Minimal residual disease (MRD) is a powerful predictor of relapse in acute myeloid leukemia (AML), yet its accuracy remains a challenge. Some patients who are MRD-positive do not relapse, while others who are MRD-negative still experience disease recurrence. At the 2024 American Society of Hematology (ASH) Annual Meeting, Professor Jerald Radich from Fred Hutchinson Cancer Center provided an in-depth analysis of how mutations, clonal architecture, and disease evolution influence treatment response, resistance, and relapse.To clarify the complex relationship between MRD and clonal dynamics, Hematology Frontier has compiled key insights from this discussion.

MRD as a Measure of Treatment Response

AML, like many cancers, can follow several distinct treatment trajectories: resistance to therapy (refractory disease, red curve), remission (green curve), or initial response followed by relapse (blue curve).

Refractory disease and complete remission often represent more homogenous tumor populations, with monoclonal or low-clonality tumors that are either highly resistant or highly sensitive to treatment, respectively.

Relapse, however, is frequently driven by tumor heterogeneity and polyclonality. In these cases, chemotherapy exerts selective pressure, eliminating sensitive cells and reducing tumor burden but failing to eradicate all malignant clones. Over time, even with continued therapy, resistant clones gain a competitive advantage.

The recurrent tumor may originate from:

• Genetically identical descendants of resistant clones present at diagnosis, which survive frontline treatment and drive relapse.
• Newly acquired subclones, which develop additional mutations under chemotherapy-induced selective pressure, leading to broader drug resistance.

This evolving clonal landscape complicates MRD interpretation, as standard MRD assessments may fail to detect emerging resistant subclones, ultimately limiting its predictive accuracy.

Why Does Clonal Origin Matter? How Do We Study It?

Clonal origin refers to the acquisition, evolution, and selection of genetic clones. Understanding the clonal origins of AML in the context of treatment response and resistance is crucial for several reasons:

1. Clonal Evolution and Heterogeneity

Cancer evolves through Darwinian selection, adapting to therapeutic pressure in ways that allow drug-resistant clones to expand. Investigating clonal origins helps identify competitive and cooperative interactions among clones, allowing us to exploit evolutionary principles to counteract aggressive and treatment-resistant clones.

2. Tracking Residual Disease

By characterizing and tracking leukemic clones from diagnosis through treatment, researchers can predict and potentially prevent relapse by identifying early signs of emerging resistant clones.

3. Targeted Therapy

With highly sensitive molecular techniques, researchers can detect persistent rare clones, which may present opportunities for targeted intervention before full-blown relapse occurs.

4. Predicting Treatment Outcomes

Refining clonal detection and characterization may enable early indicators of treatment efficacy, shortening clinical trial timelines and accelerating the development of more effective therapies. Chronic myeloid leukemia (CML) serves as a prime example of how molecular monitoring has transformed treatment strategies.

Techniques for Studying Clonal Origins and MRD

A variety of advanced technologies are used to study clonal evolution and MRD, each with distinct advantages and limitations:

1. Next-Generation Sequencing (NGS)

NGS-based approaches, including gene panels, whole-exome sequencing, and whole-genome sequencing, enable tracking of genetic alterations. However, in the context of MRD, error-correction methods must be employed to achieve the necessary sensitivity for detecting low-level mutations.

2. Single-Cell Genomics

This method allows for deconvolution of complex leukemic populations, offering deeper insights into clonal composition and dynamics. Single-cell genomics is also valuable for detecting lineage-specific mutations and gene expression patterns, although its cost and throughput limitations remain significant challenges.

3. Targeted Molecular Monitoring

Quantitative PCR (qPCR) and digital PCR (dPCR) are effective for tracking specific genetic alterations with high sensitivity and cost-efficiency. However, these methods are limited in the number of targets they can simultaneously monitor, making them less ideal for highly heterogeneous leukemias.

Which Mutations Are Clinically Significant?

2022 marked a major shift in AML classification, with updates from the European LeukemiaNet (ELN), International Consensus Classification (ICC), and the World Health Organization (WHO) reflecting advancements in our understanding of mutations and prognosis.

A key revision was the explicit recognition of TP53-mutated AML and the classification of MDS/AML with myelodysplasia-related mutations, which include ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, and ZRSR2. These mutations are common in MDS-related AML, accounting for nearly a quarter of all AML cases and almost half of newly diagnosed AML cases in adults over 60.

Applying This Knowledge to Clinical Practice

When interpreting clonal evolution and mutation data, several complexities must be considered:

1. Germline Mutations: Some mutations linked to myeloid malignancies (e.g., CEBPA, GATA2, JAK2, and NPM1) may be inherited, requiring differentiation from acquired somatic mutations.
2. Clonal Hematopoiesis of Indeterminate Potential (CHIP): CHIP-associated mutations, such as DNMT3A, TET2, and ASXL1 (DTA mutations), are common in aging populations and may confound MRD assessments. Some models suggest that excluding DTA mutations improves the predictive accuracy of MRD assessments.
3. Splicing Factor Mutations & IDH2: Recent findings suggest that excluding splicing factor mutations and IDH2 status from predictive models may enhance prognostic performance. Notably, in patients receiving IDH2 inhibitors, determining IDH2 mutational status is essential for treatment decisions.

Understanding these mutational landscapes and clonal origins is crucial for refining AML risk stratification, optimizing MRD monitoring, and advancing precision medicine strategies.

Impact of Treatment on MRD

The effect of treatment on MRD is particularly evident in allogeneic transplantation, where stronger conditioning regimens directly target AML cell populations while also benefiting from a robust graft-versus-leukemia (GVL) effect. Studies have shown that patients with persistent MRD before and after transplantation have significantly worse outcomes compared to MRD-negative patients.

Flow cytometry-based MRD monitoring has stratified post-transplant patients into four risk groups. The highest-risk group, where MRD is positive both before and after transplantation, shows relapse rates exceeding 80%, while the lowest-risk group, with MRD-negative status before and after transplant, has a relapse risk of only 20%. However, when assessing individual patient prognosis, MRD-inclusive models demonstrate only moderate predictive power, with a C-statistic of 0.7, meaning their accuracy is limited and only moderately better than random chance.

Gene sequencing-based MRD detection may offer better prognostic accuracy. Several studies indicate that MRD identified through targeted gene panels strongly correlates with relapse risk. Patients who undergo reduced-intensity conditioning transplants while being MRD-positive face a higher risk of relapse than those who receive myeloablative conditioning. Notably, the presence of mutations in DTA genes such as DNMT3A, TET2, and ASXL1 at the time of transplant does not appear to be linked to increased relapse risk. A significant finding is that the allelic burden of FLT3-ITD mutations demonstrates a clear dose-response relationship with relapse risk, an effect that has been difficult to establish using flow cytometry-based MRD. A recent study on post-transplant relapse, focusing only on patients who relapsed and excluding non-relapsing individuals, found that the dynamics of relapse depend on the functional category of mutations and their stability during molecular progression.

One promising future strategy could involve the development of cost-effective, rapid, and targeted approaches for post-transplant mutation detection, focusing only on mutations present at the time of transplant. When combined with flow cytometry-based MRD assessment, such an approach could enhance relapse prediction. In post-transplant follow-up, DTA and non-DTA mutations exhibit similar relapse dynamics, emphasizing the need for refined molecular monitoring strategies.

MRD, Mutations, and Treatment Response

AML exists within a complex cancer ecosystem, where different leukemic populations compete and cooperate under selective pressures exerted by treatment and immune responses, while the microenvironment provides both shelter and resources for tumor survival. The relative size of different clones and their response to selective treatment pressures shape the initial MRD response and the composition of residual disease, ultimately determining whether a patient experiences relapse or achieves long-term remission. The intricate nature of clonal interactions means that competition and cooperation among clones contribute to treatment outcomes in ways that are not yet fully understood.

Clinically, four general patterns of treatment response can be observed: no response, early relapse, late relapse, and long-term remission. However, these clinical categories do not fully capture the underlying tumor dynamics. Patients who respond well to treatment or exhibit primary resistance likely have low genetic diversity, with clones being either uniformly sensitive or uniformly resistant. Patients who relapse early may represent cases of Darwinian selection, where pre-existing resistant clones dominate after treatment. Late relapse cases may be driven by secondary clonal evolution, where new clones emerge under treatment pressure, leading to tumor regrowth over time.

Distinguishing between primary resistance, early relapse due to selection pressure, and late relapse caused by clonal evolution requires additional tools that can track clonal diversity and identity in real time. More advanced monitoring methods will be necessary to determine whether patients are relapsing due to the expansion of pre-existing resistant clones or the emergence of new mutations under treatment pressure.

Conclusion

MRD remains an invaluable predictor of relapse, but it is not without its limitations. False-negative results can lead to undetected residual disease, meaning that MRD-negative patients who relapse may require more sensitive detection methods. Conversely, false-positive results may indicate the presence of persistent pre-malignant clones that do not necessarily lead to relapse, highlighting the need for a better understanding of clonal dynamics.

Several unresolved questions remain, including the significance of specific mutations, clonal size, and the impact of different mutation combinations on treatment response. The sequence in which mutations are acquired may also play a crucial role in determining therapy resistance and the likelihood of relapse. Despite these challenges, MRD continues to be one of the most powerful tools in cancer treatment monitoring, and ongoing advancements in sequencing technologies and clonal evolution analysis will further refine its role in precision medicine.