KMT2A::AFF1 retains THD2 by position, so the HBO1 contact can apply to it — but the paper that established that contact tested MLL-ELL, MLL-AF6 and MLL-AF10, never the index fusion
also called does the index fusion retain THD2 · THD2 869-1124 · HBO1 contact applicability
A 2021 paper found these fusion proteins grab a histone-modifying machine through one stretch of protein, and the fusion this lab studies still contains that stretch, well before the point where the protein is cut. But the paper never tested our fusion, and the one it found most dependent on that stretch was a different fusion entirely.
The same thing, for a scientist
Takahashi 2021 (PMID 34431785, eLife, PMC8387021) showed MLL fusion proteins associate with the HBO1/KAT7 acetyltransferase complex through **trithorax homology domain 2 (THD2)**, *"mediated mainly by the ING4/5 and PHF16 subunits"* — PHF16 is JADE3's synonym, and JADE3, KAT7, ING4, MEAF6, BRPF1 and JADE2 are all in Skucha's 960-protein network, with KAT7 and JADE3 pan-fusion. **Positional answer: THD2 is retained in KMT2A::AFF1.** The paper's THD2 purification construct is *"the MLL fragment containing the residues 869–1124"*, and every leukaemogenic MLL fusion retains the CXXC domain at Q03164 **1147–1195** (UniProt, verified live). 1124 < 1147, so THD2 lies entirely N-terminal to the CXXC domain and therefore N-terminal to every breakpoint in the cluster region. **Functional answer: untested.** Figure 1B tested MLL-ELL, MLL-AF6 and MLL-AF10; MLL-ENL appears in the IP experiments. **MLL-AF4 appears once in the whole paper, in the Discussion, as context** — *"MLL-ELL associates with AEP similarly to other major MLL fusions such as MLL-AF4 and MLL-ENL"*. And the dependence is fusion-specific: *"Of the many MLL fusions, MLL-ELL **particularly** depended on the THD2-mediated association."*
Why the positional answer is worth writing down even though it is only positional
| absence class | never-measured |
|---|---|
| absence class basis | The functional half — whether KMT2A::AFF1 depends on THD2 for transformation — requires a deletion construct in a fusion nobody made for this purpose. Takahashi built MLL-ELL, MLL-AF6 and MLL-AF10 THD2 deletions and not an MLL-AF4 one. No archive can supply it. |
| shown for fusion | NO for KMT2A::AFF1 functionally. The positional retention argument applies to it; every functional result in the source paper is for MLL-ELL, MLL-AF6, MLL-AF10 or MLL-ENL. |
the positional argument
{
"thd2_construct": "MLL residues 869-1124, verbatim from the Figure 2A legend: \"A FLAG-tagged (f) GAL4 DNA-binding domain fused with the MLL fragment containing the residues 869-1124\"",
"cxxc_domain": "Q03164 1147-1195, UniProt ZF-CXXC feature, fetched live 2026-09-21",
"retention_premise": "Every leukaemogenic MLL fusion retains the CXXC domain — Cierpicki 2010 (PMID 20010842) states \"All leukemogenic MLL fusion proteins retain the CXXC domain\"",
"conclusion": "1124 < 1147, so the THD2 construct is entirely N-terminal to the CXXC domain, hence N-terminal to every breakpoint in the cluster region, hence retained in KMT2A::AFF1.",
"this_is_an_INFERENCE_not_a_measurement": "Nobody has shown THD2 is present and folded in a KMT2A::AFF1 protein. The argument is positional and rests on the CXXC retention premise, which is itself a literature generalisation rather than a per-patient observation."
}what the paper actually tested
{
"transformation_with_THD2_deletion": [
"MLL-ELL",
"MLL-AF6",
"MLL-AF10"
],
"immunoprecipitation": [
"MLL-ENL",
"full-length MLL"
],
"MLL-AF4": "NOT TESTED. Appears once in the paper, in the Discussion, as context for AEP association.",
"fusion_specificity_verbatim": "\"Of the many MLL fusions, MLL-ELL particularly depended on the THD2-mediated association with the HBO1 complex for leukemic transformation.\"",
"THD2_is_not_the_only_contact": "\"a small amount of HBO1 co-precipitated with the MLL proteins lacking the THD2 domain ... suggesting that there is a secondary binding domain for the HBO1 complex outside of THD2, which may account for the moderate effects of THD2 deletion from the full-length MLL fusion constructs\""
}- This entry supports no clinical action. It concerns whether a protein-protein contact described for other fusions can apply to this one, and its main finding is that the relevant experiment has not been done.
- HBO1/KAT7 inhibitors exist in development. Nothing here supports using one in KMT2A::AFF1 disease: the dependence was demonstrated for MLL-ELL, the index fusion was never tested, and the paper's own data show a second HBO1 contact outside the domain in question.
- Build an MLL-AF4 ΔTHD2 construct and run the HPC transformation assay Takahashi ran for MLL-ELL, MLL-AF6 and MLL-AF10. It is the direct experiment, the constructs and assay are published, and nothing about it is novel except the fusion.
- Where is the secondary HBO1 binding surface? The paper shows HBO1 still co-precipitates with THD2-deleted MLL and does not locate it. If it lies C-terminal to the breakpoint it is absent from every fusion, and if N-terminal it is another retained contact nobody has mapped.
- Does the HBO1 module's presence in Skucha's MLL-AF4 column survive a per-bait complex analysis? The complex enrichment in this graph pooled all seven baits, so 'HBO1 is in the network' is not yet 'HBO1 is in the MLL-AF4 column'.
- Give THD2 its own boundaries. Neither the paper nor UniProt states them; 869-1124 is a construct, and a domain-level entry for this graph needs the domain, not the fragment.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
THD2 lies within MLL residues 869-1124 and is therefore N-terminal to the CXXC domain at 1147-1195, which every leukaemogenic MLL fusion retains. KMT2A::AFF1 therefore retains THD2 by position, and the HBO1 contact described by Takahashi 2021 is structurally available to the index fusion.
HBO1-MLL interaction promotes AF4/ENL/P-TEFb-mediated leukemogenesis. · Takahashi S, Kanai A, Okuda H, Miyamoto R, Komata Y, Kawamura T, Yokoyama A · eLife 2021What this evidence cannot show
Full text read from PMC8387021 on 2026-09-21; the 869-1124 coordinates are the Figure 2A purification construct, which CONTAINS THD2 and may be larger than the domain itself — the paper does not give THD2's own boundaries, and UniProt does not annotate THD1 or THD2 for Q03164 at all. The conclusion is a positional inference, not an observation: nobody has shown THD2 present and correctly folded in a KMT2A::AFF1 protein. It also rests on the CXXC-retention generalisation, which is a literature statement rather than a per-breakpoint check.
Evidence that argues against this (1)
- Retention of a domain by position does not establish that it functions. The fusion replaces everything C-terminal to the breakpoint with AFF1, and domain function can depend on context the fusion destroys — the same reasoning this project applies when it refuses to transfer an isolated-domain affinity to the fusion. A positional argument is the weakest form of retention evidence and should not be upgraded by repetition. Nature Structural & Molecular Biology 2010
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The index fusion was never tested. Takahashi's transformation assays with THD2 deletion used MLL-ELL, MLL-AF6 and MLL-AF10; the immunoprecipitations used MLL-ENL and full-length MLL. MLL-AF4 appears exactly once in the paper, in the Discussion, as context for AEP association. Furthermore the dependence is explicitly fusion-specific — MLL-ELL 'particularly' depended on it — and THD2 is not the sole contact, because HBO1 still co-precipitated with MLL proteins lacking THD2, implying a second binding surface.
HBO1-MLL interaction promotes AF4/ENL/P-TEFb-mediated leukemogenesis. · Takahashi S, Kanai A, Okuda H, Miyamoto R, Komata Y, Kawamura T, Yokoyama A · eLife 2021What this evidence cannot show
Established by regex over the PMC8387021 full text for MLL-AF4, THD2 and deletion-construct terms, plus reading the surrounding passages. Figure panels and supplementary figure images were not inspected, so a construct shown only in a figure without being named in text or legend would have been missed. The title's 'AF4' refers to the AEP complex component, not to an MLL-AF4 fusion bait, which is an easy misreading of this paper.
Evidence that argues against this (1)
- Arguing the other way: the paper states the association itself holds for 'leukemic MLL fusion proteins ... in various human cell lines', and the HBO1 module is independently recovered in Skucha's MLL-AF4 column — JADE3 and KAT7 are pan-fusion across all seven baits including MLL-AF4. So the ASSOCIATION has support for the index fusion from a second, independent dataset even though the DEPENDENCE does not. Treating this entry as 'HBO1 is irrelevant to KMT2A::AFF1' would overcorrect. Nature Communications 2018
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
Answers a domain-retention question positionally: THD2 sits within MLL 869-1124, N-terminal to the CXXC domain at 1147-1195 that every leukaemogenic fusion retains, so the index fusion carries it. Whether it FUNCTIONS there is untested.
Supplies the mechanism behind JADE3's presence: PHF16 is JADE3's synonym, and Takahashi 2021 has MLL fusions reaching the HBO1 complex through THD2 'mediated mainly by the ING4/5 and PHF16 subunits'. THD2 is retained in the index fusion, so the route is structurally open to it.
Second, independent route to the same module. The complex analysis found HBO1/KAT7 the best-supported non-control complex in Skucha's network with KAT7 and JADE3 pan-fusion; this entry shows the domain that mediates that contact is retained in KMT2A::AFF1.