TNF — the only NF-kB-activating ligand with real KMT2A::AFF1 B-ALL evidence, and the evidence says the blasts are RESISTANT to it

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• Giving TNF, or sensitising to it, means deliberately raising inflammatory signalling in a child - the same axis the rest of this folder proposes to suppress. The two therapeutic directions in this beat are mutually exclusive in the same patient.
• TNF sensitisation in a transplant setting increases the risk of graft-versus-host disease along with graft-versus-leukaemia; the abstracts report GVL benefit and do not report GVHD.
• Amlexanox's mechanism here is assigned to S100A6 without excluding its known TBK1/IKK-epsilon activity. A repurposing argument built on the wrong mechanism will fail in a different way than expected.
• If NF-kB is required to hold the fusion on chromatin, and TNF activates NF-kB, then TNF exposure is predicted to strengthen fusion retention. The resistance finding and the retention finding pull in opposite therapeutic directions and nobody has put them in the same experiment.

1. Does amlexanox's TBK1/IKK-epsilon inhibition account for its S100A6 effect? If it does, an IKK-related kinase is already drugged in the index disease and nobody has connected it to the retention mechanism.
2. Is intracellular canonical NF-kB activity high or low in KMT2A::AFF1 B-ALL blasts? TNF resistance says nothing about basal pathway tone, and nobody has measured basal tone in this lineage.
3. Does IKK inhibition evict KMT2A::AFF1 from chromatin in a B-ALL line? This entry establishes that a B-lineage NF-kB-adjacent literature exists; it does not answer the retention question for this lineage.
4. Is S100A6 high in patient blasts, and does it predict transplant outcome? The mechanism is proposed as an explanation for poor allo-HSCT results and has never been tested against outcomes.

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

   MLL-AFF1-positive ALL cell lines are resistant to TNF-alpha. Upregulation of S100A6 expression and inhibition of the p53-caspase 8-caspase 3 pathway were observed only in MLL-AFF1-positive ALL cell lines in the presence of TNF-alpha; siRNA against S100A6 confirmed that S100A6 mediates the resistance by inhibiting that pathway, and the same pathway was confirmed in previously established MLL-AFF1 transgenic mice. Because TNF-alpha-mediated cytotoxic effector mechanisms contribute to the graft-versus-leukaemia effect, the authors propose this as a mechanism for the poor effect of allogeneic haematopoietic stem cell transplantation in MLL-AFF1-positive ALL.

   Resistance of MLL-AFF1-positive acute lymphoblastic leukemia to tumor necrosis factor-alpha is mediated by S100A6 upregulation. · Tamai H, Miyake K, ... Shimada T · Blood Cancer Journal 2011
   PMID 22829076DOI
   ▸What this evidence cannot show

   Abstract read; full text not retrieved. Cell lines plus a transgenic mouse, no primary patient blasts reported in the abstract, and the number of lines is not stated there. The GVL interpretation is a proposal, not a transplant outcome measurement. The paper does not measure NF-kB activity, occupancy or IKK, so it constrains what TNF does to these cells and says nothing directly about the retention mechanism.

   ▸Evidence that argues against this (1)

   • Resistance to TNF-mediated killing and dependence on intracellular NF-kB point in opposite therapeutic directions in the same cell. If IKK/NF-kB signalling is required to retain the fusion on chromatin, then TNF exposure - a canonical NF-kB activator - is predicted to strengthen retention even while failing to kill the blast. No experiment has put a TNF challenge and a fusion ChIP in the same system. Kuo HP et al. 2013
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   ALL positive for KMT2A/AFF1 translocation constitutes 60% of all infant ALL cases and has a poor prognosis even after allogeneic transplant, attributed to resistance to TNF-alpha-mediated graft-versus-leukaemia and to immune escape via upregulated S100A6. The anti-allergy drug amlexanox inhibits S100A6 expression in the presence of TNF-alpha; in KMT2A/AFF1 transgenic mice amlexanox enhanced tumour immunity and lowered the penetrance of leukaemia development, and in a NOD/SCID model of human KMT2A/AFF1 ALL it broadened graft-versus-leukaemia responses and extended survival.

   Amlexanox Downregulates S100A6 to Sensitize KMT2A/AFF1-Positive Acute Lymphoblastic Leukemia to TNFα Treatment. · Tamai H, Yamaguchi H, ... Inokuchi K · Cancer Research 2017
   PMID 28646023DOI
   ▸What this evidence cannot show

   Abstract read; full text not retrieved. Transgenic and xenograft mouse models from the same laboratory as E1; no human data. Amlexanox is separately a TBK1/IKK-epsilon inhibitor and that activity is not excluded as the operative mechanism here, so this must not be described as an IKK-complex intervention. The '60% of all infant ALL cases' figure is an abstract statement and was not checked against a cohort in this entry; other entries in this graph give roughly 50% for KMT2A::AFF1 in infant ALL, so treat the number as the paper's framing rather than as this graph's estimate.

   ▸Evidence that argues against this (1)

   • The 60% figure is series- and age-dependent and is quoted here without its denominator; infant series that use a different age cut report different proportions. The attribution of poor post-transplant outcome to TNF-alpha-mediated graft-versus-leukaemia resistance is the authors' framing of an association, not a measured mechanism. Kuo HP et al. 2013
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   In the study establishing IKK/NF-kB dependence in MLL leukaemia, the upstream stimulus used to demonstrate RELA phosphorylation and nuclear localisation in MLL leukaemia cell lines was LPS, not TNF; PANTHER analysis showed enrichment of the screen's candidate kinases and phosphatases in the Toll-like receptor signalling pathway, and Toll-like signalling transcripts that mediate p105-dependent NF-kB activation were enriched in MLL-AF1p and MLL-AF10 leukaemia cells relative to normal bone marrow.

   Epigenetic roles of MLL oncoproteins are dependent on NF-κB. · Kuo HP, Wang Z, Lee DF, Iwasaki M, Duque-Afonso J, Wong SHK, Lin CH, Figueroa ME, Su J, Lemischka IR, Cleary ML · Cancer Cell 2013
   PMID 24054986DOI
   ▸What this evidence cannot show

   The MLL-AF9 enrichment for the same transcript set did not achieve statistical significance and is reported as 'result not shown'. LPS at 10 ug/ml for 30 min after overnight serum starvation is a supraphysiological stimulus used to demonstrate pathway responsiveness, not evidence that TLR ligands drive the disease. Full text read this session (PMC3816582).

   ▸Evidence that argues against this (1)

   • This is the entry undercutting its own title: if the stimulus used was LPS and not TNF, then the TNF framing of this entry is inherited from pathway logic rather than from measurement in this disease. A PANTHER over-representation analysis names pathways, it does not identify the ligand actually operating. Wang X et al. 2012
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt reviewed human record P01375 (TNFA_HUMAN) has primary gene name TNF and length 233 aa; gene_exact:TNF AND organism_id:9606 AND reviewed:true returned exactly one row and gene_primary read back as TNF. S100A6 resolves to P06703 (S10A6_HUMAN), primary gene name S100A6, 90 aa.

   UniProtKB entry P01375 (TNFA_HUMAN), Tumor necrosis factor · UniProt Consortium · UniProt 2026
   ▸What this evidence cannot show

   Both resolved 2026-09-20 with gene_primary requested and all rows read. TNFRSF1A resolves to P19438 (455 aa) and IL1R1 to P14778 (569 aa); neither is given an entry because neither has KMT2A-specific evidence, and both are recorded as bounded absences.

   ▸Evidence that argues against this (1)

   • A reviewed database record fixes the canonical sequence and primary symbol on the date it was read, and nothing more. It does not establish that the protein is expressed in KMT2A-rearranged blasts, which isoform is present, or that the subunit acts in this disease. Accession and length are asserted so that a swap fails verify-accessions.py - a defence this graph needs because IKK-family and NF-kB subunit names collide heavily in the older literature - not because the record carries biological weight. Xiu Y et al. 2018
5. E_q95_tnf_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   TNF's topology makes a nuclear or subnuclear answer impossible, and the feature table is explicit: UniProt P01375 annotates a TRANSMEMBRANE helix at residues 36-56 described as 'Signal-anchor for type II membrane protein', a cytoplasmic topological domain at 1-35 and an EXTRACELLULAR topological domain at 57-233; the membrane form is annotated to the Cell membrane as a single-pass type II membrane protein on ECO:0000269, and the soluble form plus the two C-domain chains are annotated Secreted. So 197 of 233 residues face the extracellular space or the extracellular medium, and the only 35 residues on the cytosolic side sit on the inner face of a membrane with a signal-anchor tethering them there.

   UniProtKB P01375 (TNFA_HUMAN), SUBCELLULAR LOCATION comments for the membrane form, the soluble form and the C-domain chains with evidence codes and the topology feature table, fetched live 2026-09-22 · UniProt Consortium · Databases 2026
   ▸What this evidence cannot show

   A type II signal-anchor protein has no signal peptide and is inserted co-translationally with its N-terminus retained in the cytosol, so a cytosolic 35-residue tail genuinely exists - this claim is about the impossibility of a NUCLEAR pool for the mature ligand, not about the absence of any cytosolic sequence. The soluble form is generated by shedding, and the intramembrane processing that generates the cytoplasmic fragment is documented in activated dendritic cells by SPPL2a and SPPL2b.

   ▸Evidence that argues against this (1)

   • THE STRONGEST AVAILABLE OBJECTION IS THAT INTRAMEMBRANE PROTEOLYSIS RELEASES AN INTRACELLULAR FRAGMENT, AND IT DOES NOT RESCUE A NUCLEAR CLAIM. SPPL2a and SPPL2b promote intramembrane proteolysis of TNF-alpha in activated dendritic cells to trigger IL-12 production, so a cleaved N-terminal fragment does reach the cytosol and does have a signalling consequence - which is exactly the kind of result that in other genes (SREBF1 in this same tranche, NOTCH, APP) turns into a nuclear transcription-factor claim. It has not here: the fragment's documented function is to trigger IL-12 expression from the cytosolic side, no nuclear localisation of any TNF species is reported, and the fragment carries no DNA-binding or transactivation domain. NOT APPLICABLE is therefore a topology verdict with a named and rejected escape route, not an assumption. Friedmann E et al. 2006
6. E_q95_tnf_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Bounded 2026-09-22 and it confirms the topology verdict rather than merely failing to contradict it: against a 263,228-record gene clause, the queries that would surface a nuclear TNF pool return nothing usable. 'nuclear TNF' or 'intracellular TNF' returns 168 records and the title-checked first twelve are ALL immunology papers in which TNF is a secreted cytokine measured by intracellular cytokine staining - 'intracellular TNF' in this literature is a FLOW-CYTOMETRY METHOD, not a compartment. The twelve-term subnuclear clause returns 149 records out of 263,228, and intersected with the nine-term KMT2A/MLL fusion clause returns exactly 1, which is off topic. The membrane-form query ('transmembrane TNF', 'membrane TNF', 'membrane-bound TNF') returns 444, consistent with the annotated topology.

   NCBI E-utilities esearch db=pubmed; every count stored with the querytranslation read back, in data.localization_bounded_absence · NCBI E-utilities · Databases 2026
   ▸What this evidence cannot show

   Titles and abstracts only; supporting figures not searched (rule 80). At a gene-clause size of 263,228 no bounded absence over this symbol can be exhaustive, and the count is inflated by TNF's use as a stimulus and as a serum analyte throughout medicine - the same class of token problem the registry already records for ACTB as a loading control and PTH as an analyte. The verdict rests primarily on TOPOLOGY, with the literature bound as corroboration.

   ▸Evidence that argues against this (1)

   • AND THE ONE RECORD IN THE SUBNUCLEAR-PLUS-FUSION INTERSECTION DEMONSTRATED A RETRIEVAL TRAP WORTH RECORDING IN ITS OWN RIGHT. PubMed 38745328 came back from esummary with an EMPTY TITLE STRING, while efetch with itertext returns the full title - a neutrophil histone-post-translational-modification study in which TNF is a pre-activation stimulus and MLL1 is named as an H3K4me3 methyltransferase, containing no TNF compartment measurement. This is the brief's fourth-markup-trap family seen from the esummary side rather than the efetch side: an empty title passes any check that tests truthiness rather than content, and a citation block built from it would have had no title for verify-citations.py to compare. Every citation in this beat is built through a route that falls back to efetch itertext when esummary's title is empty, which is why this one was caught rather than stored blank. Piatek P et al. 2024
7. E_q95_tnf_loc3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   For TNF the cell-type question has two different answers and this graph needs the second one: the SOURCE cells are myeloid and lymphoid immune cells - the documented intramembrane processing is in ACTIVATED DENDRITIC CELLS, and UniProt records secretion regulated by SIRT6 - while the cell whose state matters in this disease is the TARGET, the KMT2A::AFF1 B-ALL blast, which is RESISTANT to TNF-mediated killing through S100A6 upregulation interfering with the p53-caspase-8-caspase-3 pathway, with siRNA against S100A6 restoring sensitivity and confirmation in MLL-AFF1 transgenic mice.

   Resistance of MLL-AFF1-positive acute lymphoblastic leukemia to tumor necrosis factor-alpha is mediated by S100A6 upregulation. · Tamai H, Miyake K, Yamaguchi H, Takatori M, Dan K, Inokuchi K, Shimada T · Blood Cancer J 2011
   PMID 22829076DOI
   ▸What this evidence cannot show

   The resistance result is this entry's E1 and is lymphoid and KMT2A::AFF1-specific, which is unusually well matched for this graph; the cell-type assignment recorded here is a reading of which cell each half of the biology belongs to, not a new measurement. Nothing in the TNF literature localises the protein in a blast, and for a secreted ligand that is the correct state of affairs rather than a gap.

   ▸Evidence that argues against this (1)

   • THE CONSEQUENCE FOR THIS TRANCHE'S OWN METHOD IS THAT A LOCALISATION FIELD CAN BE THE WRONG INSTRUMENT AND STILL LOOK LIKE A GAP, AND THAT IS WHY NOT APPLICABLE HAS TO BE A POSITIVE VERDICT RATHER THAN A BLANK. Nine entries in this series now carry it - HLA-A, HLA-B, NRAS, B2M, PROM1, CD33, ELOVL1, SEPTIN2 and FUT9 - and TNF makes ten. In every case a reporter counting missing localisation fields would list the entry as unmeasured, and in every case the honest answer is that the question does not apply, so recording NOT APPLICABLE keeps them out of the denominator of unmeasured things. The countervailing risk, which this entry accepts deliberately, is that NOT APPLICABLE can be used to close a question that a later result reopens: if a cleaved TNF fragment were ever shown to enter the nucleus, this verdict would be wrong. The reopening condition is therefore stated - a nuclear localisation of any TNF-derived species - rather than left implicit. Friedmann E et al. 2006

• related to →KMT2A-rearranged acute lymphoblastic leukaemia

  KMT2A::AFF1 B-ALL blasts are resistant to TNF-alpha killing via S100A6

• related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)

  TNF resistance is specific to MLL-AFF1-positive lines and transgenic mice

• supports →The NF-kB story in KMT2A-rearranged leukaemia is a myeloid story, and the lymphoid half is not a weaker version of it — it is a different subject

  the B-lineage NF-kB-adjacent literature is about evading a ligand, not depending on the pathway

• related to →The IKK complex — the unit the eviction experiment actually resolves, because the chemistry that produced the result cannot see the subunits

  canonical ligand input; but the MLL paper demonstrated pathway responsiveness with LPS, not TNF

• related to →The plasma membrane - the one compartment this graph targets with drugs and had never written down

  NOT APPLICABLE for a subnuclear answer, by TOPOLOGY: type II signal-anchor transmembrane cytokine, TM 36-56, cytoplasmic 1-35, extracellular 57-233, shed as a soluble secreted form. The one escape route was checked and rejected - SPPL2a/SPPL2b intramembrane proteolysis in activated dendritic cells releases a fragment that triggers IL-12 from the cytosolic side, with no nuclear localisation and no DNA-binding domain (Q95 2026-09-22)

• related to →Monocyte — the graph's most-discussed undefined cell identity, and the confound sitting under its own DepMap haem signal

  SOURCE versus TARGET, which must not be merged for a secreted ligand: TNF comes from myeloid and lymphoid immune cells (processing documented in ACTIVATED DENDRITIC CELLS, PMID 16829952), while the cell whose state matters here is the KMT2A::AFF1 blast, which RESISTS TNF killing via S100A6 (PMID 22829076). A localisation field on this entry would have been the wrong instrument (Q95 2026-09-22)