TCF3 (E2A) — domain architecture and residue coordinates
also called TCF3 · E2A · E12 · E47 · TCF-3 · ITF1 · P15923
TCF3 is a switch-on protein for developing B cells: one end shouts at the cell's transcription machinery, the other end grips DNA. In one childhood leukaemia the DNA-gripping end is cut off and replaced by a different protein's DNA-gripping end, so the shouting is aimed at entirely the wrong genes.
The same thing, for a scientist
TCF3 (UniProt P15923, 654 aa) is an E-protein transcription factor: a 9-amino-acid transactivation domain at 19-27 inside the broader AD1, an NLS at 170-176, a leucine-zipper region at 389-425 and a basic helix-loop-helix DNA-binding domain at 549-602. UniProt annotates the TCF3::PBX1 translocation site between residues **483 and 484**; we independently recomputed the Ensembl TCF3-201 coding exons and found exon 16 ends at codon 483, an exact match. The fusion therefore keeps both activation domains and the leucine zipper and loses the bHLH. **A correction to a common claim: the UniProt canonical P15923-1 is isoform E12, not E47** — UniProt's own alternative-products list marks E12 as displayed, and we confirmed the canonical bHLH sequence reads ...ANNARERLRVRDINEAFKELGRMCQLHL..., the E12 form, while P15923-2 (E47, 651 aa) reads ...ANNARERVRVRDINEAFRELGRMCQMHL...
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 isoform correction: canonical is E12, not E47
This one was checked directly rather than assumed 1.
- UniProt's alternative-products list for P15923 names three isoforms and marks E12 (P15923-1, 654 aa) as the displayed/canonical sequence; E47 is P15923-2 at 651 aa.
- We downloaded both sequences and aligned them. They are identical up to residue 529, diverge over the alternatively spliced bHLH-encoding exon (E12 residues 530-601, E47 residues 530-598), then share an identical 53-residue C-terminus.
- The canonical bHLH reads
...REKERRVANNARERLRVRDINEAFKELGRMCQLHLNSEKPQTKLLILHQAVSVILNLEQQ...; the E47 isoform reads...KDLRDRERRMANNARERVRVRDINEAFRELGRMCQMHLKSDKAQTKLLILQQAVQVILGLEQQ....
So the frequently repeated statement that P15923 canonical is E47 is wrong. Practically the error is small for this fusion — the divergent region is entirely inside the bHLH, which TCF3::PBX1 discards, and everything upstream of residue 529 is shared — but it is not small for anyone drawing the bHLH. Within the variable exon the two isoforms cannot be inter-converted by a constant offset; downstream of it, E47 = E12 minus 3.
The break at 483, verified twice1 deeper
UniProt annotates the TCF3-PBX1 breakpoint at 483/484 1. Independently, mapping the Ensembl TCF3-201 coding exon boundaries onto the coding sequence puts the end of coding exon 16 at codon 483 5. Two methods, one number.
Nourse and colleagues, who cloned the t(1;19) fusion, described it as an 85 kDa protein consisting of 'the amino-terminal two-thirds of E2A fused to a chromosome 1-derived protein' in which 'the DNA binding domain of E2A is replaced by the putative DNA binding domain of a homeoprotein' 4. 483 of 654 is 74% — the amino-terminal two-thirds, near enough.
So the fusion retains: AD1 with its 9aaTAD, the NLS, AD2 (unannotated, see below) and the leucine-zipper region at 389-425. It loses the bHLH at 549-602 and the last 171 residues.
AD2 is real but unannotated, so it is not in the array
The E-protein literature consistently describes a second activation domain, AD2, in the central third of TCF3, and it is retained in the fusion. UniProt annotates no AD2 feature and neither does InterPro, and we did not find a residue range we could verify in this pass. It is therefore absent from data.domains rather than given invented coordinates. The same applies to the outer extent of AD1: UniProt annotates only the 9-residue 9aaTAD at 19-27, whereas the p300 TAZ2-bound NMR construct in PDB 2MH0 (title verified as 'Solution NMR structure of the p300 Taz2:ETAD1 complex') spans TCF3 residues 1-37, and the literature commonly draws AD1 as roughly residues 1-99.
Where the sources disagree about the bHLH
A five-way spread, all recorded:
| Source | bHLH coordinates |
|---|---|
| UniProt / PROSITE PS50888 | 549-602 1 |
| Pfam PF00010 | 550-603 2 |
| SMART SM00353 | 555-608 2 |
| InterPro IPR011598 | 549-608 2 |
| CATH-Gene3D + SUPERFAMILY | 536-616 / 540-608 2 |
| CDD cd18945 | 543-627 2 |
| PDB 3U5V / 6MGN chains | 563-613 / 561-612 3 |
The array uses UniProt's 549-602 and carries the spread in the note: the honest statement is that the bHLH is somewhere in 536-627 with a well-agreed core around 555-602. Since TCF3::PBX1 loses the whole thing, the uncertainty does not affect the fusion's fate assignments — but it would badly affect any figure of wild-type TCF3.
Worth flagging separately: PDB 6MGN is titled 'mouse Id1 (51-104) - human hE47 (348-399) complex', yet UniProt maps its human chain to P15923 residues 561-612. The paper's own E47 numbering and the reference-sequence numbering differ by 213 residues. That is a construct-numbering convention, not an error, but it is precisely the sort of mismatch that corrupts a residue-level viewer.
| uniprot | P15923 |
|---|---|
| length AA | 654 |
| gene symbol | TCF3 |
| isoform note | CORRECTION: UniProt canonical P15923-1 (654 aa) is isoform **E12**, not E47. E47 is P15923-2 at 651 aa; isoform 3 is 658 aa. Verified by downloading both sequences: identical to residue 529, divergent over the alternatively spliced bHLH exon (E12 530-601 vs E47 530-598), then an identical 53-residue C-terminus. Downstream of the variable exon E47 = E12 - 3; INSIDE the variable exon there is no constant offset and the sequences differ, so bHLH coordinates cannot be converted arithmetically between isoforms. All coordinates here are canonical E12 numbering. For TCF3::PBX1 the distinction is largely moot because the break at 483 lies well upstream of residue 529, so the fusion's TCF3 moiety is identical in both isoforms. |
| assumed fusion | TCF3::PBX1, t(1;19). TCF3 contributes residues 1-483, verified by UniProt Site 483/484 and by independent Ensembl exon arithmetic. |
| alphafold ID | AF-P15923-F1 |
| domain coordinate caveat | AD1 is recorded only as the 9-residue 9aaTAD core because that is all UniProt annotates; the real activation domain is several times wider (PDB 2MH0 construct 1-37; literature ~1-99). AD2, which is retained in the fusion and is discussed throughout the E-protein literature, is DELIBERATELY ABSENT because neither UniProt nor InterPro annotates it and no residue range could be verified in this pass. The bHLH boundary is the widest disagreement in this node (536-627 across six sources) and the array uses UniProt's 549-602. The NLS and leucine-zipper are sequence-derived and soft. The breakpoint at 483 is verified twice. |
| fusion boundary residue | 483 |
| trap reverification | RE-VERIFIED 2026-09-19 and BOTH TRAPS HOLD. (1) UniProt canonical P15923-1 is isoform E12 at 654 aa; E47 is P15923-2 at 651 aa. (2) SIFTS maps PDB 6MGN's human chain to P15923 561-612 while the entry title numbers it 348-399; the difference is exactly 213 at both ends. Additional caution added: the 213-residue offset is 6MGN's construct convention and is unrelated to the 3-residue E12/E47 isoform offset. They must not be combined. |
PDB IDS
[ "2MH0", "3U5V", "6MGN", "2YPA" ]
domains — the numbers behind the chart
[
{
"name": "AD1 (9aaTAD, p300/CBP-binding)",
"start": 19,
"end": 27,
"note": "the UniProt 9aaTAD motif. The p300 TAZ2-bound NMR construct (PDB 2MH0) spans residues 1-37 and the literature commonly draws AD1 as ~1-99, so this 9-residue bar is the core, not the domain.",
"fate": "retained"
},
{
"name": "AD2 (activation domain 2)",
"start": null,
"end": null,
"note": "Named in the primary literature and named by two fusion entries here, but NO residue range is stated in the accessible text of the paper that characterises it. PMID 9742120 (Inukai T, Inaba T, Ikushima S et al., Mol Cell Biol 1998, 10.1128/MCB.18.10.6035) is titled for AD1 and AD2 and shows the mutant panel as a DIAGRAM in Fig. 4C; its full text, read via the PMC HTML route (PMC109189, 55,421 characters), states exactly one residue range in the whole paper -- 170 to 175, which is the NLS already in this atlas. Coordinates deliberately left null rather than estimated: this project stored the PML coiled coil 106 residues short and the KMT2A bromodomain as 1703-1809 when it is 1635-1765, both from estimating.",
"fate": "retained",
"boundary_precision": "none - domain named, coordinates not established from an accessible source"
},
{
"name": "NLS (nuclear localisation signal)",
"start": 170,
"end": 176,
"fate": "retained"
},
{
"name": "LZ (leucine-zipper region)",
"start": 389,
"end": 425,
"fate": "retained"
},
{
"name": "bHLH (basic helix-loop-helix)",
"start": 549,
"end": 602,
"note": "the DNA-binding domain, entirely lost in TCF3::PBX1 (break at 483). Boundary estimates across sources span 536-627; the well-agreed core is ~555-602. This is also the region where E12 and E47 differ.",
"fate": "lost"
}
]counter evidence pass
{
"queen": "Q2b-atlas2",
"cycle": 31,
"date": "2026-09-19",
"checked": [
"UniProt P15923 live record (primary gene name TCF3 confirmed against the TCF7L1 synonym trap; 654 aa; every stored interval and the 483/484 site match exactly)",
"InterPro entry/all for P15923 - all seven bHLH boundary values reproduced unchanged",
"SIFTS mappings for 6MGN (561-612), 3U5V (563-584 + 586-613) and 2MH0 (1-37)"
]
}ad2 bounded search 2026 09 22
{
"why": "check-joins.mjs reports two genuine 3D-viewer gaps and both are this domain, named by fusion-tcf3-hlf and fusion-tcf3-pbx1 and absent from this atlas.",
"queries_run": {
"(E2A[tiab] OR TCF3[tiab]) AND \"activation domain 2\"[tiab]": 0,
"(E2A[tiab] OR TCF3[tiab]) AND AD2[tiab] AND (transactivation[tiab] OR activation[tiab])": 11,
"E2A[tiab] AND \"AD2\"[tiab]": 42
},
"best_source_found": "PMID 9742120 -- 'The AD1 and AD2 transactivation domains of E2A are essential for the antiapoptotic activity of the chimeric oncoprotein E2A-HLF'. Directly on topic: E2A-HLF is one of the two fusions carrying the gap.",
"routes_tried": {
"Europe PMC fullTextXML for PMC109189": "HTTP 500",
"PMC HTML page for PMC109189": "HTTP 200, 55,421 characters of text extracted"
},
"result": "The full text names AD1 and AD2 repeatedly and describes deletion and point mutants of both, but states only ONE residue range anywhere in the paper (170-175, the NLS). AD2's boundaries exist only in the Fig. 4C schematic, which is an image.",
"what_would_close_it": "Read Fig. 4C of PMID 9742120 directly, or find the construct boundaries in a methods section or an addgene/plasmid record for the E2A-HLF mutant panel. Failing that, an AD2 range from a paper that states it numerically -- the 11-record query above is small enough to read in full and was NOT read past the top two abstracts.",
"do_not": "Do not estimate. The two worst coordinate errors in this project's history came from estimating a domain boundary that looked obvious."
}- TCF3 is essential for B-lymphopoiesis and is itself a tumour suppressor in some contexts; degrading TCF3 to remove the fusion would remove normal E-protein function from the same lineage the therapy is trying to rescue.
- AD1 works by binding the p300/CBP TAZ2 domain, a hub used by dozens of transcription factors. Blocking that interface is unlikely to be selective for the fusion, and p300 inhibition has broad transcriptional consequences.
- Anyone drawing the bHLH from a database without checking which isoform they are on is combining an E12 coordinate with an E47 sequence; within the variable exon there is no arithmetic that fixes it.
- Where does AD2 actually begin and end in residue terms? It is retained in every TCF3 fusion, is invoked constantly in the literature, and has no annotation in UniProt or InterPro. Pinning it would materially improve this node and the TCF3::HLF node alongside it.
- TCF3::PBX1 and TCF3::HLF share the same TCF3 1-483 moiety but have dramatically different clinical behaviour. Since the TCF3 contribution is identical, everything that differs must come from the partner — is that actually true at the protein level, or do the two fusions differ in TCF3 breakpoint in some cases?
- Does the retained leucine-zipper region at 389-425 still mediate any protein-protein interaction in the fusion, now that the bHLH it normally works with is gone?
- How many other entries in this graph carry a PDB coordinate taken from an entry title rather than from the SIFTS cross-reference? 6MGN is off by 213 and was caught. The check that catches it is mechanical and could be run over every pdb_ids list in the graph.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
TCF3 canonical P15923-1 is 654 residues and is isoform E12 (not E47, which is P15923-2 at 651 aa); it carries a 9aaTAD at 19-27, an NLS at 170-176, a leucine-zipper region at 389-425, a bHLH at 549-602, and an annotated TCF3-PBX1 translocation site at 483/484.
UniProtKB entry P15923 (TFE2_HUMAN): reviewed sequence, length and positional feature annotations · UniProt Consortium · UniProtKB/Swiss-Prot 2026What this evidence cannot show
The E12/E47 assignment was confirmed by downloading both isoform sequences and reading the bHLH region directly, because the canonical-isoform identity of P15923 is frequently misreported. Isoform 3 (P15923-3) is 658 aa.
Evidence that argues against this (1)
- Three of those four features have no support from any domain database, and the fourth is contested. The live InterPro call for P15923 returns only TEN entries and every one is either the bHLH or a full-length family assignment (PANTHER PTHR11793 5-654, InterPro IPR051098 5-654); nothing whatever is annotated at 19-27, 170-176 or 389-425, so the leucine-zipper region at 389-425 in particular rests on UniProt curation alone. Where they do overlap, the bHLH spreads from CDD cd18945 543-627 and CATH-Gene3D G3DSA:4.10.280.10 536-616 through SUPERFAMILY SSF47459 540-608, InterPro IPR011598 549-608 and SMART SM00353 555-608 to Pfam PF00010 550-603 - a 19-residue start spread and a 24-residue end spread around the stored 549-602. InterPro Consortium; EMBL-EBI 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
Independent domain databases place the TCF3 bHLH at 550-603 (Pfam), 555-608 (SMART), 549-608 (InterPro), 536-616 (CATH-Gene3D), 540-608 (SUPERFAMILY) and 543-627 (CDD).
InterPro/Pfam/CATH-Gene3D/SUPERFAMILY domain matches for UniProt P15923 · InterPro Consortium; EMBL-EBI · InterPro 2026What this evidence cannot show
Model-match extents, not measurements; the ~90-residue total spread reflects the different models' treatment of the flanking basic region and the C-terminal helix.
Evidence that argues against this (1)
- The five-way bHLH spread in this node was re-fetched from InterPro on 2026-09-19 and every value reproduced exactly: PROSITE PS50888 549-602, Pfam PF00010 550-603, SMART SM00353 555-608, InterPro IPR011598 549-608, CATH-Gene3D 536-616, SUPERFAMILY 540-608, CDD cd18945 543-627. The spread stands at 536-627 and no source has moved. This is the one node in the beat whose recorded disagreement needed no correction. InterPro Consortium; EMBL-EBI 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Experimental structures of the TCF3 activation and bHLH regions exist: PDB 2MH0 covers residues 1-37 bound to the p300 TAZ2 domain, and 3U5V and 6MGN cover the bHLH at 563-613 and 561-612.
PDB 2MH0 'Solution NMR structure of the p300 Taz2:ETAD1 complex' and 6MGN 'mouse Id1 (51-104) - human hE47 (348-399) complex' (titles retrieved and verified at RCSB) · RCSB Protein Data Bank (entry metadata retrieved and title verified) · RCSB PDB 2026What this evidence cannot show
6MGN's title numbers its human E47 chain 348-399 while UniProt maps the same chain to P15923 561-612, a 213-residue difference arising from the construct's own numbering. Always map PDB chains through the UniProt cross-reference rather than through a paper's figure labels.
Evidence that argues against this (1)
- The 213-residue offset is real but it is a property of PDB 6MGN's construct alone, and this node's phrasing - 'the paper's own E47 numbering and the reference-sequence numbering differ by 213 residues' - risks being read as a general E47-versus-canonical conversion. It is not. The E12-to-E47 relationship is E47 = E12 minus 3 downstream of the variable exon and is not arithmetic inside it. Anyone who combines the two offsets will be 210 residues wrong. PDBe / SIFTS; EMBL-EBI 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The t(1;19) translocation produces an 85 kDa chimeric protein consisting of the amino-terminal two-thirds of E2A fused to a chromosome-1 homeoprotein, in which the E2A DNA-binding domain is replaced by the homeoprotein's DNA-binding domain; identical junctions were found in three independent t(1;19) cell lines.
Chromosomal translocation t(1;19) results in synthesis of a homeobox fusion mRNA that codes for a potential chimeric transcription factor · Nourse J, Mellentin JD, Galili N, et al. · Cell 1990What this evidence cannot show
Three cell lines, 1990 cDNA cloning; the partner is named 'Prl' in the original paper and was subsequently renamed PBX1.
Evidence that argues against this (1)
- Functional dissection of the same chimera puts the weight entirely on the E2A side. Mutating the E2a transactivation motifs AD1 and AD2 significantly impaired transformation, while a mutant lacking the Pbx1 homeodomain altogether (E2a-PbxdeltaHD) still efficiently transformed NIH 3T3 cells and induced malignant lymphomas in transgenic mice; hyperexpressed wild-type or truncated Pbx1 was not transforming on its own. Transformation is absolutely dependent on motifs acquired from E2a and the Pbx1 homeodomain is optional, which makes the DNA-binding-domain-replacement description a correct account of the sequence and a misleading account of the mechanism. Monica K et al. 1994
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
Recomputing the Ensembl TCF3-201 (ENST00000262965, 654 aa) coding exon structure places the end of coding exon 16 at codon 483, matching the UniProt-annotated TCF3-PBX1 breakpoint at 483/484.
Ensembl transcript ENST00000262965 (TCF3): exon structure and coding-sequence coordinates, used to convert exon breakpoints into residue numbers · Ensembl; EMBL-EBI · Ensembl REST API 2026What this evidence cannot show
Computed by us; the exon ends mid-codon (+1 nt) so the junction codon is chimeric. The Ensembl canonical transcript encodes the same 654 aa sequence as the UniProt canonical, so there is no isoform mismatch here (unlike RUNX1).
Evidence that argues against this (1)
- TCF3 has the most crowded transcript model of the genes recomputed in this set, so 'exon 16' is only meaningful with the transcript named. The live Ensembl record lists 35 protein-coding TCF3 transcripts; the canonical ENST00000262965 has 19 exons and 654 aa, but the exon-count distribution across the 35 is 3, 4 (x4), 5 (x4), 6 (x2), 16 (x2), 17 (x3), 18 (x2), 19 (x15) and 20 (x2), and the longest protein-coding transcript, TCF3-234 / ENST00000931972, encodes 710 aa from 18 exons. Sources reporting a t(1;19) junction as E2A exon 16 on a 16-, 17-, 18- or 20-exon model will not convert to 483. 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-P15923-F1 covers all 654 residues.
AlphaFold Protein Structure Database entry for P15923 (model AF-P15923-F1) · Jumper J, Varadi M, et al.; EMBL-EBI/DeepMind · AlphaFold DB 2026What this evidence cannot show
UniProt predicts disorder across 31-103, 135-205, 239-268, 292-329, 343-385, 461-552 and 633-654 — i.e. most of the protein. Only the bHLH and the short AD1 helix have experimental support; everything else in the model is low confidence.
Evidence that argues against this (1)
- This is the least informative AlphaFold model of the protein entries in this beat. Per-residue pLDDT recomputed from the v6 file gives 455 of 654 residues (70%) below 50, a further 91 (14%) between 50 and 70, only 73 residues (11%) at or above 90, and a mean pLDDT of 51.1 - against 78.1 for RARA and 76.3 for MYH11 by the same calculation. Experimental coverage is correspondingly thin: PDBe lists five structures for P15923 covering only 1-37 (2MH0, NMR), 538-616, 563-613 and 561-612. More than 500 of the 654 residues have neither a confident model nor any coordinates. AlphaFold DB / EMBL-EBI; PDBe SIFTS 2026
- E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Independent re-verification on 2026-09-19: the live UniProt record for P15923 gives the primary gene name TCF3 (synonyms BHLHB21, E2A, ITF1), a length of 654 residues, the bHLH at 549-602, the leucine-zipper region at 389-425, the 9aaTAD at 19-27, the NLS at 170-176 and the TCF3-PBX1 translocation site at 483/484; and SIFTS independently maps PDB 6MGN's human chain to P15923 residues 561-612 against the entry title's 'hE47 (348-399)', confirming the 213-residue construct-numbering offset exactly.
SIFTS UniProt residue-level mappings for PDB entries 6MGN, 3U5V, 2MH0 (construct ranges read back from the PDBe mappings API) · PDBe / SIFTS; EMBL-EBI · PDBe SIFTS 2026What this evidence cannot show
Database records read back 2026-09-19; no full text read for this item. 561 minus 348 equals 213 and 612 minus 399 equals 213, so the offset is constant across the construct, which is what makes it a numbering convention rather than an error. The offset is specific to 6MGN's construct and must NOT be applied to any other E47-numbered source. Note the trap inside the trap: the 213 offset is a construct convention and is unrelated to the 3-residue E12-minus-E47 isoform offset, and the two must never be combined.
Evidence that argues against this (1)
- The offset is real but it is one of at least two incompatible E47 numbering conventions in the PDB, which is the more dangerous fact. PDBe's best-structures listing for P15923 shows 2YPA and 2YPB (the SCL:E47:LMO2:LDB1 complexes, 2.8 and 2.87 A) mapping to P15923 538-616 and 3U5V mapping to 563-613, all deposited under different construct conventions from 6MGN's. The RCSB instance record for 6MGN chain A also shows author numbering 558-609 with no unobserved residues - a third numbering for the same 52 modelled residues. A single 213-residue correction applied outside 6MGN will silently produce wrong residues, so this node's warning is necessary but not sufficient. PDBe / SIFTS et al. 2026
- E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The canonical-isoform identity of P15923 was re-confirmed from the live record: UniProt's alternative-products list marks isoform E12 as the displayed sequence at 654 aa, with E47 as P15923-2 at 651 aa and a third isoform at 658 aa.
UniProtKB entry P15923 (TFE2_HUMAN): reviewed record re-fetched and read back on 2026-09-19 · UniProt Consortium · UniProtKB/Swiss-Prot 2026What this evidence cannot show
Database record re-fetched 2026-09-19 and read back; the sequence-level alignment that originally established the E12/E47 assignment was done in cycle 11 and was not repeated in this pass. The claim re-verified here is the isoform table and the canonical length, not a fresh sequence alignment.
Evidence that argues against this (1)
- Three isoforms is a curation decision, not a measurement of what the locus makes. Ensembl's live record for TCF3 lists 35 PROTEIN-CODING transcripts, with lengths reaching 710 aa (TCF3-234 / ENST00000931972), 707 (ENST00000948466), 698, 683 (x3) and 682 (x2) - all longer than any of the three UniProt isoforms - from 20-, 19-, 18-, 17- and 16-exon models. Anyone reconciling a paper's TCF3 residue number against one of the three UniProt isoforms has roughly a one-in-twelve chance of being in the right frame if the paper used an Ensembl model. Ensembl; EMBL-EBI 2026
Counter-evidence added cycle 41: the weight is on the E2A side, not the homeodomain (Monica 1994, PMID 7969166). Separately, three of TCF3's four stored UniProt features (9aaTAD 19-27, NLS 170-176, leucine zipper 389-425) have NO support from any of the ten InterPro matches for P15923, and the bHLH spreads 536-627 across member databases against the stored 549-602.
- related to →TCF3::PBX1 (E2A-PBX1), t(1;19)(q23;p13.3)
Trap re-verified 2026-09-19 and both halves hold: P15923 canonical is isoform E12 (654 aa) not E47 (651 aa), and SIFTS maps PDB 6MGN's human chain to 561-612 against the entry title's 348-399, exactly 213 residues. New caution added: the 213-residue construct offset must never be combined with the 3-residue E12/E47 isoform offset.
- related to →TCF3::HLF (E2A-HLF), t(17;19)(q22;p13)
Same isoform and construct-numbering traps apply to the second TCF3 fusion: both fusions take TCF3 1-483, which lies upstream of the E12/E47 divergence at residue 529, so the TCF3 moiety is isoform-independent -- but any bHLH coordinate quoted for wild-type TCF3 is not.
5-prime partner domain architecture; AD1 and AD2 retained, bHLH lost
TCF3::PBX1, t(1;19)
same TCF3 5-prime contribution in TCF3::HLF, t(17;19)
TCF3's retained AD1/AD2 activation modules are redirected to PBX/HOX sites
t(1;19)(q23;p13.3)
- supports →Do different breaks give different biology? A verdict on partner identity vs breakpoint position
TCF3::PBX1 and TCF3::HLF share an identical TCF3 1-483 moiety yet behave very differently, isolating the partner as the variable
- ← related toTCF3::PBX1 (E2A-PBX1), t(1;19)(q23;p13.3)
Counter-evidence: Gow 2014 (PMID 24064250) shows the E2A AD1 p300/CBP motif is also the ETO corepressor motif, so 2MH0 does not establish AD1 as a coactivator-only surface
- ← related toTCF3::HLF (E2A-HLF), t(17;19)(q22;p13)
Counter-evidence: Baudis 2006 (PMID 16206189) shows E2A-HLF induces ABCB1 DNA-binding-independently, as do E2A-PBX1 and truncated E2A alone, undercutting the bZIP-driven transactivation model
TCF3 exon 11 or 13 contributes the N-terminal activation region; the bHLH at 549-602 is lost