RSBN1 / KDM9 (Q5VWQ0) - an H4K20me2/me3 demethylase in the AFF4 proximity record, and the only read-queue partner that is a chromatin enzyme at all

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.

• If RSBN1 were nominated as a target on the strength of the glioma result, DepMap says the cells will not care: mean gene effect -0.0372 over 1,258 lines and 0.3% below -0.5. A viability-based programme would fail before it started.
• H4K20me1 is deposited by KMT5A/SET8 and read by cell-cycle and compaction machinery genome-wide. Inhibiting the eraser raises H4K20me2/me3 everywhere, not at a leukaemia-specific locus, so the essential-function toxicity risk is the whole-genome kind rather than the lineage kind.
• RSBN1 has a paralogue (RSBN1L). Any single-gene intervention risks paralogue compensation, which is the same failure this graph has documented for the KDM3 and 14-3-3 families.
• The one human mechanism is CCND1 activation via a bromodomain protein. In a KMT2A-rearranged cell, CCND1/cell-cycle activation is already a downstream consequence of the fusion, so an RSBN1 intervention would be acting on a node the fusion also drives - redundancy rather than a chokepoint.

1. What is the H4K20me1/me2/me3 profile of a KMT2A::AFF1 cell line (SEM, RS4;11, MV4-11) against a KMT2A-wild-type B-ALL line? No H4K20 measurement of any kind has been published in a KMT2A-rearranged cell, and this graph holds no H4K20 node at all - it is the cheapest unexplored mark in the corpus.
2. Does an AFF4-directed experiment ever recover RSBN1? Both IntAct rows have RSBN1 as the bait, so the reciprocal has never been done, and it is a single tagged-AFF4 pulldown.
3. Is the BRD2 interaction reported in glioma [E5] a BRD2-specific property or a bromodomain-family one? If RSBN1 binds BRD4, the protein acquires a direct route into the KMT2A elongation axis; if it does not, the glioma result stays local to that disease.
4. Does RSBN1L, the paralogue, cover for RSBN1 in DepMap? The flat gene-effect profile is uninterpretable without the double.
5. Is there an H4K20 measurement inside a supporting figure of an existing KMT2A paper? The zero recorded here bounds titles and abstracts only, and this project has already found one mark (H3K79me2) that existed only in a figure.

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

   The C. elegans dosage-compensation complex subunit DPY-21 is a Jumonji-family demethylase whose structure and activity define a subfamily that converts H4K20me2 to H4K20me1 in worms and mammals; selective inactivation of the demethylase activity eliminates H4K20me1 enrichment, elevates X-linked gene expression and reduces X chromosome compaction and TAD formation.

   Dynamic Control of X Chromosome Conformation and Repression by a Histone H4K20 Demethylase. · Brejc K, et al. · Cell 2017
   PMID 28867287DOI
   ▸What this evidence cannot show

   SYSTEM: Caenorhabditis elegans, dosage compensation, not a mammalian cell and not leukaemia. The mammalian claim is in the abstract's own wording ("in worms and mammals") and this beat read the abstract only, not the figures, so which mammalian protein was assayed and how is taken on the abstract's word and is the weakest link in the chain. Cited for the FAMILY assignment - that a Jumonji subfamily performs H4K20me2 to me1 - not for anything about RSBN1 in human cells.

   ▸Evidence that argues against this (1)

   • The same abstract's unexpected result is that DPY-21 also acts on autosomes of germ cells in a DCC-independent manner. An enzyme whose targeting is complex-dependent in one tissue and complex-independent in another cannot have its recruitment inferred from any single interaction partner, which is exactly what an AP-MS row invites a reader to do. Brejc K et al. 2017
2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   Mouse RSBN1 demethylates both dimethyl and trimethyl H4K20, converting each to monomethyl H4K20; H4K20me2 and H4K20me3 but not H4K20me1 are absent from RSBN1-positive germ cells; and RSBN1 is NOT a testis-specific gene and is expressed in elongated as well as round spermatids, contrary to the earlier report from which its name derives.

   Comparative distributions of RSBN1 and methylated histone H4 Lysine 20 in the mouse spermatogenesis. · Wang Y, et al. · PLoS One 2021
   PMID 34185806DOI
   ▸What this evidence cannot show

   SYSTEM: mouse spermatogenesis, immunostaining of seminiferous tubules plus a demethylase assay. Not human, not haematopoietic, not leukaemia, and no KMT2A or fusion anywhere in it. Read from the abstract only. The me3 activity is this paper's own extension of the prior me2 assignment and has not, as far as this beat's queries reach, been independently reproduced. This is the ONLY record in the 18-record `RSBN1[tiab]` set that is about the enzyme's function.

   ▸Evidence that argues against this (1)

   • This paper is itself a published correction of the gene's founding characterisation - it reports that RSBN1 is not testis-specific and not restricted to round spermatids, which is what the symbol 'round spermatid basic protein 1' asserts. A gene whose name encodes a refuted expression claim is one where any expression inference drawn from the symbol is unsafe, including the inference that a somatic-cell AP-MS row must be artefactual because the protein is a testis protein. Wang Y et al. 2021
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   IntAct holds exactly two rows placing RSBN1 (Q5VWQ0) against AFF4 (Q9UHB7): PMID 28514442 and PMID 33961781, both anti-tag co-immunoprecipitation (MI:0007), both in HEK293T, both at IntAct MI-score 0.53, and in both rows the experimental role of RSBN1 is bait and the experimental role of AFF4 is prey.

   IntAct PSICQUIC, MITAB2.7 rows for Q9UHB7 (AFF4), filtered on Q5VWQ0 · EMBL-EBI IntAct 2026
   ▸What this evidence cannot show

   Queried live 2026-09-22 and the per-row fields (experimental role, host organism, detection method, MI-score) were read out of MITAB2.7 columns 19/20, 29, 7 and 15 rather than assumed. THE TWO PUBLICATIONS ARE ONE SCREENING PROGRAMME: 28514442 (Nature 2017) and 33961781 (Cell 2021) are the 2017 and 2021 BioPlex releases from the same group, so 'two publications' is not two laboratories and must not be read as independent replication. HOST is HEK293T in both rows - an embryonic kidney line, not haematopoietic, not B-lineage, and not fusion-expressing. MI-score 0.53 is BELOW IntAct's own 0.6 conventional high-confidence cut. Rule 93 applies to the endpoint: PSICQUIC returns fewer rows than IntAct REST (AFF4: 179 versus 207), so this row count is a floor.

   ▸Evidence that argues against this (1)

   • A proteome-scale anti-tag coIP in HEK293T with the candidate as bait cannot distinguish a specific contact from co-purification of an abundant nuclear complex, and this graph's own audit of the AFF4 rows from the same programme concluded that a quartet of AFF4 partners from one screen collapsed on inspection. The direction of these rows also means no AFF4-directed experiment has ever recovered RSBN1. Huttlin EL et al. 2021
4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

   RSBN1 (Q5VWQ0) is present in the 960-protein seven-bait MLL-fusion AP-MS network of Skucha 2018, recovered by one of the seven fusion baits - MLL-EEN, combined p = 0.0756, spectral counts 5|7 - and it carries no value in the MLL-AF4 (= KMT2A::AFF1) column.

   MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity. · Skucha A, Ebner J, Grebien F, et al. · Nat Commun 2018
   PMID 29777171DOI
   ▸What this evidence cannot show

   Read out of the cached Supplementary Table 5 workbook (41467_2018_4329_MOESM4_ESM.xlsx) held in analyses/skucha-mll-fusion-interactome/cache/ by this beat on 2026-09-22, by matching the gene_name column to RSBN1 and printing every bait column. RSBN1 is one of exactly ONE of the sixteen read-queue and adjudicated-queue symbols present in that network - PES1, RPS11, YWHAG, AFG2A/SPATA5, AFG2B/SPATA5L1, MDK, ARRB2, DENND2D, FAM9A, FGF12, PLCD3, CAMKV, EPB41L1, BOP1, WDR12 and H3C1 are all absent from it. SYSTEM: Jurkat human T-ALL cells expressing Strep-HA-tagged single-copy doxycycline-inducible fusion transgenes, not endogenous protein and not B-lineage. THE SELECTION IS RANK-BASED, NOT THRESHOLDED: the paper retained the top 300 interactions per bait by combined p-value, so a listed p of 0.0756 means 'inside the top 300 for MLL-EEN', not 'significant'. One bait of seven, above 0.05, and ABSENT FROM THE INDEX FUSION'S COLUMN - this evidence item is cited as a weak independent corroboration that the protein is seen in MLL-fusion purifications at all, and explicitly NOT as evidence about KMT2A::AFF1.

   ▸Evidence that argues against this (1)

   • Absence from the MLL-AF4 column in this dataset is itself weak, because the paper's own MLL-AF4 bait is the shallowest of the seven by spectral counts, so an absent MLL-AF4 value can be a property of bait depth rather than of biology. The dataset has also never been curated into IntAct, so no independent curator has re-examined these calls, and the raw spectra at PXD009338 have not been reanalysed here. ProteomeXchange/PRIDE 2018
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   In human glioma, KDM9 (= RSBN1) is highly expressed in malignant tumours, mediates demethylation of H4K20me2 at the CCND1 promoter, interacts with BRD2 to facilitate CCND1 transcription and AKT signalling, and its knockdown suppresses xenograft growth and temozolomide resistance.

   Histone demethylase KDM9 activates the CCND1/AKT pathway to promote the malignant progression of gliomas through interaction with BRD2. · Cheng J, et al. · Cell Mol Life Sci 2025
   PMID 41291235DOI
   ▸What this evidence cannot show

   SYSTEM: human glioma cell lines and xenografts. Not leukaemia, not haematopoietic, no KMT2A fusion. Read from the abstract only; the promoter-level H4K20me2 claim, the BRD2 interaction and the knockdown phenotypes are taken from the abstract and not from the figures. This is the only report in the human literature that places the enzyme at a named promoter, and it is the only one that gives it a chromatin-protein partner - which makes it the load-bearing human reference for the claim that RSBN1 acts on chromatin in a somatic cancer cell, on a single paper.

   ▸Evidence that argues against this (1)

   • A single-paper mechanism in one tumour type, published in 2025 and not yet reproduced, is the weakest form of a chromatin-enzyme claim. The BRD2 link in particular is the sort of result this project has repeatedly found to be partner-imported: BRD2 is not the bromodomain protein in the KMT2A axis (BRD4 is), and no bromodomain protein of any kind appears in the two BioPlex rows that put RSBN1 next to AFF4. Cheng J et al. 2025
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The one PubMed record in which H4K20 and MLL appear together in title or abstract reports that mammalian ASH1L occupies the transcribed region of active genes with a distribution strongly resembling methylated H3K4 but NOT that of H3K9 or H4K20, that the ASH1L SET domain methylates H3K4 in vitro, and that ASH1L and MLL1 have highly similar distributions but are recruited independently of each other in MLL-null cells.

   Mammalian ASH1L is a histone methyltransferase that occupies the transcribed region of active genes. · Gregory GD, et al. · Mol Cell Biol 2007
   PMID 17923682DOI
   ▸What this evidence cannot show

   Read from the abstract only. Cited here as a NEGATIVE: it is the whole of the H4K20-and-MLL intersection in titles and abstracts, and its finding is that the trithorax-family distribution does not track H4K20. The system is mammalian cells including MLL-null lines; no fusion. Rule 80 applies - this bounds abstracts, and an H4K20 ChIP could sit in a supporting figure of a KMT2A paper without ever reaching the abstract, so the absence of an H4K20 measurement in KMT2A-rearranged cells is a TITLE-AND-ABSTRACT absence and no full-text or supplementary search was run for it.

   ▸Evidence that argues against this (1)

   • Drosophila Ash1 had been reported to methylate H3K4, H3K9 AND H4K20, and this paper is the mammalian refutation of the H4K20 part for ASH1L specifically. It therefore also records that the H4K20-trithorax question was once open and was closed in the negative for one enzyme only - it says nothing about KMT2A itself, whose own H4K20 relationship has never been measured in any published abstract. Gregory GD et al. 2007
7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PubMed esearch, run live 2026-09-22 with querytranslation and warninglist recorded for every count: RSBN1[tiab] = 18, KDM9[tiab] = 4, RSBN1[tiab] AND H4K20[tiab] = 1, KDM9[tiab] AND H4K20[tiab] = 0 (parsed zero), RSBN1[tiab] AND KMT2A[tiab] = 0, RSBN1[tiab] AND MLL[tiab] = 0, KDM9[tiab] AND MLL[tiab] = 0, RSBN1[tiab] AND leukemia[tiab] = 0, RSBN1[tiab] AND leukaemia[tiab] = 0, KDM9[tiab] AND leukemia[tiab] = 0 - and the RSBN1 and KDM9 record sets do not intersect.

   NCBI PubMed esearch, bounded-absence queries with querytranslation recorded (counted negatives) · NCBI E-utilities 2026
   ▸What this evidence cannot show

   Every zero here carries a NON-EMPTY querytranslation and a warninglist of {"phrasesignored": [], "quotedphrasesnotfound": [], "outputmessages": ["No items found."]}, which is this project's signature of a genuine zero rather than an instrument failure. Each operand is separately tagged - no `(A OR B)[tiab]` form was used anywhere, per rule 94. BOTH KMT2A SYMBOL GENERATIONS were paired with both gene symbol generations, and the MLL arm matters: `H4K20[tiab] AND MLL[tiab]` returns 1 where the KMT2A form returns 0, so the KMT2A-only query would have produced a false total absence. NAME LIST TRIED: RSBN1, KDM9, 'round spermatid basic protein', H4K20, H4K20me1, H4K20me2, 'H4 lysine 20'. NOT tried: RSB1, the mouse spelling Rsbn1, or RSBN1L (the paralogue). Bounded by TITLES AND ABSTRACTS ONLY. data/symbol-collisions.json (147 rows, generated 2026-09-22) was grepped for RSBN1, KDM9, 'round spermatid' and H4K20 BEFORE the first query, per rule 88, and has no row for any of them.

   ▸Evidence that argues against this (1)

   • The 18-record RSBN1 set was title-checked one by one and the contamination is structural, not incidental: RSBN1 sits at 1p13.2 immediately beside PTPN22, so it is named as a locus neighbour in type 1 diabetes, juvenile idiopathic arthritis, lupus, osteoporosis and pemphigus association papers without any statement about the protein. A count of the bare operand therefore overstates the relevant literature by roughly an order of magnitude, and this is the same failure mode the registry records for other GWAS-neighbour symbols. Smyth DJ et al. 2008
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   In DepMap release 26Q2 as served by Open Targets, RSBN1 (ENSG00000081019) has a mean gene effect of -0.0372 across 1,258 screens, a minimum of -0.831, 0.3% of lines below -0.5, and a mean of -0.0192 across the 57 leukaemia-annotated screens.

   Open Targets Platform GraphQL, target.depMapEssentiality (DepMap 26Q2) · Open Targets Platform / DepMap 2026
   ▸What this evidence cannot show

   RELEASE RECORDED, per rule 83: the same API call returned meta.dataVersion = {year: 26, month: 06}, i.e. DepMap 26Q2, fetched 2026-09-22. The screen count 1,258 is this gene's own, not a corpus denominator. depMapEssentiality is a LIST OF ONE BLOCK PER TISSUE and all 29 blocks were concatenated before averaging - reading block [0] returns about 2% of the screens and is a registered trap in data/symbol-collisions.json. The Ensembl id was resolved through the HGNC primary ensembl_gene_id, per rule 78, and the readback control passed: Open Targets returned approvedSymbol RSBN1 for it. The leukaemia subset is defined by a substring match on diseaseFromSource ('leukemia', 'leukaemia', 'lymphoblastic', 'myeloid') and is NOT restricted to KMT2A-rearranged lines - no KMT2A-rearrangement stratification was performed here, so this number cannot speak to fusion selectivity in either direction.

   ▸Evidence that argues against this (1)

   • A flat DepMap profile is the expected result for a histone demethylase with a paralogue: RSBN1 has a close relative, RSBN1L, and the 14-3-3 and KDM families in this graph both show that paralogue redundancy hides single-gene phenotypes. An absence of dependency is therefore not evidence that the enzyme is dispensable, and the correct experiment is a mark readout, not a viability screen. Open Targets Platform / DepMap 2026

• co-occurs with →AFF4 — the textbook SEC scaffold that KMT2A::AFF1 cells need LESS than other leukemias

  two BioPlex rows (28514442, 33961781) - one screening programme seen twice; RSBN1 is the BAIT and AFF4 the prey in both; HEK293T; IntAct MI-score 0.53

• regulates →H4K20me2 - the abundant repair-linked H4 mark, the 53BP1 docking site, and the first H4K20 state in this graph; never once measured in a KMT2A-rearranged cell

  erases H4K20me2 (and H4K20me3) to H4K20me1 - Jumonji/DPY-21 subfamily demethylase

• binds →BRD2 - a bounded absence: zero KMT2A papers name it, it is inhibited by every BET compound used in this disease, and it is the paralogue most likely to carry the toxicity

  reported in human glioma only: KDM9-BRD2 interaction driving CCND1 transcription

• co-occurs with →MLLT1 / ENL — the reader that beat its own paralogue, with a degrader already in hand

  Skucha 2018 seven-bait MLL-fusion AP-MS: recovered by MLL-EEN only, p = 0.0756, 5|7 spectra, and ABSENT from the MLL-AF4 column

• ← supportsIn all 24 read-queue rows inspected, the candidate is the BAIT and this project's protein is the prey - nobody has ever tagged KMT2A, AFF4, menin or PSIP1 and found any of them; plus seven bounded rejections

  both RSBN1 rows read per row: candidate bait, AFF4 prey, HEK293T, MI-score 0.53, and the two publications are one BioPlex programme

• ← related toH4K20me1 - the transcription- and replication-linked H4 state, whose gene-body level tracks Pol II; it HAS been profiled genome-wide in human haematopoietic stem cells twice, and a query scoped to 'leukemia' cannot see either dataset

  RSBN1/KDM9 demethylates H4K20me2 and me3 TO me1, so the eraser of the higher states is a producer of this one. CARRIED from mark-h4k20me2 and gene-rsbn1, not independently verified in this beat