Oncofusion

start / graph / ana-turboid-of-aml-oncofusions-already-done

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q20-proximity-interactome

The fusion-oncoprotein TurboID experiment this graph proposes has already been published - for PML::RARA, RUNX1::RUNX1T1, CBFB::MYH11, NUP98::HOXA9 and SET::NUP214, and not for any KMT2A fusion

also called oncofusion TurboID · Day 2023 TurboID · NUP98-HOXA9 BioID

Two groups have already attached a biotin-tagging enzyme directly to leukaemia fusion proteins and listed everything each one touches. They did it for five fusions. None of the five was a KMT2A fusion.

The same thing, for a scientist

Day et al. fused **TurboID in frame with three AML oncofusion cDNAs - PML::RARA, RUNX1::RUNX1T1 and CBFB::MYH11 - and identified their interacting proteins in primary murine haematopoietic cells**, finding that PML::RARA and RUNX1::RUNX1T1 label common and unique nuclear repressor complexes while CBFB::MYH11's partners are largely cytoplasmic, and going on to show that the CBFB domain sequesters RUNX1 in cytoplasmic aggregates [E1]. Separately, Mendes et al. used **BioID on NUP98::HOXA9 and SET::NUP214** [E2]. So the experiment `exp-turboid-map-of-kmt2a-aff1` proposes is **not a new experiment**: it is a published protocol, validated on five leukaemia fusions in two laboratories, that nobody has pointed at KMT2A. That changes the nature of the gap from technical to selective, and it supplies a method section.

Go deeper
Why this is the most actionable finding in the beat

The standing weak spot reads as though proximity labelling of a leukaemia fusion were an unattempted frontier. It is not. Day et al. built TurboID-oncofusion cDNA constructs and expressed them in primary murine haematopoietic cells - not a cell line - and recovered a mechanism nobody had seen: CBFB::MYH11 sequesters RUNX1 in the cytoplasm, which is a localisation result that affinity purification of nuclear extract would have been poorly placed to find 1.

That is precisely the class of result this graph keeps predicting it needs and cannot get: the CBFB::MYH11 finding overturned an assumption about where a fusion acts. A KMT2A::AFF1 version of the same experiment would cost a construct and a mass-spectrometry run against an established protocol.

Which three, and why probably not KMT2A

Day et al. chose the three favourable-risk AML oncofusions 1. KMT2A rearrangements are not favourable risk and would not have belonged in that set. Mendes et al. chose NUP98::HOXA9 and SET::NUP214, two nucleoporin fusions, for a nuclear-export question 2.

Neither omission is evidence that a KMT2A fusion is hard to label. Both are evidence that the field's proximity-labelling effort has been organised around other questions.

What a KMT2A::AFF1 version would and would not settle

Would settle: whether the fusion's neighbourhood in living cells contains the SEC components affinity purification places there; whether menin is a quantitatively minor or major feature of it; whether anything unexpected sits near it, as RUNX1 did for CBFB::MYH11; and whether the neighbourhood changes under menin inhibition, which no current assay reports.

Would not settle: composition. A proximity list is not a complex roster, and the head-to-head comparison on chromatin complexes found the two methods share few partners 3. A KMT2A::AFF1 proximity map would add a list, not adjudicate the existing one - and the standing weak spot's expectation that 'a single modern proximity map could overturn several negatives at once', including the WRAD negative, is over-optimistic for that reason. In the one dataset available, the WRAD subunits split 11/9/0/0 across the same fifty baits.

Facts and figures
scriptnone (literature)
statusrun
method precedentTurboID fused in frame to oncofusion cDNA, expressed in primary murine haematopoietic cells (PMID 38061017)
result summaryThe experiment is published for five leukaemia fusions and has never been pointed at a KMT2A fusion. The gap is selective, not technical, and a validated protocol exists.
fusions with published proximity maps
[
  "PML::RARA",
  "RUNX1::RUNX1T1",
  "CBFB::MYH11",
  "NUP98::HOXA9",
  "SET::NUP214"
]
fusions without
[
  "KMT2A::AFF1",
  "KMT2A::MLLT3",
  "KMT2A::MLLT1",
  "KMT2A::MLLT10",
  "KMT2A::ELL",
  "KMT2A::AFDN"
]
publications
[
  "38061017",
  "32664447"
]
How this could bite us back
  • A published protocol makes the experiment easier, not more likely to be informative. If the KMT2A::AFF1 neighbourhood comes back as a nucleoplasmic census, the project will have spent a construct to learn the method's bias.
  • The CBFB::MYH11 result is a warning as much as an encouragement: the most interesting finding in that paper was about LOCALISATION, and this graph's model of KMT2A::AFF1 assumes nuclear chromatin throughout.
  • Tagging a fusion oncoprotein with a 35 kDa ligase and overexpressing it risks producing a neighbourhood the untagged protein never has.
Still open
  1. Would a KMT2A::AFF1-TurboID construct fold and localise correctly? The fusion is ~250 kDa before the tag, far larger than any of the five fusions already mapped.
  2. Day et al. worked in primary murine haematopoietic cells. Could the same be done in human cord-blood CD34+ cells, given that this project's cell-of-origin work argues the murine system is the wrong one?
  3. Does the neighbourhood of KMT2A::AFF1 change under revumenib? A proximity map before and after is the only assay that would show what menin inhibition does to composition rather than to output.
  4. Should the reciprocal fusion AFF1::KMT2A be mapped at the same time? It is expressed in most t(4;11) patients and has no interactome of any kind.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Day et al. fused the promiscuous biotin ligase TurboID in frame with three favourable-risk AML oncofusion cDNAs - PML::RARA, RUNX1::RUNX1T1 and CBFB::MYH11 - and identified their interacting proteins in primary murine haematopoietic cells; PML::RARA and RUNX1::RUNX1T1 labelled common and unique nuclear repressor complexes while CBFB::MYH11's partners were largely cytoplasmic, and the CBFB domain was shown to sequester RUNX1 in cytoplasmic aggregates.

    PMID 38061017DOI
    What this evidence cannot show

    Abstract read; full text not read. Primary MURINE haematopoietic cells with overexpressed fusion cDNAs carrying a 35 kDa tag, which can perturb localisation and stoichiometry. No KMT2A fusion is included. The three fusions chosen are the favourable-risk group, a selection criterion that excludes KMT2A rearrangement by definition.

    Evidence that argues against this (2)
    • In the largest human BioID survey (192 baits, HEK293) a prey's bait profile is dominated by subcellular compartment rather than by binding, so any fusion-protein neighbourhood is a localisation census before it is an interaction list. Go CD et al. 2021
    • Oncogenic mutant forms of MLL retain the ability to interact with menin but not other identified complex components; any result obtained on a different fusion, or on the wild-type protein, has to be re-shown for KMT2A::AFF1. Yokoyama A et al. 2004
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Mendes et al. used proximity-dependent biotin identification (BioID) to map the environments of the leukaemogenic NUP98-HOXA9 and SET-NUP214 fusion proteins, reporting that both interact with RNA-processing and translation-associated proteins and perturb the p53 transcriptional programme, with distinct additional processes for each.

    PMID 32664447DOI
    What this evidence cannot show

    Abstract read; full text not read. The cell system is not named in the abstract. The reported outputs are pathway-level enrichments from a proximity hit list, which is the weakest form of interpretation the method supports.

    Evidence that argues against this (2)
    • BioID and affinity purification of chromatin-associated complexes recover few shared partners and enrich for distinct processes, so a proximity map will not simply confirm or refute an affinity-purification roster - it will produce a different list. Lambert JP et al. 2015
    • Pathway enrichment over a proximity hit list inherits the compartment bias of the method: a nuclear fusion protein will enrich for nuclear pathways whatever it binds, and RNA-processing enrichment in particular is what any nucleoplasmic bait returns. Go CD et al. 2021
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    A head-to-head comparison of BioID against an affinity-purification protocol optimised for chromatin-associated complexes found the two approaches identified few shared interaction partners and enriched for distinct biological processes, while both recovered biologically meaningful interactions.

    PMID 25281560DOI
    What this evidence cannot show

    HEK293, on histones and Mediator rather than on a fusion oncoprotein. It constrains what a future KMT2A::AFF1 proximity map could settle; it is not evidence about KMT2A::AFF1.

    Evidence that argues against this (1)
    • The same paper's conclusion is that the methods are complementary, so low overlap is an argument for doing both rather than an argument that a proximity map would be uninformative. Lambert JP et al. 2015
Connected to