Oncofusion

start / graph / hyp-aff1-and-aff4-occupy-distinct-proximity-neighbourhoods

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q20-proximity-interactome

AFF1- and AFF4-containing super elongation complexes occupy physically distinct neighbourhoods in living cells, and KMT2A::AFF1 inherits the AFF1 one

also called SEC scaffold neighbourhood separation · AFF1 vs AFF4 proximity

AFF1 and AFF4 are interchangeable parts of the same machine, but in living cells they may sit in different places with different company. If so, the fusion built from AFF1 inherits AFF1's company, and treatments aimed at the machine as a whole would miss.

The same thing, for a scientist

In a 50-bait BioID2 map of K562, AFF1 and AFF4 are each recovered by four baits and share only one - ZRANB2 - despite being the alternative scaffolds of the same complex family, sampled by the same baits in the same cells [E1]. The hypothesis is that this reflects a real physical separation: AFF1-SEC and AFF4-SEC occupy different subnuclear neighbourhoods with different RNA-processing company, rather than being one interchangeable particle. **Prediction for this project:** KMT2A::AFF1 is built on the AFF1 scaffold and would inherit the AFF1 neighbourhood, so an intervention designed against SEC composition derived from AFF4-based structural work may not apply. **Prior probability 0.25**, because four against four is a small comparison and the AFF4 C-terminal homology domain is published to heterodimerise with AFF1 [E2].

Go deeper
The observation

AFF1's four baits: NPM1, RBM15, SBDS, ZRANB2. AFF4's four: GPKOW, HNRNPC, PPIL4, ZRANB2 1.

Both proteins sit at the 43.7th percentile of the dataset's baits-per-prey distribution, so neither is unusually detectable or unusually rare. The baits that reach AFF4 but not AFF1 - GPKOW, HNRNPC, PPIL4 - are spliceosome- and nascent-RNA-facing. The baits that reach AFF1 but not AFF4 - NPM1, RBM15, SBDS - are nucleolar, m6A-targeting and ribosome-maturation proteins.

Why this is not obviously true

The AFF4 C-terminal homology domain homodimerises and heterodimerises with AFF1, AFF2 and AFF3 2. If AFF1-AFF4 heterodimers are a substantial species, the two scaffolds are in the same particle some of the time and a disjoint proximity profile cannot mean physical separation.

And four against four is small. With 50 baits and a median of 5 baits per prey, two four-element subsets overlapping in one element is not a surprising outcome, and no test is offered because the baits are not a random sample 1. The hypothesis is recorded at a low prior for that reason.

What would kill it

See data.falsifiers. The decisive one is cheap: if AFF1 and AFF4 are each used as a bait in one cell and return overlapping neighbourhoods, the hypothesis is dead. That experiment has never been run for either protein.

Facts and figures
statementAFF1-containing and AFF4-containing super elongation complexes occupy physically distinct neighbourhoods in living human cells, and KMT2A::AFF1 inherits the AFF1 neighbourhood rather than a generic SEC one.
origindata-pattern
prior probability0.25
predictions
[
  "An AFF1 bait and an AFF4 bait run side by side in one cell line return neighbourhoods with less overlap than two technical replicates of either.",
  "The AFF1 neighbourhood is enriched for nucleolar and m6A-associated proteins; the AFF4 neighbourhood for spliceosomal proteins.",
  "KMT2A::AFF1-TurboID recovers the AFF1-side company (NPM1, RBM15, SBDS class) rather than the AFF4-side company (GPKOW, HNRNPC, PPIL4 class).",
  "Depleting AFF4 does not rescue AFF1-dependent phenotypes in a t(4;11) line, and vice versa."
]
experiments
[
  "exp-turboid-map-of-kmt2a-aff1",
  "exp-aff1-chd-dimerisation-separation-of-function"
]
analyses
[
  "ana-aff1-proximity-neighbourhood",
  "ana-sec-proximity-recovery-from-rbp-baits"
]
falsifiers
[
  "AFF1 and AFF4 baits in one cell return overlapping neighbourhoods. This kills the hypothesis outright and the experiment has never been run for either protein.",
  "AFF1 and AFF4 differ substantially in abundance in K562, so the disjoint four-and-four profile is a detection-threshold artefact.",
  "The overlap is one of four because of sampling: a permutation over the fifty baits shows one-element overlap is the modal outcome for two random four-element subsets.",
  "AFF1-AFF4 heterodimers are shown to be the dominant species in haematopoietic cells, making a single mixed particle the correct model.",
  "A KMT2A::AFF1 proximity map recovers the AFF4-side proteins, showing the fusion does not inherit the AFF1 neighbourhood."
]
How this could bite us back
  • If this is right, a large part of the field's structural knowledge of the SEC - which is AFF4-derived - is being applied to the wrong scaffold, and this graph inherits that substitution in several entries.
  • If it is wrong, the beat will have promoted a one-element overlap between two four-element sets into a mechanism, which is the exact failure mode this project has logged in three other contexts.
Still open
  1. Is there any orthogonal evidence - imaging, fractionation, published co-IP - that AFF1-SEC and AFF4-SEC are spatially separated in haematopoietic cells?
  2. If KMT2A::AFF1 inherits the AFF1 neighbourhood, does that predict anything about which SEC-directed interventions should work, or only about composition?
  3. Does the same disjointness appear for AFF2 and AFF3? Both are recovered by zero baits in this dataset, so the comparison cannot be made.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In the K562 50-bait BioID2 deposition, AFF1 is recovered by NPM1, RBM15, SBDS and ZRANB2, AFF4 by GPKOW, HNRNPC, PPIL4 and ZRANB2, sharing only ZRANB2; both counts place the two proteins at the 43.7th percentile of the dataset's baits-per-prey distribution.

    What this evidence cannot show

    Four against four is a small comparison over the authors' curated bona fide pair list, whose threshold is not stated. Baits are overexpressed BioID2 fusions in a cell line without a KMT2A rearrangement. Relative abundance of AFF1 and AFF4 in K562 was not measured. No statistical test is appropriate because the fifty baits are not a random sample of the proteome.

    Evidence that argues against this (2)
    • In the largest human BioID survey a prey's bait profile is dominated by subcellular compartment rather than binding, so any neighbourhood comparison must control for localisation before it means anything. Go CD et al. 2021
    • The AFF4 C-terminal homology domain both homo- and heterodimerises with AFF1, AFF2 and AFF3, so the two scaffolds can occupy the same particle; a disjoint proximity profile is compatible with heterodimers being a minority species rather than with physical separation. Chen Y et al. 2019
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The AFF4 C-terminal homology domain folds into an eight-helix domain that supports both homodimerisation and heterodimerisation among AFF family members, with AFF4 consisting of an intrinsically disordered N-terminal region that binds other super elongation complex subunits and this conserved C-terminal domain.

    PMID 31147444DOI
    What this evidence cannot show

    Structural work on isolated domains; abstract read, full text not read. It establishes that heterodimerisation is possible, not that heterodimers are abundant in cells.

    Evidence that argues against this (1)
    • A domain that can heterodimerise in vitro at the concentrations used for crystallography may not do so at cellular stoichiometry, and the proximity data are the only in-cell measurement available - which point the other way. IntAct molecular interaction database (EMBL-EBI) 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AFF1/AFF2/AFF3/AFF4 family members are the scaffold through which HIV-1 Tat and P-TEFb are organised in the super elongation complex, with the AFF4 N-terminal element resolved in complex with P-TEFb.

    PMID 24727379DOI
    What this evidence cannot show

    Structural and virological work built on AFF4 rather than AFF1. Cited here because it is an example of the substitution this hypothesis warns about: AFF4-derived structural knowledge being applied to AFF1 biology.

    Evidence that argues against this (1)
    • The AFF family N-terminal P-TEFb-binding element is conserved across AFF1-AFF4, so for that interface specifically the substitution is well justified and the hypothesis should not be read as impugning it. Chen Y et al. 2019
Connected to