CBFB — domain architecture and residue coordinates

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.

• CBFB is an obligate partner of all three RUNX proteins and is required for definitive haematopoiesis; disrupting the Runt-CBFB interface to attack the fusion would also disable normal core-binding factor, which is embryonic-lethal on knockout in mice.
• Because the fusion keeps the entire CBFB fold intact, there is no fusion-specific surface on the CBFB side at all. Selectivity would have to come from the junction or from the MYH11 rod, not from CBFB.

1. Only seventeen disordered C-terminal residues are removed by inv(16), yet the fusion behaves completely differently from wild-type CBFB. Is any of that difference attributable to losing 166-182, or is it entirely due to what is added?
2. Where does the CBFB fold really end? Sources differ between 141, 142 and 148 and no construct has been made specifically to test the boundary. The question matters because the boundary is only ~20 residues from the breakpoint.
3. The Runt-CBFB interface has two mapped hot spots on each side and has been nominated as druggable since 2000. Has any campaign actually failed, or has none been run at scale? The negative result would be as informative as a hit.
4. Does the 142-157 stretch order against RUNX1 as it does against lentiviral Vif? Every structure in which it is ordered is a Vif-ELOB-ELOC complex, and no CBFB-Runt structure has a construct long enough to answer it. If it does, the CBFB::MYH11 fusion retains a structured element this graph currently draws as disordered tail, immediately N-terminal to the junction.

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

   UniProt Q13951 is 182 residues, annotates no domain, region, motif, coiled coil, zinc finger or DNA-binding feature, has secondary-structure elements spanning residues 8-148, and annotates a Site at 165-166 as the breakpoint forming the CBF-beta-MYH11 oncogene in AML subtype M4Eo.

   UniProtKB entry Q13951 (PEBB_HUMAN): reviewed sequence, length and positional feature annotations · UniProt Consortium · UniProtKB/Swiss-Prot 2026
   ▸What this evidence cannot show

   Absence of domain annotation is informative here but is still an absence; UniProt's secondary-structure features are imported from deposited structures and therefore inherit those constructs' limits rather than proving where the fold ends.

   ▸Evidence that argues against this (2)

   • The fold boundary at 141 is soft in the direction that matters. Deposited CBFB chains are ordered to residue 157 (8CX0) and 160 (6P59), sixteen to nineteen residues past the stored end, and Pfam PF02312 models the family domain to 166. Since the CBFB::MYH11 break is at 165, this changes the picture of the fusion: the CBFB moiety is close to the whole ordered protein rather than a fold plus a 24-residue disordered tail. The 'C-terminal tail (disordered)' bar at 142-182 is only reliably disordered from about 158. RCSB Protein Data Bank (entry titles and instance observed-residue records retrieved and read back) 2026
   • This node says the 4-141 boundary was chosen 'because two independent sources give it identically'. On re-checking InterPro, only SUPERFAMILY SSF50723 gives 4-141; CATH-Gene3D gives 1-142, Pfam 1-166 and PANTHER 1-181. The PDB 1CL3 modelled chain also maps to 4-141, but that is the construct the SUPERFAMILY model was built against rather than an independent determination, so the corroboration is one source and one structure, not two independent sources. InterPro Consortium; EMBL-EBI 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Domain databases place the CBFB fold at 4-141 (SUPERFAMILY SSF50723) and 1-142 (CATH-Gene3D G3DSA:2.40.250.10), with Pfam PF02312 giving a wider family match of 1-166.

   InterPro/Pfam/CATH-Gene3D/SUPERFAMILY domain matches for UniProt Q13951 · InterPro Consortium; EMBL-EBI · InterPro 2026
   ▸What this evidence cannot show

   These are model-match extents. The Pfam family match at 1-166 is a family boundary, not a fold boundary, and should not be used as a domain coordinate.

   ▸Evidence that argues against this (1)

   • The CATH number is transferred from a mouse structure whose own ordered extent stops at 140 and has a hole in the middle. The fetched InterPro record names the CATH model generating the human 1-142 assignment as 3wttB00; SIFTS for 3WTT maps chain B to Q08024 (PEBB_MOUSE), not human Q13951, with a construct of 1-142 - and PDBe polymer coverage shows that chain has coordinates only for author residues 2-71 and 80-140, an 8-residue unmodelled gap at 72-79 and nothing at residue 1 or 141-142. So of the two fold numbers, one is a cross-species transfer whose source structure does not cover its own stated range. PDBe / SIFTS et al. 2026
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   The solution NMR structure of CBFB (PDB 1CL3, 25 models) models residues 4-141, and later Runt/CBFB/DNA ternary complexes (1E50, 1H9D) model CBFB chains over residues 2-135.

   PDB 1CL3 'MOLECULAR INSIGHTS INTO PEBP2/CBF-SMMHC ASSOCIATED ACUTE LEUKEMIA REVEALED FROM THE THREE-DIMENSIONAL STRUCTURE OF PEBP2/CBF BETA', with 1E50 'AML1/CBFbeta complex' and 1H9D 'Aml1/cbf-beta/dna complex' (all three titles retrieved and verified at RCSB) · RCSB Protein Data Bank (entry metadata retrieved and title verified) · RCSB PDB 2026
   ▸What this evidence cannot show

   Modelled chain extents are limited by construct design and by disorder in the crystal; a chain ending at residue 141 does not prove the fold ends there, only that nothing beyond it was ordered. CORRECTION 2026-09-19: this item previously called PDB 1CL3 a crystal structure. RCSB gives exptl.method = SOLUTION NMR with 25 deposited models and a null resolution field.

   ▸Evidence that argues against this (1)

   • 1CL3 is not a crystal structure. The fetched RCSB core-entry record gives exptl.method = SOLUTION NMR with 25 deposited models and zero nonpolymer entities, and a null resolution field, so 'models residues 4-141' describes an NMR ensemble. The two ternary complexes are also not continuous over 2-135: PDBe coverage gives 1E50 chain B author residues 2-73 plus 78-135 (a four-residue gap) and 1H9D chain B 2-70 plus 80-135 (a nine-residue gap), so the CBFB chain in the DNA-bound complexes is broken in the same loop in both. PDBe / SIFTS et al. 2026
4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Alanine scanning of CBFB identified the amino acids that mediate binding to the Runt domain and quantified each one's energetic contribution to heterodimerisation, showing the interface is a discontiguous surface epitope.

   Energetic and functional contribution of residues in the core binding factor beta (CBFbeta) subunit to heterodimerization with CBFalpha · Tang YY, Shi J, Zhang L, et al. · Journal of Biological Chemistry 2000
   PMID 10984496DOI
   ▸What this evidence cannot show

   In vitro binding energetics on isolated domains, with in vivo relevance assessed in cell-based assays; the study predates modern deep mutational scanning and covers only residues chosen from the structures.

   ▸Evidence that argues against this (1)

   • You can switch this interaction off from a patch of CBFbeta that the alanine scan never implicated. Low-micromolar small molecules that block RUNX1 binding to CBFbeta were docked by NMR at a site displaced from the RUNX1 binding interface - allosteric inhibitors of the protein-protein interaction - and decreased proliferation of the inv(16) human leukaemia line ME-1. The per-residue ddG map therefore defines the contact epitope but not the full set of CBFbeta residues that govern heterodimerisation. This came from the same laboratory as the alanine scan, so it is not a cross-lab reagent artefact. Gorczynski MJ et al. 2007
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The reciprocal analysis on the Runt domain found two energetic hot spots for CBFB heterodimerisation, matching hot spots identified on the CBFB side.

   Mutagenesis of the Runt domain defines two energetic hot spots for heterodimerization with the core binding factor beta subunit · Zhang L, Li Z, Yan J, et al. · Journal of Biological Chemistry 2003
   PMID 12807883DOI
   ▸What this evidence cannot show

   Same methodological limits as the CBFB-side study; hot-spot definitions depend on the mutagenesis panel chosen.

   ▸Evidence that argues against this (1)

   • The biggest energetic hot spot turns out not to be a functional requirement. The same alanine-scanning analysis is cited for T161's side chain contributing more energy (2.0 kcal/mol) than any other residue to heterodimerisation with CBFbeta; T161A then reduces CBFbeta binding 40-fold and T149A 14-fold (L148F, 100-fold), yet T161A/T161A and T149A/T149A homozygous mice were born at the expected Mendelian frequencies and had haematopoietic progenitors, making them hypomorphic rather than null. The energetic surface is also broader than two spots: T149, L148, Q158, M124, S140, K167 and T169 all cut CBFbeta binding. Matheny CJ et al. 2007
6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   In six of six inv(16) patient samples, an in-frame fusion mRNA connected the first 165 amino acids of CBFB to the tail region of smooth-muscle myosin heavy chain, and the repeated coiled coil of SMMHC was proposed to dimerise the fusion protein.

   Fusion between transcription factor CBF beta/PEBP2 beta and a myosin heavy chain in acute myeloid leukemia · Liu P, Tarle SA, Hajra A, et al. · Science 1993
   PMID 8351518DOI
   ▸What this evidence cannot show

   Six patients in the original cloning report; the dimerisation mechanism was a proposal in 1993, later supported by the assembly-competence-domain work.

   ▸Evidence that argues against this (1)

   • The coiled coil on its own does not do the job; a 28-residue assembly-competence domain inside it does. A CBFbeta-SMMHC variant lacking the 28 amino acids homologous to the skeletal-myosin assembly competence domain multimerised less effectively than wild type or a control variant lacking a different 28-residue segment, and unlike those controls it failed to inhibit CBF DNA binding, AML1-mediated reporter activation, or G1-to-S progression. A truncation retaining 149 C-terminal residues multimerised and slowed Ba/F3 proliferation, whereas CBFbeta-SMMHC(83C) - still coiled-coil sequence but only 20 aa N-terminal to the ACD - did neither, and both the deltaACD and 83C proteins were predominantly cytoplasmic. The operative element is a short defined segment plus nuclear retention, not generic heptad dimerisation. Kummalue T et al. 2002
7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   Across 36 of 37 inv(16)/t(16;16) leukaemias, four different chimeric transcripts were found, all splicing varying MYH11 positions to CBFB nucleotide 495, with all but 17 of the 182 CBFB codons retained.

   Detection of fusion transcripts generated by the inversion 16 chromosome in acute myelogenous leukemia · Claxton DF, Liu P, Hsu HB, et al. · Blood 1994
   PMID 8142642
   ▸What this evidence cannot show

   RT-PCR on 37 cases; the paper has no DOI indexed in PubMed and is cited by PMID. 182 minus 17 equals 165, which is the arithmetic that ties this to the UniProt site and to Liu 1993.

   ▸Evidence that argues against this (1)

   • A much larger, later series found more than twice as many transcript types. Characterising CBFB-MYH11 in 162 patients at diagnosis gave 128 (79.0%) type A, with nine different rare fusion genes accounting for the other 34 cases (21.0%) - ten types in one cohort - and the authors state that more than 10 different fusion transcripts have been reported. The rare types were enriched in therapy-related AML (P=0.0106), had lower median white cell count (7.8 against 35.4 G/l, P<0.0001) and atypical cytomorphology (P=0.0203), so the four-transcript picture also under-represents a clinically distinguishable subgroup. The CBFB side is not invariant either: a K-type variant carries novel breakpoints within CBFB exon 6 with a partial insertion of that exon rather than a clean junction at CBFB nucleotide 495. Schnittger S et al. 2007
8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   CBFB is ordered well past the stored fold boundary of 141 in deposited structures: in the 2.7 A cryo-EM complex 8CX0 the CBFB chain is observed for residues 5-157 (unobserved 1-4 and 158-182), and in the 2.94 A crystal 6P59 the CBFB chain is observed for residues 6-160 of a 1-165 construct.

   PDB 8CX0 'Cryo-EM structure of human APOBEC3G/HIV-1 Vif/CBFbeta/ELOB/ELOC monomeric complex' and 6P59 'Crystal structure of SIVrcm Vif-CBFbeta-ELOB-ELOC complex': observed-residue ranges for the CBFB chains · RCSB Protein Data Bank (entry titles and instance observed-residue records retrieved and read back) · RCSB PDB 2026
   ▸What this evidence cannot show

   Read from the RCSB instance records' UNOBSERVED_RESIDUE_XYZ features on 2026-09-19 (database records; no full text read). Both are CBFB bound to a lentiviral Vif-ELOB-ELOC assembly, not to RUNX1, so the ordering of residues 142-157 could in principle be induced by that partner rather than intrinsic; the 1CL3 solution structure on which the stored 4-141 is based used a construct that did not extend far enough to test it. 8CX0 confirms that the last 25 residues (158-182) are genuinely disordered.

   ▸Evidence that argues against this (1)

   • The two CBFB copies in that same crystal do not agree. PDBe polymer coverage for 6P59 gives chain A author residues 5-161 but chain C 6-160 - this item quotes the chain C number as if it were the entry's answer, and the asymmetric unit itself disagrees by one residue at each end. That is a one-residue effect, but it is the whole precision budget of a claim whose point is where order stops. The honest statement is that CBFB is ordered to somewhere in 157-161 depending on copy and on which mirror is read, not to a specific residue. The ordering past 141 stands; the endpoint does not. PDBe / SIFTS et al. 2026
9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   The profile databases do not place a boundary at 141 either: Pfam PF02312 models the core-binding factor beta subunit family as 1-166, PANTHER PTHR10276 and InterPro IPR003417 as 1-181, CATH-Gene3D G3DSA:2.40.250.10 and InterPro IPR036552 as 1-142, and only SUPERFAMILY SSF50723 gives the 4-141 this node stores.

   InterPro member-database domain matches for UniProt Q13951 (Pfam, PANTHER, CATH-Gene3D, SUPERFAMILY) · InterPro Consortium; EMBL-EBI · InterPro 2026
   ▸What this evidence cannot show

   Profile-model match extents, not measurements, and the PANTHER/Pfam family models cover whole proteins rather than folds, so 1-166 and 1-181 are family extents and not claims about a beta-barrel boundary. Retrieved 2026-09-19 (database record). The relevant observation is that of six sources only one gives 4-141, which is weaker corroboration than this node originally claimed.

   ▸Evidence that argues against this (1)

   • It is weaker still, because SSF50723's 4-141 is not an independent vote - it reproduces the construct of the structure it was built from. SIFTS maps PDB 1CL3 chain A to Q13951 residues 4-141 exactly, and SUPERFAMILY SSF50723 returns 4-141 exactly, so the independent database and the experimental structure in this node are the same measurement counted twice. Fetched InterPro gives only six matches for Q13951 in total, of which two (PF02312 1-166, PTHR10276 / IPR003417 1-181) are family models covering essentially the whole 182-residue protein and two more (G3DSA:2.40.250.10 and IPR036552, both 1-142) come from the mouse-derived CATH model 3wttB00. Stated properly there are not six independent sources but roughly two. PDBe / SIFTS et al. 2026

• refutes →CBFB::MYH11 (CBFbeta-SMMHC), inv(16)(p13;q22)

  Counter-evidence added cycle 41: the four-transcript picture is an undercount. 162 patients gave ten CBFB-MYH11 transcript types, 79.0% type A and 21.0% across nine rare types enriched for therapy-related AML with atypical cytomorphology and low white cell count (Schnittger 2007, PMID 17287858). Also, PDB 1CL3 - called a crystal structure in this entry until 2026-09-19 - is a 25-model solution NMR ensemble.

• ← related toMYH11 — domain architecture and residue coordinates

  Counter-evidence on both sides of the same junction: the MYH11 IQ motif's C-terminal boundary is 807 (InterPro/SMART) or 815 (UniProt), and the CBFB fold is ordered to residue ~157 in deposited structures rather than ending at the stored 141. The junction at CBFB 165 is therefore closer to ordered structure on the CBFB side, and less certainly placed on the MYH11 side, than either entry drew.

• related to →CBFB::MYH11 (CBFbeta-SMMHC), inv(16)(p13;q22)

  Coordinate correction (PDB 8CX0 and 6P59 observed residues): CBFB is ordered to residue 157-160 in deposited complexes and Pfam PF02312 models the family domain to 166, against the stored fold end of 141. The CBFB moiety retained by the fusion (1-165) is therefore essentially the entire ordered protein, not a fold plus a long disordered tail.

• ← bindsRUNX1 — domain architecture and residue coordinates (with the RUNX1b/RUNX1c +27 offset)

  the Runt domain (50-178) is clamped by CBFB; co-crystals 1E50 and 1H9D

• part of →CBFB::MYH11 (CBFbeta-SMMHC), inv(16)(p13;q22)

  5-prime partner domain architecture; the Runt-binding core is retained, the last 17 residues replaced

• fuses with →MYH11 — domain architecture and residue coordinates

  CBFB::MYH11, inv(16)

• binds →RUNX1 — domain architecture and residue coordinates (with the RUNX1b/RUNX1c +27 offset)

  discontiguous surface epitope on the CBFB beta-barrel binds the Runt domain; two energetic hot spots

• supports →Mechanism class 2 — transcription-factor corruption (RUNX1::RUNX1T1, PML::RARA, ETV6::RUNX1, CBFB::MYH11) — a balance, not a pure dominant negative

  the fusion retains the whole Runt-binding fold and so sequesters RUNX1 away from its targets

• supports →Hypothesis — the menin lesson generalises: every non-enzymatic fusion has an obligate partner interface that is more druggable than the fusion

  CBFB is an obligate partner of RUNX1 and the fusion hijacks that obligate interface unchanged

• ← fuses withMYH11 — domain architecture and residue coordinates

  inv(16)(p13q22) / t(16;16)(p13;q22)

• analogous to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)

  both fusions preserve a wild-type protein-protein interface intact, leaving no fusion-specific epitope on that side