Oncofusion

start / graph / drug-selinexor

Drugmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q6b-therapeutics2

Selinexor (XPO1 inhibition) — a real paediatric response rate attached to an unproven mechanism in KMT2A-rearranged disease

also called KPT-330 · Xpovio · SINE compound · selective inhibitor of nuclear export

Selinexor jams the doorway cancer cells use to push tumour-suppressor proteins out of the nucleus, trapping them where they can work. In children with relapsed leukemia it produced remissions in nearly half of those treated, but at doses that caused reversible brain toxicity.

The same thing, for a scientist

Selinexor is an oral selective inhibitor of nuclear export (SINE) that covalently binds XPO1 (exportin-1) at Cys528, retaining TP53, FOXO, IκB and eIF4E cargo in the nucleus. In the paediatric SELHEM phase 1 (NCT02212561), selinexor with fludarabine and cytarabine produced CR or CRi in 7 of 15 evaluable children (47%) with heavily pretreated relapsed/refractory acute leukemia, at a tolerable dose of up to 55 mg/m²; reversible cerebellar toxicity at 70 mg/m² was dose-limiting. Target engagement was confirmed in 15 of 16 patients. The specific case for XPO1 inhibition in KMT2A-rearranged disease rests on preclinical MYC-downregulation data and is not established.

Go deeper
The paediatric data, which are better than the field's neglect suggests

Eighteen children with relapsed/refractory acute leukemia were enrolled; 17 were evaluable for toxicity. Selinexor was given orally on days 1, 3, 8, 10, 22 and 24, escalated from 30 mg/m² through 40, 55 and 70 mg/m², with fludarabine 30 mg/m² and cytarabine 2 g/m² on days 15-19 1.

  • Dose-limiting toxicity: reversible cerebellar toxicity in two patients at 70 mg/m². Tolerable up to 55 mg/m².
  • Most common grade 3 non-haematologic toxicity: asymptomatic hyponatraemia.
  • Target engagement: ≥2-fold increase in XPO1 mRNA in 15 of 16 patients; all patients at ≥40 mg/m² showed inhibition.
  • Response: 7 of 15 evaluable (47%) achieved CR or CRi.

A 47% CR/CRi rate in heavily pretreated relapsed paediatric acute leukemia is a genuinely good number. The confounder is obvious and unavoidable: fludarabine and cytarabine alone produce remissions in this setting, and this was a single-arm dose-escalation study with no control arm. What the trial established firmly is tolerability with confirmed target engagement in children, which is the thing a paediatric phase 1 is for.

Reversible cerebellar toxicity in a child is not a footnote. In an infant whose cerebellum is still developing, a dose-limiting CNS toxicity is a different risk category than it is in an adult, and no long-term neurocognitive follow-up of these patients has been published.

Why XPO1 in KMT2A-rearranged leukemia, and how thin that case is

The mechanistic argument is MYC. KMT2A-fusion leukemia is MYC-dependent, MYC is short-lived, and XPO1 inhibition reduces MYC in several systems. The most direct published support in this genotype is preclinical: low-dose triptolide plus selinexor was synergistic in MLL-rearranged AML in vitro, ex vivo and in vivo, with MYC downregulation via Rap1/Raf/MEK/ERK and induction of mitochondrial outer membrane permeabilisation 2.

That is one paper (CORRECTED 2026-09-20: four preclinical KMT2A-intersecting XPO1 studies now exist - PMIDs 41959134, 40373471, 40340946, 35668193 - all combinations, all myeloid, none clinical), using a natural product (triptolide) that is not a clinical agent, in AML rather than infant ALL. There is no KMT2A-specific clinical signal for selinexor. The paediatric response data above are genotype-agnostic. Anyone proposing selinexor for KMT2A-rearranged infant ALL is reasoning from a mechanism, not from evidence, and should say so.

What makes it worth keeping on the list is that XPO1 inhibition is orthogonal to everything else in this beat: it is not chromatin, not apoptotic priming, not immune, and not a kinase. When menin inhibitors, venetoclax and CD19-directed therapy have all been used, the remaining options need to be mechanistically different, and there are not many.

Facts and figures
targetXPO1 / exportin-1 (covalent, Cys528); retains TP53, FOXO, IkB and eIF4E cargo in the nucleus
modalityoral small molecule (selective inhibitor of nuclear export)
statusFDA approved for multiple myeloma and DLBCL in adults; no paediatric or leukemia indication
pediatric eligibletrue
open nowfalse
pediatric tolerable dose mg m255
pediatric dltreversible cerebellar toxicity at 70 mg/m²
kmt2a specific clinical evidencenone
trials
[
  "NCT02212561"
]
key results
{
  "cr_cri_n": 7,
  "evaluable_n": 15,
  "cr_cri_pct": 47,
  "target_engagement_n": "15/16"
}
key ae
[
  "asymptomatic hyponatraemia (grade 3)",
  "reversible cerebellar toxicity",
  "anorexia and weight loss (class effect)",
  "thrombocytopenia (class effect)"
]
resistance
[
  "not characterised in KMT2A-rearranged leukemia"
]
How this could bite us back
  • Essential-function toxicity: XPO1 is the principal nuclear export receptor for hundreds of cargoes and is broadly essential; the therapeutic window is narrow by construction, not by chemistry.
  • Neurotoxicity in a developing brain: dose-limiting cerebellar toxicity in children, with no long-term neurocognitive follow-up, is a specific and unresolved late-effects risk for infants.
  • Anorexia, nausea and weight loss are the class-defining chronic toxicities of SINE compounds and are hardest to tolerate in a child whose growth is the clinical endpoint parents care about.
  • Evidence risk: the KMT2A case rests on a single preclinical paper with a non-clinical partner agent; adopting selinexor here on mechanistic grounds would repeat the reasoning error that produced the DOT1L programme.
Still open
  1. Is there any KMT2A-rearranged subgroup signal in the SELHEM cohort? Eighteen children were treated and the genotypes exist; a retrospective look is free and has never been published.
  2. Does XPO1 inhibition actually lower MYC in primary KMT2A-rearranged infant ALL cells, or only in AML lines? This is a one-week experiment that would decide whether the mechanism is real here.
  3. Would selinexor be tolerable in infants under one year, given that dose-limiting toxicity in older children was cerebellar? No infant pharmacokinetic or safety data exist.
  4. Is XPO1 inhibition synthetic-lethal with menin inhibition IN A LYMPHOID OR INFANT MODEL, and does the combination hold clinically? CORRECTED 2026-09-20: this question previously read 'the combination has not been tested in any model', which is now false. Ziftomenib plus selinexor is reported synergistic in KMT2A-rearranged and NPM1-mutant AML cell lines, primary CD34+ KMT2A-rearranged progenitors and PDX at reduced doses, with XPO1 reported to stabilise menin's binding to chromatin and its interactions with KMT2A and KMT2A/MLLT3 - i.e. the two arms may not be independent at all. PMID 41959134, DOI 10.64898/2026.03.10.710924, Uddin MH et al., bioRxiv, posted 2026-03-13: A PREPRINT, NOT PEER REVIEWED, read at abstract level, single group, all models myeloid. What remains genuinely untested is any lymphoid or infant model and any clinical combination.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the paediatric SELHEM phase 1 (NCT02212561), 18 children with relapsed/refractory acute leukemia received selinexor with fludarabine and cytarabine; selinexor was tolerable to 55 mg/m² with reversible cerebellar toxicity dose-limiting at 70 mg/m², asymptomatic hyponatraemia the most common grade 3 non-haematologic toxicity, XPO1 target inhibition confirmed in 15 of 16 patients, and CR or CRi in 7 of 15 evaluable patients (47%).

    PMID 27507877DOI
    What this evidence cannot show

    Single-arm phase 1, n=18, no control arm; fludarabine plus cytarabine alone produces remissions in this population, so the 47% CR/CRi rate cannot be attributed to selinexor. Genotype was not the selection criterion and no KMT2A-rearranged subgroup analysis is reported. Cerebellar toxicity was dose-limiting and its long-term neurocognitive consequences in children were not followed up. The promised phase 2 has not, to our knowledge, reported.

    Evidence that argues against this (1)
    • The only randomised test of selinexor in acute leukaemia was negative and numerically unfavourable. In a 2:1 randomised open-label phase 2 (NCT02088541) of selinexor 60 mg twice weekly (n=118) versus physician's choice in older patients with relapsed/refractory AML, median overall survival was 3.2 versus 5.6 months (HR 1.18, 95% CI 0.79-1.75, p=0.422), CR+CRi occurred in only a minority, and selinexor patients had more adverse events. A 47% CR/CRi rate from an uncontrolled paediatric phase 1 with a fludarabine/cytarabine backbone should be read against that result. Sweet K et al. 2021
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    Low-dose triptolide combined with selinexor was synergistic against MLL-rearranged AML in vitro, ex vivo and in vivo, acting through Rap1/Raf/MEK/ERK-mediated MYC downregulation, induction of mitochondrial outer membrane permeabilisation and DNA damage.

    PMID 40373471DOI
    What this evidence cannot show

    Preclinical only, single group, and the partner agent (triptolide) is a natural product with significant systemic toxicity and no clinical development path in leukemia. The models are AML, not infant KMT2A-rearranged ALL. The authors themselves state clinical trials are needed before any conclusion. This does not establish that selinexor monotherapy has KMT2A-specific activity.

    Evidence that argues against this (1)
    • Any selinexor combination in children has to fit inside a dose ceiling set by central nervous system toxicity, which this preclinical synergy study does not address. In the paediatric phase 1 SELHEM trial (NCT02212561, 18 enrolled, 17 evaluable for toxicity), two patients treated at 70 mg/m2 developed REVERSIBLE CEREBELLAR TOXICITY, which defined the dose-limiting toxicity; the commonest grade 3 non-haematological toxicity was asymptomatic hyponatraemia, and the tolerable dose was 55 mg/m2 with fludarabine and cytarabine. XPO1 target inhibition (a twofold or greater rise in XPO1 mRNA) was demonstrated in 15 of 16 patients at 40 mg/m2 or above, so target engagement is achievable — but a combination partner that requires higher selinexor exposure has nowhere to go. The triptolide study reports no in vivo tolerability, CNS exposure or paediatric data at all. Alexander TB et al. 2016
Connected to