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

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 sign is probably wrong for a target, and that is the entry's main message: PRMT6 gain lowers MLL1/WDR5 recruitment and HOX transcription, so PRMT6 INHIBITION is predicted to RAISE H3K4me3 and HOX output in a HOX-driven leukaemia. A probe compound exists (EPZ020411) and using it here without first establishing the sign would be the cycle-54 Elongator error repeated.
• Class inhibitors do not respect the sign: a type I PRMT inhibitor hits PRMT1 (an apparent MLL-fusion cofactor) and PRMT6 (an apparent brake) together, so the net direction of GSK3368715-like chemistry in this disease is genuinely unknown, and that agent was already terminated for lack of efficacy and dose-limiting toxicity.
• Compensation and readout ambiguity: H3R2 carries two different dimethyl states with opposite effects on WDR5. Any intervention or assay that cannot separate H3R2me2a from H3R2me2s can produce a result whose direction is uninterpretable, and the two writers sit in different enzyme classes with different inhibitors.
• Element-dependent effects mean a single-endpoint experiment can mislead: losing the mark raises KMT2A and H3K4me3 at promoters and lowers KMT2D, H3K4me1 and H3K27ac at enhancers. A study measuring only promoter H3K4me3 would report a clean activating result and miss a simultaneous enhancer collapse.
• Essential-function and expression mismatch: PRMT6 is not a dependency in any DepMap arm (mean -0.0666) and is expressed highest in kidney and testis, so an inhibitor would have little on-tumour effect to trade against whatever off-tumour effect it has.
• Evidence transfer: every mechanistic result here is from a cancer cell line, a neuronal-differentiation model or a RUNX1 system. There is no KMT2A-fusion experiment of any kind, so the entire argument - in both directions - is imported.

1. Is H3R2me2a present at KMT2A::AFF1 target loci, and is it depleted there relative to the wild-type allele's targets? The mark is selectively absent from active promoters in normal cells; a fusion that holds promoters open should show the same depletion, and if it does not, the fusion is working against a mark nobody has looked for.
2. Does PRMT6 overexpression displace a KMT2A fusion from HOXA9? The experiment exists for HoxA2 in neuronal differentiation with MLL1/WDR5 as the readout, and transplanting it to a t(4;11) line with fusion-specific occupancy would be cheap and would test the whole entry.
3. Which dimethyl state of H3R2 dominates in a KMT2A-rearranged blast? PRMT6 writes me2a, which evicts WDR5; PRMT5 and PRMT7 write me2s, which recruits it. The two predict opposite things and no measurement distinguishing them exists in this disease - and any antibody that cannot separate them cannot answer it.
4. Is PRMT6 a tumour suppressor in KMT2A-rearranged leukaemia? Everything in this entry points that way and nothing tests it; a PRMT6 knockout or knockdown in a KMT2A-fusion model with a HOXA-output readout would settle the sign before anyone proposes a type I PRMT inhibitor here.
5. Does H3R2me2a contribute to bivalency at fusion targets? RUNX1 target genes are held in an INTERMEDIATE state with this mark, this graph holds `exp-bivalency-cuttag-kmt2a-aff1` as a planned experiment, and H3R2me2a is not on that experiment's panel.
6. Does the enhancer arm matter more than the promoter arm here? Loss of enhancer H3R2me2a reduces KMT2D binding and H3K27ac, and this graph's enhancer-hijacking and BENC entries are about exactly that class of element - so PRMT6 could act on a KMT2A-rearranged cell through KMT2D at enhancers rather than through KMT2A at promoters.

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

   PRMT6 (Q96LA8, 375 aa) is a type I protein arginine methyltransferase with a strong preference for asymmetric dimethylarginine, specifically mediating asymmetric dimethylation of histone H3 Arg-2 to form H3R2me2a - which UniProt describes as a tag for epigenetic transcriptional repression that is mutually exclusive with H3K4me2 and H3K4me3 - carrying a SAM-dependent MTase PRMT-type domain at 44-374, a single Nucleus location value, and high expression in kidney and testis.

   UniProtKB Q96LA8 (ANM6_HUMAN), Protein arginine N-methyltransferase 6 - FUNCTION, CATALYTIC ACTIVITY, SUBUNIT, SUBCELLULAR LOCATION, TISSUE SPECIFICITY and feature table, fetched 2026-09-22 · UniProt Consortium 2026
   ▸What this evidence cannot show

   A database record, not a measurement. The localisation record is the weakest part of it: ONE location value (Nucleus) resting on ONE citation from 2001, for an enzyme whose chromatin mark was characterised in 2007 - rule 91 territory, and the cited paper was not read for this entry. The accession is asserted explicitly and NOT confirmed by a length readback, because PRMT6 (375 aa) and PRMT1 (371 aa) are four residues apart and both are in this corpus.

   ▸Evidence that argues against this (1)

   • UniProt's own summary that H3R2me2a 'represents a specific tag for epigenetic transcriptional repression' is more categorical than the primary literature now supports. The genome-wide study of PRMT6-dependent H3R2me2a finds the mark at ACTIVE genes, and at enhancers its loss REDUCES KMT2D binding, H3K4me1 and H3K27ac and DECREASES transcription - an activating role for the same mark at a different class of element. A database one-liner that fixes the sign of a mark is the kind of statement this graph has been burned by. Bouchard C et al. 2018
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   PRMT6 catalyses H3R2 dimethylation in vitro and controls global H3R2me2a in vivo, and the mark is MUTUALLY EXCLUSIVE with H3K4 methylation in both directions: H3K4me3 on the H3 tail prevents PRMT6 from acting, while H3R2me2a prevents H3K4 methylation and prevents binding of an ASH2/WDR5/MLL-family methyltransferase complex to the H3 tail; by ChIP, H3R2me2a lies in gene bodies and 3-prime ends irrespective of transcriptional state and is selectively depleted from active promoters where H3K4me3 is present.

   Methylation of histone H3R2 by PRMT6 and H3K4 by an MLL complex are mutually exclusive. · Guccione E, Bassi C, Casadio F, Martinato F, Cesaroni M, Schuchlautz H, Luscher B, Amati B · Nature 2007
   PMID 17898714DOI
   ▸What this evidence cannot show

   Human cells, but the systems are cancer cell lines and the promoter-level ChIP is genome-scale correlation plus in vitro reconstitution; there is no haematopoietic arm, no fusion and no leukaemia in this paper. 'An ASH2/WDR5/MLL-family complex' is the paper's own wording - it is a family statement, so which KMT2 paralogue is excluded at any given promoter is not resolved, and a KMT2A fusion that has LOST the SET domain is a different object again. Abstract read in full 2026-09-22; full text not read.

   ▸Evidence that argues against this (1)

   • Mutual exclusion on a peptide or reconstituted nucleosome need not mean mutual exclusion on chromatin in a cell, and the same laboratory later showed that the OTHER dimethylation of the same residue does the opposite: H3R2me2s, written by PRMT5 and PRMT7, ENHANCES WDR5 binding. A single residue with two dimethyl states of opposite effect means that 'H3R2 methylation blocks the MLL complex' is only true of the asymmetric form, and any antibody or mass-spectrometry readout that cannot separate me2a from me2s cannot test the claim at all. Migliori V et al. 2012
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   PRMT6 is the major in vivo H3R2 methyltransferase and methylates H3 R2 and H4/H2A R3 in vitro; H3R2 methylation inhibits H3K4 trimethylation and the recruitment of WDR5 to histone H3 in vitro; PRMT6 overexpression decreases transcription of HOX genes and MYC-dependent genes, with more H3R2me2a and PRMT6 and less H3K4me3 and less MLL1/WDR5 recruitment at the HoxA2 gene - and on retinoic-acid-induced HoxA2 activation in a neuronal-differentiation model, PRMT6 and H3R2me2a fall while H3K4me3 rises.

   PRMT6-mediated methylation of R2 in histone H3 antagonizes H3 K4 trimethylation. · Hyllus D, Stein C, Schnabel K, Schiltz E, Imhof A, Dou Y, Hsieh J, Bauer UM · Genes and Development 2007
   PMID 18079182DOI
   ▸What this evidence cannot show

   The HOX result is HoxA2 in a neuronal-differentiation model, not HOXA9 in a haematopoietic one, and the manipulation is PRMT6 overexpression rather than physiological variation. The WDR5-recruitment result is in vitro on histone H3. This is a 2007 paper and nothing has repeated it in a blood cell; there is no KMT2A-fusion arm anywhere in it. Abstract read in full 2026-09-22; full text not read.

   ▸Evidence that argues against this (1)

   • The direction this establishes is the awkward one for a target list, and it is worth stating as the counter-argument to the entry existing as a therapeutic node at all: if PRMT6 gain lowers MLL1/WDR5 recruitment and HOX transcription, then PRMT6 LOSS or inhibition should raise them - so a PRMT6 inhibitor, of which chemical probes exist (EPZ020411, 15 records in titles and abstracts, 2026-09-22), is predicted to be PRO-leukaemic in a HOX-driven disease. This is the same shape of sign error the Elongator beat found in cycle 54, where an elongation-disease framing would have proposed inhibiting a tumour suppressor. NCBI PubMed 2026
4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Genome-wide, PRMT6-dependent H3R2me2a is found at ACTIVE genes at both promoters and enhancers; loss of the mark from promoters leads to ENHANCED KMT2A BINDING and increased H3K4me3 deposition with increased target-gene transcription, while loss at enhancers reduces KMT2D binding and H3K4me1/H3K27ac deposition with decreased transcription - so the same mark interferes with adjacent histone marks in opposite directions depending on the element.

   Genomic Location of PRMT6-Dependent H3R2 Methylation Is Linked to the Transcriptional Outcome of Associated Genes. · Bouchard C, Sahu P, Meixner M, Nutzold RR, Rust MB, Kremmer E, Feederle R, Hart-Smith G, Finkernagel F, Bartkuhn M, Savai Pullamsetti S, Nist A, Stiewe T, Philipsen S, Bauer UM · Cell Reports 2018
   PMID 30232013DOI
   ▸What this evidence cannot show

   A neural-differentiation cell model with a PRMT6 knockout that has proliferation and differentiation defects - so the chromatin changes are measured in cells whose phenotype has already changed, and cause cannot be separated from consequence at the level of the abstract. No haematopoietic cell, no fusion, no leukaemia. Crucially the KMT2A result is a binding change in a PRMT6-null background, not a demonstration that H3R2me2a levels set KMT2A occupancy dynamically. Abstract read in full 2026-09-22; full text not read.

   ▸Evidence that argues against this (1)

   • A KMT2A-binding change measured in wild-type cells says nothing about a KMT2A FUSION, and the direction of the argument may not survive the transfer. Most KMT2A fusions retain the AT-hooks and the CXXC domain and lose PHD1-3, the bromodomain and the SET domain, so the chimera's chromatin address in this graph's own model is CpG-DNA-based rather than H3-tail-based (interface-kmt2a-cxxc-cpg-dna, mechanism-fusion-dna-targeting-cxxc-athooks). An H3-tail arginine mark that gates the H3-tail-binding activities of full-length KMT2A could be irrelevant to a fusion whose tail-reading modules are gone, which would make PRMT6 a regulator of the wild-type allele only. Guccione E et al. 2007
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   PRMT6 associates with the haematopoietic transcription factor RUNX1 and triggers H3R2me2a at RUNX1 target genes, where that mark negatively influences H3K4me3 and acts repressively, keeping RUNX1 target genes in an INTERMEDIATE state; on differentiation the RUNX1/PRMT6 corepressor complex is exchanged for a RUNX1 coactivator complex.

   Histone arginine methylation keeps RUNX1 target genes in an intermediate state. · Herglotz J, Kuvardina ON, Kolodziej S, Kumar A, Hussong H, Grez M, et al. · Oncogene 2013
   PMID 22777353DOI
   ▸What this evidence cannot show

   This is the primary paper behind a statement `gene-runx1` already carries from a REVIEW, whose own caveat records that the primary papers were not retrieved; the two entries agree and this one supplies the source. The system is RUNX1-driven haematopoietic differentiation, not a KMT2A fusion, and RUNX1 is not a KMT2A-fusion partner in this project's index disease. Abstract-level evidence: the abstract was read, the full text was not, so no specific target locus or effect size is claimed here beyond what the abstract states.

   ▸Evidence that argues against this (1)

   • A RUNX1-scoped result is the wrong transcription factor for this graph's index disease and may point the other way. RUNX1 is the 5-prime partner of ETV6::RUNX1 and the 3-prime partner of RUNX1::RUNX1T1, both of which this graph holds as comparator fusions with markedly BETTER outcomes than KMT2A rearrangement; a PRMT6 mechanism anchored on RUNX1 could belong to that biology rather than to the KMT2A axis. The mechanism that would actually transfer is the HOX one, and the only HOX experiment is HoxA2 in neuronal differentiation. Hyllus D et al. 2007
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PRMT6's literature is small and its KMT2A-fusion literature is almost empty: PRMT6[tiab] = 265, AND MLL[tiab] = 5, AND KMT2A[tiab] = 2, AND leukemia[tiab] = 5, AND H3K4me3[tiab] = 15, while the arginine-mark queries H3R2[tiab] AND methyl*[tiab] = 58 and H3R2me2a[tiab] = 47 - and no record intersects PRMT6 with any named KMT2A fusion.

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

   Titles and abstracts only; rule 80 applies. Registry row 124 fires as predicted - AND KMT2A returns 2 (30232013 and a prostate-carcinoma methylase-expression paper) while AND MLL returns 5 and is where the mechanism is, including the Nature and Genes and Development founding papers - so the modern symbol alone would have returned one on-topic record and a false lead. Note also that the compact mark spelling is NOT a false zero for arginine marks here: H3R2me2a[tiab] = 47 against H3R2[tiab] AND methyl*[tiab] = 58. Instrument controls through the identical code path: H3K79me2[tiab] = 224 positive, TMEM103[tiab] = 0 with the term echoed back unparsed.

   ▸Evidence that argues against this (1)

   • A small count is not the same as a small importance, and here the count understates reach in a specific and checkable way: the founding papers name the complex in their TITLES as 'an MLL complex' and 'MLL1/WDR5', which is why they are reachable through AND MLL and not through AND KMT2A - the symbol used in a 2007 title is not the symbol a 2026 query uses. Any bounded absence for PRMT6 built on the modern symbol alone would have been wrong by two Nature-and-Genes-Dev-sized papers. NCBI PubMed 2026
7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PRMT6 is NOT essential in DepMap CRISPR (Chronos): mean gene effect -0.0666 across 1,257 screened lines with 0.2% of lines depleted below -0.5, and a lymphoid-tissue mean of -0.0508 over 97 lines.

   DepMap CRISPR (Chronos) gene effect served through the Open Targets Platform GraphQL API, field depMapEssentiality, fetched 2026-09-22 with approvedSymbol read back as PRMT6 · Broad Institute DepMap via Open Targets Platform 2026
   ▸What this evidence cannot show

   Rule 83: the API does not state the DepMap release, so only the fetch date is on record; 1,257 is PRMT6's own screen count and not a corpus denominator. A near-zero gene effect in a pooled screen is also exactly what paralogue redundancy looks like - PRMT1, PRMT3 and PRMT8 are type I enzymes with overlapping substrate preference - so 'not essential' here is not evidence of dispensability. In this project's own KMT2A scan the difference is -0.0358 Chronos units at rank 4,735 of 17,728, which is noise and is recorded only so that nobody reads the rank as selectivity.

   ▸Evidence that argues against this (1)

   • A DepMap proliferation screen is the wrong assay for a gene whose predicted phenotype is a shift in differentiation state rather than a loss of viability. PRMT6 knockout in the one genome-wide study produced PROLIFERATION AND DIFFERENTIATION DEFECTS in a neural model, which a pooled dropout screen in cancer lines would register as a mild negative at best, and the mark's promoter effect is to restrain KMT2A and H3K4me3 - restraint being a phenotype that a viability screen cannot see in either direction. Bouchard C et al. 2018
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   PRMT6 has a characterised NON-histone, non-chromatin-mark substrate: it forms a complex with DNA polymerase beta and methylates it at Arg-83 and Arg-152 in vitro and in vivo, strongly stimulating polymerase activity by enhancing DNA binding and processivity without affecting single-nucleotide insertion or dRP-lyase activity, and an R83/R152K mutant fails to restore cellular resistance to a DNA alkylating agent.

   Arginine methylation regulates DNA polymerase beta. · El-Andaloussi N, Valovka T, Toueille M, Steinacher R, Focke F, Gehrig P, Covic M, Hassa PO, Schar P, Hubscher U, Hottiger MO · Molecular Cell 2006
   PMID 16600869DOI
   ▸What this evidence cannot show

   Biochemistry plus genetic complementation in Pol-beta-knockout cells; no haematopoietic system, no leukaemia, and no link to KMT2A. It is included because this beat was told not to default an epigenetic readout to not-applicable without checking, and because base excision repair matters to this project through its break-machinery entries - but the connection to KMT2A breakage is a hypothesis nobody has tested and is not claimed here.

   ▸Evidence that argues against this (1)

   • A base-excision-repair activity pulls PRMT6 towards this graph's break-machinery reasoning, and that reasoning does not obviously want it: the KMT2A breakpoint cluster region's documented mechanisms here are topoisomerase-II cleavage, replication stress and apoptotic nuclease cleavage, with repair by NHEJ and alternative end joining - none of which is base excision repair. Stimulating Pol beta would if anything be protective against alkylating damage, so the substrate is real and its relevance to KMT2A breakage is not established in either direction. UniProt Consortium 2026

• inhibits →H3K4me3 - the mark KMT2A's own SET domain writes, that the fusion cannot write, that an interface in this graph already binds, and that switches PRC2 off

  H3R2me2a is MUTUALLY EXCLUSIVE with H3K4me2/me3 and blocks its deposition

• inhibits →WDR5 — the partner that is not a partner: it binds the half of KMT2A the fusion throws away

  H3R2me2a blocks WDR5 binding to the H3 tail and lowers MLL1/WDR5 recruitment

• inhibits →KMT2A / MLL1 — the gene that breaks

  loss of promoter H3R2me2a ENHANCES KMT2A binding; the mark restrains it

• regulates →KMT2D (MLL2/MLL4) - the gene named in this graph's menin-resistance entry, written here for the first time

  loss of enhancer H3R2me2a REDUCES KMT2D binding - opposite sign to the promoter arm

• related to →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it

  PRMT6 gain represses HOX transcription (shown for HoxA2, not HOXA9)

• binds →RUNX1 — activated by KMT2A::AFF1 and then bound by the reciprocal fusion, in the one published two-oncoprotein circuit in t(4;11)

  RUNX1/PRMT6 corepressor complex deposits H3R2me2a at RUNX1 targets

• 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 H3R2, with opposite effects on WDR5

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

  type I paralogue four residues apart in length - a length readback cannot separate them

• regulates →H3K27ac - named in 40 entries with no entry of its own, and the strongest KMT2A::AFF1 measurement of it is invisible to every abstract search

  loss of enhancer H3R2me2a reduces H3K27ac and H3K4me1

• related to →EPB41L3 / DAL-1 - a membrane-skeleton tumour suppressor that binds and modulates two arginine methyltransferases, and whose own promoter methylation is a clinical test

  UniProt SUBUNIT lists an EPB41L3-PRMT6 interaction citing the PRMT3 paper; no activity effect reported

• related to →Experiment: is the primed state in KMT2A::AFF1 actually bivalent? Multi-mark CUT&Tag on primary infant blasts

  H3R2me2a holds RUNX1 targets in an INTERMEDIATE state and is not on the planned bivalency panel

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

  arginine methylation as a chromatin mark class

• related to →WDR5: an established wild-type KMT2A contact, and no fusion contact ever measured

  a mark that evicts WDR5 from the H3 tail bears directly on that adjudication

• related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)

  NO experiment in any KMT2A fusion; the relationship is inferred from wild-type KMT2A

• ← related toPRMT3 - 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

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

• causes →H3R2me2a - the asymmetric arginine mark two residues from H3K4 that LOCKS the WRAD module off the histone tail, and the first arginine mark node in this graph

  major in vivo H3R2 methyltransferase; controls global H3R2me2a levels

• 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

  methylates H4/H2A R3 IN VITRO only; degree unstated; PRMT6's in vivo target is H3R2

• ← testsMeasure H3R2 and H4R3 methylation states at KMT2A-fusion targets by isomer-resolving mass spectrometry, and test the gate by writer perturbation read out on the enzyme's product - the experiment that has never been done in any fusion

  writer of the H3R2 block; predicted to RAISE WDR5 occupancy and H3K4me3 on loss at loci where the block operates