Oncofusion

start / graph / protein-histone-h3-3

Proteinstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q37-histone-chaperones

Histone H3.3 - one protein, two genes, and the reason this graph should not file it as gene-h3f3a

also called H3.3 · H3-3A · H3-3B · H3F3A · H3F3B · histone H3.3 · P84243

H3.3 is a replacement version of a core histone that gets slotted into genes while they are being read, rather than only when DNA is copied. Two separate genes make it, and the protein they make is exactly the same, which is why this entry is about the protein and not about either gene.

The same thing, for a scientist

H3.3 is the replication-independent H3 variant, deposited by HIRA at transcribed genes and regulatory elements and by ATRX/DAXX at heterochromatin [E1]. Two human genes, **H3-3A (formerly H3F3A, 1q42) and H3-3B (formerly H3F3B, 17q25)**, encode it - and **UniProt files both under a single reviewed record, P84243, 136 aa, because the protein products are identical** [E2]. That is the reason this entry is filed as a protein entry and not a pair of gene entries (gene-h3f3a / gene-h3f3b). **The graph-hygiene reason matters as much as the biology.** This project has a documented false positive in which an instrument keeps promoting histone **H3C1** as a gap: a replication-dependent H3.1 gene, one of a 15-member cluster, that appears as a high-ranking interactome and dependency hit purely because histones are abundant and because a 15-gene cluster gives fifteen chances to rank. Filing histone variants as individual `gene-` entries creates exactly that surface. This graph already, deliberately, files histones as **marks and interfaces** - `mark-h3k79me1/2/3`, `mark-h2bk120ub`, `interface-psip1-pwwp-h3k36me-nucleosome` - and the variant belongs with them as a property of the nucleosome, not as a gene dossier. **The one place the two genes are genuinely different is oncogenic point mutation**, which is a locus fact and not a protein fact: paediatric high-grade glioma K27M and G34R/V arise predominantly in *H3-3A*, and those are comparator diseases, not this one.

Go deeper
Why gene-h3f3a would be the wrong unit, stated so it is not re-litigated

Four reasons, in decreasing order of how much they are about biology:

  1. The protein is identical. UniProt's reviewed human record P84243 carries both primary gene names H3-3A; H3-3B on one entry 2. There is no H3.3A protein and no H3.3B protein. Two gene- entries would carry the same 136 residues, the same interfaces and the same marks, and would immediately drift apart.
  2. The interesting facts are about deposition, not about the gene. Everything in this beat - which chaperone, which locus, how fast the turnover - is a statement about where the protein goes, and none of it distinguishes the two loci.
  3. The graph's existing convention. Histone content in this graph is filed as marks and interfaces. A variant is a property of a nucleosome in the same way a mark is.
  4. The instrument hazard. The H3C1 false positive is a ranking artefact that this project has already paid for once. Adding more histone gene- nodes gives that artefact more places to land and makes a future gap-finder more, not less, confident.

What would change this. If a beat needed to discuss H3-3A K27M or G34V as lesions - locus-specific, mutation-specific, glioma - then a variant- or lesion- entry keyed to the locus would be right. A gene- entry would still be wrong, because the thing being discussed is a mutation, not a gene dossier.

Why H3.3 is load-bearing for this graph even though it is not a KMT2A gene

This graph's H3K79 argument says the mark has one writer and no eraser, so its level tracks histone replacement (mechanism-h3k79-turnover-and-dilution). H3.3 is the histone of replacement.

And there is a direct, already-cited measurement linking the two that the graph carries without drawing the conclusion: in synchronised human cells resolved by H3 isoform, H3.3K79me1 levels are higher than H3.2 and H3.1, while the rate of establishing K79 methylation is the same for all three isoforms 3. The obvious reading is that H3.3 nucleosomes have had more opportunity to be methylated because they are the ones sitting in transcribed chromatin, which is exactly where DOT1L works. If that is right, then the H3K79me pool a DOT1L inhibitor has to clear is disproportionately an H3.3 pool, and its clearance rate is set by replication-independent turnover rather than by cell division.

That is a testable statement nobody has tested in this disease, and it is the core of hyp-kmt2a-aff1-alters-h3-3-deposition-at-fusion-targets.

In blood, H3.3 itself is required where its chaperone may not be: H3.3-null HSPCs exhaust prematurely, gaining H3K27me3 at stemness and lineage promoters and losing H3K9me3 at endogenous retroviral repeats 4.

Facts and figures
uniprotP84243
length AA136
filing decisionprotein-level entry, NOT gene-h3f3a / gene-h3f3b; identical protein products on one reviewed UniProt record, and this graph files histone content as marks, interfaces and variants rather than as gene dossiers
depmap noteH3-3A and H3-3B are scored separately in DepMap; H3-3A difference -0.227 at pan-essentiality -1.454, but the sign REVERSES across the lineage split (myeloid -0.472, lymphoid +0.173), so the apparent selectivity is a myeloid signal in a myeloid-enriched index arm
kmt2a literature records0
genes
[
  {
    "symbol": "H3-3A",
    "former_symbol": "H3F3A",
    "cytoband": "1q42.12"
  },
  {
    "symbol": "H3-3B",
    "former_symbol": "H3F3B",
    "cytoband": "17q25.1"
  }
]
deposition routes
[
  {
    "chaperone": "protein-huca-complex",
    "locus_class": "transcribed gene bodies, promoters, cis-regulatory elements"
  },
  {
    "chaperone": "gene-atrx + gene-daxx",
    "locus_class": "telomeres, pericentric heterochromatin, ERVs"
  },
  {
    "chaperone": "gene-asf1b",
    "locus_class": "erythroid promoters and enhancers in fetal liver"
  }
]
How this could bite us back
  • H3.3 is deposited in every transcribing cell, including every normal haematopoietic stem cell, and H3.3-null HSPCs exhaust prematurely - so a general block on H3.3 deposition has an obvious and severe marrow-failure toxicity with no described window.
  • If H3.3 turnover is what clears H3K79 methylation, then reducing H3.3 deposition PROLONGS the mark, which means any intervention in this layer works against a DOT1L inhibitor rather than with it.
  • The variant is also the substrate of the ATRX/DAXX heterochromatin route, so an H3.3-directed intervention would simultaneously perturb ERV silencing - the same axis whose loss in H3.3-null HSPCs opens interferon-regulatory-factor binding sites.
Still open
  1. Is the H3K79me2 pool at KMT2A-fusion target genes disproportionately carried on H3.3? Isoform-resolved mass spectrometry of H3K79me at HOXA loci has never been done in a KMT2A-rearranged cell and would settle whether the clearance clock is a replication clock or a turnover clock.
  2. Does a KMT2A fusion change H3.3 deposition at its own target genes, or does it occupy chromatin whose H3.3 content was already high because those genes were already transcribed?
  3. If a future beat needs H3-3A K27M or G34V as a lesion, what entry type should hold it? This entry argues for a locus-keyed variant entry, not a gene dossier, and the question is open.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Genome-wide profiling of endogenously tagged H3.3 in mouse embryonic stem and neuronal precursor cells 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 not at telomeres, where Atrx and Daxx act instead.

    Distinct factors control histone variant H3.3 localization at specific genomic regions. · Goldberg AD, Banaszynski LA, Noh KM, et al. · Cell 2010
    PMID 20211137DOI
    What this evidence cannot show

    Abstract read from efetch 2026-09-20; full text not read. Mouse ES and neural precursor cells. The profiles are of a tagged allele. No haematopoietic or leukaemic system anywhere in the study.

    Evidence that argues against this (1)
    • The chaperone assignment does not transfer to blood: in fetal liver haematopoiesis HIRA is dispensable and ASF1B determines H3.3 enrichment at erythroid regulatory elements. The routes are cell-type-specific, so an ES-cell map is a hypothesis about a blast. Liu J et al. 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Human histone H3.3 is a single reviewed UniProt record, P84243 (H33_HUMAN), 136 amino acids, whose primary gene names are BOTH H3-3A and H3-3B (synonyms H3.3A/H3F3/H3F3A and H3.3B/H3F3B); a `gene_exact` query for either symbol returns this one record and neither symbol matches it as a sole primary name.

    UniProtKB P84243 (H33_HUMAN), Histone H3.3, 136 aa, gene names H3-3A; H3-3B · UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
    What this evidence cannot show

    Fetched 2026-09-20 with `gene_primary` requested and read back per rule 62. **This is the evidence for the filing decision and it is a database-structure fact, not a biological argument**: UniProt merges records with identical sequences, so the merge is evidence that the proteins are identical and is not itself evidence about how a knowledge graph should be organised. The biological argument is made in the sections. The two loci remain distinct for genomic facts - copy number, promoter usage, somatic point mutation.

    Evidence that argues against this (1)
    • The two genes are not interchangeable for every purpose: they have different chromosomal locations, different untranslated regions and different expression control, and the oncogenic H3.3 point mutations of paediatric high-grade glioma fall predominantly in one of them. A protein-level entry cannot hold locus-level facts, and if this graph ever needs them it will need a separate, locus-keyed entry. Funato K et al. 2014
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In synchronised human cells labelled with stable isotopes on both histones and their methyl groups and resolved by H3 isoform, H3.3K79me1 levels are higher than those of H3.2 and H3.1, while the rate of establishing K79 methylation is the same for all three isoforms.

    Kinetics of re-establishing H3K79 methylation marks in global human chromatin. · Sweet SMM, Li M, Thomas PM, et al. · Journal of Biological Chemistry 2010
    PMID 20699226DOI
    What this evidence cannot show

    Abstract read from efetch 2026-09-20; full text not read. Bulk chromatin in a synchronised human cell line; no locus resolution. Established as a measurement within that system. **The interpretation offered in this entry - that the higher H3.3K79me1 reflects residence in transcribed chromatin - is this entry's inference and is not stated in the abstract.** Cell-cycle synchronisation itself perturbs chromatin metabolism. This item is already carried by `mechanism-h3k79-turnover-and-dilution` as E3; it is re-cited here from a fresh fetch, not copied. **Deliberate confidence split, recorded so it is not read as an inconsistency:** `mechanism-h3k79-turnover-and-dilution` grades this same PMID `moderate` because its claim there includes the authors' inference of positional scrambling, which goes beyond the numbers. The claim here is restricted to the isoform-resolved measurement itself, which is a result within the reported system and therefore `established` under `concept-the-confidence-grading-rule`.

    Evidence that argues against this (1)
    • Equal methylation rates across isoforms argue that DOT1L does not discriminate variants, so a higher steady-state on H3.3 must come from residence time or location rather than from enzyme preference - which means any claim that H3.3 is 'the H3K79 histone' is a claim about chromatin context and collapses if turnover at the relevant loci is not what is assumed. Chory EJ et al. 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    H3.3 prevents premature haematopoietic stem cell exhaustion and differentiation into granulocyte-macrophage progenitors; H3.3-null HSPCs show reduced expression of stemness and lineage genes with a predominant gain of H3K27me3 at their promoters and a reduction of H3K9me3 at endogenous retroviral repeats, opening IRF-family binding sites.

    PMID 34961794DOI
    What this evidence cannot show

    Abstract read from efetch 2026-09-20; full text not read. Mouse, adult HSC, normal haematopoiesis; removing H3.3 requires deleting two genes and the abstract does not state the allele configuration or residual protein. The model at the end ('enforcing chromatin adaptability') is the authors' proposal.

    Evidence that argues against this (1)
    • The same abstract states that the chaperone HIRA is NOT required for this phenotype, which means the H3.3 requirement in HSCs is met by a route other than the canonical gene-body chaperone - so an H3.3 phenotype cannot be used as evidence for a HIRA phenotype or vice versa. Murdaugh RL et al. 2021
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    As of 2026-09-20, PubMed returns 1 record for '(H3F3A OR H3F3B OR H3.3) AND (MLL OR KMT2A) AND leukemia' and it is off-topic on title check (a review of IDH1/2, KDM5C and KDM6A crosstalk with hypoxic reprogramming, PMID 31221981); '(KMT2A OR MLL) AND "H3.3"[tiab]' returns 1 record, a cortical spreading depression study (PMID 20211609); 'menin AND (H3.3 OR "histone chaperone"[tiab])' returns 1 record, a paediatric glioma H3.3K27M modelling paper.

    What this evidence cannot show

    Query translations captured; none contained a [UID] term. All returned titles were fetched and read. **'H3.3' is a difficult query token** - the period and digits make it prone to tokenisation effects, which is why three differently-shaped queries were run rather than one. A study naming the variant only as 'H3F3A' in a supplement, or discussing replication-independent deposition without naming H3.3 in title or abstract, would not be caught.

    Evidence that argues against this (1)
    • H2A.Z, the other variant in this beat, DOES have a direct KMT2A-fusion paper - Tip60 is recruited by MLL-AF10 and MLL-ENL to Hoxa9 where it acetylates H2A.Z. So variant biology per se is not absent from this disease; it is H3.3 specifically that has never been looked at. Yamagata K et al. 2021
Connected to