BAG6 / BAT3 - a DOT1L partner that is about H3K79me2 itself, reciprocally co-immunoprecipitated, and not a spoke

One database on one day, REST endpoint (brief rule 93). Two records from ONE publication, curated in both directions - an internal control, not independent replication. BAG6's 910 total records make it a high-degree node in IntAct generally, which raises the prior on non-specific recovery in any single dataset.

• 'Curated in both directions' is one figure curated twice, not two experiments. Re-read at the endpoint on 2026-09-21, EBI-7776967 and EBI-7776998 both carry the single annotation 'figure legend (f4a)' and both carry intactMiscore 0.35; they differ only in which accession holds the bait role. So the reciprocity here is a curation convention applied to one panel, not the reciprocal-co-IP control that phrase normally signals. For comparison, the PSIP1-SUMO1 rows in this same beat are annotated to three different figures (f1d, f1e, f5a/f5b) and the PSIP1-RBBP8 rows score 0.6. EMBL-EBI IntAct 2026
• The 910 that carries the promiscuity argument is itself an endpoint number, and BAG6 is one of the accessions where the endpoints diverge most. Measured the same day, PSICQUIC id:P46379 returns 311 against REST's 910 - so under the query form this project's two earlier interactome analyses used, BAG6 is a third as high-degree and the prior on non-specific recovery that this caveat rests on is a third as strong. A promiscuity argument built on a raw record count inherits whatever the endpoint happens to expose. EMBL-EBI IntAct and PSICQUIC 2026

Abstract read this session; full text not retrieved. U2OS osteosarcoma - no haematopoietic system, no KMT2A fusion, no primary sample (brief rule 73). The mechanism rests on knockdown rather than on a separation-of-function mutant of the motifs, and residue boundaries for the UBL and UIM are not in the abstract, so none are asserted here. New to this graph (checked with scripts/check-citation-reuse.py).

• The paper's own framing is that H3K79me2 is negligible during S phase and matters for 53BP1 only in G1/G2 when H4K20me2 is low, because 53BP1 binds H4K20me2 with much higher affinity. So the H3K79me2 arm of 53BP1 recruitment is conditional on cell-cycle phase, and a BAG6 effect measured outside G1/G2 would be expected to show nothing. Wakeman TP et al. 2012

Titles and abstracts only; supporting figures not searched (brief rule 80). Name list tried: DOT1L, Dot1L, BAG6, BAT3. The aliases Scythe, G3 and the phrase 'HLA-B-associated transcript 3' were NOT run - and Scythe in particular is the name much of the apoptosis literature uses, so this count is a floor. Operand control (brief rule 67): each clause alone returns a large literature, so the one is a real intersection.

• The alias gap this caveat calls a floor was measured and is not one, and the field tag is not doing what the caveat says. Re-run 2026-09-21: adding Scythe, BAG-6 and "HLA-B-associated transcript 3" to the name list leaves both counts exactly where they were - 1 and 8 - while Scythe[tiab] alone returns 65, so the instrument reaches the apoptosis literature and the intersection with DOT1L is genuinely empty beyond the source paper. Separately, PubMed's querytranslation shows that the [tiab] after each parenthesised OR group is DROPPED and both clauses expand to [All Fields], so these are all-fields counts and not the title-and-abstract counts the caveat describes - which makes the 1 a stronger absence than claimed, not a weaker one. NCBI PubMed 2026
• The query's frame, not its name list, is what hides the relevant corroboration. Bounding on 'DOT1L or leukaemia' cannot reach BAG6's independent nuclear chromatin biology: Bach et al. 2015 identify BAT3 (Scythe/BAG6) as a nuclear partner of the transcriptional repressor TRPS1 in oestrogen-receptor-positive breast cancer cells, show BAG6 silencing blocks nuclear accumulation of cathepsin D, and place both proteins on chromatin. So 'one paper, no replication in fifteen years' is true of the DOT1L pair and false of the premise behind it - that BAG6 has no business on chromatin - and this entry's promiscuous-chaperone objection is correspondingly weaker than it looks. Bach AS et al. 2015

Statement about the record. Unlike BICRA, CTCF, PARP1 and PRDM9 in this tranche, BAG6's compartment annotations are numerous and experimentally coded and this beat did NOT open them to check rule 91 - so the attributions are reported as the database gives them and have not been verified as being about localisation. The tissue comment names one cell type at protein level and is uncoded.

• Three simultaneous compartments invite a reading this entry should resist: that BAG6 is 'everywhere' and therefore its compartment carries no information. The opposite is closer to true, and the GET4 clause is why. A protein partitioned between cytosol and nucleus BY A NAMED RETENTION FACTOR has a compartment that is a variable, not a constant, and the variable has a handle on it. The honest gap is different and larger: nobody has measured the nuclear FRACTION in any cell, so 'cytosol and nucleus' could describe 99:1 or 1:1, and the entire relevance of BAG6 to this graph - supporting DOT1L's H3K79 dimethylation - depends on the nuclear pool being non-trivial in a haematopoietic cell. The U2OS knockdown result shows the nuclear pool is functionally significant THERE; that is one cell type, and a bone tumour line. Wakeman TP et al. 2012

This restates this entry's existing E2 with the localisation content foregrounded and the identifiers re-fetched rather than copied from the sibling block. U2OS osteosarcoma only. The 'co-localises with DOT1L at histone H3' phrasing is the source's; it is a biochemical co-association statement rather than a subnuclear compartment assignment, and this entry does not upgrade it into one. Abstract-level reading.

• The most useful objection is to this beat's own framing rather than to the paper. Calling 'co-localises with DOT1L at histone H3' a localisation result would be a category error of exactly the kind this tranche found in BICRA and MAML2: it is co-association with a chromatin substrate, which constrains the compartment to 'nuclear, on chromatin' - something the NLS at 1012-1054 and four ECO:0000269 nuclear attributions already establish - and adds no subnuclear address at all. That is why this entry's subnuclear field records an absolute zero rather than borrowing the DOT1L association. The separate and real limitation is the system: H3K79me2 is the mark KMT2A-fusion leukaemia lives on, the only paper connecting BAG6 to it used an osteosarcoma line, and nothing guarantees the GET4-set nuclear partition is the same in a blast. UniProt Consortium 2026