Separate AKAP95's two routes into the KMT2A axis - starting with one transfection that re-reads a thirteen-year-old control
also called AKAP95 route separation experiment
A staged experiment to find out which of AKAP95's two grips on the KMT2A machinery the leukaemia actually needs. The first stage is a single cheap test using a construct that already exists.
The same thing, for a scientist
Three stages, each gating the next, designed so that the cheapest stage can kill the programme. **Stage 0 (one transfection, days):** co-express GFP-AKAP95(101-692) - the 2013 paper's existing negative control - with mCherry-MLL1T in 293T and score co-condensation. IDR1 is 101-390, so it should condense; if it does, the 2013 coactivation defect is attributable to the DPY30 route with the condensation route intact, and the two routes are already partly separated by an existing reagent. **Stage 1 (weeks):** settle the disputed DPY30 determinant head to head in one hand - Myc-DPY30 co-IP against full-length AKAP95, AKAP95(1-210), AKAP95(101-692), AKAP95(dRII = 1-571 + 590-692) and AKAP95 I582P, all five constructs published, four of them already built by Jiang 2013. Without this the route-dead mutant cannot be designed. **Stage 2 (months):** shRNA-resistant rescue of AKAP95 knockdown in THP-1, MOLM-13 and **MV4-11 or SEM** (the index-fusion arm, which no published experiment has) with WT, DPY30-dead and condensation-dead alleles at matched expression, scoring proliferation, H3K4me3 CUT&RUN at HOXA9/MEIS1, and AKAP95 CUT&RUN.
Controls the design must carry, and why each one
- Matched expression by immunoblot for every allele. The 2013 paper repeatedly had to argue around constructs that expressed poorly (AKAP95(1-530) was 'very low', dRII 'poor'), and a rescue experiment scored on proliferation is exquisitely sensitive to allele dose.
- A non-MLLr line. K562 and HL-60 are the published negatives; a rescue phenotype that appears in them is a general growth artefact.
- A normal CD34+ arm. The whole therapeutic argument is selectivity, and neither published study exposed a normal human haematopoietic cell to an AKAP95 perturbation. This arm can also detect the configuration neither hypothesis arm expresses - that the DPY30 route matters in the normal cell and not in the blast.
- Two fusion classes. Everything published is AF9-class. The index fusion must be its own arm, not an afterthought.
- A menin-inhibitor arm at Stage 2. Revumenib co-treatment is a one-plate addition and tests the hypothesis's sharpest prediction - that displacing the fusion from chromatin should make AKAP95 dispensable. Nothing in the 92-record AKAP95 literature combines it with any menin or DOT1L inhibitor, so any result is new.
What must be pre-registered. That a partial rescue by both single-route-dead alleles refutes the hypothesis that one route carries everything, and is not to be written up as 'the condensate route dominates'. And that Stage 2 is not run unless Stage 1 produces an unambiguous DPY30 determinant, because a mutant designed against a disputed mapping is uninterpretable whichever way it comes out.
| design | Stage 0: GFP-AKAP95(101-692) + mCherry-MLL1T co-transfection in 293T, score nuclear co-condensation against GFP-AKAP95 WT and GFP-AKAP95(1-100). Stage 1: Myc-DPY30 co-IP in G-293T against five published AKAP95 constructs (WT, 1-210, 101-692, dRII, I582P) in one hand, to settle the Roeder-versus-Poreba mapping. Stage 2: shRNA-resistant AKAP95 rescue in THP-1, MOLM-13 and MV4-11 (plus SEM if available) with WT, DPY30-dead and condensation-dead alleles at matched expression; negatives K562 and HL-60; a normal CD34+ arm; and a revumenib co-treatment arm. |
|---|---|
| readout | Stage 0: co-localisation and aggregate formation by confocal. Stage 1: co-IP immunoblot, DPY30 recovered per unit AKAP95. Stage 2: proliferation over 12 days; H3K4me3 CUT&RUN at HOXA9 and MEIS1; AKAP95 CUT&RUN for genome-wide redistribution; RNA-seq of the MLL-AF9 core target set. |
| controls | Matched allele expression by immunoblot for every construct; empty-vector and scrambled-shRNA arms; non-MLLr lines K562 and HL-60; normal CD34+ arm; two fusion classes; and for Stage 1 the reciprocal DPY30 L69D allele as a positive control for loss of AKAP8 binding. |
| cost tier | low for Stage 0, low-med for Stage 1, high for Stage 2 |
| timeline | Stage 0: 1-2 weeks. Stage 1: 6-8 weeks. Stage 2: 6-9 months. |
| gating | Stage 2 does NOT run unless Stage 1 yields an unambiguous DPY30 determinant. A route-dead mutant designed against a disputed mapping is uninterpretable in either direction. |
measurements
[
{
"quantity": "count",
"subject": "published AKAP95 constructs reusable in Stage 1 without new cloning",
"value": 5,
"units": "constructs",
"n": 5,
"universe": "PMID 23995757 Methods: full length, 1-210, 101-692, dRII (1-571+590-692), I582P",
"null": null,
"note": "two are noted in that paper as expressing poorly, which may force re-engineering for a dose-matched assay"
},
{
"quantity": "count",
"subject": "published experiments combining an AKAP95 perturbation with a menin or DOT1L inhibitor",
"value": 0,
"units": "experiments",
"n": null,
"universe": "the complete 92-record AKAP8/AKAP95 title-abstract literature plus the full texts of PMID 23995757 and 41501053",
"null": null,
"note": "the revumenib arm of Stage 2 is therefore new whatever it shows"
}
]- The Stage 2 rescue reads out on proliferation, which is exactly the assay AKAP95's own biology predicts will be least sensitive - a protein that is dispensable at steady state. A negative Stage 2 could mean the route does not matter, or that the endpoint cannot see it, and the design must carry a positive control that fails if the endpoint is blind.
- This experiment is designed to attribute a phenotype to one of two routes. If AKAP95 has a third route nobody has mapped - it is a 692-residue scaffold with hnRNP, DDX5, MCM2, condensin and Pol II CTD contacts on the books - then both alleles rescue partially and the design silently attributes a third mechanism's contribution to the two it happens to be testing.
- Can a DPY30-binding-dead AKAP95 be built at all, given that the two published mappings put the determinant at opposite ends of the protein? Stage 1 exists precisely because the answer is currently no.
- Is proliferation the right Stage 2 endpoint? AKAP95's phenotype is state-dependent, and a proliferation assay in an established line is the closest thing to a steady-state readout - which is where the published biology says the phenotype is weakest. A differentiation-induction endpoint may be more sensitive and is cheaper.
- Should the normal arm be CD34+ cells or an ATRA-differentiating system? The 2013 biology says the phenotype appears during a transition, so a quiescent CD34+ culture may show nothing and give a falsely reassuring selectivity result.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The constructs required for Stage 1 already exist in the published literature: AKAP95 full length, AKAP95(1-210), AKAP95(101-692), AKAP95 dRII (1-571 + 590-692) and AKAP95 I582P were all built and used in PMID 23995757, and the Myc-DPY30 co-IP assay in G-293T cells is that paper's own Fig 5E design.
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
'Exist in the literature' means they were described in Methods thirteen years ago, not that they are obtainable. Two of the deletion constructs are noted in the paper as expressing poorly, which is a real obstacle for a dose-matched rescue and may force re-engineering.
Evidence that argues against this (1)
- The Fig 5E co-IP design this Stage 1 reuses is an over-expression co-IP in G-293T cells, where both partners are transfected. It cannot detect a weak or transient contact and it is exactly the assay class that produced the disputed mapping in the first place. Repeating it in one hand resolves which result is reproducible, not which is correct; a Kd or a structure would be needed for that, and neither is proposed here because neither is affordable at this stage. Jiang H et al. 2013
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The Stage 2 reagents are standard: THP-1 (KMT2A::MLLT3), MOLM-13 (MLL-AF9) and MV4-11 (KMT2A::AFF1) all have published AKAP95 shRNA growth curves, and K562 and HL-60 are the published non-MLLr negatives, all in PMID 41501053.
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
The expected_absent_terms record two arms this design adds that the source does not contain: no normal human CD34+ cell and no menin inhibitor appears anywhere in that paper. The published rescue was done in THP-1 and MOLM-13 only; MV4-11 has a knockdown curve but no rescue.
Evidence that argues against this (1)
- A published knockdown curve shows the line is sensitive; it does not show a RESCUE is feasible there. The published rescue was done only in THP-1 and MOLM-13. MV4-11 has a knockdown curve and no rescue, and SEM appears nowhere in that panel. A rescue that fails in the index-fusion arm for transduction or expression reasons would be indistinguishable, on a proliferation readout, from a route-specific result - which is the single most likely way this design produces a confident wrong answer. Yan X et al. 2026
three gated stages; Stage 0 is one transfection that re-reads the 2013 negative control
Stage 1 settles the Roeder-versus-Poreba mapping head to head, using five already-published constructs
the Stage 2 menin-inhibitor arm - no published experiment combines an AKAP95 perturbation with any menin or DOT1L inhibitor