Message from the void: MRD analysis from ctDNA

In this issue of Blood, Nakamura et al demonstrate that circulating tumor DNA (ctDNA) in the serum can be used for minimal residual disease (MRD) assessment in acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) patients after allogeneic hematopoietic cell transplantation and thus may relieve patients from frequent bone marrow punctures.¹ 

Compared with solid cancers, hematologists have long benefited from easily accessible tumor sampling from blood and bone marrow. This ability cemented the dogma that bone marrow sampling from the pelvis provides a representative sample to ascertain the remission status of acute leukemias and MDS. However, AML can be exclusively located at extramedullary sites at relapse or even at diagnosis, and acute lymphoblastic leukemia may relapse predominantly in lymphoid tissues. Nakamura and colleagues have now developed a method to monitor AML in patient serum using ctDNA, providing a platform to investigate important questions, including whether AML is homogeneously or distinctly distributed in the body.¹ 

Early measurement of treatment response is increasingly recognized as an important tool to predict final treatment outcome of AML patients. The European LeukemiaNet recommends refining response assessment through the measurement of residual disease (MRD).²  However, the National Comprehensive Cancer Network AML guidelines do not recommend routine MRD analysis in clinical practice.³  Flow cytometry and real-time polymerase chain reaction (PCR)–based approaches are currently recommended for MRD assessment,⁴  with next-generation sequencing (NGS) becoming more frequently used.^5,6  Nakamura and colleagues quantified MRD in AML and MDS patients after allogeneic hematopoietic stem cell transplantation (alloSCT) by droplet digital PCR (ddPCR) with a median detection limit of 0.04%. MRD positivity predicted relapse and correlated with shorter overall survival. Interestingly, MRD analysis from ctDNA identified MRD in some patients who were MRD negative in peripheral blood and/or bone marrow, indicating that ctDNA may be more representative of residual AML in some patients.

The authors screened 53 patients for gene mutations by NGS and identified at least one mutation in 51 patients. Allele-specific PCR assays were designed for each selected mutation and validated separately. MRD was monitored in ctDNA and bone marrow cells 1 and 3 months after alloSCT in AML and MDS patients who had undergone myeloablative conditioning. Variant allele frequencies (VAFs) were highly correlated between the diagnostic ctDNA and bone marrow. MRD positivity using conventional or ctDNA technologies predicted a higher relapse rate and shorter overall survival compared with MRD-negative patients both 1 and 3 months after alloSCT. The authors conclude that ctDNA monitoring provides a convenient approach to MRD monitoring with comparable sensitivity to using bone marrow and improved sensitivity to using peripheral blood cells, especially in cytopenic patients after alloSCT (see figure). This comprehensive study provides a new set of diagnostic assays for prognostication and fuels the hope that patients will require fewer bone marrow assessments after alloSCT in the future.

What is the source of the ctDNA? Is it actively secreted or shed from dying leukemic cells to the serum? This study and previous reports have observed that higher levels of cell-free DNA are found in cancer patients than in healthy controls and that circulating tumor cells do not correlate well with ctDNA levels, supporting the active secretion hypothesis⁷  and providing a possible explanation for why VAFs measured in ctDNA and bone marrow are comparable. It is surprising that 10 ng ctDNA was sufficient to detect low levels of MRD, as other DNA-based MRD assays recommend using 600 to 2000 ng DNA to achieve a minimum sensitivity of 1 leukemic cell in 100 000 normal cells. Preferential DNA secretion from tumor cells over normal cells might enrich the tumor DNA in serum and partially explain the good sensitivity from low DNA input. In contrast, there is a clear gradient of leukemic cells from bone marrow to peripheral blood, which reproducibly results in a 10-fold lower frequency of leukemic cells in peripheral blood compared with bone marrow in AML patients with low MRD. However, active DNA secretion from tumor cells remains speculative, as the mechanism by which this occurs and its regulation remains unknown.

The strong prognostic effect of MRD positivity as early as 1 month after myeloablative alloSCT demonstrated by Nakamura and colleagues is impressive. At this time point, many patients are still cytopenic, and other MRD techniques such as flow cytometry show less discriminatory potential after alloSCT compared with an assessment before alloHCT.⁸  It will be interesting to see the potential of this method in other hematologic diseases like lymphoma and myeloma, which are usually less represented in peripheral blood than acute leukemias.

A potential limitation of ddPCR is the requirement for a unique assay for each nucleotide change. It is therefore most suitable for recurrent mutations in genes like IDH1 and IDH2 and other genes with hotspot mutations. However, ctDNA can be also examined by error-corrected NGS, so the most convenient and reproducible approach will need to be established by individual laboratories.

In summary, the current study extends our tools to sensitively monitor MRD in AML and MDS and provides the technical and conceptual framework to understand asynchronous development of leukemic clones at different sites, mechanisms of DNA release from tumor cells, and potential biologic functions of circulating tumor DNA. While the serum of leukemia patients has largely been considered free of leukemia-specific information, this void is beginning to speak to us, and its next messages are eagerly expected.