KAT7 / HBO1 — an enzyme, a leukaemia-stem-cell dependency, and an inhibitor already in a crystal

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.

• Not an orthogonal mechanism: KAT7 acts through H3K14ac sustaining HOXA9/HOXA10, which is the same output menin and DOT1L inhibition target. Combinations of two agents that both lower HOXA9 should be expected to share resistance mechanisms.
• Replication-associated function: KAT7/HBO1 has roles in origin licensing, so an inhibitor may cause replication stress and myelosuppression independent of its leukaemia effect.
• Cofactor-competitive selectivity risk: the published inhibitor is an acetyl-CoA analogue, a chemotype with an inherent cross-reactivity problem across the MYST family and other acyl-CoA enzymes.
• Strong human constraint: LOEUF 0.207 and mis_z 5.92 with only 10 observed loss-of-function variants against 82.1 expected. Human genetics says this gene is protected, and cell-line data cannot see what it is protected for.
• Scaffold dependency cuts both ways: KAT7 activity requires a BRPF-family partner, so a cell could modulate activity by switching scaffold (BRPF1, BRPF2/BRD1, BRPF3) rather than by mutating the enzyme - an escape route nobody has looked for.
• Sensitivity to the HBO1 inhibitor WM-3835 depends on ATR status, because ATR phosphorylation controls how much HBO1 is on chromatin (PMID 42575441). A KMT2A-rearranged blast under replication stress - which is the state intensive chemotherapy produces - may therefore have a different amount of the target where the drug needs it, and sequencing an HBO1 inhibitor after or with a genotoxic agent is a scheduling question nobody has asked.

1. Is KAT7 a genuine KMT2A-rearrangement-selective dependency, or a general AML leukaemia-stem-cell dependency that our 10-line DepMap comparison is over-reading? The published work claims the latter; the DepMap selectivity of -0.242 suggests the former. A matched comparison in KMT2A-rearranged versus KMT2A-wild-type primary AML would settle it.
2. Does KAT7 physically contact the KMT2A fusion at all? No verified evidence exists either way. If it does not, then this is a downstream circuit node rather than an interactome member and should be reclassified.
3. Is an acetyl-CoA-competitive inhibitor selectable against the rest of the MYST family and against other acyl-CoA enzymes? The four-member acyl-CoA structural series (7D0P-7D0S) is exactly the data needed to design that selectivity and nobody has published the cross-family panel.
4. What does the mis_z of 5.92 mean for tolerability? KAT7 has replication-licensing functions, so an inhibitor may be myelosuppressive in a way the mean Chronos of -0.300 does not predict. A haematopoietic colony assay with the published inhibitor is the gating safety experiment.
5. Does KAT7 inhibition add anything to menin inhibition, given both act through HOXA9? A combination matrix with HOXA9 and H3K14ac readouts would distinguish a genuine second mechanism from a second entry point into the same one.

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   The KAT7 MYST/HAT domain (residues 336-611) is solved in eight structures bound to BRPF2/BRD1 (UniProt O95696) residues 31-80: apo (7D0O, 2.51 A), with acetyl-CoA (5GK9 2.40 A; 6MAK 2.13 A), with an inhibitor (6MAJ, 2.14 A, ligand JAV), and with propionyl-, butyryl-, crotonoyl- and succinyl-CoA (7D0P 1.80 A, 7D0Q 2.21 A, 7D0R 1.95 A, 7D0S 2.30 A). KAT7's structural scaffold partner in all of them is BRPF2/BRD1 (O95696), not BRPF1 (P55201).

   RCSB PDB entries 5GK9, 6MAJ, 6MAK, 7D0O, 7D0P, 7D0Q, 7D0R and 7D0S, with per-entity UniProt accessions, bound ligands, methods, resolutions and modelled residue ranges resolved against the RCSB search and data GraphQL APIs · RCSB Protein Data Bank · RCSB PDB 2026
   ▸What this evidence cannot show

   All identifiers, ligands, resolutions and residue ranges were read from the RCSB API in this session rather than recalled, which is how the BRPF2-versus-BRPF1 conflation was caught. Every structure is of the isolated HAT domain with a short scaffold peptide; there is no structure of the full KAT7-BRPF-JADE-MEAF6 complex, and none of these structures contains a nucleosome, KMT2A or any KMT2A fusion protein.

   ▸Evidence that argues against this (1)

   • Counting PDB entries measures how tractable a protein was to CRYSTALLISE, which correlates with being small, ordered and well-behaved rather than with being druggable. What decides design value is whether the deposited density covers the site that matters, at a resolution that supports modelling, in the relevant state — and this graph has repeatedly found structures described in the direction that overstates what is resolved. All eight entries are the MYST/HAT domain with the same BRPF2 fragment, so the coverage is one construct solved eight times rather than eight views of the protein, and the selectivity problem across KAT6A/KAT6B/KAT7 is untouched by any of them. RCSB PDB 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   ChEMBL v34 target CHEMBL3774299 (histone acetyltransferase KAT7) carries 98 activity rows at pChEMBL >= 6, with a best Ki of 20 nM (CHEMBL5812221) and a compound at IC50 30 nM in a functional assay with Kd 34 nM by binding (CHEMBL5723365), from recent (2024) records; no clinical candidate has a curated mechanism of action against it.

   ChEMBL database version 34, target CHEMBL3774299 (histone acetyltransferase KAT7): target_search, get_bioactivity and get_mechanism queries · EMBL-EBI ChEMBL · ChEMBL 2026
   ▸What this evidence cannot show

   Distinct-compound counts and 'best' values are page-local because of the 25-row API limit. An acetyl-CoA-competitive mechanism, which is what the published inhibitor uses, raises an inherent selectivity concern across the MYST family and more broadly across acyl-CoA-using enzymes; the ChEMBL record does not report cross-family selectivity.

   ▸Evidence that argues against this (1)

   • A ChEMBL activity-row count measures how much medicinal chemistry has been DONE, not how tractable the target is, and the best reported affinity belongs to the most optimised compound in whichever series someone chose to pursue. Zero rows means nobody tried; thousands of rows means a target had a decade of industrial attention. ChEMBL, EMBL-EBI 2026
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   KAT7 is not DepMap common-essential (mean Chronos -0.300 across 1,258 lines, 30.1% of lines depleted) and falls to -0.629 across 10 KMT2A-rearranged lines and -0.543 across 4 KMT2A::AFF1 lines, giving a lineage-controlled selectivity of -0.242 and rank 5 of 41; gnomAD gives pLI 1.0, LOEUF 0.207, 10 observed versus 82.1 expected loss-of-function variants and mis_z 5.92; HPA classifies it as low tissue specificity, detected in all tissues.

   DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API; loss-of-function constraint via the gnomAD GraphQL API; normal-tissue expression via the Human Protein Atlas API · Broad Institute DepMap; gnomAD; Human Protein Atlas; Open Targets Platform · Databases 2026
   ▸What this evidence cannot show

   Means were computed by us from per-line API values, not read from a DepMap release table, and the release version is not stated by the API. Only 10 KMT2A-rearranged lines (4 KMT2A::AFF1) contribute. The gnomAD profile is the most constrained in the top five candidates (mis_z 5.92), which is biologically plausible for a replication-associated acetyltransferase and is the main reason this node is not rated higher overall.

   ▸Evidence that argues against this (1)

   • Strong human missense and loss-of-function constraint (mis_z 5.92, LOEUF 0.207, 10 observed versus 82.1 expected LoF variants) argues KAT7 is functionally protected in humans, which a mean cell-line Chronos of -0.300 does not capture. gnomAD Consortium 2026
4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   shRNA and CRISPR domain-focused screens identify the HBO1/KAT7 histone acetyltransferase domain as essential for the maintenance of acute myeloid leukaemia stem cells, acting via H3K14 acetylation sustaining Hoxa9 and Hoxa10, and a potent HBO1 inhibitor acts as a competitive acetyl-CoA analogue.

   HBO1 is required for the maintenance of leukaemia stem cells · MacPherson L, Anokye J, Yeung MM, et al. · Nature 2020
   PMID 31827282DOI
   ▸What this evidence cannot show

   The claim is about AML leukaemia stem cells broadly, not specifically KMT2A-fusion or KMT2A::AFF1 leukemia; the models are an ex vivo mouse LSC system plus human AML lines and primary AML. No numeric IC50 for the inhibitor appears in the abstract. The KMT2A-rearrangement selectivity reported in this node comes from our own DepMap analysis and is a separate line of evidence, not part of this paper's claim.

   ▸Evidence that argues against this (1)

   • Domain-focused screens in AML stem cells are murine or human myeloid systems and the mechanism runs through Hoxa9, which infant t(4;11) does not express highly. The requirement may be for myeloid leukaemia rather than for the fusion. Oncofusion knowledge graph (internal analysis) 2026
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Regulation of HBO1/KAT7 acetyltransferase activity by its BRPF2 scaffold is structurally defined: a short BRPF2 N-terminal segment binds the HBO1 MYST domain and potentiates H3K14 acetylation.

   Structural and mechanistic insights into regulation of HBO1 histone acetyltransferase activity by BRPF2 · Tao Y, Zhong C, Zhu J, et al. · Nucleic Acids Research 2017
   PMID 28334966DOI
   ▸What this evidence cannot show

   The structure is the isolated MYST domain with a short scaffold peptide rather than the intact multi-subunit HBO1 complex, and no nucleosome is present. No numeric affinity for the HBO1-BRPF2 interaction appears in the abstract.

   ▸Evidence that argues against this (1)

   • The BRPF2 dependence is a complication rather than a detail: if a short BRPF2 segment potentiates H3K14 acetylation, then KAT7 activity in a cell is set by scaffold availability, and an inhibitor's effect will depend on which scaffold the complex carries in that lineage. RCSB PDB 2026
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The chromatin-bound fraction of HBO1/KAT7 is a regulated variable, not a constant: ATR is a novel HBO1 interacting partner and ATR-mediated phosphorylation of HBO1 at Ser50/Ser53 interferes with HBO1 genomic binding and alters gene expression, with an S50/53A phospho-dead mutant dampening DNA-repair gene expression and suppressing colony formation, and ATR inhibition antagonising sensitivity to the HBO1 inhibitor WM-3835.

   ATR-dependent phosphorylation of the histone acetyltransferase HBO1 suppresses chromatin binding and promotes replication stress responses. · Zong C, Zhang J, Zhang Z, Cai Y, Wang Y, Fang Y, Zheng G, Li Q, Wang L, Shen J · J Biol Chem 2026
   PMID 42575441DOI
   ▸What this evidence cannot show

   2026, single study, not replicated. 'Genomic binding' is a ChIP-level statement and therefore a genomic address rather than a spatial compartment. The systems are cancer cell lines, not leukaemia and not KMT2A-rearranged. The practical significance here is that a replication-stress state changes how much HBO1 is on chromatin, so any HBO1 occupancy measurement in a proliferating leukaemic cell is a measurement of that cell's ATR signalling as well.

   ▸Evidence that argues against this (1)

   • The only cell-cycle-resolved imaging of this complex measures the SCAFFOLD, not the enzyme. JADE1S is nuclear and chromatin-associated in interphase, dissociates from chromatin at prophase into the cytoplasm, and re-associates at telophase with nuclear envelope reassembly - and it is the cytoplasmic, not the chromatin-bound, JADE1 that is phosphorylated. Inferring HBO1's own localisation from this is an inference about a partner, and this graph has a standing record (localization-partner-baseline-compartment-is-often-miscalled) of exactly that error. Siriwardana NS et al. 2014

• ← analogous toKAT5 / Tip60 - the acetyltransferase in the only direct KMT2A-fusion histone-variant result, and both fusions tested were myeloid

  both MYST-adjacent acetyltransferases nominated in fusion leukaemia; KAT7 has a dossier and failed the pan-essentiality bar

• related to →Hypothesis: a defined subset of fusion binding sites survives saturating menin inhibition on menin-independent tethering, and that residual set is where bypass resistance comes from

  competing mechanism for residual fusion retention under menin inhibition: KAT6/KAT7 catalytic inhibition rapidly evicts MLL-FP from chromatin and overcomes primary and acquired menin-inhibitor resistance

• ← related toKAT8 (MOF) — a genuine KMT2A partner that binds the half of KMT2A the fusion discards

  MYST-family acetyltransferase contrast: KAT8 pan-essential with zero KMT2A selectivity

• ← supportsScored at complex level, coherence predicts nothing (rho -0.016 over 719 complexes) - but the Fanconi core complex is 2nd of 719 and the HBO1/MOZ acetyltransferase modules are 1st

  the HBO1/KAT7 acetyltransferase complex is position 1 of 719 by median selectivity rank

• ← supportsSix countermeasures, no head-to-head — so rank their targets: CDK6 is 13th and CDK4 is 16,052nd

  KAT7 ranks 92 of 17,728 and KAT6A 106, both inside the top 1%.

• regulates →The stem-like self-renewal state — a limited stem-cell module grafted onto a committed cell

  HBO1 sustains the leukaemia stem-cell state through H3K14 acetylation

• binds →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
• ← related toCo-dependency scan: what else does a menin-dependent cell need? (run, not planned)

  r=0.21 against MEN1

• ← supportsWhich leukemia co-dependencies survive uncertainty? (run, with p-values)
• ← supportsThe stricter null, and it partly reverses the last one (run)

  KAT7 exceeds its variance-matched null, unlike under the plain null

• ← regulatesThe HBO1/MOZ modules are not selective as modules: one paralogue per family carries the signal, and no member clears both a dependency and a pan-essentiality bar

  KAT7's KMT2A-rearranged effect is strong but it is broadly essential (pan -0.300)

• ← related toHypothesis: LEDGF licenses elongation at fusion target genes rather than anchoring the fusion, which predicts that a LEDGF agent is epistatic with BET and CDK9 inhibition and invisible in a fusion-ChIP assay

  candidate H4K16 acetyltransferase linking LEDGF to BRD4

• ← related toWhy a nucleoporin fusion needs menin-MLL1: the second menin-dependent fusion family, and what it borrows rather than carries

  KAT6A/KAT7 as a parallel, possibly competing dependency in the same disease

• ← related toMOF/KAT8 is required by both KMT2A::MLLT3 and NUP98::HOXA9 — which is why it is the wrong kind of convergence

  same untested pattern: KAT7 a dependency in both families, no comparator arm reported

• ← related toPhase-2 dossier: KAT7/HBO1 - the only enzyme here with a real active site, a co-crystal inhibitor and a co-dependency that survives lineage adjustment almost intact

  readiness layer on top of the KAT7 biology entry

• ← related to'Transcriptional memory' means two different things in this literature, and both of them are used about MLL leukaemia

  KAT7 is required for the H3K14ac that carries inducible memory in the canonical mammalian system

• related to →The soluble-nucleoplasm versus chromatin-bound axis - the one localisation readout that keeps a size dimension, and nobody has run it on a KMT2A fusion

  HBO1's chromatin-bound fraction is ATR-phospho-regulated; the soluble fraction has never been measured in this disease

• supports →A partner's 'normal' compartment is usually read off a different tissue, and for at least two KMT2A partners that reading is wrong in blood cells

  the only cell-cycle imaging of the HBO1 complex measures the JADE1 scaffold, not the acetyltransferase

• ← part ofJADE3 / PHF16 (Q92613) - the HBO1-complex scaffold with the strongest pan-fusion evidence in this beat, and an independent published contact with MLL

  JADE3 is a scaffolding subunit of the HBO1/KAT7 acetyltransferase complex; its PZP domain targets the complex to nucleosomes

• ← related toKMT2A*: a menin-blind, LEDGF-blind, SET-competent KMT2A protein made from out-of-frame reciprocal alleles

  Retains the domains needed to bind the H4K16 acetyltransferase machinery, per the recombinome's description of MOF binding.

• ← testsExperiment: time-to-resistance, not synergy index, as the readout for menin-inhibitor combinations

  includes a KAT6/KAT7 driver-eviction arm as the boundary case for the orthogonality framing