Oncofusion

start / graph / hyp-meis1-pbx3-interface-is-the-unattempted-target

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q56-homeodomain

The druggable node in the homeodomain code is the MEIS1-PBX3 interface, not the HOX-PBX one - and it is the only one nobody has attacked

also called MEIS-PBX interface hypothesis · wrong-interface hypothesis

The only drug-like molecule aimed at this biology pulls HOXA9 away from its partner PBX. But the experiment that compared the combinations found that the pairing which actually reproduces the disease is PBX3 with MEIS1, not HOXA9 with PBX. If that is right, the existing molecule is aimed at the wrong handshake.

The same thing, for a scientist

Three published results converge on a claim about which protein-protein contact is load-bearing, and it is not the contact with chemical matter against it. **(1)** PBX3+MEIS1 without any ectopic HOX gene transforms murine HSPCs, causes AML with MLL-AF9-like latency, and uniquely reproduces the MLL-fusion core transcriptome; HOXA9+PBX3 does not; and disrupting the MEIS1-PBX3 binding abolishes both the transformation and the *HOX* upregulation [E1]. **(2)** The MEIS1-PBX3 interface does three separable jobs - it protects MEIS1 from ubiquitination, it is required for high-affinity ternary DNA binding, and PBX3 raises endogenous *Meis1* transcript - and Pbx3/Meis1 dimers are rate-limiting for Hoxa9-induced leukaemia [E2]. **(3)** The HOX-PBX contact is *redundant* on the HOXA9 side, and MEIS presence makes the hexapeptide globally dispensable for trimer formation [E3]. Against that, the only agents built are HXR9 and HTL001, hexapeptide mimetics aimed at HOX-PBX, with a KMT2A evidence base of exactly one paper [E4]. **The hypothesis is that the MEIS1-PBX3 contact would be the better target, and the status is that nobody has built one.**

Go deeper
Why an interface with three jobs is a better target than a transcription factor

Neither MEIS1 nor PBX3 has an enzymatic activity, a catalytic pocket or a ligand. Individually they are the textbook undruggable transcription factor - and the DepMap numbers agree that removing either one does nothing to a leukaemia cell in vitro (PBX3 +0.0418, MEIS1 -0.1791 across 57 leukaemia lines, ana-cdk6-cdk4-lineage-split-and-the-homeodomain-code-has-no-dependency).

The interface is different, because one contact carries several non-redundant functions 2:

  1. Protein stability. Without Pbx3, Meis1 is degraded through a motif coinciding with the Pbx-binding domain; Pbx3 binding prevents the ubiquitination.
  2. DNA binding. Meis1 must bind Pbx3 to form high-affinity DNA/Hoxa9/Meis1/Pbx3 complexes.
  3. Transcriptional reinforcement. Pbx3 overexpression induces endogenous Meis1.
  4. Predicted, not measured here: nuclear retention. Family-level work shows MEINOX-class binding is what keeps PBC-class proteins nuclear, by blocking export 5. If MEIS1 does for PBX3 what PREP1 does for PBX1, disrupting the contact mislocalises the partner as well as destabilising it.

An interface inhibitor would therefore not merely block a contact - it would lower MEIS1 protein. That is a degrader-like outcome from an occupancy-based agent, and it is the reason this is worth a hypothesis rather than a sentence.

What exists, and what does not

Exists. MEIS inhibitors - reported as reducing MEIS1 transcriptional activity and LSC survival in primary leukaemia cells and inducing apoptosis 6. Real chemical matter against MEIS1, and a fact this hypothesis had to discover rather than assume: a first pass would have written no chemical matter exists, which would have been wrong.

Does not exist, as far as this beat could establish. An agent described as targeting the MEIS1-PBX3 dimerisation interface. The MEIS inhibitors in 6 are described by their effect on MEIS1 transcriptional activity, not by an interface mechanism, and nothing read this cycle tests any of them in KMT2A-rearranged disease.

Bounded, with the query and the controls. On 2026-09-20: (MEIS1[tiab] OR MEIS[tiab]) AND (PBX[tiab] OR PBX3[tiab] OR PBX1[tiab]) AND (inhibitor OR inhibitors OR inhibition OR antagonist OR antagonists OR peptide OR "small molecule")[tiab] returns 29 records, every title read; one is about MEIS inhibitors 6 and none describes a MEIS-PBX interface agent. The interface-vocabulary variant (interface OR dimerization OR dimerisation OR "protein-protein") returns 10. Operand controls: MEIS1[tiab] = 895, PBX3[tiab] = 262, MEIS1[tiab] AND PBX3[tiab] = 30. So the instrument returns plenty and the specific thing is not in it - which is the shape a bounded absence should have, and it survived precisely because every one of the 29 titles was read rather than counted.

The honest case against this hypothesis

Written out because the case for it is easy to make and the graph has published on thinner.

  1. The whole argument is murine and myeloid. 1 is murine HSPC overexpression with an AML readout; 2 is murine Hoxa9-transformed cells. The index disease is infant B-lineage KMT2A::AFF1 and nothing here has been shown in it.
  2. Sufficiency is not necessity. PBX3/MEIS1 transforming without an ectopic HOX gene does not show HOX is dispensable - the construct turns the endogenous Hoxa cluster on, so the cells have Hoxa9.
  3. The human data point the other way. HOXA9 knockdown kills human MLL-rearranged cells and co-down-regulates PBX3, putting PBX3 downstream [E4 counter-evidence]. A feed-forward loop reconciles this, and a loop has no privileged node.
  4. Normal tissue. MEIS1 and PBX3 are required for normal haematopoiesis and development. An interface inhibitor does not distinguish the leukaemic dimer from the physiological one, and the patients are infants.
  5. No assay. Neither gene is a dependency in any leukaemia line, so there is no in vitro potency readout to optimise against - which is a practical reason this target has not been attempted, and not only an oversight.
Facts and figures
statementIn KMT2A-rearranged leukaemia the load-bearing protein-protein contact in the homeodomain cofactor code is MEIS1-PBX3, not HOX-PBX. An agent disrupting MEIS1-PBX3 should lower MEIS1 protein, abolish high-affinity ternary DNA binding and collapse the HOXA programme, whereas an agent disrupting HOX-PBX competes at a contact that is redundant on the HOXA9 side and dispensable when MEIS is present.
originliterature-gap
prior probability0.25
no structureNo structure of a MEIS1-PBX3 complex was located, and no systematic RCSB search was run - recorded as unchecked, not as absent. PDB 1PUF covers the HOX-PBX contact only, and with mouse HoxA9 and human PBX1 homeodomain fragments.
predictions
[
  "a MEIS1-PBX3 interface-disrupting reagent lowers MEIS1 PROTEIN (not only activity) in KMT2A-rearranged cells, measurable by cycloheximide chase",
  "PBX3 knockdown in a human KMT2A::AFF1 line reduces MEIS1 protein half-life, reproducing the murine result",
  "the HOX-PBX peptide HXR9 does NOT lower MEIS1 protein, distinguishing the two interfaces by a single western blot",
  "disrupting MEIS1-PBX3 collapses the HOXA cluster transcript, while disrupting HOX-PBX does not",
  "PBX3 is partly cytoplasmic in cells where MEIS1 is low, if the family-level nuclear-retention mechanism holds for this pair"
]
experiments
[]
analyses
[
  "ana-cdk6-cdk4-lineage-split-and-the-homeodomain-code-has-no-dependency"
]
falsifiers
[
  "PBX3 knockdown does not change MEIS1 protein half-life in a human KMT2A-rearranged cell - kills the stability mechanism in the system that matters",
  "HXR9 lowers MEIS1 protein as much as a MEIS-PBX disruptor would - the two interfaces are not separable and the distinction this hypothesis rests on is void",
  "the MEIS1-PBX3 contact turns out to have no discrete pocket or hot spot on inspection of any structure - the target is real and undruggable, which is a different and equally useful answer",
  "PBX3/MEIS1 fails to transform or fails to reproduce the transcriptome in a B-lineage system - the founding result does not transfer"
]
chemical matter status
{
  "against_HOX_PBX": "HXR9, HTL001 - exist; one KMT2A paper",
  "against_MEIS1_activity": "MEISi compounds (PMID 37902585) - exist, no identifiers or potencies read, not interface-directed, never tested in KMT2A-rearranged disease",
  "against_the_MEIS1_PBX3_interface": "none located; bounded by the queries recorded in this entry's sections, with all 29 titles of the vocabulary query read"
}
How this could bite us back
  • This hypothesis proposes a target with no agent, no structure located, and no in vitro potency assay, in a disease where neither constituent gene is a dependency. That is three missing pieces, and the honest reading is that it is a research direction and not a programme.
  • MEIS1 and PBX3 run normal haematopoiesis and normal development. An interface inhibitor cannot tell the leukaemic dimer from the physiological one, and the patients are infants - the same therapeutic-index problem that has defeated every self-renewal target in this disease.
  • If the mechanism is protein stabilisation, the agent works at the rate MEIS1 turns over, not at the rate of occupancy. Pharmacodynamic lag is a genuine failure mode in a disease that relapses in weeks.
  • The entire argument is murine and myeloid, and the one human dataset in it points the other way. If PBX3 is downstream of HOXA9 in B-lineage cells, a menin inhibitor already reaches this node more cheaply and this hypothesis is redundant with an approved drug.
  • A hypothesis that says the field is aiming at the wrong interface is exactly the kind of claim that gets quoted without its caveats. The caveat that must travel with it: HXR9 works in cells and this proposal does not exist yet.
Still open
  1. What reagent did Garcia-Cuellar et al. use to block Meis1/Pbx3 dimerisation? The abstract does not say. If it is a peptide or a dominant-negative it is the starting point for this whole programme; if it is a point mutant, then nothing about interface druggability has been demonstrated and the hypothesis is one step further from actionable than it reads.
  2. Is there a structure of any MEIS-PBX complex to inspect for a pocket? No systematic RCSB search was run this cycle. This is a one-hour lookup and it decides whether the hypothesis is a chemistry project or a dead end.
  3. What are the MEISi compounds in PMID 37902585, and do any of them act at the PBX interface? No identifiers were read. If one does, this hypothesis is already half-tested by someone else.
  4. Does PBX3 knockdown lower MEIS1 protein in a human KMT2A::AFF1 line? A cycloheximide chase in SEM with and without PBX3 knockdown costs a week and is the cheapest test of the central mechanism.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In murine haematopoietic stem/progenitor cells, PBX3 plus MEIS1 without an ectopic HOX gene transforms cells in vitro, causes AML in vivo with latency similar to MLL-AF9, and uniquely among the tested combinations recapitulates the MLL-fusion core transcriptome including endogenous Hoxa upregulation; HOXA9/MEIS1, HOXA9/PBX3 and HOXA9 alone do not; and disrupting the binding between MEIS1 and PBX3 diminishes both transformation and HOX gene upregulation.

    PBX3 and MEIS1 Cooperate in Hematopoietic Cells to Drive Acute Myeloid Leukemias Characterized by a Core Transcriptome of the MLL-Rearranged Disease · Li Z, Chen P, Su R, Hu C, Li Y, Elkahloun AG, Zuo Z, Gurbuxani S, Arnovitz S, Weng H, Wang Y, Li S, Huang H, Neilly MB, Wang GG, Jiang X, Liu PP, Jin J, Chen J · Cancer Research 76:619-629 2016
    PMID 26747896DOI
    What this evidence cannot show

    Abstract read this cycle. Murine HSPC retroviral overexpression, myeloid AML readout, not infant B-lineage KMT2A::AFF1 and not human. Sufficiency without an ectopic HOX gene is not dispensability of HOX. The authors themselves close by proposing that targeting the MEIS1/PBX3 interaction may be a promising therapeutic strategy, so the direction of this hypothesis is theirs and what is added here is the comparison with the interface that actually has chemical matter.

    Evidence that argues against this (1)
    • In human MLL-rearranged leukaemia HOXA9 suppression kills MLL-rearranged but not MLL-germline cells and co-down-regulates PBX3, placing PBX3 downstream of HOXA9 - which would make the MEIS-PBX dimer an effector rather than the apex node. Faber J et al. 2009
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In murine Hoxa9-transformed cells, Meis1 protein is highly unstable without Pbx3, with degradation contingent on a motif coinciding with the Pbx-binding domain, and either deleting that motif or binding Pbx3 prolongs Meis1's half-life by preventing ubiquitination; Pbx3 overexpression induces endogenous Meis1 transcription; Meis1 must bind Pbx3 to form high-affinity DNA/Hoxa9/Meis1/Pbx3 complexes in vitro; blocking Meis1/Pbx3 dimerisation abrogates proliferation, colony formation and expression of the Meis1 targets Flt3 and Trib2; and Pbx3/Meis1 dimers are rate-limiting for Hoxa9-induced leukaemia.

    Pbx3 and Meis1 cooperate through multiple mechanisms to support Hox-induced murine leukemia · Garcia-Cuellar MP, Steger J, Fuller E, Hetzner K, Slany RK · Haematologica 100:905-913 2015
    PMID 25911551DOI
    What this evidence cannot show

    Abstract read this cycle. Murine Hoxa9-driven leukaemia - the fusion's output rather than the fusion - with a myeloid readout. The reagent used to block Meis1/Pbx3 dimerisation is not identified in the abstract, so whether a tractable tool compound or only a point mutant exists is unknown from what was read; that matters, because if it is a mutant then 'the interface is druggable' has not been demonstrated at all. Surname Fuller rendered without its umlaut.

    Evidence that argues against this (1)
    • If Pbx3's principal job is stabilising Meis1, an interface inhibitor works at the rate of MEIS1 protein turnover rather than at the rate of occupancy - a pharmacodynamic lag that is a real failure mode in a disease that relapses in weeks. Li Z et al. 2016
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The human HOXA9-TALE interaction relies on the redundant activity of the hexapeptide motif and two paralog-specific residues of the HOXA9 homeodomain, and the additional presence of MEIS remodels the HOX-PBX interaction so that the hexapeptide becomes globally dispensable for trimeric complex formation in the large majority of HOX proteins.

    The human HOXA9 protein uses paralog-specific residues of the homeodomain to interact with TALE-class cofactors · Dard A, Jia Y, Reboulet J, Bleicher F, Lavau C, Merabet S · Scientific Reports 9:5664 2019
    PMID 30952900DOI
    What this evidence cannot show

    Abstract read this cycle. Human proteins but in heterologous cell lines and in vitro DNA-binding assays, not leukaemia cells. It is a statement about complex FORMATION, not about transcriptional output, and it is an argument that the competing design's premise is incomplete rather than a demonstration that the competing design fails.

    Evidence that argues against this (1)
    • Hexapeptide-based peptides do kill leukaemic cells overexpressing HOXA and PBX3, so the HOX-PBX route is evidently sufficient to produce a cellular effect regardless of the binding redundancy - and a working agent at a theoretically imperfect site beats a perfect site with no agent. Li Z et al. 2013
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The only chemical matter aimed at the HOX-PBX interface is the hexapeptide-based peptide series HXR9 and its successor HTL001, and their published evidence base in MLL/KMT2A leukaemia is a single record: (HXR9 OR HTL001) AND (MLL OR KMT2A) returns 1 PubMed record, PMID 23264595, where the peptide arm is explicitly framed by the authors as proof of concept; positive controls through comparable query shapes return 71 for "HOX/PBX"[tiab] and 17 for hexapeptide[tiab] AND PBX[tiab].

    What this evidence cannot show

    Bounded by the name list HXR9, HTL001, CXR9, HOX/PBX, hexapeptide. Any compound code for this series not seen by this beat would not have been searched - the residual risk named by brief rule 72. The single record was title-checked.

    Evidence that argues against this (1)
    • A small literature is not the same as a weak agent: HTL001 is described by its developers as clinical-stage, so development may have continued in a patent and regulatory track that a PubMed count cannot see. Morgan R et al. 2025
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    HTH and PREP1 are cytoplasmic and require association with EXD/PBX for nuclear localisation; EXD and PBX1 carry both nuclear import and export signals and are actively exported in some cell contexts; and co-expression of EXD/PBX with HTH/PREP1 produces nuclear localisation of the heterodimers, proposed to work by blocking nuclear export.

    The subcellular localization of PBX1 and EXD proteins depends on nuclear import and export signals and is modulated by association with PREP1 and HTH · Berthelsen J, Kilstrup-Nielsen C, Blasi F, Mavilio F, Zappavigna V · Genes & Development 13:946-953 1999
    PMID 10215622DOI
    What this evidence cannot show

    Abstract read this cycle. PBX1 and Drosophila EXD with PREP1 and HTH, in murine fibroblasts and Schneider cells - NOT PBX3, NOT MEIS1, NOT a haematopoietic cell, and not leukaemia. It supports a PREDICTION about the MEIS1-PBX3 pair and is explicitly labelled as such in this entry's sections; it is not evidence about that pair.

    Evidence that argues against this (1)
    • PBX localisation has at least one partner-independent determinant - non-muscle myosin promotes cytoplasmic localisation of PBX - so disrupting a MEINOX partner may not mislocalise PBX3 at all. Huang H et al. 2003
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    MEIS inhibitors (MEISi) exist and reduce MEIS1 transcriptional activity and leukaemic stem cell survival by apoptosis in primary leukaemia cells; MEIS1 expression is higher in leukaemia patients and in relapsed samples, and the majority of CD123+ and CD34+ LSCs show higher MEIS1/2/3 content.

    MEIS inhibitors reduce the viability of primary leukemia cells and Stem cells by inducing apoptosis · Meric N, Albayrak E, Gulbas Z, Kocabas F · Leukemia & Lymphoma 65:187-198 2024
    PMID 37902585DOI
    What this evidence cannot show

    Abstract read this cycle. The compounds are described only as 'recently developed MEIS inhibitors (MEISi)' with NO compound identifiers, structures, potencies or selectivity data in the abstract, so nothing about their tractability is established here and the confidence is moderate for that reason. They are described by effect on MEIS1 TRANSCRIPTIONAL ACTIVITY, not as MEIS-PBX interface agents - which is why they do not close this hypothesis's gap. Primary leukaemia and LSC samples, NOT selected for KMT2A rearrangement; author surnames rendered without Turkish diacritics (PubMed gives Meric N, Gulbas Z, Kocabas F with them).

    Evidence that argues against this (1)
    • That these exist at all is a correction to the first version of this hypothesis, which was going to record 'no chemical matter against MEIS1' - the claim survived only because the 29 titles returned by the vocabulary query were read rather than counted, which is brief rule 71 working. NCBI PubMed; queries run by this project 2026
Connected to