Skip to main content
Premium Trial:

Request an Annual Quote

Researchers Uncover Molecular Signature to Predict Therapy Response in Leukemia Patient Subset


NEW YORK (GenomeWeb) – An international team led by researchers at the University of Michigan has identified a molecular signature that predicts how patients with chronic myelomonocytic leukemia (CMML) respond to at least one of the disease's standard treatments.

In a study published this week in the Journal of Clinical Investigation, the group described its discovery of a pattern of DNA methylation that distinguished those who responded to the drug decitabine from those who didn't in a small cohort of 40 CMML patients.

The group also validated the signature in an independent set of patient samples, revealing that it was 87 percent accurate in predicting response to the drug.

Decitabine, as well as its cousin azacytidine, has become a treatment of choice for CMML, a type of leukemia precursor that clinicians currently define as bridging two different subcategories: the myelodysplastic syndromes and the myeloproliferative neoplasms.

Unfortunately DNA methyltransferase inhibitors like decitabine are only effective in about half of CMML cases, and are slow to act, meaning it can take up to six months of treatment to know definitively whether the drugs are working.

Maria Figueroa, the study's senior author and a member of the University of Michigan pathology department, told GenomeWeb that the group's study was driven by the need for a better tool to estimate responsiveness upfront and avoid months of unnecessary treatment for unresponsive patients.

"The kinetics are very slow," she explained, "So if you haven’t responded in four, or six, up to 12 months, it doesn't mean you won't respond," she said. "But tolerating a drug doesn’t mean it's not causing some side effects, not to mention the costs associated, and the fact that you are denying that patient the chance to look for alternative options during that time that you are waiting those 6 to 12 months."

Figueroa and her colleagues, from both the University of Michigan Comprehensive Cancer Center and the University of Florence in Italy, initially used genome-wide next-generation sequencing and bisulfite sequencing to look for mutations or methylation signatures that aligned with drug response in a set of 40 CMML patients, half of whom were defined as responsive and half of whom were resistant to decitabine.

While the researchers did not find any somatic mutations that differentiated the responsive and unresponsive samples, they did identify a set of more than 150 differentially methylated regions (DMRs), or areas of the genome where DNA methylation differed between the more- and less-responsive CMML groups.

The team then narrowed down to the 21 most predictive of these DMRs to use as a risk classifier. With additional collaborators at the Gustave Roussy Cancer Center in France, the group then tested the signature in another 28 CMML patient samples.

The panel was up to 87 percent accurate in differentiating responsive and resistant patents in this new group, the authors reported.

According to Figueroa, the hope is that this signature can be further developed and refined into a clinical diagnostic to assess patients' likelihood of responding to DNA methyltransferase inhibitors before treatment.

"[We are hoping for] a future in which we receive these patients, classify them upfront, and if they have [poor] chances of response we could combine the standard agents with another agent to overcome resistance," she said.

Alternatives or adjuncts to decitabine or azacytidine for CMML patients are currently limited, but do exist. "One option is bone marrow transplant, at least for patients who have a donor and have overall good condition," Figueroa said.

"If not that, it's up to putting them on experimental clinical trials, and that was really another motivator to the study," she added. "We wanted not only to distinguish responders and non-responders, but also to understand why non responders don't respond — if there is something driving resistance and whether we could eventually target it."

After identifying and validating their predictive methylation signature, the researchers went on to use RNA-sequencing on samples from 14 of the original cohort — eight responders and six non-responders — to look for genes with more than a two-fold difference of expression between the responsive and non-responsive groups.

According to the authors, they found that the resistant samples were characterized by significant relative overexpression of two chemokines, CXCL4 and CXCL7, compared to the responsive samples.

The researchers focused in on these in a further set of in vitro experiments, and saw that treatment of normal CMML cells with CXCL4 and CXCL7 appeared to block the effect of decitabine, indicating that overexpression of the two genes may indeed lead to primary resistance to the drug, and that targeting the appropriate downstream pathways could potentially overcome this resistance.

Figueroa said the investigators hope to follow up in three main ways. First, they want to understand how disease- and drug-specific their methylation signature is — whether it is only predictive in the context of decitabine in CMML patients, or if it might be a broader predictor of response to other similar drugs and across other subtypes of leukemia and myelodysplasias.

Secondly, the team plans to follow up with more studies of the pathways associated with CXCL4 and CXCL7, hoping to find a way of either pre-sensitizing patients to more effective decitabine treatment, or overcoming resistance by treating with a combination of the drug with another agent.

"We think there are some approved drugs for these pathways," Figueroa said. "But we have to test this out."

Finally, the group is working on adapting its findings into a practical format for development as a clinical diagnostic test. "We can't be doing next-gen sequencing on all the patients when we are only interested essentially in 21 markers. It's not practical," Figueroa said.

These efforts to standardize their signature will involve testing different platforms such as methylation-sensitive multiplex PCR, Sequenom's MassArray, and enzyme digestion methods, among others, she added.