NG2 reads the fusion's transcriptional programme, not the DNA lesion - so a menin inhibitor should erase the graph's best diagnostic marker while the rearrangement stays put
also called NG2 is a programme marker · CSPG4 under menin inhibition · surrogate marker decoupling
NG2 is the protein used to spot a broken KMT2A gene by flow cytometry, but NG2 is switched on by the broken gene's activity rather than by the break itself. If that is right, a drug that stops the broken gene from working should make NG2 disappear while the break is still there - which would quietly break the test doctors use to track the disease.
The same thing, for a scientist
**Statement.** NG2/CSPG4 surface expression is a read-out of the KMT2A fusion's active transcriptional programme rather than of the rearrangement. Therefore **menin inhibition, which displaces the fusion from chromatin, should reduce surface NG2 in KMT2A-rearranged blasts that still carry the rearrangement** - decoupling the graph's best immunophenotypic surrogate from the genotype it stands for. **Why this is more than a guess.** NG2 is an **epigenetically regulated direct target gene of MLL-AF4**, and the mechanism is worked out: it suppresses *NR3C1* through FLT3 and AP-1 transrepression [E1]. A direct fusion target gene is by construction downstream of the fusion's chromatin activity. Three independent observations say NG2 already behaves like a state variable rather than a genotype variable: - **It is malleable.** Serial xenotransplantation shows NG2 does not enrich for leukaemia-initiating cells and is strongly upregulated in blasts infiltrating CNS and extramedullary sites - the authors' own word is *"malleable"* [E2]. - **Its optimal threshold depends on lineage and age**, from 7% in infant ALL to 28% in infant AML [E3]. A marker of a DNA lesion should not need four thresholds. - **It occurs without the lesion.** CSPG4 expression in 346 paediatric AML samples correlated with KMT2A-r and **also occurred in KMT2A-germline patients** [E4]. **And the experiment has not been done.** `NG2[tiab] AND (menin[tiab] OR revumenib[tiab] OR SNDX-5613[tiab] OR VTP-50469[tiab])` returns **0** records; `CSPG4[tiab] AND (menin[tiab] OR "menin inhibitor"[tiab])` returns **0**; the unbounded `(CSPG4 OR NG2) AND leukemia AND (menin OR revumenib OR "menin inhibitor")` returns **0**, querytranslation inspected [E5]. Meanwhile the general measurement exists: sequential flow on revumenib-treated AML found dynamic immunophenotypic change in **16 of 31 (52%)** patients, with the authors concluding that flow monitoring after menin inhibition requires recognising differentiation-associated phenotype change [E6]. **NG2 is simply not in that panel's reported markers.** **Prior probability: 0.6.** The mechanism is established, the marker's plasticity is established, and the direction is predicted by both. What is not established is magnitude - a partial fall that stays above a 7% threshold would be biologically real and clinically irrelevant.
What follows if it is true, and the sharper version for PROM1
If true:
- NG2 cannot be used for MRD under menin inhibition - it would fall as a treatment effect and over-call remission, the same failure mode as a B-lineage panel scoring a lineage-switched blast as remission.
- NG2 becomes a pharmacodynamic marker instead, and a very cheap one: same-day, single-cell, on routine equipment, where the alternative is a marrow ChIP.
- Diagnosis is unaffected but re-diagnosis is not. A relapse marrow from a previously NG2-positive patient on a menin inhibitor could be NG2-negative, and a laboratory using NG2 to confirm the same disease would be misled.
gene-prom1 is the better-powered version of the same test. CD133 is driven by an intragenic H3K79me2/3 enhancer element with direct MLL-AF4 binding, and is PRC2-repressed when off. That gives a two-drug discrimination that NG2 cannot provide:
| DOT1L inhibitor (pinometostat) | menin inhibitor (revumenib) | |
|---|---|---|
| CD133 falls under both | marker reads the programme | |
| CD133 falls only under pinometostat | marker reads the H3K79 mark | |
| NG2 falls under both | supports this hypothesis | |
| NG2 falls under neither | refutes it |
Running both markers under both drugs in one experiment is the efficient design, and it is exp-surface-marker-panel-under-menin-and-dot1l-inhibition.
| statement | NG2/CSPG4 surface expression reads the KMT2A fusion's active transcriptional programme rather than the DNA rearrangement, so menin inhibition should reduce surface NG2 in cells that retain the rearrangement. |
|---|---|
| origin | literature-gap |
| prior probability | 0.6 |
predictions
[ "Surface NG2 falls measurably in KMT2A-rearranged blasts under revumenib or another menin inhibitor, while FISH remains positive for the rearrangement.", "CD133/PROM1 falls under both DOT1L and menin inhibition, because its enhancer is H3K79-marked and fusion-bound.", "NG2 does NOT fall at lineage switch, because the fusion is retained through the switch by definition - so NG2 and CD19 dissociate in the opposite directions at switch and under menin inhibition.", "NG2 recovers as menin-inhibitor resistance emerges, tracking programme restoration." ]
falsifiers
[ "Surface NG2 is unchanged in KMT2A-rearranged blasts after a menin inhibitor at a dose that produces the expected differentiation and transcriptional response. That would place CSPG4 at a menin-independent fusion site and kill the hypothesis.", "NG2 falls equally in KMT2A-germline CSPG4-positive leukaemia cells treated with the same drug, showing the fall is a nonspecific drug effect rather than programme-dependent.", "ChIP shows the KMT2A fusion is not bound at the CSPG4 locus in primary infant KMT2A::AFF1 blasts, contradicting the direct-target premise in primary cells.", "Surface NG2 falls under menin inhibition but remains above the lineage- and age-appropriate positivity threshold in every patient, making the effect real and clinically irrelevant." ]
experiments
[ "exp-surface-marker-panel-under-menin-and-dot1l-inhibition", "exp-marker-order-at-lineage-switch-paired-flow-reanalysis" ]
- If this hypothesis is right, the most direct consequence is harmful: a diagnostic and MRD marker stops working during treatment, in the direction that looks like response. Any menin-inhibitor protocol using NG2 in its flow panel would over-call remission, and the correct response is paired lineage-independent molecular MRD, which this graph's MRD beat already found no menin trial has reported.
- The pharmacodynamic use is the attractive consequence and it is also a trap. A marker that falls early and recovers as resistance emerges would report drug effect during the honeymoon and go quiet exactly when escape begins - the same pattern the DOT1L resistance literature describes for adaptive rather than binary loss.
- If NG2 does NOT move under menin inhibition, the hypothesis fails in an informative way: CSPG4 would be a menin-independent KMT2A-fusion target site, which bears directly on the menin-interface hypothesis this whole graph rests on. The negative result is worth as much as the positive one and should be designed for, not treated as a failure.
- Does surface NG2 fall under revumenib in KMT2A-rearranged blasts that still carry the rearrangement, and by how much relative to the 7% infant-ALL threshold?
- Is CSPG4 a menin-dependent or menin-independent KMT2A-fusion binding site? A single ChIP under drug answers it and bears on the whole menin-interface argument.
- Does PMC13233299 - the full text of the revumenib immunophenotype study - contain NG2 or CSPG4 in its marker tables? One fetch could partly resolve this hypothesis with no new experiment.
- If NG2 falls, does it fall in the same cells that lose the leukaemic programme, or in a subpopulation? Single-cell surface-plus-transcriptome would distinguish a uniform shift from selection.
- 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; it negatively regulates expression of the glucocorticoid receptor NR3C1 and confers glucocorticoid resistance to MLLr B-ALL cells, mechanistically by interacting with FLT3 to produce ligand-independent FLT3 signalling and downregulation of NR3C1 via AP-1-mediated transrepression.
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**, so the strength of the direct-binding evidence is taken on the authors' word. Cell-line and model work. **Nothing in this paper involves a menin inhibitor**, so it supplies the premise of this hypothesis and none of its test.
Evidence that argues against this (1)
- Being a direct fusion target gene does not guarantee menin-dependence. Menin inhibition displaces the fusion from a subset of loci, and menin-independent KMT2A-fusion chromatin binding is an active question in this graph. If CSPG4 sits at a menin-independent site, NG2 would not move under revumenib and the hypothesis fails for a reason that is itself informative. Loghavi S et al. 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Serial xenotransplantation of primary MLL-AF4-positive leukaemias indicated that NG2 is a malleable marker that does not enrich for leukaemia-initiating cells or CNS-infiltrating cells in MLL-rearranged infant B-ALL; NG2 expression was highly upregulated in blasts infiltrating extramedullary haematopoietic sites and CNS, specific blockade of NG2 resulted in almost complete loss of engraftment, and gene expression profiling revealed a migratory signature in NG2-positive blasts. High NG2 expression in the Interfant cohort associated with lower event-free survival, more circulating blasts and more frequent CNS disease or relapse.
NG2 antigen is involved in leukemia invasiveness and central nervous system infiltration in MLL-rearranged infant B-ALL. · Prieto C, ..., Menendez P · Leukemia 2018What this evidence cannot show
Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **This paper carries a published correction (PMID 30218009) that this entry has not read.** 'Malleable' is the authors' characterisation from serial transplantation, and site-dependent upregulation is a different kind of plasticity from drug-induced downregulation - the hypothesis extrapolates from one to the other and that extrapolation is not tested.
Evidence that argues against this (1)
- Upregulation at extramedullary sites shows NG2 responds to the microenvironment, which is compatible with the marker reading a migratory state rather than the fusion's core programme. If NG2 is environment-driven rather than fusion-dose-driven, a menin inhibitor need not move it. Hoffmeister LM et al. 2021
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ROC analysis on 505 paediatric acute leukaemia patients with NG2 positivity above 1% established different optimal threshold levels for NG2 positivity by lineage and age: 7% in infants with ALL, 12% in older children with ALL, 28% in infants with AML and 14% in patients over 1 year with AML.
Quantification of NG2-positivity for the precise prediction of KMT2A gene rearrangements in childhood acute leukemia. · Zerkalenkova E, ..., Popov A · Genes Chromosomes Cancer 2021What this evidence cannot show
Abstract read in full from raw efetch XML on 2026-09-20; full text not read. Thresholds derived and evaluated in the same cohort, conditioned on NG2 above 1%. **The inference drawn here - that four thresholds imply a state variable - is this entry's, not the authors'**, and an alternative explanation is simply that antigen density differs by lineage for reasons unrelated to fusion activity.
Evidence that argues against this (1)
- Lineage- and age-dependent thresholds could reflect differences in autofluorescence, background staining and blast size between lymphoid and myeloid samples rather than differences in the marker's biology. That is a mundane explanation and the data cannot exclude it. Emerenciano M et al. 2011
- 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 within a cohort of 967 with known KMT2A-r status, a correlation between CSPG4 expression and KMT2A rearrangement was confirmed, but CSPG4 expression 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. CSPG4 was tested in only 346 of 967 with unstated selection. **AML, where NG2 specificity is known to be worst.** That CSPG4 occurs without the rearrangement supports a programme-level rather than lesion-level reading, but it is equally compatible with a second, unrelated route to CSPG4 expression in myeloid cells.
Evidence that argues against this (1)
- CSPG4 expression without KMT2A rearrangement could reflect a convergent myeloid programme with nothing to do with the fusion, in which case it says nothing about whether the fusion's programme drives CSPG4 in rearranged cases. The observation is consistent with the hypothesis and does not discriminate. Lopez-Millan B et al. 2024
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
No published record intersects NG2 or CSPG4 with menin inhibition. PubMed on 2026-09-20 returned 0 records for `NG2[tiab] AND (menin[tiab] OR revumenib[tiab] OR SNDX-5613[tiab] OR VTP-50469[tiab])`, 0 for `CSPG4[tiab] AND (menin[tiab] OR "menin inhibitor"[tiab])`, and 0 for the unbounded `(CSPG4 OR NG2) AND leukemia AND (menin OR revumenib OR "menin inhibitor")`, with querytranslation inspected confirming CSPG4 mapped to its supplementary concept rather than being dropped. The repository was also grepped and no entry in this graph cites such a measurement.
What this evidence cannot show
**A bounded absence.** Three symbol forms and four drug forms were run. **Not run: the antibody clone name `7.1` (unusable as a search token), `MCSP`, `ziftomenib`, `bleximenib`, `enzomenib`, `DS-1594` and `JNJ-75276617`.** A trial reporting an NG2 panel in a supplementary table is invisible to all of these, and so is a menin-inhibitor paper using a drug name not on the list. The absence bounds the indexed title/abstract literature, nothing more.
Evidence that argues against this (1)
- The measurement of immunophenotype under menin inhibition exists and NG2 is simply not reported in it. So 'nobody has looked' is too strong; the accurate statement is that the relevant panels have been run and NG2 was not among the reported markers, which is a different and more tractable gap. Loghavi S et al. 2026
- E6strong 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 after treatment were observed in 16 of 31 (52%) patients, characterised by a switch between myeloid/stem-like and monocytic or myelomonocytic immunophenotypes in either direction, or by substantial changes in the intensity of antigen expression or in patterns of leukaemia-associated immunophenotypes. 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, so a reader can run routes 2-4 to see whether NG2 or CSPG4 appears in the paper's marker tables** - that single check could partly resolve this hypothesis without any new experiment. Denominators differ within the abstract: 48 treated, 31 sequentially evaluable, 16 changed. Adult relapsed/refractory AML; KMT2A-rearranged and NPM1-mutant patients are not separated in the abstract.
Evidence that argues against this (1)
- That the changes are bidirectional - stem-like to monocytic AND monocytic to stem-like - argues against a simple 'the fusion programme is switched off, so its target genes fall' model. If the phenotype moves in both directions, the relationship between menin inhibition and any single fusion target gene's surface level may not be monotonic. Loghavi S et al. 2026
predicts NG2 falls under menin inhibition while FISH stays positive; falsified if NG2 is unchanged at a dose with a transcriptional response
the intervention that would decouple the marker from the genotype; zero published records intersect NG2 with any menin inhibitor
2x2 revumenib by pinometostat with NG2 and CD133 as co-primary readouts and FISH holding the lesion fixed
if NG2 is a lineage-state marker rather than a lesion marker, its behaviour under the drug belongs to that cell state