AKAP8 (AKAP95) - the only nuclear A-kinase anchoring protein, and the one that turned out to be required for MLL-rearranged leukaemia
also called AKAP95 · A-kinase anchor protein 95 kDa · AKAP 95
AKAP8, usually called AKAP95, is a nuclear scaffolding protein: it holds other proteins in the right place rather than doing chemistry itself. In 2026 it was shown that leukaemia cells carrying an MLL (KMT2A) fusion need it, while normal blood-forming cells do not.
The same thing, for a scientist
AKAP8 (UniProtKB **O43823**, **692 aa**) is the only nuclear member of the A-kinase anchoring protein family. Its canonical job is to tether the PKA RII-alpha regulatory subunit through a short C-terminal helix (**RII-binding, residues 572-589**), but essentially none of its chromatin biology runs through PKA. It carries two C2H2 zinc fingers (**392-414** and **481-504**), a bipartite NLS (**289-306**), a large N-terminal intrinsically disordered region, and a nuclear-matrix targeting site mapped by deletion to **111-140** [E1][E2]. Three lines of function are established: (i) it binds **DPY30**, the shared subunit of every SET1/MLL H3K4 methyltransferase complex, and enhances H3K4 methylation by the MLL2/KMT2B complex in vitro [E2]; (ii) it is a phase-separating splicing and transcription regulator whose zinc fingers bind intronic RNA and whose IDR drives condensation [E3][E4]; and (iii) **it is required for MLL-AF9-driven leukaemogenesis in mouse and for the growth of MLLr human lines, while Akap95-null bone marrow reconstitutes normally in competitive transplant** [E5]. It is also, separately, a recurrent fusion partner of ZNF384 in B-ALL [E6]. **Note the nomenclature trap: a PubMed search for AKAP8/AKAP95 AND KMT2A returns 0 records; the same search with MLL returns 2.**
This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.
I have never read one of these pictures before
A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.
Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.
Anything in yellow sticks is a bound small molecule, which is usually the drug.
Domain architecture, with the coordinates checked against the sequence
Every boundary below was taken from the UniProt feature table for O43823 and cross-checked against the two primary papers that defined it 12.
| region | residues | source | note |
|---|---|---|---|
| N-terminal region required for DPY30 binding and coactivation | 1-100 | Jiang 2013, by loss in AKAP95(101-692) 2 | not itself expressed as a fragment - see interface-akap8-dpy30 |
| Interaction with DPY30 (UniProt's own call) | 1-210 | UniProt feature, ECO:0000269 from PMID 23995757 1 | UniProt records 1-210, not 1-100 |
| Interaction with MCM2 | 1-195 | UniProt 1 | |
| NMTS (nuclear matrix targeting site) | 111-140 | Jiang 2013 Methods, dNMTS = 1-110 + 141-692 2 | no UniProt feature; sits inside the DDX5-interaction region 109-201 |
| Interaction with DDX5 (p68) | 109-201 | UniProt 1 | |
| IDR1, the MLL1-fragment co-condensation module | 101-390 | Yan 2026, PONDR-guided fragment series 5 | the six tyrosines mutated in the PS-mutant lie in 101-210 |
| bipartite NLS | 289-306 (UniProt) / 288-305 (Jiang dNLS) | 12 | the two definitions differ by one residue at each end |
| C2H2 zinc finger 1 (AKAP95-type) | 392-414 | UniProt 1 | |
| chromatin-binding region | 387-450 | UniProt 1 | |
| C2H2 zinc finger 2 (AKAP95-type) | 481-504 | UniProt 1 | the zinc fingers bind proximal intronic RNA 3 |
| condensin recruitment | 525-569 | UniProt 1 | |
| RII-binding (PKA anchoring) | 572-589 | UniProt 1, and Jiang dRII = 1-571 + 590-692 2 | the two agree exactly; the I582P point mutant that kills RII binding lies inside it, and residue 582 of O43823 is indeed Ile |
The internal consistency check that matters: Jiang 2013 gives its deletion constructs only as joined coordinate pairs in the Methods, so every boundary above had to be reconstructed. The RII interval reconstructed that way (572-589) is character-for-character the UniProt feature, and the I582P mutant falls inside it - which is why the other reconstructed intervals (NMTS 111-140, NLS 288-305, ZF 390-503) can be trusted.
What is NOT in this gene's record, and is worth knowing
No haematopoietic tissue in the expression annotation. UniProt's tissue-specificity line reads heart, liver, skeletal muscle, kidney and pancreas, plus mature dendritic cells 1. A reader coming to AKAP8 from a leukaemia question finds nothing in the annotation that points there; the leukaemia biology is entirely in the 2026 primary paper 5.
The PKA function is almost a red herring for this axis. AKAP8's namesake activity - anchoring PKA RII - is dispensable for its transcriptional coactivation (the I582P mutant is near-fully active, and deleting 572-589 does not impair the reporter) 2. But it is not dispensable for DPY30 binding according to a second group 7, which is the open conflict recorded in interface-akap8-dpy30.
AKAP8L is a different gene. Q9ULX6, sometimes called HA95, is a paralogue that also co-precipitates H3K4 HMT core subunits 7. Do not merge the two records.
| uniprot | O43823 |
|---|---|
| length AA | 692 |
| alphafold ID | AF-O43823-F1 |
| hgnc symbol | AKAP8 |
| cell state | The AKAP95 requirement is state-dependent in both systems where it has been examined: absent in undifferentiated mouse ESC and revealed only after ATRA differentiation (PMID 23995757); absent in normal homeostatic and regenerative haematopoiesis and revealed only under MLL-AF9 transformation (PMID 41501053). |
PDB IDS
[]
domains — the numbers behind the chart
[
{
"name": "N-terminal DPY30-binding / coactivation region",
"start": 1,
"end": 100
},
{
"name": "Interaction with DPY30 (UniProt)",
"start": 1,
"end": 210
},
{
"name": "IDR1 (MLL1-fragment co-condensation module)",
"start": 101,
"end": 390
},
{
"name": "NMTS (nuclear matrix targeting site)",
"start": 111,
"end": 140
},
{
"name": "NLS (bipartite nuclear localization signal)",
"start": 289,
"end": 306
},
{
"name": "ZF1 (C2H2 zinc finger, AKAP95-type)",
"start": 392,
"end": 414
},
{
"name": "ZF2 (C2H2 zinc finger, AKAP95-type)",
"start": 481,
"end": 504
},
{
"name": "RII-binding (PKA anchoring)",
"start": 572,
"end": 589
}
]localization
{
"compartment": "nucleus",
"subnuclear": [
"nuclear matrix (ECO:0000269, PMID 10601332)",
"nucleolus (ECO:0000269, PMID 26683827)",
"nuclear condensates/foci (endogenous mEGFP knock-in HeLa, FRAP T1/2 4.12 s, PMID 41501053)"
],
"shown_for_fusion": "PARTIAL - endogenous AKAP95 foci partially co-localise with an over-expressed MLL1 translocated fragment in HeLa knock-in cells, and with endogenous MLL1 N-terminal fragments in THP-1 (KMT2A::MLLT3); NOT shown in any KMT2A::AFF1 cell"
}essentiality
{
"mouse_ko": "Akap95-/- mice are viable and haematopoietically normal, including in 1:1 competitive transplant (PMID 41501053)"
}measurements — the numbers behind the chart
[
{
"quantity": "length",
"subject": "AKAP8 O43823 canonical sequence",
"value": 692,
"units": "aa",
"n": 1,
"universe": "UniProtKB reviewed human entry, retrieved 2026-09-24",
"null": null,
"note": "matches the 692 used by both primary papers, so their residue coordinates are directly comparable"
},
{
"quantity": "median_survival",
"subject": "secondary recipients of MLL-AF9-transformed LGMP, Akap95-/- arm",
"value": 28,
"units": "days",
"n": null,
"universe": "mouse secondary transplant, PMID 41501053 Fig 2f; WT arm 25.5 d",
"null": 25.5,
"note": "P=0.0022. A 2.5-day delay on a ~26-day median - significant but not a block; every recipient still died."
},
{
"quantity": "half_time",
"subject": "FRAP recovery of endogenous Flag-mEGFP-AKAP95 nuclear foci in HeLa knock-in cells",
"value": 4.12,
"units": "s",
"n": 8,
"universe": "8 individual cells, PMID 41501053 Fig 3a",
"null": null,
"note": "+/- 1.31 s SD. Fast recovery, consistent with a liquid-like condensate; no half-bleach control reported."
},
{
"quantity": "count",
"subject": "AKAP95 ChIP-seq binding sites, endogenous Flag tag in HeLa knock-in",
"value": 5656,
"units": "sites",
"n": 1,
"universe": "anti-Flag ChIP-seq in Flag-mEGFP-AKAP95 knock-in HeLa, PMID 41501053",
"null": null,
"note": "59.23% promoters, 17.27% introns, 12.14% intergenic"
}
]- The clean therapeutic window claimed here - MLLr cells need AKAP95, normal blood does not - was measured in a constitutive germline Akap95-/- mouse that developed normally. A germline knockout selects for compensation; an acute degron in adult marrow could look very different, and that experiment has not been done.
- AKAP95 is a general splicing regulator and a general Pol II condensate component. Any agent that dissolves AKAP95 condensates will perturb exon inclusion genome-wide, and the splicing phenotype is the one least likely to be picked up by a proliferation screen in a leukaemia line.
- AKAP95 is also the nucleolar regulator of 47S rRNA production. An inhibitor would hit ribosome biogenesis, which is the classic mechanism by which an apparently selective epigenetic agent turns out to have a narrow window in regenerating tissue.
- Is AKAP95 required in a KMT2A::AFF1 model in vivo? The entire genetic case rests on MLL-AF9; the index fusion contributes one shRNA growth curve in MV4-11.
- Does the AKAP95 requirement survive menin inhibition, or is it downstream of the same transcriptional programme? No study has combined AKAP95 loss with revumenib, ziftomenib or pinometostat in any system - a full-text-level absence, not merely an abstract one.
- Which AKAP95 pool matters for leukaemia - the DPY30/WRAD-bound pool or the MLL1-fragment co-condensate? The two use different regions of the protein (1-210 vs 101-390), so they are separable by mutant, and nobody has separated them.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProtKB O43823 (AKAP8_HUMAN) is 692 aa and its feature table records: zinc finger C2H2 AKAP95-type 1 at 392-414 and 2 at 481-504; bipartite NLS motif at 289-306; region 'Interaction with DPY30' at 1-210 with evidence ECO:0000269 PubMed 23995757; 'Interaction with MCM2' 1-195; 'Interaction with DDX5' 109-201; 'Involved in chromatin-binding' 387-450; 'Involved in condensin complex recruitment' 525-569; 'RII-binding' 572-589. Subcellular location: Nucleus (ECO:0000269 PubMed 19277197, 26683827), Nucleus matrix (ECO:0000269 PubMed 10601332), Nucleus nucleolus (ECO:0000269 PubMed 26683827), Cytoplasm (ECO:0000250). Tissue specificity: highly expressed in heart, liver, skeletal muscle, kidney and pancreas; expressed in mature dendritic cells. Residue 582 of the canonical sequence is isoleucine.
UniProtKB entry O43823 (AKAP8_HUMAN) - sequence, feature table, subcellular location, subunit and tissue specificity; retrieved live 2026-09-24 by q166 · UniProt Consortium 2026What this evidence cannot show
Rule 91: an ECO:0000269 code means a curator read an experiment, not that the cited paper is the founding or best one. Here the two localisation citations were both opened (PMID 10601332 Collas 1999 nuclear matrix/chromatin condensation; PMID 26683827 Marstad 2016 nucleolus/rRNA) and both are genuinely about localisation, so this particular record survives the rule-91 test. Residue 582 was read directly off the downloaded sequence string, not inferred.
Evidence that argues against this (1)
- UniProt records the DPY30-interaction region as 1-210, whereas the source paper's own text emphasises 1-100. Both derive from the same paper: 1-210 is the fragment that was sufficient in a cell co-IP, 1-100 is the region whose loss reduces direct binding. The database and the paper's prose therefore disagree about which number to quote, and neither is wrong. Jiang H et al. 2013
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Jiang et al. 2013 define the AKAP95 deletion constructs in their Methods as joined coordinate pairs - dNMTS: 1-110 + 141-692; dNLS: 1-287 + 306-692; dZF: 1-389 + 504-692; dRII: 1-571 + 590-692 - from which the deleted intervals are NMTS 111-140, NLS 288-305, zinc-finger region 390-503 and RII-binding 572-589. In a chromosomal Gal4-VP16 luciferase reporter in G-293T cells, the NLS and the two zinc fingers are required for AKAP95 coactivation, while the NMTS and the RII-binding region are not; the I582P point mutant that disrupts RII binding has 'only a little or marginal effect'. AKAP95(1-210) but not AKAP95(101-692) co-immunoprecipitates Myc-DPY30 from transfected 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 in full via PMC3813012 (NCBI OAI; the Europe PMC fullTextXML endpoint returned HTTP 500 on four attempts). Supplementary figures were NOT available, so every statement attributed to a Supplementary panel here is the main text's own description of it. The deletion boundaries are a reconstruction from joined coordinate pairs, validated against UniProt as described in the domain table. The expected_absent_terms entry is load-bearing: the string 'nucleol' appears ZERO times in the entire full text.
Evidence that argues against this (1)
- A deletion that abolishes a function does not prove the deleted region carries the binding site; it can destabilise the protein. The authors address this for the N-terminus (AKAP95 387-692 retains full chromatin-condensation activity; AKAP95(101-692) is recruited to the reporter promoter as well as or better than full-length) but do not address it for the zinc-finger deletion, which removes 114 residues. Jiang H et al. 2013
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AKAP95 interacts with transcription and RNA-processing factors including selective hnRNP proteins through its N-terminal region, binds preferentially to proximal intronic regions of pre-mRNAs in the human transcriptome in a manner requiring its zinc-finger domains, mainly promotes exon inclusion in coordination with hnRNP H/F and U, and directly self-interacts.
AKAP95 regulates splicing through scaffolding RNAs and RNA processing factors. · Hu J, Khodadadi-Jamayran A, Mao M, Shah K, Yang Z, Nasim MT, Wang Z, Jiang H · Nat Commun 2016What this evidence cannot show
Abstract read; full text not opened in this beat. This is cited here only to establish that AKAP95's zinc fingers are an RNA-binding module and its N-terminus a protein-scaffolding module - the domain assignment, not any quantitative claim.
Evidence that argues against this (1)
- The 'zinc fingers are the RNA-binding module' assignment is not exclusive, and the entry's domain table risks implying it is. The same C2H2 fingers carry a SELEX-demonstrated preference for GC-rich DNA, and UniProt's chromatin-binding region 387-450 overlaps zinc finger 1 at 392-414. Any zinc-finger mutant therefore loses DNA binding as well as RNA binding, which matters downstream: the 2026 paper's central ZF-mutant is interpreted throughout as an RNA-binding mutant. Marstad A et al. 2016
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AKAP95 forms phase-separated, liquid-like condensates in vitro and in the nucleus; point mutations of key residues perturb condensation in opposite directions; its splice-regulatory activity is abolished by disrupting condensation and impaired by hardening the condensates, and is regained when its condensation-mediating region is replaced by an unrelated condensation-mediating region. Its ability to support tumorigenesis requires condensates of proper liquidity and dynamicity.
Biophysical properties of AKAP95 protein condensates regulate splicing and tumorigenesis. · Li W, Li W, Hu J, Shi B, Palomba F, Digman MA, Gratton E, Jiang H · Nat Cell Biol 2020What this evidence cannot show
Abstract read; full text not opened. The tumorigenesis models in this paper are not leukaemia. This project holds `localization-condensate-evidence-standards` and `mechanism-condensates-load-bearing-or-correlative`, which set a higher bar than an abstract can meet - the specific assays behind 'liquid-like' here have not been audited against that bar by this beat.
Evidence that argues against this (1)
- An abstract cannot establish 'liquid-like', and this project's own condensate standards ask for a half-bleach and an in-cell 1,6-hexanediol test before the term is used. Neither was checked here. The stronger caution is about the claim that makes condensation look drug-like: 'mutations of key residues perturb condensation in opposite directions' implies a tunable dial, and none of the underlying assays, residues or effect sizes were audited by this beat. The tumorigenesis models in that paper are also not leukaemia. Yan X et al. 2026
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AKAP95 is required for MLL-rearranged leukaemogenesis: knockdown reduces growth of THP-1 (MLL-AF9), MOLM-13 (MLL-AF9) and MV4-11 (written MV411 in the paper, MLL-AFF1) but not K562 (BCR-ABL) or HL-60; Akap95-/- bone marrow transformed with MLL-AF9 gives longer recipient survival than wild type in primary (n=9 WT vs n=8 KO) and secondary (median survival WT 25.5 d vs KO 28 d, P=0.0022) transplant and fewer serial-replating colonies at equal MLL-AF9 protein; Akap95-/- bone marrow shows normal multilineage chimerism in 1:1 competitive transplant, so the requirement is leukaemia-specific and spares homeostatic and regenerative haematopoiesis.
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 not available, so the competitive-transplant chimerism (Supplementary Fig. 3b-d) and the MLL-AF9 protein-level control are the main text's descriptions. The survival difference in the secondary transplant is 2.5 days on a ~26-day median - statistically significant but small, and the disease still kills every recipient. This is a delay, not a block.
Evidence that argues against this (1)
- The paper's own limitations section states that the AKAP95/MLL-AF9 joint-target analysis rests substantially on HeLa cells and that the peptide's suppressive effect 'still needs to be tested on clinical patient samples'. The only index-fusion (KMT2A::AFF1) datum in the paper is a growth curve in MV4-11 after shRNA knockdown; there is no KMT2A::AFF1 in vivo model, no primary patient sample, and no ChIP or CUT&RUN in a KMT2A::AFF1 line. Yan X et al. 2026
- E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
AKAP8 is one of more than 19 reported ZNF384 fusion partners in acute lymphoblastic leukaemia, alongside EP300, CREBBP, TCF3, TAF15, EWSR1, ARID1B, SMARCA4, SMARCA2, SYNRG, CLTC, BMP2K, NIPBL, C11orf74, DDX42, ATP2C1, EHMT1 and TEX41; ZNF384-rearranged ALL generally carries a good prognosis.
ZNF384-Related Fusion Genes in Acute Lymphoblastic Leukemia. · Zhu L, Bai W, Cheng Q, Fang J · Cancer Control 2023What this evidence cannot show
This is a REVIEW cited as a map, which this project permits only as such. The primary report of an AKAP8::ZNF384 case was not located by this beat: a Europe PMC full-text search returned the UKALL14 whole-genome-sequencing cohort (PMID 36867579) and an adult BCP-ALL copy-number study (PMID 34657128) as body-text hits, but neither names AKAP8 in its abstract and neither full text was opened. The partner list is therefore unconfirmed against a primary source and this entry must not be used to assert a frequency.
Evidence that argues against this (1)
- Placing this line beside the MLLr dependency in the same summary invites a reader to weigh the two roles equally. They are not comparable in evidence mass: the leukaemia dependency rests on a mouse genetics paper with transplant, transcriptome and rescue arms, while the fusion-partner role rests on a review list plus a single WGS call in one adult patient. A gene entry should not let a partner-list mention and a full genetic study sit at the same apparent weight. Leongamornlert D et al. 2023
- E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
A second group reports that the PKA-binding (RII) domain of AKAP8 - not its N-terminus - is essential for direct interaction with DPY30, that the C-terminal Dpy-30 motif of DPY30 is required on the other side, that the single substitution DPY30 L69D disrupts DPY30 dimerisation and completely abolishes its interaction with both AKAP8 and BIG1, and that the paralogue AKAP8L also interacts with core subunits of the H3K4 HMT complexes.
PKA-binding domain of AKAP8 is essential for direct interaction with DPY30 protein. · Bieluszewska A, Weglewska M, Bieluszewski T, Lesniewicz K, Poreba E · FEBS J 2018What this evidence cannot show
Abstract read; the paper is not in PMC and the full text was not obtained, so the experiments behind the claim were not inspected. The claim is nonetheless load-bearing because it contradicts the 2013 mapping - see `interface-akap8-dpy30`, where the conflict is set out rather than resolved.
Evidence that argues against this (1)
- Jiang 2013 maps the DPY30-binding determinant to AKAP95's N-terminus and explicitly states that the AKAP95 regions responsible for binding DPY30 and RII 'seem to be different'. The two papers cannot both be right about which region is essential. Jiang H et al. 2013
- binds →DPY30 — the smallest COMPASS subunit, and the only one that behaves like a selective dependency
AKAP95 N-terminus (1-210) to the DPY30 Dpy-30 motif; the region assignment is disputed between two groups
AKAP95 IDR1 (101-390) co-condenses with the MLL1 translocated fragment - the region every KMT2A fusion retains
q160's adjudication of the same IntAct record; its shown_for_fusion and absence_class fields are now superseded by PMID 41501053
the AKAP95 N-terminal / RII determinant, disputed between two mappings
AKAP95 IDR1, residues 101-390
the enzymatic arm of AKAP95's function, and the arm that does NOT transfer to a SET-less fusion
the cell-state axis for AKAP95: silent at steady state, required during a transition
three subnuclear pools; the functionally required one is not the annotated one
the enumeration that found the leukaemia literature, and the three instrument failures that hid it
the four reconstructed residue intervals and their corroboration against UniProt
bridges AKAP95 to HSP70 and dissolves its condensates without changing its protein level
AKAP8 is the 5' partner, donating its N-terminal condensate-forming region
the bounded absence: 0 of 92 enumerated records, plus two full texts read end to end
the mechanism underneath the leukaemia phenotype: direct Pol II CTD binding, condensate size gated by nascent RNA
the two halves of AKAP8::ZNF384; written as a minimal stub so the 3D viewer can draw the fusion's 3' lane