AFF3 (LAF4) — the AFF-family scaffold of SEC-L3, and a KMT2A fusion partner in its own right

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.

• Any interface shared by all four AFF scaffolds cannot be a fusion-selective handle; an agent against it would hit normal SEC, SEC-L2 and SEC-L3 in every dividing cell.
• If a KMT2A::AFF1 cell can substitute AFF3 or AFF4 into the fusion complex when AFF1 function is blocked, scaffold-directed therapy is buffered by paralogue redundancy from the first day of treatment.

1. Does a KMT2A::AFF3 fusion behave like KMT2A::AFF1 in menin dependence and in DOT1L dependence? A head-to-head would test whether the AFF moiety matters at all beyond supplying SEC.
2. Are SEC, SEC-L2 and SEC-L3 present simultaneously in a KMT2A::AFF1 B-lymphoid blast, and do they compete for the same ENL/AF9 pool? If they do, the effective concentration of free ENL is a variable nobody has measured.
3. Would an ENL- or AF9-directed agent inhibit all four SEC species? On the published composition it must, which caps how fusion-selective such an agent can be.
4. AFF3 is thymus-maximal and AFF1 is broadly expressed. Is AFF3 protein present at all in the fetal pro-B cell where KMT2A::AFF1 arises? If it is not, the SEC-L3-versus-SEC competition this entry describes cannot be operating in the cell of origin.
5. Image all four AFF scaffolds in one cell with one tag and one speckle marker. Nobody has done it, and it would tell you in one experiment whether AFF1's 'AF4 bodies' are a real divergence or a reagent difference.
6. KMT2A::AFF3 is a recurrent fusion and has never been localised. Does the chimera go where AFF3 goes (speckles), where KMT2A goes, or somewhere neither goes - the question this graph asks of every fusion and has answered for almost none?

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

   SEC-like 2 (SEC-L2) and SEC-like 3 (SEC-L3), containing AFF2 and AFF3 respectively, were biochemically isolated in association with P-TEFb, ENL/MLLT1 and AF9/MLLT3; all SEC family members show high Pol II CTD kinase activity, but only AFF4-containing SEC is required for proper HSP70 induction on stress; genome-wide mRNA-seq showed SEC-L2 and SEC-L3 control different subsets of genes; and MYC is a direct AFF4/SEC target in acute myeloid and lymphoid leukemia cells.

   The super elongation complex family of RNA polymerase II elongation factors: gene target specificity and transcriptional output. · Luo Z, Lin C, Guest E, Garrett AS, Mohaghegh N, Swanson S, Marshall S, Florens L, Washburn MP, Shilatifard A · Molecular and Cellular Biology 2012
   PMID 22547686DOI
   ▸What this evidence cannot show

   Abstract read in full; not full text. Work is in HEK293 and Jurkat cells on wild-type complexes; no KMT2A fusion protein is tested anywhere. The abstract does not give stoichiometry or state whether a single cell contains all three SEC species simultaneously, so 'SEC-L3 exists' should not be read as 'SEC-L3 competes with SEC in a B-lymphoid blast'.

   ▸Evidence that argues against this (1)

   • Same structural objection as for AFF2, and it matters more here because AFF3 is itself a KMT2A fusion partner. The entry says AFF3 'is not a component of KMT2A::AFF1 and does not bind it'. Chen and Cramer show the AFF4 CHD mediates AFF4 homodimerisation and AFF1-AFF4 heterodimerisation and that the CHD is conserved in AFF1, AFF2, AFF3 and AFF4; the AFF1 CHD (aa 949-1209) is retained in KMT2A::AFF1. So the fusion presents a conserved dimerisation surface to which AFF3's own CHD is, on sequence and fold, a candidate partner. Nobody has tested it. This also means a therapeutic aimed at the CHD interface would engage the fusion itself, not only the free scaffolds - which sharpens rather than weakens the entry's own selectivity argument. Read: abstract in full, not full text. Chen Y et al. 2019
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

   AFF3 has DepMap mean gene effect -0.1789 across 1,208 lines with 2.6% depleted, mean -0.2657 in 11 KMT2A-rearranged leukemia lines against -0.1806 in 43 other leukemia lines (difference -0.085); AFF2 has mean +0.1052 with 0.17% depleted and a difference of -0.006; MEN1 in the same lines is -0.866.

   DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API, with KMT2A-rearrangement status from Cellosaurus; per-gene recomputation for this beat · Broad Institute DepMap; Cellosaurus; Open Targets Platform; analysis run by this project · Database and derived analysis 2026
   ▸What this evidence cannot show

   DepMap release version not stated by the API. Eleven KMT2A-rearranged lines; no permutation test applied to this difference in this entry. A -0.085 difference is within the range this project has previously shown to be indistinguishable from noise at this n.

   ▸Evidence that argues against this (1)

   • AFF3's -0.085 KMT2A-selectivity is described as 'an order of magnitude below MEN1 and should not be over-read', which is right, but the null direction is equally over-readable. A four-member family sharing the CHD dimerisation module and the same ENL/AF9/P-TEFb partners is the textbook setting for paralogue buffering in a single-gene knockout screen; Chen and Cramer supply the structural basis. Neither the small negative nor the near-zero AFF2 value licenses a conclusion about the family's requirement. Read: abstract in full, not full text. Chen Y et al. 2019
3. E_q89_aff3_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt P51826 (AFF3_HUMAN, primary gene AFF3, synonym LAF4, 1,226 aa, read back 2026-09-21) carries one subcellular location value, 'Nucleus', WITH AN EMPTY EVIDENCE LIST - no experimental, by-similarity or rule-based attribution of any kind. Its TISSUE SPECIFICITY comment reads: 'Preferentially expressed in lymphoid tissues, highest levels being found in the thymus.'

   UniProtKB P51826 (AFF3_HUMAN) SUBCELLULAR LOCATION and TISSUE SPECIFICITY comments, fetched 2026-09-21 · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Established as a statement about what the database contains, not about the protein. An uncoded location comment is the weakest annotation class this graph meets, and this project has already measured that uncoded UniProt localisations account for 10.7% of its entries. The tissue-specificity comment is likewise uncoded and gives no cell-type resolution below 'lymphoid tissue'; it is not a sorted-population or single-cell measurement.

   ▸Evidence that argues against this (1)

   • An uncoded nuclear call should not be dismissed just because it is uncoded, and here the independent support is unusually strong: Melko et al. imaged AFF3 directly and found it in nuclear speckles, which is a nuclear result arrived at without reference to UniProt. The genuine weakness is elsewhere, in the TISSUE comment rather than the location comment - 'preferentially expressed in lymphoid tissues, highest in thymus' is a bulk-tissue statement of the kind rule 45 warns about for GTEx, with no protein measurement, no cell-type resolution and no date. For a gene whose interest in this graph is competition with AFF1 inside a B-lineage progenitor, a thymus-level statement cannot tell you whether AFF3 protein is present in a pro-B cell at all, and this entry should not be read as saying it is. Melko M et al. 2011
4. E_q89_aff3_loc2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   AFF3/LAF4 LOCALISES TO NUCLEAR SPECKLES - the subnuclear stores of pre-mRNA splicing factors - shown in the same experiments that showed it for AFF4/AF5q31, and following the same group's earlier result for AFF2/FMR2. All three bind RNA with high apparent affinity for the G-quadruplex structure and modulate in-vivo splicing efficiency of a minigene containing a G-quadruplex in an alternatively spliced exon. The authors further observed that OVEREXPRESSION OF AFF2/3/4 INTERFERES WITH THE ORGANISATION AND/OR BIOGENESIS OF NUCLEAR SPECKLES, and that enlarged nuclear speckles are present in FRAXE patient fibroblasts.

   Functional characterization of the AFF (AF4/FMR2) family of RNA-binding proteins: insights into the molecular pathology of FRAXE intellectual disability. · Melko M, Douguet D, Bensaid M, Zongaro S, Verheggen C, Gecz J, Bardoni B · Hum Mol Genet 2011
   PMID 21330300DOI
   ▸What this evidence cannot show

   FRAXE intellectual-disability cell biology, not haematopoiesis: no leukaemic cell, no KMT2A fusion, no B-lineage progenitor. The abstract does not name the cell line used for the AFF3 imaging and this beat did not obtain full text, so the system is unstated (rule 73 - the claim above therefore does not name one). Graded moderate rather than strong for the reason the authors themselves supply: overexpressing these proteins distorts speckles, so the assay perturbs the structure it reports. This is the SAME citation gene-aff4 carries as E_q72_aff4_loc2; the identifiers here were re-fetched from PubMed efetch on 2026-09-21 rather than copied from that block.

   ▸Evidence that argues against this (1)

   • The paper's own finding undercuts its own assay, and the entry should say which way that cuts. If overexpressing AFF2/3/4 disorganises nuclear speckles, then a transfected AFF3 construct sitting in a speckle may be sitting in a structure it created or deformed, and the observation risks being circular. Two things keep it standing rather than collapsing it. First, the FRAXE FIBROBLAST observation is independent of transfection: enlarged speckles in patient cells lacking AFF2 is an endogenous phenotype, and it implies the family genuinely acts on speckle architecture rather than merely decorating it. Second, the alternative explanation - that these are generic overexpression aggregates - does not predict the RNA-binding specificity for G-quadruplexes or the minigene splicing effect, which are functional readouts tied to what speckles actually do. The residual and unresolved problem for THIS graph is different and larger: the one AFF-family member excluded from these experiments is AFF1, the partner in the index fusion, so the family consistency this entry reports cannot be extended to the protein the project is about. UniProt Consortium 2026
5. E_q89_aff3_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   The AFF-family compartment picture is internally consistent for three members and broken for the fourth, and the fourth is the index fusion's partner. gene-aff4 records NUCLEAR SPECKLES from PMID 21330300; this entry records NUCLEAR SPECKLES for AFF3 from the same paper; AFF2/FMR2 was shown earlier by the same group. gene-aff1 records that AFF1 was NOT included in that paper and that its only named compartment is 'AF4 bodies', foci the original authors described as explicitly non-nucleolar and 'not coincident with any known nuclear structures we have examined'.

   Functional characterization of the AFF (AF4/FMR2) family of RNA-binding proteins: insights into the molecular pathology of FRAXE intellectual disability. · Melko M, Douguet D, Bensaid M, Zongaro S, Verheggen C, Gecz J, Bardoni B · Hum Mol Genet 2011
   PMID 21330300DOI
   ▸What this evidence cannot show

   This is a synthesis across three graph entries plus one paper, not a single experiment, and it is graded moderate on that basis. The AFF1 half rests on gene-aff1's reading of a 2004 transfection study that has never been reproduced. No experiment has ever imaged two AFF-family members in the same cell with the same reagent, which is what would be needed to say the family DIVERGES rather than that it was studied unevenly.

   ▸Evidence that argues against this (1)

   • The tidy reading - three speckle proteins, one outlier - may be an artefact of who studied what, and that rival explanation is at least as likely as real divergence. AFF2, AFF3 and AFF4 were imaged together, by one group, with one construct design, in one cell system, for a FRAXE question; AFF1 was imaged separately, seven years earlier, by a different group, in a leukaemia question, with different tags. Two laboratories using different reagents on different proteins will report different compartments whether or not the proteins differ, and 'AF4 bodies' - a compartment defined by matching nothing else - is exactly what a one-off imaging experiment produces when nobody runs the marker panel. So the honest state is that the family's compartments have never been compared under a common assay, and the apparent AFF1 exception could be method rather than biology. It is also cheap to settle, and the biochemical half has already been done this way: Luo et al. purified SEC, SEC-L2 and SEC-L3 under one protocol and compared their outputs directly, which is exactly the common-assay design the imaging side lacks. Luo Z et al. 2012

• related to →KMT2A::AFF1 oligomerises through the retained AFF1 CHD, putting the commonest fusion in the oligomerisation class

  same: AFF3 carries the conserved CHD, so 'does not bind KMT2A::AFF1' is untested rather than established

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

  AFF-family paralogue; SEC-L3 scaffold against AFF1's SEC — same ENL/AF9/P-TEFb partners, different gene sets

• binds →MLLT1 / ENL — the reader that beat its own paralogue, with a degrader already in hand

  SEC-L3 contains ENL/MLLT1, the same partner AFF1 and AFF4 bind

• binds →MLLT3 / AF9 — a famous fusion partner that is not a dependency, and a hub domain that is

  SEC-L3 contains AF9/MLLT3

• fuses with →KMT2A / MLL1 — the gene that breaks

  KMT2A::AFF3, t(2;11)(q11;q23) — AFF3 is itself a recurrent 3' partner

• ← related toAFF2 (FMR2) — the SEC-L2 scaffold, and the AFF-family member with no dependency anywhere

  SEC-L2 and SEC-L3 were isolated together and control different gene subsets

• ← regulatesKDM3B - the H3K9me1/2 demethylase that sits on the KMT2A breakpoint cluster region and keeps it from rearranging, and the only leukaemia gene in this graph whose direction of effect flips between the myeloid and the lymphoid literature

  KDM3B binds across AFF3/LAF4 and its depletion causes AFF3 copy gains - transient, not inherited

• ← related toThe paper's four comparison loci, marks read against FISH calls: the H3K9 mark response agrees with the cytogenetic response at two loci and contradicts it at two - the mark is neither necessary nor sufficient

  copy gain reported without an H3K9me1 rise (0.86x over 570 kb of gene body)

• ← related toSEC-L2 and SEC-L3 contain ENL and AF9 exactly as SEC does, so AFF2 and AFF3 sit in complexes built around the same acyl-lysine readers as the index fusion's partner - and both entries carry no epigenetics annotation at all

  supplies the complex-level epigenetics record gene-aff3 lacks: SEC-L3 contains ENL and AF9

• part of →SEC-L3 — the AFF3-scaffolded complex, and the only SEC-family member with a known physiological job: genomic imprinting and X inactivation

  AFF3 is the scaffold subunit of SEC-L3

• ← related toThe SEC-family target sets, reconstructed as a full three-way Venn: 27.5% of the family's transcriptional output is invisible to AFF4 depletion — and the two 'distinct' sibling sets overlap 8-20x more than chance

  AFF3 knockdown response set, partitioned

• ← related toThree of the four AFF scaffolds have a named human Mendelian syndrome and the index fusion's partner has none — a 0/4 that is a fact about AFF1 and not about how hard anyone has looked

  AFF3 -> KINSSHIP syndrome, de novo variants in the degron

• related to →Nuclear speckles — the compartment everyone assumes KMT2A::AFF1 uses, on evidence that skips AFF1 itself

  AFF3/LAF4 localises to nuclear speckles (PMID 21330300), the same paper and the same result as AFF4; AFF1 was excluded from that study (Q89 2026-09-21)

• related to →'AF4 bodies' - a named subnuclear compartment that is non-nucleolar and coincident with no known nuclear structure, and the one place a KMT2A fusion has been shown to move a partner

  Family contrast: AFF2/AFF3/AFF4 are speckle proteins while AFF1's only named compartment is AF4 bodies, explicitly not coincident with any known structure - three laboratories' worth of assay difference, never controlled (Q89 2026-09-21)

• expressed in →Early T-cell precursor (ETP) — the thymocyte that has not yet given up being myeloid, and where KMT2A-rearranged T-ALL sits

  UniProt P51826 tissue specificity: preferentially lymphoid, HIGHEST IN THYMUS - the most lineage-restricted of the four AFF scaffolds, and an awkward competitor to place against AFF1 in a B-lineage disease (Q89 2026-09-21)