ctDNA MRD under a menin inhibitor: blood flow diverges from blood DNA, because the disease is in the spleen and marrow
also called circulating tumour DNA MRD · ctDNA infant ALL · cell-free DNA MRD · ddPCR ctDNA KMT2A fusion · peripheral blood MRD
Instead of taking marrow from a baby's hip bone repeatedly, you can look for fragments of leukaemia DNA floating in a blood sample. In mice carrying human infant leukaemia and treated with a menin inhibitor, the floating DNA and the leukaemia cells in the blood told different stories - and the DNA was the one that matched where the disease actually was.
The same thing, for a scientist
The only study of ctDNA MRD in infant KMT2A-rearranged ALL is preclinical. Three infant MLL-rearranged patient-derived xenografts (MLL-1, MLL-2, MLL-7) were engrafted in NSG mice and monitored weekly with patient-specific KMT2A fusion ddPCR assays designed to EuroMRD standards, with positivity set at two or more droplets, corresponding to 10 or more mutant copies per millilitre of plasma [E1]. Before treatment, ctDNA correlated strongly with leukaemia burden. **After treatment with the menin inhibitor SNDX-50469 the two measures diverged**: peripheral-blood leukaemia cells by flow and ctDNA by ddPCR disagreed, and the authors attribute the divergence to persistence of leukaemia cells in the spleen and bone marrow, concluding that ctDNA reflects disease dynamics in the key infiltration sites and is a superior predictor of MRD compared with peripheral-blood assessment alone, especially at low disease burden [E1]. This is the one published observation anywhere of what a menin inhibitor does to the relationship between two MRD readouts, and it says they come apart. It is a mouse study with three xenografts, and the comparator is peripheral-blood flow rather than marrow MRD. It nonetheless sits directly on the question raised by `concept-mrd-flow-cytometry-and-the-lineage-switch-blind-spot`: **under this drug class, which assay is measuring the disease?**
What was done, precisely
Read in full text. NSG mice received 2-3 x 10^6 cells from one of three infant MLL-rearranged ALL PDXs by tail vein. Engraftment was followed weekly by percentage human CD45-positive cells in peripheral blood, total cell-free DNA by TapeStation, and ctDNA by ddPCR against a patient-specific KMT2A fusion amplicon. Each qPCR assay conformed to EuroMRD guidelines, with 10-fold serial dilutions of PDX genomic DNA for optimisation and a KMT2A-fusion-negative PDX plus no-template controls to set thresholds. Positivity required at least two positive droplets, equivalent to at least 10 mutant copies per millilitre of plasma; reactions with fewer than 10,000 droplets were excluded 1.
The design point worth taking is that the ctDNA target is the genomic breakpoint - the same patient-specific junction that concept-mrd-monitoring-genomic-breakpoint describes, read on plasma rather than marrow and quantified by digital rather than real-time PCR. This is not a new marker; it is the existing marker in a new compartment, which is why it can be adopted without new target discovery.
Why the divergence is the interesting result
Before drug, ctDNA tracked burden. After SNDX-50469, blood flow cytometry and plasma ctDNA disagreed, and the explanation the authors give is anatomical: leukaemia persisted in spleen and marrow while peripheral blood cleared 1.
That has a direct clinical reading and a speculative one.
Direct. Peripheral-blood cellular MRD is not a substitute for marrow MRD under a menin inhibitor, and ctDNA may be. This matters because repeat marrow aspiration in an infant is not a trivial procedure, and because blood sampling can be done far more often than marrow sampling - the frequency argument, not the sensitivity argument, is the real case for ctDNA.
Speculative, and flagged as such. If a menin inhibitor clears circulating blasts faster than it clears the marrow and splenic compartments, then any response assessment made on blood under this drug class will overstate response. The graph already records that menin inhibition produces differentiation in myeloid disease and rapid cell death in lymphoid disease; a compartment-dependent clearance rate is a third, separable effect and it is the one this study observed. No clinical trial of a menin inhibitor has reported paired blood and marrow MRD, so whether the divergence exists in patients is unknown.
Two hard limits on reading this further. It is three PDXs in mice, with no biological replication across models reported in the abstract; and mouse plasma volume, cfDNA clearance and splenic haematopoiesis all differ from human, so the quantitative relationship between compartment burden and plasma copies does not transfer.
Where this sits against the rest of the MRD toolkit
| marrow molecular MRD | blood flow MRD | plasma ctDNA | |
|---|---|---|---|
| Procedure | marrow aspirate | venepuncture | venepuncture |
| Frequency achievable in an infant | protocol timepoints only | frequent | frequent |
| Target | patient-specific junction or Ig/TCR | phenotype | patient-specific junction |
| Survives lineage switch | yes | no | yes |
| Reflects marrow and spleen | yes | not under menin inhibition 1 | claimed yes 1 |
| Evidence in infant KMT2A-r | prospective, 249 patients 2 | prospective, 117 patients 3 | three mouse xenografts 1 |
The last row is the whole story. ctDNA has the most attractive properties and by far the weakest evidence, and the honest position is that it is a proof-of-principle result that identifies a question rather than a method ready for a protocol.
The adjacent precedent worth noting is from a different disease: in high-risk neuroblastoma, an RT-qPCR marker panel in peripheral blood was prospectively validated across 634 samples from 312 patients and correlated with paired marrow, with lower expression in blood than marrow and clear heterogeneity after treatment 4. That is what a properly powered peripheral-blood MRD validation looks like, and it took an international prospective collaboration to do it.
| assay class | droplet digital PCR against the patient-specific KMT2A genomic fusion junction in plasma cell-free DNA |
|---|---|
| analyte | plasma cell-free DNA |
| setting | preclinical only |
| host | NSG mice, 2-3e6 cells by tail vein |
| drug | SNDX-50469 (preclinical menin inhibitor) |
| positivity threshold | >=2 positive droplets, equivalent to >=10 mutant copies/mL plasma |
| key finding | ctDNA and peripheral-blood flow diverge after menin inhibitor treatment; divergence attributed to persistent disease in spleen and bone marrow |
| what is missing clinically | no menin-inhibitor trial has reported paired peripheral blood and bone marrow MRD |
models
[ "MLL-1 PDX", "MLL-2 PDX", "MLL-7 PDX" ]
qc
[ "assays conform to EuroMRD guidelines", "10-fold serial dilutions of PDX gDNA for optimisation", "KMT2A-fusion-negative PDX and no-template controls", "reactions with <10,000 droplets excluded" ]
evidence level comparison
{
"marrow molecular MRD": "prospective, 249 infants",
"marrow flow MRD": "prospective, 117 infants",
"plasma ctDNA": "three mouse xenografts"
}- If ctDNA replaces marrow sampling on the strength of three mouse xenografts, the first thing it will do is change who is intensified, and the calibration between plasma copies and the Interfant thresholds that define intensification does not exist.
- A blood test that can be repeated weekly will be repeated weekly, and each extra measurement is an extra opportunity for a borderline positive to trigger an intervention. More frequent monitoring is only a benefit if there is something to do about an early signal, and in high-MRD infant KMT2A-rearranged ALL there largely is not.
- ctDNA is released by dying cells. A therapy that kills efficiently raises ctDNA transiently, so an early post-treatment rise can mean the drug is working. Any protocol acting on ctDNA kinetics has to model that, and the preclinical study did not address it.
- The compartment divergence cuts both ways for the menin-inhibitor story. If circulating disease clears while marrow and spleen persist, then blood-based response assessment flatters the drug; but the same result also shows the genomic-junction assay detecting disease that a cellular assay missed, which is an argument for the fusion-specific target rather than against the drug.
- Does plasma ctDNA track bone marrow MRD in children on a menin inhibitor? The mouse data say blood cells and plasma DNA diverge under this drug; nobody has measured paired marrow and plasma in a patient on one.
- If a menin inhibitor clears circulating blasts faster than marrow and splenic disease, do trial response assessments made on blood overstate response? AUGMENT-101 reports no-detectable-residual-disease in 68.2% of responders without naming the assay or the tissue.
- How many plasma ctDNA copies per millilitre correspond to a given marrow MRD level in an infant? Without that calibration, a ctDNA value cannot be mapped onto the protocol thresholds that drive treatment.
- Would ctDNA sampled weekly detect relapse earlier than protocol-scheduled marrow MRD? The frequency argument is the real clinical case for the method and has never been tested in this disease.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In NSG mice engrafted with three infant MLL-rearranged ALL patient-derived xenografts (MLL-1, MLL-2, MLL-7) and monitored weekly by percentage human CD45-positive cells, cell-free DNA and patient-specific KMT2A fusion droplet digital PCR conforming to EuroMRD guidelines with positivity set at two or more droplets (10 or more mutant copies per millilitre of plasma), ctDNA correlated strongly with leukaemia burden during initial engraftment, but after treatment with the menin inhibitor SNDX-50469 leukaemic burden assessed by peripheral-blood flow cytometry and by ctDNA was discrepant, a divergence the authors attribute to persistence of leukaemia cells in the spleen and bone marrow, leading them to conclude that ctDNA is a superior predictor of MRD compared with peripheral-blood assessment alone, especially at low disease burden.
Patient-Specific Circulating Tumor DNA for Monitoring Response to Menin Inhibitor Treatment in Preclinical Models of Infant Leukemia. · Doculara L, Evans K, Gooding JJ, et al. · Cancers (Basel) 2024What this evidence cannot show
Read in full text. A preclinical study: three PDX models in immunodeficient mice, with six mice used for the engraftment studies. Mouse plasma volume, cell-free DNA clearance kinetics and splenic haematopoiesis differ substantially from human, so the quantitative relationship between compartment burden and plasma copy number does not transfer. The comparator is peripheral-blood flow cytometry, not marrow MRD, so the study does not show ctDNA outperforming the actual clinical standard. SNDX-50469 is the research analogue used preclinically rather than the clinical agent. The threshold of two positive droplets is permissive and rests on a cited convention.
Evidence that argues against this (1)
- The authors themselves state that the prognostic value of ctDNA to detect MRD in paediatric childhood ALL remains largely unknown, and that they undertook the study because clinically relevant experimental models were lacking. The paper is framed by its own authors as proof-of-principle, not as evidence for clinical use. Doculara L et al. 2024
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 249 infants with KMT2A-rearranged ALL on Interfant-06, MRD measured in bone marrow by DNA-based PCR of the KMT2A genomic breakpoint and of immunoglobulin and T-cell receptor rearrangements at end of induction and end of consolidation predicted 6-year disease-free survival, and the KMT2A genomic breakpoint fusion sequence was the preferred target.
Clinical Implications of Minimal Residual Disease Detection in Infants With KMT2A-Rearranged Acute Lymphoblastic Leukemia Treated on the Interfant-06 Protocol. · Stutterheim J, van der Sluis IM, de Lorenzo P, et al. · J Clin Oncol 2021What this evidence cannot show
Read in full text. Bone marrow only; no peripheral-blood or plasma comparison was performed, so this study establishes the standard against which ctDNA would have to be validated and contributes nothing about blood compartments. MRD was non-mandatory and analysed in 249 of 439 eligible patients.
Evidence that argues against this (1)
- Because Interfant-06 never sampled blood in parallel, it cannot rule out that a plasma assay would have performed equally well; the absence of a blood arm is an absence of testing rather than a negative result about blood. Stutterheim J et al. 2021
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
On Children's Oncology Group AALL0631, MRD by multiparameter flow cytometry in bone marrow at three timepoints was a strong independent outcome predictor in 117 infants with KMT2A-rearranged ALL but not in 58 infants with KMT2A-germline ALL.
Minimal residual disease predicts outcomes in KMT2A-rearranged but not KMT2A-germline infant acute lymphoblastic leukemia: Report from Children's Oncology Group study AALL0631. · Faulk KE, Kairalla JA, Dreyer ZE, et al. · Pediatr Blood Cancer 2023What this evidence cannot show
Bone marrow flow, not peripheral blood; used here only to give the evidence level of the established comparators. The KMT2A-germline null is underpowered.
Evidence that argues against this (1)
- Flow MRD cannot follow a clone that changes lineage, which is a failure mode ctDNA against a genomic junction does not share, so evidence strength and assay properties point in opposite directions here. Gardner R et al. 2016
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In a prospective international validation in high-risk neuroblastoma, an RT-qPCR panel of PHOX2B, TH, DDC, CHRNA3 and DBH measured in 634 peripheral blood samples from 312 patients at diagnosis, after two cycles and at end of induction correlated with paired bone marrow at diagnosis (Spearman 0.70), showed heterogeneity after two cycles (0.37) and at end of induction (0.61), had significantly lower expression in blood than marrow, and peripheral blood infiltration at or above 1% at diagnosis was prognostic for event-free survival (adjusted hazard ratio 2.37) and overall survival (2.60).
Minimal Residual Disease Detection in Peripheral Blood of Patients With High-Risk Neuroblastoma Correlates With Outcome in the International GPOH-DCOG Prospective Validation Study. · Gelineau NU, van Zogchel LMJ, De Carolis B, et al. · JCO Precis Oncol 2025What this evidence cannot show
A different disease entirely, and an mRNA marker panel rather than a patient-specific DNA junction. It is cited here only as the methodological template for what a peripheral-blood MRD validation requires, not as evidence about leukaemia. Blood-marrow correlation degraded markedly after treatment, which is the same direction of failure the mouse ctDNA study reports.
Evidence that argues against this (1)
- The blood-marrow correlation fell from 0.70 at diagnosis to 0.37 after two cycles, which argues that peripheral compartments and marrow decouple under therapy in general - a reason to be cautious about any blood-based MRD claim, including the ctDNA one this entry is built around. Gelineau NU et al. 2025
- related to →MRD by genomic breakpoint: the only assay in this disease that is fusion-specific by construction
same patient-specific junction target, read in plasma by ddPCR
menin inhibition decouples circulating blasts from plasma tumour DNA in PDX
a negative in one compartment is not a negative in another
three infant MLL-rearranged PDX models