Oncofusion

start / graph / gene-pbx3

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q56-homeodomain

PBX3 - the TALE cofactor the field says is non-redundant, and the experiment that says HOXA9 is the redundant one

also called Pre-B-cell leukemia transcription factor 3 · homeobox protein PBX3 · PBC-class TALE homeodomain protein

HOXA9 is usually described as what KMT2A-fusion leukemia makes, but HOXA9 binds DNA weakly and needs partners. PBX3 is one of those partners, and in the experiment designed to ask which part matters, PBX3 plus a second partner caused leukemia in mice without any HOXA9 at all - so the partner may be the active ingredient rather than the accessory.

The same thing, for a scientist

PBX3 is a PBC-class TALE homeodomain protein and, of the three PBX paralogues, the one consistently co-expressed with HOXA9 in cytogenetically abnormal AML including MLL-rearranged AML [E1]. Depleting Pbx3 inhibits MLL-fusion-mediated transformation, co-expressed PBX3 synergises with HOXA9 in vitro and in vivo, and CRISPR deletion of Pbx3 in established murine MLL-AF9 AML prolongs survival and depletes LSCs [E1][E3]. The headline result is stronger and more awkward than the cofactor framing: **co-expression of PBX3 and MEIS1 with NO ectopic HOX gene transforms murine HSPCs and causes AML with a latency similar to MLL-AF9, and recapitulates the MLL-fusion core transcriptome - which HOXA9/MEIS1, HOXA9/PBX3 and HOXA9 alone do not** [E2]. Independently, Pbx3 is required to stabilise Meis1 protein against ubiquitination, and Pbx3/Meis1 dimers are rate-limiting for Hoxa9-induced leukaemia [E4]. Against all of this: PBX3 is not a dependency in any human leukaemia line in DepMap - mean Chronos effect **+0.0418** across 57 leukaemia lines and +0.0297 across 1,258 lines, with 0.08% of lines below -0.5 [E5]. Every positive result above is murine, most of it myeloid.

What it looks like
loading AF-P40426-F1…
AlphaFold AF-P40426-F1 ↗
colourshape

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
The answer to the headline question, and it is not the one the brief expected

The question this beat was sent to settle was MEIS1 or PBX3 - which cofactor carries the leukaemogenic activity? The best-designed published experiment does not answer in either name. It answers both, as a dimer, and not HOXA9.

Li et al. 2016 2 asked the combinatorial question directly in murine haematopoietic stem/progenitor cells and by transplantation:

constructtransforms HSPCcauses AML in vivorecapitulates the MLL-fusion core transcriptome
HOXA9 alonenonono
HOXA9 + MEIS1yes (known)yes, rapidno
HOXA9 + PBX3--no
PBX3 + MEIS1, no ectopic HOXyesyes, latency similar to MLL-AF9yes, including upregulation of the endogenous Hoxa genes

Disrupting the MEIS1-PBX3 binding diminished both the transformation and the HOX upregulation.

Read carefully, this inverts the shorthand. The pairing the field writes as HOXA9/MEIS1 is not the pairing that reproduces the disease's transcriptome; PBX3/MEIS1 is, and it does so by turning the endogenous HOXA cluster on from underneath. HOXA9 in that experiment is downstream, not upstream.

Bounding it honestly. This is one paper, in mouse HSPCs, on a myeloid readout, from the group that had previously argued for PBX3 1. It has not been replicated in a B-lineage or human system in anything read this cycle, and a construct that works without HOXA9 does not show HOXA9 is dispensable in a cell that already has it. What it does show is that the HOXA9-MEIS1 shorthand is an assumption the field has not tested against its alternatives, and one group that did test it got a different answer.

The mechanism that makes PBX3-before-MEIS1 more than a correlation

An independent group, working in murine Hoxa9-transformed cells, found the wiring that would explain it 4:

  • In the absence of Pbx3, Meis1 protein is highly unstable. Degradation depends on a motif coinciding with the Pbx-binding domain; deleting that motif, or binding Pbx3, prolongs Meis1's half-life by preventing ubiquitination, and proteasome inhibition also blocks the breakdown.
  • Pbx3 overexpression induces endogenous Meis1 transcription, so the two cooperate genetically as well as physically.
  • Blocking Meis1/Pbx3 dimerisation abrogated proliferation and colony formation in primary cells transformed by Hoxa9, and the Meis1 targets Flt3 and Trib2 depended on the dimer.
  • Meis1 had to bind Pbx3 to form high-affinity DNA/Hoxa9/Meis1/Pbx3 complexes in vitro, and Pbx3/Meis1 dimers are rate-limiting for Hoxa9-induced leukaemia.

So PBX3 is not an alternative to MEIS1 - it is MEIS1's stabiliser. A graph entry that lists them as two parallel cofactors of HOXA9 has the topology wrong. The order is: PBX3 holds MEIS1 up, the PBX3-MEIS1 dimer is the rate-limiting unit, and HOXA9 joins a complex that is already assembled.

This also explains why targeting MEIS1 protein directly has been so unrewarding: in a Pbx3-sufficient cell, MEIS1 is a protected protein.

What PBX3 does in this disease, stated with its system
  • In cytogenetically abnormal AML including MLL-rearranged AML, PBX3 - but not PBX1 or PBX2 - is consistently co-expressed with HOXA9, and a four-gene PBX3/HOXA7/HOXA9/HOXA11 signature independently predicts poor survival 1.
  • In murine MLL-fusion transformation assays, shRNA depletion of endogenous Pbx3 significantly inhibits transformation, and co-expressed PBX3 synergises with HOXA9 in vitro and in leukaemogenesis in vivo 1.
  • In an established murine MLL-AF9 AML, CRISPR/Cas9 deletion of Pbx3 significantly prolongs survival and reduces leukaemia burden by decreasing LSC capacity and promoting LSC apoptosis; the Pbx3 locus in those LSCs carries increased H3K79me2 and decreased H3K9me3 and H3K27me3 3.
  • In human MLL-rearranged leukaemia, HOXA9 suppression co-down-regulates a programme containing HOXA10, MEIS1, PBX3 and MEF2C, and induces apoptosis specifically in MLL-rearranged cells; the in vivo arm used the t(4;11) SEMK2 line, which is the index fusion and B-lineage 6.

That last item is the only B-lineage KMT2A::AFF1 evidence touching PBX3 found this cycle, and it places PBX3 downstream of HOXA9 rather than upstream - the opposite direction from 2. Both directions are published and this graph should hold both, because the resolution is probably a feed-forward loop rather than one of them being wrong.

The dependency data disagree, and the disagreement is informative

Fetched live from DepMap on 2026-09-20 5:

genemean effect, 1,258 linesmean effect, 57 leukaemia linesfraction of lines at or below -0.5
PBX3+0.0297+0.04180.08%
PBX1+0.0110+0.03910.48%
PBX2-0.2524-0.34646.3%
HOXA9-0.1727-0.22303.8%
MEIS1-0.0528-0.17911.3%

PBX3 knockout has no fitness cost in any leukaemia line in DepMap. One line in 1,258 falls below -0.5. The paralogue with the deepest effect is PBX2, which is the one the literature says is not the relevant partner.

Four readings, in decreasing order of how much they rescue the literature:

  1. The published PBX3 phenotypes are LSC and in vivo phenotypes. 3 measures survival of transplanted mice and LSC frequency; 1 measures transformation. A 21-day CRISPR fitness screen in an immortalised line cannot see self-renewal, and this is the standard, legitimate explanation.
  2. Paralogue buffering. PBX1, PBX2 and PBX3 share a DNA-binding mode; a single-gene knockout in a cell expressing all three is buffered by construction, and the near-zero PBX1 and PBX3 scores against PBX2's -0.35 is the signature of a partially redundant family rather than of three dispensable genes.
  3. The lines are wrong. No infant-derived line exists in DepMap, and the KMT2A-r arm is myeloid-enriched.
  4. The literature is over-read. This is the reading the graph should not discard: every positive PBX3 result is murine, and the human evidence is co-expression and prognosis, which are associations.

The honest position is that PBX3's requirement in this disease is established for murine LSC maintenance and unestablished for human leukaemic cell survival, and the DepMap zero is a genuine bound on the second.

Cell biology: where PBX3 is, and what controls it

PBX nuclear localisation is not constitutive. PBX1 and EXD carry both nuclear import and nuclear export signals, and their partners HTH and PREP1 are themselves cytoplasmic and require association with EXD/PBX to enter the nucleus; co-expression of EXD/PBX with HTH/PREP drives nuclear localisation of the heterodimers in cell contexts where PBX alone is exported 7. The proposed mechanism is that heterodimerisation blocks nuclear export rather than adding an import signal.

The consequence for this beat is direct: a MEINOX-class partner is what keeps a PBC-class protein in the nucleus. If MEIS1 does for PBX3 what PREP1 does for PBX1, then the PBX3-MEIS1 dimer is simultaneously the DNA-binding unit 4, the protein-stability unit 4 and the nuclear-retention unit 7 - three separable functions of one interface, which is a much better target profile than any of them alone.

This is asserted at the level of the family, not of PBX3. 7 works on PBX1/EXD with PREP1/HTH in fibroblasts and Drosophila Schneider cells. No source read this cycle measures PBX3 subcellular distribution in a KMT2A-rearranged cell, of any lineage, in any species. That is a specific, cheap gap and it is recorded as one.

Facts and figures
uniprotP40426
gene symbolPBX3
length AA434
alphafold IDAF-P40426-F1
ensemblENSG00000167081
cytoband9q33.3
protein classPBC-class TALE (three-amino-acid-loop-extension) homeodomain transcription factor
resolution notegene_exact:PBX3 AND organism_id:9606 AND reviewed:true returned ONE reviewed row, primary gene name PBX3, entry PBX3_HUMAN, 434 aa, NO gene synonyms, fetched from the UniProt REST API on 2026-09-20 with fields=accession,id,gene_primary,gene_synonym,protein_name,length.
paralogue discriminationPBX1 P40424 (430 aa), PBX2 (not resolved this cycle), PBX3 P40426 (434 aa). PBX1 and PBX3 are FOUR RESIDUES APART and their accessions differ by two digits (P40424 vs P40426). That is tighter than the HLA-A/HLA-B gap of three residues only by one, and far tighter than brief rule 66's SMC1A/SMC3 limit of 16 - so verify-accessions.py cannot catch a PBX1-for-PBX3 swap. The literature distinction between them is the whole point of E1, so a swap here inverts the finding. Discriminate on gene_primary only.
domains noteNo domain boundaries asserted. The TALE homeodomain and the PBC-A/PBC-B regions were not read back from UniProt with residue numbers this cycle; a guessed boundary would be invisible to the 3D viewer under the residue-integer rule and wrong under the citation rule.
cofactor topology verdictPBX3 is NOT a parallel alternative to MEIS1. It is MEIS1's stabiliser and obligate dimer partner: Meis1 protein is degraded without Pbx3, Pbx3 induces endogenous Meis1 transcription, and the dimer is the rate-limiting unit for Hoxa9-induced murine leukaemia. The PBX3/MEIS1 dimer transforms murine HSPCs and reproduces the MLL-fusion core transcriptome WITHOUT an ectopic HOX gene, which no HOXA9-containing pair does.
cell context evidence classexpressed_in (high in AML, prognostic) + required_in (LSC of MLL-AF9 AML) + maintains_a_state (LSC capacity; and the interferon-suppressed, differentiation-blocked state). All three classes are present for this gene, which is unusual in the tranche, and all three are murine or myeloid.
cell context shown for fusionYES for MLL-AF9, NO for the index fusion and the index lineage. Both load-bearing studies are murine MLL-AF9, i.e. KMT2A::MLLT3, in myeloid disease. There is no KMT2A::AFF1 arm, no human B-lineage arm and no infant material in either. Given that the founding result in this area is specifically that PBX3 and not PBX1 is the rate-limiting cofactor, and that the graph's own registry records PBX3 and PBX1 as four residues apart with accessions differing by one digit, the partner-and-lineage restriction has to travel with the claim rather than sit only in a caveat.
cell context depmapMarginal, on a near-zero baseline. PBX3 is -0.044 in the 11 KMT2A-rearranged lines against +0.061 in 43 other leukaemia lines, difference -0.105, rank 1,022 of 17,728 (5.76th percentile), on a pan-line mean of +0.030. So PBX3 is preferentially scored in the index arm, but neither arm shows a meaningful absolute dependency - a small difference between two numbers that are both about zero. Cell-line viability is also the wrong readout for an LSC-capacity phenotype, which is measured by transplantation.
domains
[]
essentiality
{
  "source": "DepMap CRISPR (Chronos) via Open Targets GraphQL, retrieved live 2026-09-20",
  "n_all_lines": 1258,
  "chronos_mean_all_lines": 0.0297,
  "chronos_mean_leukaemia_lines": 0.0418,
  "n_leukaemia_lines": 57,
  "frac_lines_at_or_below_minus_0_5": 0.0008,
  "frac_lines_at_or_below_minus_1_0": 0,
  "kmt2a_r_selectivity_rank": 1022,
  "kmt2a_r_selectivity_of": 17728,
  "kmt2a_r_difference": -0.1052,
  "interpretation": "no detectable fitness requirement in any human leukaemia line; every published positive is murine and in vivo or LSC-scoped"
}
subcellular localization
{
  "compartment": "nucleus, conditionally",
  "mechanism": "PBC-class proteins carry both NLS and NES; MEINOX-class heterodimerisation blocks nuclear export and drives nuclear retention",
  "evidence_level": "family-level, from PBX1/EXD with PREP1/HTH",
  "measured_for_PBX3_in_KMT2A_r": "NO - no source read this cycle measures PBX3 subcellular distribution in a KMT2A-rearranged cell of any lineage or species",
  "second_determinant": "non-muscle myosin promotes cytoplasmic localisation of PBX (PMID 12724421)"
}
cell type
{
  "required_in": "Leukaemia stem cells of MLL-rearranged AML, in mouse. CRISPR/Cas9 deletion of Pbx3 in leukaemic cells of the MLL-AF9-induced AML mouse model significantly prolonged survival and decreased leukaemia burden by decreasing the capacity of LSCs and promoting LSC apoptosis. In the same work, ChIP-seq/qPCR in MLL-rearranged mouse models found the Pbx3 locus carrying increased H3K79me2 and decreased H3K9me3 and H3K27me3 in LSCs, which is offered as the reason Pbx3 is highly expressed there.",
  "expressed_in": "High PBX3 expression is seen in clinical AML samples and correlates with poor prognosis; the authors note differential expression between normal and malignant haematopoietic cells.",
  "how_shown": "CRISPR/Cas9 deletion in an MLL-AF9 mouse AML model with survival, burden and LSC functional readouts; ChIP-seq and ChIP-qPCR for histone marks; clinical database expression analysis",
  "systems": "murine MLL-AF9 AML plus human AML expression data",
  "nodes": []
}
cell state
{
  "maintains_state": "The leukaemia stem cell state, stated as such by the source - Pbx3 deletion decreases LSC capacity and promotes LSC apoptosis. Separately, Pbx3 maintains an immune-cold state: leukaemia progression in the MLL-AF9 model is accompanied by downregulation of type I interferon response genes, and that downregulation depends on upregulation of the MLL-AF9 transcriptional target Pbx3 IN THE PROGENITOR CELL COMPARTMENT; enhancing interferon signalling with IFN-alpha induces leukaemic cell DIFFERENTIATION, and inhibiting MLL-AF9 transcriptional activity on top of enhanced IFN signalling further delays progression.",
  "reading": "Both readouts are identity readouts rather than growth readouts - LSC capacity in one, differentiation in the other - which makes PBX3 one of the better-supported maintains-a-state calls in this tranche, in mouse and in the myeloid lineage.",
  "nodes": [
    "cell-state-stem-like-self-renewal",
    "cell-state-differentiation-block"
  ]
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.0439,
  "chronos_mean_other_leukaemia": 0.0612,
  "difference": -0.1052,
  "pan_essentiality_all_lines": 0.0299,
  "selectivity_rank": 1022,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • The intervention implied here is disrupting a protein-protein interface between two transcription factors with no enzymatic activity and no pocket. The only chemical matter that exists is a cell-penetrating peptide (see target-hox-pbx-interface-peptide-hxr9) and it targets the HOX-PBX interface, which is NOT the interface [E2] and [E4] identify as load-bearing.
  • PBX3 knockout is fitness-neutral in every human leukaemia line in DepMap [E5]. If the published requirement is genuinely LSC-scoped and in vivo, then a PBX3-directed agent would have no in vitro readout to optimise against - which is a real drug-development problem, not only an evidential one.
  • PBX1, PBX2 and PBX3 share a DNA-binding mode and PBX2 is the paralogue with the deepest DepMap effect. A PBX3-selective agent invites paralogue compensation; a pan-PBX agent hits PBX1, which is the fusion partner in TCF3::PBX1 B-ALL and is required in normal development.
  • MEIS1 and PBX3 are both required for normal haematopoietic stem cell function. Lowering the dimer in an infant lowers it in the marrow that has to regenerate, and no source read this cycle separates the leukaemic from the normal requirement quantitatively.
  • If PBX3 acts mainly by stabilising MEIS1, then a successful PBX3 intervention is a slow one - it works at the rate MEIS1 protein turns over, not at the rate of occupancy. In a disease that relapses in weeks, pharmacodynamic lag is a real failure mode.
Still open
  1. Does the PBX3/MEIS1-without-HOXA9 result reproduce in a B-lineage or human system? It is the single most consequential unreplicated finding in this beat: if it holds, the field's HOXA9/MEIS1 shorthand is wrong about which node is upstream, and the drug target moves from the HOX-PBX interface to the MEIS-PBX interface.
  2. Where is PBX3 in a KMT2A::AFF1 blast - nuclear, cytoplasmic, or partitioned? Nobody has looked, in any lineage. A fractionation blot plus immunofluorescence in SEM and RS4;11 is a week of work and would test whether MEIS1 is doing for PBX3 what PREP1 does for PBX1.
  3. Is MEIS1 protein stabilised by PBX3 in human KMT2A-rearranged cells as it is in murine Hoxa9-transformed cells? A cycloheximide chase with and without PBX3 knockdown answers it, and it determines whether lowering PBX3 is a way to lower MEIS1 protein without touching MEIS1.
  4. Is the DepMap zero real or buffered? A triple PBX1/PBX2/PBX3 knockout in a KMT2A::AFF1 line would distinguish redundancy from dispensability, and the family has never been tested combinatorially in a human leukaemia line.
  5. Every PBX3 cell-state result is murine MLL-AF9 myeloid disease. PBX3 is also expressed in KMT2A-rearranged B-ALL, where HOXA9 and MEIS1 are the same central outputs. Does Pbx3 deletion reduce leukaemia-initiating capacity in a KMT2A::AFF1 B-lineage model, or is the PBX3 requirement a myeloid-lineage property that this graph is about to generalise across a lineage boundary for the second time?
  6. IFN-alpha induces differentiation of MLL-AF9 leukaemic cells and PBX3 suppresses the interferon response. Menin inhibition independently induces an interferon/MHC-II state in this disease. Do PBX3 loss and menin inhibition converge on the same interferon programme, and does PBX3 level predict the magnitude of the menin-inhibitor-induced interferon response?
Every source behind this entry (9)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In cytogenetically abnormal AML including MLL-rearranged AML, PBX3 - but not PBX1 or PBX2 - is consistently co-expressed with HOXA9; shRNA depletion of endogenous Pbx3 significantly inhibits MLL-fusion-mediated cell transformation in murine cells; co-expressed PBX3 synergises with HOXA9 in promoting transformation in vitro and leukaemogenesis in vivo; and the HOX/PBX-disrupting peptide HXR9 selectively kills leukaemic cells overexpressing HOXA/PBX3 genes.

    PBX3 is an important cofactor of HOXA9 in leukemogenesis · Li Z, Zhang Z, Li Y, Arnovitz S, Chen P, Huang H, Jiang X, Hong GM, et al. · Blood 121:1422-1431 2013
    PMID 23264595DOI
    What this evidence cannot show

    Abstract read this cycle; full text not retrieved. The transformation and in vivo arms are MURINE and the readout is myeloid; the human component is co-expression in patient AML, which is association. The HXR9 arm is explicitly framed by the authors as proof of concept. PBX1/PBX2 are excluded on co-expression, not on a functional test.

    Evidence that argues against this (1)
    • PBX3 knockout has essentially no fitness cost in human leukaemia lines - live DepMap mean +0.0418 across 57 leukaemia lines, with 1 of 1,258 lines below -0.5 - so the 'important cofactor' claim does not extend to survival of established human leukaemic cells in vitro. Broad Institute DepMap; Open Targets Platform 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In murine haematopoietic stem/progenitor cells, co-expression of PBX3 and MEIS1 without ectopic expression of any HOX gene is sufficient for transformation in vitro and causes AML in vivo with a leukaemic latency similar to MLL-AF9; gene-expression profiling shows that PBX3/MEIS1 overexpression - but NOT HOXA9/MEIS1, HOXA9/PBX3 or HOXA9 alone - recapitulates the MLL-fusion-mediated core transcriptome including upregulation of the endogenous Hoxa genes; and disrupting the MEIS1-PBX3 binding diminishes both the transformation and the HOX 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, et al. · Cancer Research 76:619-629 2016
    PMID 26747896DOI
    What this evidence cannot show

    Abstract read this cycle; full text not retrieved. MURINE HSPCs and a MYELOID leukaemia readout - not infant B-lineage KMT2A::AFF1, and not human. It is an overexpression experiment: showing that PBX3/MEIS1 is SUFFICIENT without ectopic HOXA9 is not showing that HOXA9 is dispensable in a cell that expresses it, and the construct upregulates the endogenous Hoxa cluster, so endogenous Hoxa9 is present in the transformed cells. Same senior group as E1, so this is not independent confirmation of the PBX3 position. No replication in a B-lineage or human system was found this cycle.

    Evidence that argues against this (1)
    • In human MLL-rearranged leukaemia cells, HOXA9 suppression kills MLL-rearranged but not MLL-germline cells and co-down-regulates PBX3 among other genes - which places PBX3 downstream of HOXA9, the opposite direction to this claim. Faber J et al. 2009
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a murine MLL-AF9-induced AML model, CRISPR/Cas9 deletion of Pbx3 in established leukaemic cells significantly prolonged survival and decreased leukaemia burden by reducing leukaemic stem cell capacity and promoting LSC apoptosis; ChIP-Seq/qPCR in MLL-rearranged mouse models showed increased H3K79me2 and decreased H3K9me3 and H3K27me3 at Pbx3 in LSCs; and high PBX3 expression correlates with poor prognosis in clinical AML.

    PBX3 is essential for leukemia stem cell maintenance in MLL-rearranged leukemia · Guo H, Chu Y, Wang L, Chen X, Chen Y, Cheng H, Zhang L, Zhou Y, Yang FC, Cheng T, Xu M, Zhang X, Zhou J, Yuan W · International Journal of Cancer 141:324-335 2017
    PMID 28411381DOI
    What this evidence cannot show

    Abstract read this cycle. MURINE MLL-AF9 myeloid disease. The author list was first written from recall and was WRONG - it was replaced with the full list fetched from PubMed efetch on 2026-09-20, which is a live instance of the brief's identifier rule applying to author strings and not only to PMIDs. The epigenetic characterisation is correlative - it describes why Pbx3 is high, not what Pbx3 does. The clinical component is a database prognosis correlation, not a functional human result.

    Evidence that argues against this (1)
    • The same LSC-maintenance role is claimed for MEIS1 in this disease, and the two cofactors are obligately linked through Meis1 protein stability, so an LSC phenotype on Pbx3 deletion may be a Meis1 phenotype reported under a different gene name. Garcia-Cuellar MP et al. 2015
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In murine Hoxa9-transformed haematopoietic cells, Pbx3 and Meis1 must dimerise to support Hox-induced leukaemia: in the absence of Pbx3, Meis1 protein is highly unstable, with degradation depending 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; blocking Meis1/Pbx3 dimerisation abrogates proliferation and colony formation and the expression of the Meis1 targets Flt3 and Trib2; Meis1 must bind Pbx3 to form high-affinity DNA/Hoxa9/Meis1/Pbx3 complexes in vitro; and Pbx3/Meis1 dimers are rate-limiting for Hoxa9-induced leukaemia in transplantation assays.

    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; full text not retrieved. MURINE Hoxa9-driven leukaemia, not a KMT2A fusion model and not human - the system is Hox overexpression, which is the fusion's output rather than the fusion. The author surname Fuller is rendered without its umlaut here; PubMed gives Fuller E. Independent of the Li group, which is what makes the convergence with E2 worth something.

    Evidence that argues against this (1)
    • If Pbx3 acts mainly by stabilising Meis1, then MEIS1 and not PBX3 is the effector, and a PBX3-directed intervention would be an indirect way of lowering MEIS1 - an argument for targeting the interface rather than either protein. Li Z et al. 2016
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Fetched live on 2026-09-20, DepMap Chronos gene effect for PBX3 has mean +0.0297 across 1,258 lines and +0.0418 across 57 leukaemia-annotated lines, with only 0.08% of lines at or below -0.5 and none at or below -1.0. PBX1 is +0.0110/+0.0391 and PBX2 is -0.2524/-0.3464. PBX3 knockout therefore carries no detectable fitness cost in any human leukaemia line in the panel.

    What this evidence cannot show

    A bounded negative with a date, a query and a tool, and it is bounded ONLY on what a 21-day CRISPR fitness screen in immortalised lines can see. It cannot see self-renewal, LSC frequency, differentiation, or any in vivo phenotype, which is where every published PBX3 result lives. Three paralogues share a DNA-binding mode, so single-gene knockout is buffered by construction. No infant-derived line is in the panel, and the subset is a coarse text match on the screen's diseaseFromSource field.

    Evidence that argues against this (1)
    • This graph's own genome-wide scan ranks PBX3 1,022 of 17,728 for KMT2A-rearranged SELECTIVITY (difference -0.1052), so there is a small differential between rearranged and other leukaemia lines even though the absolute dependency is nil - a difference of nearly nothing from nothing, recorded so that nobody reconstructs it as a hit. analysis run by this project 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In human MLL-rearranged leukaemia, RNA-interference suppression of HOXA9 co-down-regulates a programme that includes HOXA10, MEIS1, PBX3 and MEF2C; HOXA9 depletion across 17 human AML/ALL lines induces proliferation arrest and apoptosis specifically in the 7 MLL-rearranged lines (P=0.007) and not the 10 MLL-germline lines; and mice transplanted with HOXA9-depleted t(4;11) SEMK2 cells had a significantly lower leukaemia burden.

    HOXA9 is required for survival in human MLL-rearranged acute leukemias · Faber J, Krivtsov AV, Stubbs MC, Wright R, Davis TN, van den Heuvel-Eibrink M, Zwaan CM, Kung AL, Armstrong SA · Blood 113:2375-2385 2009
    PMID 19056693DOI
    What this evidence cannot show

    Abstract read this cycle. HUMAN cell lines and primary AML samples, and the in vivo arm uses SEMK2, a t(4;11) KMT2A::AFF1 B-lineage line - the closest system in this entry to the index disease. But it is a knockdown, the primary-sample arm is AML, and co-down-regulation of PBX3 on HOXA9 suppression establishes co-regulation, not that PBX3 is the effector.

    Evidence that argues against this (1)
    • In this graph's genome-wide DepMap scan HOXA9 ranks 3,649 of 17,728 with a difference of -0.0479 and a live mean effect of only -0.223 across 57 leukaemia lines, so HOXA9 knockout does not reproduce the requirement this knockdown reports. Broad Institute DepMap; Open Targets Platform 2026
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Nuclear localisation of PBC-class proteins is conditional on heterodimerisation with a MEINOX-class partner: HTH and PREP1 are cytoplasmic and require association with EXD/PBX for nuclear localisation, EXD and PBX1 are nuclear in murine fibroblasts but actively exported in Drosophila Schneider cells, and co-expression of EXD/PBX with HTH/PREP1 produces nuclear localisation of the heterodimers in both contexts - proposed to work by blocking nuclear export rather than by adding an import signal.

    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, and NOT a haematopoietic cell. It is cited as family-level cell biology that makes a specific prediction about PBX3/MEIS1 in leukaemia, not as evidence about PBX3. No measurement of PBX3 subcellular distribution in any KMT2A-rearranged cell was found this cycle.

    Evidence that argues against this (1)
    • PBX localisation is also controlled cytoplasmically by an unrelated mechanism - non-muscle myosin promotes cytoplasmic localisation of PBX - so a nuclear/cytoplasmic partition attributed to MEINOX partners may have a second, partner-independent determinant. Huang H et al. 2003
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PBX3 is essential for leukaemia stem cell maintenance in MLL-rearranged leukaemia: CRISPR/Cas9 deletion of Pbx3 in leukaemic cells of the MLL-AF9-induced mouse AML model significantly prolonged survival and decreased leukaemia burden by decreasing the capacity of LSCs and promoting LSC apoptosis, and ChIP analysis in MLL-rearranged mouse models found the Pbx3 locus carrying increased H3K79me2 with decreased H3K9me3 and H3K27me3 in LSCs; high PBX3 expression correlates with poor prognosis in AML patients.

    PBX3 is essential for leukemia stem cell maintenance in MLL-rearranged leukemia. · Guo H, Chu Y, Wang L, Chen X, Chen Y, Cheng H, Zhang L, Zhou Y, Yang FC, Cheng T, Xu M, Zhang X, Zhou J, Yuan W · Int J Cancer 2017
    PMID 28411381DOI
    What this evidence cannot show

    Murine MLL-AF9 AML - KMT2A::MLLT3 in myeloid disease - with human AML expression data from a clinical database. No KMT2A::AFF1, no B-lineage and no infant arm. The prognostic correlation is association in a database, not a functional result. Read: abstract only. The graph's registry records PBX3 (P40426, 434 aa) and PBX1 (P40424, 430 aa) as four residues apart with adjacent accessions, and the founding finding in this area is specifically that PBX3 and not PBX1 is rate-limiting, so a swap would invert the claim while passing every format check.

    Evidence that argues against this (1)
  9. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Pbx3 maintains an immune-cold, differentiation-blocked state during MLL-AF9 leukaemia progression: in a CRISPR-generated murine Mll-Mllt3/Af9 AML model, progression in vivo is accompanied by downregulation of type I interferon response genes, and this depends on upregulation of the MLL-AF9 transcriptional target Pbx3 in the progenitor cell compartment; enhancing the interferon response with IFN-alpha induces leukaemic cell differentiation, and inhibiting MLL-AF9 transcriptional activity on top of enhanced IFN signalling further delays progression.

    PMID 40108441DOI
    What this evidence cannot show

    Murine CRISPR-generated MLL-AF9 model. Note the model's own caveat, which the abstract states: the MA9 fusion immortalises HSPCs in vitro but is insufficient to induce overt leukaemia rapidly in vivo, so progression here involves events beyond the fusion. No KMT2A::AFF1, no B-lineage, no human arm. Read: abstract only.

    Evidence that argues against this (1)
    • This graph records the interferon state in KMT2A-rearranged leukaemia as something a MENIN INHIBITOR creates through endogenous retrovirus induction and MHC class II, not as a baseline axis, and the present result is an IFN-alpha administration experiment in a mouse myeloid model. The two are different provenances for a shared vocabulary, and the PBX3 result additionally depends on a model the authors say does not become leukaemic from the fusion alone. Guo H et al. 2017
Connected to