DAXX and ATRX are absent from the only MLL-fusion AP-MS interactome - and PML is in it against all seven baits, alone, without a single one of its other body components
also called the PML-without-its-clients result · Skucha table 5 re-read for the H3.3 axis
There is exactly one experiment that asked what sticks to seven different versions of this fusion protein. The two chaperones this beat is about are not in its results at all - but PML, the scaffold of the nuclear compartment they normally live in, is there against every single bait, and none of PML's usual partners came with it.
The same thing, for a scientist
`gene-pml` carries an open question: *"Are other PML-body clients (DAXX, SP100, SUMO1/2, RNF4) present in the Skucha 960-protein network? A body-level signature would be far stronger than one scaffold protein."* The supplementary table is already cached in this repository (`analyses/skucha-mll-fusion-interactome/cache/41467_2018_4329_MOESM4_ESM.xlsx`, 960 rows) and had never been searched for these symbols. **The answer is no, and it is a clean no.** | protein | in the 960? | baits | MLL-AF4 p | strongest bait | |---|---|---:|---|---| | **PML** | **yes** | **7 of 7** | 0.0282 (SC 4\|4) | MLL-AF9, p 6.61e-06 (SC 22\|22) | | **TRIM28** | **yes** | **7 of 7** | 0.129 (SC 3\|2) | MLL-ENL, p 0.00347 | | USP7 | yes | 3 | - | MLL-AF9, p 0.00152 | | SUMO2 | yes | 1 | 0.147 (SC 2\|8) | MLL-AF4 only | | ASF1A | yes | 2 | 0.138 | MLL-EEN, p 0.0676 | | NASP | yes | 1 | - | MLL-AF1p, p 0.00827 | | SUPT16H | yes | 1 | - | MLL-EEN, p 0.0268 | | **DAXX** | **NO** | - | - | - | | **ATRX** | **NO** | - | - | - | | SP100 | NO | - | - | - | | RNF4 | NO | - | - | - | | SETDB1, HIRA, UBN1, CABIN1, H3-3A, H3-3B, CHAF1A, CHAF1B, MORC3, ATF7IP, ASF1B | NO | - | - | - | **Two findings, pointing in opposite directions.** 1. **Against this beat.** The DAXX/ATRX chaperone is not recovered with any of seven MLL fusions, at any rank, in the one dedicated experiment that exists. The BioPlex rows that put DAXX with AFF4, DOT1L, MLLT10 and MLLT3 are anti-tag co-IP of tagged, overexpressed baits from a single laboratory [E3]; here, with the actual fusion proteins as baits in a leukaemia cell, DAXX does not appear [E1]. 2. **For the subnuclear question, and harder to explain away.** PML is recovered against **all seven** fusions, and it is the *only* PML-nuclear-body component in the network. SP100 and RNF4 are absent; SUMO2 appears once, weakly, and only with MLL-AF4. **So this is a PML signal without a PML-body signature** - which is what `gene-pml`'s open question asked for and is the opposite of what it hoped to find [E2].
What 'absent' can and cannot mean in this table
This is a rank-thresholded table, not a detection list. This repository's own audit of it records near-identical per-bait counts (291-293 preys for every one of the seven baits) with maximum p-values ranging from 0.0164 to 0.175, which is the signature of a fixed number of top-ranked preys being taken per bait rather than a p-value cut-off.
So DAXX: absent means not in the top ~293 preys of any of seven baits, not not detected. The distinction is the difference between two absence classes, and it matters because one of them is cheap to close:
- the claim "DAXX is not a high-ranking MLL-fusion interactor in the only dedicated experiment" is established;
- the claim "DAXX was not detected with an MLL fusion" is
measured-not-reported, and the raw data are public at ProteomeXchange PXD009338. Re-searching those files for DAXX and ATRX peptides is an afternoon of compute with no wet work, and it would settle it either way.
The entry states the first and names the second as open.
Why PML-without-its-clients is the interesting half
PML sits at 7 of 7 baits with spectral counts of 22 and 22 against MLL-AF9 at p 6.61e-06 - that is not a marginal row. If the fusion were simply resident in PML nuclear bodies, the expected co-purification is the body: PML plus SP100 plus DAXX plus SUMO plus RNF4, because those proteins are recovered together routinely. Instead PML arrives with TRIM28 and nothing else from that compartment.
Two readings, both testable:
- A direct or near-direct PML contact, not a compartment. PML is a large SUMO-rich protein and MLL fusions are SUMO-modified; a SUMO-SIM contact would recover PML without recovering the clients that need an intact body.
- A SUMO-abundance artefact, which
gene-pmlalready names as an open question ("Is the co-purification SUMO-dependent?"). The single weak SUMO2 row (MLL-AF4, p 0.147) is the only evidence either way in the table and it is too weak to decide.
TRIM28 at 7 of 7 is the piece this graph has not reasoned about. TRIM28/KAP1 is a SUMO E3 ligase and an H3K9me3 co-repressor - a heterochromatin protein arriving with every MLL fusion, which is the same territory the DAXX/ATRX arm of this beat is about, reached by a different protein.
What was searched, so the absence is bounded by a name list
All 960 gene_name values were read and matched exactly against: DAXX, ATRX, PML, SP100, SUMO1, SUMO2, SUMO3, RNF4, H3F3A, H3F3B, H3-3A, H3-3B, HIRA, UBN1, CABIN1, SETDB1, TRIM28, ASF1A, ASF1B, NASP, CHAF1A, CHAF1B, SSRP1, SUPT16H, ATF7IP, MORC3, TSPYL5, USP7.
The histone-variant symbols were searched under both the current and the retired spellings (H3-3A/H3F3A, H3-3B/H3F3B) because the table is from 2018 and HGNC renamed these genes afterwards; both spellings are absent. This is the failure mode rule 69 warns about and it would have produced a false absence if only the current symbol had been tried.
Not searched: peptide-level evidence, the 769-row shRNA library in the sibling supplementary file, and the raw PRIDE deposit.
| status | run |
|---|---|
| absence class | measured-not-reported |
| what closes it | Re-searching the raw deposit ProteomeXchange PXD009338 for DAXX and ATRX peptides. The table is rank-thresholded at ~293 preys per bait, so peptide-level evidence below that rank is in the deposit and has never been looked at. No wet work. |
| result summary | DAXX and ATRX are absent from all 960 rows of the only MLL-fusion AP-MS interactome, as are SP100, RNF4, SETDB1, HIRA, UBN1, CABIN1 and both spellings of the H3.3 genes. PML is present against 7 of 7 baits and TRIM28 against 7 of 7; USP7 against 3; ASF1A 2; SUMO2, NASP and SUPT16H 1 each. |
| source pmid | 29777171 |
| source pmcid | PMC5959866 |
| source doi | 10.1038/s41467-018-04329-y |
| proteomexchange | PXD009338 |
| network size | 960 |
| n baits | 7 |
| relationship to existing entries | `analyses/skucha-mll-fusion-interactome` parsed the same table for graph-coverage statistics and never searched it for these symbols; `gene-pml` asks this exact question in its open_questions and predicted that a body-level signature would be the strong result. It is not there. It also answers gene-trim28's open question 'Is SETDB1 in the Skucha 960-protein network? A TRIM28-SETDB1 pair would be a corepressor module rather than a lone scaffold' - SETDB1 is NOT in the network, so TRIM28 is there as a lone scaffold, exactly as that entry feared. |
inputs
[ "analyses/skucha-mll-fusion-interactome/cache/41467_2018_4329_MOESM4_ESM.xlsx (Supplementary Table 5, 960 rows, cached in this repository)" ]
baits
[ "MLL-AF1p", "MLL-AF4", "MLL-AF9", "MLL-CBP", "MLL-EEN", "MLL-ENL", "MLL-GAS7" ]
symbols searched
[ "DAXX", "ATRX", "PML", "SP100", "SUMO1", "SUMO2", "SUMO3", "RNF4", "H3F3A", "H3F3B", "H3-3A", "H3-3B", "HIRA", "UBN1", "CABIN1", "SETDB1", "TRIM28", "ASF1A", "ASF1B", "NASP", "CHAF1A", "CHAF1B", "SSRP1", "SUPT16H", "ATF7IP", "MORC3", "TSPYL5", "USP7" ]
hits — the numbers behind the chart
{
"PML": {
"uniprot": "P29590",
"n_baits": 7,
"mll_af4_p": 0.0282,
"mll_af4_spectral_counts": "4|4",
"strongest": "MLL-AF9 p 6.61e-06, spectral counts 22|22",
"neg_ctrl": "0|0"
},
"TRIM28": {
"uniprot": "Q13263",
"n_baits": 7,
"mll_af4_p": 0.129,
"mll_af4_spectral_counts": "3|2",
"strongest": "MLL-ENL p 0.00347",
"neg_ctrl": "0|0"
},
"USP7": {
"uniprot": "Q93009",
"n_baits": 3,
"strongest": "MLL-AF9 p 0.00152"
},
"ASF1A": {
"uniprot": "Q9Y294",
"n_baits": 2,
"mll_af4_p": 0.138
},
"SUMO2": {
"uniprot": "P61956",
"n_baits": 1,
"mll_af4_p": 0.147,
"mll_af4_spectral_counts": "2|8",
"neg_ctrl": "0|2"
},
"NASP": {
"uniprot": "P49321",
"n_baits": 1,
"strongest": "MLL-AF1p p 0.00827"
},
"SUPT16H": {
"uniprot": "Q9Y5B9",
"n_baits": 1,
"strongest": "MLL-EEN p 0.0268"
}
}absent
[ "DAXX", "ATRX", "SP100", "RNF4", "SETDB1", "HIRA", "UBN1", "CABIN1", "ASF1B", "CHAF1A", "CHAF1B", "MORC3", "ATF7IP", "TSPYL5", "SUMO1", "SUMO3", "H3F3A", "H3F3B", "H3-3A", "H3-3B", "SSRP1" ]
measurements
[
{
"quantity": "n_of_baits_recovering_the_prey",
"subject": "PML | Skucha 7-bait MLL-fusion AP-MS",
"value": 7,
"units": "baits",
"n": 7,
"universe": "960-protein network, ~293 top-ranked preys per bait",
"null": null,
"note": "42 proteins in the network reach 7 of 7; PML is one of them and is the only PML-body component present"
},
{
"quantity": "n_of_baits_recovering_the_prey",
"subject": "DAXX | Skucha 7-bait MLL-fusion AP-MS",
"value": 0,
"units": "baits",
"n": 7,
"universe": "960-protein network, ~293 top-ranked preys per bait",
"null": null,
"note": "ATRX, SP100, RNF4, SETDB1, HIRA, UBN1 and both H3.3 gene spellings are also 0"
}
]closes open questions in
[ "gene-pml", "gene-trim28" ]
- This is the entry that most argues AGAINST a DAXX-directed intervention in this disease: the one dedicated fusion-interactome experiment does not recover it, and a dependency without a contact has no mechanism to design against.
- If the PML row is real and the fusion does sit in or near PML bodies, arsenic trioxide - which degrades PML and dissolves the bodies - is an approved drug whose effect on a KMT2A fusion nobody has tested, and the direction of that effect is not predictable from anything here.
- Search ProteomeXchange PXD009338 for DAXX and ATRX peptides below the top-293 rank cut. This converts a `measured-not-reported` absence into a real one either way, needs no wet work, and is the cheapest open item in this beat.
- Why is TRIM28 at 7 of 7 baits and unremarked anywhere in this graph? A SUMO E3 ligase and H3K9me3 co-repressor recovered with every MLL fusion reaches the same heterochromatin territory this beat argues about, by a protein nobody here has written about.
- Is the PML row SUMO-dependent? A SUMOylation-deficient fusion bait or a PML SIM mutant would separate a real contact from a SUMO-abundance artefact, and the single weak SUMO2 row is the only evidence in the table either way.
- No proximity-labelling experiment has ever been run on a KMT2A fusion. BioID or TurboID would see the transient chaperone hand-off that an affinity purification is designed to wash away, and it is the only method that could turn this absence into a real negative for DAXX.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In the 960-protein MLL-fusion interactome of Skucha et al. - the only published affinity-purification mass-spectrometry survey of MLL fusion proteins, covering seven baits - neither DAXX nor ATRX appears, nor do SP100, RNF4, SETDB1, HIRA, UBN1, CABIN1, ASF1B, CHAF1A, CHAF1B, MORC3, ATF7IP or the histone H3.3 genes under either their current (H3-3A, H3-3B) or retired (H3F3A, H3F3B) symbols.
MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity. · Skucha A, Ebner J, Schmöllerl J, et al. · Nature Communications 2018What this evidence cannot show
The support is Supplementary Table 5 (file 41467_2018_4329_MOESM4_ESM.xlsx), cached in this repository and read directly on 2026-09-24; the absence of DAXX and ATRX is a property of that table and not stated anywhere in the paper's text. The table is RANK-thresholded, not detection-thresholded - 291 to 293 preys per bait with maximum p-values from 0.0164 to 0.175 - so 'absent' means 'not in the top ~293 of any bait' and NOT 'not detected'. The baits are MLL-AF1p, MLL-AF4, MLL-AF9, MLL-CBP, MLL-EEN, MLL-ENL and MLL-GAS7, expressed constructs rather than endogenous fusions, and the cells are murine and human myeloid systems rather than infant B-lineage.
Evidence that argues against this (1)
- An AP-MS survey is biased against exactly this class of interaction. DAXX and ATRX bind H3.3-H4 as a chaperone, which is by design a transient hand-off rather than a stable complex, and a chaperone caught mid-transfer is the interaction most likely to be lost in a wash. The absence is therefore weaker evidence against a functional relationship than it looks, and the honest reading is that nobody has assayed this contact with a method that could see it - no crosslinking, no proximity labelling and no BioID has ever been run on an MLL fusion against the H3.3 chaperones. Elsässer SJ et al. 2012
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
PML is recovered against all seven MLL-fusion baits in the same table - MLL-AF9 p 6.61e-06 with 22 and 22 spectral counts, MLL-AF4 p 0.0282 with 4 and 4 - and it is the only PML-nuclear-body component present: SP100 and RNF4 are absent and SUMO2 appears against MLL-AF4 alone at p 0.147, so the network contains a PML signal without a PML-body signature.
MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity. · Skucha A, Ebner J, Schmöllerl J, et al. · Nature Communications 2018What this evidence cannot show
Read from the cached Supplementary Table 5 on 2026-09-24; the paper's text does not discuss PML. Graded strong rather than established for the SIGNATURE claim because absence in a rank-thresholded table is not absence of detection, so 'no SP100' means SP100 did not reach the top ~293 of any bait. PML is an abundant, SUMO-rich and famously sticky protein in affinity purifications and appears in many unrelated interactomes; the negative control column in this table reads 0|0 for PML, which is the only purity information available and is a single number.
Evidence that argues against this (1)
- The interpretation 'a contact, not a compartment' has a deflating alternative that the table cannot exclude: PML nuclear bodies are disrupted by the lysis and purification conditions of a standard AP-MS, so recovering the scaffold without its clients is also what a dissolved body looks like. Deciding between the two needs an in-cell method - two-colour immunofluorescence for the fusion against PML in a t(4;11) line, which `gene-pml` already proposes and which no paper has done. Ishov AM et al. 1999
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The interaction records that placed DAXX with AFF4, DOT1L, MLLT10 and MLLT3 come from BioPlex, a proteome-scale affinity-purification survey in 293T and HCT116 cells using anti-tag co-immunoprecipitation of tagged, overexpressed baits, and therefore describe DAXX's behaviour with individual wild-type fusion PARTNERS in an epithelial cell line rather than with any KMT2A fusion protein.
Dual proteome-scale networks reveal cell-specific remodeling of the human interactome. · Huttlin EL, Bruckner RJ, Navarrete-Perea J, et al. · Cell 2021What this evidence cannot show
Abstract read from efetch 2026-09-24; this entry uses the paper to characterise the METHOD behind the DAXX rows, which the abstract states, and not to make any claim about DAXX specifically, which the abstract does not mention. 293T and HCT116 are an embryonic kidney and a colorectal carcinoma line; neither is haematopoietic. The rows are prey-only and resolve to one laboratory.
Evidence that argues against this (1)
- Prey-only in one laboratory is weak, but 'proteome-scale therefore low quality' is a fallacy this project has already measured and rejected: a hand-verification of the bait-supported filter found 41 of 56 partners have a focused source and that the heuristic labelling of atlas studies was wrong five times out of nine at the top of the ranking. The right criticism of the DAXX rows is the cell type and the overexpression, not the scale of the study. Oncofusion knowledge graph (internal analysis) 2026
DAXX is not in any of the 960 rows of the only MLL-fusion AP-MS interactome, at any of seven baits
ATRX likewise absent from all 960 rows
answers gene-pml's open question: PML is at 7 of 7 baits and is the ONLY PML-body component present - no body-level signature
TRIM28/KAP1 is recovered against all seven MLL-fusion baits and is unremarked anywhere in this graph
USP7 is recovered against three of the seven baits in the same table
- related to →Histone H3.3 - one protein, two genes, and the reason this graph should not file it as gene-h3f3a
neither H3-3A nor H3-3B, under current or retired symbols, appears in the fusion interactome
a dependency without a contact: the DepMap signal survives every null and the one dedicated interactome experiment does not recover the protein
answers gene-trim28's question: SETDB1 is NOT in the 960-protein network, so TRIM28 is there as a lone scaffold rather than a TRIM28-SETDB1 module
the 7-of-7 PML co-purification is the one biochemical hint behind this otherwise empty question