H3.3 - the graph's most-named epigenetic object with no entry, 325 mentions, and not one measurement in a KMT2A-rearranged cell

Two genes, one protein, one accession. H3.3 is encoded by H3-3A (HGNC:4764, 1q42.12, previous official symbols H3F3 and H3F3A, ENSG00000163041) and H3-3B (HGNC:4765, 17q25.1, previous official symbol H3F3B, ENSG00000132475). Both resolve to the same UniProt record, P84243 (H33_HUMAN, 136 aa), whose gene_primary field reads "H3-3A; H3-3B" - a single reviewed row carrying two primary gene names 5. Rule 62's discipline (keep only the row whose primary name equals the symbol you asked for) works here, but the row it keeps is shared between two loci on two chromosomes.

Consequences that matter for search. Older literature uses H3F3A and H3F3B almost exclusively; both were renamed, so a query on the current symbols alone under-counts in exactly the way this project's septin case documented. The bounded query used here therefore ran H3.3, H3F3A, H3F3B, H3-3A and H3-3B together 3.

The variant is not the mark and not the gene. data/symbol-collisions.json already blocks H3C1 on the grounds that a histone cluster gene is the substrate rather than a private interactor. The same reasoning applies here in reverse: an entry for H3.3 is an entry for a variant protein and its deposition, and it must not be written as gene-h3f3a. This entry deliberately uses the id variant-h3-3.

A substring trap found while doing this work. A naive substring count for HIRA across the same nineteen full texts returned 9 hits; a word-boundary count returned 0. Every hit was an author surname - Hirashima, Hirai. A histone-chaperone absence claim built on substring matching would have reported the chaperone present in papers that never mention it 4.

The clearance-clock argument, which drives mechanism-h3k79-turnover-and-dilution, exp-h3k79-state-resolved-icechip-dose-series and the DOT1L-inhibitor dosing discussion, runs as follows: H3K79 methylation has no eraser, so the only route to losing it is replacing the histone. Replication-independent replacement means H3.3 deposition, which means HIRA at gene bodies 1. Therefore the rate at which a DOT1L inhibitor clears its mark at a fusion target is the rate at which H3.3 is deposited there.

hyp-kmt2a-aff1-alters-h3-3-deposition-at-fusion-targets states the open question precisely - does the fusion set the turnover rate, or inherit it? - and gives it a prior of 0.35, with exp-h3-3-turnover-at-fusion-targets-under-dot1l-inhibition as the experiment.

The premise of all of it is unmeasured in this disease. No H3.3 ChIP, no SNAP-tag or CATCH-IT turnover measurement, no time-resolved deposition assay has been reported in any KMT2A-rearranged cell, line, xenograft or patient sample 34. The dosing argument is not wrong; it is untested at its foundation, and this entry exists to make that visible from the variant's own page rather than only from a hypothesis node.

PMID 36635503 asks how menin inhibition and EZH2 inhibition interact at bivalent genes, and in doing so knocks out KMT2A and KMT2B in K-562 cells. To test whether the KMT2A/B requirement for interferon-gamma-induced MHC-I is really an H3K27me3 requirement, the authors expressed H3.3 K27M - an oncohistone variant that inhibits H3K27me3 spreading - and found it produced the same bypass as EZH2 inhibition 2.

Three things to be exact about:

1. The variant is a reagent here, not an analyte. A lentivirally delivered HA-FLAG-tagged H3.3 K27M transgene is a way of removing a methyl mark. Nothing about endogenous H3.3 deposition, turnover or distribution is measured.
2. K-562 is a BCR::ABL1 CML line, not a KMT2A-rearranged cell. This project's own record already contains a correction on exactly this point ("the atlas is K562, not HEK293"), and the caution transfers: a KMT2A-dependency experiment in K-562 is a wild-type-KMT2A experiment, not a fusion experiment.
3. It is nonetheless a rule-80 find. H3.3 appears nowhere in that paper's title or abstract, and the paper is absent from the bounded H3.3 x KMT2A abstract query. Without opening the full text this entry would have said H3.3 appears nowhere in the KMT2A literature at all, which would have been wrong.

Deposition is chaperone-split and locus-specific. Genome-wide profiling of tagged endogenous H3.3 in mouse embryonic stem cells and neuronal precursors shows H3.3 patterns depend on amino-acid sequence and change with differentiation at developmentally regulated loci; HIRA is required for H3.3 enrichment at active and repressed genes but is dispensable at telomeres and at many transcription-factor binding sites, where ATRX and DAXX act instead 1. The two systems are not redundant, so 'H3.3 deposition' is at least two measurable quantities.

H3.3 is itself a paediatric cancer gene, and not in this disease. Across 1,000 paediatric cancer genomes spanning 21 subtypes, H3F3A is among the most frequently mutated epigenetic regulatory genes, driven by the p.K27M substitution that reaches up to 78% of diffuse intrinsic pontine glioma; the epigenetic regulators that recur in paediatric ALL in the same survey are CREBBP, EED, EP300, EZH2, PHF6 and SETD2, not H3F3A 6. So the variant is a well-established oncogenic object in children, in a different organ, by a mechanism (dominant-negative blockade of H3K27 methylation) that this graph's polycomb arm would care about if it ever appeared in a leukaemia.

The graph's chaperone entries already carry the leukaemia-adjacent parts. gene-hira records that HIRA AND (MLL OR KMT2A) returns 0 while HIRA AND leukemia returns 37, including a 2026 CML result in which HIRA knockdown raises H3K9me3 through SETDB1. gene-daxx records the largest selective DepMap difference among 254 unwritten curated interactors (-0.305) with an empty KMT2A literature. Both of those absences are about the chaperones. This entry records the absence of the cargo, and the two together mean the entire replication-independent histone-deposition layer is unmeasured in this disease while three of this graph's load-bearing arguments depend on it.