STAG2 - the cohesin subunit that defines secondary AML, and the one whose loss a KMT2A fusion does not care about

{
  "chronos_mean_kmt2a_rearranged": -0.9556,
  "chronos_mean_other_leukaemia": -0.5987,
  "difference": -0.3569,
  "pan_essentiality_all_lines": -0.33,
  "selectivity_rank": 60,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}

{
  "subcellular": "NUCLEUS AND CHROMOSOME, AND THE ENTIRE RECORD IS UNSOURCED. UniProt Q8N3U4 fetched 2026-09-21 annotates 'Nucleus', 'Chromosome' and 'Chromosome, centromere' with an EMPTY EVIDENCE ARRAY on every line, plus an UNCODED note describing the prophase pathway, separase cleavage of RAD21 at anaphase and the meiotic cohesin switch. There is no ECO:0000269, no ECO:0000305 and no ECO:0000250 anywhere in the block [E_q77_loc1]. STAG2 and HOXA10 are the only two of the sixteen proteins in this tranche whose subcellular record carries no evidence code of any kind - and STAG2 is the more surprising of the two, because it is the most recurrently mutated cohesin subunit in myeloid disease.",
  "subnuclear": "CENTROMERE, SPECIFICALLY AND BY CONTRAST WITH ITS PARALOGUE - and this is a genuine subnuclear answer, so the dispatch premise that subnuclear localisation is the wrong question for cohesin does not hold for STAG2 either. Canudas and Smith showed that cohesin(SA1) and cohesin(SA2) are DIFFERENTIALLY required for TELOMERE and CENTROMERE cohesion respectively: SA1-deficient cells cannot establish or maintain cohesion between sister TELOMERES after replication, while in SA2-depleted cells telomere cohesion is normal and CENTROMERE cohesion is prematurely lost [E_q77_loc2]. So the two mutually exclusive SA subunits partition the genome by SUBNUCLEAR REGION, not merely by gene set - which is a stronger statement than 'cohesin is at loop anchors' and is the half of the compartment question the prose section of this entry did not have. UniProt Q8N3U4 does annotate 'Chromosome, centromere', but WITH NO EVIDENCE CODE AT ALL [E_q77_loc1], so the database carries the right answer with no provenance and the provenance is in a 2009 paper it does not cite.",
  "how_shown": "nothing, for the database lines [E_q77_loc1]; siRNA depletion of SA1 versus SA2 with telomere- and centromere-cohesion readouts and chromosome-morphology analysis [E_q77_loc2]; biochemical characterisation of vertebrate cohesin complexes with prophase chromatin-dissociation assays [E_q77_loc3]",
  "systems": "human cultured cells (HeLa and related) for both experimental references. NO haematopoietic localisation system, no primary blast, nothing KMT2A-rearranged - and no measurement in the secondary-AML setting where STAG2 mutation is >95% specific.",
  "shown_for_fusion_localisation": "NO - nobody has looked in any fusion. STAG2 protein has never been imaged or fractionated in a KMT2A-rearranged cell, an MLL-AF9 mouse or any other fusion system. The entry's own [E4] records the sharper version of this gap at the genomic level: in the only patient-level break map this graph holds, 517 RAD21 and 256 SMC3 sites were assayed and STAG2 SITES WERE NOT ASSAYED AT ALL.",
  "node": "localization-chromocentres-pericentromeric-kmt2a-r",
  "what_this_adds_to_the_prose_section_above": "The 'Cell type, cell state, localisation, epigenetics' section of this entry already says 'Subnuclear: at CTCF-bound loop anchors'. That remains true and is a statement about genomic occupancy. What is added here is a REGIONAL statement with a matched paralogue control: SA2 is required for CENTROMERIC cohesion and SA1 for TELOMERIC cohesion, so the two mutually exclusive rings occupy different parts of the nucleus and are not spares for one another. That strengthens the entry's existing point that STAG2 and STAG1 define distinct cohesin populations, and it gives it an experimental citation.",
  "a_partial_correction_to_my_dispatch": "My dispatch brief said of RAD21 and STAG2 that they are 'a chromatin-architecture story rather than a compartment one' and invited a bounded rejection if subnuclear localisation were the wrong question. IT IS NOT THE WRONG QUESTION FOR EITHER OF THEM. RAD21 has three coded subnuclear addresses including the nuclear matrix; STAG2 has a paralogue-discriminating centromeric one. What the premise got RIGHT is narrower and worth keeping: neither subunit has a nuclear-BODY address, and neither has been measured in a leukaemic cell, so the architecture framing is the right one for what the DATA can support - it is just not a reason to decline the question."
}

{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "identifier_note": "STAG2 is Q8N3U4, 1,231 aa, primary gene name STAG2, synonym SA2 - one clean reviewed row. data/symbol-collisions.json was grepped before the first query: there is NO STAG2 row, and THE STAG1 ROW ALREADY EXISTED and is the relevant one - gene_exact:STAG1 returns PMEPA1 first because PMEPA1 was once Solid Tumor-Associated Gene 1. This entry's own prose already documents that trap; the registry row and the entry agree, and no new row is proposed. Retrieval note for anyone re-deriving the localisation: the centromere-versus-telomere result is published under SA1 and SA2, not STAG1 and STAG2, and a current-symbol query does not reach it.",
  "expressed_in": {
    "class": 1,
    "summary": "ubiquitous. DepMap pan-essentiality is modest at -0.33, so unlike RAD21 and the SMCs this subunit is not broadly required - consistent with STAG1 covering for it in most cells."
  },
  "required_in": {
    "class": 2,
    "claim": "IN MICE, STAG2 DELETION IN HAEMATOPOIETIC STEM AND PROGENITOR CELLS INCREASES SELF-RENEWAL AND IMPAIRS DIFFERENTIATION, AND THE LINEAGE IT IMPAIRS IS THE B-CELL ONE. Stag2 loss alone decreased chromatin accessibility and transcription of LINEAGE-SPECIFICATION GENES INCLUDING Ebf1 AND Pax5, leading to increased self-renewal and REDUCED HSPC COMMITMENT TO THE B-CELL LINEAGE. ChIP sequencing showed that although Stag2 and Stag1 bind a shared set of loci, a component of Stag2 binding sites is UNOCCUPIED BY Stag1 even in Stag2-deficient HSPCs - so the paralogue does not simply substitute - while CONCURRENT loss of both abrogated haematopoiesis entirely [E_q77_ct1].",
    "pmids": [
      "31495782"
    ],
    "system_named_in_the_claim": "murine conditional Stag2 deletion in normal haematopoietic stem and progenitor cells - no fusion present",
    "nodes": [
      "cell-type-hsc-bone-marrow",
      "cell-type-pro-b"
    ],
    "why_this_is_the_most_relevant_cohesin_result_in_the_graph": "EBF1 and PAX5 are the two transcription factors that commit a progenitor to the B lineage, and closing their chromatin is a B-lineage-commitment defect in a fusion-free mouse. This project's disease is B-lineage, arises in a progenitor, and is characterised by a differentiation block. A cohesin subunit whose loss closes Ebf1 and Pax5 is therefore reaching the same endpoint by a purely architectural route, and the sibling RAD21 entry's lymphoid-selective reconstitution defect is the same signal from a different subunit."
  },
  "context": "murine normal haematopoiesis; no human material and no fusion"
}

{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "maintains_state": {
    "class": 3,
    "states": [
      "cell-state-stem-like-self-renewal",
      "cell-state-differentiation-block"
    ],
    "claim": "IN MICE WITHOUT A FUSION, STAG2 LOSS INSTALLS EXACTLY THE TWO STATES THIS GRAPH ATTRIBUTES TO THE KMT2A FUSION - increased self-renewal and impaired differentiation - and it does so by CLOSING CHROMATIN at lineage-specification genes rather than by opening an oncogenic programme [E_q77_ct1]. That is a mechanistically different route to the same phenotype, and it is the strongest available support for this entry's existing convergence argument: if cohesin loss and MLL-AF9 both produce abnormal self-renewal, a cell that has one gains little from the other.",
    "pmids": [
      "31495782"
    ],
    "system_named_in_the_claim": "murine conditional Stag2 deletion in normal HSPCs - no fusion, no leukaemia"
  },
  "the_direct_test_this_makes_available": "This entry already proposes the sharpest experiment in the beat - if the convergence is real, cohesin-mutant NON-KMT2A AML should be menin-inhibitor sensitive. The Stag2 HSPC result supplies the missing middle term: it names the mechanism as closed chromatin at lineage-specification loci, which is a MEASURABLE intermediate. So the experiment can be run as a mechanism test and not only as a drug-response test - does menin inhibition reopen Ebf1 and Pax5 in Stag2-deleted HSPCs?",
  "the_live_contradiction_inside_this_entry": "THIS ENTRY NOW CONTAINS TWO INSTRUMENTS THAT DISAGREE AND THE DISAGREEMENT SHOULD BE STATED RATHER THAN AVERAGED. The project's own genome-wide DepMap scan ranks STAG2 SIXTIETH of 17,728 for KMT2A selectivity - Chronos -0.9556 in the KMT2A-rearranged arm against -0.5987 in other leukaemia lines, difference -0.3569 against a pan-essentiality of only -0.33 - which is one of the most selective results in the whole scan and says KMT2A-rearranged lines need STAG2 MORE. The murine preprint at [E3] says complete Stag2 loss is WELL TOLERATED in MLL-AF9-expressing HSPCs with minimal effect on self-renewal. Three ways to reconcile them, none tested: the preprint is murine, myeloid and MLL-AF9 while the scan is human lines including KMT2A::AFF1; in-vitro self-renewal over days is not proliferation over three weeks of CRISPR selection; and STAG1 dosage differs between the systems, which matters because the Viny result shows STAG1 does not fully substitute. Until one of those is resolved, this graph should not quote either figure alone.",
  "confidence": "moderate",
  "caveats": "The cell-type and cell-state evidence is a single murine study in normal haematopoiesis with no fusion. Conditional deletion is not a drug. The B-lineage commitment defect is read from chromatin accessibility and transcription at Ebf1 and Pax5 plus reduced commitment, not from a full B-cell developmental staging. And the entry's existing scope statement on [E3] stands: that result is a PREPRINT, murine, MLL-AF9 rather than KMT2A::AFF1, myeloid rather than lymphoid, with an in-vitro self-renewal readout.",
  "counter_evidence": [
    {
      "note": "The paralogue is the standing counter-argument to every STAG2 loss-of-function claim, and the Viny data cut both ways on it. FOR redundancy: concurrent loss of Stag2 and Stag1 abrogated haematopoiesis, so STAG1 is carrying essential function when STAG2 is gone, and any STAG2-directed intervention in a STAG1-proficient cell is partial by construction. AGAINST redundancy: a component of Stag2 binding sites is UNOCCUPIED by Stag1 even in Stag2-deficient HSPCs, so substitution is incomplete at the locus level. The sibling RAD21 entry adds a third perspective - cohesin silencing across subunits gives expansion rather than death - which means a reader cannot infer STAG2's contribution from complex-level data.",
      "citation": {
        "title": "Cohesin Subunit RAD21 Regulates the Differentiation and Self-Renewal of Hematopoietic Stem and Progenitor Cells.",
        "authors": "Kumar P, Zhang N, Lee J, Cheng H, Kurtz K, Conneely SE, Sasidharan R, Rau RE, Pati D",
        "year": 2023,
        "venue": "Stem Cells",
        "pmid": "37534584",
        "doi": "10.1093/stmcls/sxad058",
        "type": "paper"
      }
    }
  ]
}