KMT2A-AKAP8 (AKAP95): graded 'prey-only' by this project, and it is a direct positive modulator of MLL H3K4 methyltransferase activity from a dedicated MLL paper - ADMITTED
also called KMT2A-AKAP95 · MLL1-AKAP95 · AKAP8 · A-kinase anchor protein 95 kDa · DPY30-AKAP95 · EBI-AKAP95-MLL
AKAP95 is a nuclear scaffolding protein that sits on the MLL machine and makes it better at putting the activating H3K4 mark on chromatin. Our own bookkeeping filed it as a low-grade 'passenger' hit, because the experiment that found it used a different protein as the hook.
The same thing, for a scientist
**Verdict: ADMITTED, and it is the single clearest counter-example to the premise that prey-only evidence is weak by construction.** The IntAct record is KMT2A-AKAP8, detection method **molecular sieving**, interaction type *physical association*, human, from **PMID 23995757** [E1]. That paper is Jiang, Lu, Shimada, Dou, Tang and Roeder, *Nat Struct Mol Biol* 2013, **whose title names the MLL complex** and whose subject is exactly this protein: *Regulation of transcription by the MLL2 complex and MLL complex-associated AKAP95* [E2]. Full text read. AKAP95 was recovered as a major FLAG-HA-**DPY30**-associated protein; **gel filtration on Superose 6 shows AKAP95 co-eluting with the MLL1 and MLL2 complexes but not with free or dimerised DPY30**; endogenous co-IP from HeLa nuclear extract confirms it; IP of WDR5 or ASH2L co-precipitates AKAP95 and tagged AKAP95 co-precipitates ASH2L and WDR5; and purified AKAP95 binds purified DPY30 through an N-terminal region - **UniProt O43823 records "Interaction with DPY30" as residues 1-210, and no 1-100 fragment exists in the source paper**: its Fig 5F compares full-length GST-AKAP95 against GST-AKAP95(101-692), which binds "much weaker" rather than not at all, and the 1-210 sufficiency is a Myc-DPY30 co-IP from transfected cells (Fig 5E), not a purified-protein experiment [E2]. Functionally, recombinant AKAP95 **enhances MLL2-complex H3K4 mono-, di- and tri-methylation on recombinant chromatin** while having no methyltransferase activity of its own and not binding free H3 [E2]. **The reason it scores prey-only is a bookkeeping artefact**: gel filtration has no bait, so IntAct records our protein's experimental role as `unspecified role`, and a filter that asks *was our protein the bait?* answers no. 24 of the 212 human prey-only partners are in this class [E4]. **`confidence_overall` rates the KMT2A-AKAP8 contact specifically, which is weaker than the AKAP95-MLL-complex association**: the direct binding shown is to DPY30, and the KMT2A co-elution is complex-level.
Why a 'prey-only' grade is wrong here, mechanically
This project grades a partner bait-supported if any curated record has our protein as the recorded bait, and prey-only otherwise. The grade is a proxy for did somebody set out to study our protein, and rule 92 already found the proxy misfiring in the other direction (a focused complex purification looking like a screen).
This is the third failure mode and it is cleaner than the other two: some methods have no bait at all. Across the 212 human prey-only partners, the experimental role IntAct records for our protein is prey for 188, and for the remaining 24 it is neutral component (crosslink MS, methyltransferase assay, FPS), fluorescence accept (BRET) or unspecified role (molecular sieving) 4. A gel-filtration co-elution, an ITC titration, an NMR titration and a crosslink are symmetric experiments. Neither partner is the bait; the role column is empty by the nature of the assay, and a bait/prey filter silently scores an empty column as prey.
This is the same shape as rule 69b's third clause: an instrument that supplies a value the data did not.
The top of the prey-only ranking is dominated by exactly this class. H3C1 ranks first with ITC, NMR, pull-down and BioID behind it - and ITC and NMR are bait-free biophysics on a genuine substrate. AKAP8 ranks with molecular sieving. The best evidence in the tail is bait-free evidence, and the filter is systematically blind to it.
What the paper actually shows, in order of strength
Full text of PMID 23995757 read (PMC3813012). Evidence for AKAP95 in MLL complexes, strongest last:
| observation | what it establishes |
|---|---|
| AKAP95 is a major protein co-purifying with FLAG-HA-DPY30 from Flp-In-293 nuclear extract, by MS and by anti-HA immunoblot | association, over-expressed bait |
| endogenous co-IP from HeLa nuclear extract with a DPY30 antibody | association, endogenous on both sides |
| IP of WDR5 or ASH2L co-precipitates AKAP95; tagged AKAP95 co-precipitates ASH2L and WDR5 | reciprocal, and through two other WRAD subunits |
| gel filtration (Superose 6, BC200, 0.05% NP40): AKAP95 co-elutes with the MLL1 and MLL2 complexes, and NOT with abundant free or dimerised DPY30 | the AKAP95 pool that matters is the complex-associated one - this is the KMT2A record |
| purified DPY30 binds purified AKAP95, much more weakly to AKAP95(101-692) | direct, and mapped to AKAP95 residues 1-100 |
| recombinant AKAP95, but not AKAP95(101-692), significantly enhances MLL2-complex H3 methylation on recombinant chromatin; H3K4me1, me2 and me3 all enhanced; AKAP95 has no MTase activity and does not bind free H3 | functional, and it is a direct enzymatic effect on the MLL complex |
The negative controls in that last row are the part that matters: an apparent stimulation by a contaminant methyltransferase, or by substrate sequestration, are both excluded in the same figure.
Cell type, cell state, subcellular and subnuclear localization
Subnuclear. UniProt O43823 (692 aa) carries three experimentally coded nuclear locations: Nucleus, Nucleus matrix, and Nucleus, nucleolus, plus a by-similarity Cytoplasm 3. AKAP95 is described in the source paper as the only nuclear member of the AKAP family, with a defined nuclear matrix targeting site (NMTS) and two zinc fingers 2.
And the dissection is more interesting than the annotation. Deleting the NLS or the zinc fingers abolishes AKAP95's transcriptional coactivation; deleting the NMTS does not 2. So the nuclear-matrix tether that the localisation annotation is famous for is not the part required for the MLL-linked function. A localisation field that recorded only nucleus matrix would point a reader at the wrong domain.
Cell type and cell state. The MLL-complex work is in Flp-In-293 and HeLa nuclear extract, with a Gal4-luciferase reporter in G-293T. The functional genetics is in NT2 human embryonal carcinoma cells and mouse E14 embryonic stem cells 2. The cell state is explicit and it is a differentiation transition: 4 days in differentiation medium then 4 days with 1 µM all-trans retinoic acid. AKAP95 depletion has essentially no effect on undifferentiated ESCs - no change in morphology, proliferation, pluripotency-gene levels, global H3K4me3, or DPY30 recruitment - and impairs ATRA-induced developmental gene induction, with the ATRA-driven rise in H3K4me3 at those loci diminished 2.
No haematopoietic cell appears anywhere in this paper, and no KMT2A fusion. That is the gap, and it is the same gap the corpus records for most of its load-bearing partner evidence.
Epigenetics. AKAP95 is a direct positive modulator of H3K4 me1/me2/me3 deposition by an MLL complex. Because the effect is on catalysis rather than on recruitment, it is a rate modifier on the mark this entire disease turns on.
Why this matters for KMT2A::AFF1 specifically, and the two reasons it might not
The DPY30 route is the reason to care. DPY30 is the shared WRAD subunit of every MLL/SET1 complex, gene-dpy30 is already in this graph, and the N-terminal portion of KMT2A retained in KMT2A::AFF1 does not include the SET domain - so the fusion has no H3K4 methyltransferase activity of its own and works alongside wild-type KMT2A and the SET1/MLL family. AKAP95 acts on the enzyme, not on the fusion, which makes it a candidate modifier of the wild-type allele's output in a fusion-driven cell rather than a fusion partner.
Two reasons this may not transfer:
- The MLL1 (KMT2A) evidence is co-elution, not direct binding. The direct binding demonstrated is AKAP95-DPY30. The paper's own discussion says it has not been rigorously demonstrated that AKAP95 associates with all MLL complexes, and the enzymatic stimulation was measured on MLL2C. An entry that says
AKAP95 stimulates KMT2Awould be over-reading the source. - This is the worked shape of the KAT8/MOF problem this corpus already records - a strong biochemical association with the MLL machinery that turned out not to be fusion-specific. AKAP95 is at the same stage: real biochemistry, no fusion experiment.
| partner a | gene-kmt2a |
|---|---|
| partner b | AKAP8 |
| partner b uniprot | O43823 |
| partner b length AA | 692 |
| partner a uniprot | Q03164 |
| partner a length AA | 3969 |
| verdict | ADMITTED |
| adjudicated by | q160, cycle 146, 2026-09-24 |
| motif b | AKAP95 residues 1-100 are required for direct DPY30 binding and for coactivation; the NLS and both zinc fingers are required, the NMTS is not |
| druggability | Not assessable as an interface. No structure of AKAP95 with any MLL-complex subunit exists, and the mapped determinant (residues 1-100, plus zinc fingers) is described only by deletion, not at residue resolution. The interesting druggable object here is the DPY30 dimerisation domain, whose sequence the source paper notes is strikingly homologous to the PKA RII dimerisation domain that every other AKAP binds. |
| n methods | 1 |
| n platforms | 1 |
| n publications | 1 |
| endogenous side | yes - endogenous DPY30 co-IP from HeLa nuclear extract, and endogenous WDR5 and ASH2L IPs recover AKAP95 |
| intact miscore | 0.4 |
| bait free method | true |
| why prey only is wrong here | Gel filtration has no bait. IntAct records experimentalRole 'unspecified role'; a filter asking 'was our protein the bait' scores the empty column as not-bait and the partner falls into the prey-only bin. |
| cell state | all-trans retinoic acid differentiation, 1 uM, 4 days in differentiation medium then 4 days with ATRA; the phenotype is absent in the undifferentiated state |
| epigenetics | Direct positive modulator of MLL2-complex H3K4 mono-, di- and tri-methylation on recombinant chromatin. No intrinsic methyltransferase activity and no binding to free H3, so the effect is on the enzyme rather than on the substrate. |
| evidence state | partially-measured |
| evidence state note | CORRECTED cycle 156. AKAP95 HAS been tested against KMT2A fusions: PMID 41501053 (Yan X et al., Nat Commun 2026;17(1):1457, doi 10.1038/s41467-025-68192-4) reports shRNA knockdown reducing growth of THP-1 and MOLM-13 (both KMT2A::MLLT3/MLL-AF9) and MV4-11 (KMT2A::AFF1, the index fusion), an Akap95-null mouse in which MLL-AF9 marrow gives longer survival on secondary transplant, and co-condensation of AKAP95 IDR1 (aa 101-390) with the MLL1 translocated fragment. THE ABSENCE NARROWS, IT DOES NOT CLOSE: the index fusion contributes an shRNA growth curve only; the mouse and the mechanism are MLL-AF9. |
| 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
[]
records
[ "KMT2A-AKAP8, molecular sieving, physical association, Homo sapiens, PMID 23995757, miscore 0.40, our role 'unspecified role'" ]
cell systems
[ "Flp-In-293 (FLAG-HA-DPY30 bait)", "HeLa nuclear extract (endogenous co-IP)", "G-293T Gal4-UAS x5 luciferase reporter line", "NT2 human embryonal carcinoma", "mouse E14TG2A embryonic stem cells" ]
localization
{
"compartment": "nucleus",
"subnuclear": [
"nuclear matrix (ECO:0000269)",
"nucleolus (ECO:0000269)"
],
"functionally_required_region": "NLS and zinc fingers; NOT the nuclear matrix targeting site",
"shown_for_fusion": "PARTIALLY - corrected cycle 156. Not the localisation itself, which remains a UniProt annotation in non-leukaemic cells. AKAP95 function is now shown in MLL-AF9 (mouse and THP-1/MOLM-13) and in KMT2A::AFF1 (MV4-11) as an shRNA growth curve, per PMID 41501053."
}measurements
[
{
"quantity": "count",
"subject": "distinct publications supporting the KMT2A-AKAP8 IntAct record",
"value": 1,
"units": "publications",
"n": 1,
"universe": "IntAct REST findInteractions on Q03164, 2026-09-24",
"null": null,
"note": "One publication, but a dedicated MLL-complex paper with six independent experiments inside it - which is why publication count is a bad evidence proxy in this tail."
},
{
"quantity": "count",
"subject": "human prey-only partners whose records contain no literal prey role for our protein",
"value": 24,
"units": "partners",
"n": 212,
"universe": "the 2026-09-24 IntAct REST prey-only census, human only",
"null": null,
"note": "11.3%. Bait-free assays: crosslink MS, BRET, molecular sieving, FPS, methyltransferase assay."
}
]correction cycle 156
{
"by": "q166, verified by the king",
"residue_region": "UniProt O43823 gives 1-210 for the DPY30 interaction; the 1-100 figure was wrong and the king repeated it in a log and a dispatch brief",
"localisation_provenance": "nucleus matrix and nucleolus ARE UniProt annotations (ECO:0000269) - this entry sourced them correctly; the kings brief wrongly re-described them as full-text claims",
"new_paper": "PMID 41501053, doi 10.1038/s41467-025-68192-4, full text read by q166, headline verified by the king",
"E3_untouched": true
}- If AKAP95 is a general rate modifier of MLL-family H3K4 methylation, inhibiting it would hit wild-type KMT2A, KMT2B and the SET1 complexes at the same time. The therapeutic window argument that carries menin inhibitors - the fusion needs menin and normal haematopoiesis tolerates partial loss - has no counterpart here.
- AKAP95 depletion is phenotypically silent in undifferentiated ESCs and bites only during a differentiation transition. A drug with that profile would look clean in every steady-state toxicity screen and damage exactly the regenerating compartments a leukaemia patient depends on.
- The KMT2A evidence here is co-elution. Writing this entry as though AKAP95 binds KMT2A would put a complex-level observation into a binary-interaction field, which is the error this beat is documenting in other people's work.
- Does AKAP95 depletion change H3K4me3 at HOXA9 and MEIS1 in a KMT2A::AFF1 cell? The ESC result says AKAP95 matters only during a state transition and not at steady state - a KMT2A-rearranged blast is arguably a cell locked in a failed state transition, which is the strongest reason to expect a phenotype and the reason the experiment has never been framed.
- Is the AKAP95-DPY30 contact the same surface as the PKA RII contact? The source notes the DPY30 dimerisation domain is strikingly homologous to the RII dimerisation domain, yet maps the two AKAP95 binding regions to different places. If one nuclear scaffold can be competed off MLL complexes with an RII-mimetic peptide, that is a tool compound nobody has tried.
- How many of the other 23 bait-free 'prey-only' partners are similarly mis-graded? Only AKAP8 and H3C1 have been read; the crosslink-MS set in particular has never been opened.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
IntAct REST returns a KMT2A-AKAP8 (Q03164-O43823) record from PMID 23995757, detection method 'molecular sieving', interaction type 'physical association', host organism Homo sapiens, intact-miscore 0.40, and the experimental role recorded for KMT2A is 'unspecified role' - not bait and not prey.
IntAct molecular interaction database, REST findInteractions on Q03164, retrieved 2026-09-24 by q160 · IntAct (EMBL-EBI) 2026What this evidence cannot show
A single curated record. 'Molecular sieving' is gel-filtration co-elution, which shows two proteins in the same size fraction and does not by itself show contact; the contact evidence in this entry comes from the co-IP and direct-binding experiments in the same paper, not from this row. `expected_absent_terms: [bait]` refers to the experimentalRole field of this record, not to the paper, which does use a tagged bait for other experiments.
Evidence that argues against this (1)
- Co-elution on Superose 6 puts AKAP95 in the same fractions as MLL1 and MLL2 complexes but cannot distinguish a stable subunit from a similarly sized independent assembly. The paper's own framing is that AKAP95 is 'not an integral core subunit of MLL complexes like DPY30 or RbBP5'. Jiang H et al. 2013
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
PMID 23995757 (Jiang, Lu, Shimada, Dou, Tang, Roeder, Nat Struct Mol Biol 2013) identifies AKAP95 as a major FLAG-HA-DPY30-associated protein in Flp-In-293 nuclear extract; confirms endogenous association by DPY30 co-IP from HeLa nuclear extract; shows WDR5 and ASH2L IPs co-precipitate AKAP95 and that tagged AKAP95 co-precipitates ASH2L and WDR5; shows by Superose 6 gel filtration that AKAP95 co-elutes with the MLL1 and MLL2 complexes but not with free or dimerised DPY30; maps direct binding to purified DPY30 to AKAP95 residues 1-100; and shows recombinant AKAP95, but not AKAP95(101-692), significantly enhances MLL2-complex-mediated H3K4 mono-, di- and tri-methylation on recombinant chromatin, while purified AKAP95 has no methyltransferase activity of its own and does not bind free histone H3. The NLS and the two zinc fingers, but not the nuclear matrix targeting site, are required for AKAP95 coactivation of a chromosomal Gal4-luciferase reporter. AKAP95 depletion in mouse E14 ESCs does not affect self-renewal, proliferation, pluripotency gene expression or global H3K4me3, but impairs induction of developmental genes after 1 uM all-trans retinoic acid and diminishes the ATRA-driven H3K4me3 increase at those loci; the same impairment is seen in human NT2 embryonal carcinoma cells.
Regulation of transcription by the MLL2 complex and MLL complex-associated AKAP95. · Jiang H, Lu X, Shimada M, Dou Y, Tang Z, Roeder RG · Nat Struct Mol Biol 2013What this evidence cannot show
Full text read via PMC3813012, including Methods; figure panels themselves were not inspected, so every quantitative statement here is the text's own description of its figures. The enzymatic stimulation is measured on the MLL2 (KMT2B) complex, not on MLL1/KMT2A - the KMT2A link is the gel-filtration co-elution plus the shared WRAD subunits. `expected_absent_terms` records that this paper contains no leukaemia and no fusion experiment; that absence is part of the claim.
Evidence that argues against this (1)
- The paper's discussion states explicitly that it 'is possible, but has not been rigorously demonstrated, that AKAP95 may also be commonly associated with all MLL complexes', and it notes the MLL2 complex it purified lacked any detectable HAT whereas a previously purified WDR5-MLL1 complex carried MOF - i.e. the composition of an MLL complex is tagging- and purification-strategy dependent. A co-elution result read as 'AKAP95 is in the MLL1 complex' is therefore reading past what the authors claim. Jiang H et al. 2013
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProtKB O43823 (AKAP8_HUMAN) is 692 aa and carries three experimentally coded (ECO:0000269) subcellular locations - Nucleus, Nucleus matrix, and Nucleus nucleolus - plus Cytoplasm by similarity (ECO:0000250). Tissue specificity: highly expressed in heart, liver, skeletal muscle, kidney and pancreas, and expressed in mature dendritic cells.
UniProtKB entry O43823 (AKAP8_HUMAN), subcellular location and tissue specificity, retrieved 2026-09-24 · UniProt Consortium 2026What this evidence cannot show
Rule 91 applies: an ECO:0000269 code means a curator read an experiment, not necessarily the founding or best one, and the cited papers behind these three location lines were not opened here. The tissue-specificity line names no haematopoietic tissue except mature dendritic cells, which is a genuine gap for a leukaemia argument and is not resolved by this record.
Evidence that argues against this (1)
- The nuclear-matrix annotation is the one a reader would key on and it is the one the functional dissection rules out: deleting the NMTS does not impair AKAP95's coactivation of the reporter, while deleting the NLS or the zinc fingers does. The database's most prominent localisation fact is not the functionally relevant one. Jiang H et al. 2013
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
Of the 212 human prey-only partners of the six core proteins in the 2026-09-24 IntAct REST pull, 188 have at least one record where our protein's experimental role is literally 'prey'. The remaining 24 have none: 14 are 'neutral component' (13 crosslink-MS, one methyltransferase assay), 8 are 'fluorescence accept' (BRET), one is 'unspecified role' (AKAP8, molecular sieving), and one (CREBBP) is 'neutral component' from an FPS experiment. These are symmetric assays in which no partner is a bait.
analyses/q160-the-prey-only-interactome/rest-roles.json, experimentalRole tallies over the 212 human prey-only partners, generated 2026-09-24 · this project 2026What this evidence cannot show
24 of 212 is 11.3%, which is a minority; the bait/prey filter is right about most of the population and this entry should not be read as invalidating it. What it invalidates is the inference from 'prey-only' to 'weak', because the bait-free 24 include the biophysics (ITC, NMR, gel filtration) that is the strongest evidence in the whole tail.
Evidence that argues against this (1)
- Eight of the 24 are BRET rows from the neoPPI screen (PMID 35512704), which q128 declined one by one on other grounds - single assay, no allele effect, no endogenous side. So being bait-free does not make a record good; it makes the bait/prey grade uninformative about it, which is a weaker claim. Mo X et al. 2022
AKAP95 co-elutes with the MLL1 complex on Superose 6 (molecular sieving, PMID 23995757)
- binds →DPY30 — the smallest COMPASS subunit, and the only one that behaves like a selective dependency
direct: purified AKAP95 residues 1-100 bind purified DPY30
WDR5 IP co-precipitates AKAP95 and tagged AKAP95 co-precipitates WDR5
- binds →ASH2L — the WRAD subunit that has DPY30's essentiality profile and none of its co-dependency signal
ASH2L IP co-precipitates AKAP95, reciprocally
recombinant AKAP95 enhances MLL2-complex H3K4 me1/me2/me3 on recombinant chromatin
stimulates deposition; ATRA-driven H3K4me3 rise is diminished by AKAP95 knockdown
the clearest counter-example to 'prey-only evidence is weak by construction'
AKAP8 is the partner-level worked example: bait-free, prey-only graded, and a dedicated MLL paper
q160's adjudication of the same IntAct record; its shown_for_fusion and absence_class fields are now superseded by PMID 41501053
its shown_for_fusion 'nobody has looked in any KMT2A fusion' and absence_class 'never-measured' are refuted by PMID 41501053