Oncofusion

start / graph / gene-hoxa9

Proteinstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q5c-interactome2

HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it

also called HOXA9 · homeobox protein Hox-A9 · HOX1G · ABD-B

HOXA9 is the best-known gene the KMT2A fusion switches on, and human MLL-rearranged leukemia cells die when you take it away. But mice engineered with an MLL fusion still get leukemia at the same rate without HOXA9, so it is needed to keep the disease going rather than to start it.

The same thing, for a scientist

HOXA9 (P31269, 272 aa) is the canonical transcriptional output of the KMT2A-fusion/SEC/PAFc/DOT1L machinery at the HOXA locus, not a complex member. HOXA9 knockdown induces proliferation arrest and apoptosis specifically in MLL-rearranged human cells, including in a t(4;11) SEMK2 xenograft. Against this, Mll-AF9 knock-in mice develop myeloid leukemia with the same incidence and time course with or without Hoxa9, differing only in phenotype, which argues for a redundant Hoxa5-a9 code rather than a single required gene. HOXA9 is gnomAD-unconstrained (LOEUF 0.946) and its mouse knockout is viable.

loading 1PUF…
PDB 1PUF
HOXA9 works by gripping DNA together with a partner protein, PBX1 — and even that picture comes from the mouse version.
Tell me more about this picture

1PUF: HoxA9 and Pbx1 homeodomains bound to DNA, 1.90 A X-ray.

Tell me more about this picture

The HoxA9 chain is mouse (UniProt P09631); the Pbx1 chain is human (P40424). No structure of human HOXA9 exists.

Tell me more about this picture

Pfam places the human homeodomain at 207-263 and a Hox9 activation region across 1-193, the latter largely disordered and structurally uncharacterised.

What it looks like
HOXA9272 aa
Hox9 activation region (largely disordered)Hox9 activation region (largely disordered) · residues 1–193HomeodomainHomeodomain · residues 207–2630100200
loading AF-P31269-F1…
AlphaFold AF-P31269-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 dispute is real and both sides are right

For requirement. HOXA9 depletion in 17 human AML/ALL cell lines induces proliferation arrest and apoptosis specifically in the 7 MLL-rearranged lines and not in the 10 MLL-germline lines (P = 0.007); primary MLL-rearranged AML samples show the same selectivity (P = 0.01); and mice transplanted with HOXA9-depleted t(4;11) SEMK2 cells carry significantly lower leukemia burden 1. That last point matters for this project: SEMK2 is a KMT2A::AFF1 line, so the in vivo evidence is not only myeloid.

Against requirement. Mll-AF9 knock-in mice develop myeloid leukemia with the same incidence and time course whether or not Hoxa9 is present. The leukemia in Hoxa9-null mice is simply more immature. Mll-AF9 overexpresses Hoxa5, Hoxa6, Hoxa7, Hoxa9 and Hoxa10, and the authors propose that a redundant 'Hox code' spanning Hoxa5–a9 is central to MLL leukemogenesis, with no single gene necessary 2.

These are not in conflict once the questions are separated. Faber and colleagues asked whether established human MLL-rearranged leukemia cells survive without HOXA9; Kumar and colleagues asked whether murine MLL-AF9 leukemia initiates without Hoxa9. Maintenance and initiation are different, and this graph should record both answers rather than the convenient one.

The practical consequence is a warning about redundancy. If Hoxa5–a10 are co-upregulated and functionally overlapping, then pharmacologically removing HOXA9 alone may be buffered, and the buffering will not be visible in a knockdown experiment that also depletes the programme.

Not constrained, not essential, viable knockout

HOXA9 is the least constrained gene in this beat by a wide margin: pLI 1.2 × 10⁻⁶, LOEUF 0.946 — statistically indistinguishable from neutral — with 18 observed loss-of-function variants against 19.0 expected 3. Humans tolerate heterozygous HOXA9 loss.

Hoxa9-null mice are viable, with 30–40% reductions in total leukocytes and lymphocytes, smaller spleen and thymus, reduced myeloid-erythroid and pre-B progenitors and a blunted G-CSF granulocytic response, while earlier progenitors are largely spared and heterozygotes are normal 4.

DepMap: mean Chronos -0.173, leukemia -0.191, depleted in 3.6% of 1,258 lines, not common-essential 3. Expression is tissue-enhanced and nearly restricted to bone marrow (2.4 nTPM) with essentially nothing in heart, liver or testis 3.

That is, on paper, an unusually favourable tolerability profile — the only gene in this beat that is both non-essential in cells and unconstrained in humans. The problem is not safety. The problem is that HOXA9 is a homeodomain transcription factor with no human structure (the widely used 1PUF entry is a mouse Hoxa9 homeodomain with human PBX1) 5, a large predicted-disordered activation region, and a redundant paralogue family.

A note on the identifier in 1PUF

PDB 1PUF is routinely cited as 'the HOXA9 structure'. We checked its polymer entities against the RCSB API: entity 3 is 'Homeobox protein Hox-A9' from Mus musculus, UniProt P09631; entity 4 is human PBX1, P40424 5. A structure search against the human accession P31269 returns zero entries.

The homeodomain is essentially identical between mouse and human, so the structure is usable — but this graph should say 'mouse Hoxa9 homeodomain with human PBX1 on DNA at 1.90 Å', not 'HOXA9 structure', because the distinction matters for anyone modelling the human activation region.

The chromatin state of the HOXA9 locus - and the marks HOXA9 itself does not write

HOXA9 is the most-cited fusion target in this disease and, until this amendment, this entry carried no epigenetic content at all. The distinction that makes the entry coherent is that HOXA9 has no chromatin-enzymatic activity of its own. It carries a Pfam PF04617 activation region and a PF00046 homeodomain and nothing else - no SET domain, no JmjC domain, no bromodomain, no chromodomain. So the epigenetics of HOXA9 is the epigenetics of the HOXA9 locus, as a target.

H3K79me2, and what is and is not resolved to this gene. In the KMT2A::MLLT3 (MLL-AF9) line MOLM-13, HOXA9 sits inside a ~100 kb broad H3K79me2 domain shared with neighbouring HOXA protein-coding and oncogenic lncRNA genes, with enhanced accessibility across the domain E6. In the index fusion's own literature the result is profile-level rather than gene-level: conditional murine Mll-AF4 B-precursor ALL and human MLL-rearranged ALL are distinguishable from other ALL by their H3K79 methylation profiles, and DOT1L suppression inhibits expression of critical MLL-AF4 target genes E7. That is a genuine index-fusion result - this graph's own audit found it cited only once - but it is not a HOXA9-resolved H3K79me2 measurement in a human KMT2A::AFF1 cell, and no such measurement was found in this tranche.

H3K79me2 is not the only mark activating this locus, and the second one is DOT1L-independent. In MOLM-13 and THP-1 (both KMT2A::MLLT3), SMYD3 depletion reduces H3K4 methylation and HOXA9 expression; rescue requires catalytic activity, since wild-type but not the enzymatically dead F183A SMYD3 restores it; ChIP places SMYD3 and H3K4me3 at the HOXA9 promoter; and the effect is reproduced in U937 and 293T independently of DOT1L, closing a HOXA9 -> SMYD3 -> H3K4me3 -> HOXA9 feed-forward loop E8. The same paper reports that SMYD3 and HOXA9 are elevated to a greater degree in the MLL-AF9 lines than in MV4-11, which carries the index fusion KMT2A::AFF1, and the authors read this as partner-relative specificity within the MLL-r family - so this axis must not be imported to the index fusion without test.

Cause or consequence is unresolved, and that is the entry. H3K79me2 is laid down co-transcriptionally, so the broadest domains are expected over the longest and most highly transcribed loci, and the HOXA cluster in a KMT2A-rearranged line is both. The perturbation evidence establishes that the mark's writer is required for target expression; it does not establish that the mark at HOXA9 is upstream of the transcription rather than a readout of it. This graph already carries a proposed ordering experiment and this entry does not claim the answer.

A bulk measurement that must not be read as locus-resolved. MOLM-13 carries higher global H2BK120ub and H3K79me2 than K562, and bortezomib reduces H2BK120ub and H2AK119ub together with transcripts of genes involved in MLL development E9. This is a whole-cell histone PTM measurement, not a ChIP at HOXA9, and the HOXA9 link in that work is transcript-level. It is recorded here because the H2Bub-to-DOT1L trans-histone dependency makes a global ubiquitin-pool effect a plausible route into this locus - not because the locus was assayed.

Facts and figures
uniprotP31269
gene symbolHOXA9
ensemblENSG00000078399
length AA272
alphafold IDAF-P31269-F1
cytoband7p15.2
PDB noteNo human HOXA9 structure. 1PUF is mouse Hoxa9 (P09631) with human PBX1 (P40424) on DNA, 1.90 A.
rolecanonical transcriptional output of the KMT2A fusion at the HOXA locus; not a complex member
druggabilityPoor. No human structure, a largely disordered activation region, a DNA-binding homeodomain as the functional unit, and a redundant paralogue family. The tractable object may be the HOXA9-PBX interaction rather than HOXA9 itself.
subnuclear localizationHOXA9 nuclear bodies, distinct from PML and eIF4E bodies in normal human marrow; relocated into enlarged eIF4E nuclear bodies in FAB M4/M5 AML and blast-crisis CML, reversibly under NF-kB inhibition (PMID 15657436)
shown for fusionNO - no KMT2A-rearranged specimen has been imaged for HOXA9; the FAB M4/M5 cohort in which the relocation occurs was not genotyped
epigenetic readoutTARGET LOCUS, NOT AN ACTIVITY. HOXA9 is a homeodomain transcription factor with no methyltransferase, demethylase, acetyltransferase or reader module: it writes, erases and reads no histone mark, and the epigenetic question about HOXA9 is the chromatin state OF ITS LOCUS. Measured at the locus: (1) H3K79me2 - in the KMT2A::MLLT3 line MOLM-13, HOXA9 lies inside a ~100 kb broad H3K79me2 domain shared with neighbouring HOXA genes; (2) in murine Mll-AF4 ALL and human MLL-rearranged ALL the H3K79 methylation profile over fusion targets distinguishes the disease from other ALL and DOT1L suppression collapses MLL-AF4 target expression - the index fusion's own partner, but a PROFILE-level result, not a HOXA9-resolved one; (3) H3K4me3 - SMYD3 and H3K4me3 are recruited to the HOXA9 PROMOTER by ChIP in MOLM-13, methyltransferase-activity dependent and explicitly DOT1L-INDEPENDENT, so H3K79me2 is not the only mark activating this locus. CAUSE OR CONSEQUENCE: UNRESOLVED - H3K79me2 is deposited co-transcriptionally, so breadth over a long, highly transcribed locus is the expected consequence of transcription, and no experiment has ordered mark loss against transcription loss at HOXA9 (see exp-ordering-of-elongation-and-h3k79-loss-in-kmt2a-aff1). NOT SHOWN FOR THE INDEX FUSION AT THIS LOCUS: no locus-resolved H3K79me2 or H3K4me3 ChIP at HOXA9 in a human KMT2A::AFF1 cell was found in this tranche. Bivalency is deliberately NOT recorded: cell-state-primed-bivalent-chromatin already carries the absence of any primary H3K4me3/H3K27me3 bivalent demonstration in KMT2A-rearranged leukaemia, and this tranche did not close it.
PDB IDS
[]
domains
[
  {
    "name": "Hox9 activation region (largely disordered)",
    "start": 1,
    "end": 193,
    "pfam": "PF04617"
  },
  {
    "name": "Homeodomain",
    "start": 207,
    "end": 263,
    "pfam": "PF00046"
  }
]
dispute
{
  "maintenance": "required: HOXA9 knockdown kills MLL-rearranged human cells including t(4;11) SEMK2 in vivo (Faber 2009)",
  "initiation": "not required: Mll-AF9 mice develop leukemia with equal incidence and time course without Hoxa9 (Kumar 2004)",
  "reconciliation": "different questions; a redundant Hoxa5-a10 code may buffer loss of any single paralogue"
}
essentiality
{
  "depmap_common_essential": false,
  "chronos_mean": -0.1725,
  "chronos_mean_leukemia_lines": -0.1912,
  "n_lines": 1258,
  "fraction_lines_depleted": 0.0364,
  "gnomad_pLI": 0.0000012,
  "gnomad_loeuf": 0.9457,
  "gnomad_obs_lof": 18,
  "gnomad_exp_lof": 19,
  "mouse_ko": "Viable; 30-40% reductions in leukocytes and lymphocytes, smaller spleen and thymus, reduced myeloid-erythroid and pre-B progenitors, blunted G-CSF response; heterozygotes normal",
  "hpa_ntpm": {
    "bone_marrow": 2.4,
    "thymus": 1.8,
    "classical_monocyte": 0.8,
    "naive_B_cell": 0.9,
    "spleen": 0.8,
    "lymph_node": 0.7,
    "liver": 0,
    "heart_muscle": 0,
    "testis": 0
  },
  "hpa_specificity": "Tissue enhanced"
}
codependency men1
{
  "r_leukemia_lines": 0.372,
  "ci_leukemia": [
    0.1184,
    0.5682
  ],
  "crosses_zero": false,
  "p_permutation": 0.0068,
  "p_holm": 0.08958,
  "survives_holm": false
}
localization
{
  "subcellular": "nucleus AND cytoplasm, both experimentally annotated from one source (UniProt P31269, ECO:0000269|PubMed:15657436); nuclear import is cytokine-regulated - thrombopoietin drives HOXA9 into the nucleus in UT-7/TPO and primary murine Sca-1+/c-kit+ cells without changing total protein (PMID 15254242)",
  "subnuclear": "discrete HOXA9 nuclear bodies, spatially distinct from PML bodies and eIF4E nuclear bodies in primary human normal marrow and FAB M1/M2 AML; in FAB M4/M5 AML and blast-crisis CML HOXA9 relocates almost completely into enlarged eIF4E nuclear bodies, reversed by an IkB super-repressor (PMID 15657436)",
  "how_shown": "indirect immunofluorescence and confocal microscopy on primary patient marrow, rabbit polyclonal anti-HOXA9 (Upstate) validated by antigen pre-absorption, co-stained with anti-eIF4E (BD) and anti-PML clone 5E10; plus Y11A eIF4E-binding-dead mutant and IkB-SR rescue",
  "systems": "primary human bone marrow, FAB M1/M2 and M4/M5 AML, bcCML CD34+, U937, UT-7/TPO, primary murine progenitors - none reported as KMT2A-rearranged",
  "node": "localization-hoxa9-nuclear-bodies-and-the-eif4e-body-switch"
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.2707,
  "chronos_mean_other_leukaemia": -0.2227,
  "difference": -0.0479,
  "pan_essentiality_all_lines": -0.1725,
  "selectivity_rank": 3649,
  "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
  • Redundancy: Mll-AF9 leukemia arises at normal incidence without Hoxa9 because Hoxa5-a10 are co-upregulated. A single-gene intervention against a redundant code is the classic way to get a clean pharmacodynamic effect and no clinical benefit.
  • Undruggable geometry: no human structure, a disordered activation region and DNA binding as the functional output.
  • Normal-tissue cost: Hoxa9-null mice have reduced pre-B progenitors and a blunted granulocytic stress response, both of which matter in an infant receiving intensive chemotherapy.
  • Interpretive risk: the maintenance evidence is RNAi-based and the initiation evidence is germline-knockout-based. Neither is what a drug does, and the two disagree.
  • Epigenetic escape route: SMYD3-deposited H3K4me3 activates HOXA9 independently of DOT1L in MLL-AF9 cells, so a DOT1L inhibitor can reduce H3K79me2 at this locus while a second, unaddressed writer keeps HOXA9 transcribed. A HOXA9-suppression pharmacodynamic endpoint built only on H3K79me2 could read as target engagement while the output persists.
Still open
  1. Is the Hoxa5-a10 redundancy that buffers Hoxa9 loss during initiation also present in established human KMT2A::AFF1 disease? If so, a HOXA9-directed agent would be buffered in patients in a way the knockdown experiments did not reveal.
  2. Would acute, complete degradation of HOXA9 in an established t(4;11) xenograft reproduce the knockdown result or the knockout result? A degron system separates initiation from maintenance and from compensation, and would settle a twenty-year-old dispute.
  3. Does menin inhibition reduce HOXA9 to a level below the threshold the redundant Hox code can compensate for? This graph already carries a HOXA9/MEIS1 dose-requirement entry; joining it to the redundancy argument is the natural next step.
  4. Is the HOXA9-PBX interface, visible in 1PUF, a more tractable target than HOXA9 itself?
  5. Is H3K79me2 at the HOXA9 locus cause or consequence? No experiment has ordered mark loss against transcription loss at HOXA9, and because H3K79me2 is deposited co-transcriptionally the broad domain is equally well explained as a readout of high transcription across a long locus. A degron plus a time-resolved, calibrated ChIP at HOXA9 would separate them.
  6. Has anyone measured H3K79me2 or H3K4me3 AT HOXA9, locus-resolved, in a human KMT2A::AFF1 cell? The index-fusion H3K79 evidence is profile-level (PMID 18977325) or spreading-target-level (PMID 28076791); this tranche found no HOXA9-resolved index-fusion ChIP. This reads as a measured-not-reported gap rather than a never-measured one, because the profiles exist in deposited data.
  7. Does the SMYD3-H3K4me3-HOXA9 loop operate in KMT2A::AFF1 at all? The founding paper's own data show it is weaker in MV4-11 than in the MLL-AF9 lines, and a DOT1L-independent writer sustaining HOXA9 would change combination design if it were present in the index fusion.
Every source behind this entry (9)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    HOXA9 depletion induces proliferation arrest and apoptosis specifically in MLL-rearranged human leukemia cell lines (7 MLL-rearranged versus 10 MLL-germline, P = 0.007) and in primary MLL-rearranged AML samples (P = 0.01), and mice transplanted with HOXA9-depleted t(4;11) SEMK2 cells show significantly lower leukemia burden; HOXA9 suppression co-downregulates HOXA10, MEIS1, PBX3 and MEF2C.

    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 2009
    PMID 19056693DOI
    What this evidence cannot show

    RNA interference, which is incomplete and subject to off-target seed effects. Tests maintenance of established leukemia, not initiation. Cell-line numbers are small (7 versus 10).

    Evidence that argues against this (1)
    • Mll-AF9 knock-in mice develop myeloid leukemia with the same incidence and time course with or without Hoxa9, differing only in phenotype, which argues that no single Hox gene is necessary for leukemia development. Kumar AR et al. 2004
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Mll-AF9 knock-in mice develop myeloid leukemia with the same incidence and time course in the presence or absence of Hoxa9, with only a more immature phenotype in the Hoxa9-null animals; Mll-AF9 overexpresses Hoxa5, Hoxa6, Hoxa7, Hoxa9 and Hoxa10, supporting a redundant Hoxa5-a9 code rather than a single required gene.

    Hoxa9 influences the phenotype but not the incidence of Mll-AF9 fusion gene leukemia. · Kumar AR, Hudson WA, Chen W, Nishiuchi R, Yao Q, Kersey JH · Blood 2004
    PMID 14615372DOI
    What this evidence cannot show

    Mouse Mll-AF9 knock-in, a myeloid model, addressing initiation rather than maintenance of established disease. Germline Hoxa9 loss allows lifelong compensation by paralogues in a way acute pharmacological removal would not.

    Evidence that argues against this (1)
    • Same incidence and time course without Hoxa9 is the strongest negative in this beat and it is a murine myeloid model with paralogue redundancy across Hoxa5/6/7/9/10 — so a Hoxa9-null leukaemia is not a HOXA-null leukaemia. The more immature phenotype the entry mentions is a real change the conclusion does not accommodate. Oncofusion knowledge graph (internal analysis) 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    HOXA9 is not DepMap common-essential (mean Chronos -0.173 across 1,258 lines, -0.191 across leukemia lines, 3.6% of lines depleted) and is gnomAD-unconstrained (pLI 1.2e-6, LOEUF 0.946, 18 observed against 19.0 expected loss-of-function variants); HPA reports tissue-enhanced expression nearly restricted to bone marrow at 2.4 nTPM.

    What this evidence cannot show

    DepMap release version not stated by the API. A near-neutral LOEUF in a short gene with only 19 expected loss-of-function variants is weakly powered; the honest reading is 'no evidence of constraint', not 'demonstrated tolerance'. HOXA9 is expressed in very few cell lines, which depresses its apparent essentiality.

    Evidence that argues against this (1)
    • A Chronos mean mixed across a KMT2A-rearranged subset and a comparator arm must be lineage-split before it is stated: the index arm is myeloid-enriched, so a difference between the two mixes fusion status with lineage. Ten or eleven index lines and four KMT2A::AFF1 lines also put wide intervals on every figure quoted. HOXA9 being gnomAD-unconstrained (pLI 1.2e-6) is also the expected profile of a gene with close paralogues, which constraint metrics systematically underestimate. gnomAD, Broad Institute 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Hoxa9-null mice are viable with 30-40% reductions in total leukocytes and lymphocytes, smaller spleen and thymus, reduced myeloid-erythroid and pre-B progenitors and a blunted G-CSF granulocytic response, while earlier progenitors are largely spared and heterozygotes are normal.

    Mice bearing a targeted interruption of the homeobox gene HOXA9 have defects in myeloid, erythroid, and lymphoid hematopoiesis. · Lawrence HJ, Helgason CD, Sauvageau G, Fong S, Izon DJ, Humphries RK, Largman C · Blood 1997
    PMID 9058712
    What this evidence cannot show

    A 1997 paper with no DOI (pre-DOI-era Blood record); PMID verified. Germline knockout with lifelong compensation. The pre-B progenitor deficit is directly relevant to infant B-ALL.

    Evidence that argues against this (1)
    • This is a mouse knockout describing what the gene does in normal development or normal haematopoiesis, over the animal's whole life, with the protein entirely absent. A drug gives partial reversible inhibition in an adult for weeks. The experiment bounds the biological requirement and not the therapeutic window. Nature 2012
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    There is no experimental structure of human HOXA9: a structure search against UniProt P31269 returns zero entries, and PDB 1PUF, commonly cited as the HOXA9 structure, contains mouse Hoxa9 (UniProt P09631) together with human PBX1 (P40424) on DNA at 1.90 A.

    What this evidence cannot show

    The HOXA9 homeodomain is highly conserved between mouse and human, so 1PUF is a legitimate model of the human DNA-binding module; the caution applies to the activation region and to any claim of a human structure. A negative structure search reflects deposition as of 2026-09-18.

    Evidence that argues against this (1)
    • The correction here is the valuable content: 1PUF contains MOUSE Hoxa9. This is the third instance in this project of a structure or identifier resolving to the wrong entity, and it should be read as a standing hazard rather than a one-off. RCSB PDB 2026
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In the KMT2A::MLLT3 (MLL-AF9) line MOLM-13, HOXA9 lies within a ~100 kb broad H3K79me2 domain shared with other HOXA protein-coding and oncogenic long non-coding RNA genes, and that broad domain shows enhanced chromatin accessibility throughout; genes carrying the broadest H3K79me2 domains across leukaemia cell lines are known leukaemia genes.

    Bioinformatic Analyses of Broad H3K79me2 Domains in Different Leukemia Cell Line Data Sets. · Sharma P, Sattarifard H, Fatemiyan N, et al. · Cells 2022
    PMID 36139405DOI
    What this evidence cannot show

    Abstract-level read. Re-analysis of previously published public ChIP-seq from immortalised leukaemia cell lines rather than primary blasts, with no perturbation and no test that breadth rather than height is the causal feature. Domain-breadth calls depend on sequencing depth, normalisation and peak caller, and datasets from different laboratories are compared. The line is KMT2A::MLLT3, NOT the index fusion.

    Evidence that argues against this (1)
    • Breadth is plausibly a consequence rather than a cause: H3K79me2 is deposited co-transcriptionally across gene bodies, so the broadest domains are expected over the longest and most highly transcribed loci, and the HOXA cluster in a KMT2A-rearranged line is both. The functional perturbation that exists localises DOT1L dependence to a distinct subset of ENHANCERS rather than to the broadest gene-body domains, so the broad domain over HOXA9 is not established as the dependency-bearing feature. Godfrey L et al. 2019
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In conditional murine Mll-AF4 B-precursor ALL and in human MLL-rearranged leukaemias, ChIP-chip H3K79 methylation profiles correlate with Mll-AF4-associated gene expression, human MLL-rearranged ALL is distinguishable from other ALL by its H3K79 profile, and suppression of DOT1L inhibits expression of critical MLL-AF4 target genes - establishing ectopic H3K79 methylation as a distinguishing feature of the index fusion's disease at profile level rather than resolved to HOXA9.

    H3K79 methylation profiles define murine and human MLL-AF4 leukemias. · Krivtsov AV, Feng Z, Lemieux ME, et al. · Cancer Cell 2008
    PMID 18977325DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. The murine arm is Mll-AF4, i.e. the index fusion, but in mouse; the human arm is 'MLL-rearranged', a mixture of partners rather than KMT2A::AFF1 specifically. ChIP-chip rather than ChIP-seq, so resolution and dynamic range are lower than modern profiles. Critically for this entry: the result is a genome-wide PROFILE classification, and the abstract reports no HOXA9-resolved H3K79me2 quantification, so it cannot be cited as 'H3K79me2 is at HOXA9 in KMT2A::AFF1'.

    Evidence that argues against this (1)
    • A profile that correlates with expression is not evidence that the mark drives each target. In an inducible MLL-ENL (KMT2A::MLLT1) system, fusion binding and H3K79 methylation changes are confined to a specific portion of the genome, and of 223 MLL-ENL-bound genes only 12 show a significant expression increase on fusion induction - Hoxa9 and Meis1 among them, but alongside Eya1 and Six1. On the authors' own counts, fusion binding plus an H3K79 change at a locus is usually NOT sufficient for activation, which weakens any inference from the mark's presence at HOXA9 to its causal role there. Wang QF et al. 2011
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In the KMT2A::MLLT3 lines MOLM-13 and THP-1, SMYD3 depletion reduces H3K4 methylation and HOXA9 expression, rescue requires SMYD3 methyltransferase activity (wild type but not the enzymatically dead F183A mutant), ChIP places SMYD3 and H3K4me3 at the HOXA9 promoter, and the effect is reproduced in U937 and 293T independently of DOT1L; HOXA9 in turn binds a TAATGAC motif in the SMYD3 promoter, closing a SMYD3-H3K4me3-HOXA9 feed-forward loop.

    PMID 40451434DOI
    What this evidence cannot show

    Full text read via Europe PMC (PMC12226129). The system is human but myeloid and MLL-AF9-driven, not infant B-lineage KMT2A::AFF1. The HOXA9-promoter ChIP is a single-locus ChIP in one line (MOLM-13), not genome-wide, and the DOT1L-independence arm is in U937 (CALM-AF10) and 293T, neither of which is a KMT2A-rearranged B-lineage cell. shRNA depletion carries the usual incompleteness and off-target exposure; the F183A catalytic-dead rescue is the strongest part of the causal argument.

    Evidence that argues against this (1)
    • COUNTER-EVIDENCE FROM THE SAME PAPER, which is why the axis is recorded rather than imported. The authors report that SMYD3 and HOXA9 are increased to a GREATER degree in the MLL-AF9 lines MOLM-13, THP-1 and NOMO-1 than in MV4-11, which carries the index fusion KMT2A::AFF1, and state that the regulation of SMYD3 by MLL-AF9 and HOXA9 may have some relative specificity within the MLL-r family. The loop is therefore weaker in the index fusion's own line on the paper's own data, and no KMT2A::AFF1 SMYD3 perturbation is reported anywhere in it. Zhang L et al. 2025
  9. E9weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    BULK, NOT LOCUS-RESOLVED: the KMT2A::MLLT3 line MOLM-13 carries a greater global level of H2BK120ub and H3K79me2 than the CML line K562 and a similar level of H2AK119ub, and bortezomib significantly reduces H2BK120ub and H2AK119ub together with transcript levels of genes involved in MLL development.

    Effect of the Proteasome Inhibitor, Bortezomib, on Histone Modifications in Human Leukemic Cell Lines. · Sattarifard H, Oyeyode M, Prajapati D, et al. · Int J Mol Sci 2026
    PMID 42589252DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. This is a whole-cell histone post-translational-modification measurement, NOT a ChIP at HOXA9 or anywhere else, so it establishes nothing about the HOXA9 locus directly - the HOXA9 and MEIS1 connection in that work is transcript-level and the paper's framing of those genes is introductory background. Two cell lines, one KMT2A-rearranged and one not, is a comparison confounded with everything else that differs between MOLM-13 and K562. The fusion is KMT2A::MLLT3, not the index fusion. A proteasome inhibitor depletes the free ubiquitin pool, so a fall in two ubiquitin marks is expected pharmacology and is not specific to leukaemia chromatin.

    Evidence that argues against this (1)
    • The direction of the H2Bub requirement is established but its selectivity is not, and this graph already carries that as an open dispute (hyp-h2bub-required-but-not-selective, hyp-h2bub-direction-is-unknown-in-kmt2a). RNF20, the H2B ubiquitin ligase, is required for MLL-rearranged leukaemia - which makes a bulk H2BK120ub reduction a plausible route to the fusion programme, but equally makes it non-specific: loss of a mark the leukaemia requires is not evidence that the mark was acting at HOXA9. Wang E et al. 2013
Connected to