Oncofusion

start / graph / exp-hbm-reinsertion-kmt2a-aff1

Experimentmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q-subtraction-kmt2a

Experiment: put four residues back - the minimal add-back control the PHD3 result has never had

also called HBM reinsertion · DHNY add-back · minimal insertion control

Add just four amino acids back into the leukemic protein to restore one lost contact, and see whether that alone cripples it. If four residues are enough to break it, the famous experiment that added a whole finger was probably measuring the insertion, not the biology.

The same thing, for a scientist

Insert the HCF-binding motif (DHNY, KMT2A aa 1799-1802, or the minimal Yokoyama fragment) into KMT2A::AFF1 and KMT2A::MLLT1, with an H1800A point mutant as the length-matched negative control, and score HCF-1 recruitment plus serial replating and in-vivo transformation. The experiment does two jobs at once. Job 1: it asks whether a never-tested lost module matters. Job 2, and the more important one: it is the missing methodological control for the PHD3 add-back result, because a four-residue insertion cannot plausibly impair a chimeric protein sterically. If a four-residue insert also blocks replating, the 2008 PHD3 result is substantially a construct artefact; if it does nothing while the PHD3 insert does, the PHD3 result gains real credibility for the cost of a single extra arm.

Go deeper
Why a four-residue insert is the right control

The standing objection to Chen et al. 2008 is that inserting a folded ~60-residue PHD finger into the middle of a fusion protein could impair it through folding, spacing or degron exposure rather than by restoring a regulatory switch. That objection has never been addressed, and it is the main reason the graph should not treat 'PHD3 loss is necessary' as established [see ana-phd3-addback-replication-audit].

A four-residue insertion is about as close to a null perturbation as a protein insertion can be. And it comes with an unusually clean negative control: a single H-to-A substitution inside the motif abolishes HCF binding while keeping the insert's length, position and composition otherwise identical. No add-back experiment in this literature has had a control that tight.

The motif's position is confirmed two ways: stated in the Yokoyama 2004 full text with a Delta-1799/1803 deletion mutant that loses HCF-1/HCF-2 while retaining ASH2L and WDR5 1, and as the unique [DE]Hx Y match in the whole UniProt Q03164 sequence 2.

Arms and readouts

Arms (each in KMT2A::AFF1 and KMT2A::MLLT1):

  1. Fusion, unmodified.
  2. Fusion + DHNY inserted at a defined position N-terminal to the junction.
  3. Fusion + DHNY carrying H1800A (length-matched, binding-dead).
  4. Fusion + PHD3, replicating Chen et al. in the same hands and the same cells - the arm that makes the whole comparison interpretable.
  5. Fusion + a scrambled four-residue insert.

Readouts: HCF-1/HCF-2 co-IP (does the insert work at all?); serial replating; and, for arms that show an effect, transplantation with survival. Add fusion protein level and half-life, because an insertion that reduces protein abundance would explain a replating effect without any regulatory story.

The most informative single comparison is arm 2 versus arm 3. Everything else is context.

Cycle 31: the absence this experiment exploits is real, and the confound control is now the main readout

The absence holds. This experiment's entire value rests on nobody having tested what HBM loss costs a fusion. That was searched specifically in September 2026 and nothing was found 4. Good - the experiment is still worth running, and the negative search is recorded so a future reader can tell a real gap from an unasked question.

But the design's priorities should change. The arms list 'fusion protein level and half-life' as a control, on the reasoning that an insertion which reduces abundance would explain a replating effect without any regulatory story. That is now the most interesting measurement in the experiment, not a nuisance control, because the KMT2A half's better-evidenced losses are all degradation determinants 5. Three of them - PHD2, the taspase1 sites, FYRN - sit C-terminal to the break, and disabling each in wild-type MLL1 makes the protein more abundant and more chromatin-bound.

Two consequences:

  1. Add an abundance-only arm. A four-residue HBM insert is close to a null perturbation, which is what makes it the right control for PHD3. But the panel should also carry an insert designed to lower fusion abundance without any known binding function - the simplest being a short degron - so the experiment can calibrate how much replating suppression a given percentage drop in fusion protein buys. Without that calibration, arm 4 (PHD3) remains uninterpretable even if arms 2 and 3 behave perfectly.
  2. Consider adding FYRN. The most direct test of the abundance hypothesis is to put the FYRN degron back into the fusion in cis. It is ~57 residues, its function on an unpartnered N-terminal fragment is documented, and it predicts a destabilised fusion. If a FYRN insert suppresses replating in proportion to the protein it removes while the HBM insert does nothing, the whole beat's organising claim is supported in one panel. That arm is described in hyp-kmt2a-half-loses-abundance-control-not-activity-control.
Facts and figures
designIn-cis minimal add-back panel in KMT2A::AFF1 and KMT2A::MLLT1: unmodified fusion; +DHNY; +DHNY H1800A (length-matched binding-dead); +PHD3 (replication arm); +scrambled tetrapeptide ADDED cycle 31: (6) fusion + a short heterologous degron, as an abundance-only calibration arm; (7) fusion + FYRN (aa 2018-2074) in cis, as the direct test of the degradation-determinant hypothesis.
readoutHCF-1/HCF-2 co-immunoprecipitation; serial replating colony formation; transplantation survival for arms showing an effect; and - promoted from control to primary readout - fusion protein steady-state level and half-life by cycloheximide chase in every arm, since the KMT2A half's best-evidenced losses are degradation determinants
controlsH1800A point mutant is the key control - identical length and position, no HCF binding. Scrambled insert controls for sequence-independent insertion effects. The PHD3 arm run in the same hands makes the comparison interpretable.
cost tierlow
timeline4-6 months including transplantation
decision ruleIf +DHNY blocks replating as strongly as +PHD3, the 2008 result is largely a construct artefact and the graph should downgrade the PHD3 necessity claim. If +DHNY does nothing while +PHD3 blocks, the PHD3 result survives its strongest methodological objection. Revised cycle 31: the comparison is now three-way. If replating suppression across all arms tracks fusion protein abundance, the PHD3 result and any HBM result are both abundance effects and the regulatory readings of both should be withdrawn. Only an arm that suppresses replating WITHOUT lowering fusion abundance is evidence for a regulatory mechanism.
absence rebounded cycle 58
{
  "what_the_original_search_reached": "HCFC1[tiab] = 102 of a 495-record alias union; HCFC1 intersects \"HCF-1\" in 11 records",
  "alias_sets_measured_2026_09_22": {
    "HCFC1[tiab]": 102,
    "\"HCF-1\"[tiab]": 188,
    "\"HCF1\"[tiab]": 208,
    "\"host cell factor\"[tiab]": 282,
    "VCAF[tiab]": 19,
    "union of fetched id sets": 495
  },
  "the_token_decided_the_answer": {
    "HCFC1[tiab] AND MLL[tiab]": "0 - genuine zero, parsed translation, warninglist",
    "\"HCF-1\"[tiab] AND MLL[tiab]": 6,
    "(HCFC1 OR \"HCF-1\" OR \"HCF1\") AND (MLL OR KMT2A)": 11,
    "\"HCF-binding motif\"[tiab]": 10,
    "DHNY[tiab]": "0 - genuine zero; the motif sequence itself is not an indexed term",
    "positive control KMT2A[tiab] AND leukemia[tiab]": 1087
  },
  "all_11_read_by_title_not_counted": {
    "named_by_the_original_item": [
      "15199122 Yokoyama 2004 (the founding characterisation)",
      "27009953 Dheekollu 2016 (the EBV OriP study)"
    ],
    "genuinely_relevant_and_missed": [
      "17698583 Capotosti 2007 - MLL/trithorax and HCF proteolytic maturation, species selectivity",
      "17612494 Tyagi 2007 - HCF-1 recruits MLL/Set-1 H3K4 methyltransferases to E2F promoters",
      "19763085 Tyagi 2009 - E2F1's DHQY HCF-1-binding motif, mutations modulate phenotype both ways",
      "24244196 Zhao 2013 - NSL/MLL crosstalk, context only"
    ],
    "off_topic": [
      "29875417 SETBP1 hub",
      "27259240 colon cancer",
      "20855524 E2F-MLL/ANCCA",
      "20018852 MOF HAT",
      "17643366 Wilson commentary"
    ]
  },
  "verdict": "The narrow claim SURVIVES - none of the 11 is a fusion HBM deletion or add-back, so no experiment has asked what a KMT2A fusion loses by lacking the motif. What does not survive is the bounding, and the item's description of what the search returned. It held by luck of wording."
}
How this could bite us back
  • A negative result for DHNY would not vindicate PHD3 completely - it would only exclude the crudest steric explanation, since a 60-residue folded domain can do things a tetrapeptide cannot for reasons other than restoring a switch.
  • Both arms could fail for the mundane reason that an engineered insertion site behaves differently from a natural domain boundary, which would leave the original question exactly where it is.
Still open
  1. Where exactly should the motif be inserted? The natural position (1799) does not exist in the fusion, so the insert must go somewhere chosen, and that choice is itself a variable - which is an argument for testing two positions.
  2. Does HCF-1 recruitment to a fusion even make sense given that HCF-1 normally arrives in the context of a much larger complex the fusion cannot assemble? A negative co-IP would end the experiment early and cheaply, which is a feature.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    HCF-1 and HCF-2 specifically interact with a conserved binding motif at MLL amino acids 1799-1802 fitting the (D/E)HXY consensus, and a Delta-1799/1803 mutant loses HCF-1 and HCF-2 interaction while retaining ASH2L and WDR5 binding.

    Leukemia proto-oncoprotein MLL forms a SET1-like histone methyltransferase complex with menin to regulate Hox gene expression · Yokoyama A, Wang Z, Wysocka J, Sanyal M, Aufiero DJ, Kitabayashi I, Herr W, Cleary ML · Molecular and Cellular Biology 24(13):5639-5649 2004
    PMID 15199122DOI
    What this evidence cannot show

    FULL TEXT read via PMC480881 for the residue numbers and the deletion mutant. The paper establishes the motif's necessity for HCF binding in wild-type MLL; it says nothing about whether inserting the motif into a fusion would restore recruitment, which is the assumption this experiment tests.

    Evidence that argues against this (1)
    • A deletion mutant losing HCF binding while retaining ASH2L and WDR5 shows the motif is necessary for that contact and that the complex has not globally collapsed. It does not show nothing else at that surface changed, and a re-insertion experiment built on it assumes the motif is sufficient to restore the contact in a fusion context where the surrounding sequence is absent. Molecular and Cellular Biology 2004
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The tetrapeptide DHNY occurs at residues 1799-1802 of the UniProt Q03164 canonical sequence and is the only [DE]Hx Y match in the entire 3,969-residue protein, confirming the motif's position and uniqueness.

    What this evidence cannot show

    Accession verified by reading back the primary gene name (KMT2A) and length (3,969 aa). The scan is arithmetic on a public sequence; it confirms position and uniqueness, not function.

    Evidence that argues against this (1)
    • Sequence uniqueness in the canonical isoform is not functional uniqueness, and the search covered one isoform of one protein. The design needs the motif to be both necessary and sufficient in the fusion; uniqueness speaks to neither. Molecular and Cellular Biology 2004
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Inserting the third PHD finger alone into MLL-ENL blocks the fusion's haematopoietic stem cell immortalisation potential in serial plating colony assays while PHD1+PHD2 insertion does not - the result this experiment is designed to control.

    Loss of MLL PHD finger 3 is necessary for MLL-ENL-induced hematopoietic stem cell immortalization · Chen J, Santillan DA, Koonce M, Wei W, Luo R, Thirman MJ, Zeleznik-Le NJ, Diaz MO · Cancer Research 68(15):6199-6207 2008
    PMID 18676843DOI
    What this evidence cannot show

    ABSTRACT ONLY; single unreplicated in-vitro study in one fusion. Confidence set to moderate for that reason. Arm 4 of this experiment is a direct replication attempt of it.

    Evidence that argues against this (1)
    • The methodological objection this experiment was built to answer has a second, sharper form. The concern was steric: a folded 60-residue insert might impair a chimeric protein. The better-supported concern is now about abundance, because the KMT2A half's measured losses are degradation determinants and any insertion that lowers fusion protein level will lower replating on its own. A four-residue HBM insert controls for sterics but NOT for abundance, so the panel needs an abundance-only calibration arm before either the HBM or the PHD3 arm is interpretable. Wang J et al. 2012
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    A dedicated PubMed search in September 2026 for any functional test of HCF-binding-motif loss in a KMT2A fusion - HCFC1 with MLL and leukemia, HCF-1 with MLL transformation - returned only the founding Yokoyama 2004 characterisation and an unrelated Epstein-Barr virus study. No experiment has asked what the fusion loses by lacking the HBM. **RE-BOUNDED CYCLE 58 AND THE SEARCH BEHIND THIS SENTENCE WAS WRONG, THOUGH THE SENTENCE ITSELF SURVIVES.** The original queries were untagged AND-chains on the weakest available token. Measured on ID sets: HCFC1[tiab] = 102, "HCF-1"[tiab] = 188, "HCF1"[tiab] = 208, "host cell factor"[tiab] = 282, union of the fetched sets = 495, and **HCFC1 intersects "HCF-1" in only 11 records** - so the original operand reached roughly a fifth of the literature. The token decided the answer on the fusion side too: `HCFC1[tiab] AND MLL[tiab]` returns **0** (a genuine zero, parsed translation with warninglist) while `"HCF-1"[tiab] AND MLL[tiab]` returns **6**, and the alias union against (MLL OR KMT2A) returns **11**. All 11 were read by title. Two are the records this item names; **four are genuine MLL-HCF papers it never surfaced** (PMIDs 17698583, 17612494, 19763085, 24244196). None of the 11 is a fusion HBM deletion or add-back, so the claim in this sentence holds - but it held by luck of wording, not by bounding.

    What this evidence cannot show

    A negative search, bounded by query wording and by indexing; preprints not covered. Read as 'not findable by these queries', not as proof of non-existence. Recorded because the value of this experiment depends entirely on the absence being real. SUPERSEDED IN PART, cycle 58: this caveat correctly said 'read as not findable by these queries', and the re-bounding showed those queries were reaching about a fifth of the literature. The absence is now bounded on the alias UNION with every zero checked for a parsed querytranslation, and all 11 records in the fusion intersection were read rather than counted. Still abstract-and-title level: a motif experiment buried in a supplementary figure would not be seen, and preprints remain uncovered.

    Evidence that argues against this (2)
    • A counted absence is only as good as its queries, and this project has been wrong about exactly this class repeatedly - HCF-1 publishes under HCFC1, HCF1, VP16-accessory protein and C1 factor, and an alias set that misses the older names under-counts the literature where a functional test would most likely sit. The absence may be real; the search as described does not establish it. Molecular and Cellular Biology 2004
    • The re-bounding weakens this item's own rationale rather than the absence. This experiment is partly justified on the grounds that a four-residue insertion is a novel and minimally perturbing manipulation whose behaviour is unknown, which is why it can serve as the methodological control for the PHD3 add-back. In the same motif class the manipulation is not unknown: Tyagi 2009 mutated E2F1's DHQY HCF-1-binding sequence and found it modulates phenotype in BOTH directions. That makes the experiment more feasible and less novel at the same time, and it means a null result would be harder to interpret as 'the motif does not matter'.
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Three independent degradation determinants of MLL1 - the PHD2 CDC34-dependent E3 ligase, the taspase1 cleavage sites and the FYRN domain that marks a free MLL N-terminal fragment for destruction - all lie C-terminal to the breakpoint cluster region and are absent from every der(11) fusion; disabling each of them in wild-type MLL1 makes the protein more abundant and more chromatin-bound.

    A subset of mixed lineage leukemia proteins has plant homeodomain (PHD)-mediated E3 ligase activity · Wang J, Muntean AG, Wu L, Hess JL · Journal of Biological Chemistry 287(52):43410-43416 2012
    PMID 23129768DOI
    What this evidence cannot show

    ABSTRACT ONLY for the PHD2 component (PMID 23129768); the taspase1 and FYRN components are PMIDs 30573454 (full text read) and 21670200 (abstract only), cited in full in `ana-kmt2a-lost-module-ledger` [E6, E8]. Cited here for one design consequence: this experiment already plans to measure fusion protein level and half-life as a confound control, and that measurement is now the most informative thing in the whole design rather than an afterthought.

    Evidence that argues against this (1)
    • Calling the three degradation determinants INDEPENDENT overstates what is shown. PHD2-CDC34 activity, taspase1 cleavage and the FYRN mark were characterised in different systems and no experiment orders them or shows they act separately; a protein with three nominated degrons may have one dominant route and two that are conditional. Journal of Biological Chemistry 2012
  6. E_k58_hcf_motif_precedentmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The four-residue HCF-binding-motif manipulation this experiment proposes has validated precedent in another protein, with a nearly identical motif. E2F1 associates with HCF-1 through a short **DHQY** sequence - the same D-H-x-Y class as KMT2A's **DHNY** at aa 1799-1802, differing at one position - and sequence changes in that E2F1 HCF-1-binding site 'can modulate both up and down the ability of E2F1 to induce apoptosis' (PMID 19763085). The same group had shown that during the G1-to-S transition HCF-1 recruits MLL and Set-1 H3K4 methyltransferases to E2F-responsive promoters and induces histone methylation and transcriptional activation (PMID 17612494) - so HCF-1 is an established recruiter of the MLL family, not only an MLL binding partner. Separately, human and mouse HCF-1 maturation is **taspase-1 independent** while MLL is cleaved by taspase 1, and the two pathways are species-selective (PMID 17698583).

    PMID 19763085DOI
    What this evidence cannot show

    All three papers were fetched and read at ABSTRACT level only; no figure or methods section was inspected, so the exact E2F1 mutations, their magnitudes and their controls are not established here. The central inference - that a DHQY result transfers to a DHNY motif in a different protein - is an analogy across proteins, which is why this is moderate and not strong: the motifs differ at one residue, they sit in unrelated sequence contexts, and E2F1's readout (apoptosis) is not this experiment's readout (HCF-1 recruitment, serial replating, transformation). Nothing here is in a leukaemia cell or a KMT2A fusion; PMID 19763085 is about E2F1 biology and mentions leukaemia only in the expansion of the MLL acronym.

    Evidence that argues against this (2)
    • The precedent may transfer in the unhelpful direction. In E2F1 the motif mutations modulate phenotype BOTH up and down, so if a DHNY insertion into KMT2A::AFF1 changes replating, the result would not cleanly distinguish 'the fusion regained a lost function' from 'a four-residue insertion perturbed the construct' - which is exactly the ambiguity this experiment was designed to resolve for the PHD3 result. The length-matched H1800A point mutant control becomes essential rather than optional.
    • Capotosti 2007 cuts against reasoning from model organisms here. It shows the MLL/trithorax and HCF maturation pathways are species-selective, with human taspase 1 preferring MLL and Drosophila taspase 1 preferring Trithorax, and human HCF-1 maturation being taspase-1 independent entirely. So any HBM result obtained in a fly or mouse system may not transfer, and this graph's habit of supporting KMT2A claims with murine MLL-AF9 data is specifically exposed on this axis.
Connected to