Oncofusion

start / graph / risk-xpo1-pan-essentiality-and-p2ry2-akt-escape

Riskstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q61-export-axis

Two reasons XPO1 is a hard target: every cell needs it, and blocking it switches on a survival pathway

also called XPO1 common essentiality · P2RY2-AKT adaptive resistance to selinexor · SINE therapeutic window

XPO1 is needed by essentially every human cell, so there is no built-in safety margin from the target itself. And when you block it, leukaemia cells turn on a survival signal that partly rescues them - a resistance mechanism that appears with the first dose rather than after months.

The same thing, for a scientist

**Risk 1 - essential-function toxicity, quantified.** In this project's own genome-wide DepMap scan, XPO1's mean gene effect is **-2.142 across 1,208 cell lines** [E1]. For scale, this graph's index dependency MEN1 sits at **-0.363** pan-essential and is called selective because KMT2A-rearranged lines reach -1.263. **XPO1 is six times more essential across all lines than MEN1 is in the lines that depend on it.** Its KMT2A-selectivity difference is **-0.070**, rank 2,228 of 17,728. **And the honest caveat cuts both ways.** At a pan-essentiality of -2.14 XPO1 is in the **floor-compression** regime this graph's brief names: there is almost no dynamic range left for selectivity to appear, so the near-zero difference is **uninterpretable, not a measured null**. HCFC1 (-0.073 difference, 100% of lines depleted) and KAT8 (+0.004, 99.5%) are the worked examples. So the correct statement is *XPO1 shows no measurable KMT2A-selectivity and the measurement could not have detected it* - not *XPO1 is not selective*. **Risk 2 - adaptive resistance from the first dose.** Selinexor **activates PI3K-gamma-dependent AKT signalling in AML by up-regulating the purinergic receptor P2RY2**; inhibiting that axis potentiates selinexor in AML cell lines, patient-derived primary cultures and multiple mouse models, and in a **syngeneic MLL-AF9 mouse AML model** selinexor plus ipatasertib outperformed both chemotherapy and chemotherapy plus selinexor [E2]. This is **drug-induced**, not selected - it is a pharmacodynamic consequence of XPO1 inhibition, present immediately. **Risk 3 - the clinical record.** The only randomised test of selinexor in acute leukaemia was **negative and numerically unfavourable**: 2:1 open-label phase 2 (NCT02088541) in older relapsed/refractory AML, median OS **3.2 versus 5.6 months** (HR 1.18, 95% CI 0.79-1.75, p=0.422), with more adverse events on selinexor [E3]. And the paediatric phase 1 found **reversible cerebellar toxicity dose-limiting at 70 mg/m2** in children [E4]. **Risk 4 - schedule.** Weekly or twice-weekly selinexor disrupts the XPO1-NPM1c interaction only **temporarily**, which is the published explanation for why it produced no benefit in NPM1-mutated patients [E5]. A schedule that engages the target intermittently is a different intervention from one that engages it continuously, and the toxicity ceiling constrains which you can give.

Go deeper
What a therapeutic window would have to come from, since it cannot come from the target

Three routes are available in principle and each has a specific problem here:

  1. Greater reliance. The leukaemia uses the platform and normal progenitors use it less. Untested - nobody has measured chromatin-bound XPO1 in a normal haematopoietic progenitor, which is prediction 6 of hyp-xpo1-is-a-shared-address-for-hoxa-driven-leukaemia.
  2. Schedule. Intermittent dosing that spares normal tissue. Contradicted by the NPM1c schedule result, which says intermittent dosing is why the drug failed.
  3. Combination at reduced dose. The ziftomenib-plus-selinexor preprint explicitly reports efficacy at reduced doses of each [see mechanism-xpo1-stabilises-menin-chromatin-binding]. This is the only route with any supporting data and it is a preprint.

There is no fourth route, because a chromatin-selective XPO1 ligand does not exist even as a chemical concept - every agent binds Cys528 in the NES groove, which is the export surface.

Facts and figures
kindessential-function
severityhigh - no therapeutic index available from the target's biology
mitigationcombination at reduced dose is the only route with supporting data, and that data is a preprint; greater tumour reliance on the chromatin pool is the theoretically attractive route and is entirely unmeasured
no chromatin selective ligandevery agent binds Cys528 in the NES groove, the export surface; a chromatin-pool-selective XPO1 ligand does not exist as a concept
affects
[
  "drug-selinexor",
  "drug-eltanexor"
]
quantified
{
  "depmap_pan_essentiality": -2.1422,
  "depmap_n_lines": 1208,
  "kmt2a_selectivity_difference": -0.0697,
  "kmt2a_selectivity_rank": "2228 of 17728",
  "comparator_men1_pan_essentiality": -0.3631,
  "comparator_men1_rank": 1,
  "floor_compression": "at -2.14 the measurement cannot detect selectivity; the null is uninterpretable"
}
adaptive resistance
{
  "mechanism": "drug-induced upregulation of P2RY2 driving PI3K-gamma-dependent AKT signalling",
  "timing": "immediate pharmacodynamic consequence, not selected over time",
  "named_combination_partner": "ipatasertib (AKT inhibitor), outperformed chemotherapy plus selinexor in a syngeneic MLL-AF9 mouse model"
}
clinical record
{
  "randomised_aml_phase2": "NCT02088541, median OS 3.2 vs 5.6 months, HR 1.18 (0.79-1.75), p=0.422, more AEs on selinexor",
  "paediatric_dlt": "reversible cerebellar toxicity at 70 mg/m2; tolerable dose 55 mg/m2",
  "schedule_problem": "weekly/twice-weekly dosing engages the target only temporarily"
}
How this could bite us back
  • This entry IS the bite-back for the whole export axis. Every attractive mechanism in this beat has to be read against a pan-essential target, a negative randomised trial, a paediatric CNS dose-limiting toxicity and an immediate adaptive escape pathway.
  • The floor-compression caveat is double-edged: it prevents this graph from claiming XPO1 is non-selective, which means the reassurance that 'it is clearly a bad target' is not available either. The honest state is that the measurement cannot see the answer.
  • Combination at reduced dose is the only mitigation with data, and combinations are also where toxicity profiles add.
Still open
  1. Is chromatin-bound XPO1 present in normal haematopoietic progenitors? That single measurement decides whether a therapeutic window from differential reliance is even possible.
  2. Does the P2RY2-AKT escape occur in lymphoid leukaemia, or is it myeloid? The paper is AML throughout.
  3. Would eltanexor's more frequent schedule widen or narrow the window? More continuous target engagement is what the NPM1c work says is needed and is also more cumulative normal-tissue exposure.
  4. Has anyone measured whether the negative randomised AML result would have been different in an NPM1-mutant or address-using subgroup? The trial was genotype-unselected and the samples presumably exist.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    In a genome-wide DepMap scan over 17,728 protein-coding genes, XPO1 has a pan-essentiality of -2.142 across 1,208 cell lines, a mean gene effect of -2.332 in 11 KMT2A-rearranged leukaemia lines against -2.263 in 43 other leukaemia lines, a selectivity difference of -0.070 and rank 2,228 of 17,728; for comparison MEN1 ranks 1st with a difference of -0.866 and a pan-essentiality of -0.363.

    DepMap KMT2A-selective dependency scan (full genome), analyses/depmap-kmt2a-selective-scan-full · Oncofusion Knowledge Graph project · internal analysis over DepMap CRISPR gene effect 2026
    What this evidence cannot show

    This graph's own scan, read 2026-09-20. FLOOR COMPRESSION: at -2.14 pan-essentiality there is almost no dynamic range in which selectivity could appear, so the near-zero difference is uninterpretable rather than a measured absence of selectivity. The 11 KMT2A-rearranged lines are myeloid-enriched and none is infant-derived. A gene effect is a CRISPR knockout, which is complete and permanent; a drug gives partial, reversible occupancy, so pan-essentiality is not by itself a prediction of clinical toxicity - it is a statement about the biology's margin.

    Evidence that argues against this (1)
    • Floor compression cuts the other way too: at a pan-essentiality of -2.14 this measurement could not have detected KMT2A-selectivity even if it existed, so the DepMap number is not evidence that XPO1 is a bad target - only that the scan cannot say. And a CRISPR knockout is complete and permanent whereas a drug gives partial reversible occupancy, so gene-effect scores systematically overstate the toxicity implication of inhibiting an essential gene. A 2026 preprint reports efficacy at REDUCED doses of both agents in combination, which is exactly the regime a pan-essentiality score cannot model. Uddin MH et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Selinexor activates PI3K-gamma-dependent AKT signalling in AML by upregulating the purinergic receptor P2RY2; inhibiting this axis potentiates selinexor in AML cell lines, patient-derived primary cultures and multiple mouse models; and in a syngeneic MLL-AF9-driven mouse model of AML, selinexor plus ipatasertib outperformed both standard-of-care chemotherapy and chemotherapy with selinexor.

    PMID 35668193DOI
    What this evidence cannot show

    Abstract-level for this beat. Murine syngeneic MLL-AF9 myeloid leukaemia plus human AML lines and primary cultures - not infant B-lineage KMT2A::AFF1. The finding is framed by the authors as actionable rather than prohibitive, and it is both: an immediate adaptive escape and a named combination partner.

    Evidence that argues against this (1)
    • NAMED METHODOLOGICAL DEFEATER: the P2RY2 upregulation is attributed to XPO1 inhibition without the drug-resistant-allele control that this target uniquely supports. Neggers et al. built XPO1-C528S leukaemia lines by CRISPR/Cas9 and showed that a single heterozygous substitution confers selinexor resistance dominantly, which makes an isogenic resistant line the definitive test of whether any selinexor-induced transcriptional change is on-target. Without it, 'selinexor upregulates P2RY2' is compatible with an off-target effect of the compound, and the combination result with ipatasertib - which would still work if P2RY2 induction were off-target - does not distinguish the two. Neggers JE et al. 2016
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a 2:1 randomised open-label phase 2 (NCT02088541) of selinexor 60 mg twice weekly versus physician's choice in older patients with relapsed or refractory AML, median overall survival was 3.2 versus 5.6 months (HR 1.18, 95% CI 0.79-1.75, p=0.422) and selinexor patients had more adverse events.

    PMID 34323164DOI
    What this evidence cannot show

    Recorded in `drug-selinexor` as counter-evidence and repeated here because it is the single most important constraint on the whole export axis. Older adults, unselected by genotype, twice-weekly schedule - and the schedule is exactly what the NPM1c work later identified as insufficient, so this negative is confounded with the dosing question. Citation identifiers taken from the existing `drug-selinexor` entry's prose and re-checked against PubMed in this beat.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT from the same class of agent in a different setting, and it cuts against reading this negative as the single most important constraint on the export axis. Selinexor 60-80 mg orally twice weekly with daunorubicin and cytarabine in 21 previously untreated poor-risk AML patients gave CR/CRi in 8 of 19 evaluable patients (53%), MTD not reached, median overall survival 10.3 months and 38% alive at 28.9 months' median follow-up. The relapsed/refractory older-adult monotherapy comparison and the untreated poor-risk combination are two different questions, and the axis fails one and not the other. The honest statement is that selinexor has no demonstrated benefit as monotherapy in relapsed/refractory AML, not that XPO1 inhibition has failed in AML. Sweet K et al. 2020
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the paediatric SELHEM phase 1 (NCT02212561), selinexor with fludarabine and cytarabine was tolerable to 55 mg/m2 with reversible cerebellar toxicity dose-limiting at 70 mg/m2 and asymptomatic hyponatraemia the most common grade 3 non-haematologic toxicity.

    PMID 27507877DOI
    What this evidence cannot show

    n=18, single arm. A dose-limiting CNS toxicity in a child is a different risk category from the same toxicity in an adult, and no long-term neurocognitive follow-up has been published. No infant was enrolled.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT showing the dose-limiting toxicity is a property of the COMPOUND, not of XPO1 blockade. Leptomycin B (elactocin), the prototype XPO1 inhibitor, went to a human phase 1 in which the dose-limiting toxicity across five different schedules was profound anorexia and malaise - not cerebellar toxicity and not hyponatraemia - with no responses and an explicit recommendation against further trials. Two agents against the same target, two different dose-limiting toxicities. That matters for how this risk entry generalises: reversible cerebellar toxicity in a child is a selinexor finding, and a second-generation SINE designed for reduced central-nervous-system penetration is a different risk profile, while the anorexia/malaise signature suggests the class also carries a non-CNS ceiling that the paediatric trial's 55 mg/m2 did not reach. Newlands ES et al. 1996
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Selinexor given once or twice per week disrupts the XPO1-NPM1c interaction only temporarily, which the authors give as the reason it did not translate into clinical benefit in NPM1-mutated patients.

    Prolonged XPO1 inhibition is essential for optimal antileukemic activity in NPM1-mutated AML. · Pianigiani G, Gagliardi A, Mezzasoma F, et al. · Blood Advances 2022
    PMID 36037515DOI
    What this evidence cannot show

    Abstract-level. NPM1-mutated AML; the mechanistic claim about schedule is preclinical and the clinical half summarises prior experience.

    Evidence that argues against this (1)
    • AUTHORS' OWN STATED LIMITATION, taken from the structure of the paper rather than from a caveat sentence. The result reported in this work is that ELTANEXOR, dosed more frequently, produced irreversible HOX down-regulation, terminal differentiation and prolonged survival in NPM1-mutated MOUSE models; the selinexor half - that once- or twice-weekly dosing only temporarily disrupts the XPO1-NPM1c interaction and did not translate into clinical benefit - is the authors' framing of PRIOR clinical experience, not a trial reported here, and no patient in this paper received either schedule. So the sentence this graph repeatedly cites for 'schedule explains the clinical failure' is the paper's motivation, not its result, and it is being carried at strong confidence in three separate entries. Pianigiani G et al. 2022
Connected to