Oncofusion

start / graph / mechanism-kmt2a-aff1-cannot-be-sequestered-in-the-7sk-snrnp

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q35-ptefb-reservoir

The fusion begins at AFF1 residue 347 and the cyclin T1-binding sequence ends at 308, so KMT2A::AFF1 is the one AFF-family protein that cannot enter the reservoir

also called fusion excluded from 7SK snRNP · AFF1 CBS lost in KMT2A::AFF1

Normal AFF1 carries a short stretch near its front end that lets it dock onto the elongation enzyme, and that same stretch is what pulls AFF1 into the off-switch particle. The leukaemia fusion protein starts after that stretch, so it can never be switched off that way - it has to borrow the enzyme from a normal copy of AFF1 or from AFF4 instead.

The same thing, for a scientist

Lu and colleagues mapped AFF1's entry into the 7SK snRNP to its N-terminal cyclin T1-binding sequence: AFF1 lacking residues 1-308 loses binding to P-TEFb, HEXIM1 and LARP7 while keeping its ELL2 and ENL contacts, and AFF1 1-77 or 38-72 alone is sufficient to join the particle. AFF1 contacts the rest of the snRNP only through P-TEFb and 7SK RNA [E1]. The KMT2A::AFF1 fusion does not contain that region. All three recurrent breakpoint architectures splice to the same AFF1 acceptor and produce the same AFF1 head peptide QTYSNEVH [E2], which locates to **residue 347** of UniProt P51825 [E3]. Residues 1-346 of AFF1 - the whole cyclin T1-binding sequence and the ALF/N-terminal homology domain - are absent from the fusion protein. Three consequences follow, and none of them is stated anywhere else in this graph: the fusion cannot be sequestered in the reservoir; the fusion cannot extract P-TEFb from the reservoir by the route AFF1 uses; and the fusion's access to P-TEFb must run in trans, through wild-type AFF1 or through AFF4 bound at the AFF-family CHD heterodimerisation interface [E4].

Go deeper
The arithmetic, laid out so it can be checked
quantityvaluesource
AFF1 canonical length1,210 aa (P51825)UniProt, fetched 2026-09-20 3
AFF1 region required to join the 7SK snRNP1-308, minimal element 38-72Lu et al. deletion series 1
AFF1 head peptide of the fusionQTYSNEVHthis repository's junction-peptide analysis 2
Position of QTYSNEVH in P51825347string search of the fetched sequence 3
AFF1 residues absent from the fusion1-346subtraction

The same AFF1 acceptor is used by all three recurrent KMT2A architectures (breaks at KMT2A ~1362/1363, ~1406/1407 and ~1444/1445), so this conclusion does not depend on which breakpoint a patient carries 2.

Two existing entries disagree on this and one of them is wrong

This is reported rather than edited, per the standing rule.

  • interface-aff4-ptefb-ccnt1-n-terminal already reaches the right conclusion - that the AFF4 2-73-equivalent region is not in the fusion - but states the fusion retains AFF1 'from roughly residue 100 onwards'. The measured figure is 347.
  • fusion-kmt2a-aff1 lists ALF / N-terminal homology domain under domains_retained_3 for the bp-kmt2a-intron-11-aff1-e4 breakpoint and, from the same field structure, for the others. Other entries in this graph place the ALF/NHD at approximately AFF1 aa 1-100. If that placement is right, the ALF domain is lost, not retained, and the domains_retained_3 / domains_lost_3 arrays are inverted for that module.

The citation that settles both is the pair 2 + 3: the junction head peptide, computed in this repository from the RefSeq transcripts, matched against the fetched UniProt sequence. This reader ran the string match; it is not taken from another entry.

What the fusion can and cannot do to P-TEFb

Cannot: be captured into the 7SK snRNP; bind cyclin T1 directly through an AFF-family CBS; extract P-TEFb by the mechanism Lu et al. describe for AFF1, in which AFF1's CBS raises Tat's affinity for cyclin T1 1.

Can: retain the pSER region and the C-terminal homology domain, which mediate AFF-family homo- and hetero-dimerisation and the ENL/AF9 contact 4. So the fusion's plausible route to P-TEFb is to bind AFF4 or wild-type AFF1 through the CHD and inherit their cyclin T1 contact - a two-step, in-trans recruitment.

This matters for drugs. A compound aimed at the AFF-family/cyclin T1 interface would hit the wild-type protein that the fusion borrows, not a fusion-specific surface, and would therefore act on the fusion only indirectly. It also predicts that the fusion's P-TEFb occupancy should depend on AFF4 dosage, which is testable and, as far as this beat can find, untested.

Honest limits

The deletion mapping was done in HeLa nuclear extracts with overexpressed AFF1 fragments, in an HIV transactivation study 1. The residue arithmetic is exact but it establishes an absence of a known module, which is weaker than a measured absence of binding. Nobody has immunoprecipitated a KMT2A::AFF1 fusion protein and blotted for HEXIM1, LARP7 or 7SK. That experiment is the direct test and it is written up in exp-glycerol-gradient-ptefb-partition-kmt2a-aff1.

Facts and figures
mechanism classmodule-loss consequence
fusionfusion-kmt2a-aff1
aff1 first retained residue347
aff1 residues lost1-346
aff1 cyclin t1 binding sequence1-308 (minimal 38-72)
conclusionthe fusion protein cannot join the 7SK snRNP and cannot extract P-TEFb by the AFF1 route; access to P-TEFb is necessarily in trans via AFF4 or wild-type AFF1
breakpoint sidethe lost side of AFF1 - this is a 3' partner module-loss finding, and the graph had no such finding for the reservoir
direct test absentno published co-immunoprecipitation of a KMT2A fusion protein blotted for HEXIM1, LARP7 or 7SK
corrections flagged
[
  {
    "entry": "interface-aff4-ptefb-ccnt1-n-terminal",
    "field": "summary_expert",
    "wrong": "KMT2A::AFF1 retains AFF1 from roughly residue 100 of AFF1 onwards",
    "right": "from residue 347",
    "settled_by": "analyses/junction-peptides-all-breakpoints (AFF1 head peptide QTYSNEVH) matched against UniProt P51825 fetched 2026-09-20"
  },
  {
    "entry": "fusion-kmt2a-aff1",
    "field": "data.breakpoints[].domains_retained_3",
    "wrong": "lists 'ALF / N-terminal homology domain' as retained",
    "right": "if the ALF/NHD is aa ~1-100 as other entries in this graph state, it is lost, not retained",
    "settled_by": "same"
  }
]
related entries
[
  "fusion-kmt2a-aff1",
  "gene-aff1",
  "gene-aff4",
  "interface-aff4-ptefb-ccnt1-n-terminal",
  "interface-aff1-aff4-chd-heterodimer",
  "protein-7sk-snrnp"
]
How this could bite us back
  • If the fusion reaches P-TEFb only in trans, then any drug aimed at the AFF-family/cyclin T1 interface hits the wild-type protein, and its therapeutic index is the index of blocking normal SEC assembly.
  • The absence of a CBS also means the fusion cannot be switched off by anything that works through the reservoir, so raising HEXIM1 or 7SK can only act on the shared free P-TEFb pool, never on the fusion specifically.
Still open
  1. Does an antibody against the fusion's KMT2A half co-precipitate HEXIM1, LARP7 or 7SK from SEM or RS4;11 nuclear extract? A clean negative would confirm this by measurement rather than by arithmetic.
  2. Does the fusion's chromatin occupancy at HOXA9 depend on AFF4 dosage, as the in-trans model requires?
  3. Do the reciprocal der(4) product and the other AFF-family fusions (KMT2A::AFF3, KMT2A::AFF4) retain their CBS? If any of them does, that fusion would be sequestrable and the comparison would be informative.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    AFF1 requires its N-terminal cyclin T1-binding sequence to associate with the 7SK snRNP: deletion of residues 1-308 abolishes binding to P-TEFb, HEXIM1 and LARP7 while preserving the ELL2 and ENL contacts, AFF1 1-308, 1-77 and 38-72 each suffice to join the particle, and AFF1 contacts the rest of the complex only through P-TEFb and 7SK RNA.

    AFF1 is a ubiquitous P-TEFb partner to enable Tat extraction of P-TEFb from 7SK snRNP and formation of SECs for HIV transactivation. · Lu H, Li Z, Xue Y, Schulze-Gahmen U, Johnson JR, Krogan NJ, Alber T, Zhou Q · Proc Natl Acad Sci U S A 2014
    PMID 24367103DOI
    What this evidence cannot show

    Co-immunoprecipitation with overexpressed deletion constructs in HeLa nuclear extracts, within an HIV-1 Tat transactivation study. AFF1(38-72) showed the weakest binding of the three sufficient fragments. No leukaemia cell and no KMT2A fusion appears anywhere in the paper.

    Evidence that argues against this (1)
    • A deletion series cannot exclude a second, weaker contact below co-IP sensitivity, and overexpressed fragments can occupy a complex differently from endogenous full-length protein. Schulze-Gahmen U et al. 2013
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    All three recurrent KMT2A breakpoint architectures splice to the same AFF1 acceptor (nt 1175 of NM_005935.4) and produce the same AFF1-derived head peptide QTYSNEVH.

    What this evidence cannot show

    An internal computation in this repository from RefSeq transcripts, not an experimental sequencing of patient fusion transcripts. It covers the three UniProt-annotated recurrent breakpoints and not rare variant junctions.

    Evidence that argues against this (1)
    • This graph's own fusion entry records one case in 1,262 in which the genomic break lies upstream of AFF1 exon 1, so the acceptor is essentially but not absolutely invariant. Oncofusion Knowledge Graph et al. 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The peptide QTYSNEVH occurs once in the fetched UniProt P51825 (AFF1_HUMAN, 1,210 aa) canonical sequence, beginning at residue 347, so the KMT2A::AFF1 fusion lacks AFF1 residues 1-346 and therefore the entire cyclin T1-binding sequence at 1-308.

    What this evidence cannot show

    A sequence string match performed in this session against the FASTA fetched the same day. It assumes the canonical UniProt isoform corresponds to the transcript used in the junction computation; the AFF1 Ensembl canonical transcript ENST00000395146 encodes 1,218 aa rather than 1,210, so isoform choice shifts residue numbering by a small offset. That offset is far smaller than the 38-residue gap between 308 and 347, so the conclusion is robust to it.

    Evidence that argues against this (1)
    • Residue arithmetic establishes that a module is absent from the protein, not that the protein fails to bind. Only a co-IP of the fusion against HEXIM1, LARP7 or 7SK can convert this into a measured negative, and none has been published. Lu H et al. 2014
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The AFF-family N-terminal region binds P-TEFb by draping over cyclin T1 without stable contact with CDK9, and AFF-family proteins homo- and hetero-dimerise through their C-terminal homology domains, which the fusion retains.

    The AFF4 scaffold binds human P-TEFb adjacent to HIV Tat. · Schulze-Gahmen U, Upton H, Birnberg A, Bao K, Chou S, Krogan NJ, Zhou Q, Alber T · Elife 2013
    PMID 23471103DOI
    What this evidence cannot show

    The crystal structure is of AFF4 fragments with P-TEFb at 2.94 A; the CHD dimerisation statement is drawn from the AFF-family literature this graph already carries at interface-aff1-aff4-chd-heterodimer and is not itself a result of the cited structure. No structure of AFF1 with cyclin T1 exists.

    Evidence that argues against this (1)
Connected to