Oncofusion

start / graph / gene-rbm15b

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q6-interactome-rna-layer

RBM15B — the bridge from the DOT1L mark to the RNA layer: it reads H3K79me2 and steers m6A to start codons

also called OTT3 · RBM15-like · RNA binding motif protein 15B

RBM15B reads the same chemical mark on chromatin that DOT1L writes, and uses it to decide which messenger RNAs get tagged near their start signal so they are translated more. It is the first direct mechanical link found between this disease's central chromatin mark and its RNA machinery.

The same thing, for a scientist

RBM15B modulates m6A deposition in 5'UTRs and around start codons, and that site selectivity is **guided by H3K79me2** - the histone mark this graph's entire DOT1L axis is built on. Residue H47 of RBM15B is identified as key for H3K79me2 recognition; the H3K79me2-RBM15B axis enhances translation efficiency of oncogenic transcripts and promotes leukaemic stem-cell self-renewal and leukaemia maintenance, and blocking the axis inhibits leukaemia cell survival, promotes differentiation and impairs haematological malignancy in MLL-rearranged leukaemia [E1]. DepMap gives RBM15B a Chronos mean of -0.041 with 0.2% of 1,208 lines depleted and a **positive** index-arm difference (+0.058 / +0.059), the 86.3rd null percentile [E2] - a complete disagreement between a published mechanism and the dependency data, in a gene whose reported phenotype is self-renewal.

Go deeper
Why this closes a gap the graph has had all along

This graph holds a large DOT1L axis: H3K79me2 spreading at fusion target loci, pinometostat, the aberrant elongation programme. It also now holds an RNA layer. Until this paper there was no mechanism connecting them, and the two were parallel stories about the same disease.

RBM15B connects them in one step: a histone mark on chromatin determines where on the transcript m6A is deposited, and therefore which messages are translated efficiently 1. That makes the RNA layer a readout of the chromatin layer rather than an alternative to it, and it makes an obvious prediction - DOT1L inhibition should redistribute m6A, not merely change transcription. No published work tests that, and it is cheap: m6A-seq before and after pinometostat in an MLL-rearranged line.

It also makes the entry's own weakness sharp. If the axis is real and it matters, DOT1L inhibitors and m6A-pathway inhibitors are acting on the same mechanism at two points, and combining them would be expected to be redundant rather than additive.

The disagreement with DepMap, stated without resolving it

RBM15B's DepMap row is as close to nothing as this beat found: mean -0.041, 0.2% of 1,208 lines depleted, and a positive difference in both halves of the KMT2A-rearranged index arm 2.

Four readings, and this entry does not choose.

  1. Assay mismatch. The reported phenotype is leukaemic stem-cell self-renewal and leukaemia maintenance 1, which a two-to-three-week proliferation dropout cannot see. This is the same argument gene-igf2bp3 makes and it is the most likely explanation.
  2. Paralogue buffering. RBM15 and RBM15B are paralogues, and RBM15 has a real dependency row in the same scan (mean -0.267, 18.9% of lines depleted) as well as being independently named as an AML dependency with KMT2A-rearranged bias 3. If RBM15 covers for RBM15B, single-gene CRISPR shows nothing.
  3. The paper is very recent and has not been independently reproduced.
  4. The effect is real and small in culture.

What this entry should not do is quote the DepMap row as refutation. A 0.2% depletion rate in a gene whose claimed function is self-renewal is uninformative, not contradictory.

Speckles, the nuclear envelope, and an explicit nucleolar exclusion (Q87, 2026-09-21)

Subcellular. Nucleus, and NONSHUTTLING (PMID 16129689) - a stronger statement than 'nuclear', and a slightly awkward one for a protein that binds NXF1 and Aly/REF. HPA v25.1 IF, reliability Enhanced: nucleoplasm.

Subnuclear, three statements including a negative. The splicing factor compartment (nuclear speckles) and the nuclear envelope (PMID 19586903), granular nucleoplasmic staining, and EXCLUDED FROM THE NUCLEOLI. The exclusion is a real datum: it rules out the compartment a reader would otherwise assume for an RNA-binding protein.

The unreconciled axis. This entry's headline is that RBM15B reads H3K79me2 and steers m6A - a chromatin claim - and no localisation study places RBM15B at chromatin. The compartment literature was written by people studying mRNA export. Compare AQR in this same tranche, where BOTH the speckle and the chromatin addresses were measured.

Cell type and state. Neither measured. The only state-adjacent fact is that RBM15 knockdown is embryonically lethal in mouse and RBM15B cannot compensate.

Shown for fusion. No, in the strong sense. Both localisation papers publish under the name OTT3, which a current-symbol-only query would have missed entirely.

Facts and figures
gene symbolRBM15B
rna functionm6A adaptor directing site-selective methylation to 5'UTRs and start-codon regions; reported to read H3K79me2 via residue H47
chromatin linkH3K79me2 (the DOT1L mark) guides RBM15B-directed m6A deposition
paralogueRBM15 (Chronos mean -0.267, 18.9% of lines depleted)
uniprotQ8NDT2
length AA890
uniprot resolutiongene_exact:RBM15B AND organism_id:9606 AND reviewed:true; one hit; primary gene name read back as RBM15B; length 890 aa checked. Its paralogue RBM15 is Q96T37 at 977 aa, resolved the same way.
subnuclear localizationNUCLEAR SPECKLES (the splicing factor compartment) and the NUCLEAR ENVELOPE, with a granular nucleoplasmic staining pattern and EXPLICIT EXCLUSION FROM THE NUCLEOLI (UniProt Q8NDT2; PMID 19586903 for speckles and envelope, PMID 16129689 for the nonshuttling nuclear character). The envelope address goes with binding to the mRNA export factors NXF1 and Aly/REF through the C-terminal region. Note that this entry's H3K79me2-reader claim predicts a chromatin address that no localisation study reports, and the two pictures have never been tested together.
cell stateNone published. No cell-cycle, differentiation or stress-dependent relocalisation of RBM15B has been reported. The one state-adjacent fact is comparative rather than about RBM15B itself: genetic knockdown of RBM15 in mouse is embryonically lethal and OTT3/RBM15B cannot compensate, so the two paralogues have distinct biological roles despite sharing the speckle and envelope addresses (PMID 19586903). Recorded as an absence rather than left blank.
shown for fusionNO - nobody has looked in any fusion. Measured 2026-09-21: (RBM15B[tiab] OR OTT3[tiab]) AND the eight-term fusion clause AND the ten-term localisation clause returns 0 records. The retired/alternative name OTT3 was included, and it is the name under which both localisation papers publish - a current-symbol-only query would have missed the entire compartment literature for this gene.
pubmed counts
(MLL OR KMT2A) AND RBM15B1.000(MLL OR KMT2A) AND RBM15B: 1H3K79me2 AND m6A2.000H3K79me2 AND m6A: 2RBM15B AND (leukemia OR l…3.000RBM15B AND (leukemia OR leukaemia): 3
essentiality
{
  "depmap_common_essential": false,
  "chronos_mean": -0.0414,
  "pct_lines_depleted": 0.2,
  "n_lines": 1208,
  "kmt2a_r_lymphoid_diff": 0.0581,
  "kmt2a_r_myeloid_diff": 0.0589,
  "null_percentile": 86.27
}
pubmed counts — the numbers behind the chart
{
  "RBM15B AND (leukemia OR leukaemia)": 3,
  "(MLL OR KMT2A) AND RBM15B": 1,
  "H3K79me2 AND m6A": 2
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.0286,
  "chronos_mean_other_leukaemia": -0.0872,
  "difference": 0.0586,
  "pan_essentiality_all_lines": -0.0414,
  "selectivity_rank": 15361,
  "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."
}
localization
{
  "subcellular": "NUCLEUS, and NON-SHUTTLING. Transfected OTT3 (RBM15B) is described as a nonshuttling nuclear protein (PMID 16129689), which is a stronger statement than 'nuclear' - it says the protein does not cycle through the cytoplasm, which for an mRNA-export cofactor is a substantive and slightly surprising result. UniProt Q8NDT2 lists Nucleus/nucleoplasm (ECO:0000269 PubMed:16129689), Nucleus speckle and Nucleus envelope (both ECO:0000269 PubMed:19586903), with the note 'Localized in the nucleoplasm with a granular staining pattern and excluded from the nucleoli'. HPA v25.1 IF, reliability Enhanced, gives a single main location, Nucleoplasm.",
  "subnuclear": "THREE resolvable statements, two positive and one NEGATIVE, and the negative is the one most often lost. (1) THE SPLICING FACTOR COMPARTMENT (nuclear speckles): OTT3/RBM15B shares with RBM15 the association with the splicing factor compartment (PMID 19586903). (2) THE NUCLEAR ENVELOPE: it shares that association too, and binds the mRNA export factors NXF1 and Aly/REF through its C-terminal region - an export-competent address that fits the envelope localisation. (3) EXCLUDED FROM THE NUCLEOLI, with a granular nucleoplasmic staining pattern. A recorded exclusion is a real subnuclear datum and is worth as much as a positive assignment, because it rules out the compartment a reader might otherwise assume for an RNA-binding protein.",
  "the_h3k79me2_reader_claim_is_a_separate_axis": "This entry's headline - that RBM15B reads H3K79me2 and steers m6A - is a chromatin-binding claim, and it would predict a CHROMATIN address that none of the localisation literature above reports. Nobody has imaged RBM15B at chromatin, and the speckle/envelope/nucleoplasm picture was built by researchers studying mRNA export, not chromatin reading. The two pictures have never been reconciled in one experiment.",
  "how_shown": "immunofluorescence of transfected OTT3 with EB2 co-localisation and heterokaryon-style shuttling assessment (PMID 16129689); subcellular localisation studies plus NXF1/Aly-REF interaction mapping (PMID 19586903); HPA antibody IF",
  "node": "localization-nuclear-speckles-kmt2a-r"
}
cell type
{
  "haematopoietic_answer": "NONE MEASURED for localisation. RBM15B's compartment work is in transfected human epithelial-class lines in an Epstein-Barr virus mRNA-export context, with no haematopoietic system anywhere. HPA v25.1 records low RNA tissue specificity. The haematopoietic relevance of this gene in the graph comes from its paralogue relationship with RBM15 - the RBM15-MKL1 fusion of acute megakaryoblastic leukaemia - and from the m6A axis, neither of which is a localisation result.",
  "expression_atlas": "HPA v25.1: RNA tissue specificity Low; IF reliability Enhanced (nucleoplasm)."
}
How this could bite us back
  • If H3K79me2 steers m6A, then DOT1L inhibitors already act partly through the RNA layer, and a separate RNA-layer drug may buy nothing on top.
  • A 0.2% depletion rate across 1,208 lines is the weakest dependency signal in this beat. Building a target argument on it requires the assay-mismatch explanation to be true, and that explanation is being invoked repeatedly in this beat without independent support.
  • RBM15 and RBM15B are constantly conflated in reviews. A programme aimed at 'RBM15' could be pursuing the wrong paralogue for the mechanism it cites.
Still open
  1. Does DOT1L inhibition with pinometostat redistribute m6A in an MLL-rearranged line? The mechanism predicts it and nobody has run m6A-seq before and after.
  2. Is the H3K79me2-RBM15B axis present in KMT2A::AFF1 B-lymphoid cells, or only in the myeloid models the MLL-r literature is built from?
  3. Do RBM15 and RBM15B buffer each other, which would explain RBM15B's null DepMap row and make the double knockout the necessary experiment?
  4. If chromatin and RNA layers are one mechanism at two points, is a DOT1L inhibitor plus an m6A inhibitor redundant rather than additive?
Every source behind this entry (6)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    RBM15B modulates m6A modification in 5'UTRs and around start codons of transcribed mRNAs, guided by H3K79me2 histone methylation, with residue H47 of RBM15B key for H3K79me2 recognition; the H3K79me2-RBM15B axis enhances translation efficiency of oncogenic transcripts and promotes leukaemic stem cell self-renewal and leukaemia maintenance, and blockade of the axis inhibits leukaemia cell survival, promotes differentiation and impairs haematological malignancies in MLL-rearranged leukaemia.

    PMID 41629530DOI
    What this evidence cannot show

    Abstract read; full text not read. Published 2026 and not independently reproduced. 'MLL-r leukemia' in the title is not lineage-specified in the abstract, so whether any B-lymphoid KMT2A::AFF1 work is included cannot be determined without the full text. The H47 residue assignment and the H3K79me2 recognition are structural claims that need the paper's own data to assess.

    Evidence that argues against this (2)
    • DepMap gives RBM15B a Chronos mean of -0.041 with 0.2% of 1,208 lines depleted and a positive difference in the KMT2A-rearranged index arm in both lineage halves. A gene reported to maintain leukaemia produces no dependency signal anywhere in a 1,208-line screen, which needs the assay-mismatch or paralogue-buffering explanation to stand. Oncofusion knowledge graph (internal analysis) 2026
    • A histone mark guiding a nuclear RNA-binding protein to a transcript is a co-localisation-based mechanism, and co-localisation between chromatin and nascent RNA is exactly the setting in which the graph's condensate caveat applies: spatial clustering and direct recognition are hard to separate, and the standard assays do not distinguish them. Oncofusion knowledge graph (internal analysis) 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    DepMap CRISPR (Chronos) gives RBM15B a mean gene effect of -0.041 across 1,208 lines with 0.2% depleted below -0.5, and index-arm differences of +0.058 (four B-lymphoid) and +0.059 (seven myeloid) against 43 other leukaemia lines, at the 86.3rd percentile of a 466-gene random null. Its paralogue RBM15 is -0.267 with 18.9% of lines depleted and index-arm differences of -0.036 / -0.115.

    What this evidence cannot show

    A near-zero row in a gene with a self-renewal phenotype is uninformative rather than contradictory; the screen's readout cannot measure what the paper reports. The paralogue comparison is suggestive of buffering and does not establish it.

    Evidence that argues against this (1)
    • The same paralogue-buffering explanation is being used here and for the IGF2BP family, and it is the standard rescue for a disappointing screen result. Used twice in one beat it should be read as a pattern in the beat's reasoning as much as in the biology. Oncofusion knowledge graph (internal analysis) 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A domain-focused CRISPR screen across 341 RNA-associated proteins identified the m6A reader RBM15, alongside PTBP1, as an AML dependency with a bias toward the KMT2A-rearranged subtype.

    PMID 42637878DOI
    What this evidence cannot show

    Abstract read; full text not read. AML. RBM15 is named in the abstract as a hit but only PTBP1 is described as validated in detail, so RBM15's evidence within that paper is weaker than PTBP1's.

    Evidence that argues against this (1)
    • This is RBM15, not RBM15B. The two are paralogues and are routinely conflated in secondary literature; the H3K79me2 mechanism is reported for RBM15B and the CRISPR-screen dependency for RBM15, and no published work shows both properties in the same protein. Chen TQ et al. 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB Q8NDT2 (RB15B_HUMAN) is human RBM15B, 890 amino acids, and its paralogue RBM15 is Q96T37 at 977 amino acids; both resolved by gene_exact query with the primary gene name read back and the lengths checked.

    What this evidence cannot show

    Two separate single-hit queries with matching primary gene names. Recorded together because the two paralogues are routinely conflated and the beat cites a different paper for each.

    Evidence that argues against this (1)
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    RBM15B/OTT3 shares with its paralogue RBM15 the association with the SPLICING FACTOR COMPARTMENT and with the NUCLEAR ENVELOPE, and binds the mRNA export factors NXF1 and Aly/REF directly through its C-terminal region; OTT3 and RBM15 also interact with each other in vivo. UniProt Q8NDT2 records the resulting terms - Nucleus speckle and Nucleus envelope on ECO:0000269 PubMed:19586903 - together with the note that the protein is localised in the nucleoplasm with a granular staining pattern and EXCLUDED FROM THE NUCLEOLI.

    The RNA-binding motif protein 15B (RBM15B/OTT3) acts as cofactor of the nuclear export receptor NXF1. · Uranishi H, Zolotukhin AS, Lindtner S, Warming S, Zhang GM, Bear J, et al. · J Biol Chem 2009
    PMID 19586903DOI
    What this evidence cannot show

    The work is on transfected protein in human non-haematopoietic lines and the comparison is with RBM15 throughout, so several statements are 'shares with RBM15' rather than measured independently for RBM15B. 'Splicing factor compartment' is the authors' phrase for what is conventionally the nuclear speckle, and the paper does not resolve the speckle against adjacent bodies. Nothing haematopoietic; nothing about a fusion.

    Evidence that argues against this (1)
    • The subnuclear picture recorded here was built entirely by people studying mRNA EXPORT, and it does not contain the address this graph's own headline for RBM15B requires. If RBM15B reads H3K79me2 and steers m6A deposition at fusion-target genes, it must occupy CHROMATIN, and no localisation study reports it there - the reported addresses are speckles, the nuclear envelope and a granular nucleoplasm, with the nucleoli explicitly excluded. A protein can of course do both, and the sibling case in this tranche shows exactly that: AQR is a nuclear speckle protein AND occupies promoter-proximal chromatin in an AFF4-ELL2-EAF1 assembly. But for AQR both addresses were measured; for RBM15B the chromatin one is inferred from a mark-reading claim and has never been imaged, which is a weaker position than the entry's headline implies. Xiang Y et al. 2026
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Transfected OTT3/RBM15B is a NONSHUTTLING nuclear protein - it does not cycle between nucleus and cytoplasm - and is ubiquitously expressed as a 120-kDa protein; it co-localises in the nucleus with the co-transfected Epstein-Barr virus mRNA export factor EB2. UniProt Q8NDT2's Nucleus/nucleoplasm term rests on this work (ECO:0000269 PubMed:16129689).

    PMID 16129689DOI
    What this evidence cannot show

    Transfected, overexpressed protein in a virology system, and the nonshuttling call is made in the context of comparing OTT3 with a viral export factor rather than as a dedicated shuttling study. The paper's own conclusion about OTT3 is mainly negative - it finds no evidence for a role in transcriptional regulation - which is the opposite emphasis from the chromatin-reader framing this gene carries in this graph.

    Evidence that argues against this (1)
    • A nonshuttling nuclear protein is a genuinely awkward fact for the export story the same literature tells, and the tension should be recorded rather than smoothed over: RBM15B binds NXF1 and Aly/REF and is associated with the nuclear envelope, which is what an export adaptor looks like, yet it is reported not to accompany its cargo out. The reconciliation would be that RBM15B hands transcripts off at the envelope rather than escorting them, but no experiment in either paper tests that - and the same pair of papers is the entire evidential basis for both the speckle and the envelope assignments, so this is not two independent lines of evidence converging. Uranishi H et al. 2009
Connected to