What is lost: the PHD3/CYP33 switch that lets wild-type KMT2A turn itself off
also called PHD3 CYP33 switch · PPIE prolyl isomerase MLL · MLL repression domain · activator-to-repressor switch
Normal KMT2A has an off-switch: a small finger domain that, when a partner enzyme twists a nearby bond, flips the protein from switching genes on to switching them off. Every leukemic fusion throws that finger away, so the protein can only ever say 'on'.
The same thing, for a scientist
Wild-type KMT2A carries a three-PHD-finger cassette plus bromodomain between the CXXC domain and the C-terminus. PHD3 binds H3K4me3 preferentially and also binds the RRM domain of CYP33/PPIE, a prolyl isomerase — and these two bindings are mutually inhibitory, making PHD3 a molecular switch between KMT2A-mediated activation and repression. CYP33 isomerises a proline in the PHD3-bromodomain linker and downregulates KMT2A targets including HOXA9 in an isomerase-dependent manner, with reduced H3K4me3 and H3 acetylation. The PHD cassette is absent from every leukemogenic fusion. Re-inserting PHD3 alone into MLL-ENL restores CYP33/HDAC1 recruitment and blocks the fusion's immortalisation potential; re-inserting PHD1+PHD2 does nothing. The switch is therefore not incidental collateral of the breakpoint — its removal is required.
The switch, biochemically1 deeper
CYP33 (PPIE) mediates downregulation of KMT2A targets — HOXC8, HOXA9, CDKN1B and C-MYC — in a proline-isomerase-dependent manner, correlating with reduced H3K4me3 and reduced H3 acetylation. Structures of PHD3 and of the CYP33 RRM show that PHD3 binds H3K4me3 (preferentially) and the CYP33 RRM at distinct sites, and the two bindings are mutually inhibitory — so PHD3 is a molecular switch for the transition between activation and repression 1.
The downstream repressive machinery is also mapped: HDAC1 binds the KMT2A repression domain, and CYP33 binding to the adjacent PHD domain potentiates that binding. The Polycomb proteins HPC2 and BMI-1 and the corepressor CtBP also bind the repression domain, and exogenous BMI-1 potentiates its activity 2.
Note a detail in 2 that is easy to miss and important: the repression domain itself is retained in the leukemogenic fusion and is required for transforming activity. So the fusion keeps the repression platform while losing the switch that engages it — which is precisely why the fusion can still perform targeted repression (e.g. of BIM via CBX8/PRC1) while being unable to shut its own activating function off.
A third layer: RNA competes for the same interface
CYP33 also binds AU-rich RNA. RNA-SELEX identified an enriched motif (AAUAAUAA) and aptamers with affinities comparable and competitive with the CYP33-PHD3 interaction; NMR shows overlapping but non-identical RNA and PHD3 interfaces, so RNA can disrupt the repressive CYP33-KMT2A complex 3.
This is in-vitro work and its cellular relevance is unestablished, but it is mechanistically suggestive: if local RNA can titrate CYP33 away from PHD3, then transcription itself would feed back to keep KMT2A in the activating state — a positive feedback loop that the fusion has hard-wired by deleting the switch entirely.
The functional proof: putting PHD3 back kills the fusion
The decisive experiment is a rescue-by-re-insertion. The three-PHD-finger cassette is absent in all fusion proteins. Re-inserting PHD3 into MLL-ENL restores CYP33 and HDAC1 recruitment and mediates CYP33-dependent HOXC8 downregulation. And adding either the whole PHD cassette or the third PHD finger alone into MLL-ENL blocks the haematopoietic-stem-cell immortalisation potential of the fusion in serial-replating colony assays — while inserting only PHD1 and PHD2 has no such effect 4.
That is an unusually clean domain-resolution result. It means:
- PHD3 loss is necessary, not incidental. A translocation that spared PHD3 would produce a much weaker or non-oncogenic protein. This is a mechanistic explanation for why the KMT2A breakpoint cluster region sits where it does (see
breakpoint-kmt2a-bcr-architecture): breaks C-terminal to PHD3 would not transform. - The fusion is a constitutive transactivator by construction — the authors' own framing: loss of the PHD fingers plus gain of the partner's activation domain makes it one.
- There is a therapeutic idea buried here that nobody has pursued. If restoring the PHD3 switch disables the fusion, then a molecule or construct that recruits CYP33/HDAC1 to the fusion — a chemically induced proximity approach, not an inhibitor — should phenocopy PHD3 re-insertion. This is a genuinely under-explored modality for this disease and it targets a fusion-adjacent vulnerability rather than a shared surface.
Superseded: the switch is not mutual exclusivity, it is an RNA-triggered allosteric relay
This entry carried, with no counter-note, the claim that H3K4me3 binding and Cyp33-RRM binding to PHD3 are mutually inhibitory. That was the standing model from Park 2010 and it was overturned in 2023.
PMID 37075125 (Blatter et al., Sci Adv 2023) determined solution structures of the Cyp33 RRM free, bound to RNA, bound to MLL1 PHD3, and bound to both MLL1 and H3K4me3 - the last being a trimolecular complex that strict exclusivity forbids (PDB 7ZEZ). From the abstract: "a conserved alpha helix, amino-terminal to the RRM domain, adopts three different positions facilitating a cascade of binding events. These conformational changes are triggered by Cyp33 RNA binding and ultimately lead to MLL1 release from the histone mark."
So the switch input is RNA, not competition for PHD3. The exclusive pair is RNA versus PHD3 on the RRM beta-sheet; the release of MLL1 from H3K4me3 is downstream of that, through an allosteric relay in the N-terminal helix.
What survives from Park 2010: the Kd of 14.7 uM for the RRM-PHD3 interaction and the V1617 hot spot. What does not survive is the exclusivity conclusion built on them.
This entry is about a switch the fusion LOSES, since PHD3 is discarded. The correction does not change that, but it changes what is lost: not a simple competitive toggle, but an RNA-sensing relay - which is a more specific thing to have removed, and raises the question of which RNA.
| chain | mechanism-kmt2a-fusion-causal-chain |
|---|---|
| retained repressive element | KMT2A repression domain (retained in fusion; required for transformation) |
| mutual exclusivity | SUPERSEDED 2026-09-20. Strict mutual exclusivity between H3K4me3 and Cyp33-RRM binding at PHD3 does not hold: a TRIMOLECULAR Cyp33-RRM / MLL1-PHD3 / H3K4me3 structure exists (PDB 7ZEZ, PMID 37075125). The genuinely exclusive pair is RNA versus PHD3 on the RRM beta-sheet, and RNA binding is the switch input that releases MLL1 from the histone mark. |
| domain resolution | PHD3 alone is sufficient to block immortalisation when re-inserted; PHD1+PHD2 are not |
| phd3 loss status | necessary for transformation, not incidental to the breakpoint |
| untested modality | chemically induced proximity to recruit CYP33/HDAC1 to the fusion, phenocopying PHD3 re-insertion |
| mutual exclusivity previous text | H3K4me3 binding and CYP33-RRM binding to PHD3 are mutually inhibitory |
lost domains
[ "PHD1", "PHD2", "PHD3", "bromodomain", "FYRN", "FYRC", "TAD", "SET" ]
switch components
{
"reader": "PHD3 binds H3K4me3",
"regulator": "CYP33/PPIE RRM binds PHD3; prolyl isomerase acts on PHD3-bromodomain linker",
"effectors": [
"HDAC1",
"HPC2",
"BMI-1",
"CtBP"
]
}- A proximity-based approach that recruits HDAC1/CYP33 would also act on wild-type KMT2A if it engages shared N-terminal sequence, reintroducing the selectivity problem the strategy was meant to avoid.
- The repression domain is retained and required for transformation, so a therapy that broadly disables KMT2A-associated repression could paradoxically impair the fusion's own oncogenic program or de-repress other loci unpredictably.
- CYP33/PPIE is a general prolyl isomerase and spliceosome-associated protein; manipulating it has wide consequences beyond KMT2A.
- Restoring repression at HOXA9 also represses CDKN1B, which is tumour-suppressive — the switch does not distinguish good targets from bad ones.
- Can the PHD3 switch be restored pharmacologically? A chemically induced proximity agent that recruits CYP33 or HDAC1 to the fusion should phenocopy PHD3 re-insertion. This is, in this Queen's view, the most interesting unexplored modality in the whole KMT2A programme, because it exploits what the fusion LACKS rather than what it shares with wild-type KMT2A.
- Does CYP33/PPIE expression level vary across KMT2A-r subgroups, and does low CYP33 phenocopy PHD3 loss in cells that retain wild-type KMT2A? If so, CYP33 status could modulate the wild-type allele's contribution.
- Is the PHD3 re-insertion result reproducible in vivo and with KMT2A::AFF1 rather than KMT2A::MLLT1? The domain-resolution claim rests on a single 2008 in-vitro immortalisation study.
- Does the retained repression domain explain the fusion's targeted repression of genes such as BIM, and can that arm be blocked selectively while leaving activation alone?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
CYP33/PPIE downregulates KMT2A targets HOXC8, HOXA9, CDKN1B and C-MYC in a proline-isomerase-dependent manner with reduced H3K4me3 and H3 acetylation; structural work shows KMT2A PHD3 binds H3K4me3 preferentially and binds the CYP33 RRM at a distinct site, and the two interactions are mutually inhibitory, making PHD3 a molecular switch between KMT2A-mediated activation and repression.
The PHD3 domain of MLL acts as a CYP33-regulated switch between MLL-mediated activation and repression · Park S, Osmers U, Raman G, Schwantes RH, Diaz MO, Bushweller JH · Biochemistry 49(31):6576-6586 2010What this evidence cannot show
Structural and biochemical work on isolated domains plus cell-line expression experiments; the switch model is an inference from mutual exclusivity of binding rather than a direct observation of the transition in chromatin. CYP33 overexpression is the main cellular perturbation used.
Evidence that argues against this (1)
- The transcriptional half of this is overexpression work: CYP33/PPIE was supplied above its endogenous level and target genes followed. Mutual inhibition between H3K4me3 binding and CYP33-RRM binding is measured on isolated domains and peptides, so the switch is demonstrated as a biochemical possibility rather than as a regulated event at physiological stoichiometry. Cell Reports 2014
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
HDAC1 binds the KMT2A repression domain and CYP33 binding to the adjacent PHD domain potentiates that binding; the Polycomb proteins HPC2 and BMI-1 and the corepressor CtBP also bind the repression domain, exogenous BMI-1 potentiates its activity, and the repression domain is retained in the leukemogenic fusion protein and required for transforming activity.
MLL repression domain interacts with histone deacetylases, the polycomb group proteins HPC2 and BMI-1, and the corepressor C-terminal-binding protein · Xia ZB, Anderson M, Diaz MO, Zeleznik-Le NJ · Proceedings of the National Academy of Sciences USA 100(14):8342-8347 2003What this evidence cannot show
Co-immunoprecipitation and reporter assays in cell lines with overexpressed constructs; binding is shown but stoichiometry, chromatin context and endogenous relevance are not established.
Evidence that argues against this (1)
- The repression domain binds HDAC1, HPC2, BMI-1 and CtBP — four partners with different activities at one region, which is the profile of a promiscuous low-affinity surface as much as of a regulated module. The crucial claim, that the repression domain is retained in the leukaemogenic fusion, is a statement about the construct: retention of a domain does not establish that its partners still assemble on it in the fusion's context. Cell Reports 2014
- E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper
CYP33 binds AU-rich RNA via an extended interface overlapping but not identical to its PHD3-binding interface; SELEX-derived RNA aptamers with an AAUAAUAA-enriched motif bind with affinities comparable to and competitive with the CYP33-KMT2A-PHD3 interaction and can disrupt the gene-repressive CYP33-KMT2A complex in vitro.
Cyp33 binds AU-rich RNA motifs via an extended interface that competitively disrupts the gene repressive Cyp33-MLL1 interaction in vitro · Lloyd NR, Wuttke DS · PLoS ONE 16(2):e0237956 2021What this evidence cannot show
Explicitly in vitro throughout — NMR, SELEX and binding competition with purified components. No cellular or leukemia data. Whether endogenous RNA reaches the local concentrations required is unknown, so this is a mechanistic possibility rather than a demonstrated regulatory layer.
Evidence that argues against this (1)
- SELEX-derived aptamers are selected in vitro to bind tightly and are not ligands the cell contains. Showing that an artificial AU-rich RNA competes with the CYP33-PHD3 interaction establishes that the interface is RNA-accessible, not that an endogenous RNA regulates the switch — and AU-rich elements are abundant enough that an overlapping interface is weak evidence of a dedicated mechanism. Cell Reports 2014
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The three-PHD-finger cassette is absent in all KMT2A fusion proteins; re-inserting PHD3 into MLL-ENL restores CYP33 and HDAC1 recruitment and CYP33-dependent HOXC8 downregulation, and inserting either the full PHD cassette or the third PHD finger alone blocks the haematopoietic-stem-cell immortalisation potential of the fusion in serial-replating colony assays, whereas inserting only PHD1 and PHD2 has no such effect.
Loss of MLL PHD finger 3 is necessary for MLL-ENL-induced hematopoietic stem cell immortalization · Chen J, Santillan DA, Koonce M, et al. (Zeleznik-Le NJ, Diaz MO, senior authors) · Cancer Research 68(15):6199-6207 2008What this evidence cannot show
Murine serial-replating immortalisation rather than an in-vivo leukemia endpoint, and a gain-of-function rescue-by-re-insertion design in which the engineered protein differs from any natural fusion. KMT2A::MLLT1 only. Re-inserted PHD3 could impair function by steric or folding effects rather than by restoring the switch, although the CYP33/HDAC1 recruitment data argue against that.
Evidence that argues against this (1)
- Loss of the PHD cassette is not necessary for KMT2A-driven leukaemia in general, which limits how much of the disease this switch can explain. The MLL partial tandem duplication - a duplication of exons 5 to 11 in the absence of any partner gene, found in roughly 4-7% of cytogenetically normal AML and associated with poor prognosis - RETAINS the full PHD3/CYP33 module, and a knock-in Mll(PTD/WT) mouse still develops aberrant Hox expression through specific epigenetic alterations. A KMT2A lesion that keeps the off-switch and still deregulates HOX genes shows the switch can be bypassed rather than only deleted. Dorrance AM et al. 2006
- E-trimolecularstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article
A trimolecular Cyp33-RRM / MLL1-PHD3 / H3K4me3 complex has been determined (PDB 7ZEZ), so H3K4me3 and Cyp33 binding at PHD3 are not mutually exclusive. The exclusive pair is RNA versus PHD3 on the RRM beta-sheet, and RNA binding triggers an allosteric cascade through an N-terminal helix that releases MLL1 from the histone mark.
RNA binding induces an allosteric switch in Cyp33 to repress MLL1-mediated transcription · Blatter M, Meylan C, Clery A, et al. · Science Advances 2023What this evidence cannot show
Read from the abstract 2026-09-20; the structures are solution NMR of isolated domains, not the full-length proteins or chromatin. 'Release from the histone mark' is demonstrated structurally and biochemically, not in cells. The identity of the physiological RNA input is not established by this work.
Evidence that argues against this (1)
- A ternary structure shows the three components CAN co-exist; it does not show that the ternary state is populated at cellular concentrations, nor that the older competition experiments were misread. Park 2010's competition data still need an account, and this entry now asserts the newer model without having reconciled the older observations against it. Blatter M et al. 2023
Counter-evidence to this entry's own PHD1 mechanism: CYP33/PPIE binds MLL through its RRM to PHD3, not PHD1 (Park 2010, PMID 20677832), and inserting PHD1+PHD2 into MLL-ENL has no effect on immortalisation while PHD3 alone abolishes it (Chen 2008, PMID 18676843). An intron-11 break cannot plausibly act by impairing CYP33 binding via a truncated PHD1.
supplies the reader affinity and the aromatic-cage residue that entry lacked
that entry asserts mutual exclusivity; PMID 20541251 concludes a simultaneous ternary complex
- part of →KMT2A / MLL1 — the gene that breaks
the brake the break throws away
modifier, not a step - what the fusion loses rather than gains
- ← supportsLost from the KMT2A half: PHD3 and the bromodomain are one cassette, not two modules (aa 1566-1765)
extends it: PHD3 and the bromodomain are one cassette, and the switch is a ternary platform rather than a mutually exclusive toggle
caps the necessity claim: restoration blocks replating in one fusion; that is not evidence that PHD3 loss is required