Oncofusion

start / graph / mechanism-acetylation-is-a-cycle-not-a-mark

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q39-acetylation-cycle

Acetylation is a cycle, not a mark - and this graph holds one complete cycle, on a residue the fusion throws away

also called writers readers erasers · acetylation turnover · the eraser half of the acetylation layer

An acetyl tag is not a permanent label, it is a balance: one enzyme adds it, another reads it, a third removes it, continuously. This graph spent many cycles collecting the enzymes that add tags and never collected a single one that removes them - and the one place where it now has the whole loop, the tag sits on a piece of protein the cancer version has deleted.

The same thing, for a scientist

An acetyl-lysine is a **steady state between a writer and an eraser, read continuously by a third protein** - not a stable annotation. Histone acetylation turns over on a timescale of minutes, far faster than methylation, which means the occupancy at any locus is set by the **ratio** of writer to eraser activity and not by either alone. This graph's acetylation layer was, before this beat, **writers and one reader with no eraser at all**: `gene-kat6a`, `gene-kat7`, `gene-kat8`, `gene-ep300`, `gene-crebbp`, `gene-brd4` and `gene-sirt1`, with `HDAC` appearing in the text of 81 files and in the id of two entries while no HDAC entry existed. This entry records what closing that half revealed. **Three findings.** (1) **The graph now holds exactly one complete writer-reader-eraser cycle, and it is not on a histone.** PCAF/KAT2B writes acetyl-K339 on AF9; the mark's reader function is AF9's own oligomerisation and TFIID contact; HDAC5 erases it; and the switch is thrown by CaMKII phosphorylating HDAC5 and exporting it from the nucleus [E1]. **K339 lies N-terminal of AF9 residue 376 and is absent from the fusion in about 92% of KMT2A::MLLT3 patients** [E2] - so the one complete cycle is one the oncoprotein does not have. (2) **On histones, the two halves have never been measured in the same experiment.** Tip60 acetylates H2A.Z at *Hoxa9* [E3]; **no paper in this beat measures H2A.Z acetylation, or any fusion-target-locus acetylation, under any HDAC inhibitor.** The writer half and the eraser half of the histone cycle in this disease are two literatures that do not touch. (3) **Consequently, every 'HDAC inhibitor works here' result is compatible with several mechanisms that are not histone acetylation at all** - and four distinct ones are now on record.

Go deeper
The four mechanisms on record for one drug class, none of them attributable to an enzyme
mechanismagent it was shown withis it histone acetylation?measured in a patient?
bulk histone acetylationassumed for the classyesno
suppression of the RNF20/RNF40/WAC ligase complex, depleting H2B ubiquitinationpanobinostatno - a different histone mark, reached transcriptionallyno
acute release of free P-TEFb from the 7SK snRNP, with HEXIM1 induction as feedbackSAHA / vorinostatno - transcriptional elongation machineryno
suppression of cytidine deaminase, potentiating cytarabineromidepsinno - drug metabolismno

Four mechanisms, four agents, and the column that should discipline all of them is empty. No HDAC-inhibitor trial in this disease has measured any pharmacodynamic marker - not histone acetylation in blasts, not H2Bub, not HEXIM1, not CDA. So a negative trial cannot distinguish 'the mechanism is wrong' from 'the target was never engaged', and a positive one cannot say which of four things it did.

The third row is the one most likely to be read backwards. In a disease whose lesion is excess transcriptional elongation, an agent whose acute action is to release free P-TEFb is, for the first hours of each dose, pushing in the direction the disease already goes. gene-hexim1 records that the sequence is release first and HEXIM1 induction afterwards, so 'give an HDAC inhibitor to raise HEXIM1 and brake elongation' inverts the order of events. Nobody has looked for this in a leukaemia cell.

Why the ratio, not the enzyme, is the quantity that matters

If acetyl-lysine occupancy is a writer/eraser steady state, several things follow that the single-enzyme framing hides:

  • Inhibiting an eraser and activating a writer are the same intervention on the mark, and completely different interventions on everything else those enzymes do. They have different toxicity and different resistance.
  • A cell can restore occupancy by moving either side. Resistance to an HDAC inhibitor could be writer downregulation with the eraser untouched, which no target-site resistance assay would detect. Nothing in this disease has looked.
  • A dependency screen on either enzyme alone under-reads the system. Knock out the writer and the eraser now has no substrate; knock out the eraser and the writer's output accumulates. The phenotypes are not symmetric and neither measures the balance.
  • Compartment beats catalysis, sometimes. The one complete cycle here is controlled by exporting the eraser from the nucleus, not by changing its activity 1. A catalytic inhibitor and a localisation change produce the same substrate outcome by unrelated routes, and only one is a druggable event.

The practical proposal this generates is exp-catalytic-dead-versus-degrader-test-of-class-iia-hdac-function-in-kmt2a-r-all: measure both halves at the same loci in the same cells.

What the eraser half turned out to contain, once written

The brief for this beat expected the eraser half to look like the writer half - a set of enzymes with myeloid mouse-model results and no lineage-matched comparator. It does not. The eraser literature in this disease is, unexpectedly, the more lymphoid and the more clinical of the two halves:

  • The index fusion silences an HDAC: KMT2A::AFF1 with EZH2 represses HDAC7 in t(4;11) infant B-ALL, and the proposed therapy is induction, not inhibition (gene-hdac7).
  • The only patient-cohort result in the whole acetylation layer is an eraser result: high HDAC9 marks KMT2A-rearranged paediatric leukaemia across 211 patients, independent of age, partner and lineage (gene-hdac9).
  • The only clinical trial in newly diagnosed infants is an eraser-directed agent (drug-vorinostat).
  • The only unbiased screen on primary infant ALL cells returned this drug class (drug-romidepsin).
  • And the only per-enzyme genetic dissection in an MLL-fusion model is an eraser study (gene-hdac3).

So the graph's asymmetry was not merely an absence of entries. It was an absence of the half of the layer where the human, lymphoid and clinical evidence actually lives. The writer half is where the mechanism is; the eraser half is where the patients are.

Facts and figures
principlean acetyl-lysine is a writer/eraser steady state read continuously by a third protein, not a stable annotation; occupancy is set by the RATIO of writer to eraser activity
graph state before this beatwriters and one reader, no eraser of any kind; HDAC appeared in the text of 81 files and in the ids of two entries with no HDAC entry existing
complete cycles in the graph1
histone cycle statuswriter half measured at a locus (Tip60 on H2A.Z at Hoxa9); eraser half never measured at any fusion target locus under any HDAC inhibitor
pharmacodynamic marker measured in any hdac trial in this diseasefalse
consequence of thata negative trial cannot distinguish a wrong mechanism from an unengaged target, and a positive one cannot say which of four things it did
asymmetry findingthe eraser half of this layer is the more lymphoid, more human and more clinical half - it contains the index-fusion silencing result, the only 211-patient cohort, the only infant trial, the only unbiased primary-cell screen, and the only per-enzyme genetic dissection
the one complete cycle
{
  "writer": "gene-kat2b (PCAF)",
  "substrate": "AF9 K339 - a protein residue, not a histone",
  "reader_function": "AF9's own oligomerisation and TFIID contact",
  "eraser": "gene-hdac5 (HDAC4 also active in vitro)",
  "control_point": "CaMKII phosphorylation of HDAC5 at S259/S498, exporting the eraser from the nucleus - localisation, not catalysis",
  "present_in_the_fusion": false,
  "why_not": "K339 is N-terminal of AF9 residue 376; about 92% of KMT2A::MLLT3 breakpoints fall at or after 376"
}
mechanisms on record for hdac inhibition in this disease
[
  {
    "mechanism": "bulk histone acetylation",
    "agent": "assumed for the class",
    "is_histone_acetylation": true,
    "measured_in_a_patient": false
  },
  {
    "mechanism": "RNF20/RNF40/WAC suppression depleting H2B ubiquitination",
    "agent": "panobinostat",
    "is_histone_acetylation": false,
    "measured_in_a_patient": false
  },
  {
    "mechanism": "acute P-TEFb release from the 7SK snRNP",
    "agent": "SAHA / vorinostat",
    "is_histone_acetylation": false,
    "measured_in_a_patient": false
  },
  {
    "mechanism": "cytidine deaminase suppression potentiating cytarabine",
    "agent": "romidepsin",
    "is_histone_acetylation": false,
    "measured_in_a_patient": false
  }
]
corollaries
[
  "inhibiting an eraser and activating a writer are identical on the mark and unrelated in toxicity and resistance",
  "resistance could be writer downregulation with the eraser untouched, which no target-site assay would detect",
  "a single-enzyme dependency screen cannot measure a ratio",
  "compartment can substitute for catalysis - the one complete cycle here is controlled by exporting the eraser"
]
How this could bite us back
  • If occupancy is a ratio, then an eraser inhibitor can be defeated by writer downregulation - a resistance route that requires no mutation in any drug-binding site and that no assay in this field currently looks for.
  • Four distinct mechanisms are on record for HDAC inhibition in this disease and none has been measured in a patient. A trial designed around the assumed mechanism (histone acetylation) could succeed or fail for one of the other three and nobody would know which, which means neither outcome would inform the next trial.
  • The one complete acetylation cycle this graph holds is absent from the oncoprotein. Presenting the cycle framing as a source of targets, rather than as a source of discipline, would repeat the beat's own central error at a higher level of abstraction.
  • Because writers and erasers act on the same residues, combining a writer inhibitor with an eraser inhibitor - which the field will eventually propose - is pharmacologically incoherent on the mark while doubling the toxicity of both classes in an infant.
  • The vorinostat P-TEFb finding means an HDAC inhibitor may push the disease's core lesion in the wrong direction for the first hours of every dose. If real, that is a schedule problem invisible to any endpoint measured at the end of a cycle.
Still open
  1. What happens to acetylation at KMT2A-fusion target loci under an HDAC inhibitor? The single most obvious experiment in this layer, and nobody in this literature has published it.
  2. Is H2A.Z acetylation at Hoxa9 - the one locus-resolved writer result in the disease - reversed by any clinically used HDAC inhibitor?
  3. Can resistance to an HDAC inhibitor arise by writer downregulation rather than by anything happening to the eraser? If so, every resistance assay in the field is looking in the wrong place.
  4. In a disease of excess transcriptional elongation, does the acute P-TEFb release produced by vorinostat transiently accelerate the fusion programme in the first hours of each dose?
  5. Which of the four recorded mechanisms actually operated in the 50 infants who received vorinostat on NCT02553460? With no pharmacodynamic marker measured, the trial cannot say - and the assays to say it are cheap and exist.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PCAF acetylates AF9 at K339, reducing AF9 oligomerisation and TFIID interaction; HDAC5 deacetylates that site; and CaMKII-mediated phosphorylation of HDAC5 at S259 and S498, which controls its nuclear-cytoplasmic shuttling, is required for genotoxic-stress-dependent modulation of AF9 acetylation, with an HDAC5(S259A,S498A) phospho-defective mutant blocking the change - a complete writer-substrate-eraser cycle whose control point is the eraser's localisation rather than its catalysis.

    PMID 38750015DOI
    What this evidence cannot show

    HEK293T with ionising radiation, wild-type AF9, not a leukaemia cell and not a fusion protein. HDAC4 is also active on the site in vitro so the eraser is not uniquely HDAC5. The 'reader' in this cycle is the substrate's own oligomerisation state rather than a separate reader protein, which is a looser use of the term than the histone case and is flagged rather than glossed.

    Evidence that argues against this (1)
    • The one complete cycle this graph holds is on a residue absent from the fusion in the great majority of KMT2A::MLLT3 patients, so it demonstrates the principle without supplying a target. An entry that presented it as a therapeutic opportunity would be describing wild-type AF9 biology. Hu H et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    K339 lies N-terminal of AF9 residue 376 and is therefore absent from the fusion protein in approximately 92% of KMT2A::MLLT3 patients, as is the Poly-Ser domain whose oligomerisation the acetylation controls.

    PMID 37295550DOI
    What this evidence cannot show

    A review, cited for the breakpoint-distribution fact only. The 92% is a population statistic and a minority of KMT2A::MLLT3 breakpoints retain sequence N-terminal of residue 376. It says nothing about other partners.

    Evidence that argues against this (1)
    • If the acetylation site is absent from the fusion in most patients, then an acetylation CYCLE on that residue cannot be the mechanism in those patients - which cuts against the entry it is filed under rather than supporting it. The honest reading is that this residue is a wild-type AF9 mechanism that the fusion mostly deletes. Shi J et al. 2013
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Tip60 is recruited by MLL-AF10 and MLL-ENL fusions to the Hoxa9 locus where it acetylates H2A.Z, promoting Hoxa9 gene expression, and conditional deletion of Tip60 prevented the development of both leukaemias - establishing the writer half of a histone-acetylation cycle at a fusion target locus.

    Tip60 activates Hoxa9 and Meis1 expression through acetylation of H2A.Z, promoting MLL-AF10 and MLL-ENL acute myeloid leukemia. · Yamagata K, Shino M, Aikawa Y, Fujita S, Kitabayashi I · Leukemia 35(10):2840-2853 2021
    PMID 33967269DOI
    What this evidence cannot show

    Established within its reported system only: two myeloid mouse fusion models, no lymphoid arm, no KMT2A::AFF1 arm, no lineage-matched non-fusion comparator. Cited here specifically for the writer half of a locus-resolved histone-acetylation cycle, and the point of citing it in this entry is that the matching eraser measurement does not exist anywhere.

    Evidence that argues against this (1)
    • No corresponding eraser measurement exists. A search of this beat's literature found no paper measuring H2A.Z acetylation, or acetylation at any KMT2A-fusion target locus, under any HDAC inhibitor - so the writer half of the histone cycle is measured at a locus and the eraser half is measured only in bulk, if at all. The two halves have never appeared in the same experiment in this disease. NCBI E-utilities 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    HDAC inhibitors including SAHA acutely release the positive transcription elongation factor b (P-TEFb) from its inhibitory 7SK snRNP complex, with HEXIM1 induction following as downstream feedback - so an HDAC inhibitor in a disease driven by transcriptional elongation has a mechanism that is not histone acetylation and that acutely points in the direction the disease already goes.

    Histone deacetylase inhibitors (HDACis) that release the positive transcription elongation factor b (P-TEFb) from its inhibitory complex also activate HIV transcription. · Bartholomeeusen K, Fujinaga K, Xiang Y, Peterlin BM · The Journal of Biological Chemistry 288(20):14400-14407 2013
    PMID 23539624DOI
    What this evidence cannot show

    HIV latency in T-cell lines; no leukaemia cell and no KMT2A-rearranged system was studied. The biochemical action on the 7SK reservoir is what is carried here, and this graph's P-TEFb beat treats it as established for HMBA, SAHA and JQ1. Whether it matters in KMT2A-rearranged leukaemia is entirely unmeasured, and this entry states that as the open question rather than as a mechanism. Identifiers fetched from PubMed in this session, not copied from the sibling entry gene-hexim1.

    Evidence that argues against this (1)
    • Every compound known to release P-TEFb produces a TRANSIENT rise followed by reassembly, and three independent negative feedbacks operate on the axis, so the sign of any effect depends entirely on the timescale observed. An acute release measured over hours does not license a claim about chronic dosing over weeks. Bartholomeeusen K et al. 2012
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Romidepsin reduces expression of cytidine deaminase, the enzyme that deactivates cytarabine, and enhances the DNA damage response to cytarabine, enhancing cytarabine activity in MLL-rearranged infant ALL cells in vitro with a further reduction in leukaemia burden in xenografted mice - a fourth mechanism for this drug class in this disease, and one that is about drug metabolism rather than chromatin.

    Systematic chemical and molecular profiling of MLL-rearranged infant acute lymphoblastic leukemia reveals efficacy of romidepsin. · Cruickshank MN, Ford J, Cheung LC, Heng J, Singh S, Wells J, et al. · Leukemia 31(1):40-50 2017
    PMID 27443263DOI
    What this evidence cannot show

    Cell lines and xenografts. The cytidine deaminase mechanism is a gene-expression observation; the abstract does not report that CDA knockdown phenocopies the potentiation, so the causal link is not closed. Romidepsin is class I-selective across HDAC1, HDAC2 and HDAC3 and nothing here attributes to an enzyme.

    Evidence that argues against this (1)
    • Reducing cytidine deaminase and enhancing the DNA damage response are effects on the PARTNER drug's pharmacology and on repair, not on the acetylation cycle this entry is about. Cited here, it supports a combination rationale while saying nothing about whether acetylation turnover is the mechanism - and a pan-HDAC inhibitor cannot localise the effect to any enzyme in the cycle. Matthews GM et al. 2015
Connected to