KDM2B — 513 potent ChEMBL rows that are mostly patent bins, and no structure of the domain they claim to inhibit

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.

• Chemistry that looks solved and is not: an automated ligandability score would rank KDM2B third in this interactome on 513 potent rows. Those rows are censored patent bins, the real literature point is 630 nM, and there is no co-crystal of the target domain. A programme started on that basis would discover the problem after spending the money.
• Selectivity is the class problem: 2-oxoglutarate-mimetic inhibitors act at a site conserved across the Jumonji demethylases and across other 2-OG-dependent dioxygenases. Off-target demethylase and dioxygenase inhibition would arrive with the on-target effect, and nothing in the data reports it.
• Essential-function toxicity: mis_z 5.52 and LOEUF 0.240, with 26 observed loss-of-function variants against 150 expected, make KDM2B one of the most constrained genes in this dossier. Missense constraint is the relevant kind for a catalytic-site inhibitor, and the reassuring flat DepMap profile comes from immortalised lines over three weeks.
• Compensation: PRC1.1 can be assembled with BCORL1 in place of BCOR (5JH5 versus 8HCU), and KDM2A is a close paralogue of KDM2B. Both routes let the complex survive a single-agent perturbation, which is consistent with the flat knockout data and predicts a flat drug response too.
• No therapeutic index to find: mean Chronos -0.040 with 0.6% of lines depleted and a selectivity of -0.020 means there is no measured gain to trade against the constraint risk. This is the one node in the batch where the recommendation is not 'needs a better experiment' but 'the numbers argue against starting'.

1. Is there any real, measured, selective KDM2B inhibitor? The 513 potent rows collapse to one ~630 nM literature point once censored patent bins are removed, and with no co-crystal of the KDM2B catalytic site there is no published basis for a selectivity claim against other Jumonji demethylases. A cross-KDM panel on CHEMBL3786952 would be the first honest datapoint.
2. Does KDM2B contact a KMT2A fusion protein at all, directly or through PRC1.1? Nothing verified this cycle shows that it does. Every link in this node runs through BCOR, and BCOR's fusion-binding segment is separable from its PRC1.1 domain, so the whole rationale could be an inherited association.
3. If the KDM2B-BCOR-PRC1.1 tether is phase-separation-dependent and calcium-modulated (PMID 39256555), what does 'inhibition' even mean quantitatively? Condensate behaviour has thresholds, so an isolated-domain affinity may mispredict nuclear occupancy by orders of magnitude and dose-response may be non-linear. No affinity has been published for any interaction in the complex.
4. Why is KDM2B under mis_z 5.52 missense constraint — the highest in this dossier — while whole-gene knockout does essentially nothing to 1,208 cancer cell lines? A catalytic-site inhibitor is a missense-like perturbation, not a truncation, so the constraint figure may be the more relevant predictor of drug toxicity than the knockout is of efficacy.
5. Does catalytic activity matter at all here, or only the C-terminal PRC1.1 platform? All four verified structures are of the platform and none of the active site. A catalytically dead knock-in would separate the two functions and would tell a programme which end of the protein to work on — and if the answer is the platform, the entire 2-OG-mimetic chemical literature is aimed at the wrong domain.

1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   KDM2B is not DepMap common-essential; mean Chronos across 1,208 lines is -0.040 with 0.6% of lines depleted, the leukemia mean is -0.107, the 10 KMT2A-rearranged lines average -0.127 and the 4 KMT2A::AFF1 lines -0.155, giving selectivity_vs_leukemia -0.020 and selectivity_vs_all_lines -0.087; all 10 KMT2A-rearranged lines are negative but none reaches the -0.5 depletion threshold (range -0.323 at SEM to -0.023); gnomAD gives pLI 1.0, LOEUF 0.240, 26 observed versus 150 expected loss-of-function variants and mis_z 5.52; HPA classifies KDM2B as low tissue specificity, detected in all tissues, low immune cell specificity, tissue-enriched at protein level, nucleoplasmic.

   DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API; loss-of-function constraint via the gnomAD GraphQL API; normal-tissue expression via the Human Protein Atlas API · Broad Institute DepMap; gnomAD; Human Protein Atlas; Open Targets Platform · Databases 2026
   ▸What this evidence cannot show

   Means were computed by us from per-line API values rather than read from a DepMap release table, and the API does not state the release version. Only 10 KMT2A-rearranged lines (4 KMT2A::AFF1) contribute; a selectivity of -0.020 is an order of magnitude inside the n=10 noise band of 0.15 and carries no information. That all ten lines share a negative sign leaves open a real but very small general-leukemia requirement, which is not a therapeutic signal. Whole-gene knockout removes the demethylase activity and the PRC1.1 assembly platform together and cannot separate catalytic from scaffolding function. mis_z 5.52 is the highest missense-constraint value in this dossier but measures selection in humans over evolutionary time, which is a different question from cancer-cell proliferation over three weeks.

   ▸Evidence that argues against this (1)

   • gnomAD constraint is severe and points the other way: LOEUF 0.240 with 26 observed against 150 expected loss-of-function variants, plus mis_z 5.52. A gene under that much missense constraint is doing something a three-week proliferation assay does not measure, and a catalytic-site inhibitor is a missense-like perturbation rather than a truncation. gnomAD Consortium 2026
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   All four verified KDM2B structures cover the C-terminal PRC1.1 assembly region or a zinc-finger region and none covers the catalytic demethylase domain: 5JH5 (X-ray 2.55 A) contains KDM2B 1059-1336 with PCGF1 150-255, SKP1 and BCORL1 1668-1785; 8HCU (X-ray 2.20 A) contains KDM2B 1103-1336 with PCGF1 151-255 and BCOR 1607-1748; 6BVA (X-ray 2.66 A) contains KDM2B 1060-1104 with SKP1 and a ubiquitin variant; 4O64 (X-ray 2.13 A) contains KDM2B zinc fingers 607-723.

   RCSB PDB entries 5JH5, 8HCU, 6BVA and 4O64, with per-entity UniProt accessions and modelled residue ranges resolved against the RCSB search and data APIs · RCSB Protein Data Bank · RCSB PDB 2026
   ▸What this evidence cannot show

   Identifiers, resolutions and residue ranges were resolved against the RCSB API in this session; ranges are RCSB's UniProt-mapped modelled segments rather than curated domain boundaries. The union of modelled residues is 607-723 plus 1059-1336 out of 1,336, so roughly half the protein including the catalytic region is structurally uncharacterised in KDM2B itself, and we deliberately do not supply catalytic-domain boundaries from memory. 5JH5 contains BCORL1 and not BCOR, a distinction routinely lost in secondary literature. None of these entries carries a published affinity in the abstracts verified.

   ▸Evidence that argues against this (1)

   • Counting PDB entries measures how tractable a protein was to CRYSTALLISE, which correlates with being small, ordered and well-behaved rather than with being druggable. What decides design value is whether the deposited density covers the site that matters, at a resolution that supports modelling, in the relevant state — and this graph has repeatedly found structures described in the direction that overstates what is resolved. None of the four covering the catalytic demethylase domain is the finding: the enzymatic activity that gives the protein its name has no experimental structure here, so any catalytic-site strategy starts from a model. RCSB PDB 2026
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   ChEMBL v34 target CHEMBL3779760 (Lysine-specific demethylase 2B) carries 513 activity rows at pChEMBL >= 6, but the nominal best values are exact IC50 ties at 100 nM across CHEMBL3786803, CHEMBL3787664 and CHEMBL3786952 that are censored patent bins rather than measurements; the cleanest literature point is CHEMBL3786952 at IC50 630 nM; the chemotype is generic 2-oxoglutarate-mimetic KDM chemistry with known cross-family selectivity problems; and there is no curated mechanism of action.

   ChEMBL database version 34, target CHEMBL3779760 (Lysine-specific demethylase 2B): target_search, get_bioactivity and get_mechanism queries · EMBL-EBI ChEMBL · ChEMBL 2026
   ▸What this evidence cannot show

   This is the strongest ligandability trap in the interactome: exact ties at a round number across unrelated compounds indicate a patent reporting activity as a bin edge, so 513 potent rows must not be read as 513 potent compounds. Counts and 'best' values are additionally page-local under a 25-row API limit, so they reflect one sampled page likely dominated by a single patent family, not a potency-ranked view of the whole target. 2-OG-mimetic compounds act at an active site conserved across the Jumonji family and more broadly across 2-OG-dependent dioxygenases, and ChEMBL rows do not carry cross-KDM selectivity panels. There is no KDM2B co-crystal of the catalytic site to support structure-based selectivity claims.

   ▸Evidence that argues against this (1)

   • A ChEMBL activity-row count measures how much medicinal chemistry has been DONE, not how tractable the target is, and the best reported affinity belongs to the most optimised compound in whichever series someone chose to pursue. Zero rows means nobody tried; thousands of rows means a target had a decade of industrial attention. Exact IC50 ties at 100 nM across three compounds are the signature of an assay floor or a reporting convention rather than of three genuinely equipotent ligands, so the 'best value' is not a measurement. ChEMBL, EMBL-EBI 2026
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   KDM2B recruitment of the Polycomb group complex PRC1.1 requires cooperation between PCGF1 and BCORL1, with KDM2B's C-terminal region serving as the assembly platform.

   KDM2B recruitment of the Polycomb group complex, PRC1.1, requires cooperation between PCGF1 and BCORL1 · Wong SJ, Gearhart MD, Taylor AB, et al. · Structure 2016
   PMID 27568929DOI
   ▸What this evidence cannot show

   The partner here is BCORL1, not BCOR; the equivalent assembly with BCOR is 8HCU from a different study. The work does not involve KMT2A fusions and says nothing about the AF9/ENL AHD. No affinity measurement is reported in the abstract, so no Kd is attributed to the KDM2B-PCGF1 or KDM2B-BCORL1 interactions.

   ▸Evidence that argues against this (1)

   • This is the BCORL1 result again, and BCORL1 is a paralogue of BCOR. The KDM2B-PRC1.1 assembly evidence therefore runs through a protein that is not the one this graph connects to the fusion. RCSB PDB 2026
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   Calcium modulates the tethering of the BCOR-PRC1.1 enzymatic core to KDM2B via liquid-liquid phase separation, so the KDM2B-BCOR connection behaves as a regulated condensate-dependent tether rather than a fixed stoichiometric complex.

   Calcium modulates the tethering of BCOR-PRC1.1 enzymatic core to KDM2B via liquid-liquid phase separation · Chen R, Zhang Y, Wang X, et al. · Communications Biology 2024
   PMID 39256555DOI
   ▸What this evidence cannot show

   This is a calcium and phase-separation study, not a crystallography paper; it is cited for the phase-separation mechanism and for contributing the 8HCU structure, and the primary structural citation for PRC1.1 assembly remains PMID 27568929. Phase-separation results are assay-sensitive and in vitro condensate behaviour does not straightforwardly transfer to a nucleus. Confidence is moderate for that reason. If the mechanism holds, affinities measured on isolated domains will mispredict nuclear occupancy and an inhibitor's dose-response may be non-linear, which is a real problem for any programme on this interface.

   ▸Evidence that argues against this (1)

   • Calcium-modulated tethering via liquid-liquid phase separation is exactly the claim this project's standing rule constrains: puncta, FRAP recovery and aliphatic-alcohol sensitivity cannot distinguish phase separation from clustered binding to a template, and only half-bleach internal mixing does. A regulated condensate-dependent tether is a possible reading of the data, not the established one. eLife 2019
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The connection between KDM2B/PRC1.1 and KMT2A-fusion biology runs entirely through BCOR, whose AF9/ENL AHD-binding segment (1176-1207, in 6B7G) is a different region of the protein from its PRC1.1 PUFD (1634-1748), and whose requirement for MLL-AF9 leukemia is itself contested.

   BCOR binding to MLL-AF9 is essential for leukemia via altered EYA1, SIX, and MYC activity · Schmidt CR, Achille NJ, Kuntimaddi A, et al. · Blood Cancer Discovery 2020
   PMID 32954361DOI
   ▸What this evidence cannot show

   This paper concerns BCOR and MLL-AF9; it does not study KDM2B, and nothing verified this cycle shows KDM2B contacting a KMT2A fusion protein at all. Because BCOR's AF9-binding segment and its PUFD are separable regions, PRC1.1 membership and fusion binding are separable functions, and a KDM2B-directed drug would not be expected to break the fusion contact. The paper's own positive claim is not supported by DepMap (BCOR mean Chronos +0.025) and is contradicted on the CBX8 arm by PMID 22094252.

   ▸Evidence that argues against this (1)

   • If the whole connection runs through BCOR, and BCOR's AHD-binding segment is a different region from its PUFD, then KDM2B's relevance to the fusion is two proteins and two interfaces away. The entry states the chain and should also state its length. Nature 2012
7. E_q77_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   Human JHDM1B/KDM2B is a NUCLEOLAR protein: it preferentially binds the transcribed region of ribosomal DNA and represses rRNA gene transcription, and the repression requires the JmjC domain and specific demethylation of H3K4me3 in the nucleolus; a JmjC-dependent negative effect on cell size and proliferation follows.

   JHDM1B/FBXL10 is a nucleolar protein that represses transcription of ribosomal RNA genes. · Frescas D, Guardavaccaro D, Bassermann F, Koyama-Nasu R, Pagano M · Nature 2007
   PMID 17994099DOI
   ▸What this evidence cannot show

   Rule 91 check performed and PASSED: the paper UniProt cites at ECO:0000269 for 'Nucleus, nucleolus' has the annotated fact in its title, and it is the founding observation rather than a later paper that happened to contain an experiment. Limits: human cultured cells, no haematopoietic material, and the nucleolar localisation is of endogenous and tagged JHDM1B in interphase - no measurement in a leukaemic blast and none under a menin inhibitor. The search-recall point matters for anyone re-deriving this: the paper publishes under JHDM1B and FBXL10 and is unreachable from a KDM2B-only query.

   ▸Evidence that argues against this (1)

   • A SECOND EXPERIMENTALLY CODED SUBNUCLEAR STATEMENT IN THE SAME UNIPROT BLOCK POINTS SOMEWHERE ELSE, and the two are not reconciled anywhere. The 'Nucleus' and 'Chromosome' annotations rest on a DNA-damage paper in which KDM2B is at double-strand breaks, not at rDNA. Both can be true - constitutive nucleolar, inducibly recruited to damage - but nothing published measures the partition between them, so a reader who takes 'KDM2B is a nucleolar protein' as the whole answer will be surprised by a damage experiment. Jiang Y et al. 2015
8. E_q77_loc2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The 'Nucleus' and 'Chromosome' annotations on UniProt Q8NHM5 both rest on a single paper whose subject is METABOLIC control of DNA repair, not KDM2B localisation: Jiang et al. 2015 show that locally generated fumarate promotes DNA repair by inhibiting KDM2B-mediated histone H3 demethylation at damage sites.

   Local generation of fumarate promotes DNA repair through inhibition of histone H3 demethylation. · Jiang Y, Qian X, Shen J, Wang Y, Li X, Liu R, Xia Y, Chen Q, Peng G, Lin SY, Lu Z · Nat Cell Biol 2015
   PMID 26237645DOI
   ▸What this evidence cannot show

   Rule 91 FIRES, in the mild form. The evidence code is genuine and the paper does place KDM2B on chromatin, but it is a fumarate/DNA-repair study and its KDM2B chromatin association is a supporting element rather than the subject; it is not a general localisation study and it is not the founding observation for KDM2B being nuclear. Recorded as a condition-scoped statement - chromatin at induced breaks - and deliberately NOT generalised to 'KDM2B is on chromatin'. Human cancer lines, no haematopoietic material.

   ▸Evidence that argues against this (1)

   • Taking this annotation at face value would displace the better one. If a reader filtered Q8NHM5 to its ECO:0000269 rows and kept the most recent, they would get 'nucleus and chromosome, from a fumarate paper' and lose the nucleolar/rDNA result that actually characterises this protein - the same failure mode rule 91 documents for KMT2A, where filtering on the evidence code keeps the least useful line. Frescas D et al. 2007
9. E_q77_ct1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In MICE, transgenic over-expression of Fbxl10/Kdm2b in HAEMATOPOIETIC STEM CELLS causes myeloid or B-lymphoid leukaemia with COMPLETE PENETRANCE; the transgenic HSCs show an accelerated G0/G1-to-S transition with normal G0-to-G1 entry and pleiotropic progenitor expansion, raised Nsg2, marked enrichment of mitochondrial oxidative-phosphorylation genes and increased cellular ATP, and ChIP-seq places Fbxl10 directly on the regulatory regions of Nsg2 and of the OXPHOS genes.

   Fbxl10 overexpression in murine hematopoietic stem cells induces leukemia involving metabolic activation and upregulation of Nsg2. · Ueda T, Nagamachi A, Takubo K, Yamasaki N, Matsui H, Kanai A, Nakata Y, Ikeda K, Konuma T, Oda H, Wolff L, Honda Z, Wu X, Helin K, Iwama A, Suda T, Inaba T, Honda H · Blood 2015
   PMID 25872778DOI
   ▸What this evidence cannot show

   Murine, transgenic OVER-EXPRESSION, and normal haematopoiesis with no fusion present - so it grades what excess KDM2B can do, not what endogenous KDM2B is required for. Forced expression of a chromatin protein in HSCs is a design that produces leukaemia for many genes, and the specific mechanism claimed (Nsg2, OXPHOS) rests on transcriptional enrichment plus a bulk ATP measurement rather than on flux. The paper is reachable only under Fbxl10/Jhdm1b, not KDM2B.

   ▸Evidence that argues against this (1)

   • The mouse genetics goes the other way when the endogenous gene is removed rather than added, and specifically when the CxxC DOMAIN is removed: Kdm2b-CxxC-deficient haematopoietic cells give lethal, largely NOTCH1-dependent T-ALL, with KDM2B acting as a tumour suppressor. Over-expression-causes-leukaemia and deletion-causes-leukaemia in the same tissue is not a contradiction in the data - it is a statement that the informative variable is dose and lineage, and that neither arm licenses a directional claim about this gene in KMT2A-rearranged disease. Isshiki Y et al. 2019
10. E_q77_ct2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In MICE lacking the zinc-finger CxxC DNA-binding domain of KDM2B in haematopoietic cells, animals develop lethal T-cell acute lymphoblastic leukaemia mostly in a NOTCH1-dependent manner, phenocopying Bcor insufficiency; thymocyte ChIP-seq places KDM2B at promoters where BCOR and EZH2 colocalise, and KDM2B target genes overlap markedly with NOTCH1 targets in human T-ALL cells, so non-canonical PRC1.1 appears to antagonise NOTCH1-mediated gene activation.

    KDM2B in polycomb repressive complex 1.1 functions as a tumor suppressor in the initiation of T-cell leukemogenesis. · Isshiki Y, Nakajima-Takagi Y, Oshima M, Aoyama K, Rizk M, Kurosawa S, Saraya A, Kondo T, Sakaida E, Nakaseko C, Yokote K, Koseki H, Iwama A · Blood Adv 2019
    PMID 31471323DOI
    ▸What this evidence cannot show

    Murine, T-lineage, no fusion. The deleted element is the CxxC domain rather than the whole gene, which is the right experiment for the CpG-targeting question and means the phenotype does not bound what catalytic-domain loss would do. The NOTCH1 target overlap is computed in HUMAN T-ALL cells against MURINE thymocyte ChIP, a cross-species join that the paper makes explicitly and that should not be read as a single-system result. T-ALL is not this project's index disease.

    ▸Evidence that argues against this (1)

    • The tumour-suppressor reading has to live beside a complete-penetrance oncogene result from over-expressing the same gene in HSCs in the same species, and beside a flat CRISPR profile in 1,208 human cancer lines in which KDM2B is depleted in 0.6% of them. Three instruments, three answers. What none of them supports is a directional therapeutic claim about KDM2B in KMT2A-rearranged leukaemia, which is the claim a reader of this entry is most likely to want to make. Ueda T et al. 2015

• ← part ofPCGF1 — the PRC1.1 scaffold whose LOSS causes menin-inhibitor resistance, so it is a restraint on the leukaemia rather than a dependency of it

  PCGF1 aa 150-255 is required for interaction with the KDM2B-SKP1 heterodimer within PRC1.1

• ← bindsBCORL1 — the BCOR paralogue that is in the only crystal structure of the PRC1.1 core, and is the one PRC1.1 subunit nobody has tested in KMT2A-rearranged leukaemia

  composite BCORL1-PCGF1 surface contacts KDM2B; unique to PCGF1-containing PRC1.1

• ← bindsSKP1 — a pan-essential housekeeping adaptor that is also a structural subunit of PRC1.1, which makes it the one PRC1.1 member that can never be a target

  SKP1 F-box-binding region (aa 104-163) docks the KDM2B F-box; the same motif that would build an SCF ligase

• ← part ofPRC1.1 — the complex this graph has been leaning on, now enumerated subunit by subunit, and the enumeration changes the conclusion

  ZF-CxxC DNA-binding subunit targeting non-methylated CpG islands

• part of →Polycomb in KMT2A-r leukemia — core PRC2 required, EZH2 and BMI1 dispensable, and CBX8 acting outside its repressive role
• related to →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it

  placed in the interactome beat, but its graph-relevant claim is chemical-matter quality against the HOXA program

• ← related toH3K9me3 - measured in KMT2A::AFF1 cells, in a supplementary figure, and it rises GENERALLY where H3K27ac falls SPECIFICALLY in the same experiment

  proposed H3K79me2/3 demethylase whose action recruits SIRT1 and gains H3K9me3 - untested in KMT2A-r

• ← related toRe-bound at full text: still no publication puts KDM2B's rDNA repression together with a KMT2A fusion's nucleolar residence - 27 records, 10 of them congress abstract books, the rest reviews - but KDM2B ChIP-seq in MLL-AF9 cells is now deposited, so the rDNA re-map has a second and better dataset

  KDM2B ChIP-seq now exists in MLL-AF9 cells (GSE260742) but was never read at rDNA

• ← related toKDM2B, the PRC1.1 CpG-island recruiter, sits at the 99.7th percentile of its own distribution on the fusion's strongest sites in DMSO and does not gain share under either a menin inhibitor or a DOT1L inhibitor - it holds position while comparable bins drift down

  KDM2B is at the 99.685th percentile of its own distribution at the fusion's top sites in DMSO and its share does not move under either drug

• ← depends onFour matched input tracks, used for the two things an input is actually for: no chromosome-scale copy-number difference between the CRISPR clone and its control (max 1.0988, min 0.9470 in both processing states), and the knockout's H2AK119ub depletion from CpG islands is REAL because its own genomic DNA is enriched there, not depleted

  the CxxC subunit whose unmethylated-CpG targeting the test assumes

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

  KDM2B's measured constitutive subnuclear address is the nucleolus at rDNA (Frescas 2007) - the compartment where this graph's KMT2A-fusion coverage is thinnest

• related to →The CXXC domain (aa 1147-1195, PDB 2KKF) is a chromatin address, not a compartment address — CpG mutants lose transformation but keep their subnuclear position

  same chemistry as the KMT2A CxxC address: KDM2B reads unmethylated CpG, and its CxxC-domain deletion is the informative murine genetics

• expressed in →Bone marrow haematopoietic stem cell — the postnatal comparator, and why the same fusion behaves differently here

  murine HSC transgenic over-expression gives fully penetrant myeloid or B-lymphoid leukaemia; CxxC-domain deletion in haematopoietic cells gives T-ALL

• related to →The OXPHOS-dependent state — established in KMT2A-rearranged AML, and inverted in KMT2A-rearranged B-ALL where OXPHOS inhibition makes leukaemia-initiating cells appear

  Fbxl10-transgenic HSCs enrich mitochondrial OXPHOS genes with raised ATP, and KDM2B ChIP sits on those regulatory regions

• related to →The stem-like self-renewal state — a limited stem-cell module grafted onto a committed cell

  accelerated G0/G1-to-S with normal quiescence exit, producing pleiotropic progenitor expansion in murine HSCs

• ← part ofThe human CXXC/CpG-reading family is exactly twelve proteins, and TET2 is not one of them

  KDM2B is one of the twelve.

• ← related toKDM2A - the CXXC reader that antagonises KMT2A leukaemogenesis, and displaces wild-type MLL rather than the fusion

  Paralogue with the same domain architecture - JmjC, CXXC, PHD, F-box - and a different complex (PRC1.1 versus SCF).

• ← related toThree 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

  KDM2B's constitutive nucleolar rDNA address, put next to the fusion's nucleolar address for the first time.

• ← testsDegrade one CXXC reader, ChIP the fusion: the one experiment that settles whether the address is contested

  Third reader: the PRC1.1 targeting subunit, and the one carrying the nucleolar rDNA address.

• ← related toThe measurement everybody said was missing exists: in SEM cells KMT2A::AFF1 occupies 20% of unmethylated CpG sites, CXXC1 50% and KDM2B 89%, and only the fusion tracks menin (r2 = 0.96)

  KDM2B occupies 89% of unmethylated CpG sites in SEM, the widest of the three readers measured, and does not spread.

• ← inhibitsIDH2 - the hypermethylating co-mutation that IS enriched in KMT2A-rearranged AML (16.0% vs 10.4%), at p = 0.0004 and not the p < 0.0001 the abstract claims

  2-HG inhibits KDM histone demethylases as well as TET, so a 2-HG-high cell may impair two of the fusion's competing CXXC readers at the same time as it raises 5mC.

• ← inhibits2-hydroxyglutarate competitively inhibits the alpha-KG dioxygenases - and it hits the TET demethylases AND the KDM histone demethylases, which is why its net effect on a CXXC reader is not predictable

  CLASS INFERENCE ONLY, not measured: KDM2B is an alpha-KG-dependent JmjC demethylase and occupies 89% of unmethylated CpG sites in SEM.

• ← testsPut mutant IDH2 into SEM and ask three questions at once: does the unmethylated-CpG pool shrink, does the fusion move, and does its competitor move

  Reads out the competitor as well as the fusion, because 2-HG inhibits JmjC demethylases and KDM2B is on 89% of unmethylated CpG.

• ← related toHypothesis: menin inhibition does not clear the fusion from chromatin but redistributes it onto the four fifths of the unmethylated CpG address it normally ignores

  KDM2B occupies 89% of uCpG sites, the upper bound on the available address

• ← related toHalf the PRC1.1 intersection was free all along: Zhou's locus set is a publisher supplementary table, and 159 of this graph's gene entries are in it

  SPLIT FROM ITS OWN COMPLEX: KDM2B is menin-specific here while PCGF1, BCOR, BCORL1 and RNF2 are H2AK119ub-only.