AKAP95 IDR1 co-condenses with the KMT2A N-terminal fragment - the one AKAP95 contact that lands inside the region every KMT2A fusion retains
also called AKAP95-MLL1T · AKAP8-KMT2A N-terminus · AKAP95 IDR1-MLL1T co-condensate
The part of KMT2A that survives in every leukaemia fusion is a floppy, shapeless stretch. AKAP95 has a floppy stretch of its own that sticks to it, and the two form droplets together inside the nucleus - which is how AKAP95 gets pulled onto the genes the fusion switches on.
The same thing, for a scientist
**This is the contact that transfers.** Endogenous AKAP95 co-immunoprecipitates with over-expressed FLAG-HA-**MLL1T** (the MLL1 translocated fragment) from 293T cells; in HeLa cells carrying an endogenous **Flag-mEGFP-AKAP95 knock-in**, AKAP95 foci partially co-localise with mCherry-MLL1T; and in rescued **THP-1** (KMT2A::MLLT3) cells, endogenous MLL1 N-terminal fragments partially co-localise with AKAP95 WT but **not** with the zinc-finger or phase-separation mutants [E1]. Mapping a GFP-fragment series (aa1-100, 101-390, 391-550, 551-692) against mCherry-MLL1T identifies **IDR1, aa 101-390**, as the only fragment that co-localises and forms large nuclear aggregates; replacing six tyrosines in **aa 101-210** with serine (the PS-mutant) abolishes the co-condensation [E1]. The proposed driver is charge complementarity: AKAP95 IDR1 has a net negative charge per residue and MLL1T a net positive one, with similar fractions of charged residues overall [E1]. **Note what this is not: it is not a defined binding site, there is no Kd, no structure, and the MLL1T construct boundaries are not stated in the main text.**
Why this one crosses the fusion boundary and the DPY30 one does not
KMT2A::AFF1 and every other KMT2A fusion keep the N-terminal portion of KMT2A and lose the SET domain. This project's standing structural question - is the contact in a region the fusion keeps? - has a clean answer here for once: MLL1T is, by definition, the retained fragment. The partner is the piece that survives the translocation.
Contrast the two AKAP95 routes:
| route | AKAP95 region | KMT2A side | survives the fusion? |
|---|---|---|---|
| via DPY30 / WRAD | 1-210 (or 572-589; disputed) | none - DPY30 docks ASH2L, not KMT2A | the question does not apply; WRAD reaches the fusion only as a second-shell passenger (hyp-wrad-reaches-kmt2a-fusions-only-as-a-second-shell-passenger) |
| via MLL1T co-condensation | IDR1, 101-390 | the retained N-terminal fragment itself | yes, by construction |
So the entry this graph already held - which reasoned that AKAP95 acts on the enzyme and therefore on the wild-type allele rather than on the fusion - was reasoning correctly from the only evidence then in the corpus, and the 2026 paper supplies a second, different contact that does reach the fusion. The two readings are not in conflict; they are two surfaces.
The regions overlap only in 101-210. IDR1 is 101-390; the DPY30 region is 1-210. That 110-residue overlap contains the six tyrosines whose mutation kills co-condensation, so a PS-mutant is not a clean separation-of-function reagent for the DPY30 route. A construct that is clean is AKAP95(1-100) - sufficient for neither, necessary for DPY30 binding - and AKAP95(211-390), which has never been made.
What the co-condensation actually buys, in genomic terms
AKAP95 binds 5,656 sites genome-wide by anti-FLAG ChIP-seq on the endogenous knock-in tag, 59.23% of them promoters 1. A core set of 177 MLL-AF9 target genes is significantly enriched among AKAP95-bound promoters/genes (n = 56, P < 0.0001, Fisher exact), and AKAP95 binding intensity is higher at the joint sites than elsewhere. When AKAP95 is lost, the joint AKAP95/MLL-AF9 targets (n = 18) go down significantly more than other AKAP95 targets 1.
That last comparison is the one that carries information, because it is internally controlled: it asks whether the fusion-associated subset is preferentially sensitive, not merely whether AKAP95 targets overlap fusion targets at all. Overlap between two sets of active promoters is close to free; differential sensitivity within one protein's own target set is not.
And the null nobody wrote down. The paper reports the overlap P value but not what fraction of all expressed promoters would be recovered by chance under matched expression. AKAP95 binds 59% promoters and MLL-AF9 targets are highly expressed genes, so a promoter-biased factor will overlap a highly-expressed target list. The enrichment is probably real - the within-set n=18 comparison does not have this problem - but the headline Fisher P is not a clean number.
| partner a | gene-akap8 |
|---|---|
| partner b | gene-kmt2a |
| partner a uniprot | O43823 |
| partner a length AA | 692 |
| partner b uniprot | Q03164 |
| partner b length AA | 3969 |
| motif b | MLL1T, the MLL1 translocated (N-terminal) fragment - residue boundaries NOT stated in the source |
| druggability | Targeted indirectly and successfully: the JD-PI95 peptide bridges AKAP95 to HSP70 and dissolves the condensate rather than blocking a pocket. There is no pocket to block - the interaction is between two disordered regions driven by charge complementarity. |
| region retained by fusion | YES - MLL1T is by definition the fragment KMT2A fusions retain |
| shown for fusion | KMT2A::MLLT3 (THP-1 co-localisation and rescue), MLL-AF9 mouse (genetics and transcriptome), MLL-AF9/MOLM-13; KMT2A::AFF1 only as an shRNA growth curve in MV4-11. NOT shown: any KMT2A::AFF1 imaging, co-IP, ChIP or in vivo model. |
| evidence state | partially-measured |
| evidence state note | The contact is measured for MLL-AF9 and KMT2A::MLLT3. For the index fusion KMT2A::AFF1 it is measured only as a proliferation phenotype in one cell line. Closing it needs a co-IP and an imaging co-localisation in SEM or RS4;11, which is a two-week experiment with off-the-shelf reagents. |
| evidence state why not absence class | This entry is not an absence. It reached for absence_class to record an evidence state that vocabulary cannot express - 'partially-measured' - so the value moved to evidence_state rather than being folded into an absence token that would misdescribe it. See q190: only 69 of the 529 entries carrying absence_class were absence entries. |
interface residues a
[]
structures
[]
ligands
[ "drug-jd-pi95" ]
region on akap8
{
"name": "IDR1",
"start": 101,
"end": 390
}measurements — the numbers behind the chart
[
{
"quantity": "count",
"subject": "AKAP95 ChIP-seq binding sites, endogenous Flag-mEGFP knock-in HeLa",
"value": 5656,
"units": "sites",
"n": 1,
"universe": "anti-Flag ChIP-seq, HeLa knock-in, PMID 41501053",
"null": null,
"note": "59.23% promoters, 17.27% introns, 12.14% intergenic"
},
{
"quantity": "count",
"subject": "core MLL-AF9 target genes also found among AKAP95-bound promoters/genes",
"value": 56,
"units": "genes",
"n": 177,
"universe": "a 177-gene published core MLL-AF9 target set intersected with AKAP95 ChIP-seq promoters/genes in HeLa, PMID 41501053",
"null": null,
"note": "P < 0.0001 two-sided Fisher. The paper does NOT report an expression-matched null, and AKAP95 is 59% promoter-biased, so the marginal overlap is inflated; the within-set n=18 sensitivity comparison is the clean result."
},
{
"quantity": "count",
"subject": "joint AKAP95/MLL-AF9 targets whose down-regulation on AKAP95 loss exceeds that of other AKAP95 targets",
"value": 18,
"units": "genes",
"n": null,
"universe": "AKAP95-depleted vs control, PMID 41501053 Fig 4f",
"null": null,
"note": "internally controlled comparison - the denominator is AKAP95's own target set, so promoter bias cancels"
},
{
"quantity": "count",
"subject": "MLL-fusion proteins recovering AKAP95 in a published fusion interactome",
"value": 4,
"units": "fusions",
"n": null,
"universe": "MLL-AF9, MLL-EEN, MLL-ENL, MLL-GAS7; reported at one remove via PMID 41501053 reference 20, source not opened",
"null": null,
"note": "KMT2A::AFF1 is NOT among the four"
}
]- A condensate interaction has no pocket, so the only tractable modality is what the authors did - forced proximity to a disaggregase. That is a genetically encoded peptide, and the delivery problem for a nuclear peptide in a marrow blast is unsolved.
- If AKAP95 reaches the fusion through the retained N-terminal fragment, then it also reaches wild-type KMT2A, which carries the same fragment. Every fusion-directed argument here has an unaddressed wild-type allele on the other side of it.
- The one fusion class with no data is the index fusion. Building a programme on 'AKAP95 is an MLLr dependency' and then discovering the effect is AF9-specific would be the same failure this corpus records for partner-imported evidence generally.
- Where does MLL1T end? The construct's boundaries are not stated, so it cannot be checked against the breakpoint intervals this graph records for KMT2A::AFF1 - and whether the AKAP95-contacting stretch lies 5' or 3' of the common intron 9-11 breakpoint region is unknown and decidable from the plasmid map alone.
- Does AKAP95 co-condense with KMT2A::AFF1, which carries the AFF1 pSER and CHD domains and joins the super elongation complex, or only with fusions whose partner is a YEATS/DOT1L-type? All four interactome hits and all the imaging are AF9/ENL/EEN/GAS7-class.
- Is the co-condensation required for the leukaemia phenotype, or merely correlated with it? The PS-mutant fails to rescue, but the PS-mutant also fails to make condensates at all, so it cannot separate 'AKAP95 needs to condense' from 'AKAP95 needs to condense WITH MLL1T'.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Endogenous AKAP95 co-immunoprecipitates with over-expressed Flag-HA-MLL1T in 293T cells; endogenous Flag-mEGFP-AKAP95 foci in knock-in HeLa cells partially co-localise with mCherry-MLL1T; a GFP fragment series identifies AKAP95 aa101-390 (IDR1), and not aa1-100, aa391-550 or aa551-692, as the fragment that co-localises with MLL1T and forms large nuclear aggregates; replacing six tyrosines in aa101-210 with serine (PS-mutant) abolishes co-condensation; in rescued THP-1 cells, endogenous MLL1 N-terminal fragments partially co-localise with AKAP95 WT but not the ZF- or PS-mutants, and only AKAP95 WT rescues the growth defect of AKAP95-knockdown THP-1 and MOLM-13. AKAP95 has 5,656 genome-wide binding sites (59.23% promoters); 56 of a 177-gene core MLL-AF9 target set are enriched among AKAP95-bound promoters/genes (P < 0.0001, two-sided Fisher), and the 18 joint AKAP95/MLL-AF9 targets are down-regulated significantly more on AKAP95 loss than other AKAP95 targets.
AKAP95 condensates regulate transcription and can be targeted in MLL-fusion driven oncogenesis. · Yan X, Luan Y, Zhu Y, Cai Y, Ye F, Yang L, Zhu K, Hu J, Luo S, Lin J, Wang Y, Huang Y, Sun P, Yang Z, Shi B, Jiang H, et al. · Nat Commun 2026What this evidence cannot show
Full text read via PMC12886973; supplementary figures and Source Data were not available. THREE specific limits. (1) The residue boundaries of the MLL1T construct are NOT stated anywhere in the main text or Methods, so 'the retained fragment' is the authors' label, not a coordinate this beat can verify against Q03164. (2) The co-IP is endogenous AKAP95 against an OVER-EXPRESSED MLL1T in 293T - not an endogenous-endogenous pair, and not in a leukaemia cell. (3) The fragment mapping is an imaging co-localisation, not a binding assay; a fragment can fail to co-localise for folding or expression reasons and no fragment co-IP is shown in the main text. Confidence is capped at moderate for those reasons and not because the result is doubted.
Evidence that argues against this (1)
- The same paper reports that AKAP95 binding intensity at MLL-AF9 target sites in THP-1 was higher than at other sites but that THE DIFFERENCE WAS NOT SIGNIFICANT (Supplementary Fig.) - i.e. the preferential-binding result that is clean in HeLa does not reach significance in the leukaemia line where it matters. The paper's own limitations section concedes that the joint-target analysis rests substantially on HeLa. Yan X et al. 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
AKAP95 was identified as a hit against four different MLL-fusion proteins - MLL-AF9, MLL-EEN, MLL-ENL and MLL-GAS7 - in a previously published MLL-fusion-protein interactome, which is what prompted the 2026 study.
AKAP95 condensates regulate transcription and can be targeted in MLL-fusion driven oncogenesis. · Yan X, Luan Y, Zhu Y, Cai Y, et al. · Nat Commun 2026What this evidence cannot show
This is a secondary statement: the 2026 paper's reference 20 is the interactome, and that source was NOT resolved or opened by this beat. The four-fusion recurrence is therefore reported at one remove and should be re-sourced before it is used as an independent line of evidence. It is recorded here because it is the stated provenance of the whole study.
Evidence that argues against this (1)
- The four fusions named are MLL-AF9, MLL-EEN, MLL-ENL and MLL-GAS7, and KMT2A::AFF1 is absent from the list. If the source interactome included an AFF1 fusion among its baits, that absence is a NEGATIVE for the index fusion, not an untested case - and the entry would then be recording a recurrence that specifically excludes the fusion this project is built on. Which of the two it is cannot be determined, because the source was cited at one remove and never resolved to a PMID. Yan X et al. 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
This graph's prior entry on the same protein recorded the localisation field as shown_for_fusion 'NO - nobody has looked in any KMT2A fusion' and the absence class as 'never-measured', on the basis of PMID 23995757 alone.
interface-kmt2a-akap8-a-prey-only-grade-on-a-direct-mll-coactivator, data.localization.shown_for_fusion and data.absence_class, written cycle 146 by q160 · this project 2026What this evidence cannot show
Recorded here so the correction is visible from the new entry. q160's reasoning was sound on the evidence it had; the miss is a nomenclature artefact, quantified in `ana-q166-the-akap95-leukaemia-absence-was-an-abstract-level-artefact`. The URL field is a placeholder for an internal artefact with no external locator.
Evidence that argues against this (1)
- Citing this project's own prior field as evidence is circular for anything except the narrow claim 'the corpus said X on this date', and it is used for no more than that. The risk is in how it reads: q160 reasoned correctly from the only source then in the corpus, and a later reader encountering this item beside the 2026 paper may take it as an indictment of that reasoning. The failure was in three search instruments, not in the inference, and the accompanying analysis entry is where that is measured. this project 2026
AKAP95 IDR1, residues 101-390
co-condensation with the MLL1 translocated fragment, charge-complementarity driven
endogenous MLL1 N-terminal fragments co-localise with AKAP95 WT in THP-1, and only WT rescues the knockdown growth defect
the second, non-overlapping AKAP95 route into this axis; the two regions share only residues 101-210
partially - AKAP95 reaches the fusion directly through the retained fragment, not through WRAD, so a WRAD-adjacent protein can have a first-shell route
the paper the absence missed
the condensate route is what the peptide dissolves; the DPY30 route is never assayed
the same AKAP8 region is a recruited coactivator in MLLr leukaemia and a fusion moiety in ZNF384-rearranged B-ALL
Route B, the co-condensation route, which survives the fusion by construction
the same PS- and ZF-mutants that break RNA-gated condensation are the ones that fail to co-localise with MLL1 N-terminal fragments and fail to rescue MLLr growth