The AFF-family N-terminus on cyclin T1: Kd 10 nM, a 2.94 A crystal, and every number in it measured on AFF4 rather than on AFF1
also called AFF4 2-73 P-TEFb · SEC P-TEFb contact · 4IMY
The scaffold protein of the super elongation complex grips the CDK9 kinase through its first seventy amino acids, and the grip is very tight. The catch for this project is that the crystal and the affinity are for AFF4, the sister protein, not for AFF1, the one inside the infant leukaemia fusion.
The same thing, for a scientist
PDB 4IMY (2.94 A) contains CDK9, cyclin T1 and an AFF4 fragment whose deposited sequence aligns to Q9UHB7 residues ~2-73. Purified AFF4 2-73 binds P-TEFb with Kd = 10 +/- 1 nM, and AFF4 2-363 with 7 +/- 1 nM; a shorter AFF4 32-67 peptide binds cyclin T1 alone at 104 nM, and alanine scanning of cyclin T1 identifies W207 (6,050 nM), W210 (2,190 nM) and R165 (1,592 nM) as the dominant hot spots. The interface is entirely within the AFF4 N-terminus. KMT2A::AFF1 retains AFF1 from residue **347** onwards - **corrected 2026-09-20 from 'roughly residue 100', which was wrong by about 250 residues**, so the region homologous to AFF4 2-73 is on the AFF1 N-terminal side and is NOT part of the fusion protein - the fusion reaches P-TEFb through AFF1's other motifs or through a partner, not through this contact. That distinction is not drawn anywhere else in this graph.
What the structure actually contains
Fetched from RCSB on 2026-09-20: three polymer entities - CDK9 (P50750, 332 residues), cyclin T1 (O60563, 264 residues) and AFF4 (Q9UHB7, 75 residues) - at 2.94 A 3. Three related deposits (4OGR, 4OR5, 5L1Z, 6CYT) add HIV-1 Tat and TAR RNA to the same core. Every one of the five is an HIV structural-virology paper. The SEC's P-TEFb contact has been solved five times, always in the context of Tat.
AFF4 is not AFF1, and the graph should stop eliding them
The entire quantitative and structural description of the SEC's P-TEFb contact is AFF4. The PDB contains exactly one structure of any part of AFF1 - the AF9 chimera 2LM0 - and none of AFF1 with P-TEFb, ELL, ELL2 or AFF4 5.
For the index fusion this matters twice over. First, the fusion contains the AFF1 C-terminus, and the P-TEFb contact characterised here is an N-terminal one. Second, AFF1 and AFF4 heterodimerise through their C-terminal homology domains, so the fusion's most plausible route to this interface is in trans, through a bound AFF4 molecule - which is a different mechanism with different drug consequences than a direct contact.
Correction: the AFF1 side of the junction starts at 347, and the conclusion gets stronger
This entry said the fusion retains AFF1 "from roughly residue 100 ... depending on breakpoint". It is residue 347, and it does not depend on the breakpoint.
Verified by the coordinator on 2026-09-20 rather than taken on report: the junction head peptide QTYSNEVH - identical across all three recurrent KMT2A architectures per analyses/junction-peptides-all-breakpoints - was matched against the canonical P51825 FASTA (1,210 aa) fetched from UniProt. It begins at index 346, residue 347, in the context ...LKVPAKAKLTKLKMPSQSVE|QTYSNEVH|CVEEILKEMTHS.... The three recurrent breakpoints vary on the KMT2A side (1362/1363, 1406/1407, 1444/1445); the AFF1 side does not move.
This entry's conclusion was right and is now much stronger. The region homologous to AFF4 2-73 was already argued to be absent from the fusion. At 347 rather than 100 the fusion also lacks the whole of AFF1 1-308, which is the cyclin T1-binding sequence (minimal 38-72) by which AFF1 enters the 7SK snRNP and reaches P-TEFb; an AFF1 delta-1-308 construct loses P-TEFb, HEXIM1 and LARP7 while keeping ELL2 and ENL.
So the fusion has no direct route to P-TEFb at all, not merely no AFF4-homologous one, and its access is necessarily in trans through AFF4 or wild-type AFF1. The mirror statement is the interesting one: that module travels on the reciprocal AFF1::KMT2A product, which is a mechanical reading of why AF4-MLL drives flavopiridol-sensitive elongation where MLL-AF4 does not.
| partner a | gene-aff4 |
|---|---|
| partner b | protein-ptefb-cdk9-ccnt1 |
| motif a | AFF4 N-terminal region, residues 2-73 (crystallised) with the minimal cyclin T1-binding peptide at 32-67 |
| motif b | cyclin T1 surface adjacent to the Tat-binding site |
| kd NM | 10 |
| fate in fusion | not applicable directly - the characterised contact is in the AFF4 (or homologous AFF1) N-terminus, which is on the 5' side of the AFF1 breakpoint and therefore NOT in KMT2A::AFF1; the fusion's plausible route to this contact is in trans via an AFF1-AFF4 heterodimer |
| druggability | a nanomolar peptide-in-groove contact on cyclin T1 with well-defined tryptophan hot spots; no reported small-molecule ligand, and the same surface is the HIV-1 Tat therapeutic target |
| resolution note | PDB metadata and entity sequences fetched from data.rcsb.org 2026-09-20; AFF4 span aligned to UniProt Q9UHB7 fetched the same day. Absences established by RCSB search API, queries recorded in the evidence caveats. |
interface residues b
[ "CCNT1 W207", "CCNT1 W210", "CCNT1 R165", "CCNT1 F176", "CCNT1 D169", "CCNT1 E211", "CCNT1 Y175" ]
kd measurements — the numbers behind the chart
[
{
"ligand": "AFF4 2-73 + P-TEFb",
"kd_nM": 10,
"error_nM": 1,
"method": "quantitative in vitro binding",
"source_pmid": "23471103"
},
{
"ligand": "AFF4 2-363 + P-TEFb",
"kd_nM": 7,
"error_nM": 1,
"method": "quantitative in vitro binding",
"source_pmid": "23471103"
},
{
"ligand": "AFF4 2-73 + Tat-P-TEFb",
"kd_nM": 0.85,
"error_nM": 0.15,
"method": "quantitative in vitro binding",
"source_pmid": "23471103"
},
{
"ligand": "AFF4 32-67 + cyclin T1 (wild type)",
"kd_nM": 104,
"error_nM": 17,
"method": "binding assay, Table 3",
"source_pmid": "23471103"
},
{
"ligand": "AFF4 32-67 + cyclin T1 W207A",
"kd_nM": 6050,
"error_nM": 871,
"method": "binding assay",
"source_pmid": "23471103"
},
{
"ligand": "AFF4 32-67 + cyclin T1 W210A",
"kd_nM": 2190,
"error_nM": 246,
"method": "binding assay",
"source_pmid": "23471103"
},
{
"ligand": "AFF4 32-67 + cyclin T1 R165A",
"kd_nM": 1592,
"error_nM": 171,
"method": "binding assay",
"source_pmid": "23471103"
}
]structures
[ "4IMY", "4OGR", "4OR5", "5L1Z", "6CYT" ]
structure details
[
{
"pdb": "4IMY",
"title_fetched": "The AFF4 scaffold binds human P-TEFb adjacent to HIV Tat",
"method": "X-ray",
"resolution_A": 2.94,
"contents": "CDK9 (P50750) + cyclin T1 (O60563) + AFF4 (Q9UHB7 ~2-73)",
"fetched": "2026-09-20"
}
]structures absent
[ "no structure of AFF1 with CDK9 or cyclin T1", "no structure of a KMT2A fusion with P-TEFb", "all five deposits of this interface are HIV-1 structural-virology studies" ]
- Wrong molecule: every number here is AFF4, and the index fusion contains AFF1; assuming equivalence is exactly the substitution this graph flags elsewhere as partner-imported evidence.
- Wrong end: the contact is N-terminal and the fusion keeps the AFF1 C-terminus, so a ligand for this interface may not touch the fusion at all.
- Essential function: P-TEFb is required for the great majority of RNA polymerase II elongation, so disrupting its scaffold contact has a broad toxicity floor.
- Redundancy: cyclin T2 substitutes for cyclin T1 in P-TEFb, so a cyclin T1-surface ligand has an immediate paralogue escape.
- Does AFF1 bind P-TEFb with the same 10 nM affinity as AFF4, or does the family differ? No AFF1 measurement exists.
- In a KMT2A::AFF1 cell, is P-TEFb held by the fusion directly, by a co-bound AFF4 molecule, or by ENL/AF9 through the AHD? The three routes predict different consequences for an AHD-directed drug.
- Would a cyclin T1 W207/W210 pocket ligand, developed for HIV, have any activity in KMT2A-rearranged leukaemia? Nobody has tried.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
AFF4 2-73 binds P-TEFb with Kd 10 nM and AFF4 2-363 with 7 nM; AFF4 32-67 binds cyclin T1 with Kd 104 nM, and cyclin T1 W207A, W210A and R165A weaken it 58-, 21- and 15-fold respectively.
The AFF4 scaffold binds human P-TEFb adjacent to HIV Tat. · Schulze-Gahmen U, Upton H, Birnberg A, et al. · eLife 2013What this evidence cannot show
All measurements are on recombinant AFF4 fragments and recombinant P-TEFb in vitro; the paper's framing and its tightest measurements involve HIV-1 Tat. No measurement in it involves AFF1 or any KMT2A fusion. INTERACTION PROVENANCE (Q76, 2026-09-21, live IntAct REST), which shows the asymmetry is not confined to the structures: AFF4-CDK9 is 13 IntAct records all psi-mi:MI:1060 spoke-expanded, AFF4-CCNT1 is 8 records of which 7 are spoke-expanded and one is MI:0915 physical association - but the AFF4 arm is rescued by structure, sharing PDB 4IMY, 4OGR, 4OR5, 5L1Z and 6CYT with both kinase subunits. The AFF1 arm has neither: AFF1-CDK9 is 16 records, ALL spoke-expanded, AFF1-CCNT1 is 8 records, all spoke-expanded, no MI:0407 record on either, and no shared PDB entry. Across all of IntAct, AFF1 has exactly one partner with a direct-interaction record (MLLT3) against menin's eight. So the entire evidential basis for an AFF-family N-terminal P-TEFb interface is AFF4, in both databases, and the transfer to AFF1 has never been tested binary. See ana-aff1-has-one-observed-binary-interaction-and-forty-two-residues-of-structure and hyp-aff1-reaches-ptefb-only-through-aff4.
Evidence that argues against this (1)
- AFF4 dimerises through its C-terminal homology domain, and dimerisation itself modulates activation of proviral transcription, so an affinity measured on a monomeric N-terminal fragment is not the avidity of the assembled scaffold. Tang D et al. 2020
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AFF4 dimerisation through the C-terminal domain affects its activation of transcription, and a separate crystal form of the AFF4 C-terminal domain has been solved.
Structural and functional insight into the effect of AFF4 dimerization on activation of HIV-1 proviral transcription. · Tang D, Chen C, Liao G, et al. · Cell Discov 2020What this evidence cannot show
An HIV-1 proviral transcription system; the functional readout is viral, not leukaemic.
Evidence that argues against this (1)
- Chen and Cramer's independent structure of the same domain attributes dimerisation to helices alpha5/alpha6 and a 1,655 square-angstrom interface, and the two papers describe the domain's behaviour with different emphases. Chen Y et al. 2019
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PDB 4IMY is a 2.94 A crystal of CDK9, cyclin T1 and an AFF4 fragment mapping to Q9UHB7 ~2-73.
PDB 4IMY: The AFF4 scaffold binds human P-TEFb adjacent to HIV Tat · Schulze-Gahmen U, Upton H, Birnberg A, et al. · RCSB Protein Data Bank 2013What this evidence cannot show
Title, method, resolution and all three entity sequences fetched from data.rcsb.org 2026-09-20; the AFF4 span is this cycle's alignment to the UniProt Q9UHB7 sequence fetched the same day. The deposited AFF4 entity begins with a three-residue expression artefact that must be stripped before alignment.
Evidence that argues against this (1)
- The four related deposits all include HIV-1 Tat, which binds adjacent to AFF4 and mutually increases affinity roughly elevenfold, so the Tat-free geometry is represented by one structure and the Tat-bound geometry by four. Schulze-Gahmen U et al. 2013
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
No structure in the PDB contains AFF1 together with CDK9, cyclin T1, ELL, ELL2 or AFF4.
PDB 2LM0: Solution structure of the AF4-AF9 complex · Leach BI, Kuntimaddi A, Schmidt CR, et al. · RCSB Protein Data Bank 2013What this evidence cannot show
Established by RCSB search on 2026-09-20 using exact_match on reference_sequence_identifiers.database_accession for P51825 paired in turn with P50750, O60563, P55199, O00472 and Q9UHB7; each pairing returned an empty result set (HTTP 204). A search-based absence is bounded by the query, and structures deposited without a UniProt cross-reference would not be found.
Evidence that argues against this (1)
- 2LM0 does contain AFF1, fused to AF9 in a single chain, so AFF1 is not structurally untouched - it is structurally represented once, as part of a chimera. Leach BI et al. 2013
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The single PDB entry containing AFF1 is 2LM0, a one-chain AFF1/AF9 chimera annotated to both P51825 and P42568.
PDB 2LM0: Solution structure of the AF4-AF9 complex · Leach BI, Kuntimaddi A, Schmidt CR, et al. · RCSB Protein Data Bank 2013What this evidence cannot show
Fetched from data.rcsb.org 2026-09-20; the AFF1 portion aligns to P51825 ~737-779 and the AF9 portion to P42568 490-568. Being a single polypeptide, the deposit cannot report on the entropic cost of the bimolecular association.
Evidence that argues against this (1)
- The affinity for the same pair was measured separately by fluorescence anisotropy on an MBP-AHD fusion at KD 0.17 nM, so the chimeric structure is backed by a bimolecular affinity even though the structure itself is not bimolecular. Leach BI et al. 2013
- E-junction-347established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The KMT2A::AFF1 junction head peptide QTYSNEVH begins at residue 347 of the canonical AFF1 sequence (UniProt P51825, 1,210 aa), so the fusion lacks AFF1 1-346 for all three recurrent KMT2A-side breakpoints.
UniProtKB P51825 (AFF1_HUMAN) canonical sequence, 1,210 aa, FASTA, with the KMT2A::AFF1 junction peptide QTYSNEVH located by string match · UniProt Knowledgebase, reviewed (Swiss-Prot) 2026What this evidence cannot show
String match against the fetched canonical FASTA on 2026-09-20, run by the coordinator independently of the agent that reported it. **It is a fact about the canonical isoform**: AFF1 has other isoforms, and a fusion transcript spliced to a different AFF1 first exon would place the junction elsewhere. The peptide's invariance across the three architectures comes from this project's own junction analysis, not from an external source.
Evidence that argues against this (1)
- An AFF1 residue number is not a domain boundary. Whether AFF1 1-346 contains every module the fusion needs is a separate question from where the junction sits, and the delta-1-308 experiment that motivates the reading is a deletion construct in an overexpression system, not the fusion. PNAS 2014
AFF4 2-73 on P-TEFb, Kd 10 nM
cyclin T1 hot spots W207, W210, R165
no equivalent AFF1 structure or affinity exists anywhere
the SEC P-TEFb contact is entirely AFF4-derived evidence imported onto AFF1
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the characterised contact lies N-terminal to the AFF1 breakpoint and is not in the fusion
in-trans route by which a fusion lacking the N-terminal P-TEFb contact could still hold P-TEFb
confirms that entry's conclusion and corrects its residue number from ~100 to 347