Oncofusion

start / graph / gene-daxx

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.

DAXX - the strongest of five selective dependencies among 254 unwritten interactors but it FAILS Bonferroni correction (p=0.0128, rank 4 of 254), and not one paper connects it to this disease

also called DAXX · DAP6 · hDaxx · EAP1 · Q9UER7 · death domain-associated protein 6

DAXX helps load a special version of a histone protein onto DNA and lives in the same nuclear bodies this project already studies. Cancer cells carrying this fusion need it more than other leukaemia cells do - and nobody has ever written a paper putting the two together.

The same thing, for a scientist

DAXX (Q9UER7, 740 aa) entered this graph from an instrument, not from reading. It has the **largest selective DepMap difference of the 254 unwritten curated interactors** - **-0.305**, index mean -0.718 against -0.414 in other leukaemia lines, pan-essentiality -0.374, and most required in the lymphoid half at **-0.89** [E1]. It appears against **four baits** (AFF4, DOT1L, MLLT10, MLLT3) [E2]. **And the literature intersection with this disease is empty.** `DAXX AND MLL` returns 0. `DAXX AND (MLLT3 OR AF9 OR MLLT1 OR ENL OR AFF4 OR DOT1L)` returns 0. `DAXX AND (MLL OR KMT2A) AND leukemia` returns 0. `DAXX AND KMT2A` returns 2 and both are irrelevant - an adrenocortical carcinoma case report and a pancreatic neuroendocrine fusion survey, co-mentioning the genes in a panel [E5]. Every query translation was recorded and none was rewritten, so these are bounded zeros rather than parser artefacts. **What DAXX does is squarely inside this graph's subject matter.** With ATRX it is the chaperone that deposits **histone variant H3.3** and maintains **H3K9me3** at heterochromatin [E3], and that deposition is organised by **PML** in megabase heterochromatic domains, where losing PML shifts the H3 methylation balance from K9me3 to K27me3 [E4]. This graph has a subnuclear beat, a PML entry, a histone-mark beat and a long argument about H3K79 - and no entry for the chaperone that puts the variant in place.

Go deeper
Why this entry exists, stated plainly

It is not because the evidence is strong. The database evidence is one platform: every IntAct row for DAXX against these four baits comes from BioPlex, affinity purification of tagged, overexpressed baits, which this project has already shown is two papers and one observation 2. The dependency evidence is a screen difference on eleven cell lines with no permutation null and a lymphoid mean resting on four lines 1.

It is because the two weak signals are of different kinds and neither knows about the other, and because the third element - the literature - is a counted zero rather than a weak positive. A protein that (a) co-purifies with four separate KMT2A fusion partners, (b) is the most selectively required unwritten interactor in the set, and (c) has never been studied in this disease is the shape of a gap, not the shape of a finding.

The honest prior is that it is an artefact. BioPlex recovers abundant nuclear proteins against many baits; DAXX is abundant and nuclear; and 254 genes ranked without correction will put something at the top. The entry is written so that a future cycle can kill it cheaply rather than so that it can be built on.

The one mechanistic reason it is worth more than its evidence

DAXX is the H3.3 chaperone. This graph's central mechanism is a fusion that drives transcriptional elongation through gene bodies, and H3.3 is deposited at actively transcribed gene bodies by HIRA and at heterochromatin by ATRX/DAXX 3. A fusion that forces elongation forces histone turnover, and histone turnover is what determines the valency of a mark with no eraser.

That connects to something this graph already argues. exp-h3k79-state-resolved-icechip-dose-series rests on the fact that H3K79 has a single writer and no known demethylase, so its clearance tracks nucleosome replacement. Whoever controls variant deposition controls that clock. The graph records this for turnover in the abstract and has no entry for either chaperone.

The subnuclear angle is the second one. PML organises the domains where ATRX/DAXX deposit H3.3, and losing PML flips those domains from H3K9me3 toward H3K27me3 4. This graph carries protein-pml-domains, a subnuclear beat and a condensate-evidence standard. A KMT2A fusion's relationship to PML bodies has not been examined here, and DAXX is the protein that would carry it.

None of that is evidence. It is the reason the gap is worth a query rather than a shrug.

Cell type, cell state, localisation, epigenetics
  • Cell type. Broadly expressed. Nothing in a KMT2A-rearranged cell, by the counted absence in 5; the DepMap numbers are the only disease-adjacent data and they are a screen, not an experiment.
  • Cell state. Not characterised in this context. The lymphoid-versus-myeloid split in the screen (-0.89 against -0.62) is the only state-adjacent signal and rests on four lymphoid lines.
  • Subcellular. Nucleus.
  • Subnuclear. PML nuclear bodies, and DAXX's H3.3 deposition is organised by PML in megabase PML-associated domains 4. This is the most specific localisation claim available and it is from a non-leukaemic system.
  • Epigenetics. With ATRX, deposits H3.3 and maintains H3K9me3 at heterochromatin genome-wide 3. In PML-associated domains, loss of PML shifts the balance to H3K27me3 while HIRA-dependent H3.3 loading is preserved 4.
Update: DAXX does not only maintain H3K9me3, it establishes it before deposition

This entry's chaperone evidence carried the older model - ATRX/DAXX maintains H3K9me3 at heterochromatin. The newer result is upstream of that: DAXX adds a de novo H3.3K9me3 deposition pathway, recruiting methyltransferases to mark new H3.3-H4 before it reaches chromatin (PMID 36868228, DAXX adds a de novo H3.3K9me3 deposition pathway to the histone chaperone network, Mol Cell 2023; fetched and read back).

That matters for the reason this entry was written at all. A chaperone that maintains a mark acts after deposition and is a downstream reader; a chaperone that establishes the mark on soluble histone before deposition is setting chromatin state at the point of supply. If a fusion that forces elongation changes histone supply, the second model is the one where that would matter.

This entry's open question 'write gene-hira before building anything on DAXX' is now closed - gene-hira exists, written by the histone-chaperone beat. The question it was protecting against stands: HIRA is the gene-body chaperone and DAXX is the heterochromatin one, and a fusion driving elongation through gene bodies has the more obvious relationship with the first.

Facts and figures
uniprotQ9UER7
gene symbolDAXX
length AA740
subnuclear localizationND10/PML nuclear bodies and pericentromeric heterochromatin, ALTERNATING and cell-cycle-gated: ND10 residence requires SUMO-1-modified PML, heterochromatin residence is confined to a window from the end of S phase to early G2 and requires ATRX phosphorylation (PMID 10525530, PMID 15252119). DAXX is the best-resolved subnuclear record in this tranche and one of the best in the graph.
shown for fusionNO, and the absence is counted. PubMed 2026-09-20: daxx[tiab] AND (KMT2A[tiab] OR MLL[tiab] OR MLL-AF4[tiab] OR MLL-AF9[tiab]) returns 2 records and BOTH were title-checked and are off-beat (PMID 42158873, an adrenocortical carcinoma case report; PMID 36690805, ACTH-secreting neuroendocrine neoplasms). Zero in titles and abstracts; supporting figures not searched. No DAXX localisation - or any DAXX result - has been published in a KMT2A-rearranged cell.
cell context evidence classmaintains_a_state, cleanly - the one entry in this tranche where the loss-of-identity readout is the primary measurement rather than an inference from growth. NOT required_in: the progenitors survive.
cell context shown for fusionNO. Measured 2026-09-21: DAXX[tiab] AND (KMT2A[tiab] OR MLL[tiab] OR "mixed-lineage leukemia"[tiab]) returns 2 records. Operand counts DAXX[tiab] = 1,039 and the KMT2A/MLL operand = 8,160, so the near-zero is a genuine empty intersection of two large sets rather than a non-composing query. Zero relevant in titles and abstracts; supporting figures not searched. No DAXX cell-type or cell-state experiment exists in any KMT2A-rearranged system.
cell context depmapWorth flagging because nobody has. DAXX is rank 87 of 17,728 - the 0.49th percentile - on this project's KMT2A-selectivity scan, at -0.718 in the 11 KMT2A-rearranged lines against -0.414 in 43 other leukaemia lines, difference -0.305, on a pan-line mean of -0.375. That places DAXX among the most KMT2A-selective genes this project has scanned, and it has no published KMT2A literature at all. The scan has not been lineage-split for this gene, so per the standing rule no selectivity CLAIM is made here - it is a flagged candidate for the lineage-split analysis.
localization
{
  "subcellular": "nucleus, in two alternating intranuclear domains rather than one. A cytoplasmic pool is separately documented in solid tumour and is functionally opposite to the nuclear pool.",
  "subnuclear": "ND10/PML NUCLEAR BODIES and HETEROCHROMATIN, alternating with the cell cycle. Daxx is recruited to ND10 by the C-terminal interaction with SUMO-1-modified PML, and in PML-null cells it is instead highly enriched in condensed chromatin - the two sites are in equilibrium and the equilibrium is set by SUMOylated PML. Daxx is actively recruited to heterochromatin at the END OF S PHASE together with the FACT subunit SSRP1, in a manner dependent on ATRX phosphorylation, and both are released early in G2 with Daxx returning to ND10. Its residence at heterochromatin therefore occupies a very short temporal window.",
  "how_shown": "cell lines lacking individual ND10 proteins, PML-null rescue by transient expression and cell fusion, SUMO-1 modification dependence (PMID 10525530); cell-cycle-resolved immunofluorescence with ATRX phosphorylation dependence (PMID 15252119)",
  "systems": "mammalian cell lines including PML-/- fibroblasts; no haematopoietic or leukaemic system in either study",
  "node": "localization-pml-nuclear-bodies-pmlrara"
}
cell type
{
  "maintains_lineage_output_in": "Haematopoietic progenitors, and it is a B-versus-myeloid decision. Loss of the H3.3 chaperone Daxx alters the chromatin landscape, H3.3 distribution and histone marks of haematopoietic progenitors, engaging a Pu.1-dependent transcriptional programme for MYELOPOIESIS AT THE EXPENSE OF B-CELL DIFFERENTIATION; this causes neutrophilia and inflammation and predisposes mice to an autoinflammatory skin disease. Animals lacking both Pu.1 and Daxx partially revert the phenotype but their progenitors carry unique chromatin and transcriptome alterations of their own.",
  "reading": "This is a maintains-a-state result in the strict sense used in this tranche: removing DAXX does not kill the progenitor, it changes what the progenitor becomes. In a graph about a B-lineage disease, a chromatin factor whose loss shifts progenitors out of the B lineage is worth more attention than its degree currently gets.",
  "how_shown": "conditional Daxx loss in murine haematopoiesis with ATAC/ChIP chromatin profiling, H3.3 distribution, transcriptomics, blood counts and a Pu.1 double-knockout epistasis arm",
  "systems": "mouse; the DAXX context also notes that DAXX is inactivated in myeloid leukaemia and other neoplasms",
  "nodes": [
    "cell-type-pro-b",
    "cell-type-gmp"
  ]
}
cell state
{
  "maintains_state": "Lineage output of the progenitor compartment - specifically the balance between the Pu.1-driven myeloid programme and B-cell differentiation. The mechanism proposed is defective silencing of retrotransposable elements through H3.3 loading.",
  "candidate_link_not_demonstrated": "Therapy-driven lineage switch in KMT2A-rearranged leukaemia sends cells to a GMP-like state, and DAXX loss sends normal progenitors along a Pu.1-driven myeloid route at the expense of B. The shapes rhyme. Nothing connects them experimentally, and the graph records this as a candidate parallel and not a finding.",
  "nodes": [
    "cell-state-gmp-like"
  ]
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.7184,
  "chronos_mean_other_leukaemia": -0.4136,
  "difference": -0.3048,
  "pan_essentiality_all_lines": -0.3745,
  "selectivity_rank": 87,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
Still open
  1. Is the DAXX dependency real? The cheapest test is a permutation null on the same 11-versus-43 comparison, which this triage did not compute, followed by the lineage-split checks this graph applies to every other dependency claim.
  2. Does KMT2A::AFF1 occupancy overlap PML-associated domains, and does the fusion perturb ATRX/DAXX H3.3 deposition there? Neither question has been asked and both are ChIP-scale.
  3. If H3.3 deposition matters to a fusion that forces elongation, HIRA - not DAXX - is the gene body chaperone. Write gene-hira before building anything on DAXX. **CLOSED 2026-09-20: `gene-hira` now exists.** The point it was protecting stands - HIRA is the gene-body chaperone, DAXX the heterochromatin one, and a fusion forcing elongation through gene bodies has the more obvious relationship with the first.
  4. Search the DAXX absence again under the older vocabulary ('mixed lineage leukemia'), under DAXX synonyms, and in supplements rather than titles. The current zero is bounded by title-and-abstract indexing of current symbols.
  5. DAXX is at the 0.49th percentile of this project's KMT2A-selectivity scan with a difference of -0.305 and has essentially no KMT2A literature - 2 records in titles and abstracts. Does that signal survive a lineage split of the 11-line index arm into its myeloid and B-lineage halves? Given how many apparent selectivities in this graph have dissolved at that step, this is the first thing to run.
  6. Daxx loss pushes normal murine progenitors out of the B lineage and into a Pu.1-driven myeloid programme. Therapy-driven lineage switch in KMT2A-rearranged B-ALL sends blasts to a GMP-like state. Is DAXX level or H3.3 loading altered in switched blasts compared with their paired diagnostic sample? Paired material exists in the lineage-switch literature.
Every source behind this entry (10)
  1. E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.analysis

    In a DepMap triage of 262 curated interactors with no entry in this graph, DAXX has the largest selective difference: -0.305 between 11 KMT2A-rearranged lines (mean -0.718) and 43 other leukaemia lines (-0.414), at a pan-essentiality of -0.374 across 1,208 lines, with an index myeloid mean of -0.62 and lymphoid mean of -0.89.

    What this evidence cannot show

    Run 2026-09-20. **Moderate deliberately.** No permutation null was computed - it is a triage, not a test - and a five-item list drawn from 254 genes without multiple-testing correction contains false positives by construction. The index arm is 11 lines and the lymphoid half is **four**. **DepMap contains no infant KMT2A::AFF1 line**, so this is a statement about adult and paediatric cell lines. The threshold that admitted DAXX is this project's convention, not a derived cut-off. **Null computed 2026-09-20 and this item does not survive it** (`ana-the-null-kills-my-own-shortlist`): 50,000 label permutations give p = 0.0128 (rank 4 of 254), and Bonferroni over the 254 genes the shortlist was drawn from leaves nothing below 0.05. DAXX is the strongest of the five and still fails correction. The literature absence recorded elsewhere in this entry is unaffected - it was never a dependency claim.

    Evidence that argues against this (1)
    • The ranking that produced this number is the same one that put histone H3C1 second on the evidence-ordered queue, a false positive by construction. An instrument that has a known false positive at the top of one ranking has not earned trust at the top of another. Huttlin EL et al. 2021
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.journal-article

    Every IntAct interaction row placing DAXX with AFF4, DOT1L, MLLT10 or MLLT3 derives from BioPlex, by anti-tag co-immunoprecipitation of tagged, overexpressed baits, resolving to a single research group after clustering publications by author overlap.

    PMID 33961781
    What this evidence cannot show

    Read from the IntAct PSICQUIC service and the graph's own lab-clustering on 2026-09-20. Established as a fact about the database's provenance. **Four baits recovering the same prey on one platform is one observation repeated, not four independent ones** - and BioPlex baits are overexpressed and tagged, a system in which abundant nuclear proteins are recovered promiscuously.

    Evidence that argues against this (1)
    • Dismissing a single-platform hit is how IGF2BP3 stayed invisible to this project for thirty cycles. The correct posture is that a one-platform hit is a reason to look and not a reason to conclude, which is what this entry is. Oncofusion knowledge graph (internal analysis) 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article

    Components of the ATRX/DAXX/histone H3.3 complex are required for heterochromatin silencing at multiple genomic regions, and the proposed model is that the ATRX/DAXX chaperone complex deposits H3.3 to maintain the H3K9me3 modification at heterochromatin throughout the genome.

    PMID 26773061DOI
    What this evidence cannot show

    Read from the abstract on 2026-09-20. **This is a review proposing a model from other groups' studies**, so it is graded `strong` as a generalisation rather than `established` as a result - the underlying experiments are in embryonic stem cells, telomeres and repeat regions, none of them leukaemia, and none of them KMT2A-rearranged. **Superseded in part, 2026-09-20**: the maintenance model here is joined by a de novo pathway in which DAXX recruits methyltransferases to mark new H3.3-H4 before deposition (PMID 36868228). The maintenance claim is not wrong; it is no longer the whole mechanism.

    Evidence that argues against this (1)
    • H3.3 has two deposition pathways and DAXX runs only one of them: HIRA deposits H3.3 at transcribed gene bodies, which is where a KMT2A fusion does its work. If the fusion's relationship to H3.3 matters, HIRA may be the relevant chaperone and DAXX the irrelevant one - and this graph has no entry for either. Delbarre E et al. 2017
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.journal-article

    PML organises megabase heterochromatic PML-associated domains and is required for H3.3 deposition by ATRX and DAXX in those domains; loss of PML shifts the histone H3 methylation balance in them from K9me3 to K27me3 while preserving the HIRA-dependent H3.3 loading function.

    PMID 28341773DOI
    What this evidence cannot show

    Read from the abstract on 2026-09-20 (PMC5453325 exists; full text not fetched for this entry). Established as a result within the reported system, which is not leukaemia. The relevance here is entirely by analogy to this graph's subnuclear beat and its PML entry, and analogy is not evidence about KMT2A::AFF1. **Superseded in part, 2026-09-20**: the maintenance model here is joined by a de novo pathway in which DAXX recruits methyltransferases to mark new H3.3-H4 before deposition (PMID 36868228). The maintenance claim is not wrong; it is no longer the whole mechanism.

    Evidence that argues against this (1)
    • This graph's PML material exists because PML::RARA is a comparator fusion, not because PML bodies have been implicated in KMT2A biology. Using a PML result to motivate a DAXX entry risks importing a mechanism from the comparator into the index disease, which is an error this graph has made before in the other direction - inheriting myeloid framing for a lymphoid disease. Voon HP et al. 2016
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    As of 2026-09-20, PubMed returns 0 records for 'DAXX AND MLL', 0 for 'DAXX AND (MLLT3 OR AF9 OR MLLT1 OR ENL OR AFF4 OR DOT1L)', 0 for 'DAXX AND MLLT3', 0 for 'DAXX AND AFF4' and 0 for 'DAXX AND (MLL OR KMT2A) AND leukemia'. 'DAXX AND KMT2A' returns 2 records, both irrelevant on title check: an adrenocortical carcinoma case report and a survey of gene fusions in ACTH-secreting neuroendocrine neoplasms.

    What this evidence cannot show

    Every query translation was captured and **none contained a [UID] term**, which is this project's documented signature of a parser rewrite producing a false zero. The two returned titles were fetched and read. **A bounded absence is bounded by its queries**: a paper studying DAXX in this disease under a different vocabulary - 'mixed lineage leukemia', a partner not in the list, or DAXX under a synonym such as DAP6 - would not be caught.

    Evidence that argues against this (1)
    • This project has logged that a symbol-driven absence cannot see a prey named only in a supplement, and that `AFF1` appears nowhere in the body text of a 2024 map that nonetheless recovers it. A zero over title-and-abstract indexing is weaker than it looks for exactly the kind of large-scale study that produced the BioPlex rows in [E2]. Huttlin EL et al. 2021
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    DAXX is UniProt Q9UER7 (DAXX_HUMAN), death domain-associated protein 6, 740 amino acids, also known as hDaxx, DAP6 and ETS1-associated protein 1.

    UniProtKB Q9UER7 (DAXX_HUMAN), Death domain-associated protein 6, 740 aa · UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
    What this evidence cannot show

    Fetched 2026-09-20 with the primary gene name read back, after rule 62. No residue coordinates are stored in this entry because it makes no structural argument, and this project's worst errors have come from coordinates that were never fetched.

    Evidence that argues against this (1)
    • The name is a fossil: DAXX was characterised as a Fas death-domain-associated protein and its chromatin role came later. An entry that presents it only as a histone chaperone is as incomplete as one that presents it only as an apoptotic adaptor, and the apoptotic literature is not assessed here at all. Voon HP et al. 2016
  7. E-daxx-denovoestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.journal-article

    DAXX provides a de novo H3.3K9me3 deposition pathway within the histone chaperone network, recruiting methyltransferase activity to newly synthesised H3.3-H4 before chromatin deposition, as part of an interactomics map of human H3-H4 chaperone complexes that also predicts an ASF1-SPT2 co-chaperone structure.

    PMID 36868228DOI
    What this evidence cannot show

    Fetched 2026-09-20; title, journal, year and DOI read back; **abstract only, full text not read**. Established as a result within the reported system, which is human cell-line interactomics and not leukaemia. **Nothing here involves KMT2A, a fusion, or a haematopoietic cell**, and this entry's own headline remains that zero papers connect DAXX to this disease.

    Evidence that argues against this (1)
    • This strengthens DAXX as a mechanistic entity and does nothing for its relevance here. The dependency evidence in this entry failed a permutation null the same week it was written, and a better model of what DAXX does does not make it a KMT2A-rearranged dependency. Oncofusion knowledge graph (internal analysis) 2026
  8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Daxx is an ND10/PML-nuclear-body protein whose recruitment to that compartment requires SUMO-1 modification of PML and is mediated by the Daxx C-terminus; in PML-null cells Daxx is not merely dispersed but is highly ENRICHED IN CONDENSED CHROMATIN, and increasing SUMO-1-modified PML pulls it back to ND10 - a variable equilibrium between two intranuclear sites.

    PML is critical for ND10 formation and recruits the PML-interacting protein daxx to this nuclear structure when modified by SUMO-1. · Ishov AM, Sotnikov AG, Negorev D, Vladimirova OV, Neff N, Kamitani T, Yeh ET, Strauss JF, Maul GG · J Cell Biol 1999
    PMID 10525530DOI
    What this evidence cannot show

    Mammalian cell lines, no haematopoietic system. The PML-null rescue is by transient expression and by cell fusion, which is a strong design for its era. This establishes the compartment and the mechanism of entry; it does not establish the fraction of endogenous DAXX in either site in any particular cell type.

    Evidence that argues against this (1)
    • DAXX is not obligatorily nuclear. In gastric cancer a cytoplasmic DAXX pool exists whose biological function is OPPOSITE to the nuclear pool - cytoplasmic DAXX inhibits proliferation and promotes apoptosis while nuclear DAXX does the reverse - the nuclear/cytoplasmic ratio tracks prognosis across 323 tumours, and SUMO-2/3 knockdown shifts the partition. A localisation record that stops at the nucleus gets the sign of DAXX's effect wrong half the time. Chen C et al. 2020
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Daxx occupies ND10 and heterochromatin as ALTERNATE sites under cell-cycle and phosphorylation control: ATRX is at heterochromatin throughout the cycle, Daxx is actively recruited there only at the end of S phase - simultaneously with the FACT subunit SSRP1 and dependent on ATRX phosphorylation - and both are released early in G2, with Daxx returning to ND10.

    PMID 15252119DOI
    What this evidence cannot show

    Immunofluorescence across a synchronised cell cycle, with the usual synchronisation caveat. Not haematopoietic. The recruitment of ATRX to ND10 is Daxx-dependent while Daxx's recruitment to heterochromatin is ATRX-dependent, so the two proteins cannot be cleanly separated by this design.

    Evidence that argues against this (1)
    • The alternation makes any single-timepoint DAXX localisation experiment uninterpretable. Because the heterochromatin window runs only from late S to early G2, an asynchronous population will show mostly ND10 and a reader will conclude DAXX is a PML-body protein. Conversely, in a PML-null or PML-disrupted cell the default is condensed chromatin, not dispersal. Any future DAXX imaging in KMT2A-rearranged cells must be cell-cycle resolved or it will produce a confidently wrong compartment. Ishov AM et al. 1999
  10. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In mice, loss of the H3.3 chaperone Daxx alters the chromatin landscape, H3.3 distribution and histone marks of haematopoietic progenitors, engaging a Pu.1-dependent transcriptional programme for myelopoiesis at the expense of B-cell differentiation; this causes neutrophilia and inflammation and predisposes the animals to an autoinflammatory skin disease, and the phenotype is partly but not wholly reverted in Pu.1/Daxx double-deficient animals, whose progenitors carry unique chromatin and transcriptome alterations.

    Aberrant chromatin landscape following loss of the H3.3 chaperone Daxx in haematopoietic precursors leads to Pu.1-mediated neutrophilia and inflammation. · Gerber JP, Russ J, Chandrasekar V, Offermann N, Lee HM, Spear S, Guzzi N, Maida S, Pattabiraman S, Zhang R, Kayvanjoo AH, Datta P, Kasturiarachchi J, Sposito T, Izotova N, Händler K, Adams PD, Marafioti T, Enver T, Wenzel J, Beyer M, Mass E, Bellodi C, Schultze JL, Capasso M, Nimmo R, Salomoni P · Nat Cell Biol 2021
    PMID 34876685DOI
    What this evidence cannot show

    Mouse, normal haematopoiesis, no leukaemia arm and no KMT2A-rearranged material. The proposed upstream mechanism - defective silencing of retrotransposable elements - is inferred from the chromatin and transcriptome data rather than demonstrated causally. The double-knockout arm only partially reverts, so Pu.1 is not the whole explanation. Read: abstract only.

    Evidence that argues against this (1)
Connected to