In NUP98-rearranged leukaemia, who responds to a menin inhibitor is decided by the co-mutations, not by the fusion
also called co-mutation gating hypothesis · NUP98 menin response predictors · FLT3-ITD WT1 menin response
Every NUP98 patient is currently treated as one group for menin drugs. The evidence suggests the fusion only makes the cell eligible to respond, and whether it actually responds is decided by the other mutations it happens to carry.
The same thing, for a scientist
**Statement.** In NUP98-rearranged leukaemia the fusion establishes *susceptibility* to menin inhibition but does not determine *response*; the determinant is the co-mutation landscape — principally FLT3-ITD, *WT1*, RAS-pathway lesions and MEIS1 output. The hypothesis is built from a set of observations that are individually weak and jointly awkward: the founding menin study's **non-responding** NUP98-NSD1 PDX carried *WT1*, *ASXL1*, *IDH1*, *BCORL1* and **FLT3-ITD**, giving 2 of 3 PDX responders [E1]; a second laboratory reports the **opposite** for the same co-lesions, finding monotherapy sufficient in NUP98-rearranged AML with FLT3-ITD **and** *WT1* [E2]; the only published NUP98 complete remission ended in **RAS-pathway** escape [E3]; **MEIS1 re-expression abrogates sensitivity** outright [E1]; and 74–91% of NUP98::NSD1 patients carry FLT3-ITD, so an unstratified NUP98 cohort is effectively a dual-positive cohort [E4]. **No trial stratifies NUP98-rearranged patients by any of these.** The hypothesis is cheap to test on samples that already exist: the SAVE trial enrolled ~9 NUP98-rearranged patients with molecular characterisation, and every menin-inhibitor programme has banked baseline sequencing.
The four gating candidates, ranked by how testable they are
- FLT3-ITD. Highest prevalence, most disputed. Two laboratories report opposite effects in NUP98-NSD1 PDX 12, and the only combination paper in the family pairs menin inhibition with a FLT3 inhibitor 5. Directly testable in banked trial samples: response by FLT3-ITD status within the NUP98 subgroup.
- WT1. Present in 42–50% of NUP98::NSD1 4 and present in both the responding and the non-responding PDX 1, which makes it a poor single candidate but a plausible modifier in combination with FLT3-ITD.
- MEIS1 output. Mechanistically the most direct — enforced MEIS1 abrogates sensitivity 1 — but requires RNA rather than DNA and is not collected.
- RAS-pathway lesions. Observed at escape rather than at baseline in the one reported case 3; whether baseline RAS lesions predict primary non-response is untested.
The ranking matters because the first two are already in every banked sequencing dataset and the analysis could be run without a new patient.
Why this competes with the partner-identity explanation
This graph also holds a partner-based account: menin dependency is graded across NUP98 partners (VTP50469 IC50 84 nM for KDM5A against 517 nM for HOXA9) and abolished in NUP98::KMT2A (hyp-menin-dependency-tracks-available-mbm).
The two hypotheses are not mutually exclusive but they are confounded in every existing dataset, because partner and co-mutation travel together: NSD1 carries FLT3-ITD at 74–91% while other partners carry it at 18% 4. So any observed partner difference in drug response could be a co-mutation difference, and vice versa.
The discriminating analysis is a two-way split — partner × FLT3-ITD status — which needs more NUP98 patients than any single trial has enrolled. That is an argument for pooling across menin-inhibitor programmes, and it is the single most valuable thing the field could do with data it already owns.
What would kill it
- A NUP98 cohort in which response is uniform across co-mutation strata. If FLT3-ITD-positive and negative NUP98 patients respond at the same rate, co-mutation gating is dead and partner identity or fusion biology takes over.
- An isogenic model in which adding FLT3-ITD to a NUP98-fusion background does not change menin-inhibitor sensitivity. This is the clean experiment: the two published PDX comparisons are across laboratories, models and co-mutation constellations, which is why they disagree.
- Demonstration that the two published PDX results differ for a technical reason — dose, schedule, engraftment — rather than a biological one. That would remove the observation the hypothesis is built on.
- A NUP98::KMT2A patient who responds. That would favour the co-mutation account over the MBM account, since NUP98::KMT2A's resistance is attributed to the fusion's structure.
| statement | In NUP98-rearranged leukaemia the fusion establishes susceptibility to menin inhibition but the co-mutation landscape (FLT3-ITD, WT1, RAS-pathway lesions, MEIS1 output) determines response. |
|---|---|
| origin | literature-gap |
| prior probability | 0.4 |
| confounded with | hyp-menin-dependency-tracks-available-mbm; partner identity and co-mutation travel together because NSD1 carries FLT3-ITD at 74-91% while other partners carry it at 18% |
| discriminating analysis | two-way split of response by partner AND FLT3-ITD status, which needs more NUP98 patients than any single trial has enrolled - an argument for pooling across menin-inhibitor programmes |
| testable on existing samples | true |
predictions
[ "Within a NUP98-rearranged menin-inhibitor cohort, response rate differs by FLT3-ITD status.", "Adding FLT3-ITD to an isogenic NUP98-fusion background changes menin-inhibitor sensitivity.", "Baseline MEIS1 expression is inversely associated with depth of response.", "Baseline RAS-pathway lesions predict primary non-response, not only relapse." ]
experiments
[ "exp-mbm-addback-and-trans-deletion-panel" ]
analyses
[ "ana-depmap-nup98-men1-power" ]
falsifiers
[ "A NUP98 cohort in which response is uniform across FLT3-ITD strata.", "An isogenic model in which adding FLT3-ITD does not change menin-inhibitor sensitivity.", "Demonstration that the two published PDX results differ for a technical reason (dose, schedule, engraftment) rather than a biological one.", "A NUP98::KMT2A patient who responds to a menin inhibitor, which would favour co-mutation gating over the structural MBM account and simultaneously undercut this hypothesis's framing of the fusion as merely establishing susceptibility." ]
no trial currently stratifies by
[ "FLT3-ITD within NUP98-rearranged patients", "WT1", "MEIS1 expression" ]
- If response is co-mutation-gated, a partner-blind and co-mutation-blind NUP98 trial could return a modest overall response rate that conceals a highly responsive subgroup, and the drug would be abandoned for the patients it works in.
- Conversely, if the hypothesis is wrong and response is uniform, stratifying would fragment an already tiny population into subgroups too small to report - the cost of over-stratifying a rare disease is real.
- Every gating candidate except FLT3-ITD requires an assay nobody runs at baseline, so acting on this hypothesis means adding correlative science to trials in a population where accrual is already the binding constraint.
- Why do two laboratories reach opposite conclusions about FLT3-ITD and menin-inhibitor response in NUP98-NSD1 PDX, and has anyone asked them to compare protocols?
- Could the SAVE trial's nine NUP98-rearranged patients be analysed by FLT3-ITD status? The samples and the sequencing exist; only the analysis is missing.
- Is MEIS1 expression measurable reliably enough in a diagnostic marrow to serve as a stratification variable?
- Do baseline RAS-pathway lesions predict primary non-response, or only relapse? These are different questions and only the second has any data.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the founding menin-dependency study, 2 of 3 NUP98-rearranged PDX models responded while one NUP98-NSD1 model carrying WT1, ASXL1, IDH1, BCORL1 and FLT3-ITD did not, and Meis1 overexpression or CRISPR activation abrogated menin-inhibitor sensitivity.
The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML · Heikamp EB, et al. · Blood 139(6):894-906 2022What this evidence cannot show
Three PDX models is a very small denominator, and the non-responder carried five co-mutations at once so no single lesion can be assigned responsibility. Read via a delegated full-text pass in this session. A non-responding xenograft can also fail for engraftment or pharmacokinetic reasons unrelated to biology.
Evidence that argues against this (1)
- The same study's responding PDX was also WT1-mutant, so WT1 alone cannot be the gate, and 2 of 3 responding is more consistent with a generally sensitive disease with occasional refractory cases than with a co-mutation switch. Rasouli M et al. 2023
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
A second laboratory reported that menin-inhibitor monotherapy was sufficient to suppress NUP98-rearranged AML with co-occurring FLT3-ITD and WT1 mutations, the opposite conclusion about the same co-lesions.
The MLL-Menin Interaction is a Therapeutic Vulnerability in NUP98-rearranged AML · Rasouli M, et al. · HemaSphere 7(8):e935 2023What this evidence cannot show
Different laboratory, different models, different compound (revumenib rather than VTP50469), different endpoints. The disagreement with the founding study may be methodological rather than biological, and no head-to-head comparison has been attempted.
Evidence that argues against this (1)
- The same paper characterises the effect as differentiation and suppression of proliferation rather than apoptotic death, so 'sufficient to suppress' is a weaker claim than it sounds and is not straightforwardly the opposite of the founding study's non-response. Heikamp EB et al. 2022
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The only published complete remission of NUP98-rearranged AML on a menin inhibitor progressed with emergent CEBPA, KRAS G13D, NRAS G13D, WT1 and RUNX1 variants and no MEN1 binding-site mutation.
Complete remission of NUP98 fusion-positive acute myeloid leukemia with the covalent menin inhibitor BMF-219, icovamenib · Carraway HE, et al. · Haematologica 110(4):1041-1046 2025What this evidence cannot show
A single patient, post-transplant, on a drug whose oncology programme has since been terminated. Escape co-mutations at progression are not the same thing as baseline co-mutations gating response, and this hypothesis is about the latter.
Evidence that argues against this (1)
- RAS-pathway subclones emerge under many selective pressures in relapsed AML, so this patient's escape may say nothing specific about menin inhibition, and using an escape genotype to argue about baseline response gating conflates two different questions. Issa GC et al. 2026
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
NUP98::NSD1 co-occurs with FLT3-ITD in 74-91% and with WT1 mutation in 42-50% across cohorts, against 18% and 16% for other NUP98 partners, so partner identity and co-mutation status are confounded in every existing dataset.
Comprehensive molecular and clinical characterization of NUP98 fusions in pediatric acute myeloid leukemia · Bertrums EJM, et al. · Haematologica 108(8):2044-2058 2023What this evidence cannot show
The 74-91% range spans cohorts with different detection methods and eras. The 18%/16% comparison for non-NSD1 partners comes from a separate adult cohort, so the contrast pools two populations.
Evidence that argues against this (1)
- Because partner and co-mutation are confounded, an equally good explanation of every response difference is partner identity acting through the menin tether, which this graph holds as a separate hypothesis. Neither can be preferred on current data. Fisher JN et al. 2020
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Menin inhibition in NUP98-translocated leukaemia shows limited single-agent sensitivity and is combined with CDK6 or FLT3 inhibition for effect, which presupposes that a co-lesion limits the response.
Combination of menin and kinase inhibitors as an effective treatment for leukemia with NUP98 translocations · Miao H, et al. · Leukemia 38(8):1674-1687 2024What this evidence cannot show
A combination benefit does not establish that the co-lesion gates the monotherapy response; additive activity of two independently active drugs produces the same result. The primary data were not read in this session.
Evidence that argues against this (1)
- CDK6 inhibition is not a co-mutation-directed intervention in most NUP98 leukaemias, so the combination benefit may reflect general cell-cycle vulnerability rather than anything about co-mutation gating - which would make the combination result irrelevant to this hypothesis. Michmerhuizen NL et al. 2025
an unstratified NUP98 cohort is effectively a dual-positive cohort
FLT3-ITD as the highest-prevalence, most-disputed gating candidate
nine NUP98-rearranged patients with banked molecular characterisation
confounded rival: partner and co-mutation travel together in every dataset
RAS-pathway escape in the only published NUP98 complete remission