Oncofusion

start / graph / mark-h4r3me2s

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

H4R3me2s - the repressive symmetric arginine mark that recruits DNMT3A, blocks a KMT2 enzyme's own PHD module, and is the ONLY arginine mark ever measured in a KMT2A-fusion system

also called H4R3me2s · symmetric dimethylation of histone H4 arginine 3 · H4 Arg-3 symmetric dimethylation · H4R3me2s/H2AR3me2s

H4R3me2s is a silencing tag on histone H4 that physically pulls in the enzyme which adds methyl groups to DNA itself, locking a gene off. It is also the one arginine tag anybody has actually measured in a cell carrying a KMT2A fusion - it fell when the fusion's partner complex was removed.

The same thing, for a scientist

H4R3me2s is the symmetric dimethyl mark on histone H4 arginine 3. It is **repressive by a defined mechanism**: PRMT5-deposited H4R3me2s is a **direct binding target for DNMT3A**, engaged through the ADD domain containing the PHD motif, and is required for subsequent DNA methylation at the human beta-globin locus - knocking down PRMT5 loses the mark, loses DNMT3A binding, loses DNA methylation and activates the gene [E1]. In **primary erythroid progenitors from adult human bone marrow** the mark sits on the inactive methylated globin genes coincident with PRMT5 [E1], which makes it the only one of these four arginine marks with a primary human haematopoietic measurement. Its second mechanism is the trans-tail one: **H4R3me2s inhibits binding of the tandem PHD(4-6) of KMT2D**, the module KMT2D needs for nucleosomal methyltransferase activity, and **PRMT7 - but not PRMT5 - represses KMT2D target genes by raising H4R3me2s**, with PRMT7 knockdown increasing KMT2D-catalysed H3K4me3 and falling H4R3me2s tracking rising H3K4me3 at many HOXA and HOXB genes during differentiation [E2]. UniProt assigns H4R3me2s deposition specifically to **PRMT7** and separately records PRMT5's symmetric histone activities [E3][E4]. **And this is the one arginine mark with a KMT2A-fusion datum**: in murine MLL-AF9 leukaemia, PRMT5-mediated H4R3me2s **falls** after conditional excision of the PAF-complex subunit Cdc73, placing the mark downstream of a complex that contacts MLL fusion proteins directly [E5]. `"H4R3me2s"[tiab] AND KMT2A[tiab]` returns exactly that one record [E6]. **No eraser is established** [E7][E8].

Go deeper
The one arginine-mark measurement that exists in a KMT2A-fusion cell, and exactly what it does and does not say

Across all four marks in this beat, one record pairs an arginine mark with KMT2A in a title or abstract in a leukaemic system: PMID 28945229 5.

What it says. Conditional excision of Cdc73 in murine MLL-AF9 leukaemia reduces expression of an early haematopoietic progenitor program that prevents differentiation. Prmt5 is a direct PAFc target, positively regulated by a transcriptional unit comprising the PAFc, MLL1, HOXA9 and STAT5 in leukaemic cells. PRMT5-mediated H4R3me2s is reduced following excision of Cdc73, placing the histone modification downstream of the PAFc. Prmt5 knockdown or pharmacologic inhibition causes G1 arrest and extends leukaemic latency in vivo.

What it does not say, and the distinction is the whole value of recording it here.

  • It is not a locus-resolved measurement at fusion target genes. H4R3me2s is reported as falling; the abstract does not place it at HOXA9, MEIS1 or any fusion-bound site.
  • It is downstream of the writer's transcription, not downstream of the fusion's chromatin activity. The chain is PAFc → Prmt5 gene expression → PRMT5 protein → H4R3me2s. Removing Cdc73 lowers the enzyme; the mark follows. That is a gene-dosage result, not a demonstration that the fusion controls the mark on chromatin.
  • It is murine MLL-AF9, myeloid. The index fusion of this project is human KMT2A::AFF1 in infant B-lineage disease, and this graph has measured that 0 of 24 sampled load-bearing experiments rest on a dedicated human KMT2A::AFF1 experiment.
  • The mark's direction is repressive, and it FALLS when the leukaemic program is disrupted - so lower H4R3me2s accompanies differentiation here, consistent with the mark's repressive assignment and with PMID 23249737's differentiation series.

So the field's single arginine-mark datum in this disease is a writer-abundance result in a mouse myeloid model, and every entry that cites it should say so in the claim rather than the caveat.

Two mechanisms, and they point the same way

(1) DNA methylation coupling. H4R3me2s is a direct docking site for DNMT3A through its ADD domain (the domain containing the PHD motif). The order is established in the paper's own framing: the histone mark is required for subsequent DNA methylation, not the reverse - and DNMT3A is therefore "both a reader and a writer of repressive epigenetic marks" 1. Loss of the mark by PRMT5 knockdown loses DNMT3A binding, loses DNA methylation, and activates the gene.

gene-dnmt3a exists in this graph. Nothing in it reasons about an arginine mark, and the DNMT3A reader function here is a mechanism by which a PRMT can hand a locus to the DNA-methylation machinery - a route from the arginine layer into a layer this project already treats as central.

(2) KMT2 blockade. H4R3me2s inhibits the tandem PHD(4-6) of KMT2D, which that enzyme requires for nucleosomal H3K4 methylation; PRMT7 raises the mark and represses KMT2D targets; PRMT7 knockdown raises KMT2D-catalysed H3K4me3; and during NT2/D1 differentiation falling H4R3me2s tracks rising H3K4me3 at many HOXA and HOXB genes 2.

Both mechanisms are repressive, both are independent of menin and of the WIN motif, and the second one is independent of WDR5 as well - it acts on a PHD cassette inside the catalytic subunit. Three separate control points on KMT2-family catalysis now exist in this graph: the menin tether, the WIN/WDR5 contact, and the enzyme's own reading of the H4 tail.

Writers: PRMT5 and PRMT7, and the division of labour is measured rather than assumed

This is the one place in the beat where the two type-II-ish enzymes are experimentally separated rather than lumped.

contextwritersource
human beta-globin locus, K562 and primary adult BM erythroid progenitorsPRMT5 - shRNA knockdown loses the mark1
murine MLL-AF9 leukaemiaPRMT5 - the mark falls when Prmt5 transcription falls5
KMT2D target genes, NT2/D1 differentiationPRMT7, and explicitly NOT PRMT52
UniProt curated functionPRMT7 "specifically mediates the symmetric dimethylation of histone H4 Arg-3 to form H4R3me2s"; recruited by CTCFL at the H19 imprinted control region3
UniProt curated functionPRMT5's H4R3me2s and H3R8me2s activities are annotated BY SIMILARITY, i.e. weaker in human cells than the textbook implies (recorded in gene-prmt5)4

The consequence is a specific and cheap prediction. If the locus-dependent division holds - PRMT5 at globin and in MLL-AF9, PRMT7 at KMT2D targets - then a PRMT5 inhibitor and a PRMT7 loss should move H4R3me2s at different loci in the same cell. A class-level statement about "symmetric arginine methylation" cannot capture that, and a PRMT5 inhibitor would not be expected to relieve the KMT2D blockade at all.

And it sharpens the PRMT7 question this beat inherited. The dispatch brief described PRMT7 as a second H3R2me2s writer. That attribution rests on one sentence in PMID 22231400, is contradicted in principle by a same-year study finding PRMT7 to be a type III monomethyl-only enzyme 9, and is absent from UniProt. The PRMT7 mark that UniProt names, and that a cellular knockdown experiment supports, is H4R3me2s - this one. So PRMT7 belongs in the graph, and it belongs here rather than on the H3R2 entry.

The WDR5/globin thread, which runs the other way and is worth keeping

One further haematopoietic result touches this mark from an unexpected direction 10. In K562 and primary human bone-marrow erythroid progenitors, WDR5 interacts with PRMT5 - a known repressor of gamma-globin - and is required for H3K4me3 at the gamma-globin promoter; enforced WDR5 reduces gamma-globin expression and WDR5 knockdown increases it, with histone H4R3 and H3K9 methylation decreased in the knockdown. WDR5 was found to interact with HDAC1 and with ING2, an H3K4me3 reader, to enhance repression.

Two reasons to record it.

  1. It is a WDR5 result in which WDR5 is a repressor, in primary human erythroid cells, with an arginine mark in the mechanism. This graph already carries concept-a-wdr5-result-is-not-automatically-a-kmt2a-axis-result; this is a clean example, and it is about the mark this entry is about.
  2. ING2 has an entry here (gene-ing2, written the same cycle by the q101 beat) and this is the paper that puts ING2 in a PRMT5/WDR5/H4R3 complex. The thread already exists in the corpus and nothing connected it to an arginine mark, because there was no arginine mark node.

The limit. The H4R3 methylation change is reported without a degree assignment in the abstract, so whether it is the symmetric state specifically is not established by this paper; it is cited for the axis, not for the isomer.

Bounded absence, with translations, warninglists, re-runs and the symbol-scoping control

Run 2026-09-22, translation read back from each response, every zero re-run identical.

querynids / warninglist
"H4R3me2s"[tiab]92-
H4R3[tiab]109-
"H4R3me2s"[tiab] AND KMT2A[tiab]128945229
"H4R3me2s"[tiab] AND MLL[tiab]128945229
"H4R3me2s"[tiab] AND leukemia[tiab]423249737, 28945229, 30930442, 32042864
"H4R3me2s"[tiab] AND "MLL-AF9"[tiab]0outputmessages: ["No items found."]; re-run 0
"H4R3me2s"[tiab] AND menin[tiab]0same; re-run 0
H4R3[tiab] AND KMT2A[tiab]0same; re-run 0
H4R3[tiab] AND MLL[tiab]317891136, 22689669, 26766589
"H4R3me2s"[tiab] AND antibod*[tiab]321072194, 25092918, 36230689
operand control KMT2A[tiab]1652

Note the shape of the one-record result. "H4R3me2s" AND KMT2A = 1 and "H4R3me2s" AND "MLL-AF9" = 0, and they are the same paper - it is a murine MLL-AF9 study whose abstract writes "MLL fusion" and "KMT2A" but not "MLL-AF9". A bounded absence written against the fusion NAME would have returned zero and been false, which is rule 71/72's shape exactly and is why the name list must include the modern symbol, the historical symbol and the fusion aliases.

Identifier trap, on ID sets. H4R3[tiab] = 109; union of "H4R3me2a", "H4R3me2s", "H4R3me1"[tiab] = 152; intersection 24 (Jaccard 10.1%). 75 of the 92 "H4R3me2s" records are invisible to H4R3[tiab]. The two spellings are nearly disjoint in both directions; neither bounds the other.

Facts and figures
markH4R3me2s
histoneH4 (and H2A R3, the same N-terminal sequence)
residueR3 (Arg-3), N-terminal tail
chemistrysymmetric omega-N,N'-dimethylarginine (sDMA) - one methyl on each terminal guanidino nitrogen. Isobaric with H4R3me2a, and the two have OPPOSITE effects on the KMT2D PHD(4-6) module.
writer division of labourMEASURED, NOT ASSUMED, and locus-dependent: PRMT5 at globin and in MLL-AF9; PRMT7 and NOT PRMT5 at KMT2D target genes (PMID 23249737). PREDICTION: a PRMT5 inhibitor and a PRMT7 loss should move this mark at DIFFERENT loci in the same cell, and a PRMT5 inhibitor should not relieve the KMT2D blockade. Class-level reasoning about 'symmetric arginine methylation' cannot represent this.
eraser statusNO ESTABLISHED ERASER. JMJD6 was the only claim, for H3R2 and H4R3 (PMID 17947579), and was reassigned as a lysyl-5-hydroxylase (PMID 19574390; review PMID 25997831). PADI4 citrullination cannot remove a dimethyl state (PMID 15339660). Purified JmjC KDMs show in vitro arginine-demethylase capability with bona fide RDMs called controversial by their own authors (PMID 27337104). Removal is by histone turnover or nucleosome eviction - the same structural consequence as for the other three arginine marks in this beat, and it means no intervention can 'erase' this mark enzymatically.
kmt2a fusion datumTHE ONLY ONE IN THE WHOLE BEAT. PMID 28945229: PRMT5-mediated H4R3me2s is REDUCED following conditional excision of Cdc73 in murine MLL-AF9 leukaemia, placing the mark downstream of the PAF complex. READ THE LIMITS INTO THE CLAIM: murine, myeloid, MLL-AF9 and not the index fusion; a bulk/global reduction and NOT a locus-resolved measurement at fusion targets; and mechanistically a writer-ABUNDANCE result, because the chain runs PAFc -> Prmt5 transcription -> PRMT5 protein -> the mark. It is not a demonstration that the fusion controls H4R3me2s on chromatin.
antibody constraintLOAD-BEARING. H4R3me2s and H4R3me2a are isobaric and have opposite effects on KMT2D's PHD(4-6), so an assay that cannot separate them cannot test either direction. Every distribution and perturbation result carried on this entry is antibody-based and none reports an isomer-competition panel: PMID 19234465, 23249737, 28945229, 22689669 and 21072194 alike. 'H4R3me2s'[tiab] AND antibod*[tiab] = 3 (21072194, 25092918, 36230689) and none is a specificity study of this reagent class. Base rates: >25% of 200+ histone-modification antibodies fail specificity by dot/western and >20% of the specific ones fail in ChIP (PMID 21131980); 52 antibodies against the H3K4 methylforms often fail to distinguish states and change the conclusion (PMID 30244833). The Histone Antibody Specificity Database (PMID 26212453) exists and was NOT queried in this beat. Isomer-resolving mass spectrometry is the arbiter.
relationship to menin and the win motifINDEPENDENT OF BOTH, and of WDR5 in the KMT2D case - PHD(4-6) is a module inside the catalytic subunit. The DNMT3A arm is independent of KMT2 entirely and runs into the DNA-methylation layer. So this mark supplies two menin-independent control points, not one.
cell type and stateMeasured in: K562 and PRIMARY human adult bone-marrow erythroid progenitors, on the inactive methylated globin genes (PMID 19234465) and at the gamma-globin promoter (PMID 22689669); murine MLL-AF9 leukaemic cells (PMID 28945229); NT2/D1 differentiation (PMID 23249737); murine neural precursors, where it predominates in the UNDIFFERENTIATED, symmetrically dividing neuroepithelium (PMID 21072194). Cell STATE: consistently associated with the UNDIFFERENTIATED or repressed state, falling as differentiation proceeds. NOT measured in fetal liver HSC, LMPP, pro-B, or any human KMT2A-rearranged line or patient sample.
shown for fusionPARTIAL, and it should be named rather than graded: MURINE MLL-AF9, as a bulk reduction downstream of Cdc73 excision and of Prmt5 transcription (PMID 28945229). NOTHING in human KMT2A::AFF1, nothing in any human KMT2A-rearranged line or patient sample, and no locus-resolved measurement at any fusion target in any species.
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, an operand control and an ID-set comparison'
writers
[
  "PRMT5 - established at the human beta-globin locus in K562 and primary adult bone-marrow erythroid progenitors (PMID 19234465), and the writer whose abundance sets the mark in murine MLL-AF9 (PMID 28945229); UniProt annotates PRMT5's H4R3me2s activity BY SIMILARITY",
  "PRMT7 - the writer UniProt names specifically for H4R3me2s (Q9NVM4), and the one that raises the mark at KMT2D target genes where PRMT5 explicitly does not (PMID 23249737)"
]
erasers
[]
readers
[
  "DNMT3A - direct binding target; engaged through the ADD domain containing the PHD motif; required for subsequent DNA methylation at the beta-globin locus (PMID 19234465). This is the strongest reader assignment of any of the four arginine marks in this beat."
]
anti readers
[
  "KMT2D tandem PHD(4-6) - binding INHIBITED by H4R3me2s; the module is required for KMT2D's nucleosomal methyltransferase activity (PMID 23249737)"
]
mechanisms
{
  "dna_methylation_coupling": "H4R3me2s -> DNMT3A ADD/PHD binding -> DNA methylation -> silencing; PRMT5 knockdown reverses the whole chain and activates the gene (PMID 19234465)",
  "kmt2_blockade": "H4R3me2s -> inhibited KMT2D PHD(4-6) binding -> reduced nucleosomal H3K4 methylation; PRMT7 knockdown raises KMT2D-catalysed H3K4me3; falling H4R3me2s tracks rising H3K4me3 at many HOXA and HOXB genes during differentiation (PMID 23249737)"
}
literature bounds
{
  "date": "2026-09-22",
  "operand_control": {
    "query": "KMT2A[tiab]",
    "n": 1652
  },
  "counts": [
    {
      "query": "\"H4R3me2s\"[tiab]",
      "translation": "\"H4R3me2s\"[Title/Abstract]",
      "n": 92
    },
    {
      "query": "H4R3[tiab]",
      "translation": "\"H4R3\"[Title/Abstract]",
      "n": 109
    },
    {
      "query": "\"H4R3me2s\"[tiab] AND KMT2A[tiab]",
      "translation": "\"H4R3me2s\"[Title/Abstract] AND \"KMT2A\"[Title/Abstract]",
      "n": 1,
      "ids": [
        "28945229"
      ]
    },
    {
      "query": "\"H4R3me2s\"[tiab] AND MLL[tiab]",
      "n": 1,
      "ids": [
        "28945229"
      ]
    },
    {
      "query": "\"H4R3me2s\"[tiab] AND leukemia[tiab]",
      "n": 4,
      "ids": [
        "23249737",
        "28945229",
        "30930442",
        "32042864"
      ]
    },
    {
      "query": "\"H4R3me2s\"[tiab] AND \"MLL-AF9\"[tiab]",
      "n": 0,
      "warninglist": "phrasesignored [], quotedphrasesnotfound [], outputmessages [\"No items found.\"]",
      "rerun_n": 0,
      "note": "FALSE-ABSENCE TRAP: the single KMT2A record IS an MLL-AF9 study whose abstract never writes MLL-AF9"
    },
    {
      "query": "\"H4R3me2s\"[tiab] AND menin[tiab]",
      "n": 0,
      "warninglist": "outputmessages [\"No items found.\"]",
      "rerun_n": 0
    },
    {
      "query": "H4R3[tiab] AND KMT2A[tiab]",
      "n": 0,
      "warninglist": "outputmessages [\"No items found.\"]",
      "rerun_n": 0
    },
    {
      "query": "H4R3[tiab] AND MLL[tiab]",
      "n": 3,
      "ids": [
        "17891136",
        "22689669",
        "26766589"
      ]
    },
    {
      "query": "\"H4R3me2s\"[tiab] AND antibod*[tiab]",
      "n": 3,
      "ids": [
        "21072194",
        "25092918",
        "36230689"
      ]
    }
  ],
  "identifier_trap": "ID-SET comparison, retmax=500 so nothing truncated: H4R3[tiab] = 109; union of \"H4R3me2a\"/\"H4R3me2s\"/\"H4R3me1\"[tiab] = 152; INTERSECTION = 24 (Jaccard 10.1%). 75 of the 92 \"H4R3me2s\" records are NOT returned by H4R3[tiab]; 85 H4R3[tiab] records are returned by no compact spelling. Neither form bounds the other.",
  "bound_type": "titles and abstracts only"
}
How this could bite us back
  • PRMT5 inhibitors have reached myeloid patients and the trial stopped: NCT03614728 (GSK3326595 with and without 5-azacitidine, phase 1/2, MDS and AML) was TERMINATED with 30 actual participants 'due to internal review of clinical data in context of indication under investigation'. Any programme proposing to lower H4R3me2s pharmacologically in leukaemia is re-entering a space where the class has already failed to clear a benefit-risk review in this exact indication.
  • PRMT5 is DepMap near-pan-essential (mean Chronos -1.1947 across 1,258 lines, 94.8% of lines depleted, per gene-prmt5) and its dominant measured activities are methylosome function and snRNP core assembly. UniProt annotates its H4R3me2s activity only BY SIMILARITY. So the most likely mechanism of any anti-leukaemic effect is splicing, not this mark, and a pharmacodynamic built on H4R3me2s would be measuring a bystander while the therapeutic window is set by an essential housekeeping function.
  • The direction may be actively wrong. This mark is repressive, falls as cells differentiate, and falls when the leukaemic program is disrupted by Cdc73 loss. Lowering it further is therefore the direction the disease already moves in when it responds - which sounds favourable until the DNMT3A arm is considered: lowering H4R3me2s removes a DNA-methylation targeting signal, and DNMT3A-pathway perturbation in myeloid disease is not a benign axis.
  • If PRMT7 rather than PRMT5 is the writer at KMT2-family target genes, then the available drug hits the wrong enzyme for the mechanism that matters here, and no PRMT7-selective clinical agent exists. A combination built on 'inhibit symmetric arginine methylation' would engage the essential enzyme and miss the relevant one.
Still open
  1. Is H4R3me2s present at KMT2A-fusion target loci, and is it locus-resolved anywhere? The one fusion datum is a bulk reduction in a mouse. A CUT&RUN or calibrated ChIP at HOXA9, MEIS1 and a menin-inhibitor-responsive gene set in SEM or MV4;11, or better an isomer-resolving mass-spectrometry read on fractionated chromatin, would be the first locus-level arginine-mark measurement in this disease.
  2. Does the DNMT3A hand-off operate at fusion target genes? DNMT3A reads this mark through its ADD/PHD domain and couples it to DNA methylation, DNMT3A has an entry in this graph, and nothing in the corpus connects the two layers through an arginine mark. If it operates, PRMT5 inhibition should reduce DNA methylation at those loci - a prediction with an existing drug and an existing assay.
  3. Does PRMT7 loss relieve a KMT2-family blockade in a KMT2A-rearranged cell, as it does for KMT2D in NT2/D1? PRMT7 is not a DepMap dependency (mean gene effect -0.1513 across 1,258 lines, 1.67% of lines depleted below -0.5; lymphoid tissue mean -0.1039 over 97 lines, fetched 2026-09-22) and had no entry in this graph before this beat, so nobody has looked.
  4. Which isomer does each published H4R3 antibody actually see? Every result on this entry depends on the answer and none of the papers reports it. Querying the Histone Antibody Specificity Database for the H4R3me2s and H4R3me2a reagents is a desk job of under an hour and would either license or retire a decade of distribution claims.
Every source behind this entry (11)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Using the human beta-globin locus as a model, symmetric methylation of histone H4 arginine 3 (H4R3me2s) by PRMT5 is required for subsequent DNA methylation; H4R3me2s serves as a direct binding target for DNMT3A, which interacts through the ADD domain containing the PHD motif, and loss of the mark by shRNA knockdown of PRMT5 leads to reduced DNMT3A binding, loss of DNA methylation and gene activation; in primary erythroid progenitors from adult bone marrow, H4R3me2s marks the inactive methylated globin genes coincident with localisation of PRMT5, defining DNMT3A as both a reader and a writer of repressive marks.

    PMID 19234465DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. One locus, the beta-globin cluster, in K562 plus primary adult bone-marrow erythroid progenitors - genuinely primary and genuinely human haematopoietic, but erythroid and not lymphoid, and with no leukaemic fusion present. The chromatin readout is antibody-based and no clone or isomer-specificity panel is named in the abstract, which for an isobaric arginine isomer is the binding constraint. The epistatic ordering (histone mark before DNA methylation) rests on a knockdown, so an indirect route through PRMT5's many non-histone substrates is not excluded by this design as described.

    Evidence that argues against this (1)
    • The same residue's symmetric mark is attributed by UniProt specifically to PRMT7, not PRMT5, and a Genes Dev study reports that PRMT7 BUT NOT PRMT5 raises H4R3me2s at KMT2D target genes. So the writer of this mark appears to be locus- and context-dependent, and a PRMT5 knockdown result at globin should not be generalised into 'PRMT5 writes H4R3me2s' genome-wide. Dhar SS et al. 2012
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The binding strength of the tandem plant homeodomain PHD(4-6) of MLL4 - the enzyme the paper also calls MLL2 and ALR, i.e. KMT2D - is inhibited by H4R3 symmetric dimethylation (H4R3me2s), a gene-repressive mark; PHD(4-6) is required for MLL4's nucleosomal methyltransferase activity; PRMT7, but not PRMT5, represses MLL4 target genes by upregulating H4R3me2s levels and antagonises MLL4-mediated differentiation; PRMT7 knockdown increases MLL4-catalysed H3K4me3 levels; and during differentiation, decreased H4R3me2s levels are associated with increased H3K4me3 levels at a cohort of genes including many HOXA and HOXB genes.

    PMID 23249737DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read, and NO residue numbers are asserted for PHD(4-6) because none were fetched from a sequence. NOMENCLATURE: the paper's 'MLL4' is KMT2D, settled by two tells in its own abstract - the aliases MLL2 and ALR, which are KMT2D-only, and the Kabuki-syndrome attribution. It is NOT KMT2B, whose MLL4 alias travels with WBP7. The system is NT2/D1 embryonal carcinoma differentiation: not haematopoietic, not leukaemic, and the enzyme is KMT2D rather than KMT2A. The HOXA/HOXB correlation is an association between two marks during differentiation, not a causal locus-level perturbation.

    Evidence that argues against this (1)
    • The result is about KMT2D and its PHD(4-6) cassette. KMT2A's homologous PHD1-3 plus bromodomain cassette is LOST from every 5'-KMT2A fusion, and whether KMT2A's PHD1-3 reads H4R3 at all has not been tested anywhere found in this beat. Reading this as a statement about KMT2A, let alone about a KMT2A fusion, would be a paralogue substitution. UniProt 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt Q9NVM4 (ANM7_HUMAN, PRMT7, 692 aa, two tandem SAM-dependent MTase PRMT-type domains at 14-345 and 358-684, Cytoplasm/cytosol and Nucleus both ECO:0000269:PubMed:15494416) records that PRMT7 'specifically mediates the symmetric dimethylation of histone H4 Arg-3 to form H4R3me2s' and that it plays a role in gene imprinting by being recruited by CTCFL at the H19 imprinted control region and methylating histone H4 to form H4R3me2s, possibly leading to recruitment of DNA methyltransferases at these sites.

    What this evidence cannot show

    A curated annotation, not a measurement; accession fetched live rather than recalled. The same record's annotated CATALYTIC ACTIVITY reaction is the MONOmethyl one, and the FUNCTION text says PRMT7 forms sDMA 'with a preference for the formation of MMA' - so the database simultaneously assigns a symmetric dimethyl histone mark and annotates a monomethyl-only reaction, which is a sign of an unsettled assignment. The DNMT-recruitment clause is hedged by UniProt itself ('possibly').

    Evidence that argues against this (1)
    • A dedicated enzymology study found full-length human PRMT7 forms ONLY omega-N-monomethylarginine on every substrate tested including recombinant human histones H2A, H2B, H3 and H4, concluding PRMT7 is a type III enzyme. If that is right in cells, PRMT7 cannot deposit a symmetric dimethyl mark directly and the cellular H4R3me2s phenotypes attributed to it would have to be indirect. Zurita-Lopez CI et al. 2012
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    UniProt O14744 (ANM5_HUMAN, PRMT5, 637 aa) is a type II protein arginine methyltransferase forming monomethyl- and symmetric dimethylarginine, a component of the methylosome with WDR77 and CLNS1A, and carries a Chromosome subcellular-location value with ECO:0000269 experimental evidence - the annotation that licenses reading PRMT5 as a chromatin enzyme rather than only a cytoplasmic methylosome component; its H4R3me2s and H3R8me2s histone activities are annotated BY SIMILARITY rather than experimentally in human cells.

    What this evidence cannot show

    A curated annotation, not a measurement. The BY SIMILARITY status of PRMT5's H4R3me2s annotation is material: it means the human H4R3me2s attribution in the database is inferred, and the experimental support in human cells comes from the primary papers (E1, E5) rather than from the record. This evidence item is cited to record the annotation's strength class, and the accession/length were verified against the entry already in this graph (gene-prmt5) which fetched them live on 2026-09-22.

    Evidence that argues against this (1)
    • A BY SIMILARITY histone annotation for a near-pan-essential enzyme whose dominant measured activities are snRNP core assembly and methylosome function is weak ground for a chromatin argument. Any anti-leukaemic effect of PRMT5 inhibition is more plausibly a splicing effect than an H4R3 effect, and attributing it to this mark would misdirect the pharmacodynamic. Serio J et al. 2018
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In murine MLL-AF9 leukaemia, conditional loss of the PAF-complex subunit Cdc73 reduces an early haematopoietic progenitor gene program that prevents differentiation; Prmt5 is a direct PAFc target positively regulated by a transcriptional unit including the PAFc, MLL1, HOXA9 and STAT5 in leukaemic cells; PRMT5-mediated H4R3me2s was REDUCED following excision of Cdc73, placing this histone modification downstream of the PAFc; and knockdown or pharmacologic inhibition of Prmt5 causes G1 arrest and reduced proliferation with extended leukaemic disease latency in vivo.

    PMID 28945229DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. MURINE MLL-AF9, MYELOID - not human, not B-lineage, not the index fusion KMT2A::AFF1. The mark is measured as a global/bulk reduction, NOT at named fusion target loci, so this is not a locus-resolved chromatin result. The causal chain runs PAFc to Prmt5 TRANSCRIPTION to PRMT5 protein to the mark: it is a writer-abundance result, not a demonstration that the fusion controls H4R3me2s on chromatin. The antibody used and its isomer specificity are not stated in the abstract.

    Evidence that argues against this (1)
    • Because the mark falls as a consequence of the writer's gene being downregulated, this result is equally consistent with H4R3me2s being an inert bystander of PRMT5 abundance. The G1 arrest and latency phenotypes are PRMT5 phenotypes and PRMT5 is near-pan-essential with dominant roles in snRNP assembly, so nothing here attributes any leukaemic phenotype to the histone mark itself. Zhao Q et al. 2009
  6. E6strong 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 and every zero re-run: '"H4R3me2s"[tiab]' = 92; H4R3[tiab] = 109; '"H4R3me2s"[tiab] AND KMT2A[tiab]' = 1 (28945229); AND MLL[tiab] = 1 (28945229); AND leukemia[tiab] = 4 (23249737, 28945229, 30930442, 32042864); AND 'MLL-AF9'[tiab] = 0; AND menin[tiab] = 0; AND antibod*[tiab] = 3 (21072194, 25092918, 36230689); H4R3[tiab] AND KMT2A[tiab] = 0; H4R3[tiab] AND MLL[tiab] = 3. Operand control KMT2A[tiab] = 1652. ID-SET comparison at retmax=500: intersection of H4R3[tiab] (109) with the union of the three compact spellings (152) is 24, Jaccard 10.1%; 75 of the 92 '"H4R3me2s"' records are NOT returned by H4R3[tiab].

    What this evidence cannot show

    Bounds TITLES AND ABSTRACTS only; no Europe PMC full-text sweep. THE MOST IMPORTANT ARTEFACT HERE: '"H4R3me2s" AND "MLL-AF9"' returns 0 while '"H4R3me2s" AND KMT2A' returns 1 and the record IS an MLL-AF9 study - its abstract says 'MLL fusion' and 'KMT2A' but never 'MLL-AF9'. An absence written against the fusion name alone would have been false. Also, 23249737 enters the 'AND leukemia' set only because the enzyme's legacy name contains 'leukemia'; counting it as leukaemia literature would be a symbol artefact.

    Evidence that argues against this (1)
    • This project has a measured case of a four-spelling bounded zero on a histone mark being false because the measurement sat in a supporting figure rather than the abstract. The correct reading of these zeros is 'no paper announces the measurement', and the founding KMT2A-fusion chromatin papers' supplementary tables have not been searched for arginine marks. Van Vlierberghe P et al. 2008
  7. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review

    No bona fide eraser is established for this mark: the only enzyme ever claimed to demethylate H4R3 is JMJD6, reported in 2007 to demethylate H3R2 and H4R3 in biochemical and cell-based assays, subsequently shown to be a lysyl-5-hydroxylase of the splicing factor U2AF65, and described in a dedicated review of the enzyme as having an arginine-demethylase assignment 'subject to conflicting reports'.

    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 a controversy rather than as primary evidence. Abstract read from the eutils record 2026-09-22. Ten years old and cannot bound the later literature; nothing found in this beat's searches assigns a cellular H4R3 eraser since. The primary papers on both sides are cited in full on mark-h3r2me2a (E7, E8, E15).

    Evidence that argues against this (1)
    • A subset of JmjC lysine demethylases can catalyse arginine demethylation in purified form with formaldehyde release, so the chemistry is possible and an enzyme may yet be assigned; the graph's position must be 'no established eraser', not 'no eraser can exist'. Walport LJ et al. 2016
  8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PADI4 deiminates arginine residues in histone tails and antagonises arginine methylation, but dimethylation of arginines PREVENTS deimination by PADI4 while monomethylation still allows it - so citrullination cannot remove a symmetric dimethylarginine mark, and in the absence of any established demethylase, H4R3me2s can only be lost by histone turnover or nucleosome eviction.

    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 residues explicitly named as PADI4 substrates in this paper are H3 R2, R8, R17 and R26 - H4R3 is NOT among them here; the H4 Arg-3 PAD4 target comes from the companion Science paper (PMID 15345777). What transfers to H4R3 is the general degree restriction on deimination, which is a chemical constraint on the reaction rather than a site-specific result, and this entry relies on it in that general form only.

    Evidence that argues against this (1)
    • The companion Science paper does name the PRMT1 H4 Arg-3 site among PAD4 targets and reports that PAD4 activity lowers histone arginine methylation in HL-60 granulocytes, so PAD4 does affect this residue - just not at the dimethyl state. An experiment seeing H4R3me2s fall after raising PAD4 would most likely be seeing the monomethyl substrate pool removed upstream of the writer. Wang Y et al. 2004
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Full-length human PRMT7 expressed as a GST fusion with robust activity produced only omega-N-monomethylarginine on synthetic peptides, GST-GAR, myelin basic protein and recombinant human histones H2A, H2B, H3 and H4, regardless of incubation time, reaction volume or substrate concentration, while control GST-PRMT1 and Myc-PRMT5 dimethylated their substrates as expected - so PRMT7 is a type III methyltransferase forming only monomethylarginine under the conditions tested.

    PMID 22241471DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. In vitro with a bacterially expressed GST fusion; the authors note a prior study characterised PRMT7 fusions as type II/type I and qualify their own conclusion with 'under the conditions tested'. An in vitro negative on recombinant histone components does not exclude symmetric dimethylation of a nucleosomal substrate in a cell with cofactors present. This item is included so that the graph's PRMT7 attributions carry their own strongest objection.

    Evidence that argues against this (1)
    • A cellular knockdown experiment reports that PRMT7, but not PRMT5, raises H4R3me2s at KMT2D target genes and that PRMT7 knockdown increases KMT2D-catalysed H3K4me3 - a symmetric-dimethyl phenotype in cells that this in vitro negative does not account for. Either PRMT7 makes the mark in cells under conditions not reproduced in vitro, or it acts indirectly through another enzyme. Dhar SS et al. 2012
  10. E10moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In K562 cells and primary human bone-marrow erythroid progenitors, WDR5 interacts with PRMT5 - a known repressor of gamma-globin expression - and is essential for generating H3K4me3 at the gamma-globin promoter; enforced WDR5 reduces gamma-globin expression while WDR5 knockdown increases it, with histone H3 and H4 acetylation increased and histone H4R3 and H3K9 methylation decreased at the promoter; WDR5 also interacts with HDAC1 and with ING2, an H3K4me3 reader, to enhance transcriptional repression, and WDR5 is highly enriched on the gamma-promoter relative to other globin promoters in human bone-marrow cells.

    The role of WDR5 in silencing human fetal globin gene expression. · Xu Z, He Y, Ju J, Rank G, et al. · Haematologica 2012
    PMID 22689669DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. The abstract reports 'histone H4R3 methylation' WITHOUT a degree assignment, so whether the measured species is the symmetric state specifically is NOT established by this paper - it is cited for the PRMT5/WDR5/H4R3 axis in primary human erythroid cells, not for the isomer. Erythroid, not lymphoid; no leukaemic fusion present. The proposed model in the abstract is a chain of interactions rather than a demonstrated order of events.

    Evidence that argues against this (1)
    • This is a WDR5 result in which WDR5 acts as a REPRESSOR, which cuts against reading any WDR5 finding as a KMT2A-axis activation finding - a caution this graph already carries as its own entry. It also means the H4R3 change here could be a consequence of the repressive complex assembling rather than a cause of it, and the abstract cannot separate the two. Zhao Q et al. 2009
  11. E11moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Immunohistochemistry of the developing mouse cortex shows the proliferative neuroepithelium, during the stage of mainly symmetric expansive divisions, is characterised by the prevalence of H4R3me2s with almost no detectable H4R3me2a, while later, once post-mitotic cortical layers begin forming, both marks are detected in post-mitotic neurons and oligodendrocyte precursors - proposing H4R3me2s as part of the histone code of undifferentiated neural precursors.

    PMID 21072194DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Single-author study, mouse cortex, immunohistochemistry only, no locus resolution, no quantification. The entire result is a comparison between two antibodies against isobaric isomers of one residue, with no competition panel reported - so a reagent affinity difference would reproduce the apparent developmental switch with no change in chromatin. Not haematopoietic.

    Evidence that argues against this (1)
    • The direction agrees with the KMT2D trans-tail differentiation series - H4R3me2s with the undifferentiated state, falling as differentiation proceeds - but that study is also antibody-based, so the two are not independent of the antibody problem. Isomer-resolving mass spectrometry would settle it and has not been applied to this question in any system read for this beat. Egelhofer TA et al. 2011
Connected to