Oncofusion

start / graph / gene-aff1

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q5b-interactome

AFF1 / AF4 — the 3' half of the fusion, and a target with almost no structural footing

also called AFF1 · AF4 · MLLT2 · AF4/FMR2 family member 1 · proto-oncogene AF4

AFF1 is the protein whose tail end gets welded onto KMT2A in the commonest infant leukemia fusion. Despite being half the cancer protein, almost nothing is known about its shape: the entire Protein Data Bank contains one small piece of it.

The same thing, for a scientist

AFF1 (P51825, 1,210 aa) is the 3' partner of KMT2A::AFF1, contributing its C-terminal region including the pSER domain and the C-terminal homology domain that scaffolds the super elongation complex. It is a founding member of the AEP/SEC description of KMT2A-fusion complexes. DepMap ranks it sixth on lineage-controlled selectivity (-0.207) with a near-neutral mean Chronos of +0.069 — a favourable profile — but the structural and chemical position is the weakest of any major partner: exactly one PDB entry covering residues 738-779, no ChEMBL matter that is not a chemoproteomic artefact, and no published affinity for any AFF1 interaction.

AFF1 drawn to scale against what we actually know its shape to be: one short stretch out of 1,210 amino acids.
Tell me more about this picture

The only experimentally determined region of AFF1 is residues 738-779, in complex with the AF9 AHD (PDB 2LM0, NMR).

Tell me more about this picture

The C-terminal homology domain, which is retained in KMT2A::AFF1 and scaffolds the super elongation complex, has never been solved for AFF1; the AFF4 equivalent has three crystal structures.

Tell me more about this picture

An AlphaFold model exists but AFF1 is predicted to be largely disordered, so it marks absence of structure rather than providing a designable surface.

Tell me more about this picture

This is the structural position for the 3' half of the commonest infant leukemia fusion.

What it looks like
loading AF-P51825-F1…
AlphaFold AF-P51825-F1 ↗
colourshape

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
Half the oncoprotein, one structure1 deeper

AFF1's entire experimental structural footprint is 2LM0, a solution NMR structure of AFF1 residues 738-779 bound to the AF9/MLLT3 AHD (residues 490-568) 1. That is 42 residues out of 1,210.

For comparison, in this same beat WDR5 has 60 PDB entries, BRD4 has 60 and DOT1L has 48. The 3' half of the most common infant leukemia fusion has one, and it is a peptide complex with a partner rather than a structure of AFF1 itself.

An AlphaFold model exists (AF-P51825-F1) and resolves, which is more than can be said for full-length KMT2A (AF-Q03164-F1 returns HTTP 404). But AFF1 is predicted to be largely intrinsically disordered, and for a disordered protein an AlphaFold monomer prediction tells you mainly where order is absent. The practical consequence is that no structure-based campaign against the AFF1 half of the fusion is currently possible, and closing that gap — even by co-folding predictions of the AFF1 C-terminal homology domain against its partners — is a concrete, cheap deliverable. The paralogous AFF4 C-terminal homology domain has been crystallised three times (6K7P 2.4 Å, 6KN5 2.2 Å, 6R80 2.2 Å), so a homology-based starting model for the AFF1 equivalent is available.

The literature gap is as striking as the structural one

We searched specifically for a primary paper whose central claim is that AFF1/AF4 knockdown kills KMT2A::AFF1 leukemia cells, and could not find one 2. The requirement for AFF1 in this disease is supported indirectly: by the AEP description, in which AF4-family and ENL-family proteins with P-TEFb form higher-order complexes that KMT2A fusions hijack 3; and by the SEC description, in which AFF4 is required for SEC stability and AFF4 knockdown reduces KMT2A-chimera target gene expression 4.

Both of those are about the complex and the paralogue. Neither is a dedicated AFF1-dependency paper for KMT2A::AFF1. For the single commonest fusion in infant ALL that is a remarkable hole, and the KG should record it as a hole rather than paper over it.

Correction 2026-09-19. That search was too narrow in one direction and correct in another. It remains true that no dedicated AFF1-knockdown-kills-KMT2A::AFF1 paper was found. But a whole line of work was missed: Bursen et al. and the Marschalek group argue that the reciprocal AF4::MLL, not MLL::AF4, is the leukaemogenic product in t(4;11), on the basis that AF4.MLL alone induced pro-B ALL in mice while MLL.AF4 alone induced none over 13 months; the same hypothesis failed in human cord-blood CD34+ cells over 36 weeks [counter-evidence to E2]. The hole in the literature is therefore not only 'nobody has knocked AFF1 down' but 'the field does not agree which derivative product to knock down'. See the counter-argument sweep section.

The dependency numbers are better than the literature implies, and the AFF4 contrast is the finding
AFF1AFF4
mean Chronos, all lines+0.069-0.056
all leukemia lines-0.052-0.085
10 KMT2A-rearranged lines-0.259+0.026
4 KMT2A::AFF1 lines-0.337+0.024
selectivity vs leukemia-0.207 (rank 6)+0.111 (rank 38)
common-essentialnono
gnomAD pLI / LOEUF0.7645 / 0.5021.0 / 0.166

Two readings 5.

AFF1 behaves like a fusion dependency and AFF4 does not. AFF1's requirement deepens specifically in KMT2A-rearranged lines and deepens further in the four KMT2A::AFF1 lines, which is the pattern expected if the CRISPR guides are cutting the fusion allele as well as the wild-type one. AFF4 goes the other way: KMT2A-rearranged lines need it less than other leukemia lines do. The standard picture in which AFF4 is the SEC scaffold underlying KMT2A-fusion transcription does not survive contact with the lineage-controlled data.

There is a confound that has to be stated. In a KMT2A::AFF1 line, guides targeting AFF1 exons retained in the fusion will disrupt both the fusion and wild-type AFF1. So AFF1's selective score is partly a measurement of fusion dependency rather than of AFF1-as-a-partner dependency — and for target-selection purposes that is arguably more relevant, but it is not the same claim. Distinguishing the two requires guides restricted to AFF1 exons excluded from the fusion, and we have not seen that analysis published.

AFF1 also has the loosest human constraint of the well-scoring candidates: pLI 0.7645 and LOEUF 0.502, with 48 observed loss-of-function variants against 121.7 expected, and a mis_z of -0.12 indicating no missense constraint at all. Compared with MEN1 (LOEUF 0.204), PSIP1 (0.373) and AFF4 (0.166), AFF1 looks the most tolerant of loss in humans of any of them.

Chemical matter: a cautionary tale about reading ChEMBL naively

ChEMBL v34 has an AFF1 target (CHEMBL5724736) with 2 activity rows at pChEMBL ≥ 6 and an apparent best Kd of 13 nM 6. Taken at face value that reads as a ligandable protein with nanomolar matter.

It is not. Both rows come from a single 2011 chemoproteomics document, and the compound is molibresib (CHEMBL1232461), a BET bromodomain inhibitor. The measurement is affinity-capture co-enrichment of the super elongation complex on a BET-inhibitor matrix — AFF1 was pulled down because it travels with BRD proteins, not because the compound binds AFF1. The same artefact inflates AFF4, ELL, PAF1, SETD1A and ASH2L in this database 6.

So the honest statement is that AFF1 has no chemical matter at all. Six genes in this interactome would be mis-scored as ligandable if these rows were taken at face value, and this node records the artefact explicitly so the graph does not propagate it.

Counter-argument sweep (2026-09-19): two things this entry asserted that the literature contests

1. 'No dedicated AFF1-dependency paper exists' was the wrong search. There is a substantial body of work arguing that the relevant oncoprotein in t(4;11) is not MLL::AF4 at all but the reciprocal AF4::MLL. Bursen et al. transduced murine Lin-/Sca1+ progenitors and found that AF4.MLL alone, or AF4.MLL plus MLL.AF4, produced pro-B ALL, biphenotypic or mixed-lineage leukaemia, while MLL.AF4 alone produced no disease over 13 months [counter-evidence to E2]. On that reading the AFF1 C-terminal region this entry is built around is not the leukaemogenic moiety.

The dispute is unresolved and cuts the other way in human cells: AF4-MLL in cord-blood CD34+ HSPCs enhanced engraftment but produced no leukaemia in 36 weeks and no serial self-renewal, which the authors attribute to species-specific determinants [counter-evidence to E2]. This is the same mouse-versus-human split the project keeps finding.

Consequence for E5. The DepMap confound already recorded here - that guides against fusion-retained AFF1 exons cut the fusion as well as wild-type AFF1 - is worse than stated, because in a t(4;11) line those guides may also cut the reciprocal AF4::MLL transcript. The proposed discriminating experiment (guides inside versus outside the fusion) should be extended to guides that distinguish the two derivative products.

2. 'AFF1/AFF4 is a component of the SEC' is now an underdetermined claim. Native chromatin proteomics in human cells identifies ECAS, a bipartite assembly joining an AFF4-ELL2-EAF1 elongation module to a U2 snRNP-associated intron-binding complex, reported as molecularly distinct from canonical P-TEFb-containing SEC [counter-evidence to E4]. The SEC evidence cited here is 2004-2012 affinity purification, and presence in a purification no longer identifies which assembly is meant. The clean way to state this entry's position is: AFF-family proteins scaffold at least two compositionally distinct promoter-proximal elongation assemblies, the fusion has been shown to associate with AFF-family proteins, and which assembly it depends on has not been established for KMT2A::AFF1.

The epigenetic dimension: AFF1 carries no chromatin activity, and the one route by which it might deliver one has a paper this graph had never cited

This entry described half the index oncoprotein for eleven cycles with no epigenetic content, in a disease defined by a histone mark. The reason it is worth stating carefully rather than filling in is that AFF1 has no chromatin-enzymatic domain of any kind - its Pfam complement is PF05110, PF18875 and PF18876, and it carries no SET, JmjC, HAT, HDAC, bromodomain, chromodomain, Tudor, PHD or CXXC module. So the only epigenetic question about AFF1 is whether it delivers an enzyme, and that question is genuinely open.

The positive claim, which was missing from this graph entirely. Bitoun, Oliver and Davies showed that mouse Af4 acts as a positive regulator of P-TEFb and, in complex with the MLL fusion partners Af9, Enl and Af10, as a mediator of histone H3-K79 methylation by recruiting Dot1 to elongating Pol II; in the ataxic robotic mouse, an Af4-overexpression model, both phosphorylated Pol II and methylated H3-K79 are increased in vivo E7. That is a primary, peer-reviewed, in-vivo epigenetic readout for AFF1's own capacity, and a search of this corpus on 2026-09-21 found PMID 17135274 cited in zero entries. The project's mechanism-kmt2a-aff1-dot1l-recruitment-route records the complex-level DOT1L claim as resting on PMID 26171280, which it correctly marks weak and DOI-less; the stronger source for the same claim was not in the file.

Two limits on it, before anyone treats this as closing the gap. It is mouse Af4, wild type, over-expressed - not human AFF1 and not a fusion; and it is a complex-level attribution in which Af4 is one of four named components. Nothing in it maps a residue of AFF1 to DOT1L, so this entry's and the mechanism entry's structural statement - that KMT2A::AFF1 has no mapped DOT1L interface - survives intact.

The reconciliation, which is inference and is labelled as such. The reconstitution result that anchors this graph's sceptical position is that DOT1 forms a stable complex with AF9 and does not interact with AF9-AF4-P-TEFb complexes E8. Set against Bitoun, that reads as a flat contradiction. It may not be one, and the reason is a component list: the positive complex contains Af10, whose octapeptide-motif/leucine-zipper region is the annotated DOT1L-binding module, and the reconstitution that returned the negative contained AF9, AF4 and P-TEFb and no AF10. This is my inference from the two component lists and neither paper states it. If it is right, it does not rescue a fusion-directed route: it says AFF1's access to H3K79 methylation runs through MLLT10, which is precisely the endogenous DOT1L unit that hyp-index-fusion-h3k79-is-entirely-endogenous proposes. The reconciliation strengthens that hypothesis rather than weakening it, and it is falsifiable - an AF10-free Af4/Af9/Enl complex that still methylates H3K79 would kill it.

What is actually demonstrated to be modified is AFF1, not a histone. P-TEFb-dependent phosphorylation of Af4, Af9 and Enl controls their transactivation activity and/or protein stability E7, and this entry already records that PARP1 binds and PARylates AFF1 at DNA damage sites. Both are non-histone post-translational modifications on a chromatin-associated protein, and neither is a mark AFF1 places on chromatin. They are recorded here so that a later reader does not promote them into the epigenetic column.

A partner-import trap specific to this protein. The complex inventory that does contain histone-modifying enzymes belongs to AF4-MLL, the reciprocal derivative, which is a different protein from KMT2A::AFF1: affinity purification gives at least 16 constituents including P-TEFb, H3K4(me3) and H3K79(me3) histone methyltransferases, a protein arginine N-methyltransferase and a histone acetyltransferase, against 11 partners for the AF4 complex alone E9. This graph already carries five entries on that paper and an unresolved dispute about which derivative drives the disease; the point for the epigenetic column is narrower and easy to get wrong - the writer-rich inventory is the reciprocal's, and attaching it to the index fusion would manufacture a chromatin mechanism out of a naming collision.

Facts and figures
uniprotP51825
gene symbolAFF1
length AA1210
protein nameAF4/FMR2 family member 1
alphafold IDAF-P51825-F1
alphafold verifiedtrue
PDB verified against rcsbtrue
n PDB entries1
structural gap42 of 1,210 residues experimentally resolved; the C-terminal homology domain retained in the fusion has never been solved for AFF1. Homology templates available from AFF4: 6K7P, 6KN5, 6R80.
role in fusion3' partner of KMT2A::AFF1; contributes the AFF1 C-terminal region including the pSER and C-terminal homology domains, which scaffold AFF-family promoter-proximal elongation assemblies. Note two qualifications added 2026-09-19: (a) whether the relevant oncoprotein in t(4;11) is MLL::AF4 or the reciprocal AF4::MLL is disputed (PMID 20194896 in mouse for AF4::MLL; PMID 29137234 against it in human CD34+ cells); (b) 'the super elongation complex' is not a single entity - AFF4-ELL2-EAF1 also occurs in ECAS, reported as molecularly distinct from canonical P-TEFb-containing SEC (PMID 42748909) - so co-purification does not identify which assembly the fusion depends on.
subnuclear localization'AF4 BODIES' - discrete nuclear foci where AF4 and AF9 interact, explicitly NON-NUCLEOLAR and, in the authors' words, 'not coincident with any known nuclear structures we have examined' [E_q72_aff1_loc3]. That is the only named compartment AFF1 has, it comes from one 2004 transfection experiment, it has never been reproduced or given a marker, and it disagrees with the one subnuclear description of the MLL-AF4 chimera, which is PREDOMINANTLY NUCLEOLAR in an inducible expression system [E_q72_aff1_loc4]. See localization-af4-bodies-a-named-compartment-matching-nothing-known. [EXPLICIT NEGATIVE] AFF1 IS THE AFF-FAMILY MEMBER NOBODY PUT IN A SPECKLE. The paper that established nuclear-speckle localisation for the family covered AFF2/FMR2, AFF3/LAF4 and AFF4/AF5q31 and DID NOT INCLUDE AFF1 - the one that is actually fused to KMT2A in infant leukaemia [E_q72_aff1_loc1 counter-note]. No KMT2A fusion has ever been co-stained with a speckle marker. Separately, NO STEADY-STATE ENDOGENOUS LOCALISATION OF AFF1 OR OF KMT2A::AFF1 HAS EVER BEEN MEASURED BY AN ASSAY THAT CAN TELL THE TWO APART: the 1998 imaging is unattributable by construction (shared-epitope polyclonal, fusion the minority species), and the size dimension that does attribute on a blot is unavailable in a microscope. UNPUBLISHED RATHER THAN NOT-APPLICABLE. See localization-nuclear-speckles-kmt2a-r, ana-attribution-audit-subnuclear-localisation-kmt2a-aff1 and concept-anti-aff1-antibodies-what-they-can-and-cannot-attribute.
epigenetic readoutWRITES, ERASES AND READS NOTHING ITSELF - AFF1 cross-references only PF05110 (AF-4), PF18875 (AF4_int) and PF18876 (AFF4_CHD) and has no SET, JmjC, HAT, HDAC, bromo, chromo, Tudor, PHD or CXXC module. Its epigenetic role, if it has one, is as a SCAFFOLD THAT MAY DELIVER A WRITER, and that is contested. THE CLAIMED ROUTE (new to this graph): mouse Af4 is a positive regulator of P-TEFb and, IN COMPLEX WITH Af9, Enl AND Af10, a mediator of histone H3-K79 methylation by recruiting Dot1 to elongating Pol II; the robotic Af4-overexpression mouse shows increased phosphorylated Pol II AND increased methylated H3-K79 in vivo (PMID 17135274 - mouse, WILD-TYPE Af4 overexpression, NOT a fusion). THE OPPOSING RESULT: in reconstitution from defined components DOT1 forms a stable complex with AF9 and DOES NOT interact with AF9-AF4-P-TEFb complexes (PMID 21896721), which is why this graph holds KMT2A::AFF1 to have no mapped DOT1L interface. INFERENCE, NOT A SOURCE CLAIM: the two need not conflict, because the complex in the positive result CONTAINS Af10 - whose OM-LZ region is the annotated DOT1L-binding module - while the reconstitution that returned the negative contained AF9, AF4 and P-TEFb and no AF10. If AF10 is the DOT1L-binding moiety, then AFF1's route to H3K79 methylation IS the endogenous MLLT10-directed unit, which supports hyp-index-fusion-h3k79-is-entirely-endogenous rather than refuting it. NON-HISTONE MODIFICATIONS OF AFF1 ITSELF, which are the only modifications actually demonstrated on this protein: P-TEFb-dependent phosphorylation of Af4 controls its transactivation activity and/or protein stability (PMID 17135274), and PARP1 binds and PARylates AFF1 at DNA damage sites (already in this entry as E_q72_aff1_loc2). Neither is a histone mark. DO NOT IMPORT FROM THE RECIPROCAL: the AF4-MLL complex - a DIFFERENT protein from KMT2A::AFF1 - carries H3K4me3 and H3K79me3 methyltransferases, a protein arginine N-methyltransferase and a histone acetyltransferase among at least 16 constituents (PMID 21030982); that inventory belongs to the reciprocal derivative and has repeatedly been mistaken for the index fusion's.
PDB IDS
[
  "2LM0"
]
key structures
[
  {
    "pdb": "2LM0",
    "what": "AFF1 738-779 + MLLT3/AF9 AHD 490-568",
    "method": "NMR",
    "resolution_A": null,
    "pmid": "23260655",
    "title": "Leukemia fusion target AF9 is an intrinsically disordered transcriptional regulator that recruits multiple partners via coupled folding and binding.",
    "authors": "Leach BI, Kuntimaddi A, Schmidt CR, et al.",
    "venue": "Structure",
    "year": 2013,
    "title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
  }
]
essentiality
{
  "depmap_common_essential": false,
  "chronos_mean": 0.069,
  "n_lines": 1258,
  "fraction_lines_depleted": 0.129,
  "chronos_mean_leukemia": -0.052,
  "chronos_mean_kmt2ar": -0.259,
  "chronos_mean_kmt2a_aff1": -0.337,
  "selectivity_vs_leukemia": -0.207,
  "selectivity_vs_all_lines": -0.328,
  "rank_in_interactome": 6,
  "confound": "in KMT2A::AFF1 lines, guides against fusion-retained AFF1 exons disrupt the fusion as well as wild-type AFF1, so this score partly measures fusion dependency. Extended 2026-09-19: in a t(4;11) line those guides may additionally disrupt the reciprocal AF4::MLL product, which one body of work argues is the actual leukaemogenic protein (PMID 20194896). Separating the three requires guides that distinguish wild-type AFF1, MLL::AFF1 and AFF1::MLL, and no such analysis has been published.",
  "gnomad_pLI": 0.7645,
  "gnomad_loeuf": 0.502,
  "gnomad_obs_lof": 48,
  "gnomad_exp_lof": 121.7,
  "gnomad_mis_z": -0.12,
  "mouse_ko": "not established in this node; no verified citation obtained this cycle",
  "hpa_rna_tissue_specificity": "Low tissue specificity",
  "hpa_rna_tissue_distribution": "Detected in all",
  "hpa_rna_blood_cell_specificity": "Low immune cell specificity"
}
paralogue comparison
{
  "AFF4": {
    "chronos_mean": -0.056,
    "chronos_mean_kmt2ar": 0.026,
    "selectivity_vs_leukemia": 0.111,
    "rank": 38,
    "gnomad_loeuf": 0.166
  }
}
chemical matter
{
  "chembl_target": "CHEMBL5724736",
  "chembl_pref_name": "AF4/FMR2 family member 1",
  "n_activity_rows_pchembl_ge_6": 2,
  "apparent_best_activity": {
    "molecule": "CHEMBL1232461",
    "molecule_name": "molibresib",
    "type": "Kd",
    "value": 13,
    "unit": "nM"
  },
  "artefact_warning": "Both rows are BET-inhibitor chemoproteomic co-enrichment of the super elongation complex, not direct AFF1 binding. Effective chemical matter against AFF1: none.",
  "curated_mechanism_drugs": []
}
interfaces
[
  "interface-aff1-mllt3-ahd",
  "interface-af9-ahd-dot1l-three-sites-2mv7"
]
pointer repairs
[
  "2026-09-20: interface references re-pointed to their existing type-correct nodes; the originals were never-written ids, not missing biology"
]
cell type
{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "expressed_in": {
    "class": 1,
    "summary": "broadly expressed; not used here as evidence of requirement"
  },
  "required_in": {
    "class": 2,
    "kmt2a_rearranged_lines_mean_gene_effect": -0.257,
    "all_1258_lines_mean_gene_effect": 0.068,
    "pct_all_lines_depleted": 0.2,
    "floor_compressed": false,
    "myeloid_kmt2ar_mean": -0.181,
    "blineage_kmt2ar_mean": -0.391,
    "verdict": "AFF1 whole-gene loss is close to inert everywhere - depleted below -0.5 in 0.2% of 1,258 lines, the lowest in the worklist - INCLUDING in the KMT2A::AFF1 lines that carry it as the index fusion",
    "why_the_null_cannot_be_read_as_dispensability": "DepMap cuts the whole AFF1 locus, so the wild-type allele and the fusion allele are hit together; and the transforming moiety is the 3' AFF1 portion retained in the fusion, whose loss should be lethal if the fusion drives the line. A dependency screen keyed on the gene is the wrong instrument for a 3' fusion partner, and the striking zero is an artefact of the instrument, not a biological statement.",
    "source": "DepMap via Open Targets fetched 2026-09-20; ana-q60-cell-type-versus-fusion"
  },
  "cell_of_origin": {
    "class": 3,
    "claim": "In a murine system, Mll-AF4 confers enhanced self-renewal and lymphoid potential only during a RESTRICTED DEVELOPMENTAL WINDOW, and full transforming capacity is interlinked with lymphoid lineage commitment rather than with the stem cell itself.",
    "pmids": [
      "27396339",
      "28637661"
    ],
    "human_arm": "In human cord-blood CD34+ cells the AF4-MLL reciprocal transiently augments multilineage engraftment but is NOT sufficient to initiate leukaemia (PMID 29137234)."
  },
  "context": "fusion context throughout; murine for the window and the lineage-commitment result (brief rule 73)",
  "graph_links": [
    "cell-type-pro-b",
    "cell-type-fetal-preprob-cd10-negative",
    "cell-type-lmpp"
  ]
}
cell state
{
  "maintains": {
    "class": 3,
    "states": [
      "cell-state-differentiation-block",
      "cell-state-fetal-program-retention",
      "cell-state-stem-like-self-renewal"
    ],
    "claim": "Ontogeny-specific induction of KMT2A::AFF1 drives development of a distinct CD24-positive PRE-leukaemic state - the state the fusion creates before leukaemia, defined by a surface marker, in a defined developmental context.",
    "pmids": [
      "40646135",
      "27396339"
    ]
  },
  "bounded_query": {
    "date": "2026-09-20",
    "tool": "NCBI esearch db=pubmed",
    "query": "(AFF1[tiab] OR \"AF4\"[tiab] OR \"MLLT2\"[tiab] OR \"MLL-AF4\"[tiab] OR \"MLL::AF4\"[tiab]) AND <cell-state clause> AND <leukaemia clause>",
    "count": 72,
    "controls": {
      "gene_clause_alone": 1409,
      "state_clause_alone": 96743
    },
    "title_check_note": "a large fraction of the 72 are minimal-residual-disease method papers in which AFF1 appears only as an RT-PCR target, not as a gene whose cell-state function is under study. The state clause includes MRD terms deliberately, and for the fusion-partner genes that term dominates the recall."
  }
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.2575,
  "chronos_mean_other_leukaemia": -0.0103,
  "difference": -0.2472,
  "pan_essentiality_all_lines": 0.0687,
  "selectivity_rank": 160,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
localization
{
  "subcellular": "NUCLEUS. UniProt P51825 (fetched 2026-09-21) carries exactly one location value on exactly one experimental attribution - ECO:0000269 to PMID 41062835, a 2026 paper on transcriptional restart after DNA damage - plus an uncoded note from the same paper that AFF1 associates with chromatin and is recruited to DNA-damage sites in a PARP1-dependent manner [E_q72_aff1_loc1][E_q72_aff1_loc2]. No NLS motif is annotated. The 1998 paper whose title is about AF4 subnuclear compartments is not cited by UniProt for this field.",
  "subnuclear": "'AF4 BODIES' - discrete nuclear foci where AF4 and AF9 interact, explicitly NON-NUCLEOLAR and, in the authors' words, 'not coincident with any known nuclear structures we have examined' [E_q72_aff1_loc3]. That is the only named compartment AFF1 has, it comes from one 2004 transfection experiment, it has never been reproduced or given a marker, and it disagrees with the one subnuclear description of the MLL-AF4 chimera, which is PREDOMINANTLY NUCLEOLAR in an inducible expression system [E_q72_aff1_loc4]. See localization-af4-bodies-a-named-compartment-matching-nothing-known.",
  "explicit_no": "AFF1 IS THE AFF-FAMILY MEMBER NOBODY PUT IN A SPECKLE. The paper that established nuclear-speckle localisation for the family covered AFF2/FMR2, AFF3/LAF4 and AFF4/AF5q31 and DID NOT INCLUDE AFF1 - the one that is actually fused to KMT2A in infant leukaemia [E_q72_aff1_loc1 counter-note]. No KMT2A fusion has ever been co-stained with a speckle marker. Separately, NO STEADY-STATE ENDOGENOUS LOCALISATION OF AFF1 OR OF KMT2A::AFF1 HAS EVER BEEN MEASURED BY AN ASSAY THAT CAN TELL THE TWO APART: the 1998 imaging is unattributable by construction (shared-epitope polyclonal, fusion the minority species), and the size dimension that does attribute on a blot is unavailable in a microscope. UNPUBLISHED RATHER THAN NOT-APPLICABLE. See localization-nuclear-speckles-kmt2a-r, ana-attribution-audit-subnuclear-localisation-kmt2a-aff1 and concept-anti-aff1-antibodies-what-they-can-and-cannot-attribute.",
  "how_shown": "UniProtKB P51825 fetched 2026-09-21 and read for ECO codes and feature table; PARylation, chromatin association and damage-site recruitment (PMID 41062835); size-resolved Western plus confocal immunofluorescence on t(4;11) and non-t(4;11) leukaemic lines (PMID 9808577); tagged-protein transfection and co-localisation (PMID 14603337); inducible MLL-AF4 expression with nucleolar readouts (PMID 35468859).",
  "systems": "human cells under genotoxic stress (2026); t(4;11) and non-t(4;11) human leukaemic cell lines, imaged but unattributable (1998); transfection systems (2004, 2022). NO ENDOGENOUS, JUNCTION-SPECIFIC LOCALISATION EXPERIMENT IN INFANT B-LINEAGE KMT2A::AFF1 EXISTS.",
  "shown_for_fusion_localisation": "NAME WHICH, and the answer is worse than a bare yes. For KMT2A::AFF1 / MLL-AF4 itself: (1) punctate nuclear in t(4;11) patient-derived lines - but the SAME pattern appears in non-t(4;11) cells with the same reagent, so the image is UNATTRIBUTABLE and the authors' own blot, not their microscope, is what identified the fusion (PMID 9808577); (2) predominantly NUCLEOLAR under inducible overexpression in a non-leukaemic host (PMID 35468859); (3) the fusion MAINTAINS the AF4-AF9 interaction and ALTERS AF9's subnuclear localisation in trans (PMID 14603337) - which is the only positive, attributable localisation phenotype of this fusion anywhere, and it is about where the PARTNER goes, not where the fusion is. Nothing endogenous, nothing junction-specific, nothing in a patient sample.",
  "node": "localization-af4-bodies-a-named-compartment-matching-nothing-known"
}
How this could bite us back
  • Paralogue redundancy: AFF2, AFF3 and AFF4 are all present and AFF4 is the canonical SEC scaffold, so an AFF1-directed agent could be buffered by family members in a way single-gene CRISPR data partially conceals.
  • No druggable surface is currently known. With 42 of 1,210 residues resolved and a largely disordered predicted structure, there is nothing to design against; this is a discovery problem, not an optimisation problem.
  • The database trap: naive reading of ChEMBL would suggest 13 nM chemical matter against AFF1. It is a BET-inhibitor pulldown artefact, and acting on it would waste a programme.
  • Targeting AFF1 means targeting a wild-type protein with normal functions in transcription elongation and, through the AFF family, in neurodevelopment; the tolerance implied by gnomAD is for heterozygous loss, not for pharmacological inhibition.
  • Because the selective CRISPR score is confounded by cutting the fusion itself, the apparent therapeutic index of AFF1 may be substantially better in the data than it would be for a drug that spares the fusion and hits only wild-type AFF1.
  • The premise may be wrong at the level of which protein to target. A line of mouse work says the reciprocal AF4::MLL, not MLL::AF4, is the leukaemogenic product in t(4;11) (PMID 20194896); a human cord-blood experiment says AF4-MLL is insufficient (PMID 29137234). A programme aimed at the AFF1 half of MLL::AF4 is betting on one side of an unresolved dispute.
  • The complex the fusion is assumed to hijack is not a single entity. AFF4-ELL2-EAF1 sits in ECAS, reported as distinct from canonical P-TEFb-containing SEC (PMID 42748909), so an agent designed against 'the SEC' may be designed against the wrong assembly, and a negative result would be uninterpretable without composition-resolved proteomics of the fusion complex itself.
  • If the reconciliation offered here is right - that AFF1 reaches DOT1L only through AF10/MLLT10 - then a DOT1L inhibitor in t(4;11) is acting on the cell's endogenous elongation machinery rather than on a fusion-specific module, and the therapeutic index that argument implies is worse, not better. The clinical record is consistent with that reading: pinometostat's two complete remissions in 51 adults were both in t(11;19), not t(4;11).
Still open
  1. Which derivative chromosome product drives t(4;11) disease? AF4.MLL alone induced pro-B ALL in mice while MLL.AF4 alone induced none over 13 months (PMID 20194896), yet AF4-MLL was insufficient in human cord-blood CD34+ cells over 36 weeks (PMID 29137234). This entry, and this graph's fusion entries generally, assume MLL::AF4 is the target. No human experiment settles it.
  2. Is there a dedicated AFF1-dependency experiment in KMT2A::AFF1 leukemia at all? We could not verify one. Guides restricted to AFF1 exons excluded from the fusion versus guides against fusion-retained exons would separate AFF1-as-partner from AFF1-as-half-the-oncoprotein.
  3. Why do KMT2A-rearranged lines need AFF4 less than other leukemia lines do, when AFF4 is described as the SEC scaffold? The hypothesis that the fusion supplies its own AFF-family scaffold is ours and untested; comparing SEC composition in KMT2A::AFF1 versus KMT2A::MLLT3 lines would test it.
  4. Can the AFF1 C-terminal homology domain be modelled from the AFF4 crystal structures (6K7P, 6KN5, 6R80) well enough to be useful? This is a co-folding job that costs nothing and would give the project its first designable AFF1 surface.
  5. Does the fusion's AFF1 portion make any contact that wild-type AFF1 does not? If so that is a genuinely fusion-specific surface, which is the thing the whole project is looking for and which the junction epitope node approaches from the other side.
  6. AFF1 has the loosest human loss-of-function constraint of any well-scoring candidate here (LOEUF 0.502, mis_z -0.12). Is that a real tolerance signal that makes it safer to target than menin, or an artefact of paralogue redundancy with AFF2, AFF3 and AFF4?
  7. Does KMT2A::AFF1 depend on canonical SEC or on ECAS? AFF4-ELL2-EAF1 occurs in an assembly reported as distinct from P-TEFb-containing SEC (PMID 42748909). Composition-resolved proteomics of the fusion complex in a human KMT2A::AFF1 line, rather than of AFF4 pulldowns, would say which.
  8. Do 'AF4 bodies' exist at endogenous expression? The compartment has been named since 2004 on one transfection experiment, has no marker, and has never been sought in a t(4;11) cell - an endogenously tagged AFF1 knock-in would answer it in one imaging session.
  9. Is AFF1 in nuclear speckles? Three of the four AFF-family members are; the one fused to KMT2A was never tested, and the speckle hypothesis for this disease rests on that untested link.
  10. Can an Af4/Af9/Enl complex mediate H3K79 methylation WITHOUT Af10? This is the single experiment that would resolve the contradiction between PMID 17135274 (positive, in a four-component complex containing Af10) and PMID 21896721 (negative, in an AF9-AF4-P-TEFb reconstitution without AF10). If AF10 is required, AFF1's only access to DOT1L is the endogenous MLLT10 unit and hyp-index-fusion-h3k79-is-entirely-endogenous gains a mechanism; if it is not, KMT2A::AFF1 has an unmapped route and the interface argument weakens.
  11. Does human AFF1, at endogenous level, in a t(4;11) cell, associate with DOT1L at all? Everything positive is mouse, wild-type and over-expressed; everything negative is reconstituted. Neither arm has been run in the index fusion's own cells.
Every source behind this entry (13)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    AFF1's entire experimental structural coverage is a single PDB entry, 2LM0, a solution NMR structure of AFF1 residues 738-779 bound to the AF9/MLLT3 AHD (residues 490-568); an AlphaFold model AF-P51825-F1 exists and resolves, whereas full-length KMT2A has no AlphaFold model (AF-Q03164-F1 returns HTTP 404). The paralogous AFF4 C-terminal homology domain has three crystal structures (6K7P 2.4 A, 6KN5 2.2 A, 6R80 2.2 A) that could seed a homology model of the AFF1 equivalent.

    What this evidence cannot show

    The RCSB search was by UniProt accession cross-reference, so an entry containing AFF1 without that cross-reference annotated would be missed. The AlphaFold model was confirmed only to return HTTP 200; AFF1 is predicted to be largely disordered, so a monomer prediction indicates where order is absent rather than providing a designable surface.

    Evidence that argues against this (1)
    • A solution NMR structure of an isolated domain is the weakest structural evidence in this dossier class: the construct is chosen to be soluble and monomeric, the interaction partners are absent, and the model is an ensemble rather than a density map. For AFF1 specifically the single entry covers 42 residues of a 1,210-residue protein, so 96% of it has no experimental structure at all. RCSB PDB 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    No primary paper whose central claim is that AFF1/AF4 knockdown is required for KMT2A::AFF1 leukemia cell survival could be found; the requirement for AFF1 in this disease rests on complex-level and paralogue-level evidence rather than on a dedicated AFF1-dependency study.

    PubMed literature search for primary studies of AFF1/AF4 dependency in MLL-AF4 / KMT2A::AFF1 leukemia, returning no paper with that central claim · PubMed (US National Library of Medicine), searched via the PubMed E-utilities API · PubMed 2026
    What this evidence cannot show

    A negative literature search is weak evidence: the paper may exist under terminology we did not use, may be a figure within a larger study, or may be in the pre-2000 AF4 literature that is poorly indexed. This should be read as 'we could not verify one' rather than 'none exists'. It is recorded because the project's quality bar forbids citing sources we could not verify.

    Evidence that argues against this (2)
    • The negative literature search recorded in E2 missed a whole line of work that argues this entry is looking at the wrong protein. Bursen et al. transduced Lin-/Sca1+ murine progenitors with MLL.AF4, with AF4.MLL, or with both. Recipients of AF4.MLL-transduced or double-transduced cells developed pro-B ALL, B/T biphenotypic acute leukaemia or mixed-lineage leukaemia; recipients of MLL.AF4-transduced or mock-transduced cells developed NO disease over 13 months of observation. Their conclusion is that the reciprocal AF4.MLL fusion protein is capable of inducing ALL even in the absence of MLL.AF4, and that t(4;11) leukaemia may rest on two oncoproteins. If that holds, then the 3' half of MLL::AF4 - the AFF1 C-terminal region this entry is built around - is not the leukaemogenic product, and the DepMap selectivity in E5 (which cuts the fusion allele) may be measuring a dependency on the wrong one of the two derivative chromosomes. Bursen A et al. 2010
    • The counter to the counter, and the reason this should be recorded as a live dispute rather than a correction. Prieto et al. expressed AF4-MLL in human cord-blood CD34+ HSPCs and found enhanced long-term haematopoietic reconstitution in primary immunodeficient recipients but no self-renewal on serial transplantation, and NO phenotypic or haematological sign of disease within 36 weeks - AF4-MLL was not sufficient to initiate leukaemia in human cells. The authors attribute the discrepancy with the murine result to species-specific intrinsic determinants. So the reciprocal-product hypothesis is supported in mouse and unsupported in human, which is exactly the kind of split this project has repeatedly found between MLL-AF9 mouse work and human KMT2A::AFF1 biology, and it leaves the question of which derivative matters genuinely open. Prieto C et al. 2017
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A higher-order complex containing AF4-family and ENL-family proteins with P-TEFb (AEP) facilitates both oncogenic and physiological KMT2A-dependent transcription, and KMT2A fusions with AEP components form constitutive KMT2A/AEP hybrid complexes that sustain target gene expression and transform haematopoietic progenitors.

    PMID 20153263DOI
    What this evidence cannot show

    This is a complex-level description across AF4-family and ENL-family proteins rather than a demonstration that AFF1 specifically is the required member; the transformation readouts are murine haematopoietic progenitors. No affinity measurements are reported.

    Evidence that argues against this (1)
    • Complex composition is not requirement. The AEP/SEC papers establish what co-purifies and what a reconstituted complex does; they do not establish that each subunit is rate-limiting in a patient's leukaemia, and this project's own dependency data separates the members by an order of magnitude. Nature 2012
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Purification of multiple KMT2A chimeras identifies a super elongation complex containing ELL, P-TEFb and AFF4; AFF4 is required for SEC stability, and AFF4 knockdown reduces KMT2A-chimera target gene expression in leukemic cells.

    PMID 20159561DOI
    What this evidence cannot show

    This is the AFF4 paralogue, not AFF1, and it is the central tension in this node: the biochemistry says AFF4 scaffolds the complex while the lineage-controlled DepMap data say KMT2A-rearranged lines need AFF4 less than other leukemia lines do (selectivity +0.111). The most likely reconciliation is that in KMT2A::AFF1 the fusion supplies its own AFF-family scaffold, but that reconciliation is our inference, not a published result.

    Evidence that argues against this (1)
    • The SEC framing that this entry inherits from 2004-2012 affinity purification underdetermines which complex is meant, and a 2026 paper says so directly. Xiang et al. used native chromatin proteomics in human cells to identify ECAS, a bipartite assembly linking an AFF4-ELL2-EAF1 elongation module to a U2 snRNP-associated intron-binding complex, occupying promoter-proximal chromatin and reported as molecularly DISTINCT from canonical P-TEFb-containing SEC. AFF4 and ELL2 therefore appear in at least two compositionally different assemblies. Every statement in this entry of the form 'AFF1 or AFF4 is a component of the SEC, therefore the fusion hijacks the SEC' rests on co-purification, and co-purification no longer identifies a single complex. This does not refute the AFF4 requirement reported in E4; it means the entity that requirement belongs to has not been established, and the AFF1-versus-AFF4 DepMap asymmetry in E5 may be an asymmetry between assemblies rather than between paralogues. Xiang Y et al. 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    AFF1 is not DepMap common-essential and has a mean Chronos of +0.069 across 1,258 lines, falling to -0.259 across 10 KMT2A-rearranged lines and -0.337 across 4 KMT2A::AFF1 lines (lineage-controlled selectivity -0.207, rank 6 of 41), whereas the paralogue AFF4 scores +0.111 (rank 38), i.e. KMT2A-rearranged lines require AFF4 less than other leukemia lines do; gnomAD gives AFF1 pLI 0.7645, LOEUF 0.502, 48 observed versus 121.7 expected loss-of-function variants and mis_z -0.12, the most tolerant profile among the well-scoring candidates.

    What this evidence cannot show

    A specific and important confound applies to AFF1 and not to the other partners: in a KMT2A::AFF1 cell line, CRISPR guides targeting AFF1 exons retained in the fusion disrupt both the fusion transcript and wild-type AFF1, so AFF1's selective score partly measures fusion dependency rather than partner dependency. Separating the two requires guides restricted to AFF1 exons excluded from the fusion and we have not seen that published. Means were computed by us from per-line API values and only 10 KMT2A-rearranged lines (4 KMT2A::AFF1) contribute.

    Evidence that argues against this (1)
    • A Chronos mean mixed across a KMT2A-rearranged subset and a comparator arm must be lineage-split before it is stated: the index arm is myeloid-enriched, so a difference between the two mixes fusion status with lineage. Ten or eleven index lines and four KMT2A::AFF1 lines also put wide intervals on every figure quoted. A pan-line mean of +0.069 falling to −0.337 in four KMT2A::AFF1 lines is a difference computed from four numbers. Oncofusion knowledge graph (internal analysis) 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ChEMBL v34 target CHEMBL5724736 (AFF1) shows 2 activity rows at pChEMBL >= 6 with an apparent best Kd of 13 nM, but both come from a single 2011 chemoproteomics document in which the compound is the BET bromodomain inhibitor molibresib (CHEMBL1232461); these are affinity-capture co-enrichment readouts of the super elongation complex, not direct AFF1 binding. The same artefact inflates AFF4, ELL, PAF1, SETD1A and ASH2L.

    What this evidence cannot show

    The identification of these rows as a chemoproteomic co-enrichment artefact rests on the source document and the compound identity rather than on an explicit ChEMBL flag, so a reader who trusts the database's data_validity fields alone would not see it. It remains possible that molibresib has some genuine direct affinity for an AFF1 surface, but nothing in the record supports that and the parsimonious reading is complex co-capture.

    Evidence that argues against this (1)
    • The quoted affinity is an OFF-TARGET measurement. A chemoproteomics document reports what a compound pulled down, so a nanomolar number attributed to this protein from a BET bromodomain probe is evidence that the compound binds something else and touched this too. It is not chemical matter and should not enter a tractability count. ChEMBL, EMBL-EBI 2026
  7. E_q72_aff1_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    AFF1'S ENTIRE EXPERIMENTAL SUBCELLULAR ANNOTATION IN UNIPROT IS ONE PAPER FROM 2026, AND IT IS A DNA-DAMAGE PAPER. Fetched from UniProtKB 2026-09-21, P51825 (AFF1_HUMAN, primary gene AFF1, 1,210 aa) carries a single SUBCELLULAR LOCATION value, 'Nucleus', with a single ECO:0000269 attribution to PMID 41062835, plus one uncoded-scope note attributed to the same paper: 'Associates with chromatin and is recruited to DNA damage sites in a PARP1-dependent manner'. There is NO nuclear-localization-signal motif in the feature table and NO subnuclear value of any kind - no speckle, no nucleolus, no chromosome.

    What this evidence cannot show

    A database state on a date. The oddity worth carrying is that the founding AF4 localisation paper (PMID 9808577, 1998, whose TITLE is about subnuclear compartments) is not cited by UniProt for this field, and the paper that is cited is a 2026 study of transcriptional restart after DNA damage. Same shape as menin's record: ECO:0000269 certifies that a curator read an experiment, not that the cited paper is the founding or the best one.

    Evidence that argues against this (1)
    • The absent speckle value is defensible rather than an omission. The AFF-family speckle paper demonstrated speckle localisation for AFF2/FMR2, AFF3/LAF4 and AFF4/AF5q31 and DID NOT include AFF1 - so there is no AFF1 speckle result for UniProt to annotate. This graph already reasons about that gap in localization-nuclear-speckles-kmt2a-r and it should not be re-derived. Melko M et al. 2011
  8. E_q72_aff1_loc2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    IN HUMAN CELLS UNDER GENOTOXIC STRESS, AFF1 ASSOCIATES WITH CHROMATIN AND IS RECRUITED TO DNA DAMAGE SITES IN A PARP1-DEPENDENT MANNER: PARP1 binds and PARylates AFF1 in a region targeted by the E3 ligase Siah1, preventing AFF1 ubiquitination and stabilising it, and that stabilisation supports transcriptional restart after damage. Cells resistant to genotoxic stress carry elevated PARP1 activity and AFF1 levels, and AFF1 depletion impairs DNA repair and survival.

    Stabilization of AFF1 by PARylation ensures transcriptional restart after DNA damage. · Zhu F, Fu H, Zhu W, Li C, Yang M, Feng Z, Shao Y, Li Z, Guo Y, Huang M, Zhang Y, He Y, Lu Y, Wu Z, Zhou Q, Jiao H, Huang J, Zhang L, Xue Y, Lu H · Nat Chem Biol 2026
    PMID 41062835DOI
    What this evidence cannot show

    Abstract read in full; full text not read. NOT A LEUKAEMIA SYSTEM and not stated to be haematopoietic in the abstract; no KMT2A fusion is involved (brief rule 73). This is DAMAGE-INDUCED recruitment - it describes where AFF1 goes under stress, and does not establish a steady-state compartment. 'Associates with chromatin' as a UniProt note is not a subnuclear compartment. NOTE the collateral relevance and the risk in it: this graph carries a menin-PARP experiment and a menin-FANCD2 interface, so an AFF1-PARP1 axis will look like convergence - it is not, because nothing here connects PARP1 to the fusion, to menin, or to leukaemia, and treating it as convergence would be exactly the partner-imported inference this project keeps catching.

    Evidence that argues against this (1)
    • A stress-recruitment result cannot carry the steady-state claim the UniProt line is being read for. The only steady-state images of AF4 protein in leukaemic cells are from 1998 and are UNATTRIBUTABLE: anti-AF4 polyclonals resolve the fusion at ~240 kDa from wild-type AF4 at 125 and 145 kDa on a blot, but in the microscope t(4;11) and non-t(4;11) cells show the SAME punctate nuclear pattern, with the fusion the less abundant species. So the compartment question for AFF1 in this disease has never been answered at steady state by any assay that can tell the two proteins apart. Li Q et al. 1998
  9. E_q72_aff1_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    AFF1'S ONE NAMED SUBNUCLEAR COMPARTMENT MATCHES NOTHING KNOWN. AF4 and AF9 interact, and that interaction is restricted to discrete nuclear foci the authors named 'AF4 BODIES', which are explicitly NON-NUCLEOLAR and 'not coincident with any known nuclear structures we have examined'. The AF4-AF9 interaction is maintained by the MLL-AF4 fusion protein, and expression of MLL-AF4 CAN ALTER THE SUBNUCLEAR LOCALIZATION OF AF9.

    MLL fusion partners AF4 and AF9 interact at subnuclear foci. · Erfurth F, Hemenway CS, de Erkenez AC, Domer PH · Leukemia 2004
    PMID 14603337DOI
    What this evidence cannot show

    Abstract read in full; full text not read. TRANSFECTION AND TAGGED PROTEINS, not endogenous t(4;11) cells - so the species carrying the readout is unambiguous but the expression level is not physiological, which is the condition under which discrete puncta are least informative. This graph already carries the compartment as localization-af4-bodies-a-named-compartment-matching-nothing-known and that entry, not this field, is where the reasoning lives (rule 82). 'AF4 bodies' have never been reproduced, never been given a marker, and never been looked for in a patient sample - in 22 years.

    Evidence that argues against this (1)
    • At physiological expression, low-complexity-domain interactions that control transcription form DYNAMIC HUBS WITHOUT DETECTABLE PHASE SEPARATION, so a discrete punctate structure in an overexpressing cell is weak evidence about the endogenous protein. That applies with full force to AF4 bodies, which exist only in an overexpression system. Chong S et al. 2018
  10. E_q72_aff1_loc4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    THE FUSION AND THE NAMED COMPARTMENT DISAGREE ABOUT THE NUCLEOLUS. AF4 bodies are explicitly non-nucleolar [E_q72_aff1_loc3], but in an inducible expression system human MLL-AF4 is reported as PREDOMINANTLY LOCALISING INSIDE THE NUCLEOLUS, interfering with Pol I transcription and ribosome biogenesis and turning on the nucleolar stress pathway, while a murinized pSer variant localises more to the nucleoplasm. Both results come from expression systems and both are about where the protein sits; they cannot both be the whole story.

    MLL-AF4 and a murinized pSer-variant thereof are turning on the nucleolar stress pathway. · Siemund AL, Hanewald T, Kowarz E, Marschalek R · Cell Biosci 2022
    PMID 35468859DOI
    What this evidence cannot show

    Abstract read in full; full text not read. INDUCIBLE OVEREXPRESSION in a non-leukaemic mammalian host, not endogenous t(4;11) protein (rule 73). The comparison here is not like-for-like - AF4 bodies are wild-type AF4 with AF9, the nucleolar result is the MLL-AF4 CHIMERA - so this is a tension to record rather than a contradiction to resolve, and the honest statement is that the two available subnuclear descriptions of this axis were produced in different constructs, different hosts and eighteen years apart. This graph already carries the nucleolar result in localization-nucleolus-kmt2a-fusion and localization-endogenous-mll-af4-punctate-negative. Confidence is deliberately weak.

    Evidence that argues against this (1)
    • Nucleolar accumulation of an overexpressed basic, disordered protein is a well-known artefact class, and the 1998 endogenous experiment saw punctate nucleoplasmic staining rather than a nucleolar one - though that experiment cannot distinguish fusion from wild-type, so it does not settle the point either. Neither observation has ever been repeated with an endogenously tagged, junction-specific reagent, which is the experiment this graph keeps arriving at. Li Q et al. 1998
  11. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In mouse, Af4 functions as a positive regulator of P-TEFb and, in complex with the MLL fusion partners Af9, Enl and Af10, as a mediator of histone H3-K79 methylation by recruiting Dot1 to elongating Pol II; these pathways are regulated by P-TEFb-dependent phosphorylation of Af4, Af9 and Enl, which controls their transactivation activity and/or protein stability, and the ataxic robotic mouse, an Af4 over-expression model, shows increased levels of phosphorylated Pol II and methylated H3-K79.

    PMID 17135274DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. MOUSE Af4, WILD TYPE and OVER-EXPRESSED, in a cerebellar/ataxia model - not human AFF1, not a leukaemia, and not a fusion protein, so rule 73 applies and the first clause of any downstream claim must say so. The H3-K79 mediation is attributed to a four-component complex (Af4, Af9, Enl, Af10) and no residue of Af4 is mapped to Dot1, so this does NOT supply the missing KMT2A::AFF1-DOT1L interface and must not be cited as though it did. The in vivo readout is bulk methylated H3-K79 in an over-expression mutant, not locus-resolved ChIP at fusion targets. NOT CITED ANYWHERE in this corpus before 2026-09-21, verified with scripts/check-citation-reuse.py and a repo grep.

    Evidence that argues against this (1)
    • The direct opposing result, and the one this graph's sceptical position rests on: in reconstitution from defined, functionally active minimal complexes, DOT1 forms a stable complex with AF9 and DOES NOT interact with AF9-AF4-P-TEFb complexes, while AF4 and AFF4 independently bridge P-TEFb to AF9. Read against Bitoun this is a flat contradiction about whether an AF4-containing complex can carry DOT1L. The component lists suggest a reconciliation - the positive complex contains Af10 and the reconstitution did not - but that is an inference drawn here and stated by neither paper, and until it is tested the honest position is that AFF1's route to H3K79 methylation is CONTESTED, not established. Biswas D et al. 2011
  12. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In reconstitution from defined, functionally active minimal complexes, DOT1 forms a stable complex with AF9 and does not interact with AF9-AF4-P-TEFb complexes; AF4 and AFF4 independently mediate association of P-TEFb with AF9 through direct interactions with AF9 and cyclin T1, and P-TEFb provides the link by which ELL, EAF1 and EAF2 associate with AF4.

    PMID 21896721DOI
    What this evidence cannot show

    Abstract read from the eutils record; already cited in 21 entries of this corpus, so it is recorded here only to hold the two sides of the AFF1 epigenetic question in one place rather than to add a new conclusion. This is a RECONSTITUTION from defined components - a tube experiment. Its negative bounds what a minimal AF9-AF4-P-TEFb assembly does, not what a nuclear AF4-containing complex does in a cell, and specifically it does not test an assembly containing AF10. Its functional arm is MLL-AF9, not the index fusion.

    Evidence that argues against this (1)
    • The reconstitution negative is about protein-protein association, and the functional partition measured in cells is compatible with either answer. CRISPR-ChIP resolves H3K79 methylation into an oncogenic DOT1L unit whose catalysis is directed by the MLL fusion in a menin-dependent manner and a separate endogenous unit, so a fusion CAN direct DOT1L without any of the contacts this reconstitution tested - and that measurement was made in a murine MLL-AF9 system, where an AHD exists. Neither result settles the index fusion, which is why this graph carries the question as a hypothesis with a prior of 0.35 rather than as a finding. Gilan O et al. 2023
  13. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    PARTNER-IMPORT TRAP: the histone-modifying inventory belongs to the RECIPROCAL derivative, not to the index fusion. Affinity purification of the AF4-MLL complex gives at least 16 constituents including P-TEFb kinase, H3K4(me3) and H3K79(me3) histone methyltransferases, a protein arginine N-methyltransferase and a histone acetyltransferase, whereas the AF4 complex alone consists of 11 binding partners with P-TEFb-mediated activation of promoter-arrested Pol II in conjunction with several chromatin-modifying activities.

    PMID 21030982DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read, and this graph records the AF4-MLL affinity purification as unreadable by all five routes (absence-af4-mll-affinity-purification-unreadable-by-all-four-routes), so the component inventory cannot be checked against the paper's own tables here. AF4-MLL is the RECIPROCAL fusion and a different protein from KMT2A::AFF1; the two are routinely conflated in prose because the token order is the only thing distinguishing them. The murine system is haematopoietic progenitor/stem cells expressing AF4-MLL, and whether AF4-MLL is even expressed in patients is contested in this corpus. 'Histone methyltransferases' identified by affinity purification are co-purifying proteins, not demonstrated activities on the fusion's targets.

    Evidence that argues against this (1)
    • A component detected in an affinity purification is not a component of a defined complex, and this project has been burned by exactly that step. The reconstitution study's whole point is that fusion-partner complexes purified through different partners have overlapping components, different elongation activities and different cofactor associations that 'suggest dynamic interactions' - so a sixteen-member inventory is a superset of several assemblies rather than a machine. Reading H3K4 and H3K79 methyltransferase activity into the reciprocal fusion's mechanism from a purification list is the same inference this graph now refuses for the canonical SEC. Biswas D et al. 2011
Connected to