Does the diagnostic marker survive the drug? Surface NG2, CD133, CD10 and CD33 under menin and DOT1L inhibition, with the rearrangement held fixed
also called NG2 pharmacodynamic experiment · marker decoupling experiment · CSPG4 under revumenib
This experiment treats KMT2A-rearranged leukaemia cells with each of two drugs that interrupt the broken gene's activity in different ways, and measures whether the surface markers doctors use to recognise the disease are still there afterwards. The point is to find out whether the diagnostic test survives the treatment.
The same thing, for a scientist
**Design.** KMT2A-rearranged human cell lines (at least one KMT2A::AFF1 B-lineage and one KMT2A::MLLT3 myeloid) plus at least two primary infant KMT2A::AFF1 patient-derived xenograft samples, treated in a 2x2 with **revumenib** (menin-KMT2A) and **pinometostat** (DOT1L) at doses producing the published transcriptional response, plus vehicle and a combination arm. Readout at 0, 48, 96 and 168 hours. **Primary readout:** surface **NG2/CSPG4** by flow, reported as percent-positive against the lineage- and age-appropriate threshold (7% infant ALL, 12% child ALL, 14-28% AML) **and** as MFI and as molecules per cell by calibrated beads - because this graph's own surrogate-marker entry shows that percent, MFI and site density are three different quantities that get conflated. **Co-primary readout:** surface **CD133/PROM1**, which is the discriminating arm. CD133 is driven by an intragenic **H3K79me2/3 enhancer element** with direct MLL-AF4 binding and is PRC2-repressed when off [E2]. That gives a two-drug truth table NG2 alone cannot supply: | CD133 falls under | interpretation | |---|---| | both drugs | the marker reads the **programme** | | pinometostat only | the marker reads the **H3K79 mark** | | neither, at a dose with a transcriptional response | the enhancer mechanism does not control surface protein | **Secondary markers:** CD10/MME (predicted to *rise* if its promoter methylation is programme-maintained), CD33, CD117, CD19, CD34, CD38, CD371 and monocytic markers CD14/CD64 - the last group because a differentiation-driven monocytic shift is the documented effect of menin inhibition [E3] and must be distinguished from marker-specific loss. **Controls that make it interpretable.** (1) **FISH for KMT2A rearrangement at every timepoint** - the entire claim is that the marker moves while the lesion does not. (2) A **KMT2A-germline CSPG4-positive** line, to show any NG2 fall is programme-dependent rather than a nonspecific drug effect [E4]. (3) Parallel RT-qPCR for *CSPG4* and *PROM1* transcript, to separate transcriptional loss from surface-protein turnover. (4) **Viability and apoptosis gating**, because a marker can appear to fall simply because the brightest cells died. **And the free preliminary step.** The revumenib immunophenotype study [E3] is open access as **PMC13233299**. Fetching its marker tables to see whether NG2 or CSPG4 appears there costs one request and could partly answer the question with no experiment at all.
What each outcome would mean
- NG2 falls under revumenib while FISH stays positive -
hyp-ng2-tracks-the-transcriptional-programme-not-the-rearrangementis supported. Immediate consequence: NG2 must be removed from, or re-calibrated in, any MRD panel used during menin-inhibitor therapy, and this graph's MRD beat's demand for paired lineage-independent molecular MRD gets a second concrete justification. - NG2 unchanged at a dose with a clear transcriptional and differentiation response - the hypothesis is falsified, and the informative reading is that CSPG4 is a menin-independent KMT2A-fusion binding site. That bears directly on the menin-interface hypothesis the whole graph rests on, because it would be a documented fusion target gene that menin inhibition does not reach.
- NG2 falls equally in the KMT2A-germline control - the fall is a nonspecific drug effect and neither reading applies.
- CD133 falls under pinometostat only - the H3K79 mark, not the programme, is what surface CD133 reads, and CD133 becomes a specific DOT1L pharmacodynamic marker rather than a general one.
- CD10 rises under either drug - the promoter methylation at MME is programme-maintained rather than fixed, which would be a genuinely new result and would connect
gene-mmeto the hypomethylating-agent combinations already in use.
Cost and feasibility. Low. Cell lines, two drugs with published dosing, a flow panel of markers all available as clinical-grade reagents, and FISH. The PDX arm raises cost to medium and is what makes the result about infants rather than about cell lines.
| design | 2x2 factorial: revumenib x pinometostat, plus vehicle and combination, in >=2 KMT2A-rearranged human cell lines (one B-lineage KMT2A::AFF1, one myeloid) and >=2 primary infant KMT2A::AFF1 PDX samples; timepoints 0/48/96/168 h |
|---|---|
| readout | surface NG2/CSPG4 and CD133/PROM1 by flow, reported three ways (percent-positive against lineage- and age-appropriate threshold, MFI, molecules per cell by calibrated beads); secondary panel CD10, CD33, CD117, CD19, CD34, CD38, CD371, CD14, CD64; parallel RT-qPCR for CSPG4 and PROM1 transcript |
| controls | FISH for KMT2A rearrangement at every timepoint (the lesion must be shown not to move); KMT2A-germline CSPG4-positive line for drug-nonspecificity; viability and apoptosis gating; dose matched on transcriptional response not concentration |
| cost tier | low (cell lines) to med (with PDX arm) |
| timeline | 3-4 months for the cell-line arm; 8-10 months including PDX |
| zero cost first step | fetch PMC13233299 (routes 2-4) and check whether NG2/CSPG4 appears in the revumenib immunophenotype study's marker tables |
tests hypotheses
[ "hyp-ng2-tracks-the-transcriptional-programme-not-the-rearrangement" ]
- A positive result immediately invalidates a diagnostic tool that is in clinical use, in the direction that looks like response. Publishing 'NG2 falls under menin inhibition' without simultaneously offering a replacement MRD anchor would leave laboratories with a marker they distrust and nothing to use instead - so a lineage-independent molecular MRD arm should be designed in from the start.
- The result will be read as a reason to stop using NG2 in menin-inhibitor trials even if the fall is small and stays above threshold. Reporting percent-positive against the age- and lineage-appropriate cut-off, not only MFI, is what prevents that misreading.
- Cell-line results here may not transfer. NG2 is documented to be microenvironment-responsive and upregulated at extramedullary sites, so a dish measurement could miss the dominant in vivo determinant of its level. The PDX arm is not optional if the answer is meant to apply to patients.
- Does PMC13233299 already contain NG2 or CSPG4 in its reported marker tables? This should be checked before any laboratory work begins.
- What is the right dose equivalence between revumenib and pinometostat for a fair 2x2? Matching on transcriptional response rather than on concentration is the only defensible choice and requires a pilot.
- Is there a KMT2A-germline CSPG4-positive leukaemia line? The specificity control depends on it and the literature found here does not name one.
- Should ziftomenib, bleximenib or enzomenib be included? Class effects and compound-specific effects are not separable with one menin inhibitor.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
NG2 is an epigenetically regulated direct target gene of the MLL-AF4 fusion protein, negatively regulates NR3C1 and confers glucocorticoid resistance to MLLr B-ALL cells through FLT3 and AP-1-mediated transrepression; NG2 expression is described as a hallmark of MLLr B-ALL and as minimally expressed in healthy haematopoietic cells.
NG2 is a target gene of MLL-AF4 and underlies glucocorticoid resistance in MLLr B-ALL by regulating NR3C1 expression. · Lopez-Millan B, ..., Bueno C · Blood 2024What this evidence cannot show
Abstract read in full from raw efetch XML on 2026-09-20; full text not read. Supplies the premise of the experiment, not a pilot result: **no menin or DOT1L inhibitor appears anywhere in this work.** Cell-line and model systems. If the full text specifies which lines and which fusion, the experiment should use them for comparability.
Evidence that argues against this (1)
- A direct target gene is not necessarily a menin-dependent one, and the experiment is designed so that a null result is interpretable rather than a failure. The 2x2 with pinometostat exists precisely so that 'the drug did not work' and 'the marker is menin-independent' can be told apart. Godfrey L et al. 2021
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In MLL-AF4 leukaemia cells, aberrant PROM1/CD133 expression is essential for leukaemic cell growth and is mediated by direct binding of MLL-AF4, controlled by an intragenic H3K79me2/3 enhancer element that increases enhancer-promoter interactions between PROM1 and the nearby gene TAPT1; in non-expressing cells the PROM1 locus is repressed by PRC2 binding. MLL-AF4 promotes leukaemogenesis mainly through recruitment of DOT1L and increased H3K79me2/3.
H3K79me2/3 controls enhancer-promoter interactions and activation of the pan-cancer stem cell marker PROM1/CD133 in MLL-AF4 leukemia cells. · Godfrey L, ..., Milne TA · Leukemia 2021What this evidence cannot show
Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **The paper measures transcript and chromatin, not surface protein by flow**, so the experiment's assumption that surface CD133 tracks the KEE is an assumption and is one of the things being tested. Cell lines, not primary infant material.
Evidence that argues against this (1)
- Surface protein and transcript can diverge substantially for a heavily glycosylated pentaspan protein with a long half-life, and the AC133 flow epitope is glycosylation-dependent. A null surface result would therefore not cleanly refute the chromatin mechanism, which is why parallel RT-qPCR is in the design. Schneider P et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In sequential bone marrow specimens from 48 patients with relapsed or refractory AML treated with revumenib, dynamic changes in the immunophenotype were observed in 16 of 31 (52%) patients, characterised by switching between myeloid/stem-like and monocytic or myelomonocytic immunophenotypes in either direction, or by substantial changes in antigen expression intensity or leukaemia-associated immunophenotype patterns. The authors conclude that flow-cytometric treatment monitoring after menin inhibition requires recognition of phenotypic changes associated with differentiation.
Immunophenotypic changes following menin inhibition in acute myeloid leukemia. · Loghavi S, ..., Issa GC · Leukemia 2026What this evidence cannot show
Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **PMCID PMC13233299 exists and reading its marker tables is the zero-cost first step of this experiment.** Denominators differ in the abstract (48 treated, 31 evaluable, 16 changed). Adult relapsed/refractory AML; the design here adds the B-lineage infant arm this study lacks. That the changes are bidirectional is the reason viability gating and monocytic markers are in the design.
Evidence that argues against this (1)
- If a large fraction of the phenotype change under menin inhibition is a differentiation-driven shift of the whole population toward monocytic identity, then ANY marker's apparent fall could be a composition effect rather than locus-specific regulation. Monocytic markers and viability gating are in the design specifically to catch that, and without them the experiment would produce a confident wrong answer. Kovach AE et al. 2026
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Among 346 paediatric AML patients tested for CSPG4 expression, CSPG4 expression correlated with KMT2A rearrangement but also occurred in patients without KMT2A rearrangement and had no significant prognostic impact on event-free or overall survival.
Impact of KMT2A Rearrangement and CSPG4 Expression in Pediatric Acute Myeloid Leukemia. · Hoffmeister LM, ..., Schneider M · Cancers (Basel) 2021What this evidence cannot show
Abstract read in full from raw efetch XML on 2026-09-20; full text not read. Cited here to justify the KMT2A-germline CSPG4-positive control arm. **The abstract does not name any specific CSPG4-positive KMT2A-germline cell line**, so identifying a suitable control line is itself a preparatory step and may be the hardest part of the design.
Evidence that argues against this (1)
- If no KMT2A-germline CSPG4-positive line can be found, the specificity control has to be a KMT2A-rearranged line with CSPG4 knocked out and re-expressed from a constitutive promoter - a harder construct, and one whose surface turnover may differ from the endogenous protein. Hoffmeister LM et al. 2021
2x2 revumenib by pinometostat with NG2 and CD133 as co-primary readouts and FISH holding the lesion fixed
CD133 is the discriminating arm: falling under both drugs means the marker reads the programme; under pinometostat only means it reads the H3K79 mark
THE DISCRIMINATOR ALREADY EXISTS AND CLEC12A IS ALREADY IN ITS PANEL: that experiment lists CD371 in its secondary flow panel and uses CD133/PROM1 - an intragenic H3K79me2/me3 enhancer with direct MLL-AF4 binding - as its epigenetic positive control. Adding the occupancy arm at the CLEC12A locus would settle whether 'menin inhibition primes CLEC12A' is a chromatin claim or a differentiation claim.