape” of ARCH and pre-AML is very different.

“The most predictive mutations were in U2AF1 (2 codons), SRSF2 (1 codon), IDH1 (1 codon), and IDH2 (2 codons), lead researcher Liran Shlush MD, PhD, a visiting physician at the Princess Margaret Hospital, University Health Network in Toronto, Canada, told Cancer Therapy Advisor. “Other hot spots in other genes were also predictive if they were more recurrently reported in blood malignancies in the past,” Dr Shlush added. He said mutated IDH1/2 inhibits TET2, which is an epigenetic regulator, while U2AF1 and SRSF2 are part of the spliceosome machinery.

“The fact that the early events are associated with different players of the epigenetic machinery suggest that downstream epigenetic changes give selective advantage to HSPCs without dramatically influencing their differentiation capacity,” Dr Shlush said.

Also of interest was that recurrent CEBPA mutations, which are a hallmark of de novo AML in 10% of cases, was absent across ARCH and pre-AML groups, “suggesting that driver events in this gene may also be late events in AML evolution.”

In addition, the researchers found a significant association (P = .0016) between higher blood cell distribution width (RDW) and risk of progression to AML. (This proves interesting, as RDW is typically only measured in the evaluation of anemia.) Next, the team used machine learning to create an AML prediction tool, which they determined was able to forecast AML 6 to 12 months before diagnosis with an overall specificity of 98.2%.

The team concluded that using basic clinical data, such as RDW, could inform physicians who would be good candidates for subsequent genetic screening for AML.

And while the method described in the paper could be helpful for early diagnosis, Dr Shlush said the method would not be considered a liquid biopsy, as liquid biopsies measure DNA extracted from blood plasma, and they looked at DNA from blood cells. Dr Shlush said that while liquid biopsies are getting more accurate for diagnosing cancers in more advanced stages, they “are still not even close to what we propose, which is early detection 7 years before current diagnosis.”


  1. Abelson S, Collord G, Ng SWK et al. Prediction of acute myeloid leukemia risk in healthy individuals [published July 9, 2018]. Nature. doi: 10.1038/s41586-018-0317-6