FTSJ3 (Q8IY81) - the pre-rRNA 2'-O-methyltransferase at the weakest edge of the MLL-fusion nucleolar block: two baits of seven, not the index fusion, and never screened

THE INDEX FUSION DID NOT RECOVER THIS PROTEIN. FTSJ3 is absent from the MLL-AF4 column of Supplementary Table 5 entirely -- there is no p-value and no spectral count for it under the index bait. It is an MLL-FUSION interactor, not a demonstrated KMT2A::AFF1 interactor, and nothing in this entry should be read as a KMT2A::AFF1 contact. The 2 baits that did recover it are MLL-CBP, MLL-EEN. Negative control: ZERO spectra. THIS IS THE WEAKEST ROW OF THE ELEVEN PROTEINS THIS BEAT WROTE, on every axis at once: fewest baits (2 of 7), absent from the index column, below the screen threshold, and no independent connection to leukaemia in any system. It is recorded to BOUND the nucleolar block rather than to support it. THE HEADLINE OF THIS ENTRY: this protein is NOT in the MLL-AF4 column. It is an MLL-FUSION interactor and NOT a demonstrated KMT2A::AFF1 interactor. Six of the eleven nucleolar-block proteins written by this beat are in that position - RRP12, WDR18, UTP14A, RRP9, UTP18 and FTSJ3. The 'thirteen-protein nucleolar block' is a POOLED statement over seven baits; in the index fusion's own column there are seven members and four below p = 0.05. AND IT IS BELOW THE SCREEN THRESHOLD TOO: the source paper's shRNA library covers only the 128 interactors shared by at least five baits, so this protein was never functionally tested against any MLL fusion either. Both records - binding and function - are empty for the index fusion. FUNCTIONAL SCREEN -- NOT SCREENED. The shRNA library (MOESM5) covers the 128 interactors shared by at least five of the seven baits: 768 rows, 128 symbols, exactly six shRNAs each, counted with openpyxl by this beat on 2026-09-21. This protein's symbol is NOT in that library, because it falls below the five-bait threshold. So unlike the seven-plus-bait members of this block it has never been functionally tested against any MLL fusion, even in the source paper. That is a stronger absence than measured-not-reported. THE BLOCK, re-measured by this beat on 2026-09-21 by re-running analyses/mll-fusion-interactome-complexes/run.py. A 26-subunit nucleolar / ribosome-biogenesis set returns 15 of 26 present in the 960-protein network against 1.25 expected by chance, hypergeometric p = 6.99e-14, q(BH) = 8.39e-13 across the twelve complexes tested -- the TOP-RANKED complex in the whole interactome, ahead of the CCT/TRiC negative control (q = 1.4e-09) and the SEC positive control (q = 1.5e-09). The fifteen are FBL, FTSJ3, GNL2, LYAR, NOLC1, NOP2, NOP58, PA2G4, RRP12, RRP9, TBL3, TCOF1, UTP14A, UTP18, WDR18. BUT THE BACKGROUND IS THE ~20,000-PROTEIN PROTEOME, not the ~4,600 proteins the study actually recovered before filtering, so every q here is an UPPER BOUND on significance and it is loosest for a complex of abundant nuclear proteins -- which this one is, more than any other row in that table. AND THE POOLED NUMBER IS NOT THE PER-BAIT NUMBER. Subunits of the block recovered per bait, out of 26: MLL-AF1p 8, MLL-AF4 7, MLL-AF9 12, MLL-CBP 10, MLL-EEN 12, MLL-ENL 11, MLL-GAS7 7. The index fusion is TIED LAST. In its column there are seven members and only four at p<0.05 -- GNL2, NOP2, PA2G4 and TBL3; LYAR (0.141), NOLC1 (0.134) and TCOF1 (0.163) are above it. So the strongest complex-level signal in this entire interactome is weakest in the fusion this project studies. SYSTEM: the AP-MS host is JURKAT, a human T-ALL line, while KMT2A::AFF1 disease is B-lineage infant leukaemia; a partner requiring a B-cell or infant proteome cannot appear here and a T-lineage partner can appear spuriously. The bait is a Strep-HA-tagged, SINGLE-COPY, doxycycline-induced transgene at an FRT/Flp cassette, not the endogenous fusion. Prey came from bulk nuclear lysate after a single STREP-Tactin step, so co-purification is NOT evidence of a direct contact. INCLUSION RULE: the paper kept the top 300 interactions per bait by combined p-value rather than applying a p threshold (Methods, verbatim: "All the identified bait-prey pairs were ranked by the combined p-value. The 300 most significant interactions per bait were retained."). In the MLL-AF4 column 186 of 291 listed p-values (63.9%) exceed 0.05; the pooled rate over all baits is 26.6%, and the pooled rate must NOT be quoted for a per-bait claim. BAIT DEPTH: MLL-AF4 is the shallowest of the seven baits by spectral counts, so a low or absent MLL-AF4 count is a property of the bait (transgene expression, purification depth, normalisation) and must not be read as a weaker association. ABUNDANCE IS THE LEADING ALTERNATIVE EXPLANATION. The nucleolus is the densest compartment in the nucleus, every bait here is a tagged nuclear protein, and the complex-level analysis in this repo exists to make exactly that point: its strongest per-bait signature in the index fusion's own column is CCT/TRiC, a canonical AP-MS contaminant, at 8 of 9 subunits with 6 at p<0.05 -- BETTER represented in MLL-AF4's column than the nucleolar block is. Complex-wide enrichment is not by itself evidence of specific binding. THE ARGUMENT ON THE OTHER SIDE: all thirteen members of the nucleolar block, this protein included, have a negative control of ZERO spectra. A non-specific sticky abundant protein would be expected to appear in the control purification too, and none of them does. That is the strongest single fact about this block and it is not explained by abundance alone. SELECTIVITY: the paper's own named NON-selectivity controls are RFC1, SF3A3 and CCT3 -- knockdown inhibited MLL-wild-type K562 and HL-60 as much as MOLM-13 -- against MEN1 as the selective positive. This protein is not among them. That is NOT evidence of selectivity: only 40 of the 128 entered the MLL-r versus MLL-wild-type re-screen and the paper does not say which 40, so absence from the control list is equally consistent with never having reached the second screen. Recorded as UNKNOWN. This entry does not nominate a target. Per-bait membership re-derived by this beat on 2026-09-21 by re-running analyses/skucha-mll-fusion-interactome/run.py (both controls PASS: 128 shared by >=5 baits, 42 by all seven, MEN1 present with all seven) and analyses/mll-fusion-interactome-complexes/run.py. n_fusions and the MLL-AF4 p-value were supplied pre-verified by the coordinator from the cached MOESM4 xlsx and were independently reproduced by this beat's re-run of the complex analysis.

• IntAct has never curated this dataset, so no independent curator has re-examined these calls and no interactome-coverage metric built on IntAct can see them; the raw spectra at ProteomeXchange PXD009338 have not been reanalysed here. The authors themselves removed seven proteins from the shared set by hand as frequently observed contaminants, which concedes that a rank cutoff lets contaminants through. ProteomeXchange/PRIDE 2018
• The same paper that motivates the nucleolar reading supplies its null hypothesis: MLL-AF4 is reported predominantly nucleolar, so every abundant nucleolar protein becomes a high-probability co-purification partner on compartment grounds alone. And the prediction that reading makes fails -- if nucleolar residency of the bait drove recovery, the index fusion should pull MORE of the block than the other six baits, and it pulls the FEWEST (7 of 26, tied last with MLL-GAS7). The report is also an ectopic-overexpression system, which this project treats as weak evidence about an endogenous protein. Siemund AL et al. 2022

HUMAN proteins via a YEAST TWO-HYBRID screen plus follow-up, 2011 PLoS One. Not leukaemia, not a fusion, not haematopoietic. Read from the abstract only. A two-hybrid interaction is a heterologous-system binary assay with well-known false-positive rates, and the paper's own framing is that human NIP7 has a different partner set from yeast Nip7p - so this is a genuine but modest functional credential. It is also one of the two ECO:0000269 citations behind UniProt's nucleolar localisation for FTSJ3, so localisation and function again rest partly on one source.

• FTSJ3's own UniProt record carries its nucleolar assignment partly by HAMAP rule (ECO:0000255|HAMAP-Rule:MF_03163), which is automatic annotation propagated from a family profile rather than a measurement. So the strength of 'FTSJ3 is nucleolar' is two experimental citations plus a rule, which is thinner than the compartment call sounds, and no sub-compartment exists at all. UniProt Knowledgebase 2026

HUMAN cells and in vitro, 2019 Nature, with RiboMethSeq mapping. Read from the abstract only. NOT RELEVANT TO KMT2A BIOLOGY and deliberately not used as though it were: it is recorded because it is the reason FTSJ3 has a substantial literature, which is what makes a naive literature count on this gene misleading. It does establish the enzymatic activity - FTSJ3 is a genuine 2'-O-methyltransferase - which supports the pre-rRNA methyltransferase assignment by mechanism even though the substrate here is viral.

• Recorded to block an inference this entry declines to make: a 2'-O-methyltransferase that acts on a viral genome through an adaptor is a promiscuous enzyme, and promiscuity is a reason to expect it in many protein lists rather than a reason to read its presence in one as meaningful. It also means that FTSJ3's 124-record literature is overwhelmingly virological, so the zero on-topic count here is not the absence of a field - it is the absence of an intersection with a field that exists. Morello LG et al. 2011

A counted negative with a date, the exact query, the querytranslation and the tool. Bounded by TITLES AND ABSTRACTS ONLY - supporting figures not searched (rule 80). NOTE A RECALL HAZARD recorded rather than resolved: 'SPB1' is a short token that collides with other entities and was included for recall; the intersection is empty so nothing needed title-screening, but a non-zero SPB1 result would have had to be screened one by one. The FTSJ family is also a trap in the other direction - FTSJ1, FTSJ2 and FTSJ3 are three different genes with FTSJ2 being mitochondrial, so no FTSJ* wildcard was used.

• As for UTP14A, RRP9 and UTP18, the 'screened but not reported' counter does NOT apply: FTSJ3 is at two baits, far below the five-bait library threshold, so it was never functionally tested against an MLL fusion at all. Absence class never-measured. Skucha A et al. 2018

Database record fetched live 2026-09-21. The location comment is 'Nucleus, nucleolus' with ECO:0000255|HAMAP-Rule:MF_03163 alongside ECO:0000269 PubMed 12429849 and 22195017 - so part of this assignment is AUTOMATIC ANNOTATION from a family profile, not a measurement, which per rule 91 must be distinguished and is. This beat opened only 22195017. NOTE the entry name is SPB1_HUMAN and the protein name leads with 'pre-rRNA 2'-O-ribose RNA methyltransferase FTSJ3' - reading the entry name back alone would not confirm the symbol. NO sub-nucleolar assignment: a bounded absence.

• A compartment call that leans on a HAMAP family rule is a call about what proteins of this family usually do. For a protein whose best-characterised activity is on a viral RNA in the cytoplasmic TRBP complex, the possibility of a substantial non-nucleolar pool is real and unaddressed, and would weaken its membership of a nucleolar block in the same way the zebrafish-derived nucleoplasmic note weakens WDR18's. Ringeard M et al. 2019