NPM1c and XPO1: one drug, one output, two incompatible mechanisms - is the mutant needed in the cytoplasm or on the chromatin?
also called NPM1c NES · NPM1c XPO1 dependence · NPM1c HOX maintenance
The commonest mutation in adult AML creates a new export tag that hands the protein to XPO1. Blocking XPO1 switches the HOX genes off - but two well-regarded laboratories disagree completely about why, and their explanations are mirror images of each other.
The same thing, for a scientist
The NPM1 exon-12 frameshift creates a **C-terminal leucine-rich NES** and destroys the nucleolar localisation signal, so NPM1c is aberrantly cytoplasmic. Both camps agree that **XPO1 inhibition collapses HOX expression and drives differentiation**. They disagree about the mechanism, and the disagreement is not a matter of emphasis: - **Oka et al. (address model).** Chromatin-bound CRM1 at the HOX clusters **recruits NPM1c onto that chromatin through the NES-CRM1 interaction**. XPO1 inhibition removes NPM1c from HOX chromatin, and HOX falls [E1]. On this reading NPM1c does its oncogenic work **in the nucleus, on DNA**. - **Brunetti et al. (cytoplasmic-requirement model).** Loss of NPM1c **from the cytoplasm** - by nuclear relocalisation or by targeted degradation - causes immediate HOX down-regulation and differentiation; XPO1 inhibition works precisely **because it relocalises NPM1c to the nucleus** [E2]. On this reading the nucleus is where NPM1c stops working. **The same intervention, the same readout, opposite required localisations.** This graph does not resolve it, and it should not be smoothed: the two models predict opposite results for a nuclear-retained NPM1c mutant that cannot bind CRM1, and that construct decides the question. **A third reading has arrived and it may dissolve the first two.** The 2025 C-body work reports NPM1c forming nuclear condensates - in human lines, mouse models and **primary patient samples** - that are necessary for gene expression and leukaemic expansion, and that recruit NUP98 and KMT2A [E3]. Condensates are a nuclear phenomenon, which sits with Oka and against Brunetti; but 'necessary for expansion' is a statement about an assembly, not about a compartment. **All of this is AML.** NPM1 mutation is essentially absent from infant KMT2A-rearranged B-ALL, and none of these experiments is in a lymphoid cell.
Why this graph cares about NPM1c at all
Three reasons, all structural to the project rather than incidental:
- It is the third lesion on the same address. NPM1c reaches HOXA through the NES-XPO1 contact, SET::NUP214 through FG-XPO1, NUP98::HOXA9 through FG-XPO1. If the address is real, NPM1c is the version of it that has no nucleoporin in it at all - the cleanest test that the FG repeats are not the point.
- It is a menin-inhibitor-enrolling genotype. Revumenib and ziftomenib are being developed in NPM1-mutant AML alongside KMT2A-rearranged disease, so anything true of the NPM1c address is immediately relevant to the trials this graph tracks.
- It is where the menin and XPO1 stories collide. The C-body work puts menin, KMT2A, NUP98 and XPO1 in one object 3, and a 2026 preprint reports XPO1 stabilising menin on chromatin in both KMT2A-rearranged AND NPM1-mutant AML (see
mechanism-xpo1-stabilises-menin-chromatin-binding).
The schedule result, which is the most actionable thing in this entry
Selinexor given once or twice weekly produced no clinical benefit in NPM1-mutated patients, and the reason turns out to be mechanical rather than biological: that schedule only temporarily disrupts the XPO1-NPM1c interaction 4. The second-generation SINE eltanexor, which can be dosed more frequently, produced irreversible HOX down-regulation, terminal differentiation and prolonged survival in NPM1-mutated AML models 4.
This is a rare case where a negative clinical result has a published mechanistic explanation and a specific fix. It also reframes every XPO1-inhibitor result in this beat: an intermittent schedule tests a different intervention from a continuous one, and a preclinical synergy measured under continuous exposure may not survive a weekly clinic.
| mechanism class | chromatin targeting / mislocalisation |
|---|---|
| lesion | NPM1 exon-12 frameshift (NPM1c) |
| grip module | C-terminal leucine-rich NES created by the frameshift; nucleolar localisation signal destroyed |
| receptor | gene-xpo1 |
| output | HOXA/HOXB and MEIS1 expression; differentiation block |
| unresolved contradiction | Oka places the requirement on NUCLEAR chromatin-bound NPM1c; Brunetti places it on the CYTOPLASMIC pool and treats nuclear relocalisation as therapeutic. Both report HOX down-regulation under XPO1 inhibition. |
| deciding experiment | an NPM1c mutant with an intact nuclear localisation but a dead NES, or a cytoplasm-restricted degron versus a nucleus-restricted degron, scored on HOX expression |
| schedule finding | weekly or twice-weekly selinexor disrupts the XPO1-NPM1c interaction only temporarily and gave no clinical benefit; eltanexor's more frequent dosing gave irreversible HOX down-regulation preclinically |
| relevance to infant b all | none directly - NPM1 mutation is essentially absent from infant KMT2A-rearranged B-ALL; relevance is via the shared address and via shared menin-inhibitor trials |
systems
[ "human NPM1-mutant AML lines including OCI-AML3", "murine Npm1-mutated leukaemia", "primary patient samples (C-body work only)" ]
- If Brunetti is right, XPO1 inhibition works in NPM1c AML by putting the mutant back in the nucleus - which is the opposite of the rationale that would be used for SET::NUP214, where the aim is to take the fusion off chromatin. One drug cannot be justified by two opposite mechanisms without someone checking which applies in which disease.
- The weekly-schedule failure in NPM1-mutant patients is a published clinical negative for exactly this rationale. Any new XPO1 proposal has to say how its schedule differs.
- NPM1c biology is adult myeloid biology. Importing it into an infant B-ALL argument is the generalisation failure this project's rule 73 exists to prevent.
- Cytoplasm or chromatin? A nuclear-retained, NES-dead NPM1c separates the two models in one construct and neither group appears to have published it.
- Are the NPM1c condensates chromatin-associated or nucleoplasmic? The abstract does not say, and the answer decides whether the condensate model supports Oka or is orthogonal to both.
- Does the Oka recruitment result reproduce in a laboratory other than Oka's? Both the 2016 and 2019 chromatin-bound CRM1 papers are from one group.
- Is there any lymphoid system in which NPM1c-XPO1 biology has been tested? Everything read here is myeloid.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Chromatin-bound CRM1 recruits NPM1c onto HOX clusters via the NES-CRM1 interaction in human leukaemia cells, alongside its recruitment of SET-Nup214, causing aberrant HOX expression.
Chromatin-bound CRM1 recruits SET-Nup214 and NPM1c onto HOX clusters causing aberrant HOX expression in leukemia cells. · Oka M, Mura S, Otani M, et al. · eLife 2019What this evidence cannot show
Human AML cell lines; abstract and partial full text (PMC6874418) read by earlier beats of this project. The NPM1c arm of the paper is a companion result to the SET-Nup214 work, not its main subject, so the depth of the NPM1c characterisation is lower than for SET-Nup214. Nothing in it is in a lymphoid cell or an infant.
Evidence that argues against this (1)
- Brunetti et al. report the opposite requirement: NPM1c must be CYTOPLASMIC, and XPO1 inhibition works by relocalising it to the nucleus. The two models cannot both be right about where NPM1c has to be. Brunetti L et al. 2018
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In NPM1-mutant AML cells, loss of NPM1c from the cytoplasm - by nuclear relocalisation or by targeted degradation - causes immediate down-regulation of HOX genes followed by differentiation; XPO1 inhibition relocalises NPM1c to the nucleus, promotes differentiation and prolongs survival of Npm1-mutated leukaemic mice.
Mutant NPM1 Maintains the Leukemic State through HOX Expression. · Brunetti L, Gundry MC, Sorcini D, et al. · Cancer Cell 2018What this evidence cannot show
Human NPM1-mutant AML cells and a murine Npm1-mutated model; abstract-level for this beat. The degron arm is the strongest part - it removes the protein without moving it - but the abstract does not say whether degradation was cytoplasm-restricted, and a degron that destroys the whole protein cannot by itself establish that the CYTOPLASMIC pool was the requirement. That is the crux of the disagreement with [E1] and it is not resolved by either abstract.
Evidence that argues against this (1)
- Oka et al. place the oncogenic action of NPM1c on HOX chromatin in the nucleus, recruited there by chromatin-bound CRM1 - the reverse polarity. Oka M et al. 2019
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
NPM1c forms nuclear condensates in human cell lines, mouse models and primary patient samples; its phase separation is necessary and sufficient to recruit NUP98 and KMT2A to those condensates; and the condensates are necessary for regulating gene expression, promoting in vivo leukaemic expansion and maintaining the undifferentiated state.
Disparate leukemia mutations converge on nuclear phase-separated condensates. · Datar GK, Khabusheva E, Anand A, et al. · Cell 2025What this evidence cannot show
Abstract-level. This graph's standing caveat on phase-separation assays applies: hexanediol and classical FRAP cannot discriminate LLPS from binding to spatially clustered sites. The abstract does not say whether the condensates are nucleoplasmic or chromatin-associated, which is exactly the datum the Oka/Brunetti disagreement needs. Primary patient samples in the system list is a genuine strength that neither E1 nor E2 has.
Evidence that argues against this (1)
- NAMED METHODOLOGICAL DEFEATER for 'phase separation is necessary and sufficient'. Muzzopappa et al. benchmarked the standard assays against controlled model systems in vitro and in living cells and found that 1,6-hexanediol treatment and classical FRAP BOTH fail to distinguish liquid-liquid phase separation from molecules binding spatially clustered binding sites without phase-separating; only preferential internal mixing measured by calibrated half-bleaching discriminates. 'Necessary and sufficient to recruit NUP98 and KMT2A' is a strong causal claim about a mechanism that the commonest assays cannot identify - and the rival, binding to clustered sites, is the model this very entry needs to adjudicate between Oka and Brunetti, because clustered binding is chromatin-anchored and phase separation need not be. Muzzopappa F et al. 2022
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Selinexor administered once or twice per week results in only a temporary disruption of the XPO1-NPM1c interaction and did not translate into clinical benefit in patients with NPM1 mutations; the second-generation XPO1 inhibitor eltanexor, which can be given more frequently, caused irreversible HOX down-regulation, induced terminal AML differentiation and prolonged the survival of leukaemic mice in NPM1-mutated models.
Prolonged XPO1 inhibition is essential for optimal antileukemic activity in NPM1-mutated AML. · Pianigiani G, Gagliardi A, Mezzasoma F, et al. · Blood Advances 2022What this evidence cannot show
Abstract-level for this beat. NPM1-mutated AML models, adult disease, no lymphoid or paediatric data. The clinical statement is a summary of prior selinexor experience rather than a trial reported in this paper. 'Irreversible' is the authors' word for the HOX effect and is a strong claim resting on preclinical models.
Evidence that argues against this (1)
- PUBLISHED CONFLICTING RESULT against generalising the schedule argument. Selinexor given TWICE WEEKLY with daunorubicin and cytarabine in 21 previously untreated poor-risk AML patients gave CR/CRi in 8 of 19 evaluable patients, MTD not reached, median overall survival 10.3 months. The intermittent schedule that failed to sustain disruption of the XPO1-NPM1c interaction is therefore not clinically inert in AML generally. What the Blood Advances work supports is the narrow claim that intermittent dosing does not maintain one protein-protein interaction long enough to collapse HOX in an NPM1c model - not the broad claim that weekly dosing cannot produce antileukaemic benefit. Sweet K et al. 2020
NPM1c instance of the route
the exon-12 frameshift mutant of this gene
NPM1c C-terminal NES binds XPO1
HOX maintenance collapses under XPO1 inhibition
- related to →NPM1 (nucleophosmin) domain architecture
where the new NES sits in the protein
the NES-only instance, with no nucleoporin in the lesion
- ← treated byEltanexor (KPT-8602) - the second-generation SINE whose whole point is that you can give it more often
prolonged XPO1 inhibition gives irreversible HOX down-regulation where weekly selinexor does not
'weekly selinexor only transiently disrupts XPO1-NPM1c and therefore fails' is a narrow claim about one protein-protein interaction; the same intermittent schedule with 7+3 produced 53% CR/CRi and 10.3 months median overall survival