Molecular disease monitoring using circulating tumor DNA in myelodysplastic syndromes

Current diagnostic and prognostic algorithms in myelodysplastic syndromes (MDSs) rely heavily on peripheral blood (PB) counts and bone marrow (BM) morphology,^1-3  which often have interobserver variability.⁴  Recurrent chromosomal aberrations have helped refine current MDS prognostic models,^1,2,5  and more recently, the sequencing of MDS cancer genomes has also identified several recurrent mutations^6-12  that give prognostic insights at diagnosis.^7,8,11-15  However, there is little evidence of how the clonal and subclonal architecture is influenced by therapy.

Hypomethylating agents such as azacitidine and decitabine remain the mainstay of MDS treatment.^16,17  However, response may take months to achieve, and only half of patients respond to therapy.¹⁸  Repeated BM biopsies to monitor response can be invasive, resource-demanding, and associated with procedure-related complications. These limitations have, in part, compromised our ability to more regularly assess response to therapy and study clonal evolution in MDS. In this regard, we assessed the role of cell-free circulating tumor DNA (ctDNA) as a novel and minimally invasive biomarker to monitor therapeutic response and clonal evolution in MDSs.

Serial BM (n = 89) and plasma samples (n = 83) were collected from 12 patients with MDS who received azacitidine and eltrombopag as part of a phase 1 clinical trial (Table 1). Targeted deep sequencing (TS) was performed on DNA derived from BM and plasma using a customized panel of 55 genes known to be recurrently mutated in MDS and acute myeloid leukemia (see supplemental Table 1, available on the Blood Web site). Sequencing of BM samples identified putative driver mutations in 10 of 12 patients (Figure 1A). Digital polymerase chain reaction (dPCR) was performed to validate the TS results. Quantification of mutant allele fraction (MAF) showed excellent correlation (supplemental Figure 1A) between TS and dPCR.

Further methods are detailed in supplemental Methods.

The majority of cell-free DNA is derived from hematopoietic cells.¹⁹  As MDS is a clonal disorder characterized by ineffective hematopoiesis, we postulated that the malignant MDS genome would be well represented in circulating cell-free DNA. We first sought to understand how mutation detection in ctDNA compared with other hematopoietic compartments in a patient with MDS without significant cytopenia (Figure 1B). Here, the MAF of a TP53 and U2AF1 mutation from plasma ctDNA was comparable to DNA from PB neutrophils, whole blood, and BM aspirate mononuclear (MNL) cells. The lower MAF in the PB MNL cells most likely reflects the fact that many circulating lymphocytes may not be derived from the malignant MDS clones. These data are representative of our findings with several pathogenic mutations in patients with MDS (data not shown).

PB cytopenias are common in MDS; as such, molecular assessment of neutrophils and other PB cells is not always feasible. Therefore, the current “gold standard” for molecular testing in MDS is from BM aspirate DNA. Indeed, much of our understanding of the MDS genomic landscape has been through sequencing this compartment.⁶  In all cases (n = 10), the main driver mutations were detected in both BM and ctDNA (Table 1). Importantly, there was an excellent correlation (r² = 0.84; P < .0001) between the MAF of these mutations in BM and ctDNA across multiple matched time points (n = 75) (Figure 1C). This is the highest correlation reported between a tumor compartment and ctDNA. Importantly, this correlation was preserved even when the patients were leukopenic (supplemental Figure 1C-D), and there was no correlation between ctDNA MAF and PB white cell count (supplemental Figure 1E). Together, these data confirm the fact that ctDNA accurately reflects the genomic architecture of BM blasts in MDS regardless of the PB white cell count. Not infrequently, BM biopsies provide poor-quality specimens due to technical difficulties such as BM fibrosis or hypoplasia. In situations where a suboptimal, blood-dilute BM aspirate was obtained, ctDNA analysis provided molecular information that was equal, and in some cases superior, to BM sampling (Figure 1D; supplemental Figure 2).

Overall assessment of therapeutic response in MDS is measured using standard criteria.²⁰  However, these criteria fail to appreciate the tumor heterogeneity in MDS and do not capture the clonal dynamics and evolutionary changes observed with therapeutic pressure. Therefore, to better understand this, serial TS and/or dPCR analysis was performed on BM and ctDNA in the 10 patients with detected mutations (Figure 1; supplemental Figure 2). In each case, ctDNA dynamics closely followed that of BM DNA.

Studies have shown that TET2 mutations predict response to azacitidine,^21,22  although benefit appears confined to patients with a MAF >10%. Furthermore, not all clones that harbor TET2 mutations show sensitivity to azacitidine.²¹  In case AZA004, there were 2 TET2 mutations present at baseline with a MAF >10%. Consistent with the findings of Bejar et al,²¹  a response to therapy was achieved, which was paralleled by a reduction in the MAF of both TET2 mutations in plasma ctDNA (Figure 1D). Importantly, we also found cases with distinct TET2 mutations where the TET2 mutant clones were not suppressed by azacitidine therapy (supplemental Figure 2E).

Despite initial response to azacitidine-based therapies, progression invariably occurs. In case AZA009, despite initial clinical stability during treatment, ctDNA demonstrated an expanding malignant subclone containing the NRAS mutation, which ultimately resulted in the patient’s progression to acute myeloid leukemia (Figure 1E). Importantly, in several patients who progressed after an initial response to therapy, ctDNA reflected dynamic changes in tumor burden (Figure 1F,H; supplemental Figure 2B-F).

Together, these findings highlight the ability of ctDNA analysis to mirror the genomic changes observed in the BM and track multiple driver mutations throughout therapy. Importantly, ctDNA shows the differential response of the malignant subclones during therapy and can be used to identify and preempt disease progression.

Although used in MDS prognostic models,²  the difficulty in using karyotypic abnormalities as a monitoring tool is that they are present in <50% of cases.^17,23  This may be further compounded by lack of sensitivity of metaphase cytogenetics. To address this, we used LC-WGS to monitor chromosomal aberrations in plasma from 3 MDS patients throughout therapy (Figure 1; supplemental Figure 3).

Case AZA011 highlights a patient who had a del9q alteration detected in the BM by conventional cytogenetics (28/30 metaphase cells) (Table 1). Prior to treatment, this del9q was clearly identifiable in plasma ctDNA (Figure 1G). Notably, following clinical response to therapy, the copy-number alteration was markedly reduced in plasma, suggesting that azacitidine was able to suppress the malignant clone and allow hematopoiesis to be restored from other hematopoietic stem and progenitor cells.²⁴  The patient achieved a prolonged response (>4 years) but eventually progressed in the BM with an increasing myeloblast count (day 1492). Interestingly, progression was associated with a re-emergence of the del9q clone, which ctDNA detected almost 3 months before confirmation by BM cytogenetic analysis (Figure 1G). Importantly, while no cytogenetic evolution was noted at progression, TS clearly showed clonal evolution with the emergence of a new ASXL1 mutation in ctDNA at this time (Figure 1H). ASXL1 mutations confer a poor outcome in the myeloid malignancies, potentially negating the positive influence of TET2 mutations in mediating response to azacitidine therapy.^21,22  It is currently unclear if azacitidine primarily acts by suppressing or differentiating the dominant malignant clone.^11,25,26  Although the data in this case would suggest the former, it remains entirely possible that in other cases, azacitidine results in a restoration of blood counts by differentiating the malignant clone.

There is growing evidence supporting the importance of molecular assessment in MDS. Here, we show that ctDNA mirrors the genomic information from BM, accurately reflects the dynamic clonal changes seen in response to therapy, and can predict treatment failure. Together, these data support the use of ctDNA analysis as a noninvasive biomarker to compliment existing monitoring strategies for MDS.

The authors thank Gisela Mir Arnau, Timothy Semple, Sreeja Gadipally, and Timothy Holloway for assistance with LC-WGS; Michelle McBean for assistance with sample retrieval; and Meaghan Wall for providing cytogenetic information on patients.

This work was funded by a Senior Leukaemia Foundation Australia Fellowship and VESKI Innovation Fellowship (M.A.D.), a National Breast Cancer Foundation and Victorian Cancer Agency Fellowship (S.-J.D.). National Health and Medical Research Council of Australia grants 1128984, 1106444, and 1106447 (M.A.D.) and 1107126 and 1104549 (S.-J.D.) partly funded this research. This work was also supported by the Klempfner Epigenetics Fellowship administered by the Snowdome Foundation and the Haematology Society of Australia and New Zealand young investigator grant (P.Y.). The clinical study patient recruitment was funded by the Victorian Cancer Agency, Celgene, and GlaxoSmithKline.

Contribution: P.Y., S.-J.D., and M.A.D. designed the project, interpreted data, and wrote the manuscript; M.D. recruited study participants; M.D., P.Y., D.W., and P.B. provided patient samples and clinicopathological data; S.-J.D., M.A.D., P.Y., and R.A. developed the targeted gene panel with helpful input from P.B. and D.W.; P.Y., S.F., D.S., R.V., S.Q.W., and C.Y.F. performed the experimental work; P.Y. and T.H. analyzed data with input from K.D.; and all authors approved the final version of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Sarah-Jane Dawson, Peter MacCallum Cancer Centre, 305 Grattan St, Melbourne, VIC 3000, Australia; e-mail: sarah-jane.dawson@petermac.org; and Mark A. Dawson, Peter MacCallum Cancer Centre, 305 Grattan St, Melbourne, VIC 3000, Australia; e-mail: mark.dawson@petermac.org.