The third category: modules lost from the fusion protein but supplied in trans by the wild-type KMT2A allele
also called wild-type allele supplies in trans · trans-complementation KMT2A · third category subtraction · MLL1 prebinding requirement
The graph has been sorting KMT2A's parts into two boxes, kept and thrown away. There is a third box: parts the fusion threw away but still needs, borrowed from the intact second copy of the gene that every patient keeps.
The same thing, for a scientist
A KMT2A translocation is heterozygous, so the leukemic cell retains an intact allele encoding every module this beat catalogues as lost. The graph's subtraction framework was two-category - retained or discarded - until this entry. The missing third category is **lost in cis, supplied in trans**, and the PHD fingers are its clearest instance: Milne et al. identify an essential MLL1 recruitment domain spanning the CXXC domain AND the PHD fingers, and show that MLL1 fusion proteins, lacking the PHD fingers, **require prebinding of a wild-type MLL1 complex** together with CXXC recognition of DNA for stable HoxA9 association [E4]. On that evidence, PHD loss is not only a brake removed; it is a function outsourced. Thiel et al. independently report the wild-type Mll allele as required for MLL-AF9 leukemogenesis and maintenance [E1], and MM-401, which blocks an interface only the wild-type allele possesses, selectively kills MLL-rearranged cells [E3]. Against all three, Chen et al. find endogenous MLL1 dispensable with MLL2/KMT2B mattering more [E2]. The category is real and under-treated; whether the wild-type allele is strictly required is genuinely unresolved.
The dispute, laid out without a verdict1 deeper
| Study | Model | Finding |
|---|---|---|
| Thiel et al. 2010 1 | mouse, MLL-AF9 | Menin recruits both wild-type MLL and MLL-AF9 to HOX loci; wild-type MLL controls distinct fusion-induced histone methylation; the wild-type Mll allele is required for leukemogenesis and for maintenance of transformed cells |
| Chen et al. 2017 2 | mouse, MLL-rearranged AML, 'rigorous, independent animal models' | Endogenous MLL1 is dispensable; Mll2 deletion alone significantly impairs survival of MLL-AF9-transformed cells; MLL1 and MLL2 collaborate by regulating distinct pathways, not redundantly |
| Cao et al. 2014 3 | human MLL-r cell lines, mouse cells | MM-401, which blocks MLL1-WDR5 assembly - an interface the fusion does not have - selectively blocks MLL-r cell proliferation |
The graph's mechanism-set-loss-neomorph-not-haploinsufficiency entry already records that 2 contradicts 1, and calls the requirement claim 'directly contradicted'. That is a defensible reading of 2 alone, but it does not account for 3, which is pharmacological evidence pointing the other way and which that entry does not cite. The honest status is unresolved, three-way, not settled against Thiel.
Why the disagreement may be real rather than a mistake
Paralogue buffering is the obvious candidate and 2 names it: MLL2/KMT2B. If MLL2 can substitute for some of MLL1's functions, then whether MLL1 looks required depends on MLL2 levels in the particular model and cell state - and 2 found Mll2 deletion alone had a significant impact, which is exactly the signature of a buffered dependency.
The graph already has hyp-paralogue-buffering-hides-dependencies. This dispute is that hypothesis's best worked example and should be linked to it.
Two axes, not one - and the framework gap this exposes1 deeper
Until this entry the graph classified every KMT2A module as retained or discarded. That is not exhaustive, and nothing anywhere in the graph systematically treats the remainder. Each lost module needs classifying on two independent axes:
Axis 1 - is the loss demonstrated to matter? demonstrated / tolerated / unmeasured.
Axis 2 - is the function supplied in trans by the intact allele? supplied-in-trans / not-recoverable / unknown.
The cross-product is what actually distinguishes the objects:
| supplied in trans | not recoverable in trans | |
|---|---|---|
| loss matters | the fusion outsources the function and depends on the wild-type allele for it - PHD fingers, per Milne 4 | the function is genuinely gone; nothing in this beat is confidently here |
| loss tolerated | irrelevant twice over - SET domain, on the genetic evidence | the fusion simply does not need it |
| unmeasured | most of the ledger: HBM, TAD, FYRN/FYRC, taspase1 sites, PHD4 | - |
The cell that matters most is the top-left. A module that is lost, load-bearing, and supplied in trans is not a subtraction at all in the functional sense; it is a dependency relocated from cis to trans. And a relocated dependency is a therapeutic opportunity of a completely different shape from a deleted one, because the thing the fusion needs is now a separate molecule that can be targeted without touching the fusion.
Milne et al. make the point themselves: 'wild-type MLL1 and oncogenic MLL1 fusion proteins have overlapping yet distinct recruitment mechanisms, this creates a window of opportunity that could be exploited for the development of targeted therapies' 4.
What this does to the PHD3 'lost brake' reading
If the fusion requires a prebound wild-type MLL1 complex for stable association at HoxA9 4, then the PHD fingers are present at the locus the fusion occupies - on the other molecule. The fusion has not created a PHD-free environment; it has created a PHD-free oncoprotein that docks onto chromatin already marked by a PHD-containing complex.
That is the strongest available counter-argument to reading PHD3 loss as simply 'a brake removed', and it belongs beside the add-back result rather than after it. Both can be true: re-inserting PHD3 in cis blocks transformation [see ana-phd3-addback-replication-audit] AND the wild-type PHD fingers are required in trans for the fusion to bind stably. The reconciliation would be that the fusion must be exempt from the brake while still using the platform the braked protein builds - which is a considerably more specific and more testable model than 'the fusion loses restraint'.
What this does to the subtraction framing
Three consequences, in order of how much they change the beat's conclusions:
1. 'Lost' is a statement about a molecule, 'absent' is a statement about a cell. The HBM is lost from the fusion; HCF-1 recruitment to KMT2A is not absent from the cell. Every entry in this beat has been written to respect that distinction, and the loss_class field in each records it.
2. A lost module can be a drug target. MM-401 and the taspase1/MLL-nc tool both act on modules the fusion does not have, in cells the fusion is driving. This inverts the usual assumption that the drug target must be in the retained half. The retained half is where fusion-selective targets must live; the lost half is where cell-required targets may live, at the price of hitting a gene under extreme constraint.
3. It weakens, but does not kill, the 'loss of restraint' reading. If the wild-type allele still supplies PHD3-mediated repression and PHD2-mediated self-limitation at the same loci, then the fusion has not removed those brakes from the locus - it has produced a molecule that is immune to them while leaving them operating on its wild-type counterpart. That is a subtly different and arguably more interesting claim: not 'the brake is gone' but 'the oncoprotein is exempt from it'.
| claim type | framing correction |
|---|---|
| dispute status | unresolved, three-way |
| untested stratifier | retention versus loss of the wild-type KMT2A allele in patient samples, as a predictor of response to agents acting on wild-type-only interfaces |
| framework gap | The graph's subtraction framework was two-category (retained / discarded) until cycle 29. The third category - lost in cis, supplied in trans by the intact allele - had no systematic treatment anywhere in the graph. This entry names it; it is not yet applied across the partner-side subtraction beat. |
positions
[
{
"position": "wild-type allele required",
"evidence": "PMID 20159607",
"model": "mouse MLL-AF9"
},
{
"position": "wild-type allele dispensable; MLL2 matters more",
"evidence": "PMID 28609655",
"model": "independent mouse models"
},
{
"position": "pharmacological block of a wild-type-only interface kills MLL-r cells",
"evidence": "PMID 24389101",
"model": "human MLL-r lines"
},
{
"position": "the fusion requires a prebound wild-type MLL1 complex for stable HoxA9 association",
"evidence": "PMID 20541448",
"model": "domain-dissection recruitment assays at HoxA9"
}
]modules present in the cell but absent from the fusion
[ "PHD1-PHD4 and the bromodomain", "HCF-binding motif (1799-1802)", "transactivation domain (2829-2883)", "taspase1 cleavage sites (2666, 2718)", "FYRN/FYRC clamp", "WIN motif (3762-3767) and WRAD assembly", "SET domain (3829-3945)", "MOF recruitment contact" ]
two axis classification
{
"axis_1_loss_demonstrated_to_matter": [
"demonstrated",
"tolerated",
"unmeasured"
],
"axis_2_supplied_in_trans_by_wild_type_allele": [
"supplied-in-trans",
"not-recoverable",
"unknown"
],
"clearest_supplied_in_trans_case": "PHD fingers - the fusion requires prebinding of a wild-type MLL1 complex for stable HoxA9 association (PMID 20541448)",
"note": "A module that is lost, load-bearing AND supplied in trans is a dependency relocated from cis to trans, not a subtraction in the functional sense."
}- If the wild-type allele is required, then agents acting on it are effective for a reason that predicts on-target toxicity in normal haematopoiesis and development - the same reason KMT2A carries LOEUF 0.07 and heterozygous loss causes Wiedemann-Steiner syndrome.
- If the wild-type allele is dispensable and MLL2/KMT2B is the real dependency, then a whole class of proposed KMT2A-directed strategies is aimed at the wrong paralogue, and the menin-interface strategy's advantage over them widens.
- Cases that have lost the wild-type allele would be intrinsically resistant to any wild-type-directed agent, and nobody is measuring allele status, so such failures would appear as unexplained non-response.
- Apply the two-axis classification to the PARTNER-side subtraction beat as well. Nothing there has been asked whether a discarded partner module is supplied in trans by the untranslocated partner allele, which is exactly the same question and has the same answer structure.
- Which patients retain the wild-type KMT2A allele? Chen et al. note it is 'most commonly, but not always' retained. Nobody has used allele retention as a stratifier for any therapy, although it is directly readable from existing sequencing data.
- Can the Thiel/Chen dispute be resolved by MLL2 status? If MLL2 buffers MLL1, then MLL1 requirement should track inversely with MLL2 expression across models - a reanalysis of existing data, not a new experiment.
- Does the wild-type allele actively repress at fusion target loci through PHD3/CyP33 while the fusion activates the same loci? If so, the fusion is not operating in the absence of a brake; it is competing with one, and the ratio of fusion to wild-type occupancy becomes the quantity that matters.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Menin recruits both wild-type MLL and the oncogenic MLL-AF9 fusion protein to HOX gene loci to activate transcription; wild-type MLL both catalyses histone methylation at key target genes and controls distinct MLL-AF9-induced histone methylation; and the wild-type Mll allele is required for MLL-AF9-induced leukemogenesis and for maintenance of MLL-AF9-transformed cells.
MLL-AF9-induced leukemogenesis requires coexpression of the wild-type Mll allele · Thiel AT, Blessington P, Zou T, Feather D, Wu X, Yan J, Zhang H, Liu Z, Ernst P, Koretzky GA, Hua X · Cancer Cell 17(2):148-159 2010What this evidence cannot show
ABSTRACT ONLY. Mouse MLL-AF9 model. Confidence is set to 'moderate' rather than 'strong' NOT because the sourcing is weak but because the finding is directly contradicted by E2 in independent models; the citation itself is solid and resolves correctly.
Evidence that argues against this (1)
- Independent animal models found endogenous MLL1 dispensable for MLL-rearranged leukemia, with the paralogue MLL2 mattering more. Chen Y et al. 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Using rigorous, independent animal models, endogenous MLL1 is dispensable for MLL-rearranged leukemia; deleting the closest paralogue Mll2 alone had a significant impact on survival of MLL-AF9-transformed cells and additional Mll1 loss further reduced viability and proliferation; MLL1 and MLL2 collaborate by regulating distinct pathways rather than redundantly.
MLL2, Not MLL1, Plays a Major Role in Sustaining MLL-Rearranged Acute Myeloid Leukemia · Chen Y, Anastassiadis K, Kranz A, Stewart AF, Arndt K, Waskow C, Yokoyama A, Jones K, Neff T, Lee Y, Ernst P · Cancer Cell 31(6):755-770.e6 2017What this evidence cannot show
ABSTRACT ONLY. Mouse AML models; the abstract notes the wild-type allele is 'most commonly, but not always, retained' in patients, which is itself an under-used stratifier. Does not address lymphoid KMT2A::AFF1 disease. A genetic deletion tests requirement over a long timescale and can be rescued by adaptation in ways an acute pharmacological block cannot.
Evidence that argues against this (1)
- This is the paper that contradicts the result five other entries in this beat rest on, and the two have never been placed side by side. Cancer Cell 2010 reports that MLL-AF9 leukemogenesis REQUIRES the wild-type Mll allele; Cancer Cell 2017 reports that endogenous MLL1 is DISPENSABLE and that MLL2 sustains the disease. Both are in this graph. Whichever is right, the beat cannot treat the wild-type-requirement result as settled, and this entry's own framing - lost from the protein but not from the cell - depends on which answer holds. Cancer Cell 2017
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MM-401 blocks the MLL1-WDR5 interaction and hence MLL1 core-complex assembly and H3K4 methyltransferase activity, and specifically blocks proliferation of MLL-rearranged cells by inducing cell-cycle arrest, apoptosis and myeloid differentiation without general toxicity to normal bone marrow or non-MLL cells.
Targeting MLL1 H3K4 methyltransferase activity in mixed-lineage leukemia · Cao F, Townsend EC, Karatas H, et al. (Wang S, Dou Y, senior authors) · Molecular Cell 53(2):247-261 2014What this evidence cannot show
ABSTRACT ONLY. The inference that MM-401 must be acting on the wild-type allele - because the fusion has no WIN motif to block - is this Queen's, combining this paper with the WIN-motif position (UniProt Q03164, aa 3762-3767) and the breakpoint cluster position. The paper does not make that argument, and an off-target contribution is not excluded by the abstract.
Evidence that argues against this (1)
- MM-401 blocks MLL1 core-complex assembly via WDR5, so it removes a shared COMPLEX function rather than testing whether the cell still has the modules the fusion lost. Its selectivity for MLL1 over other family members is also the point at issue: if MLL2 is what sustains MLL-rearranged AML, an MLL1-selective probe is testing the wrong paralogue. Cancer Cell 2017
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A systematic analysis of MLL1 domain recruitment activity identifies an essential MLL1 recruitment domain that includes the CXXC domain and the PHD fingers, controlled by direct interactions with the PAF elongation complex and H3K4me2/3; MLL1 fusion proteins lack the PHD fingers and require prebinding of a wild-type MLL1 complex, plus CXXC-domain recognition of DNA, for stable HoxA9 association; wild-type MLL1 and oncogenic MLL1 fusion proteins therefore have overlapping yet distinct recruitment mechanisms.
Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis · Milne TA, Kim J, Wang GG, Stadler SC, Basrur V, Whitcomb SJ, Wang Z, Ruthenburg AJ, Elenitoba-Johnson KSJ, Roeder RG, Allis CD · Molecular Cell 38(6):853-863 2010What this evidence cannot show
ABSTRACT ONLY. Domain-dissection and recruitment assays at a single locus (HoxA9) in cell lines and mouse cells; 'prebinding of a wild-type MLL1 complex' is an inference from the recruitment data in that system, not a direct imaging or single-molecule observation of sequential occupancy. It concerns recruitment and stable association, not transformation, so it establishes that the fusion depends on wild-type MLL1 for stable binding at HoxA9, not that it depends on it for leukemogenesis - that is E1's claim and E2 disputes it.
Evidence that argues against this (1)
- A recruitment domain is a statement about targeting, and this entry needs a statement about function-in-trans. Establishing that CXXC plus PHD fingers are required for MLL1 to reach HOXA9 does not establish that the wild-type allele supplies those activities TO the fusion, which is the claim being made. Molecular Cell 2010
a lost KMT2A interface with no prior entry in this graph
- related to →KMT2A / MLL1 — the gene that breaks
the untranslocated allele still encodes every module catalogued as lost
the Thiel/Chen dispute is that hypothesis's best worked example: MLL2 buffers MLL1
that entry calls the wild-type requirement 'directly contradicted'; MM-401 is pharmacological evidence the other way that it does not cite, so the status is unresolved rather than settled
- supports →Lost from the KMT2A half: the WIN motif and the whole WRAD interface (aa 3762-3767, Arg3765)
the clearest instance: a wild-type-only interface that is nonetheless a validated dependency in KMT2A-r cells
menin recruits both the fusion and the wild-type protein to the same HOX loci
the second axis of the ledger is defined by that entry's third category
- refutes →Lost from the KMT2A half: PHD3 and the bromodomain are one cassette, not two modules (aa 1566-1765)
the PHD function is outsourced, not removed: the fusion requires a prebound wild-type MLL1 complex for stable HoxA9 association
- related to →Subtraction is a partner class, not a curiosity — and the class is defined by compartment, not by sequence
the same two-axis test should be applied to the partner-side subtraction beat, where it has never been asked
CXXC DNA recognition plus prebound wild-type complex are jointly required for stable fusion association at HoxA9
the third category is what forces 'exemption' over 'absence'