Clonal evolution of acute myeloid leukemia relapsed after 19 years of remission

To the Editor

Although the standard intensive chemotherapy induces complete remission (CR) in 60–80% of patients with de novo acute myeloid leukemia (AML), significant numbers of patients experience relapse. While majority of AML relapses occur within 3 years of remission, late relapse, arbitrarily defined as relapse occurring after 3–5 years, has been described in 1–3% of patients with AML [1,2]. Compared to early relapse, late AML relapse is associated with better response to salvage therapy and more favorable outcome. However, the interpretation of late relapse can be confounded by new primary or therapy-related AML and its clinical distinction as a true relapse is often difficult [1,2]. Although, morphology and cytogenetic analysis can assist in differentiating true relapse from new primary or t-AML, deeper genetic characterization is needed for precise understanding of clonal origin in late relapse. To better understand clonal origin of late relapse, we performed whole exome sequencing (WES) on longitudinal samples obtained from a patient with AML who relapsed after 19 years of remission. Detailed case description can be found in the Supporting Information Appendix. Three samples were collected from the patient: (1) original AML bone marrow (BM), (2) relapse AML BM, and (3) 2nd CR BM. Methods of sample collection, sequencing, and bioinformatic analysis are described in the Supporting Information method. The mean coverage of WES was 127-fold. Forty-seven and 54 exonic variants were detected in the primary and relapse samples, respectively, with nonsynonymus/synonymus ratio of 1.7 and 1.5, respectively. Fifteen exonic variants were shared between the two samples. C/G > T/A transition was the most frequent alteration in both samples, while frequency of transversion was higher in the relapse sample (Supporting Information Fig. S3). In both samples, WES detected a canonical IDH1 p.R132L mutation that has been well characterized as a driver mutation in AML. A canonical SF3B1 p.K700E mutation was also detected in the relapse sample but not in the primary sample. We also identified a ZEB2 p.R524G variant in both samples as one of the potential putative driver mutations. ZEB2 has been implicated to have crucial role in hematopoietic stem cell differentiation, mobilization, and homing [3]. Conditional overexpression of ZEB2 in mice has been reported to induce T-cell leukemia and deleterious mutations of ZEB2 have been identified in other leukemias [4,5], further suggesting the possible association between altered ZEB2 function and leukemogenesis. These are consistent with ZEB2 being a likely driver mutation in this case. The PCR capillary electrophoresis (PCR-CE) assay detected an NPM1 p.W288fs and several different sizes of FLT3-ITD in both samples (Supporting Information table).

We next inferred a model of clonal evolution by tracing cancer cell fraction (CCF) of the detected variants (Fig. 1). Both primary and relapse AML shared the same founder clone with IDH1, ZEB2, and most likely the NPM1 mutation. Although the method of variant allelic fraction (VAF) calculation is different between WES and PCR-CE, VAF of NPM1 mutation on PCR-CE was stable around 0.5. Further, a previous study has shown that NPM1 mutation is almost always an early founding event in AML, consistent with the NPM1 mutation as early clonal event in this case. We did not incorporate FLT3-ITD into our model because association between respective ITD sizes and clonality has not been well understood. However, as a whole, FLT3-ITD was clearly detected at two time points, suggesting that this mutation persisted in the dominant clone.

Figure 1.

A, B: Two dimensional density plot for cancer cell fraction (CCF) of the detected variants in (A) primary vs. late relapse AML samples and (B) late relapse vs. 2nd remission samples. CCF was estimated by ABSOLUTE algorithm. Each dot represents the variant with corresponding CCF and increasing density indicates the region of high posterior probability of a cluster and thus implies respective clones. C: The inferred model of clonal evolution in a case of AML relapse after 19 years of remission. Y axis shows bone marrow (BM) blast percentage. X axis indicates time. The number in parentheses indicates clone from 2D density plots above. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Our model suggests that the founder clone persisted after initial therapy and relapsed 19 years later with additional mutations acquired. Overall, it is consistent with one of the models that were proposed by Ding et al., who performed whole genome sequencing on eight relapsed AML cases, all of which relapsed within 3 years of remission [6]. Our report differs in that our case had a larger fraction of relapse-specific mutations and fewer shared mutations between primary and relapse AML. This would be consistent with the much longer period before relapse and accumulation of additional mutations over this time period.

We also observed the emergence of a minor population with an SF3B1 mutation at relapse. CCF of the SF3B1 mutation did not follow that of the founder clone after salvage therapy (Fig. 1). SF3B1 mutation is frequently associated with MDS but rare in AML. The studied patient was suspected to have MDS 3 years before she experienced relapse (Supporting Information Appendix). Taken together, it is likely that the clone with SF3B1 mutation represents the co-occurrence of MDS in the context of a relapsing AML.

In summary, longitudinal genomic characterization of an individual with a late relapse of AML revealed that the founder clone of the primary AML persisted after treatment and constituted the basis of relapsed disease 19 years later, hence confirming “true” relapse. More cases of late relapse in AML need to be examined to better characterize the mechanisms of relapse and disease latency.