Single cell multiomics reveal clonal and functional dynamics of MDS stem/progenitor cells during hypomethylating therapy

This study utilizes single-cell multiomics to reveal that azacitidine treatment in myelodysplastic neoplasms drives partial hematopoietic regeneration through the expansion of genetically stable stem/progenitor cells while suppressing malignant clones, yet eventual disease progression is fueled by the persistence of refractory sub-populations whose in vivo resistance is modulated by cell-extrinsic factors.

The Gist

The Big Picture: A Garden in Trouble

Imagine the human body's bone marrow as a garden where blood cells are grown. In a healthy person, this garden is full of diverse, strong plants (blood cells) that keep the body running.

In Myelodysplastic Syndromes (MDS), the garden is invaded by weeds. These aren't just ordinary weeds; they are "mutated" plants that have bad instructions in their DNA. They grow uncontrollably, crowd out the good plants, and produce weak, useless fruit (blood cells). This leads to anemia, infections, and bleeding.

The doctors treat this garden with a special fertilizer called Azacitidine (AZA). Sometimes, this fertilizer works wonders: the weeds die back, and the garden looks healthy again. But often, the weeds come back stronger later, turning into a full-blown forest fire (Leukemia).

The Question: Why does the fertilizer work for a while, and why does it eventually fail? What is happening inside the garden at the microscopic level?

The Study: A High-Definition Time-Lapse

The researchers didn't just look at the garden from a distance; they used a super-powered microscope (single-cell multiomics) to take a "time-lapse video" of the garden before, during, and after the fertilizer treatment. They looked at individual cells one by one to see who was growing, who was dying, and who was hiding.

Here are the four main discoveries they made:

1. The "Good" Plants Make a Comeback (The Resurrection)

When the fertilizer worked, the researchers saw something surprising. The garden didn't just get rid of the bad weeds; it actually regrew healthy plants.

• The Analogy: Imagine a forest fire. Usually, you think the fire just burns the bad trees. But here, the fire actually woke up a hidden stash of healthy seeds that were sleeping underground. These new plants looked and acted exactly like the healthy plants from a normal, non-diseased garden.
• The Catch: These healthy plants were "clean." They didn't have the bad DNA mutations or the broken chromosomes (karyotypic abnormalities) that the weeds had. The fertilizer didn't kill the bad weeds instantly; it just helped the few remaining healthy seeds take over the garden temporarily.

2. The "Bad" Weeds Have Many Faces (The Chameleons)

The researchers found that the bad weeds weren't all the same. They were divided into different "clans" or sub-clones.

• The Analogy: Think of the weeds as a criminal gang. Some members are loud and aggressive (expanding rapidly), while others are quiet and hiding.
• The Twist: Even when the patient looked healthy (clinical response), some of these criminal gangs were still there, just hiding in the shadows. They weren't dying; they were just waiting. When the patient eventually got sick again (progression), these specific gangs were the ones that had taken over the whole garden.

3. The "Bad" Weeds Are Actually Sensitive (The Trap)

This was the most confusing and interesting part. The researchers took the specific "bad" weed clones that were causing the cancer to come back and grew them in a petri dish in the lab.

• The Analogy: They put the "super-weeds" in a test tube with the fertilizer. Shockingly, the weeds died easily in the test tube! They were actually very sensitive to the medicine.
• The Mystery: So why didn't the medicine work in the patient's body? The researchers suspect the garden soil (the bone marrow environment) is protecting the weeds. In the patient's body, the weeds are hiding in a "safe house" or getting help from their neighbors (other cells in the bone marrow) that shields them from the fertilizer. In the test tube, they are naked and vulnerable.

4. The "TP53" Problem (The Unfixable Bug)

They found that if a patient had a specific mutation called TP53, the "good" plants rarely made a comeback.

• The Analogy: If the garden has a specific type of weed with a "super-shield" (TP53 mutation), the fertilizer can't wake up the healthy seeds. The garden stays dominated by the bad weeds, and the treatment usually fails quickly.

The Takeaway: What Does This Mean for Patients?

1. Hope: The body can regenerate healthy blood cells during treatment. The goal of therapy is to help these healthy cells grow faster than the bad ones.
2. The Enemy is Hidden: Just because a patient looks healthy doesn't mean the bad cancer cells are gone. They are often still there, waiting for the right moment to strike.
3. The Environment Matters: The reason the medicine stops working isn't always because the cancer cells become "immune" to the drug. It's often because the surrounding environment protects them.
4. Future Treatments: To cure MDS, doctors might need to do two things at once:
   • Give the fertilizer (AZA) to help the healthy cells grow.
   • Change the "soil" (the bone marrow environment) so it stops protecting the bad weeds, making them vulnerable to the medicine again.

In short: The study shows that MDS treatment is a battle between a regenerating army of healthy cells and a shapeshifting, hidden army of cancer cells. The cancer cells are actually weak on their own, but they are very good at hiding in the body's defenses. To win the war, we need to figure out how to drag them out of their hiding spots.

Technical Summary

1. Problem Statement

Myelodysplastic neoplasms (MDS) are heterogeneous malignancies driven by somatic mutations in hematopoietic stem cells (HSCs). High-risk patients are often treated with hypomethylating agents (HMAs) like azacitidine (AZA). While AZA can improve blood counts and delay progression to acute myeloid leukemia (AML), responses are rarely durable, and the mechanisms driving relapse remain poorly understood.
Key knowledge gaps include:

• Whether clinical response is driven by the eradication of mutated clones or the regeneration of healthy cells.
• The specific cellular and genetic dynamics of subclones during treatment and progression.
• The role of the bone marrow microenvironment versus cell-intrinsic factors in AZA resistance.
• How to distinguish between transient epigenetic states and distinct genetic clones that drive disease progression.

2. Methodology

The study employed a longitudinal, single-cell multiomics approach on bone marrow (BM) samples from high-risk MDS patients treated uniformly with AZA in a clinical trial (NCT03493646).

• Sample Collection: Longitudinal CD34+ hematopoietic stem and progenitor cells (HSPCs) were collected from 9 patients (at diagnosis, response, and progression) and 5 healthy donors.
• CITE-seq (Cellular Indexing of Transcriptomes and Epitopes by Sequencing):
  • Performed on CD34+ cells to simultaneously capture transcriptomic (gene expression) and proteomic (surface protein) data.
  • Used a custom panel of 273 antibodies (TotalSeqC) to define immunophenotypes.
  • Samples were pooled and processed to minimize batch effects.
• Genomic and Clonal Analysis:
  • InferCNV: Used chromosomally binned gene expression to infer Copy Number Variations (CNVs) at the single-cell level.
  • SNP Genotyping & Microarrays (SNP-CMA): Validated CNVs using bulk SNP genotyping and single-nucleotide polymorphism chromosome microarrays on flow-sorted cell populations.
  • TP53 Genotyping: Retrospectively genotyped TP53 mutations in single cells using Oxford Nanopore sequencing of 10X cDNA amplicons.
  • Mitochondrial Variants: Tracked mitochondrial SNVs to resolve clonal phylogenies and historical branch points.
• Functional Assays:
  • Ex Vivo Co-culture: Sorted specific transcriptomic clusters were cultured with murine MS-5 stromal cells and treated with varying concentrations of AZA to assess drug sensitivity and differentiation capacity (loss of CD34 expression).

3. Key Contributions & Results

A. Identification of "Shared" vs. "Patient-Restricted" Clusters

• Shared Clusters: Cells with transcriptomic profiles resembling healthy HSPCs (HSC/MPP, LMPP, GMP, MEP, etc.). These clusters emerged during clinical response and were depleted of CNVs and TP53 mutations.
• Patient-Restricted Clusters: Unique cell populations found only in specific patients, often located at the periphery of UMAP embeddings. These clusters possessed unique surface immunophenotypes and distinct genetic alterations.
• Dynamics: Clinical response correlated with an expansion of "shared" clusters and a reduction in "patient-restricted" clusters. Conversely, disease progression was marked by the dominance of expanding patient-restricted clones.

B. Genetic Architecture of Response and Relapse

• Regeneration Source: In patients with karyotypic abnormalities, the cells driving clinical response (the "shared" clusters) were derived from a residual, karyotypically normal, unmutated pool of HSPCs.
• TP53 and CNVs: TP53 mutations strongly correlated with abnormal CNV profiles. TP53-mutant clones were largely depleted during the response phase but re-emerged or expanded upon progression.
• Clonal Evolution: Phylogenetic reconstruction revealed complex evolutionary trees. Patient-restricted clusters often represented distinct genetic clones rather than just alternate epigenetic states. Minor clones present at diagnosis frequently expanded to become dominant at progression.

C. The Paradox of AZA Sensitivity

• Counterintuitive Finding: Patient-restricted clones that expanded in vivo during treatment and dominated at progression were often highly sensitive to AZA in ex vivo co-culture assays.
• Implication: This suggests that the lack of response in vivo is not necessarily due to intrinsic drug resistance of the clone, but rather is mediated by cell-extrinsic factors (e.g., the bone marrow microenvironment, stromal interactions, or soluble factors) that protect these clones from AZA-induced cytotoxicity or differentiation.

D. Pathway Enrichment

• Receding Clusters: Enriched for immune-related pathways (IFNγ response, TNFα signaling).
• Persisting/Expanding Clusters: Enriched for DNA repair and oxidative phosphorylation pathways, suggesting mechanisms of survival and metabolic adaptation.

4. Significance and Implications

• Mechanism of Action: The study clarifies that AZA response in MDS is primarily driven by the partial regeneration of hematopoiesis from residual, genetically normal stem cells, rather than the eradication of all mutated clones.
• Biomarker Development: The identification of unique surface immunophenotypes for patient-restricted clusters offers a potential pathway for Measurable Residual Disease (MRD) tracking using flow cytometry, which is more feasible for clinical diagnostics than single-cell sequencing.
• Therapeutic Insight: The disconnect between in vivo progression and in vitro sensitivity highlights the critical role of the bone marrow microenvironment in conferring resistance. This suggests that future therapies may need to target the niche or combine AZA with agents that disrupt protective microenvironmental signals.
• Precision Medicine: The findings support the concept that MDS is a collection of distinct genetic subclones with varying dependencies. Successful treatment strategies may require individualized approaches that target specific resistant clones or their microenvironmental support systems.

Conclusion

This paper provides a high-resolution map of MDS evolution under AZA treatment. It demonstrates that clinical improvement is a transient state of hematopoietic regeneration from a normal stem cell pool, while disease progression is driven by the selection and expansion of genetically distinct, patient-restricted clones that are protected by the in vivo microenvironment despite retaining intrinsic drug sensitivity.