Oncofusion

start / graph / subtraction-kmt2a-phd3-bromo-cassette

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q-subtraction-kmt2a

Lost from the KMT2A half: PHD3 and the bromodomain are one cassette, not two modules (aa 1566-1765)

also called PHD3-Bromo cassette · MLL PHD3 bromodomain linker · CyP33 proline isomerisation

The part of KMT2A that reads an 'on' mark and the part next to it work as a single hinged unit, and a helper enzyme bends the hinge to flip the protein from switching genes on to switching them off. The fusion throws away the whole hinged unit.

The same thing, for a scientist

KMT2A PHD3 (UniProt Q03164, aa 1566-1627) and the bromodomain (aa 1635-1765) are separated by a short linker and behave as a single regulatory platform. CyP33/PPIE's PPIase domain isomerises a proline within the PHD3-Bromo linker, which disrupts the PHD3-Bromo interface and licenses the CyP33 RRM to bind PHD3; H3K4me3/2 and the CyP33 RRM engage different surfaces of PHD3 and can bind simultaneously as a ternary complex; and the MLL1-CyP33 interaction is required for repression of HOXA9 and HOXC8 in vivo [E1]. Re-inserting PHD3 alone into MLL-ENL restores CyP33 and HDAC1 recruitment and blocks haematopoietic stem cell immortalisation, while PHD1+PHD2 do not [E2]. The cassette is C-terminal to the breakpoint cluster region and is absent from every fusion. The entry's point is structural: treating 'PHD3' and 'the bromodomain' as two independently lost items in a domain ledger double-counts one mechanism, and the graph's existing bromodomain coordinates are wrong by ~65 residues.

Go deeper
Why the bromodomain does not get its own entry1 deeper

The previous beat's list of lost KMT2A modules treats PHD1, PHD2, PHD3 and 'bromodomain' as four separate deletions. Structurally that is not what the region is. Wang et al. show the PHD3-Bromo pair operating as one regulatory platform: the PPIase domain of CyP33 acts on a proline in the PHD3-Bromo linker, and the consequence of that isomerisation is to disrupt the PHD3-Bromo interface, freeing PHD3 to bind the CyP33 RRM 1. The bromodomain is not a bystander; it is the clamp that holds PHD3 shut until CyP33 opens it.

So the correct subtraction unit is PHD3 + linker + bromodomain, aa 1566-1765. A fusion that somehow retained PHD3 without the bromodomain would not be a fusion that kept the switch; it would be a fusion with a switch stuck open.

A second finding from the same paper matters for the 'brake' reading: H3K4me3/2 and CyP33-RRM target different surfaces of PHD3 and can bind simultaneously to form a ternary complex 1. That is a refinement of the older 'mutually inhibitory' model recorded elsewhere in the graph. PHD3 is not a toggle that reads either the activating mark or the repressor; it can hold both at once, which makes it a recruitment platform whose output depends on what else is present, not a binary switch.

Correction owed: the graph's bromodomain coordinates

gene-kmt2a (data.domains) gives the KMT2A bromodomain as 1703-1809. UniProt Q03164's feature table gives Bromo 1635-1765 3. The stored range is shifted C-terminal by roughly 65 residues and, as stored, it overlaps the HCF-binding motif at 1799-1802, which is not inside the bromodomain. gene-kmt2a carries a general domain_coordinate_caveat about approximate boundaries, but a 65-residue shift is larger than that caveat covers and it changes which features the domain contains. Reported, not edited - the file belongs to the King.

What is demonstrated and what is assumed

Demonstrated: CyP33 binding to PHD3 is required for repression of HOXA9 and HOXC8 in vivo 1. Re-inserting the third PHD finger alone into MLL-ENL restores CyP33 and HDAC1 recruitment, mediates CyP33-dependent HOXC8 downregulation, and blocks the fusion's serial-replating immortalisation potential; PHD1+PHD2 do not 2. This is the only lost KMT2A module with a published add-back phenotype.

Assumed and not demonstrated: that the loss of this cassette is why KMT2A fusions transform. The add-back experiment shows that putting PHD3 back breaks the fusion. It does not show that removing PHD3 from an otherwise intact KMT2A makes one. Those are different claims, and the literature supports only the first. See ana-phd3-addback-replication-audit for how thin the base under even the first claim is.

Contradicted in part: the MLL partial tandem duplication retains the whole PHD3 cassette and still deregulates Hox genes and causes disease, so the switch can be bypassed rather than only deleted 4.

Facts and figures
modulePHD3-Bromo regulatory cassette
uniprotQ03164
start1566
end1765
fate in fusionlost
regulatorPPIE / CyP33 (UniProt Q9UNP9, 301 aa, primary gene name PPIE confirmed)
demonstrated functional consequence of lossadd-back of PHD3 alone into MLL-ENL blocks serial-replating immortalisation (single study, 2008)
loss classlost in cis but the PHD function is supplied in trans; the only lost KMT2A module with a published in-cis add-back phenotype
supplied in transsupplied-in-trans (PMID 20541448: the fusion requires a prebound wild-type MLL1 complex for stable HoxA9 association)
sub features
PHD31,566PHD3: 1566H3K4me3 interaction region1,584H3K4me3 interaction region: 1584Bromodomain1,635Bromodomain: 1635
sub features — the numbers behind the chart
[
  {
    "name": "PHD3",
    "start": 1566,
    "end": 1627,
    "source": "UniProt Q03164"
  },
  {
    "name": "H3K4me3 interaction region",
    "start": 1584,
    "end": 1600,
    "source": "UniProt Q03164"
  },
  {
    "name": "Bromodomain",
    "start": 1635,
    "end": 1765,
    "source": "UniProt Q03164"
  }
]
effectors recruited when engaged
[
  "HDAC1"
]
correction owed
{
  "entry": "gene-kmt2a",
  "field": "data.domains[] entry named 'Bromodomain'",
  "stored": "start 1703, end 1809",
  "should_be": "start 1635, end 1765",
  "source": "UniProt Q03164 feature table (Domain: Bromo 1635-1765)",
  "read": "database record, fetched and read directly"
}
How this could bite us back
  • A chemically induced proximity agent that restores CyP33/HDAC1 recruitment to the fusion would, if it engages shared N-terminal sequence, do the same to wild-type KMT2A - and the wild-type allele's PHD3 is already competent, so the drug would be adding a brake where one exists and none where it matters.
  • Repression restored at HOXA9 also reaches CDKN1B, a tumour suppressor, in the same CyP33-dependent programme; the switch does not discriminate between targets you want off and targets you need on.
Still open
  1. If the fusion needs a prebound wild-type MLL1 complex at HoxA9, is the PHD3/CyP33 brake actually operating at that locus on the wild-type molecule while the fusion activates it? That would make the fusion exempt from a brake rather than free of one, and predicts that raising CyP33 should repress the locus even in a fusion-driven cell.
  2. If the bromodomain is the clamp that holds PHD3 shut, does a PHD3-only add-back (Chen 2008) restore a switch or install a permanently open reader? The two would look identical in a replating assay and would have opposite implications for a proximity-based therapeutic.
  3. Does the PHD3-CyP33-H3K4me3 ternary complex form on chromatin in a KMT2A-rearranged cell that still carries a wild-type allele, and is the wild-type allele therefore still capable of repressing the same HOXA9 locus the fusion is activating?
  4. Is there a reported patient breakpoint anywhere in exons 14-21, which would spare part of this cassette? If such fusions exist and behave differently, that is the natural experiment nobody has looked for.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The PPIase domain of CyP33 regulates MLL1 conformation by proline isomerisation within the PHD3-Bromo linker, disrupting the PHD3-Bromo interface and facilitating binding of MLL1-PHD3 to the CyP33 RRM; H3K4me3/2 and CyP33-RRM target different surfaces of MLL1-PHD3 and can bind simultaneously to form a ternary complex; and the MLL1-CyP33 interaction is required for repression of HOXA9 and HOXC8 genes in vivo.

    Pro isomerization in MLL1 PHD3-bromo cassette connects H3K4me readout to CyP33 and HDAC-mediated repression · Wang Z, Song J, Milne TA, Wang GG, Li H, Allis CD, Patel DJ · Cell 141(7):1183-1194 2010
    PMID 20541251DOI
    What this evidence cannot show

    ABSTRACT ONLY. Structure-function work on isolated MLL1 fragments (crystallography and NMR) plus cell-line experiments; 'in vivo' in the abstract means in cells, not in an animal. The ternary-complex result is from purified domains and does not establish occupancy on chromatin. Nothing in this study concerns a fusion protein.

    Evidence that argues against this (1)
    • The conformational mechanism is elegant and it concerns a cassette every major-BCR fusion has already deleted, since PHD3-bromo lies in exons 14-16, 3' of the breakpoint cluster. What the entry documents is therefore a regulatory layer the fusion never had rather than one it escaped, and those are different claims about restraint. Cell 2010
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The three-PHD-finger cassette is absent in all MLL fusion proteins; inserting the third PHD finger of MLL into MLL-ENL allows recruitment of CyP33 and then HDAC1, mediates CyP33-dependent downregulation of HOXC8, and blocks the haematopoietic stem cell immortalisation potential of the fusion in serial plating colony assays, whereas insertion of only the first and second PHD fingers 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, Wei W, Luo R, Thirman MJ, Zeleznik-Le NJ, Diaz MO · Cancer Research 68(15):6199-6207 2008
    PMID 18676843DOI
    What this evidence cannot show

    ABSTRACT ONLY (PMID fetched and title/abstract read in full; the figures and methods were not). Murine serial-replating colony assay, not an in-vivo leukemia endpoint; one fusion only (MLL-ENL); a gain-of-function insertion that creates a protein no patient carries. Inserting a folded 60-residue domain could impair the fusion sterically, although the CyP33/HDAC1 recruitment data argue against a purely steric explanation. The paper's own title asserts necessity, which is stronger than the replating experiment can establish.

    Evidence that argues against this (1)
    • The strongest counter to reading PHD3 loss as simply 'a brake removed'. Milne et al. place the PHD fingers inside an essential MLL1 recruitment domain and show that MLL1 fusion proteins, lacking the PHD fingers, REQUIRE prebinding of a wild-type MLL1 complex for stable HoxA9 association. On that evidence the PHD function is not removed from the locus at all - it is outsourced to the wild-type allele, and the fusion depends on it. ABSTRACT ONLY. Milne TA et al. 2010
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The UniProt Q03164 feature table places the KMT2A bromodomain at residues 1635-1765, PHD-type zinc finger 3 at 1566-1627, and an 'Interaction with histone H3K4me3' region at 1584-1600.

    What this evidence cannot show

    Accession verified by reading back the primary gene name (KMT2A) and sequence length (3,969 aa) from the fetched entry, not from a synonym-matching search. UniProt domain boundaries are themselves curatorial calls derived from Pfam/PROSITE profiles and can differ from a crystallographic construct by a few residues; they should not be treated as experimental measurements. The discrepancy flagged against gene-kmt2a is ~65 residues, well beyond that tolerance.

    Evidence that argues against this (1)
    • A UniProt feature table is a curated model rather than a measurement. Profile-based domain calls for KMT2A differ between UniProt, Pfam, SMART, InterPro and CATH, and the sibling atlas beat was corrected last cycle for quoting one source's range as though it were observed. Quoting coordinates without the spread states a choice as a fact. Cell 2010
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The Mll partial tandem duplication, which retains the full PHD3 region because it duplicates exons 5-11 in the absence of any partner gene, still induces aberrant Hox expression in vivo through specific epigenetic alterations in a knock-in mouse.

    Mll partial tandem duplication induces aberrant Hox expression in vivo via specific epigenetic alterations · Dorrance AM, Liu S, Yuan W, et al. · Journal of Clinical Investigation 116(10):2707-2716 2006
    PMID 16981007DOI
    What this evidence cannot show

    ABSTRACT ONLY. A mouse knock-in of a different KMT2A lesion; it shows Hox deregulation, not leukemia, in that model, and the PTD is a predominantly adult-AML lesion whereas the fusions this graph concerns are predominantly infant. It establishes that Hox deregulation does not require PHD3 loss, which is the relevant point, but it does not establish that PHD3 loss is irrelevant to the fusions.

    Evidence that argues against this (1)
    • The partial tandem duplication is the natural experiment for this entry and it runs against the simple reading: the PTD retains the full PHD3 region, has no partner gene at all, and still drives aberrant Hox expression in vivo. If PHD3 retention is compatible with Hox deregulation, losing PHD3 cannot be the explanation for it in the fusions. Journal of Clinical Investigation 2006
Connected to