The 110 MLL-AF4-specific preys are 14.5% mitochondrial and 34% nuclear; the 42 pan-fusion preys are 2.4% mitochondrial and 88% nuclear — Fisher p = 0.042 on a nuclear-extract pulldown
also called mitochondrial contaminant signature · compartment of the Skucha singleton tail · ACADM GLUD1 ECHS1 PDHA1
The pulldowns were done on nuclear extracts, so anything that lives in mitochondria should not be there. One in seven of the proteins said to be unique to MLL-AF4 is a mitochondrial enzyme, against one in forty of the proteins shared by all seven fusions. That pattern is what mass-spectrometry background looks like.
The same thing, for a scientist
Skucha 2018 purified Strep-HA-tagged MLL-fusions from **nuclear extracts** of Jurkat cells (benzonase- and RNase-treated, 420 mM salt extraction adjusted to 150 mM, single-step StrepTactin, three washes) [E1]. Scoring both prey classes against live UniProt subcellular-location comments [E2]: of the 110 MLL-AF4 singletons, **16 (14.5%) carry a mitochondrial location and only 37 (33.6%) a nuclear one**; of the 42 pan-fusion proteins, **1 (2.4%) is mitochondrial — PML, which is dual-annotated — and 37 (88.1%) are nuclear**. Fisher exact two-sided p = 0.0417 on the mitochondrial share. The 16 are ACADM, ACSF3, ATPAF2, C1QBP, CHCHD4, ECHS1, FDXR, GLS, GLUD1, LARS2, LIPT1, NUBPL, PDHA1, SLC25A25, TRAF3 and TUFM — a fatty-acid-oxidation, TCA and mitochondrial-translation signature with no plausible route to a chromatin bait. **This is the single most interpretable quality signal available for the AF4-specific list, and it says the list is substantially background.**
The measurement
| class | n | mitochondrial | nuclear | neither | unresolved |
|---|---|---|---|---|---|
MLL-AF4 singletons (n_fusions == 1) | 110 | 16 (14.5%) | 37 (33.6%) | 58 | 0 |
MLL-AF4 pan-fusion (n_fusions == 7) | 42 | 1 (2.4%) | 37 (88.1%) | 5 | 0 |
Fisher exact, two-sided, on the mitochondrial share: p = 0.0417. All 152 accessions resolved against live UniProt; none was dropped 2.
The single pan-fusion "mitochondrial" call is PML, which UniProt annotates to nucleus, PML body and mitochondrion. It is a dual-compartment protein and an established chromatin factor, not a matrix enzyme. Removing it makes the pan-fusion mitochondrial count zero.
The nuclear column is the mirror image and is arguably the stronger half: 88.1% of the shared core is nuclear and only 33.6% of the specific tail is. In a nuclear-extract pulldown that difference is the experiment discriminating signal from background, and the AF4-specific class is on the wrong side of it.
The sixteen
ACADM (medium-chain acyl-CoA dehydrogenase, matrix), ACSF3, ATPAF2, C1QBP, CHCHD4, ECHS1 (enoyl-CoA hydratase, matrix), FDXR, GLS (glutaminase), GLUD1 (glutamate dehydrogenase 1, matrix), LARS2 (mitochondrial leucyl-tRNA synthetase), LIPT1, NUBPL, PDHA1 (pyruvate dehydrogenase E1 alpha, matrix), SLC25A25, TRAF3, TUFM (mitochondrial elongation factor Tu).
Four of these — ACADM, ECHS1, GLUD1, PDHA1 — are matrix enzymes of beta-oxidation, the TCA cycle and pyruvate entry. C1QBP and TUFM are among the most frequently reported non-specific pulldown contaminants in the CRAPome-era literature. Two of the six p<0.05 dispatch targets for this beat, ACADM and GLUD1, are in this list, and each is adjudicated in its own entry.
What this does NOT show. It does not show that every mitochondrial protein in a nuclear pulldown is background — GLS and C1QBP have reported non-canonical nuclear pools, and moonlighting metabolic enzymes on chromatin are a real, if contested, literature. It shows that the rate of such calls is six-fold higher in the specific tail than in the shared core in the same experiment, which is a property of the tail and not of any individual protein.
| script | analyses/q180-is-af4-specificity-a-depth-artefact/compartment_of_the_tail.py |
|---|---|
| status | run |
| result summary | MLL-AF4 singletons: 16/110 mitochondrial (14.5%), 37/110 nuclear (33.6%). MLL-AF4 pan-fusion core: 1/42 mitochondrial (2.4%, and that one is PML, dual-annotated), 37/42 nuclear (88.1%). Fisher two-sided p = 0.0417. |
| shown for fusion | YES for MLL-AF4 as one of seven baits in one experiment, in a Jurkat T-ALL host from a tagged single-copy transgene. |
| epigenetic readout | none - a compartment audit of a proteomic prey list. |
| absence class | measured-not-analysed |
| absence class note | The peptide-level evidence for every one of these sixteen calls is public in PXD009338 and has never been reopened. Whether any of them rests on a single borderline PSM is answerable from the deposit with no wet work. |
inputs
[ "analyses/skucha-mll-fusion-interactome/cache/41467_2018_4329_MOESM4_ESM.xlsx", "UniProtKB REST, subcellular-location comments, fetched 2026-09-24" ]
measurements
[
{
"quantity": "proportion_mitochondrial",
"subject": "MLL-AF4 singleton preys",
"value": 0.1455,
"units": null,
"n": 110,
"universe": "the 110 preys of Skucha 2018 Suppl Table 5 recovered under MLL-AF4 and no other bait, scored on UniProt cc_subcellular_location",
"null": 0.0238,
"note": "null column is the same quantity in the 42 pan-fusion preys of the same bait; Fisher exact two-sided p = 0.0417"
},
{
"quantity": "proportion_nuclear",
"subject": "MLL-AF4 singleton preys",
"value": 0.3364,
"units": null,
"n": 110,
"universe": "same 110 preys, same annotation source",
"null": 0.881,
"note": "null column is the pan-fusion class in the same bait; the pulldown was from nuclear extract"
}
]mitochondrial members of the af4 specific tail
[ "ACADM", "ACSF3", "ATPAF2", "C1QBP", "CHCHD4", "ECHS1", "FDXR", "GLS", "GLUD1", "LARS2", "LIPT1", "NUBPL", "PDHA1", "SLC25A25", "TRAF3", "TUFM" ]
- If a future beat proposes targeting a metabolic enzyme on the strength of 'it is in the KMT2A::AFF1 interactome', the enzyme is most likely in a class that is six-fold enriched for mitochondrial proteins in a nuclear-extract pulldown. Metabolic dependencies in KMT2A-r leukaemia may well be real, but this dataset is not evidence for a physical contact.
- Is the mitochondrial rate in the AF4-specific tail different from the mitochondrial rate in the OTHER six baits' specific tails? If all seven tails are equally mitochondrial the signature is a property of the preparation and says nothing about MLL-AF4; if MLL-AF4's is worse, that would need explaining. This was not computed and is one pass over the same workbook.
- C1QBP and GLS both have reported non-canonical nuclear pools. Are either of them in the corpus already as anything other than a metabolic enzyme? GLS appears in ana-q109-serine-panel-enumeration-what-was-declined-and-the-measurement-that-declined-it and has no entry of its own.
independently consistent: the declined survivors ACADM and GLUD1 are both mitochondrial enzymes
an independent axis - compartment, not p-value - reaching the same verdict on the same 110 preys
the single mitochondrion-annotated protein in the 42-member pan-fusion core, and it is dual-annotated to nucleus and PML body
having removed the negative-control column as a discriminator, the compartment audit is the only quality axis left that separates the two prey classes
the second of the four mechanisms: a reader skimming the tail correctly declines its mitochondrial members and stops reading
the decline rests on the class-level mitochondrial enrichment measured there, not on ACADM's annotation alone
same class-level compartment argument; GLUD1 adds abundance and hexameric assembly
the compartment enrichment is the second of the four measurements the hypothesis rests on
the compartment audit carries the quality argument because the negative-control column was shown there to be unable to separate the two prey classes