Oncofusion

start / graph / hyp-oligomerisation-interfaces-as-mutation-resistant-targets

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q11-comparative

Hypothesis — forced-oligomerisation interfaces are the mutation-resistant drug targets the fusion field has been ignoring

also called oligomerisation interface hypothesis · NHR2 analogy · self-association target class

Several fusion proteins only work because the partner glues copies of them together, and the idea here is that a drug that unglues them could not be escaped by a simple mutation. One case has now been checked against patients and it went the other way: arsenic, which acts on exactly such a glue in a different leukaemia, was defeated by a single amino-acid change while the cancer protein kept working.

The same thing, for a scientist

Across four independent fusion families the 3-prime partner contributes a self-association module that is strictly required for oncogenic function: the ETO NHR2 tetramerisation domain in RUNX1::RUNX1T1, the PML RBCC coiled-coil in PML::RARA, the BCR coiled-coil in BCR::ABL1, the NPM1 oligomerisation domain in NPM1::ALK and the ETV6 SAM/PNT polymerisation domain in ETV6::NTRK3. We hypothesise that these interfaces form a distinct and undervalued target class with a property no active-site or adaptor-interface target has: the drug-binding surface and the function-critical surface are the same, so escape mutations that abolish drug binding also abolish oncogenicity. If correct, oligomerisation-interface inhibitors should be intrinsically resistance-resistant, and the reason none has reached the clinic is chemistry and neglect rather than biology. The cycle-33 counter-pass found one adverse precedent that the original version missed: arsenic trioxide acts at the PML RBCC and is escaped in patients by PML B2-domain missense mutation, with the fusion intact and the disease progressing (PMID 21613260). It also answered the druggability question: exactly one oligomerisation-interface small molecule is approved anywhere in medicine -- tafamidis, which STABILISES the transthyretin tetramer -- and -- **CORRECTED cycle 114** -- one DOES disrupt an oncogenic oligomer, though none is approved. Compound **7.44** binds the NHR2 domain of RUNX1::RUNX1T1 (RUNX1/ETO) and shifts the tetramer to a dimer, Kd 3.75 +/- 1.22 uM, restoring repressed gene expression and reducing tumour growth in mice [PMID 35986043, Sci Rep 2022, 10.1038/s41598-022-17913-6]; non-charged analogues reach Kd 39-114 uM with cellular IC50 33-77 uM [PMID 40399488, Sci Rep 2025]. It is a small molecule, not a peptide, metalloid or stabiliser. **The APPROVAL half of the original claim stands and the conclusion survives for a better reason:** the class is a chemistry programme not because no disruptor exists but because the best one after a decade sits at tens of micromolar. The hypothesis is therefore untested rather than supported, and its prior has been reduced from 0.45 to 0.2.

Go deeper
The recurring structural motif nobody has systematised1 deeper

Look at what the 3-prime partners actually contribute, stripped of family history:

FusionOligomerisation moduleEvidence it is required
RUNX1::RUNX1T1ETO NHR2 (tetramer)Interfering peptide dissolves the complex, restores target genes, kills t(8;21) cells 12
PML::RARAPML RBCC coiled-coil (dimer)Forced RARA homodimers alone prime mice for APL-like disease 3
BCR::ABL1BCR coiled-coilLigand-independent trans-autophosphorylation requires it (established kinase biology)
NPM1::ALKNPM1 oligomerisation domainSame activation logic as BCR::ABL1
ETV6::NTRK3ETV6 SAM/PNT polymerisationSame activation logic

Four of five are coiled-coil or helical-bundle self-association interfaces. These are not the flat, featureless surfaces that gave protein-protein interactions their undruggable reputation; helical-bundle interfaces have been drugged elsewhere (BCL-2 family, HIV gp41). The field has treated them as structural trivia because attention went to whatever the other half did.

The asymmetry that makes this class special

For an active-site inhibitor, the drug site and the catalytic site overlap but are not identical, which leaves room for mutations that keep catalysis and lose drug binding — imatinib gatekeeper mutants, and the MEN1 interface mutants that keep menin-MLL1 binding while evicting revumenib.

At a self-association interface there is far less room. The residues that mediate the coiled-coil pairing are the function. A mutation that stops a small molecule intercalating in that groove is likely to destabilise the oligomer. The prediction is therefore quantitative and testable: a deep mutational scan of an oligomerisation interface should show a much smaller set of variants that are simultaneously drug-resistant and oncogenically competent than an equivalent scan of the menin pocket.

Does KMT2A::AFF1 have one? This is the actionable question

Unknown, and it is the single cheapest high-value experiment in this memo. AFF1 and AFF4 are long scaffolds with predicted helical and low-complexity segments; the super elongation complex assembles around them 4. Correction, 2026-09-19. This section previously said the AFF1 self-association question 'has not been tested with interface point mutants'. That is wrong as written, and the error mattered because it made the experiment look harder than it is. The interface is solved and the point mutants already exist - they have simply never been put into a KMT2A fusion.

Chen et al. solved the AFF4 C-terminal homology domain at 2.2 A and showed it mediates AFF4 homodimers and AFF1-AFF4 heterodimers 8. Tang et al. solved the same region (which they call THD) at 2.4 A and went further: single alanine substitutions AFF4 F1014A and Y1096A abolish dimer formation and cripple transactivation, and the equivalent AFF1 residues Phe1063 and Tyr1145 do the same for AFF1-dependent transactivation - while leaving the interaction of AFF1/4 with the other SEC subunits intact 9. That last clause is what makes this a separation-of-function reagent rather than a blunt deletion: the mutant still assembles the complex, it just cannot dimerise.

So the true statement is narrower and much more actionable: AFF1 F1063A / Y1145A have never been tested in the KMT2A::AFF1 fusion context, in HIV transactivation or anywhere else in a leukaemia. The experiment is now a two-residue mutant knocked into an existing fusion construct, with a published positive control for the mutation's effect on dimerisation, and it reads out on HOXA9/MEIS1 induction and menin docking.

If KMT2A::AFF1 does depend on an AFF1 oligomerisation interface, the lab gains a second target that (a) is on the fusion side rather than on a wild-type host protein, so normal-tissue toxicity is lower, and (b) belongs to this putatively mutation-resistant class. That would be a strictly better target than menin on both axes, and it would be partner-specific rather than partner-agnostic — worth knowing either way.

Counter-pass (cycle 33): the resistance argument has a clinical counterexample, and the druggability question now has an answer

1. The resistance argument fails where it should be strongest. The hypothesis's load-bearing claim is that at an oligomerisation interface the drug-binding surface and the function-critical surface are the same, so escape mutations are self-defeating. The one fusion oligomerisation module with a clinically used agent aimed at it -- the PML RBCC in PML::RARA, targeted by arsenic trioxide -- has been escaped in patients by PML B2-domain missense mutation, with arsenic-induced SUMOylation, oligomerisation and insolubility all abolished while the fusion persisted and the leukaemia progressed 6. PML A216V is now detectable by ddPCR before overt resistance. The prediction that this class 'should be intrinsically resistance-resistant' therefore has a published counterexample inside one of its own four exemplar families.

The honest defence is that arsenic is a cysteine-reactive metalloid that promotes PML oligomerisation rather than an interface inhibitor that disrupts it, so the geometry is not identical. That defence narrows the hypothesis to: interfaces where the drug acts by steric disruption of the oligomer itself. No such molecule exists in oncology, so the claim is currently untested rather than supported.

2. Has any oligomerisation interface ever yielded a clinical molecule? The answer, after a deliberate search, is yes once, in the wrong direction, outside oncology: tafamidis binds the transthyretin tetramer interface, prevents dissociation, and in the randomised ATTR-ACT trial lowered all-cause mortality (29.5% vs 42.9%) and cardiovascular hospitalisation; it is approved 7. It stabilises a native tetramer. Nothing in oncology has been approved that disrupts an oncogenic oligomer. Within the families named in this entry: ETO NHR2 has peptides only and no drug-like molecule 12; the BCR coiled coil has never been drugged (CML's second site is the ABL1 myristoyl pocket, an autoinhibitory site, not the BCR oligomerisation interface); the NPM1 and ETV6 oligomerisation domains have no chemical matter; and PML has arsenic, which is the case above.

3. The one interface this graph has measured is large and distributed. Cycle-32 work found PML self-association dominated by a long anti-parallel coiled coil with at least three separate hydrophobic pockets, where single point mutants lose only 74-84% of self-interaction [see protein-pml-domains]. That cuts both ways and the two directions are not independent: interfaces hard to escape with one mutation are hard to occupy with one molecule. The hypothesis wants the first property without paying for the second.

Net. The hypothesis is not refuted, but it has been moved from 'undervalued target class' to 'untested target class with one adverse precedent and no oncology molecule'. data.prior_probability is reduced accordingly and a new falsifier records the arsenic case.

Facts and figures
statementPartner-supplied forced-oligomerisation interfaces (ETO NHR2, PML coiled-coil, BCR coiled-coil, NPM1 oligomerisation domain, ETV6 SAM/PNT) constitute a distinct fusion drug-target class in which the drug-binding surface and the function-critical surface coincide, making acquired resistance by point mutation far less available than for active-site or adaptor-interface drugs.
origindata-pattern
prior probability0.2
why this is cheap to testAll of the required constructs are point mutants of existing fusion cDNAs, and the readouts (colony replating, HOXA9/MEIS1 qPCR, immunoprecipitation for complex integrity) are standard; no new chemistry is needed to falsify the biological half of the hypothesis. Updated 2026-09-19: cheaper than previously stated for the KMT2A arm, because the AFF1 interface residues are published and validated (F1063/Y1145, PMID 32128251) - the work is a two-residue knock-in, not an interface hunt.
prior probability noteReduced from 0.3 in the cycle-33 counter-pass. The reduction is driven by one adverse clinical precedent (PML B2-domain mutation defeating arsenic while the oncoprotein persists, PMID 21613260) and by the absence of any approved oligomerisation-DISRUPTING small molecule in oncology after decades of interest.
predictions
[
  "A deep mutational scan of an oligomerisation interface will yield far fewer variants that are simultaneously drug-resistant and transformation-competent than an equivalent scan of the menin drug pocket.",
  "Interface point mutants (not deletions) in ETO NHR2, PML coiled-coil or the NPM1 oligomerisation domain will abolish transformation in colony and transplant assays.",
  "If KMT2A::AFF1 depends on an AFF1 self-association or AFF1-AFF4 heteromerisation interface, the published interface point mutants AFF1 F1063A / Y1145A (PMID 32128251) placed in the fusion will abolish HOXA9/MEIS1 induction while leaving menin docking and SEC subunit binding intact, cleanly separating the two targets.",
  "Helical-bundle oligomerisation interfaces in this set will score as more ligandable in computational hot-spot analysis than the diffuse IDR interactions of the condensate class.",
  "If the class is real, a ligand occupying one coiled-coil groove should abolish function. For PML the assembly is multivalent across neighbouring dimers (PMID 39587079), so this predicts single-site occupancy will give partial loss only, and the experiment that decides it is a titration of occupancy against program output rather than a binary disruption assay."
]
experiments
[
  "exp-cross-family-interface-disruption-panel"
]
analyses
[
  "ana-comparative-interface-druggability-scan"
]
falsifiers
[
  "Deep mutational scanning of an oligomerisation interface recovers an abundant set of variants that retain oligomerisation and transformation while losing small-molecule binding, showing the class is no more resistance-proof than any other.",
  "Interface point mutants that demonstrably abolish oligomerisation in vitro nonetheless leave transformation intact, showing oligomerisation is dispensable and the whole target class is illusory.",
  "AFF1 F1063A / Y1145A interface mutants in KMT2A::AFF1 leave the HOXA9/MEIS1 program and transformation intact, removing the KMT2A relevance even if the class is real elsewhere.",
  "Computational and fragment-screening assessment shows these coiled-coil grooves have no ligandable hot spots, making the class biologically real but chemically inaccessible.",
  "ALREADY OBSERVED, in a weakened form: a clinically used agent acting at a fusion oligomerisation module (arsenic trioxide at the PML RBCC) is defeated in patients by a point mutation at its binding surface, with the oncoprotein intact and the disease progressing (PMID 21613260). The hypothesis survives only if 'oligomerisation interface' is narrowed to steric disruption of the oligomer, for which no molecule exists.",
  "A structural survey shows that fusion oligomerisation interfaces are typically large and distributed -- as PML's is, with single mutants retaining 16-26% of self-association -- so that buried surface areas fall outside the range small molecules can occupy in most families. Then mutation-resistance and undruggability are the same property and the class has no therapeutic value.",
  "The class's two showcase precedents do not belong to it. Arsenic at the PML RBCC INDUCES oligomerisation rather than blocking it (PMID 20378816), and tafamidis occupies transthyretin's natural thyroxine sites and STABILISES the tetramer against dissociation (PMID 22645360). If no existing approved or clinical agent actually disrupts an obligate oligomerisation interface, the hypothesis has no precedent and its prior should fall accordingly."
]
has any oligomerisation interface yielded a clinical molecule
{
  "answer": "Yes once, and in the opposite direction, outside oncology.",
  "case": "tafamidis binds the transthyretin tetramer interface and prevents dissociation; ATTR-ACT (NCT01994889) all-cause mortality 29.5% vs 42.9% on placebo",
  "pmid": "30145929",
  "why_it_does_not_support_the_hypothesis": "it STABILISES a native oligomer rather than disrupting an oncogenic one, and it is not an oncology drug",
  "oncology_status": "no approved small molecule disrupts an oncogenic oligomerisation interface; ETO NHR2 has peptides only, the BCR coiled coil has never been drugged (asciminib targets the ABL1 myristoyl autoinhibitory pocket, not the BCR oligomerisation interface), and NPM1/ETV6 oligomerisation domains have no chemical matter",
  "nearest_oncology_case": "arsenic trioxide at the PML RBCC -- a cysteine-reactive metalloid that promotes oligomerisation, and is escaped by PML B2-domain point mutation (PMID 21613260)",
  "title": "Tafamidis Treatment for Patients with Transthyretin Amyloid Cardiomyopathy.",
  "authors": "Maurer MS, Schwartz JH, Gundapaneni B, et al.",
  "venue": "N Engl J Med",
  "year": 2018,
  "title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
aff1 interface point mutants available
{
  "aff4": [
    "F1014A",
    "Y1096A"
  ],
  "aff1_equivalent": [
    "F1063A",
    "Y1145A"
  ],
  "domain": "THD / C-terminal homology domain (CHD)",
  "published_effect": "abolish AFF1/4 dimerisation and transactivation without disrupting binding to other SEC subunits",
  "sources": [
    "32128251",
    "31147444"
  ],
  "never_tested_in": "any KMT2A fusion",
  "recorded": "2026-09-19"
}
How this could bite us back
  • Oligomerisation domains are frequently shared with wild-type partner proteins that have normal cellular roles (PML in nuclear bodies, NPM1 in ribosome biogenesis, ETV6 in haematopoiesis), so a drug that blocks the fusion oligomer may also poison the normal assembly.
  • If resistance cannot arise by point mutation it will arise another way — gene amplification to overwhelm the drug, bypass signalling, or lineage switching — so 'resistance-resistant' may only mean 'resistant differently'.
  • Helical-bundle PPI drugs have historically needed very high potency to compete with high-avidity self-association, which translates into large doses and off-target burden in small children.
  • Mutation-resistance and undruggability may be the same property. An interface that one point mutation cannot break is an interface with a large, distributed buried surface, which is the kind small molecules occupy worst.
  • The only approved drug that binds an oligomerisation interface (tafamidis) works by stabilising the oligomer. If the tractable pharmacology at these surfaces is stabilisation rather than disruption, an oncology programme built on disruption is choosing the harder direction of the only precedent it has.
  • If an oligomerisation interface is multivalent - many equivalent contacts across neighbouring protomers rather than one compact two-helix groove - then partial occupancy buys partial loss of function, and the dose needed for a phenotype approaches complete target engagement. That is the same narrow-window problem the condensate class has, arriving by a different route.
  • The interface this hypothesis wants to drug in the KMT2A row (AFF4/AFF1 self-association) is the normal assembly interface of the normal super elongation complex, used by unmodified cells during ordinary transcriptional induction. A drug against it is not fusion-selective by construction, and its therapeutic index would have to come entirely from differential dependency.
Still open
  1. Does KMT2A::AFF1 require any AFF1-mediated self-association or AFF1-AFF4 heteromerisation, and has anyone made the interface point mutants?
  2. Do coiled-coil grooves in this set contain ligandable hot spots by fragment screening, or is the class biologically obligate but chemically inert?
  3. Because BCR::ABL1 has both an active site and an oligomerisation interface, could an oligomerisation-interface inhibitor be tested in imatinib-resistant CML as the fastest clinical proof of the whole concept?
  4. ANSWERED YES, cycle 114: compound 7.44 disrupts the RUNX1::RUNX1T1 NHR2 tetramer (PMID 35986043), a small molecule rather than a peptide, metalloid or stabiliser. The follow-on question is now about POTENCY, not existence: the best analogues after a decade are at Kd 39-114 uM and cellular IC50 33-77 uM (PMID 40399488). Is there any oncogenic oligomerisation interface anywhere with a genuine steric-disruptor small molecule, as opposed to a peptide, a metalloid or a stabiliser? If the answer after a full search is no, the class's problem is chemistry and the hypothesis should be reframed as a chemistry programme, not a biology one.
  5. Does the PML B2/arsenic escape generalise -- i.e. for each exemplar family, is there a residue set that can be mutated to abolish drug binding while retaining enough self-association for oncogenicity? PML's 74-84% loss per single mutant suggests there is headroom.
Every source behind this entry (9)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Oligomerisation through the ETO NHR2 domain is essential for RUNX1/ETO oncogenic activity, and expressing a competing NHR2-containing peptide disrupts the high-molecular-weight complex, restores target-gene transcription, reverts the differentiation block and causes cell death.

    PMID 17332359DOI
    What this evidence cannot show

    Intracellularly expressed 128-residue peptide in cell lines, no in vivo efficacy, and no demonstration that a small molecule can occupy the same groove.

    Evidence that argues against this (1)
    • The graph's own answer to 'has any oligomerisation interface ever yielded a clinical molecule' is a qualified yes, and the qualification is fatal to the proposed mechanism. Tafamidis binds the transthyretin tetramer interface, prevents tetramer dissociation and reduced all-cause mortality and cardiovascular hospitalisation in a 441-patient randomised trial; it is approved. But it works by STABILISING a native oligomer, not by disrupting an oncogenic one, and it is not oncology. No oligomerisation-DISRUPTING small molecule has been approved for any cancer. Maurer MS et al. 2018
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    An exogenously delivered cell-penetrating TAT-NHR2 polypeptide reduced proliferation and increased apoptosis in the t(8;21) line Kasumi-1, with specificity demonstrated against an unrelated BCR oligomerisation-domain control peptide.

    PMID 23865046DOI
    What this evidence cannot show

    One cell line, TAT-mediated delivery with known artefacts, no animal data, low-impact venue; supports interface relevance only, not the modality.

    Evidence that argues against this (1)
    • The readout that makes this look like delivered intracellular target engagement is the one with a documented artefact. Apparent energy-independent entry of TAT and other arginine-rich cell-penetrating peptides was traced to redistribution of peptide and cargo during the fixation step of fluorescence microscopy, and to strong ionic binding at the cell surface that overestimates cell-associated signal; uptake is endocytic and energy-dependent. A TAT-NHR2 polypeptide that reduces Kasumi-1 proliferation may therefore be acting with very little peptide free in the nucleoplasm, and the experiment does not establish that an oligomerisation interface is drug-accessible in a cell - which is the step this hypothesis most needs. Vives E et al. 2003
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Forced RARA homodimerisation is by itself sufficient to prime mice for APL-like leukaemia, showing that partner-supplied oligomerisation can be the essential oncogenic contribution of a fusion.

    Forced retinoic acid receptor alpha homodimers prime mice for APL-like leukemia · Sternsdorf T, Phan VT, Maunakea ML, et al. · Cancer Cell 2006
    PMID 16473276DOI
    What this evidence cannot show

    Engineered constructs with long latency that prime for rather than fully reproduce human APL; sufficiency of dimerisation does not prove that dimerisation is the only drug-relevant requirement.

    Evidence that argues against this (1)
    • PML's self-association interface is not a compact hot spot. Cycle-32 work in this graph on Tan 2024 (PMID 39587079) found a long anti-parallel coiled coil with at least three separate hydrophobic pockets (L240/L247/I279 against L297-L359), in which single point mutants lose only 74-84% of self-interaction. A distributed interface of that size is hard to escape with one mutation -- which supports the hypothesis -- and equally hard to occupy with one small molecule, which undercuts it. The two halves of the claim trade off against each other rather than reinforcing. Tan Y et al. 2024
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The super elongation complex shared by all KMT2A chimeras is nucleated by the AFF1/AFF4 scaffold, whose depletion reduces KMT2A-chimera target gene expression.

    PMID 20159561DOI
    What this evidence cannot show

    Establishes scaffold dependence, not self-association; no interface point-mutant experiment on AFF1 in the fusion context has been published, so the KMT2A arm of this hypothesis rests on analogy alone.

    Evidence that argues against this (2)
    • This evidence item undercuts the hypothesis it is filed under. If the shared SEC scaffold is what matters, the target is AFF1/AFF4 and not an oligomerisation interface; and if oligomerisation interfaces are the target, then the interfaces are demonstrably NOT interchangeable, which removes the mutation-resistance argument. MLL-SEPT6 lacking its coiled coil failed to immortalise with or without 4-OHT, and only the intact fusion self-associated - the authors require 'proper whole structure'. Heterologous RA domains also failed to substitute for AFDN RA1. One drug per interface, not one drug for many fusions. Journal of Clinical Investigation 2005
    • The second substitution failure is independent of the first and in a different partner class, which is what makes it hard to dismiss as a construct artefact: replacing the AFDN RA1 domain with heterologous Ras-association domains did not reconstitute transformation, even though RA1's own Ras-binding-dead mutants still transform. The self-association surface is partner-specific. Blood 2010
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Resistance to a protein-interface drug can arise from mutations that abolish drug binding while preserving the biological interaction, which is the failure mode this hypothesis predicts oligomerisation interfaces will avoid.

    MEN1 mutations mediate clinical resistance to menin inhibition · Perner F, Stein EM, Wenge DV, et al. · Nature 2023
    PMID 36922589DOI
    What this evidence cannot show

    Demonstrates the failure mode for one adaptor interface; whether oligomerisation interfaces genuinely have less mutational room is the untested core of this hypothesis.

    Evidence that argues against this (2)
    • The hypothesis's central claim has a direct clinical counterexample, and it is in one of its own four exemplar families. Arsenic trioxide acts on the PML RBCC oligomerisation module of PML::RARA; two of fifteen relapsed APL patients treated with arsenic were clinically resistant, and both carried missense mutations in the PML B2 domain. The mutant fusion no longer SUMOylated, no longer oligomerised on arsenic and no longer shifted into the insoluble fraction -- yet it remained an intact oncoprotein and the leukaemia progressed. So a mutation at an oligomerisation-domain drug site abolished drug action WITHOUT abolishing oncogenicity. That is exactly the escape route this hypothesis predicts oligomerisation interfaces cannot take. Goto E et al. 2011
    • PML A216V is detectable by droplet digital PCR before overt clinical arsenic resistance, so the escape clone behaves like any other drug-binding-surface resistance mutation. Alfonso V et al. 2019
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Arsenic resistance in APL arises from PML B2-domain missense mutations that abolish arsenic-induced oligomerisation, SUMOylation and insolubility of PML-RARA while the oncoprotein remains intact and the leukaemia progresses.

    Missense mutations in PML-RARA are critical for the lack of responsiveness to arsenic trioxide treatment. · Goto E, Tomita A, Hayakawa F, Atsumi A, Kiyoi H, Naoe T · Blood 2011
    PMID 21613260DOI
    What this evidence cannot show

    Abstract only. 2 of 15 relapsed/refractory APL patients; mechanism reconstructed by overexpression in HeLa. Small n. Arsenic is also not an interface inhibitor in the medicinal-chemistry sense -- it is a cysteine-reactive metalloid that PROMOTES oligomerisation -- so this is a counterexample to the hypothesis's resistance argument rather than to its druggability argument. It is nonetheless the only case where a clinically used agent acts at a fusion oligomerisation module, and it was escaped by point mutation.

    Evidence that argues against this (1)
    • This item is carried as a weakened already-observed falsifier, and it is weaker still than that, because arsenic is not an oligomerisation-interface DISRUPTOR. Arsenic binds directly to cysteine residues in the zinc fingers of the PML RBCC domain, and that binding INDUCES PML oligomerisation, which increases UBC9 recruitment, SUMOylation and degradation. The B2 missense mutants therefore defeat a drug-induced GAIN of oligomerisation, not a blockade of an obligate one; they say nothing about whether a ligand occupying a coiled-coil groove could be escaped by point mutation. The honest statement is that the PML row is not a test of this hypothesis's modality at all, and the hypothesis currently has no clinical test case. Zhang XW et al. 2010
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    An oligomerisation interface has yielded an approved small molecule -- tafamidis, which binds the transthyretin tetramer interface, prevents dissociation, and lowered all-cause mortality (29.5% vs 42.9%) and cardiovascular hospitalisation in the randomised ATTR-ACT trial -- but it acts by stabilising a native oligomer outside oncology, and no oligomerisation-disrupting small molecule has been approved for any cancer.

    Tafamidis Treatment for Patients with Transthyretin Amyloid Cardiomyopathy. · Maurer MS, Schwartz JH, Gundapaneni B, et al. · New England Journal of Medicine 2018
    PMID 30145929DOINCT01994889
    What this evidence cannot show

    Abstract only. Non-oncology, and the direction of action is the opposite of what this hypothesis proposes: stabilisation of a physiological tetramer rather than disruption of an oncogenic one. It establishes that oligomerisation interfaces are in principle addressable by oral small molecules, and nothing more.

    Evidence that argues against this (1)
    • Tafamidis is the wrong precedent in two directions and both matter. It does not bind an oligomerisation interface: it occupies the two normally unoccupied THYROXINE-binding sites of the transthyretin tetramer - a pre-existing natural small-molecule pocket - with Kds of about 2 nM and 200 nM and negative cooperativity. And it does not disrupt the oligomer: it kinetically STABILISES the tetramer, the crystal structure showing it braces the weaker dimer-dimer interface against dissociation, because dissociation is the rate-limiting step of amyloidogenesis. The therapeutic logic is the opposite of this hypothesis's. TTR is also an abundant secreted plasma protein with an evolved ligand site; a fusion partner's intracellular coiled coil has neither property, so 'an oligomerisation interface has yielded an approved small molecule' does not transfer. Bulawa CE et al. 2012
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The X-ray structure of the human AFF4 C-terminal homology domain (CHD) at 2.2 A shows a novel eight-helix fold distantly related to TPR motifs, and biochemical analysis shows the CHD mediates formation of an AFF4 homodimer or an AFF1-AFF4 heterodimer.

    PMID 31147444DOI
    What this evidence cannot show

    Recombinant domains in vitro; no cellular test, no leukaemia cell and no KMT2A fusion. Establishes that the interface exists and where it is, not that a fusion needs it.

    Evidence that argues against this (1)
    • The AFF4 CHD dimerisation surface is not partner-supplied and not fusion-specific: it is the normal assembly interface of a normal super elongation complex, and the SEC assembles and functions this way in unmodified cells during ordinary rapid transcriptional induction, with ENL's disordered region initiating droplet formation and AFF4 fluidising it within minutes of serum exposure. Whatever selectivity window a CHD-directed ligand has would therefore come from dependency, not from the interface being a neomorphic feature of the fusion - which is the property this hypothesis uses to argue that the drug-binding surface and the function-critical surface coincide in a fusion-selective way. For KMT2A::AFF1 specifically the relevant AFF1 self-association mutants have not been placed in the fusion, so this row is a structural observation awaiting a functional test. Guo C et al. 2020
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The crystal structure of human AFF4-THD at 2.4 A shows helices a4-a6 mediate dimerisation; single alanine mutations AFF4 F1014A or Y1096A impair dimer formation and impair transactivation, the corresponding AFF1 residues Phe1063 and Tyr1145 affect transactivation of the HIV-1 provirus, and these mutations do not affect the interaction of AFF1/4 with other SEC subunits.

    Structural and functional insight into the effect of AFF4 dimerization on activation of HIV-1 proviral transcription · Tang D, Chen C, Liao G, Liu J, Liao B, Huang Q, Chen K, Feng J, Zhu Z, Teng M, Niu L, Zhang J, Zhu C · Cell Discovery 2020
    PMID 32128251DOI
    What this evidence cannot show

    The functional readout is HIV-1 proviral transactivation in a reporter setting, not leukaemic transformation, and the AFF1 residues were tested in that same HIV context. Whether KMT2A::AFF1 needs AFF1 dimerisation is exactly what remains untested - but the reagent and its published phenotype are in hand.

    Evidence that argues against this (1)
    • Two problems, and the first is internal. If single alanine substitutions (AFF4 F1014A, Y1096A) IMPAIR dimer formation and transactivation, then this interface is demonstrably sensitive to single point substitutions - which is the opposite of the property the hypothesis claims makes the class resistance-proof, and it means the mutational landscape around such an interface is rich rather than empty. The second problem is structural resolution. For the comparable PML coiled coil, the assembly is not a single compact two-helix contact but a multivalent 'octopus-like' arrangement in which each coiled-coil helix of one dimer engages a coiled-coil helix of a NEIGHBOURING dimer in anti-parallel configuration, and the evidence for it is a 5.3 A cryo-EM map of the RBCC dimer supplemented by AlphaFold2 prediction and crosslinking mass spectrometry - not an atomic structure a medicinal chemist can design into. A distributed, many-contact interface resolved at 5.3 A is a poor candidate for the compact single-site occupancy this hypothesis assumes. Tan Y et al. 2024
Connected to