RUNX1 — domain architecture and residue coordinates (with the RUNX1b/RUNX1c +27 offset)
also called RUNX1 · AML1 · AML1A · CBFA2 · PEBP2A2 · Q01196
RUNX1 is the master switch for making blood cells: one compact module grips DNA, and a partner protein clamps onto that same module to make the grip stronger. Two different official versions of the protein are in circulation and their residue numbers differ by twenty-seven, which is enough to put the DNA-gripping module in the wrong place on a diagram.
The same thing, for a scientist
RUNX1 (UniProt canonical Q01196-1, 453 aa, designated AML-1B by UniProt and matching RefSeq NP_001001890 'isoform AML1b') carries a single structured module, the Runt domain at residues 50-178, which binds both DNA and the CBFB subunit. C-terminally it carries interaction regions for KAT6A (291-371), KAT6B (307-400) and FOXP3 (362-402), and ends in the VWRPY motif (449-453) that recruits TLE/Groucho co-repressors. **Two numbering systems are in wide use**: RUNX1b (453 aa, the UniProt canonical) and RUNX1c (RefSeq NP_001745, 480 aa, which is also the Ensembl canonical transcript). We aligned the two sequences and confirmed they converge at RUNX1b residue 6 / RUNX1c residue 33, giving a constant **+27** offset, so the Runt domain is 50-178 in RUNX1b numbering and 77-205 in RUNX1c numbering. In ETV6::RUNX1 the RUNX1 contribution begins at its second coding exon and retains essentially the whole protein including the Runt domain.
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.
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.
The +27 offset, verified by alignment
This one is easy to get wrong because the two isoforms have different N-terminal sequences, not merely different lengths 1.
- RUNX1b = UniProt canonical Q01196-1 = RefSeq NP_001001890 = 453 aa, beginning
MRIPVDASTSRRF... - RUNX1c = RefSeq NP_001745 = 480 aa, beginning
MASDSIFESFPSYPQCFMRECILGMNPSRDVHDASTSRRF... - The shared sequence begins at
DASTSRRF, which is RUNX1b residue 6 and RUNX1c residue 33. The two proteins are then identical to the C-terminus. - Offset = +27 from RUNX1b residue 6 onwards. Runt domain 50-178 (b) = 77-205 (c). VWRPY 449-453 (b) = 476-480 (c).
A trap worth flagging: Ensembl's canonical RUNX1 transcript encodes the 480 aa RUNX1c, while UniProt's canonical is the 453 aa RUNX1b. Any pipeline that mixes an Ensembl-derived coordinate with a UniProt-derived domain boundary will be 27 residues out. This node uses RUNX1b/UniProt numbering throughout. Note also that the naming is genuinely inconsistent between databases: RefSeq labels the 250 aa short form NP_001116079 as 'isoform AML1a', while UniProt's isoform named 'AML-1A' (Q01196-2) is 472 aa.
One domain does two jobs
The Runt domain is simultaneously the DNA-binding domain and the CBFB-binding domain, which is why the CBFB::MYH11 fusion is able to sequester RUNX1 without touching RUNX1 itself. Alanine scanning of the Runt domain mapped the DNA-contacting residues and quantified each one's energetic contribution to DNA binding with and without CBFB 3, and a companion study identified two energetic hot spots on the Runt domain for CBFB heterodimerisation, mirroring hot spots found on the CBFB side 4. UniProt independently annotates DNA-contacting sub-regions at 80-84, 135-143 and 168-177 and specific DNA-binding residues at 112, 116, 139 and 170 1.
That is an unusually well-mapped interface for a transcription-factor complex, and it is the reason the Runt-CBFB interaction has been repeatedly proposed as druggable.
What ETV6::RUNX1 keeps, and the contrast with RUNX1::RUNX1T1
ETV6::RUNX1 joins ETV6's polymerising PNT domain to essentially the whole of RUNX1 2. All RUNX1 domains in the array below are therefore marked retained: the fusion is a full RUNX1 with a repression polymer bolted onto its N-terminus.
That contrasts sharply with the other common RUNX1 fusion. UniProt annotates a translocation site at 177/178 for RUNX1::RUNX1T1 (AML1-ETO) and several other partners 1 — a break that lands at the last residue of the Runt domain, keeping the DNA-binding module and discarding everything C-terminal to it including the KAT6A/KAT6B interaction regions and the VWRPY motif. Same gene, two breakpoints, two quite different molecules. The graph carries that contrast as a link to mech-breakpoint-position-consequences.
| uniprot | Q01196 |
|---|---|
| length AA | 453 |
| gene symbol | RUNX1 |
| isoform note | Coordinates are RUNX1b numbering = UniProt canonical Q01196-1 = 453 aa = RefSeq NP_001001890. The other common convention is RUNX1c (RefSeq NP_001745, 480 aa), which is ALSO the Ensembl canonical transcript. Verified by alignment: the two sequences differ at the N-terminus and converge at RUNX1b residue 6 / RUNX1c residue 33, so RUNX1c = RUNX1b + 27 from that point on. Runt domain: 50-178 (b) = 77-205 (c). Database isoform NAMES for this gene are inconsistent (UniProt 'AML-1A' is 472 aa; RefSeq 'isoform AML1a' is 250 aa), so match on length and accession, never on name. |
| numbering offset runx1c minus runx1b | 27 |
| assumed fusion | ETV6::RUNX1 (TEL-AML1), t(12;21). The RUNX1 moiety begins at its second coding exon and retains essentially the whole protein, so all domains below are retained. For the contrasting RUNX1::RUNX1T1 fusion the UniProt-annotated break at 177/178 would make everything C-terminal to the Runt domain lost. |
| alphafold ID | AF-Q01196-F1 |
| domain coordinate caveat | The Runt domain 50-178 is pinned by multiple crystal structures whose modelled chains run 50-183, so it is reliable to about +/-5 residues. The KAT6A, KAT6B and FOXP3 interaction regions are functional mapping intervals with no structure and overlapping boundaries; they should be drawn as regions, not folds, and are soft to tens of residues. The VWRPY motif is exact (sequence-verified). Every number is RUNX1b/UniProt numbering; add 27 for RUNX1c literature and for Ensembl-derived coordinates. |
PDB IDS
[ "1E50", "1H9D", "1LJM", "1CMO" ]
domains — the numbers behind the chart
[
{
"name": "RD (Runt domain, DNA + CBFB binding)",
"start": 50,
"end": 178,
"note": "= 77-205 in RUNX1c numbering. Crystal constructs span 50-183 (1E50, 1H9D). Binds DNA and the CBFB subunit through distinct surfaces.",
"fate": "retained"
},
{
"name": "KAT6A-interaction region",
"start": 291,
"end": 371,
"note": "functional region, no structure; boundaries are from interaction mapping, not from a fold.",
"fate": "retained"
},
{
"name": "transactivation domain (TAD)",
"start": 291,
"end": 371,
"aliases": [
"TAD",
"transactivation domain",
"KAT6A-interaction region"
],
"feature_type": "Region",
"boundary_precision": "soft",
"note": "UniProt Q01196 annotates 291-371 as the KAT6A-interaction region, which is the segment referred to as the transactivation domain in the fusion literature. Soft boundaries: this is a functional call over a partly disordered region, and it overlaps the NMTS, so the two should not be drawn as disjoint blocks.",
"source": "UniProt Q01196"
},
{
"name": "KAT6B-interaction region",
"start": 307,
"end": 400,
"note": "overlaps the KAT6A region; both are functional assignments.",
"fate": "retained"
},
{
"name": "nuclear matrix targeting signal (NMTS)",
"start": 351,
"end": 381,
"aliases": [
"NMTS",
"nuclear matrix targeting signal"
],
"feature_type": "Region",
"boundary_precision": "firm",
"note": "A 31-residue segment near the C terminus, necessary and sufficient to target the transcriptionally active AML-1B to the nuclear matrix, and distinct from nuclear localisation signals. Fusing it to GAL4(1-147) redirects GAL4 to the nuclear matrix. Its loss is a frequent consequence of t(8;21) and t(3;21). Added because three fusion entries reference it and the 3D viewer could not draw those lanes without coordinates.",
"source_pmid": "9192636"
},
{
"name": "FOXP3-interaction region",
"start": 362,
"end": 402,
"fate": "retained"
},
{
"name": "VWRPY (TLE/Groucho-binding motif)",
"start": 449,
"end": 453,
"note": "the five C-terminal residues; sequence verified directly as ...RLEEAVWRPY. = 476-480 in RUNX1c numbering. Lost in RUNX1::RUNX1T1, retained in ETV6::RUNX1.",
"fate": "retained"
}
]other annotated breakpoints
[
{
"site": "177/178",
"fusion": "RUNX1-RUNX1T1 (ETO), RUNX1-MECOM, RUNX1-CBFA2T3 and others"
},
{
"site": "241/242",
"fusion": "RUNX1-EAP, RUNX1-CBFA2T2"
}
]- RUNX1 is haploinsufficient in humans: germline heterozygous loss causes familial platelet disorder with predisposition to myeloid malignancy. Any agent that degrades RUNX1 alongside a RUNX1-containing fusion is walking into a known human loss-of-function phenotype.
- The +27 isoform offset is a live hazard rather than a theoretical one, because Ensembl and UniProt disagree about which isoform is canonical; a pipeline that joins the two sources without checking will silently mis-place the Runt domain by 27 residues.
- ETV6::RUNX1 retains the VWRPY/TLE-recruiting motif while RUNX1::RUNX1T1 discards it. Is that difference responsible for any of the large clinical divergence between the two diseases, or is it swamped by the difference in partner?
- The Runt-CBFB interface has two mapped energetic hot spots on each side and has been called druggable for twenty years without a drug. Is the failure chemical, or is it that inhibiting the interface phenocopies a tumour suppressor loss?
- Does the RUNX1c-specific N-terminal 32 residues do anything? If it does, then which isoform a fusion is built on is a biological question and not merely a numbering one.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProt canonical RUNX1 (Q01196-1, 453 aa, designated AML-1B) has a Runt domain at 50-178, DNA-contact sub-regions at 80-84, 135-143 and 168-177, KAT6A-interaction 291-371, KAT6B-interaction 307-400, FOXP3-interaction 362-402, and annotated translocation sites at 177/178 and 241/242.
UniProtKB entry Q01196 (RUNX1_HUMAN): reviewed sequence, length and positional feature annotations · UniProt Consortium · UniProtKB/Swiss-Prot 2026What this evidence cannot show
UniProt's own isoform naming for RUNX1 does not match RefSeq's: UniProt calls Q01196-2 'AML-1A' at 472 aa, whereas RefSeq's 'isoform AML1a' (NP_001116079) is 250 aa. Treat isoform *names* for this gene as unreliable and use lengths and accessions instead.
Evidence that argues against this (1)
- The Runt boundaries wobble, and more importantly the member databases annotate a C-terminal DOMAIN that UniProt does not have. The live InterPro call returns Pfam PF00853 52-179, InterPro IPR013524 50-179, IPR008967 52-178, SUPERFAMILY SSF49417 52-178 and CATH-Gene3D G3DSA:2.60.40.720 48-189 against UniProt/PROSITE's 50-178 - an 11-residue C-terminal spread. Separately, Pfam PF08504 / InterPro IPR013711 define a Runx inhibition domain / Runx C-terminal domain at 362-453, and CATH-Gene3D G3DSA:4.10.770.10 defines a folded central domain at 303-346, neither of which corresponds to any UniProt feature. The node's C-terminus is drawn as interaction regions where two independent methods see structured domains. InterPro Consortium; EMBL-EBI 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The t(12;21) fusion joins the N-terminal self-associating domain of TEL/ETV6 to the DNA-binding and transactivation domains of AML1/RUNX1, so the RUNX1 moiety is retained essentially in full.
Fusion of the TEL gene on 12p13 to the AML1 gene on 21q22 in acute lymphoblastic leukemia · Golub TR, Barker GF, Bohlander SK, et al. · PNAS 1995What this evidence cannot show
Two index cases; the retained RUNX1 extent is inferred from the cDNA junction rather than from a protein-level measurement of the fusion.
Evidence that argues against this (1)
- Retained in sequence, inverted in function, and the retained RUNX1 region is itself a repression module inside the fusion. Dissecting TEL-AML1 on the AML1-responsive IL-3 promoter shows it REPRESSES basal promoter activity in lymphoid cells, where AML1B and AML1A activate it, and that repression requires three distinct domains: the TEL pointed/HLH motif, the runt homology domain, and the 74 amino acids immediately downstream of the runt homology domain. Freshly isolated human TEL-AML1 ALL cells do not express IL-3. 'Retains the transactivation domains' is therefore misleading about what those residues do in the chimera. Uchida H et al. 1999
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Alanine-scanning mutagenesis of the Runx1 Runt domain quantified the energetic contribution of each DNA-contacting residue to DNA binding, in the presence and absence of CBFB.
Energetic contribution of residues in the Runx1 Runt domain to DNA binding · Li Z, Yan J, Matheny CJ, et al. · Journal of Biological Chemistry 2003What this evidence cannot show
In vitro binding energetics on an isolated Runt domain; residue numbering in the paper follows the 453 aa (RUNX1b) sequence used here, but readers should confirm this before importing individual residue numbers.
Evidence that argues against this (1)
- Per-residue energies assume additivity, and the Runt domain is not additive. Solving the RUNX1 Runt domain and its DNA site in their FREE states and comparing them with the DNA-bound and CBFbeta-bound structures showed that the Runt domain undergoes an ALLOSTERIC TRANSITION on DNA binding and that the free DNA target already adopts a bent-helical conformation compatible with the complex. When binding is coupled to a global conformational change plus DNA pre-bending, a single alanine substitution redistributes energy across the transition rather than subtracting one contact, so the tabulated per-residue values bound the total but cannot be read contact by contact. Bartfeld D et al. 2002
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Two energetic hot spots on the Runt domain mediate heterodimerisation with CBFB, mirroring hot spots previously mapped on the CBFB subunit.
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 2003What this evidence cannot show
Energetics measured on isolated domains; in vivo relevance was tested only by transient transfection reporter assays.
Evidence that argues against this (1)
- The same allosteric result gives a different reading of what CBFbeta is doing. The Runt domain's DNA-bound conformation is reached by an allosteric transition that is FURTHER STABILISED by CBFbeta, which is how CBFbeta enhances Runt-domain binding to DNA. On that account the two energetic hot spots are the sites where a conformational equilibrium is shifted rather than two independent interface patches, and hot-spot mutations report both interface loss and loss of allosteric stabilisation. The alanine-scanning work tested in-vivo relevance only by transient-transfection reporter assays, which cannot separate the two. Bartfeld D et al. 2002
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
RUNX1b (453 aa, RefSeq NP_001001890, UniProt canonical) and RUNX1c (480 aa, RefSeq NP_001745) differ in their N-terminal sequence and converge at RUNX1b residue 6 / RUNX1c residue 33, giving a constant +27 numbering offset thereafter; Ensembl's canonical RUNX1 transcript encodes the 480 aa form.
RefSeq protein records NP_001001890 (RUNX1 isoform AML1b, 453 aa), NP_001745 (isoform AML1c, 480 aa) and NP_001116079 (isoform AML1a, 250 aa), with pairwise sequence comparison · NCBI RefSeq; comparison computed by us · NCBI RefSeq / Ensembl 2026What this evidence cannot show
The offset was computed by us by locating the first shared 8-residue window and verifying that the C-terminal sequences are identical. It holds only from RUNX1b residue 6 onwards; residues 1-5 of RUNX1b have no counterpart in RUNX1c.
Evidence that argues against this (1)
- The Ensembl half is right but it does not pin a single transcript, and the two-isoform framing is an undercount. The live Ensembl record lists 13 protein-coding RUNX1 transcripts with exon counts of 3, 3, 4, 5, 5, 5, 6, 6, 7, 8, 9, 9 and 13, and THREE separate transcripts encode 480 aa - ENST00000675419 (the canonical, 9 exons), ENST00000300305 (8 exons) and ENST00001131757 (9 exons) - alongside 467, 453, 389, 364 and 256 aa forms. 'Ensembl canonical encodes 480 aa' is not sufficient to identify which transcript a source used, and the +27 rule holds only between the two named RefSeqs. Ensembl; EMBL-EBI 2026
- E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
An AlphaFold model AF-Q01196-F1 covers all 453 residues, and crystal structures of the Runt domain with CBFB and DNA exist (1E50, 1H9D, both titles verified at RCSB).
AlphaFold Protein Structure Database entry for Q01196 (model AF-Q01196-F1) · Jumper J, Varadi M, et al.; EMBL-EBI/DeepMind · AlphaFold DB 2026What this evidence cannot show
Every experimental structure covers only ~50-183; the entire C-terminal two-thirds of RUNX1 is predicted-disordered and has no experimental coordinates, so its depiction is schematic.
Evidence that argues against this (1)
- Neither crystal structure covers the range this implies. SIFTS maps both 1E50 and 1H9D chain A to Q01196 50-183, but the RCSB instance records show 1E50 chain A leaving entity residues 1-9 and 125-134 unobserved - UniProt 50-58 and 174-183 - so it models only 59-173, and 1H9D leaves 1-4 and 130-134 unobserved, modelling 57-178. The observed Runt domain is 115-122 residues, not 134, and residues 50-56 and 179-183 have never been seen in either entry. On the AlphaFold side, per-residue pLDDT from the v6 file gives 248 of 453 residues (55%) below 50 with a mean of 61.0. RCSB Protein Data Bank; PDBe / SIFTS et al. 2026
- E_nmtsstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Transcriptionally active AML-1B (RUNX1) binds the nuclear matrix via a 31-amino-acid nuclear matrix targeting signal near its C terminus, between residues 351 and 381, which is independent of DNA binding, distinct from nuclear localisation signals, and sufficient to redirect a heterologous GAL4(1-147) protein to the nuclear matrix. Loss of the RUNX1 C-terminal domain is a frequent consequence of t(8;21) and t(3;21).
Identification of a nuclear matrix targeting signal in the leukemia and bone-related AML/CBF-alpha transcription factors · Zeng C, van Wijnen AJ, Stein JL, Meyers S, Sun W, Shopland L, Lawrence JB, Penman S, Lian JB, Stein GS, Hiebert SW · PNAS 94:6746-51 1997What this evidence cannot show
Read as abstract only by the coordinator; the interactome beat reports reading PMC full text. The abstract gives 'between 351 and 381' and describes a 31-amino-acid segment, which is consistent. The nuclear matrix as a structure is itself contested elsewhere in this graph as a possible preparation artefact, so an NMTS should be read as a biochemical fractionation behaviour rather than as evidence for a physical scaffold.
Evidence that argues against this (1)
- This graph separately records that the nuclear matrix is contested as a preparation artefact. A signal that targets a protein to an operationally defined fraction is a real and reproducible behaviour, but naming it after a structure whose existence is disputed imports an assumption the experiment does not test. Zeng C et al. 1997
Counter-evidence added cycle 41: 'retains the DNA-binding AND transactivation domains' is misleading about function. TEL-AML1 represses the AML1-responsive IL-3 promoter in lymphoid cells where AML1B activates it, and repression requires the TEL pointed motif, the runt homology domain AND the 74 residues immediately downstream of it (Uchida 1999, PMID 10023677).
Counter-evidence on per-residue energetics: the Runt domain undergoes an allosteric transition on DNA binding and its DNA target is pre-bent, so an alanine substitution redistributes energy across the transition rather than subtracting one contact (Bartfeld 2002, PMID 12377125). The same result reframes the CBFbeta 'hot spots' as sites where a conformational equilibrium is shifted.
ETV6::RUNX1, t(12;21)
3-prime partner domain architecture; the Runt DNA-binding domain is retained
5-prime partner in RUNX1::RUNX1T1, t(8;21); same Runt domain, opposite orientation
the Runt domain (50-178) is clamped by CBFB; co-crystals 1E50 and 1H9D
two RUNX1 breakpoints (whole protein vs 1-177) produce two very different oncoproteins from one gene
t(12;21)(p13;q22)
discontiguous surface epitope on the CBFB beta-barrel binds the Runt domain; two energetic hot spots
- ← related toETV6::RUNX1 (TEL-AML1), t(12;21)(p13;q22)
Counter-evidence: Wolf-Watz 1999 (PMID 10103057) shows the Runt-domain constructs behind the clamp model are Cys-to-Ser mutants whose N-terminal truncation itself enhances DNA binding
3' partner domain coordinates