ILF3 / NF90 - a nucleolar dsRNA-binding coactivator whose chromatin residency was mapped in the same leukaemia line that proximity-labelled KMT2A

No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.

• ILF3 is a DepMap common essential at -1.089 across 1,208 lines, so any intervention against it is an essential-function liability in every tissue, not a selective one - and it is marginally LESS required in KMT2A-rearranged lines than in other leukaemias, which is the wrong direction for a target.
• ILF3 functions as an obligate heterodimer with ILF2, so a single-agent degrader or inhibitor would remove a pair, and the screen evidence that exists treats ILF2/ILF3 as one unit - meaning selectivity cannot be engineered at the level of one partner.
• The one leukaemia-context result places ILF2/ILF3 as a MODIFIER of response to a GSPT1 degrader. A modifier of drug response is as likely to be a resistance mechanism as a target: loss of ILF2/ILF3 could blunt an otherwise effective degrader rather than help.
• ILF3 is an antiviral effector with PKR-dependent regulation and a documented role in HIV-1 RNA stabilisation, so perturbing it in an already immunosuppressed infant carries an infection-control liability that has nothing to do with leukaemia biology.

1. Does NF90/NF110 occupy KMT2A::AFF1 target promoters in a t(4;11) line? Its K562 occupancy is at immediate-early-gene promoters, and the fusion's programme is HOXA9/MEIS1 - a CUT&RUN would either make the proximity row a colocalisation or kill it.
2. Is the KMT2A that ILF3 biotinylated the nucleolar pool? ILF3 has an ECO:0000269 nucleolar assignment and this graph holds an open absence that nobody has put the nucleolar KMT2A fusion and its nucleolar neighbours together. Fractionated BioID would separate them.
3. Which Mediator subunit carries the NF45-NF90 cooperation? The in vitro result names Mediator as a whole; MED19 is written in this same tranche from the Mediator side, and nothing connects the two claims on evidence.
4. Which ILF3 isoform is in the proximity record? NF90 and NF110 are isoforms of one gene, the BioID construct's identity is not resolvable from the IntAct row, and their chromatin behaviours are reported together rather than separately.
5. Does the ILF2/ILF3 modifier effect on GSPT1-degrader response hold in KMT2A-rearranged AML or ALL? The screen was in AML without stated fusion status, and KMT2A-rearranged leukaemia is the subtype where a differentiation-releasing agent would matter most.

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt Q12906 is the single reviewed human ILF3 record at 894 aa (95,338 Da) with a DZF domain (5-378), two dsRNA-binding motifs (398-467, 524-590), a bipartite NLS (371-389) and a PRMT1-interaction region (609-894), no chromatin writer/eraser/reader module, three experimentally coded locations including Nucleus-nucleolus, and phospho-regulated nucleocytoplasmic partitioning controlled by ILF2.

   UniProtKB Q12906 (ILF3_HUMAN) features, comments, PTMs and subcellular-location evidence codes; HGNC:6038 prev_symbol and alias_symbol via rest.genenames.org; retrieved 2026-09-22 · UniProt Consortium / HGNC 2026
   ▸What this evidence cannot show

   Retrieved 2026-09-22. SIX isoforms are annotated (Q12906-1 to -6) and 894 aa is the canonical; NF90 and NF110 are different isoforms of this one gene, so a claim about 'ILF3' may be true of one isoform and not the other, and the ChIP work is reported for NF90/NF110 together. Domain coordinates are on the UniProt canonical frame; RefSeq was not checked. Absence of an annotated chromatin module is absence of an annotation.

   ▸Evidence that argues against this (1)

   • UniProt's coded experimental annotations frequently point at a paper that is not about the annotated fact, and filtering on evidence code discards isoform-scoped comments - which matters unusually here because NF90 and NF110 are isoforms of this entry and a single location comment cannot distinguish them. UniProt Consortium 2025
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   NF45/ILF2 and NF90/NF110/ILF3 are constitutively expressed chromatin-interacting proteins localised predominantly in the nucleus, and NF90/NF110 ChIP-seq in K562 erythroleukaemia cells shows enriched association with chromatin at active promoters and strong enhancers, with specific occupancy of the immediate-early-gene promoters EGR1, FOS and JUN.

   Inducible expression of immediate early genes is regulated through dynamic chromatin association by NF45/ILF2 and NF90/NF110/ILF3. · Wu TH, Shi L, Lowe AW, et al. (5 authors; senior author Kao PN) · PLoS One 2019
   PMID 31022259DOI
   ▸What this evidence cannot show

   The K562 ChIP-seq is cited by this paper as previously reported rather than generated in it; the new ChIP here is in serum-starved HEK293. K562 is a BCR::ABL1 chronic myeloid leukaemia line with NO KMT2A rearrangement, so occupancy there is not occupancy in a fusion cell, and the occupied loci are immediate-early-gene promoters rather than the fusion's HOXA9/MEIS1 programme. ChIP of an RNA-binding protein can report RNA-mediated rather than DNA-mediated association. Read as abstract and title.

   ▸Evidence that argues against this (1)

   • Chromatin immunoprecipitation of an abundant nucleic-acid-binding protein at active promoters is the classic false-positive geometry - highly transcribed, open, hyper-accessible regions are recovered non-specifically in ChIP from every cell type, which is why the field requires a matched input and a sonication-bias control before calling promoter occupancy specific. Park D et al. 2013
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The NF45-NF90 and NF45-NF110 complexes function as direct transcriptional coactivators at the c-fos gene, purified in an unbiased way from a defined in vitro transcription system, acting through both the upstream enhancer and the core promoter, independently of their double-stranded RNA-binding activities, and cooperating with the general coactivators PC4 and Mediator.

   The RNA binding complexes NF45-NF90 and NF45-NF110 associate dynamically with the c-fos gene and function as transcriptional coactivators. · Nakadai T, Fukuda A, Shimada M, et al. (5 authors; senior author Hisatake K) · J Biol Chem 2015
   PMID 26381409DOI
   ▸What this evidence cannot show

   A reconstituted in vitro transcription system on one promoter, plus knockdown in mouse cells; not a genome-wide result and not a leukaemia system. The Mediator cooperation is a functional statement in a purified system, not a mapped subunit contact, so it does not identify which Mediator subunit is involved and cannot be joined to this tranche's MED19 entry on evidence. Read as abstract and title.

   ▸Evidence that argues against this (1)

   • A coactivator activity purified from a defined in vitro transcription system on a single promoter template may not reflect the protein's genome-wide role: the same complex's in vivo chromatin association is reported as DYNAMIC and stimulus-dependent, which means the constitutive coactivator function measured in vitro and the inducible occupancy measured in cells are not the same claim. Wu TH et al. 2019
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   The single publication behind ILF3's presence in the interactome read queue is a 50-RBP BioID2 proximity map in K562 cells, deposited to IMEx as IM-30059, in which no super elongation complex component, no KMT2A-complex component and no fusion protein is used as a bait - KMT2A, MEN1, PSIP1, AFF1, AFF4, ELL, MLLT1, CDK9, CCNT1 and DOT1L all appear only in the prey column.

   Systematic identification of post-transcriptional regulatory modules. · Khoroshkin M, Buyan A, Dodel M, et al. (22 authors; senior author Goodarzi H) · Nat Commun 2024
   PMID 39251607DOI
   ▸What this evidence cannot show

   A single publication, one cell line, an overexpression BioID2 map, and the curated IMEx rows are the authors' filtered bona fide set rather than the raw hit lists. K562 is a BCR::ABL1 CML line with no KMT2A rearrangement. This paper is already cited in 12 entries in this graph, including a dedicated analysis of the deposition, so the platform-level facts here restate an existing entry deliberately rather than re-deriving them; what is new is the per-bait ILF3 reading.

   ▸Evidence that argues against this (1)

   • Proximity recovery is organised by subcellular compartment rather than by binding - the largest published proximity map of a human cell shows a prey's bait profile is dominated by where it lives - so recovery of nuclear elongation factors by a nuclear RNA-binding bait is the expected consequence of shared compartment occupancy alone. Go CD et al. 2021
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   IntAct REST serves three binary ILF3 rows against this graph's core scaffold and elongation proteins - KMT2A, PSIP1 and ELL - all from one spoke-expanded n-ary proximity record (EBI-56904857), all MI:1314 proximity-dependent biotin identification typed MI:2364 proximity at MI-score 0.27 in K562, and in all three ILF3 is the BAIT.

   IntAct REST findInteractions/Q12906 (987 indexed interactions), filtered to Q03164/O00255/O75475/Q9UHB7/Q8TEK3/P51825/Q03111/P42568/P55199/P55197 with experimentalRole, detectionMethod and hostOrganism read per row; retrieved 2026-09-22 · IntAct molecular interaction database (EMBL-EBI) 2026
   ▸What this evidence cannot show

   Retrieved 2026-09-22. MI-score 0.27 is the lowest band IntAct assigns, appropriate to a single-method single-publication proximity row. The three rows are ONE n-ary record spoke-expanded, so they are one experiment and not three observations. The read queue lists two baits (KMT2A, PSIP1) and not ELL, because the queue's bait set omits ELL, MLLT1, MLLT3 and MLLT10 - all canonical entries in this graph - so the queue's bait counts are floors for every partner in it.

   ▸Evidence that argues against this (1)

   • This graph's own audit of the same deposition found that most preys are recovered by four or fewer of the 50 baits with a mean of 7.23 and a maximum of 40, so a prey recovered by one bait is at the sparse end of a distribution in which promiscuity is common - and the same audit found KMT2A recovered by 25 of 50 baits, which is the reading that undercuts any specific claim about ILF3's neighbourhood containing KMT2A. IntAct molecular interaction database (EMBL-EBI) 2026
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

   ILF3 is a DepMap common essential - mean gene effect -1.089 across 1,208 lines - and lies in the anti-selective half for KMT2A rearrangement at rank 15,542 of 17,728, being marginally LESS depleted in 11 KMT2A-rearranged leukaemia lines (-1.1716) than in 43 other leukaemia lines (-1.2331).

   analyses/depmap-kmt2a-selective-scan-full/results.json - 17,728 protein-coding genes, 11 Cellosaurus-annotated KMT2A-rearranged versus 43 other leukaemia lines, computed 2026-09-19 · DepMap (Broad Institute), via this repository's scan 2026
   ▸What this evidence cannot show

   The informative number here is the ABSOLUTE essentiality (-1.089, a strong common essential), not the rank: about half of any gene set lies above the median of the selectivity scan by construction, this project's own base rate is 45.3%, and reporting 'anti-selective at rank 15,542' as a finding would be reporting that null. The difference between the two leukaemia groups is +0.0615, which is within noise at these group sizes (11 versus 43). The stored scan records a compute date and NOT a DepMap release, so it cannot be told from quarterly drift.

   ▸Evidence that argues against this (1)

   • A single-gene CRISPR knockout cannot resolve an obligate heterodimer: ILF3's nuclear retention and its coactivator activity both require ILF2, so knocking out ILF3 alone measures the pair's dependency and a common-essential call says nothing about which half of the dimer the requirement sits in. Nakadai T et al. 2015
7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   A genome-wide CRISPR-Cas9 screen for effectors of response to CC-90009 - a first-in-class GSPT1-selective cereblon E3 ligase modulator that rapidly induces AML apoptosis and reduces leukaemia engraftment and leukaemia stem cells across 35 independent primary AML patient xenografts - identified the ILF2/ILF3 heterodimeric complex as a determinant of response.

   CC-90009, a novel cereblon E3 ligase modulator, targets acute myeloid leukemia blasts and leukemia stem cells. · Surka C, Jin L, Mbong N, et al. (29 authors; senior author Lu G) · Blood 2021
   PMID 33197925DOI
   ▸What this evidence cannot show

   The disease is AML, not B-lineage ALL, and the fusion status of the models is not stated in the abstract - so this is a leukaemia result and not a KMT2A-rearranged result. ILF2/ILF3 appears here as a drug-RESPONSE modifier in a degrader screen, not as an independent dependency and not in connection with KMT2A. The mechanism connecting the heterodimer to GSPT1 degradation is not established by a screen hit. Read as abstract and title; the screen's hit table was not opened.

   ▸Evidence that argues against this (1)

   • A genome-wide CRISPR modifier screen recovers regulators of the drug's delivery, the ligase machinery and general translational stress as readily as biology specific to the target, and ILF2/ILF3 is a broadly essential, highly abundant RNA-binding pair - so its appearance as a response modifier is compatible with a general fitness or translation effect rather than a specific role in GSPT1 degradation. DepMap (Broad Institute), via this repository's scan 2026
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   ILF3 has no KMT2A and no histone-mark literature under its own symbol - ILF3[tiab] AND KMT2A[tiab] = 0, AND MLL[tiab] = 0, AND H3K4[tiab] = 0, AND H3K9[tiab] = 0, AND HDAC[tiab] = 0 - the apparently substantial counts (histone 10, chromatin 17, epigenetic 11) are dominated by off-beat records, and the single ILF3-AND-H3K27 hit is a RETRACTED publication about the antisense lncRNA ILF3-AS1 rather than about the protein.

   NCBI E-utilities esearch, db=pubmed, querytranslation read back from every response, with PubMed publication types read from efetch for PMID 34141456; run 2026-09-22 · NCBI PubMed 2026
   ▸What this evidence cannot show

   All queries are one tag per term and never a parenthesised group; each querytranslation was read back and all were clean. The off-beat judgement rests on FOUR abstracts sampled from the ten-record histone set (PMIDs 27462403 adipocyte nuclei, 23856356 spermatozoal RNA, 34012635 gastric-cancer chemoresistance, 31747963 renal-carcinoma STAT3), not on all ten - so 'dominated by off-beat records' is a sampled judgement and is recorded as one. Instrument control through the identical code path: H3K79me2[tiab] = 224, NF90[tiab] AND chromatin[tiab] = 15. These zeros bound titles and abstracts only.

   ▸Evidence that argues against this (1)

   • The ILF3 symbol's own literature is the wrong index for this protein's chromatin biology, because the founding work is written under the NF90/NF110 names: NF90[tiab] AND chromatin[tiab] returns 15 and NF45[tiab] AND chromatin[tiab] returns 8, with different membership from the ILF3 queries, so a bounded absence on the approved symbol understates what has been published. NCBI PubMed 2026

• related to →KMT2A / MLL1 — the gene that breaks

  ILF3-bait BioID2 in K562 recovered KMT2A as prey (MI:1314 proximity, MI-score 0.27) - a proximity row, not a contact

• related to →PSIP1 / LEDGF p75 — the tether with the best therapeutic index in the whole interactome

  ILF3-bait BioID2 in K562 recovered PSIP1 as prey, same n-ary record EBI-56904857

• related to →ELL — rank 7 on selectivity and the deepest common-essential in the top ten, with one NMR fragment to its name

  ILF3-bait BioID2 in K562 recovered ELL as prey - a bait the read queue's six-bait set omits

• supports →The 50-RBP BioID2 proximity map in K562 - the first modern proximity dataset this graph can reach, and AFF1 is in it as prey

  ILF3 is one of the 50 RBP baits; per-bait reading of a deposition this graph had characterised only at platform level

• supports →The super elongation complex, sampled from the RNA side: ELL, ENL, CDK9 and cyclin T1 are all in the proximity record - ELL2, EAF1/2, AF9 and AF10 are at zero

  ILF3 is one of the baits behind the ELL recovery count in that analysis

• related to →Three CXXC-pathway activities converge on rDNA - KDM2B represses Pol I, KMT2A and SETD1A bind the whole locus, MLL-AF4 is predominantly nucleolar - and no paper contains two of them

  a third nucleolar protein (ECO:0000269) in the same neighbourhood, which labelled KMT2A in a leukaemia cell

• related to →The nucleolus — the one compartment with direct published images of a KMT2A fusion inside it

  ILF3 nucleolar by ECO:0000269; whether the KMT2A it biotinylated was the nucleolar pool is unresolved

• related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here

  chromatin-resident coactivator at active promoters and strong enhancers; writes and reads no mark; a PRMT1 substrate

• related to →KMT2A-rearranged acute myeloid leukaemia in children and infants

  ILF2/ILF3 is a CRISPR-screen modifier of GSPT1-degrader response in AML; fusion status of the models unstated

• related to →MED19 - the Mediator subunit curated against an engineered MLL-AFF4 fusion protein, and the one REST uses to deliver G9a

  NF45-NF90 coactivation cooperates with Mediator; MED19 reaches the same axis from the Mediator side and nothing joins the two claims

• ← related toBUD13 - a RES-complex splicing factor 1.66 Mb from KMT2A on 11q23.3, whose one chromatin paper puts it in the soluble fraction

  two of the 50 K562 BioID2 baits whose neighbourhoods contain the KMT2A scaffold; BUD13 recovered KMT2A+MEN1, ILF3 recovered KMT2A+PSIP1+ELL

• ← related toTwo of seven interactome read-queue partners should not get a gene entry, and the reasons are a pseudokinase with no nuclear pool and a membrane-skeleton protein whose two leukaemia hits are a gene-name artefact

  records the retracted ILF3-AS1/H3K27me3 paper this entry declines to cite