PRMT3 - the mostly-cytoplasmic ribosome methyltransferase that also writes the ACTIVATING mark on the residue PRMT5 represses, has an acute-leukaemia degrader, and is invisible to this project's own mark-query rule

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.

• The modality that works is degradation, and degradation of a protein whose knockout does nothing in 1,258 DepMap lines should be read cautiously: the PROTAC recruits MDM2, so part of any phenotype may be p53-axis engagement rather than PRMT3 loss, and the abstract reports no control separating them.
• Paralogue compensation is severe here and runs through a better-evidenced enzyme: PRMT1 is the dominant H4R3me2a writer with 41 records against PRMT3's 6, is near-pan-essential, and already has its own failed clinical programme in this graph. Any H4R3me2a-based rationale for PRMT3 has to explain why PRMT1 does not simply cover it.
• The mark's direction is contested at the residue level: PRMT3 writes the activating asymmetric mark on H4R3 and PRMT5 writes the repressive symmetric mark on the same residue. An intervention on one writer shifts the balance towards the other, and an antibody or mass-spectrometry readout that cannot separate me2a from me2s cannot tell which way it moved.
• Lineage and system transfer is the weakest part: the only locus-specific histone experiment is a microRNA promoter in osteogenic mesenchymal stem cells, and the only leukaemia experiment names no cell line, no subtype and no genotype. Nothing here is B-lineage, infant, or fusion-driven.
• Cytoplasmic residence caps the whole chromatin argument: PRMT3 is predominantly cytoplasmic with a nuclear pool that depends on an uncharacterized partner. If ZNF200 is low in a blast, a PRMT3-directed agent is acting on a ribosomal and retinoic-acid-signalling enzyme, not on chromatin - which is a different toxicity profile and a different mechanism.
• An assay hazard that this entry is the worked example of: a bounded absence for PRMT3's mark written in this project's standard residue-plus-wildcard form returns ZERO, while the compact spelling returns six real records. Any absence recorded for an arginine mark in one query form alone is unsafe.

1. Is there any nuclear PRMT3 in a leukaemic blast, and does it depend on ZNF200? The nuclear pool is reported to require a ZNF200 interaction that also protects PRMT3 from degradation, and ZNF200 has never been measured in a haematopoietic cell - so the prior question to every chromatin question about PRMT3 here is whether the protein gets into the nucleus at all.
2. Which mark wins on H4R3 in a KMT2A-rearranged cell: the activating asymmetric me2a or the repressive symmetric me2s? PRMT1 and PRMT3 write one, PRMT5 writes the other, PRMT5's product falls when the PAF complex is removed from MLL-AF9 cells, and no measurement resolving the two states at this residue exists in this disease.
3. Does the PRMT3 degrader's effect on acute leukaemia cells survive an MDM2-only control? The warhead recruits MDM2, the readouts include MYC, E2F and apoptosis, and a PRMT3 knockout does nothing across 1,258 DepMap lines - so the simplest explanation available is that some of the phenotype is not PRMT3.
4. Is the relevant PRMT3 function catalytic at all? The degrader beats the inhibitor, the paper says small molecules have struggled to lower PRMT3-deposited ADMA, and PRMT3 inhibits ALDH1A1 without using its methyltransferase activity. If the answer is no, this entry's epigenetic readout is a description rather than a mechanism.
5. Does EPB41L3 silencing raise PRMT3 activity in leukaemia? EPB41L3 inhibits PRMT3 and its own promoter is hypermethylated in many cancers, which would make PRMT3 de-repressed by an epigenetic event on another gene - a two-step mechanism nobody has looked for in a blood cell.
6. Is PRMT3 nucleolar? Its dominant substrate is a small-subunit ribosomal protein, this graph holds a nucleolar layer including localization-nucleolus-kmt2a-fusion and a nucleolar-block folder, and no subnuclear assignment for PRMT3 exists in any source.

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

   PRMT3 (O60678, 531 aa) is a type I protein-arginine N-methyltransferase catalysing monomethylation and ASYMMETRIC dimethylation, whose characterised substrate is the Arg/Gly-rich region of small ribosomal subunit protein uS5/RPS2 at multiple sites, which also CONTRIBUTES TO methylation of histone H4 Arg-3 - a specific tag for epigenetic transcriptional ACTIVATION - and which inhibits ALDH1A1 aldehyde dehydrogenase activity by direct interaction independently of its methyltransferase activity; it carries a C2H2-type zinc finger at 48-71 annotated as responsible for substrate specificity and a SAM-dependent MTase PRMT-type domain at 217-531, it is located in Cytoplasm/cytosol and Nucleus with experimental evidence for both, and it interacts with EPB41L3 via that protein's FERM domain in an interaction that is direct and INHIBITS PRMT3's methyltransferase activity.

   UniProtKB O60678 (ANM3_HUMAN), Protein arginine N-methyltransferase 3 - FUNCTION, CATALYTIC ACTIVITY, SUBUNIT, SUBCELLULAR LOCATION, DOMAIN and feature table, fetched 2026-09-22 · UniProt Consortium 2026
   ▸What this evidence cannot show

   A database record, not a measurement. The wording on the histone activity is deliberately weaker than for the ribosomal one - 'Contributes to methylation of histone H4 Arg-3' against 'Catalyzes the asymmetric arginine dimethylation at multiple sites in the Arg/Gly-rich region of small ribosomal subunit protein uS5/RPS2' - and this entry preserves that asymmetry rather than flattening it. The accession is asserted explicitly and not confirmed by length: PRMT3 at 531 aa is far from PRMT1 (371) and PRMT6 (375), but the family as a whole defeats length checks and the habit should not be relaxed for one member.

   ▸Evidence that argues against this (1)

   • The record's own weighting argues that PRMT3 is not a chromatin enzyme in any leading sense, and this graph has already taken that position in print: gene-epb41l3 states that 'PRMT3 is predominantly cytoplasmic and its principal characterised substrate is a ribosomal protein, not a histone', and uses that to argue that the EPB41L3-PRMT3 contact is coherent as a CYTOPLASMIC event and that a chromatin interpretation is unsupported. That reading remains correct about what is PRINCIPAL. What this entry adds is only that the histone activity is not absent - three experimental citations and one locus-specific experiment exist - so the epigenetic readout must not be recorded as not-applicable. Oncofusion Knowledge Graph 2026
2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   PRMT3's nuclear pool and its stability are both conditional on ZNF200: the interaction is direct, runs through PRMT3's C2H2-type zinc finger, and is required BOTH to localise PRMT3 to the nucleus AND to inhibit its proteasomal degradation.

   The uncharacterized protein ZNF200 interacts with PRMT3 and aids its stability and nuclear translocation. · Gupta S, Verma M, Kadumuri RV, Chutani N, Khan MIK, Chavali S · Biochemical Journal 2024
   PMID 39513743DOI
   ▸What this evidence cannot show

   Identified from the UniProt SUBUNIT and SUBCELLULAR LOCATION evidence codes, where this single paper supplies the cytosol value, the nucleus value AND the ZNF200 mechanism - so one 2024 paper carries three separate fields of the record, which is the concentration of dependence rule 91 warns about. The abstract was read at title-and-summary level only via esummary; the full text was not read, so the cell systems and the size of the localisation shift are not characterised here. ZNF200 is described as UNCHARACTERIZED in the paper's own title, has no entry in this graph, and has never been measured in any haematopoietic or leukaemic system.

   ▸Evidence that argues against this (1)

   • If the nuclear pool depends on a partner that nothing in this disease has measured, then every chromatin claim about PRMT3 here is conditional on an unmeasured variable, and the same 24-residue zinc finger is annotated as responsible for SUBSTRATE SPECIFICITY - so ZNF200 binding and substrate selection may not be independent. A tumour suppressor binding PRMT3's FERM-accessible surface and inhibiting its activity (EPB41L3), a zinc-finger partner controlling its nuclear entry and stability (ZNF200), and a direct non-catalytic inhibition of ALDH1A1 together describe a protein whose activity is set mostly by what is holding it, which makes any single-gene perturbation of PRMT3 hard to interpret. UniProt Consortium 2026
3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   PRMT3 mediates asymmetric arginine dimethylation of histone H4 Arg-3 (H4R3me2a) in the promoter region of microRNA miR-3648 to promote its transcription and osteogenesis, in human mesenchymal stem cells.

   Asymmetrical methyltransferase PRMT3 regulates human mesenchymal stem cell osteogenesis via miR-3648. · Min Z, Xiaomeng L, Zheng L, Yangge D, Xuejiao L, Longwei L, et al. · Cell Death and Disease 2019
   PMID 31378783DOI
   ▸What this evidence cannot show

   This is the ONE locus-specific PRMT3 histone experiment identified, and its system - human mesenchymal stem cells differentiating along the osteogenic lineage - has no relationship to haematopoiesis, to B-lineage cells or to any fusion. The locus is a microRNA promoter, not a protein-coding gene, and this graph's noncoding entries record how differently microRNA loci behave. Identified from the UniProt FUNCTION evidence codes and confirmed by title via esummary; neither the abstract in full nor the paper was read, so this evidence item is title-level plus a database assertion and is graded accordingly.

   ▸Evidence that argues against this (1)

   • One microRNA promoter in one differentiation system is a thin basis for calling PRMT3 an H4R3me2a writer in general, and PRMT1 is the far better-evidenced writer of the same mark: H4R3me2a[tiab] AND PRMT1[tiab] returns 41 records against 6 for PRMT3, and PRMT1 already has an entry in this graph with its own MLL-fusion literature. The parsimonious reading is that H4R3me2a in any leukaemic cell is PRMT1's product and PRMT3's contribution is minor or context-restricted - which is what UniProt's 'contributes to' wording implies and what this entry assumes. NCBI PubMed 2026
4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   PRMT3 is being drugged in acute leukaemia by degradation rather than inhibition: a first-in-class MDM2-based PRMT3-targeted PROTAC selectively reduced both PRMT3 protein and PRMT3-deposited asymmetric dimethylarginine, inhibited acute-leukaemia cell growth MORE EFFECTIVELY than the PRMT3 inhibitor SGC707, induced intrinsic apoptosis and endoplasmic-reticulum stress signalling while downregulating E2F, MYC and oxidative-phosphorylation programs, and synergised with the glycolysis inhibitor 2-DG by further reducing ATP production.

   Discovery of PRMT3 Degrader for the Treatment of Acute Leukemia. · Zou W, Li M, Wan S, Ma J, Lian L, Luo G, Zhou Y, Li J, Zhou B · Advanced Science 2024
   PMID 39120042DOI
   ▸What this evidence cannot show

   The abstract names no cell line, no leukaemia subtype, no KMT2A or cytogenetic status, and no in vivo experiment - 'acute leukemia cell growth' is as specific as it gets, and the authors' own stated purpose is that the compound is 'a powerful chemical tool for investigating PRMT3 protein functions' rather than a therapeutic candidate. The degrader is MDM2-based, which means the warhead engages a p53 regulator and any phenotype has an MDM2-dependent component that the abstract does not control for. The MYC and E2F downregulation and OXPHOS effects are global expression changes in an unnamed system. Abstract read in full 2026-09-22; full text not read.

   ▸Evidence that argues against this (1)

   • The paper's own framing undercuts the target rationale in a specific and useful way: it states that although PRMT3 inhibitors exist, there has been 'little success in removing the cellular levels of PRMT3-deposited ADMA with small molecules', and that 'the non-enzymatic functions of PRMT3 remain required to be clarified'. A degrader working better than an inhibitor is therefore evidence that the relevant PRMT3 function may be NON-CATALYTIC - consistent with its direct, methyltransferase-independent inhibition of ALDH1A1 and with its activity being set by partners that hold it. If so, PRMT3's epigenetic readout is the wrong handle on it entirely, and the H4R3me2a story is not what the drug is acting through. Verma M et al. 2021
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The DAL-1/4.1B (EPB41L3) FERM domain binds PRMT3 directly and inhibits its ability to methylate substrates in vitro and in vivo - which is the reason this graph reached for PRMT3 at all.

   DAL-1/4.1B tumor suppressor interacts with protein arginine N-methyltransferase 3 (PRMT3) and inhibits its ability to methylate substrates in vitro and in vivo. · Singh V, Miranda TB, Jiang W, Frankel A, Roemer ME, Robb VA, et al. (10 authors; senior author Newsham IF) · Oncogene 2004
   PMID 15334060DOI
   ▸What this evidence cannot show

   Rule 82 applies and was checked: this PMID is already cited twice in this graph (gene-epb41l3 evidence[1] and [3], and ana-q99-read-queue-bounded-rejections evidence[4]), and this entry carries it only as the PRMT3-side statement of a fact gene-epb41l3 already reasons about at length - the reasoning is NOT restated here. The substrates in the methylation assays are PRMT3's characterised substrates and NO HISTONE substrate was assayed, so this establishes methyltransferase regulation and not a histone-mark effect; gene-epb41l3 records that limit precisely and names the missing experiment. The authors are as recorded in that entry rather than independently re-derived for this one.

   ▸Evidence that argues against this (1)

   • A tumour suppressor that inhibits PRMT3 predicts that losing the suppressor RAISES PRMT3 activity - and EPB41L3's own promoter is hypermethylated and silenced in many cancers, which is why it is used as a clinical methylation biomarker. If that holds in a leukaemic cell, PRMT3 would be de-repressed by an epigenetic event on a different gene, and a PRMT3 degrader would be correcting a downstream consequence of EPB41L3 silencing. None of that has been measured in any haematopoietic cell: gene-epb41l3 records EPB41L3[tiab] AND leukemia at two records and DAL1[tiab] AND leukemia at zero. Oncofusion Knowledge Graph 2026
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PRMT3 has no KMT2A-fusion literature and its own mark literature is only reachable in the compact spelling: PRMT3[tiab] = 135, AND KMT2A[tiab] = 0, AND MLL[tiab] = 1 (an assay-development paper), AND MLL-AF9[tiab] = 0, AND AML[tiab] = 1, AND leukemia[tiab] = 2, AND CDC73[tiab] = 0, AND PAFc[tiab] = 0, AND RPS2[tiab] = 11, AND histone[tiab] = 22, AND SGC707[tiab] = 18, AND degrader[tiab] = 2 - and critically AND H4R3[tiab] = 0 against AND "H4R3me2a"[tiab] = 6, which inverts this project's standing residue-plus-wildcard rule for arginine marks.

   NCBI E-utilities esearch, db=pubmed, counts with querytranslation and warninglist read back, every zero run twice, 2026-09-22 · NCBI PubMed 2026
   ▸What this evidence cannot show

   Titles and abstracts only, and E7 shows exactly what that costs here. Every zero was re-run and returned a FULLY PARSED querytranslation together with warninglist.outputmessages = ['No items found.'] - the signature of a genuine compound-query zero, which is NOT the unparsed echo that registry row 123 describes for a genuine zero; that echo signature applies to an unrecognised single token (TMEM103[tiab] -> 'TMEM103[tiab]', confirmed as the control here) and not to a compound query. Three of the counts in this beat were first returned with an EMPTY translation and were discarded and re-run. Instrument control through the identical code path: H3K79me2[tiab] = 224.

   ▸Evidence that argues against this (1)

   • The comparison that shows the rule inverting is the part to carry, because a single pair could be coincidence. Measured across all three arginine writers in this beat: PRMT5 AND H4R3 = 56 against AND H4R3me2s = 75; PRMT6 AND H3R2 = 19 against AND H3R2me2a = 39; PRMT3 AND H4R3 = 0 against AND H4R3me2a = 6. The compact form wins in all three, and only in the PRMT3 case does the residue form cross from understatement into a false zero. Standalone: H4R3 = 109, H4R3me2a = 63, H4R3me2s = 92, H3R2 = 75, H3R2me2a = 47 - so the residue token is not broken, it simply fails to match the indivisible compact token. The safe practice is to run BOTH forms and compare, not to prefer either. NCBI PubMed 2026
7. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The single positive PRMT3 datum in a KMT2A-fusion system is full-text only: the PAF-complex study reports that PRMT5, PRMT1 AND PRMT3 were all significantly downregulated following loss of either CDC73 (conditional excision in murine MLL-AF9 cells) or PAF1 (knockdown in human THP-1), in a comparison of the two expression datasets - a result that appears in the body text and accompanying table and not in the abstract.

   The PAF complex regulation of Prmt5 facilitates the progression and maintenance of MLL fusion leukemia - full text via Europe PMC PMC5785415, Figure 2E-F and accompanying text, read 2026-09-22 · Serio J, Ropa J, Chen W, Mysliwski M, Saha N, Chen L, Wang J, Miao H, Cierpicki T, Grembecka J, Muntean AG · Oncogene 2018
   PMID 28945229DOI
   ▸What this evidence cannot show

   This is a transcript-level observation about the PRMT3 GENE inside a paper about PRMT5, with no PRMT3 protein, activity, occupancy or functional experiment anywhere in it - the authors did not follow PRMT3 up and neither does this entry. The CDC73 arm is murine MLL-AF9 and the PAF1 arm is human THP-1 (also MLL-AF9-driven); there is no MLL-AF4, no B-lineage and no infant material. It is recorded because it is the ONLY positive PRMT3 datum in a KMT2A-fusion system that exists, and because it is unreachable by any PRMT3 query - PRMT3[tiab] AND CDC73[tiab] = 0, AND PAFc[tiab] = 0, and the paper is not in PRMT3[tiab] at all.

   ▸Evidence that argues against this (1)

   • A gene being downregulated alongside two paralogues when a transcriptional co-activator complex is removed is close to uninformative on its own: the PAF complex is a general RNA-Pol-II-associated elongation factor, its excision downregulated most PRMT-family genes with RPKM above 5, and the authors themselves single out PRMT5 for follow-up rather than PRMT3. Reading this as a PRMT3-specific PAFc target would be the same error as reading a pan-essential gene as a fusion dependency. It bounds an absence rather than establishing a relationship. Serio J et al. 2018
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PRMT3 is not a DepMap dependency: mean gene effect +0.0419 across 1,258 screened cell lines with 0.2% of lines depleted below -0.5, and a lymphoid-tissue mean of +0.0776 over 97 lines; in this project's own KMT2A scan both arms are positive at rank 5,198 of 17,728.

   DepMap CRISPR (Chronos) gene effect served through the Open Targets Platform GraphQL API, field depMapEssentiality, fetched 2026-09-22 with approvedSymbol read back as PRMT3 · Broad Institute DepMap via Open Targets Platform 2026
   ▸What this evidence cannot show

   Rule 83: the API does not state the DepMap release; 1,258 is PRMT3's own screen count. The rank of 5,198 ranks a 0.031-unit difference between two POSITIVE means and carries no information. A knockout screen also tests only the consequences of losing the protein, which is what a degrader does - so unlike the catalytic-inhibitor case, here the screen and the drug modality are matched, and the screen says losing PRMT3 does not impair proliferation in 1,258 bulk lines.

   ▸Evidence that argues against this (1)

   • The degrader result is the direct contradiction and it is the better-matched experiment for a degradation modality, so the disagreement is real rather than a modality mismatch: an MDM2-based PRMT3 PROTAC inhibited acute-leukaemia cell growth and outperformed the inhibitor SGC707. The most likely reconciliation is not biological but pharmacological - an MDM2-recruiting degrader has an MDM2-dependent, p53-axis component that a PRMT3 knockout does not, and the abstract reports no PRMT3-independent control. Until that control exists, 'PRMT3 degradation kills leukaemia cells' and 'PRMT3 knockout does nothing in 1,258 lines' should both stay on record. Zou W et al. 2024

• ← inhibitsEPB41L3 / DAL-1 - a membrane-skeleton tumour suppressor that binds and modulates two arginine methyltransferases, and whose own promoter methylation is a clinical test

  DAL-1/4.1B FERM domain binds PRMT3 directly and inhibits its methyltransferase activity

• related to →PRMT5 - the type II arginine methyltransferase this graph has reached for sixteen times, whose mark on H3R2 RECRUITS WDR5 to chromatin

  opposite dimethyl states of the SAME residue H4R3 - me2a activating against me2s repressive

• related to →PRMT1 — an MLL-fusion cofactor whose first-in-human inhibitor was terminated for lack of efficacy and toxicity

  PRMT1 is the dominant H4R3me2a writer, 41 records against 6 for PRMT3

• related to →PRMT6 - the enzyme whose mark is MUTUALLY EXCLUSIVE with the MLL complex on the H3 tail, and which shuts HOX genes down by evicting WDR5

  type I paralogue; both write an asymmetric arginine mark that opposes a PRMT5 symmetric one

• related to →CDC73 / parafibromin — the PAF1C subunit that was genetically tested, and a second tumour suppressor required by the fusion

  PRMT3 transcript falls on CDC73 excision in MLL-AF9 cells - a full-text-only datum

• related to →PAF1 — pan-essential subunit of a complex the fusion needs, where the window is in the interface and not in the protein

  PRMT3 transcript falls on PAF1 knockdown in human THP-1

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

  dominant substrate is the small-subunit ribosomal protein uS5/RPS2; no nucleolar assignment exists

• related to →KMT2A-rearranged acute myeloid leukaemia in children and infants

  an MDM2-based PRMT3 PROTAC inhibits acute-leukaemia cell growth; subtype and genotype unstated

• related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here

  a writer whose principal substrate is not a histone - and whose histone activity is real anyway

• analogous to →CPT1A - the outer-mitochondrial-membrane fatty-acid gatekeeper that is also a lysine succinyltransferase, fused to KMT2A exon 3 in the one case that is tether WITHOUT address

  both are enzymes this graph files with a NON-HISTONE activity as a real epigenetics answer

• causes →H4R3me2a - the activating arginine mark that a KMT2-family enzyme's OWN tandem PHD module reads, where the Kabuki-syndrome mutations sit, and which no KMT2A-fusion study has ever measured

  deposits H4R3me2a at the miR-3648 promoter in human MSCs - one locus, one cell type