Oncofusion

start / graph / ana-phd3-addback-replication-audit

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q-subtraction-kmt2a

Audit, CORRECTED: the PHD add-back result was replicated in 2008 by a second group in a different fusion, and this graph already cited it

also called PHD3 add-back audit · PHD3 reinsertion replication search

The graph's most exciting claim about what KMT2A fusions throw away was recorded as resting on a single 2008 experiment that nobody repeated. That is wrong. A second group did the same thing the same year in a different fusion and got the same answer — and this graph was already citing their paper in four other entries.

The same thing, for a scientist

The original audit concluded that PHD3 add-back rested on Chen 2008 (PMID 18676843, MLL-ENL, Cancer Res) alone. Muntean 2008 (PMID 18796627, MLL-AF9, Blood, Hess laboratory) is an independent test of the same design by a different group: inclusion of the PHD fingers in MLL-AF9 blocked immortalisation of haematopoietic progenitors, and two or more fingers reduced association with the Hoxa9 locus and suppressed Hoxa9 up-regulation. The general claim — PHD inclusion blocks fusion-mediated transformation — therefore has two concordant independent reports in two different partners. What remains single-sourced is the GRANULAR claim: that PHD3 specifically, via Cyp33, is the operative finger, with PHD1+2 alone having no effect. That is Chen 2008 only.

Go deeper
Correction (2026-09-19): the replication exists, and the graph already held it

This entry was written on a negative literature search and the negative was an artefact. The searches recorded in the section below returned zero because the PubMed MCP tool ANDs every keyword, so a natural-language query fails on topics with substantial literature. The original author flagged exactly this risk in the same section — "PubMed's query parser in this tool ANDs every term, so a differently worded replication could be missed" — and was right.

Muntean 2008 (PMID 18796627, Blood 112:4690-3, Hess laboratory): "inclusion of the PHD fingers in the MLL fusion protein MLL-AF9 blocked immortalization of hematopoietic progenitors. Inclusion of 2 or more PHD fingers reduced association with the Hoxa9 locus and suppressed Hoxa9 up-regulation."

Against Chen 2008 (PMID 18676843, Cancer Res, MLL-ENL, Loyola/Chicago) this is a different group, a different fusion partner and a different readout of the same manipulation, published within weeks. That is replication in the sense that matters.

What this changes. The add-back finding is considerably more robust than the audit said, and the graph's derived claims about the discarded PHD cassette rest on firmer ground. What survives of the audit is narrower and still worth keeping: the specificity claim — PHD3 in particular, acting through Cyp33, with PHD1+2 alone inert — is Chen 2008 alone. Muntean tested the cassette and "two or more" fingers, not PHD3 in isolation, so the two papers agree on the phenomenon and only one of them localises it.

How it was found. Not by re-reading the literature, but by scripts/check-absence-vs-library.py, which asks whether an entry claims nobody has studied something the graph already cites. Muntean 2008 was cited by four other entries at the time this audit said no replication existed. It was the checker's first real catch.

What the one study actually did1 deeper

Chen et al. inserted (a) the whole three-PHD cassette and (b) the third PHD finger alone into MLL-ENL, expressed the constructs in murine haematopoietic cells, and scored serial replating colony formation. Both insertions blocked immortalisation; PHD1+PHD2 did not. In parallel they showed the PHD3-containing fusion recruited CyP33 and HDAC1 and downregulated HOXC8 in a CyP33-dependent way 1.

Four limitations are in the design, not in anyone's reading of it:

  1. Serial replating is not leukemia. It is a surrogate for self-renewal in a dish. No transplantation, no latency, no in-vivo endpoint was reported.
  2. One fusion. MLL-ENL (KMT2A::MLLT1). Nothing in KMT2A::AFF1, the commonest infant fusion and the one this project cares most about.
  3. A gain-of-function insertion. The construct is a protein that exists in no patient. Adding a folded ~60-residue domain into the middle of a chimeric protein can impair it for reasons that have nothing to do with the switch - folding, spacing, degron exposure. The CyP33/HDAC1 recruitment data make the steric explanation less likely but do not exclude it.
  4. The title asserts necessity. 'Loss of MLL PHD finger 3 is necessary for MLL-ENL-induced immortalization' is a stronger statement than a replating assay can support. Showing that re-adding X breaks the system does not show that removing X from wild-type would build it.
Why the fourth point is the one that matters

The King's brief flags this: the graph has previously confused 'retained' with 'required'. The mirror error here is reading 'lost' as 'load-bearing'. Chen et al. establish sufficiency of restoration to block, which is a real and interesting result. The graph elsewhere describes it as showing that 'its removal is required', which is the necessity claim, and that is an over-read of this experiment.

The fully general version of the confusion: a module can be (i) lost and irrelevant, (ii) lost and incidentally advantageous to the oncogene, (iii) lost and necessarily absent for transformation. Only (iii) is a mechanistic explanation of the breakpoint's position, and only (iii) would predict that no patient fusion ever retains PHD3. The replating result is equally consistent with (ii).

The replication search

Searches run against PubMed on 2026-09-18 (this Queen, via the PubMed MCP tools):

  • PHD3 reinsertion MLL fusion protein abrogates transformation - 0 results
  • MLL PHD finger insertion fusion protein leukemia transformation (with and without year terms) - 0 results
  • MLL1 PHD finger 1 CYP33 intramolecular interaction transformation - 0 results
  • CYP33 overexpression MLL leukemia cells apoptosis HOXA9 repression - 0 results

The PubMed related-articles set for PMID 18676843 was also inspected; it contains the expected mechanistic neighbours (PMIDs 20677832, 20541251, 12829790, 20460131, 33606679, 16981007) but no study that repeats the insertion-and-transformation design 2.

This is a negative literature result, and negative literature results are weak. PubMed's query parser in this tool ANDs every term, so a differently worded replication could be missed; a replication buried in a figure panel of a paper about something else would certainly be missed. The claim here is narrow: no replication surfaced under four independent phrasings plus a related-articles sweep. It is not proof that none exists.

What would settle it, cheaply

Two experiments, in increasing order of cost:

  1. exp-hbm-reinsertion-kmt2a-aff1 - the four-residue HCF-binding motif add-back. This is the control Chen et al. never had: a minimal insertion at a nearby position that restores a different lost contact. If a four-residue insert also blocks replating, the 2008 result is at least partly a construct artefact. If it does nothing, the PHD3 result gains a great deal of credibility for almost no money.
  2. exp-phd3-addback-kmt2a-aff1-in-vivo - repeat the original design in KMT2A::AFF1 with a transplantation endpoint.

Note the asymmetry: experiment 1 is the cheaper and more informative of the two, and it has never been proposed, because the field has treated the 2008 result as settled rather than as unreplicated.

Facts and figures
statusrun
result summaryNo replication found. The add-back claim rests on one 2008 in-vitro study in one fusion. Claim strength should be capped at 'restoration is sufficient to block replating in MLL-ENL', not 'PHD3 loss is required for transformation'.
inputs
[
  "PMID 18676843",
  "PubMed E-utilities esearch",
  "PubMed ELink pubmed_pubmed"
]
queries run
[
  "PHD3 reinsertion MLL fusion protein abrogates transformation",
  "MLL PHD finger insertion fusion protein leukemia transformation 2015",
  "MLL1 PHD finger 1 CYP33 intramolecular interaction transformation",
  "CYP33 overexpression MLL leukemia cells apoptosis HOXA9 repression"
]
recommended confidence ceiling
{
  "experiment_was_performed_as_described": "strong",
  "phd3_loss_is_required_for_transformation": "moderate"
}
correction 2026 09 19
{
  "withdrawn": "that the PHD add-back result rests on one paper and has never been replicated",
  "replication": {
    "pmid": "18796627",
    "fusion": "MLL-AF9",
    "group": "Hess (Michigan)",
    "title": "The PHD fingers of MLL block MLL fusion protein-mediated transformation.",
    "authors": "Muntean AG, Giannola D, Udager AM, et al.",
    "venue": "Blood",
    "year": 2008,
    "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."
  },
  "original": {
    "pmid": "18676843",
    "fusion": "MLL-ENL",
    "group": "Zeleznik-Le / Diaz (Loyola, Chicago)",
    "title": "Loss of MLL PHD finger 3 is necessary for MLL-ENL-induced hematopoietic stem cell immortalization.",
    "authors": "Chen J, Santillan DA, Koonce M, et al.",
    "venue": "Cancer Res",
    "year": 2008,
    "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."
  },
  "still_single_sourced": "that PHD3 specifically, via Cyp33, is the operative finger (Chen 2008 only)",
  "root_cause": "the PubMed MCP search tool ANDs every keyword, so the replication search returned 0",
  "found_by": "scripts/check-absence-vs-library.py - Muntean 2008 was already cited by 4 other entries"
}
How this could bite us back
  • If the PHD3 story is weaker than the graph has assumed, then the 'restore the brake by induced proximity' modality that the graph calls its most interesting unexplored idea is resting on one unreplicated 2008 replating assay - a poor foundation for programme-level investment.
  • Conversely, over-correcting and discarding the result would be an error too: nobody has contradicted it either, and 'unreplicated' is not 'refuted'.
Still open
  1. Does a minimal four-residue insertion at a nearby position also impair the fusion in a replating assay? If yes, the 2008 result is partly a construct artefact and the graph's headline claim collapses.
  2. Has anyone catalogued whether any patient KMT2A fusion retains PHD3? The necessity claim predicts none should. The KMT2A recombinome dataset could answer this from existing data at zero experimental cost.
  3. Is the field's silence since 2008 evidence of consensus or of a failed replication that was never published? A direct enquiry to the originating laboratory (collab-ernst-colorado and the Loyola/Zeleznik-Le group) would settle it faster than any experiment.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Inserting the MLL three-PHD-finger cassette, or the third PHD finger alone, into MLL-ENL blocks the fusion's haematopoietic stem cell immortalisation potential in serial plating colony assays, restores CyP33 and HDAC1 recruitment and mediates CyP33-dependent HOXC8 downregulation, while inserting only PHD1 and PHD2 has no such effect.

    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 - PMID fetched from PubMed and the abstract read verbatim; figures, methods and statistics were not inspected, so colony counts, replicate number and effect size are unknown to this entry. In-vitro murine serial replating; one fusion (MLL-ENL); engineered insertion constructs. The confidence 'strong' applies to the proposition that this experiment was performed and reported as described, not to the necessity claim in the title.

    Evidence that argues against this (1)
    • A KMT2A lesion that RETAINS the entire PHD3 cassette - the partial tandem duplication of exons 5-11 with no partner gene - still deregulates Hoxa7/a9/a10 in a knock-in mouse via increased H3/H4 acetylation and H3K4 methylation at the Hoxa7 and Hoxa9 promoters. So Hox deregulation does not require losing PHD3, which is the specific thing the necessity claim would predict. Dorrance AM et al. 2006
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    A PubMed search on 2026-09-18 using four independent phrasings of 'PHD finger re-insertion into an MLL fusion and transformation', plus inspection of the PubMed related-articles set for PMID 18676843, surfaced no study replicating the PHD3 add-back result, no in-vivo version of it, and no version in a fusion other than MLL-ENL.

    What this evidence cannot show

    A negative literature search, which is the weakest class of evidence in this graph. The search tool ANDs all supplied terms, so recall is poor for differently worded work; non-English, preprint, conference-abstract and buried-figure replications would all be missed. The correct reading is 'not found under these queries', never 'does not exist'. Recorded so that a future agent can re-run the same queries rather than re-deriving them.

    Evidence that argues against this (1)
    • This is a counted absence and its four phrasings are all modern. PHD-finger work on MLL predates the KMT2A symbol and much of it indexes under MLL, ALL-1, HRX and HTRX1, while 're-insertion' is described in the older literature as add-back, restoration or reconstitution. An audit concluding the replication does not exist should record the alias and synonym set it used, because this project has repeatedly found that the bounded, well-formed absence claims are the ones that turn out wrong. Journal of Clinical Investigation 2006
  3. E_muntean_replicationstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Inclusion of the PHD fingers in MLL-AF9 blocked immortalisation of haematopoietic progenitors; inclusion of two or more PHD fingers reduced association with the Hoxa9 locus and suppressed Hoxa9 up-regulation. An independent group, a different fusion partner and a different readout from Chen 2008, published the same year - so the general PHD-add-back result is replicated, and this entry's premise that it was not is withdrawn.

    The PHD fingers of MLL block MLL fusion protein-mediated transformation · Muntean AG, Giannola D, Udager AM, Hess JL · Blood 112(12):4690-3 2008
    PMID 18796627DOI
    What this evidence cannot show

    Read as abstract only, re-fetched from eutils on 2026-09-19. Muntean tested the PHD cassette and 'two or more' fingers; it does NOT isolate PHD3, so it does not replicate Chen's specificity claim that PHD3 is the operative finger and PHD1+2 inert. Both studies are retroviral overexpression in mouse progenitors with serial-plating or immortalisation readouts, so they share a design limitation: neither tests add-back in an established human KMT2A::AFF1 leukaemia, and neither uses the AFF1 partner this project cares about.

    Evidence that argues against this (1)
    • Two papers from 2008 using the same overexpression paradigm are less independent than the word 'replication' suggests. They share an era, a technique and a set of assumptions, and both measure immortalisation in mouse progenitors rather than maintenance of established disease. The audit's deeper point - that a load-bearing claim in this graph rests on a narrow and old evidence base - survives its specific error about the count. Chen J et al. 2008
Connected to