Oncofusion

start / graph / mark-h3r2me2a

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q103-arginine-marks

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

also called H3R2me2a · H3R2me2 · asymmetric dimethylation of histone H3 arginine 2 · H3 Arg-2 asymmetric dimethylation

H3R2me2a is a chemical tag sitting two amino acids away from the spot the KMT2A machine works on, and while it is there the machine cannot grip the histone tail at all. It is written by a different enzyme, PRMT6, and it is the reason the tag next door - the 'gene on' mark - cannot be made.

The same thing, for a scientist

H3R2me2a is the asymmetric dimethyl mark on histone H3 arginine 2, written principally by PRMT6 [E1][E2][E3], and it is the only histone modification for which the published mechanism of action is *direct occlusion of the WRAD module* this project spends most of its effort on. Two independent 2007 reports established mutual antagonism with H3K4 methylation in both directions: H3K4me3 on the tail prevents PRMT6 acting, and H3R2me2a prevents H3K4 methylation **and prevents binding of an ASH2/WDR5/MLL-family methyltransferase complex to the H3 tail** [E2]; the second group showed the same block on WDR5 recruitment, with PRMT6 gain reducing MLL1/WDR5 occupancy and HOX transcription at HoxA2 [E3], and a third showed the yeast equivalent runs through occlusion of Spp1, the Set1-complex subunit needed for trimethylation [E4]. **The quantitative structure of that block was revised in 2020**: crystallography plus quantitative binding shows the switch is BINARY rather than graded - H3R2me0, me1 and me2s are all *equally* permissive for WDR5 and only me2a prevents the interaction [E5]. And the simple reading of the mark as repressive is contradicted by the genome-wide work: PRMT6-dependent H3R2me2a sits at ACTIVE genes, at promoters **and** enhancers, and its loss at enhancers REDUCES KMT2D binding [E6] - the opposite sign from its promoter behaviour, and the opposite of the UniProt annotation that calls it "a specific tag for epigenetic transcriptional repression" [E1]. **No eraser is established** [E7][E8][E9]. **Nobody has measured this mark at a KMT2A fusion target in any fusion, in any lineage**: `"H3R2me2a"[tiab] AND KMT2A[tiab]` returns exactly 1 record and it is the neural-differentiation paper [E6], `AND "MLL-AF4"[tiab]` = 0 and `AND menin[tiab]` = 0, both re-run [E10].

Go deeper
Why an arginine mark is a WRAD problem and not an epigenetics footnote

This graph contains gene-wdr5, gene-rbbp5, gene-ash2l, gene-dpy30, interface-wdr5-win-kmt2a-arg3765 and three concept-*-what-is-established-for-wild-type-kmt2a-and-what-for-the-fusion adjudications, because what actually brings WRAD to a KMT2A fusion is the open question in the beat. Every one of those entries reasons about the protein-protein contact: the WIN motif, the RBBP5 VDV peptide, the ASH2L-DPY30 interface.

H3R2me2a is a different kind of lever on the same module. It does not compete for the WIN pocket; it changes the substrate. Guccione et al. state it plainly - the H3R2me2a mark "prevented methylation of H3K4 as well as binding to the H3 tail by an ASH2/WDR5/MLL-family methyltransferase complex" 2. Hyllus et al. measured the same thing as a loss of WDR5 recruitment to histone H3 in vitro, and then in cells as reduced MLL1 and WDR5 occupancy at HoxA2 with reduced H3K4me3 and reduced transcription 3.

So there exists a chromatin state in which the whole WRAD/KMT2 module is excluded from the tail with the WIN motif intact and menin untouched. That is what makes it worth four node entries rather than a sentence inside gene-prmt6.

And note which direction the leukaemia-relevant arrow points. PRMT6 gain lowers HOX transcription 3; loss of promoter H3R2me2a raises KMT2A binding and H3K4me3 6. In a disease that lives on a HOXA program, H3R2me2a is a brake, and an inhibitor of the enzyme that writes it would be releasing that brake. gene-prmt6 already records this direction warning for the enzyme; it is recorded here for the mark because the mark is what an experiment measures.

The 2020 correction, which this graph did not have and which changes the claim

The version of the switch that reached this graph through gene-prmt5 is Migliori et al.'s: H3R2me2a impedes WDR5 recruitment while H3R2me2s ENHANCES it 11. Lorton et al. 2020 tested that with quantitative binding and high-resolution crystal structures of WDR5 with unmodified, me1 and me2s arginine and with the H3R2me1 peptide, and reported an interaction "regulated by a binary arginine methylation switch: H3R2me2a prevents interaction with WDR5, whereas H3R2me0, -me1, and -me2s are equally permissive" 5.

The difference matters for exactly the experiment this beat would propose. Under the graded model, a rise in H3R2me2s is a gain-of-recruitment event and should raise WRAD occupancy above baseline. Under the binary model it is not: converting me2a to me2s relieves a block and returns the tail to the me0 baseline, and a rise in H3R2me2s over unmodified chromatin predicts no change at all.

So the honest statement of the switch is one-sided. The established half is the OFF state, H3R2me2a, replicated by three groups on two continents in two organisms 234 and independently confirmed as the only non-permissive state in the structural re-examination 5. The ON state is the contested half. This is recorded as a correction to a claim in gene-prmt5 and gene-prmt6 rather than a defect in them: their citation 11 says what they say it says, and the paper that revises it is four years later and was not cited there.

The antibody problem, which is not a caveat here but the binding constraint on every experiment

H3R2me2a and H3R2me2s are the same residue, the same number of methyl groups, and the same mass. They differ only in whether the two methyls sit on one terminal nitrogen of the guanidino group or on both. They predict opposite consequences for WRAD occupancy 511. Therefore:

An assay that cannot separate me2a from me2s cannot test the switch, in either direction, and its result is uninterpretable rather than merely noisy. A ChIP that pulls both states reports the sum of an inhibitory and a permissive mark.

The general evidence that this is a live risk rather than a theoretical one:

  • More than 25% of over 200 antibodies raised against 57 histone modifications failed specificity tests by dot or western blot, and more than 20% of the specific ones failed in ChIP 12.
  • The Histone Antibody Specificity Database exists precisely because commercial histone antibody behaviour has to be checked per lot by peptide microarray 13.
  • For the neighbouring residue, 52 commercial antibodies against the H3K4 methylforms were characterised by peptide array and internally calibrated ChIP, and "many widely used antibodies poorly distinguish the methylforms", with high- and low-specificity reagents giving "dramatically different biological interpretations" 14. That is the mark two residues away from this one, measured in the same chromatin, and the reason the H3K4me-versus-H3R2me literature needs reading with the reagent in view.

Bounded honestly: no study read for this entry reports a head-to-head me2a-versus-me2s cross-reactivity panel for H3R2. "H3R2me2a"[tiab] AND antibod*[tiab] returns 1 record, which is Kirmizis et al. using "an antibody specific for H3R2me2a" for ChIP-on-chip without a published symmetric-state competition 104. So the correct statement is: the discrimination has not been demonstrated for this antibody class, and the field's base rate for histone antibodies makes assuming it unsafe. Mass spectrometry distinguishes the isomers; antibodies are the ones that must prove it.

Genomic distribution: three papers, two incompatible pictures, and the reason both may be right
studysystemwhere H3R2me2a is
Guccione 2007 2human promoters, ChIPgene body and 3' end, regardless of transcriptional state; selectively and locally depleted from active promoters coincident with H3K4me3
Kirmizis 2007 4S. cerevisiae, genome-wide ChIP-on-chipenriched throughout heterochromatic loci and inactive euchromatic genes; at the 3' end of moderately transcribed genes; pattern mutually exclusive with H3K4me3
Bouchard 2018 6murine neural differentiation, PRMT6 knockoutat ACTIVE genes, at promoters and enhancers; promoter peaks behave repressively, enhancer peaks co-localise with H3K4me1 and H3K27ac and behave as activating

The first two agree and are the basis of the textbook reading. The third is the one this beat was told to take seriously, and it does not simply overturn them - it splits the mark by genomic compartment. Bouchard et al.'s own summary is that loss of promoter H3R2me2a gives enhanced KMT2A binding, more H3K4me3 and more transcription (repressive), while loss of enhancer H3R2me2a gives reduced KMT2D binding, less H3K4me1/H3K27ac and less transcription (activating) 6.

That is a single mark with opposite signs on two KMT2-family enzymes in one experiment, and it is the strongest published warning against writing "H3R2me2a is repressive" as a fact. UniProt does write it as a fact 1. The disagreement is recorded in counter_evidence on E1 rather than resolved.

Erasers: what is claimed, what was withdrawn, and the one route that provably cannot work

No arginine demethylase is established for this mark. The record, in order:

  1. JMJD6 was reported as a histone arginine demethylase acting on H3R2 and H4R3 in biochemical and cell-based assays 7.
  2. Two years later the same enzyme was shown to be a lysyl-5-hydroxylase of the splicing factor U2AF65 8, and a 2015 review of the enzyme is titled "a case study in conflicting assignments" and states the demethylase role "is also subject to conflicting reports" while the lysyl hydroxylation of splicing regulators and histones does occur 9.
  3. A subset of JmjC lysine demethylases can catalyse arginine demethylation in purified form, with formaldehyde release, on histone and non-histone fragments - but the authors' own framing is that the existence of bona fide N-omega-methylarginine demethylases "is controversial" and that their result should "stimulate efforts to detect biologically relevant RDM activity" 15. That is an in vitro capability, not an assigned cellular eraser.

And the route everyone reaches for is ruled out for the dimethyl states specifically. PADI4/PAD4 converts methyl-arginine to citrulline, releasing methylamine, and antagonises arginine methylation at H3 R2, R8, R17 and R26 1617. But Cuthbert et al. measured the substrate requirement: "Dimethylation of arginines prevents deimination by PADI4 although monomethylation still allows deimination to take place" 17. So citrullination can remove H3R2me1 and unmodified H3R2 from the pool; it cannot remove H3R2me2a.

The consequence for the switch model is structural, not cosmetic. If there is no eraser, H3R2me2a is removed by histone turnover, replication-coupled dilution, or nucleosome eviction - not by a signalling enzyme. A therapeutic hypothesis of the form "remove the block" therefore has no enzyme to agonise, and the only pharmacological handle is the writer, PRMT6, which is not a DepMap dependency (gene-prmt6, mean Chronos -0.0666) and has no clinical agent.

What is known in a KMT2A-rearranged context: essentially nothing, bounded

Every count run 2026-09-22 with the querytranslation read back; each zero re-run and each re-run returned the same zero with a non-empty translation and warninglist.outputmessages = ["No items found."]. Operand control: KMT2A[tiab] = 1652, so the conjunct works.

querynnote
"H3R2me2a"[tiab] AND KMT2A[tiab]1PMID 30232013, murine neural differentiation - not leukaemia
"H3R2me2a"[tiab] AND MLL[tiab]417898714, 22231400, 22777353, 27323813 - the two 2007 mechanism papers, the RUNX1 paper, a prostate-cancer paper; none is a leukaemia chromatin measurement
H3R2[tiab] AND KMT2A[tiab]1same single record
"H3R2me2a"[tiab] AND "MLL-AF4"[tiab]0 (re-run 0)
"H3R2me2a"[tiab] AND "MLL-AF9"[tiab]0 (re-run 0)
"H3R2me2a"[tiab] AND menin[tiab]0 (re-run 0)
"H3R2me2a"[tiab] AND revumenib[tiab]0 (re-run 0)
"H3R2me2a"[tiab] AND leukemia[tiab]322777353 (RUNX1), 36574771, 38424136 (AML LSC, PRMT6 via circFAM193B)

The absence is bounded to titles and abstracts. An H3R2me2a ChIP could exist as a supporting panel in a KMT2A-fusion paper and be invisible to every query above; that is this project's rule-80 failure mode and it fires on measurements more often than on entities. What is bounded is that no paper announces such a measurement, which is a weaker and true claim.

The nearest thing to a leukaemia result is not about a KMT2A fusion at all: PRMT6 associates with RUNX1 and triggers H3R2me2a at RUNX1 target genes, where the mark negatively influences H3K4me3 and holds those genes in an intermediate state (gene-runx1, gene-prmt6, PMID 22777353). RUNX1 is a haematopoietic transcription factor with its own entry here, so the mark has a haematopoietic foothold and no KMT2A foothold.

Facts and figures
markH3R2me2a
histoneH3
residueR2 (Arg-2), N-terminal tail, two residues N-terminal of K4
chemistryasymmetric omega-N,N-dimethylarginine (aDMA) - both methyl groups on ONE terminal guanidino nitrogen. Isobaric with the symmetric isomer H3R2me2s; the two are indistinguishable by mass and predict OPPOSITE consequences for WRAD recruitment.
writers notePRMT6 is established as the major in vivo H3R2 methyltransferase by both overexpression and knockdown (PMID 18079182) and controls global H3R2me2a levels (PMID 17898714). PRMT6 also makes H3R2me1 (PMID 32207970). No second human H3R2me2a writer was established in any source read for this entry.
eraser statusNO ESTABLISHED ERASER. (i) JMJD6 was claimed for H3R2 and H4R3 (PMID 17947579, Science 2007) and the assignment has not held - the same enzyme was shown to be a lysyl-5-hydroxylase of U2AF65 (PMID 19574390) and a dedicated review calls the demethylase role 'subject to conflicting reports' (PMID 25997831). (ii) A subset of JmjC KDMs can demethylate arginine in PURIFIED form (PMID 27337104) with the authors themselves calling bona fide RDMs controversial. (iii) PADI4 citrullination provably CANNOT remove this mark: dimethylation prevents deimination while monomethylation still permits it (PMID 15339660). CONSEQUENCE: removal is by histone turnover or nucleosome eviction, and there is no enzyme to agonise therapeutically.
reader noteThe mechanism of action of this mark is OCCLUSION, not recruitment. No positive reader of H3R2me2a on histone H3 was found in any source read for this entry. The one KMT2-family domain reported to BIND an asymmetric arginine mark is the tandem PHD(4-6) of KMT2D/MLL4, and that is on H4R3, not H3R2 (PMID 23249737; see mark-h4r3me2a).
effect on h3k4 methylationBLOCKS. Mutual exclusion in both directions: H3K4me3 on the tail prevents PRMT6 acting, and H3R2me2a prevents H3K4 methylation (PMID 17898714). In yeast the block is on trimethylation specifically, via Spp1 occlusion (PMID 17898715).
switch stateOFF state of the H3R2 binary switch. H3R2me0, me1 and me2s are equally permissive for WDR5; only me2a prevents the interaction (PMID 32207970).
antibody constraintLOAD-BEARING, not a caveat. H3R2me2a and H3R2me2s are isobaric isomers of the same residue predicting opposite WRAD outcomes, so an assay that cannot separate them cannot test the switch in either direction and returns an uninterpretable sum. Base rates: >25% of 200+ histone-modification antibodies failed specificity by dot/western blot and >20% of the specific ones failed in ChIP (PMID 21131980); 52 commercial antibodies against the H3K4 methylforms - the residue two positions away - often fail to distinguish methylforms, with reagent choice changing the biological conclusion (PMID 30244833). BOUNDED: no head-to-head me2a-versus-me2s H3R2 antibody panel was found ('H3R2me2a'[tiab] AND antibod*[tiab] = 1, 2026-09-22). The Histone Antibody Specificity Database (PMID 26212453) exists and WAS NOT QUERIED in this beat - that is a named next step, not a finding. Mass spectrometry resolves the isomers; antibodies must prove it per clone and per lot.
relationship to menin and the win motifINDEPENDENT OF BOTH. The occlusion acts on the histone substrate, not on the KMT2A polypeptide: the WIN motif can be intact and menin fully engaged while the WRAD/KMT2 module is excluded from the H3 tail. This is a PARALLEL control point on the same module, not evidence against menin dependence - see mechanism-arginine-methylation-gates-kmt2-access-independently-of-menin-and-the-win-motif.
cell type and stateMeasured in: HeLa and human promoter panels (PMID 17898714); S. cerevisiae (PMID 17898715); a neuronal-differentiation cell model with retinoic-acid-induced HoxA2 (PMID 18079182); murine neural precursors under PRMT6 knockout (PMID 30232013); 32D/haematopoietic RUNX1 target genes (PMID 22777353, via gene-prmt6 and gene-runx1). NOT measured in fetal liver, LMPP, pro-B, or any KMT2A-rearranged line or patient sample. The only haematopoietic foothold is the RUNX1 axis.
shown for fusionNO - nobody has looked in any fusion. Not in KMT2A::AFF1, not in KMT2A::MLLT3, not in any KMT2A rearrangement, not in mouse and not in human. The single KMT2A-co-occurring record (PMID 30232013) is wild-type KMT2A in a murine neural-differentiation model.
in this graphFIRST ARGININE MARK NODE. Before 2026-09-22 this corpus held 14 mark-* nodes and every one was a lysine or ubiquitin mark, while H3R2 and H4R3 were named across 11 entries with no node behind them.
absence classliterature-absence-bounded
absence class noteMIGRATED 2026-09-23, mechanically and not by judgement: the class string itself begins 'literature'. The field previously held the whole verdict as free text and that string is preserved here verbatim because it says more than the token can: 'literature - bounded with counts, querytranslations, warninglists, re-runs and an operand control'
writers
[
  "PRMT6 (major in vivo H3R2 methyltransferase; type I)"
]
erasers
[]
readers
[]
anti readers
[
  "WDR5 - binding to the H3 tail is PREVENTED by H3R2me2a and is permissive for me0/me1/me2s (PMID 32207970, 22231400, 18079182)",
  "the ASH2/WDR5/MLL-family methyltransferase complex - binding to the H3 tail prevented (PMID 17898714)",
  "Spp1 in S. cerevisiae (human counterpart CFP1/CXXC1, a SET1A/B-complex subunit and NOT a WRAD member) - occluded, blocking Set1 trimethylation (PMID 17898715)"
]
genomic distribution
{
  "guccione_2007_human": "gene body and 3' end regardless of transcriptional state; selectively and locally DEPLETED from active promoters coincident with H3K4me3",
  "kirmizis_2007_yeast": "throughout heterochromatic loci and inactive euchromatic genes; 3' end of moderately transcribed genes; mutually exclusive with H3K4me3",
  "bouchard_2018_mouse_neural": "at ACTIVE genes, at promoters AND enhancers; enhancer peaks co-localise with H3K4me1 and H3K27ac",
  "contradiction": "The 2007 pair and the 2018 study are not reconcilable as a single distribution. The graph holds both. UniProt's 'specific tag for epigenetic transcriptional repression' follows the 2007 pair and is contradicted by the 2018 enhancer arm."
}
effect on kmt2 family recruitment
{
  "kmt2a_wild_type_promoters": "loss of H3R2me2a ENHANCES KMT2A binding and H3K4me3 deposition (PMID 30232013) - so the mark restrains wild-type KMT2A",
  "kmt2d_enhancers": "loss of H3R2me2a REDUCES KMT2D binding and H3K4me1/H3K27ac (PMID 30232013) - so the mark supports KMT2D at enhancers",
  "mll1_wdr5_hoxa2": "PRMT6 gain reduces MLL1 and WDR5 occupancy and H3K4me3 at HoxA2 with reduced HOX transcription (PMID 18079182)",
  "kmt2a_fusion": "NEVER MEASURED, in any fusion, in any lineage - see literature_bounds"
}
literature bounds
{
  "date": "2026-09-22",
  "operand_control": {
    "query": "KMT2A[tiab]",
    "n": 1652
  },
  "counts": [
    {
      "query": "\"H3R2me2a\"[tiab]",
      "translation": "\"H3R2me2a\"[Title/Abstract]",
      "n": 47
    },
    {
      "query": "H3R2[tiab]",
      "translation": "\"H3R2\"[Title/Abstract]",
      "n": 75
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND KMT2A[tiab]",
      "translation": "\"H3R2me2a\"[Title/Abstract] AND \"KMT2A\"[Title/Abstract]",
      "n": 1,
      "ids": [
        "30232013"
      ]
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND MLL[tiab]",
      "n": 4,
      "ids": [
        "17898714",
        "22231400",
        "22777353",
        "27323813"
      ]
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND leukemia[tiab]",
      "n": 3,
      "ids": [
        "22777353",
        "36574771",
        "38424136"
      ]
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND WDR5[tiab]",
      "n": 4,
      "ids": [
        "17898714",
        "22231400",
        "22777353",
        "32207970"
      ]
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND \"MLL-AF4\"[tiab]",
      "translation": "\"H3R2me2a\"[Title/Abstract] AND \"MLL-AF4\"[Title/Abstract]",
      "n": 0,
      "warninglist": "phrasesignored [], quotedphrasesnotfound [], outputmessages [\"No items found.\"]",
      "rerun_n": 0
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND \"MLL-AF9\"[tiab]",
      "n": 0,
      "warninglist": "outputmessages [\"No items found.\"]",
      "rerun_n": 0
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND menin[tiab]",
      "n": 0,
      "warninglist": "outputmessages [\"No items found.\"]",
      "rerun_n": 0
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND revumenib[tiab]",
      "n": 0,
      "warninglist": "outputmessages [\"No items found.\"]",
      "rerun_n": 0
    },
    {
      "query": "\"H3R2me2a\"[tiab] AND antibod*[tiab]",
      "n": 1,
      "ids": [
        "17898715"
      ]
    }
  ],
  "identifier_trap": "For arginine marks the residue form and the compact form retrieve NEARLY DISJOINT PMID SETS, not nested ones. Measured 2026-09-22 by ID-set comparison, not by counts: H3R2[tiab] = 75; union of \"H3R2me2a\"/\"H3R2me2s\"/\"H3R2me1\"[tiab] = 78; the INTERSECTION is 15. 60 records are reachable only by the residue form and 63 only by a compact form (Jaccard 15/138 = 10.9%). Either form alone under-reports by roughly 5x. EVERY bounded absence on an arginine mark must run BOTH and take the union.",
  "bound_type": "titles and abstracts only; no Europe PMC full-text sweep was run, and this project has a measured precedent (SET-NUP214/H3K79) of a mark measurement hiding in a supporting figure"
}
How this could bite us back
  • Any type I PRMT inhibitor inhibits a BRAKE as well as a cofactor. PRMT6 writes H3R2me2a, which restrains wild-type KMT2A recruitment and lowers HOX transcription (PMID 18079182, 30232013); PRMT1 is treated in this graph as an MLL-fusion cofactor. A class inhibitor such as GSK3368715 (NCT03666988, terminated - benefit-risk did not support continuation, 31 participants) moves both, and the net direction on a HOXA program is not predictable from either arm alone. Nobody measured H3R2me2a in that trial.
  • A pharmacodynamic built on an H3R2 antibody could report the wrong sign. If a reagent cross-reacts between me2a and me2s, a drug that converts one state into the other shows NO CHANGE while the biology has inverted; and if it detects both, a real fall in the inhibitory state can be masked by a rise in the permissive one. A trial that declares target engagement on such a readout has measured nothing.
  • Because there is no eraser, any intervention aiming to REMOVE this block has no enzyme to agonise. The only handles are the writer, PRMT6 - not a DepMap dependency, no clinical agent, and predicted to be the wrong direction - and histone turnover, which is not druggable with specificity. An H3R2-based therapeutic hypothesis is a writer-inhibition hypothesis whether or not it is framed that way.
  • PRMT6 loss causes proliferation and differentiation defects in a normal neural-differentiation model (PMID 30232013) and PRMT6 has non-histone substrates including DNA polymerase beta, where its methylation stimulates base-excision repair (PMID 16600869, via gene-prmt6). Inhibiting it to move a histone mark also removes a DNA-repair stimulator, in a disease already treated with genotoxic chemotherapy.
Still open
  1. Is H3R2me2a present at KMT2A-fusion target loci at all? Nobody has measured it in any fusion, in any lineage. The cheapest version is mass-spectrometry quantification of bulk H3R2 methylation states in SEM or RS4;11 (KMT2A::AFF1) against a non-rearranged B-ALL line - no antibody required, one experiment, and it would either open this axis or close it.
  2. Does the mark restrain the FUSION as it restrains wild-type KMT2A? The fusion has no SET domain and no WIN motif, so the promoter result of PMID 30232013 - loss of H3R2me2a enhances KMT2A binding - may not transfer. If the fusion's chromatin engagement runs through CXXC/menin rather than through the H3 tail, H3R2me2a should restrain the WILD-TYPE allele and leave the fusion untouched, which is a testable asymmetry within one cell.
  3. Which of the H3R2 states does PRMT6 inhibition actually move, and in which direction does the leukaemia go? PRMT6 is not a DepMap dependency and PRMT6 gain LOWERS HOX transcription, so the predicted sign of PRMT6 inhibition in a HOXA-driven leukaemia is pro-leukaemic. No PRMT6-selective agent has reached a trial, but the type I PRMT class inhibitor that reached humans would hit PRMT6 as well as PRMT1 - and nobody measured H3R2me2a in that trial.
  4. Can the enhancer-versus-promoter sign inversion of PMID 30232013 be reproduced outside neural differentiation? It is the single most consequential result for this mark and it rests on one lineage. If it holds in a haematopoietic system, then 'H3R2me2a is repressive' has to be retired corpus-wide, including from UniProt's annotation.
Every source behind this entry (17)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt Q96LA8 (ANM6_HUMAN, PRMT6, 375 aa) records that PRMT6 specifically mediates the asymmetric dimethylation of histone H3 'Arg-2' to form H3R2me2a, and states that 'H3R2me2a represents a specific tag for epigenetic transcriptional repression and is mutually exclusive with methylation on histone H3 Lys-4 (H3K4me2 and H3K4me3)', citing PubMed 17898714 and 18077460; the same FUNCTION block also records that PRMT6 methylates histone H2A and H4 'Arg-3'.

    What this evidence cannot show

    This is a curated annotation, not a measurement, and the accession was fetched live rather than recalled. The phrase 'a specific tag for epigenetic transcriptional repression' is UniProt's editorial summary of two 2007 papers, both of which examined promoters; it is not a genome-wide statement and the curator's sentence is stronger than either source supports.

    Evidence that argues against this (1)
    • The annotation is contradicted by a genome-wide study of the same mark. Bouchard et al. find PRMT6-dependent H3R2me2a at ACTIVE genes at both promoters and enhancers, and at enhancers its loss REDUCES KMT2D binding and H3K4me1/H3K27ac with decreased transcription - i.e. an activating function for the same mark UniProt calls a specific repressive tag. A reader taking the UniProt sentence at face value would predict the wrong sign at half the genome. 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 levels in vivo; H3R2 methylation by PRMT6 was prevented by the presence of H3K4me3 on the H3 tail, and conversely the H3R2me2a mark prevented methylation of H3K4 as well as binding to the H3 tail by an ASH2/WDR5/MLL-family methyltransferase complex; ChIP showed H3R2me2a distributed within the body and at the 3' end of human genes regardless of transcriptional state, and selectively and locally depleted from active promoters coincident with H3K4me3.

    Methylation of histone H3R2 by PRMT6 and H3K4 by an MLL complex are mutually exclusive. · Guccione E, Bassi C, Casadio F, Martinato F, et al. · Nature 2007
    PMID 17898714DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. The complex-binding block is an in vitro result on H3 tails, not a demonstration on nucleosomes and not in a leukaemic cell. 'An ASH2/WDR5/MLL-family' complex is the paper's own wording and does not identify which KMT2 paralogue; nothing here is a measurement on a KMT2A FUSION protein, which has no SET domain and would not be recruited by the same logic. The promoter-depletion ChIP is 2007-era uncalibrated ChIP, which normalises away global changes by construction.

    Evidence that argues against this (1)
    • The genome-wide follow-up eleven years later disagrees about where the mark is and what it does: PRMT6-dependent H3R2me2a is found at active genes, and at enhancers it behaves as an activating mark whose loss reduces KMT2D recruitment. The mutual-exclusion mechanism can be correct while the promoter-centric distribution picture is incomplete. Bouchard C et al. 2018
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PRMT6 methylates histone H3 at R2 and histones H4/H2A at R3 in vitro and is the major in vivo H3R2 methyltransferase; H3R2 methylation inhibits H3K4 trimethylation and the recruitment of WDR5, a subunit of the MLL K4 methyltransferase complex, to histone H3 in vitro; on PRMT6 overexpression, HOX and Myc-dependent gene transcription decreases, coinciding at HoxA2 with increased H3R2 methylation and PRMT6 occupancy and reduced H3K4me3 and MLL1/WDR5 recruitment, while retinoic-acid-induced HoxA2 activation is accompanied by loss of PRMT6 and H3R2 methylation and a rise in H3K4me3.

    PRMT6-mediated methylation of R2 in histone H3 antagonizes H3 K4 trimethylation. · Hyllus D, Stein C, Schnabel K, Schiltz E, et al. · Genes Dev 2007
    PMID 18079182DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. The WDR5 block is in vitro on histone H3. The cellular arm is overexpression and knockdown in a neuronal differentiation model at a single locus (HoxA2), not a KMT2A-rearranged cell and not a fusion protein. 'MLL1/WDR5 recruitment' is wild-type MLL1; the leukaemic allele has no SET domain and no WIN motif.

    Evidence that argues against this (1)
    • Quantitative re-examination of the same interaction thirteen years later finds the WDR5 block is specific to the asymmetric state and BINARY, not a graded affinity effect: H3R2me0, me1 and me2s are equally permissive. Statements of the form 'H3R2 methylation inhibits WDR5 recruitment' are therefore too broad - H3R2me1 does not inhibit it, and PRMT5 makes H3R2me1 as well as me2s. Lorton BM et al. 2020
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    H3R2 is methylated in budding yeast, and by ChIP-on-chip with an antibody specific for H3R2me2a the mark is enriched throughout heterochromatic loci and inactive euchromatic genes and present at the 3' end of moderately transcribed genes, in a pattern mutually exclusive with H3K4me3; methylation at H3R2 abrogates H3K4 trimethylation by the Set1 methyltransferase, and the specific effect on H3K4me3 results from occlusion of Spp1, a Set1-complex subunit necessary for trimethylation.

    Arginine methylation at histone H3R2 controls deposition of H3K4 trimethylation. · Kirmizis A, Santos-Rosa H, Penkett CJ, Singer MA, et al. · Nature 2007
    PMID 17898715DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Saccharomyces cerevisiae. Yeast has no PRMT6 orthologue and no MLL-family enzyme - the writer is not identified in the abstract and the reader occluded is Spp1, whose human counterpart is CFP1/CXXC1, a SET1A/SET1B-complex subunit and NOT a WRAD component. So this paper supports the occlusion PRINCIPLE on the same residue and does not transfer the identity of the occluded subunit to human KMT2A. The antibody is described as specific but the abstract reports no competition panel against the symmetric state.

    Evidence that argues against this (1)
    • The subunit occluded in yeast, Spp1, maps to human CFP1/CXXC1, which this graph already records as a SET1-complex CXXC protein and NOT a member of WRAD. Transferring the yeast result onto the WRAD module would be a paralogue substitution of exactly the kind this project's cxxc-readers beat was written to prevent; the human WDR5 block is established separately by Guccione and Hyllus and should be cited from those. Hyllus D et al. 2007
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Quantitative binding analysis with high-resolution crystal structures of WDR5 in complex with unmodified, monomethyl and symmetric-dimethyl L-arginine and with the H3R2me1 peptide supports an interaction model regulated by a binary arginine methylation switch: H3R2me2a prevents interaction with WDR5, whereas H3R2me0, me1 and me2s are equally permissive; H3R2 methylarginine isoforms comprise PRMT5-catalysed me1 and me2s and PRMT6-catalysed me1 and me2a.

    A Binary Arginine Methylation Switch on Histone H3 Arginine 2 Regulates Its Interaction with WDR5. · Lorton BM, Harijan RK, Burgos ES, Bonanno JB, et al. · Biochemistry 2020
    PMID 32207970DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read, and the PDB accessions for the new structures were NOT fetched, so no structure id is asserted here. This is isolated recombinant WDR5 with free amino acids and short peptides - not WDR5 inside an assembled WRAD module, not on a nucleosome, and not in a cell. Affinity to a free peptide need not predict occupancy of chromatin by a six-subunit complex, and the paper's model does not address whether H3R2me2s changes WRAD residence time or RBBP7 competition, which is the other half of Migliori's claim and is not tested here.

    Evidence that argues against this (1)
    • The earlier study this revises reported, from SILAC MS profiling of H3-tail interactors plus a WDR5/H3R2me2s co-crystal, that H3R2me2s ENHANCES WDR5 binding and that the interaction distinguishes me2s from me2a - a graded rather than binary model. Both papers agree that me2a is the non-permissive state; they disagree about whether me2s is a gain over unmodified, and that disagreement decides whether a rise in H3R2me2s is predicted to raise WRAD occupancy or merely to restore it. Migliori V et al. 2012
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a murine neural-differentiation model with PRMT6 knockout, PRMT6-dependent H3R2me2a is detected 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 at enhancers H3R2me2a peaks co-localise with H3K4me1 and H3K27ac and loss of the mark reduces KMT2D binding and H3K4me1/H3K27ac deposition with decreased transcription - so the same mark exerts both repressive and activating functions depending on genomic location.

    PMID 30232013DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Murine neural differentiation, not haematopoiesis and not leukaemia; this is the ONLY record returned by 'H3R2me2a'[tiab] AND KMT2A[tiab]. The KMT2A and KMT2D ChIP are on wild-type alleles in a non-malignant differentiation model; no fusion protein is present, so the enhanced-KMT2A-binding result cannot be read as a statement about a fusion. A PRMT6 knockout removes the writer rather than the mark, so indirect effects through PRMT6's non-histone substrates (DNA polymerase beta, HMGA1, SIRT7, CRTC2) are not excluded by the design as described in the abstract.

    Evidence that argues against this (1)
    • This result contradicts both 2007 papers' distribution picture and the UniProt annotation, all of which place H3R2me2a on inactive chromatin and at the 3' end of genes and call it repressive. The enhancer arm in particular reverses the sign. Because this is a single study in one lineage while the repressive picture has two independent 2007 reports plus a yeast genome-wide map, the graph holds both and should not present either distribution as settled. Kirmizis A et al. 2007
  7. E7weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    JMJD6 was reported to be a JmjC-containing iron- and 2-oxoglutarate-dependent dioxygenase that demethylates histone H3 at arginine 2 (H3R2) and histone H4 at arginine 3 (H4R3) in both biochemical and cell-based assays - the first claimed eraser for either mark.

    JMJD6 is a histone arginine demethylase. · Chang B, Chen Y, Zhao Y, Bruick RK · Science 2007
    PMID 17947579DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Confidence is set to WEAK deliberately despite the venue, because the assignment has not held: see E8 and E9. This entry cites the paper to record what was claimed and by whom, NOT to assert that an eraser exists.

    Evidence that argues against this (1)
    • Two years later the same enzyme was shown to catalyse lysyl-5-hydroxylation of the splicing factor U2AF65, and a 2015 review of Jmjd6 titled 'a case study in conflicting assignments' states the arginine-demethylase role is subject to conflicting reports while the lysyl hydroxylase activity does occur. The graph should treat H3R2me2a as having NO established eraser. Webby CJ et al. 2009
  8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The splicing factor U2AF65 undergoes post-translational lysyl-5-hydroxylation catalysed by the Fe(II)- and 2-oxoglutarate-dependent dioxygenase Jmjd6, which changes alternative RNA splicing of some but not all endogenous and reporter genes - supporting a role for Jmjd6 in splicing regulation rather than in histone arginine demethylation.

    Jmjd6 catalyses lysyl-hydroxylation of U2AF65, a protein associated with RNA splicing. · Webby CJ, Wolf A, Gromak N, Dreger M, et al. · Science 2009
    PMID 19574390DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Demonstrating a hydroxylase activity does not formally exclude a second demethylase activity on a different substrate; what it does is remove the assumption that the 2007 assignment was correct, which is why E9 rather than this paper is cited for the field's current position. This is an in vitro and reporter-based study, not a chromatin measurement.

    Evidence that argues against this (1)
    • Detailed biochemistry shows that a subset of JmjC lysine demethylases CAN act as arginine demethylases in purified form, on histone and non-histone fragments, with formaldehyde release - so arginine demethylation is chemically possible and the question is whether any of it is biologically relevant. That paper's own framing is that bona fide arginine demethylases remain controversial. Walport LJ et al. 2016
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    A review of Jmjd6 assignments states that Jmjd6 was initially assigned a role as the phosphatidylserine receptor, which now seems unlikely, and was subsequently assigned a role in catalysing N-methyl-arginine demethylation on the N-termini of human histones H3 and H4, but that this function 'is also subject to conflicting reports'; what Jmjd6 does catalyse is 2-oxoglutarate-dependent C-5 hydroxylation of lysine residues in mRNA splicing-regulatory proteins and histones.

    The oxygenase Jmjd6 - a case study in conflicting assignments. · Böttger A, Islam MS, Chowdhury R, Schofield CJ, et al. · Biochem J 2015
    PMID 25997831DOI
    What this evidence cannot show

    A review, cited as a map of the controversy rather than as primary evidence, per the QUEEN_BRIEF rule on reviews. Abstract read from the eutils record 2026-09-22. It is ten years old; nothing found in this beat's searches has since assigned a cellular arginine demethylase for H3R2, but a review cannot bound the subsequent literature and this entry does not claim it does.

    Evidence that argues against this (1)
    • The original demethylase claim is a Science paper with biochemical and cell-based assays and has not been formally retracted. A reader should treat 'no established eraser' as the field's current position rather than as a refutation, and note that this graph's confidence assignment (weak on E7) is a judgement drawn from the conflict, not from a withdrawal. Chang B et al. 2007
  10. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed esearch, run 2026-09-22 with querytranslation and warninglist read back for every query and every zero re-run: 'H3R2me2a'[tiab] AND KMT2A[tiab] = 1 (PMID 30232013); AND MLL[tiab] = 4; AND leukemia[tiab] = 3; AND 'MLL-AF4'[tiab] = 0; AND 'MLL-AF9'[tiab] = 0; AND menin[tiab] = 0; AND revumenib[tiab] = 0; AND WDR5[tiab] = 4; AND antibod*[tiab] = 1. Operand control KMT2A[tiab] = 1652. Every zero returned a non-empty translation with warninglist.outputmessages = ['No items found.'], and the same zero on re-run.

    What this evidence cannot show

    A PubMed zero bounds TITLES AND ABSTRACTS only. An H3R2me2a ChIP or mass-spectrometry panel could exist as a supporting figure or supplementary table in a KMT2A-fusion paper and be invisible to every query here; this beat did not run a Europe PMC full-text sweep. The non-zero counts are raw and were title-checked but not read in full. The 4 'AND MLL[tiab]' records are 17898714, 22231400, 22777353 and 27323813, none of which is a leukaemia chromatin measurement - 27323813 is prostate cancer.

    Evidence that argues against this (1)
    • This project has a measured precedent for exactly this absence being false: a bounded SET-NUP214 query for H3K79 returned a clean zero with four spellings while PMID 18299449 contains an H3K79me2 ChIP in Figure 6A. The correct reading of these zeros is 'no paper announces the measurement', not 'the measurement does not exist'. Before any entry leans on them, the founding KMT2A-fusion chromatin papers should be opened and their supplementary tables searched for arginine marks. Bernt KM et al. 2011
  11. E11moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    H3R2 is symmetrically dimethylated by PRMT5 and PRMT7 to give H3R2me2s, present in euchromatic regions; SILAC MS profiling of H3-tail interactors showed H3R2me2s excludes binding of RBBP7, a central component of the co-repressor complexes Sin3a, NURD and PRC2, and conversely enhances binding of WDR5, a common component of the coactivator complexes MLL, SET1A, SET1B, NSL1 and ATAC; the interaction of histone H3 with WDR5 distinguishes H3R2me2s from H3R2me2a, which impedes the recruitment of WDR5 to chromatin.

    Symmetric dimethylation of H3R2 is a newly identified histone mark that supports euchromatin maintenance. · Migliori V, Müller J, Phalke S, Low D, et al. · Nat Struct Mol Biol 2012
    PMID 22231400DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. HeLa and NT2 differentiation, not haematopoiesis. Confidence is MODERATE rather than established for the 'enhances WDR5' half specifically, because a later quantitative and crystallographic study (E5) reports me2s is merely permissive rather than enhancing; the 'me2a impedes WDR5' half is corroborated by three other papers and is established. The RBBP7 exclusion has not, as far as this beat found, been independently re-tested.

    Evidence that argues against this (1)
    • The 2020 binary-switch study reports H3R2me0, me1 and me2s are equally permissive for WDR5, which removes the gain-of-function reading of H3R2me2s. This matters here because it is the half of the two-way switch this graph's PRMT5 entry leans on. Lorton BM et al. 2020
  12. E12established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    More than 200 antibodies raised against 57 different histone modifications were tested in Drosophila, C. elegans and human cells; although most performed well, more than 25% failed specificity tests by dot blot or western blot, and among the specific antibodies more than 20% failed in chromatin immunoprecipitation - so rigorous per-reagent testing is required before use.

    An assessment of histone-modification antibody quality. · Egelhofer TA, Minoda A, Klugman S, Lee K, et al. · Nat Struct Mol Biol 2011
    PMID 21131980DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. This is a base-rate statement across 57 modifications and does NOT report the failure rate for H3R2 methylarginine antibodies specifically, nor whether any tested antibody was assayed for me2a-versus-me2s discrimination. It is cited to establish that assuming specificity is unsafe, not to assert that a particular H3R2 reagent fails.

    Evidence that argues against this (1)
    • Most antibodies in that panel performed well, so a blanket presumption of failure would also be wrong. The actionable form of this constraint is per-reagent: check the peptide-array behaviour of the specific lot in the Histone Antibody Specificity Database or run the competition panel, and report the clone. Rothbart SB et al. 2015
  13. E13established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The Histone Antibody Specificity Database (http://www.histoneantibodies.com) is an online, expanding resource cataloguing the behaviour of widely used commercially available histone antibodies by peptide microarray.

    An Interactive Database for the Assessment of Histone Antibody Specificity. · Rothbart SB, Dickson BM, Raab JR, Grzybowski AT, et al. · Mol Cell 2015
    PMID 26212453DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read, and THE DATABASE ITSELF WAS NOT QUERIED IN THIS BEAT. This entry therefore does not assert what the database says about any H3R2 or H4R3 reagent - only that the resource exists and is the right place to look. A later beat should query it for the arginine-mark antibodies and record what it finds, because that is the measurement that would convert this from a general caution into a specific one.

    Evidence that argues against this (1)
    • A peptide microarray reports binding to a synthetic peptide, which is not the same question as behaviour on a nucleosome in a cell. Internally calibrated ChIP with nucleosome standards is the stronger test and gives different answers for some reagents that pass on peptides. Shah RN et al. 2018
  14. E14established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Fifty-two commercial antibodies purported to distinguish the H3K4 methylforms were characterised by histone peptide array and internally calibrated ChIP (ICeChIP); many widely used antibodies poorly distinguish the methylforms, and high- versus low-specificity reagents yield dramatically different biological interpretations, producing substantial divergence from the literature for numerous H3K4 methylform paradigms.

    Examining the Roles of H3K4 Methylation States with Systematically Characterized Antibodies. · Shah RN, Grzybowski AT, Cornett EM, Johnstone AL, et al. · Mol Cell 2018
    PMID 30244833DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. This is about H3K4 methylforms, NOT about H3R2 methylarginine isomers - it is the mark two residues away, cited to establish that discrimination failures among closely related methyl states of one residue are common and consequential. It does not report H3R2 cross-reactivity, and no claim here says it does.

    Evidence that argues against this (1)
    • Because this study is about a lysine and about me1/me2/me3 rather than about symmetric versus asymmetric dimethylarginine, it is an argument by analogy. It would be refuted as a constraint on this beat if a published panel showed H3R2me2a and H3R2me2s antibodies discriminate cleanly; no such panel was found (1 record for 'H3R2me2a'[tiab] AND antibod*[tiab]), so the constraint stands as unresolved rather than proven. NCBI PubMed 2026
  15. E15moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Detailed biochemical studies show that, in purified form, a subset of JmjC lysine demethylases can also act as arginine demethylases on both histone and non-histone fragments with release of formaldehyde, although the authors state that the existence of bona fide N-omega-methylarginine demethylases is controversial and that their result should stimulate efforts to detect biologically relevant activity.

    Arginine demethylation is catalysed by a subset of JmjC histone lysine demethylases. · Walport LJ, Hopkinson RJ, Chowdhury R, Schiller R, et al. · Nat Commun 2016
    PMID 27337104DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Purified enzymes and peptide substrates, including artificial sequences in which the enzyme's target lysine was substituted by a methylated arginine - i.e. an in vitro capability demonstration, explicitly not a cellular assignment. Which JmjC enzymes, and whether any acts on H3R2me2a in chromatin, is not asserted here.

    Evidence that argues against this (1)
    • A purified-enzyme capability on a designed peptide is the weakest form of evidence for a cellular eraser, and the field has once already been misled by exactly that pattern with JMJD6. The graph should continue to record H3R2me2a as having no established eraser until a cellular loss-of-enzyme experiment shows the mark rising. Böttger A et al. 2015
  16. E16established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Human peptidylarginine deiminase 4 (PAD4) regulates histone arginine methylation by converting methyl-arginine to citrulline and releasing methylamine, targeting multiple sites in histones H3 and H4 including those methylated by CARM1 (H3 Arg-17) and PRMT1 (H4 Arg-3); a decrease in histone arginine methylation with a concomitant increase in citrullination requires PAD4 activity in human HL-60 granulocytes.

    Human PAD4 regulates histone arginine methylation levels via demethylimination. · Wang Y, Wysocka J, Sayegh J, Lee YH, et al. · Science 2004
    PMID 15345777DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Demethylimination is NOT demethylation - it destroys the arginine rather than restoring it, so it cannot reset the residue to the unmodified state and produces a third chemical species, citrulline, with its own consequences. The abstract does not state which methylation degree is the substrate; the companion Cell paper (E17) does, and it excludes the dimethyl states.

    Evidence that argues against this (1)
    • The same-year companion study measured the substrate requirement directly and found dimethylation of arginines PREVENTS deimination by PADI4 while monomethylation still allows it. So this route cannot remove H3R2me2a or H3R2me2s, and citing PAD4 as an eraser of the dimethyl marks would be wrong. Cuthbert GL et al. 2004
  17. E17established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Peptidyl arginine deiminase 4 (PADI4) specifically deiminates arginine residues R2, R8, R17 and R26 in the H3 tail and thereby prevents arginine methylation by CARM1; dimethylation of arginines PREVENTS deimination by PADI4, although monomethylation still allows deimination to take place, and PADI4 is recruited to the pS2 promoter following hormone induction when the gene is transcriptionally downregulated.

    Histone deimination antagonizes arginine methylation. · Cuthbert GL, Daujat S, Snowden AW, Erdjument-Bromage H, et al. · Cell 2004
    PMID 15339660DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. The deimination-substrate result is in vitro on H3 tail peptides/histones; the cellular arm is a single oestrogen-responsive promoter in MCF-7, not haematopoietic and not KMT2A-related. H3R2 is named among the deiminated residues, which makes the monomethyl-only constraint directly relevant to this mark. Nothing here addresses H4R3.

    Evidence that argues against this (1)
    • Because deimination acts only up to the monomethyl state, PADI4 activity at H3R2 would deplete the pool of substrate available to PRMT6 rather than remove existing H3R2me2a - a competing-writer effect, not an eraser. An experiment that raises PADI4 and sees H3R2me2a fall would therefore be seeing reduced deposition, and could not distinguish that from demethylation without a pulse-chase or a mass-spectrometry read of the citrulline species. Wang Y et al. 2004
Connected to