TBXT / brachyury - the notochord master regulator, and one of only two of 102 transcription-factor baits whose neighbourhood contained DOT1L
also called T-box transcription factor T · brachyury · T · TBXT · brachyury protein · T-box protein T
Brachyury is the transcription factor that tells an embryo where its spine goes, and in adults it is the single gene the bone cancer chordoma cannot live without. In a screen of 102 tagged transcription factors it was one of only two whose neighbourhood contained DOT1L - the enzyme this project measures constantly - which makes its row the most selective single observation in either proximity dataset this analysis walked.
The same thing, for a scientist
TBXT (HGNC:11515, previous symbol **T**, ENSG00000164458, UniProt **O15178**, 435 aa, 6q27) is the founding T-box transcription factor, required for mesoderm formation and notochord development; a homozygous T-box missense H171R causes sacral agenesis with a persistent notochordal canal [E4]. It is **the top selectively essential gene in chordoma**, sits in a 1.5-Mb super-enhancer region, and chordoma lines are killed by transcriptional CDK inhibitors against **CDK7/12/13 and CDK9** [E5]. It enters this project on one curated row: IntAct **IM-28767** (Göös 2022), **MI:1314 BioID**, **MI:1060 spoke**, **TBXT bait / DOT1L prey**, MI-score 0.27, host HEK293 Flp-In [E1][E2]. **What makes that row worth an entry is the denominator, re-derived here rather than quoted:** DOT1L is recovered by **2 of the 102 BioID baits** in that deposition against KMT2A's 11, and under a null holding both the bait's reach (TBXT labelled 111 preys) and the prey's recoverability, **TBXT→DOT1L has p = 0.039 - the second smallest of all 62 bait-to-core-protein pairs across both depositions**, and the only one of the 22 partners under adjudication to clear the pre-registered PROMOTED threshold [E3]. **It is still one spoke-expanded proximity row in an embryonic kidney cell, it does not survive correction for 62 tests, and no paper anywhere links brachyury to KMT2A, MLL or H3K79** [E6].
No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.
What the interaction record actually is
One IntAct row, IMEx IM-28767. Detection MI:1314 proximity-dependent biotin identification; expansion MI:1060 spoke; experimental roles MI:0496 bait (TBXT) / MI:0498 prey (DOT1L); host taxid:9606(Homo sapiens HEK293flp-in embryonic kidney cell); MI-score 0.27, the deposition-wide value, carrying no pair-specific information 2.
Proximity, over-expression, embryonic kidney, spoke-expanded. The binary TBXT-DOT1L pair was never measured; it was inferred by joining every prey in the TBXT pulldown to the bait. Brachyury is a notochordal transcription factor and HEK293 is a kidney line: this is an ectopic construct in a cell that does not normally make it. That is the reason the entry exists and it is not the reason to believe anything.
Why this row is not interchangeable with the other twelve - the measurement, not the intuition
This graph already records that eleven TF baits in IM-28767 recover KMT2A and only two recover DOT1L (GATA3, TBXT), and gene-gata3 argues from it that a 2-of-109 prey is more selective than an 11-of-109 prey. That argument was second-hand; this beat re-derived the denominators from the deposition and then tested them 3.
Re-derived: the BioID arm resolves to 102 distinct bait symbols and 946 human preys. DOT1L is recovered by 2. KMT2A by 11. The size-and-abundance model predicts 11.5 baits for DOT1L (1,537 aa, 0.273 ppm) and it is found by 2 - a ratio of 0.17. The assumption-free version is milder and is reported because it is the honest one: against a pool of 42 preys matched to DOT1L on both length (+/-25%) and abundance (+/-0.5 log10), whose median is 6 baits, DOT1L sits at the 21st percentile 3.
So the strong statement - DOT1L is systematically missed by the TF panel - is a model statement and the empirical pool softens it to DOT1L is in the bottom quintile of comparable preys. Both point the same way, and the pair null that uses neither assumption gives TBXT→DOT1L p = 0.039. That is the number this entry rests on.
The honest ceiling on that p-value
62 pairs were tested. Bonferroni's threshold is 0.0008. TBXT→DOT1L at 0.039 does not survive it, and neither does anything else in either deposition 3. The pre-registered PROMOTED threshold (p < 0.05) was fixed before any number was seen and is applied as declared, but it was declared as a way of ordering an unorderable tail, not as a significance test.
The null also assumes preys are drawn independently given their degrees. Baits that sit in a shared complex share preys, so the real null is wider than the simulated one, in the same direction as the overdispersion this beat found elsewhere. The correct reading is: of 62 pairs, this is the one the data flags first - not that the data establishes it.
Localization, subnuclear compartment, cell type and cell state
Subcellular (UniProt): Nucleus, ECO:0000269 experimental, cited to PMID 22611028 (brachyury immunohistochemistry in lung carcinoma) and PMID 24253444 (the sacral-agenesis pedigree, in which the H171R mutant's protein transport was assayed) 4. Both codes are real experiments, and rule 91 applies: neither paper is about brachyury's localisation, so the annotation is a pointer.
Subnuclear (HPA immunofluorescence): nucleoplasm, main location, reliability Supported, with no additional compartment 7. DOT1L's own HPA call is also nucleoplasm 7. The two share a compartment call, which is compatible with the proximity row and does not evidence it - a shared compartment is what makes an uninformative proximity row look informative.
Domain architecture: a single T-box DNA-binding domain, residues 51-219, and a disordered region 279-308; no chromatin reader, no enzymatic domain 4.
Cell type of the observation: HEK293 Flp-In embryonic kidney - not notochord, not blood, not leukaemic, not KMT2A-rearranged. Cell state: an over-expressed tagged construct in a proliferating immortalised line.
Expression in blood: HPA classes TBXT's cell-line RNA as group enriched with detectable expression in leukaemia (2.4 nTPM), lymphoma (6.1) and lung cancer (8.2) 7. That is low, it is bulk RNA across a cell-line panel rather than a measurement in KMT2A-rearranged blasts, and it is recorded as a pointer for anyone who wants to check whether brachyury is expressed in the index disease at all.
Epigenetics, and the one genuine mechanistic bridge
TBXT has no chromatin-modifying or chromatin-reading activity. It is a sequence-specific T-box factor. Any epigenetic role is indirect.
The substantive bridge runs the other way, through transcriptional CDKs. Sharifnia et al. found by genome-scale CRISPR screening that T/brachyury is the top selectively essential gene in chordoma, that it is the most highly expressed super-enhancer-associated transcription factor in that disease, sitting in a 1.5-Mb super-enhancer region, and that CDK7/12/13 and CDK9 inhibitors potently suppress chordoma proliferation and preferentially downregulate brachyury protein, with tumour-growth reduction in vivo 5.
CDK9 is the kinase of P-TEFb, the module the super elongation complex delivers and which this graph carries as protein-ptefb-cdk9-ccnt1 and gene-cdk9. So chordoma and KMT2A-rearranged leukaemia are two diseases whose addiction is to a transcription-elongation-dependent lineage programme rather than to a kinase or a receptor, and both are reported sensitive to CDK9 inhibition. That is an analogy between diseases, not evidence for the proximity row, and this entry keeps them separate.
Other fusion families, and a bounded absence with both operands
No paper links brachyury to this project's subject matter at all. Run with scripts/pubmed_safe.py, which raises rather than returning an unexplained zero, and with both the modern and the historical gene name as the rule requires 6:
| query | translation | n |
|---|---|---|
(TBXT[tiab] OR brachyury[tiab]) AND KMT2A[tiab] | ("TBXT"[Title/Abstract] OR "brachyury"[Title/Abstract]) AND "KMT2A"[Title/Abstract] | 0 |
(TBXT[tiab] OR brachyury[tiab]) AND MLL[tiab] | ("TBXT"[Title/Abstract] OR "brachyury"[Title/Abstract]) AND "MLL"[Title/Abstract] | 0 |
(brachyury OR TBXT) AND (DOT1L OR H3K79 OR KMT2A OR MLL), all [tiab] | non-empty, returned by the same instrument | 0 |
control: brachyury[tiab] AND (DOT1L OR H3K79 OR chordoma OR "super-enhancer")[tiab] | - | 270 |
All zeros are genuine: non-empty querytranslation, warninglist reporting No items found. The control returns 270, so the instrument works. Bounded by titles and abstracts only - a supporting figure in a chordoma paper could contain an H3K79me2 track and this query would not see it.
Other fusion families: nobody has looked, in any of them. TBXT is not a recurrent fusion partner in leukaemia and this entry does not suggest it is. The chordoma link is a disease analogy, not a fusion analogy.
| hgnc ID | HGNC:11515 |
|---|---|
| symbol | TBXT |
| ensembl gene ID | ENSG00000164458 |
| uniprot | O15178 |
| length AA | 435 |
| chromosomal location | 6q27 |
| entrez ID | 6862 |
| identifier provenance | HGNC REST fetch/symbol/TBXT and UniProtKB O15178 retrieved live 2026-09-23; Ensembl id taken from the HGNC record's ensembl_gene_id field, not from recall. The previous symbol T was word-boundary matched against the full corpus id set and node text and resolves to no existing entry. |
| subnuclear location hpa | Nucleoplasm (main, reliability Supported, no additional location) |
| cell type of the observation | HEK293 Flp-In embryonic kidney - not notochord, not blood, not leukaemic, not KMT2A-rearranged |
| cell state of the observation | over-expressed tagged construct in a proliferating immortalised line |
| epigenetic role | None direct. TBXT is a sequence-specific T-box transcription factor with no chromatin-modifying or chromatin-reading domain. The indirect link is that TBXT is itself super-enhancer-driven in chordoma and is depleted by CDK7/12/13 and CDK9 inhibition. |
| shown for fusion | NO - nobody has looked in any fusion. The row is wild-type DOT1L in HEK293 Flp-In, and PubMed returns zero for brachyury against KMT2A and zero against MLL. |
| evidence class | single publication, single method, spoke-expanded BioID proximity, no orthogonal validation; promoted on the rarity of its prey, not on the strength of its row |
prev and alias symbols
[ "T" ]
domains
[
{
"name": "T-box (DNA binding)",
"start": 51,
"end": 219
}
]interaction with project baits
[
{
"core_protein": "DOT1L",
"core_uniprot": "Q8TEK3",
"partner_experimental_role": "bait",
"core_experimental_role": "prey",
"method": "MI:1314 proximity-dependent biotin identification",
"expansion": "MI:1060 spoke",
"mi_score": 0.27,
"imex": "IM-28767",
"pmid": "35140242",
"host_cell": "HEK293 Flp-In embryonic kidney",
"n_rows": 1,
"n_publications": 1
}
]adjudication
{
"verdict": "PROMOTED",
"p_pair_both_margins_held": 0.039,
"rank_among_62_pairs": 2,
"bait_preys_recovered": 111,
"bait_selectivity_rank_in_panel": "17 of 102 (1 = widest labeller)",
"prey_baits_recovering_it": 2,
"of_panel": 102,
"multiplicity": "Bonferroni threshold over 62 tests is 0.0008; this does NOT survive correction and is an ordering, not a test",
"source": "analyses/interactome-tail-size-and-abundance-adjudication/pair_null.json"
}subcellular location uniprot
[ "Nucleus (ECO:0000269, cited to PMID 22611028 and PMID 24253444 - neither is a localisation study)" ]
expression in blood lines hpa ntpm
{
"leukemia": 2.4,
"lymphoma": 6.1,
"lung_cancer": 8.2,
"note": "HPA 'RNA cell line specific nTPM' summary field, group enriched; low, bulk, and not measured in KMT2A-rearranged blasts"
}bounded absence
{
"queries": [
"(TBXT[tiab] OR brachyury[tiab]) AND KMT2A[tiab] -> 0",
"(TBXT[tiab] OR brachyury[tiab]) AND MLL[tiab] -> 0",
"(brachyury[tiab] OR TBXT[tiab]) AND (DOT1L[tiab] OR H3K79[tiab] OR KMT2A[tiab] OR MLL[tiab]) -> 0"
],
"querytranslations": [
"(\"TBXT\"[Title/Abstract] OR \"brachyury\"[Title/Abstract]) AND \"KMT2A\"[Title/Abstract]",
"(\"TBXT\"[Title/Abstract] OR \"brachyury\"[Title/Abstract]) AND \"MLL\"[Title/Abstract]"
],
"positive_control": "brachyury[tiab] AND (DOT1L OR H3K79 OR chordoma OR super-enhancer)[tiab] -> 270",
"bound": "titles and abstracts only; supporting figures not searched"
}- If a rationale for testing a DOT1L inhibitor in chordoma were ever built on this entry, it would rest on one spoke-expanded proximity row in a kidney cell that does not survive correction for 62 tests. This entry should be cited as the reason to look, never as the reason to dose.
- Brachyury is required for normal axial development and a homozygous hypomorph causes sacral agenesis in humans. Any agent that depletes brachyury protein - which is what the transcriptional CDK inhibitors do - carries a developmental-toxicity question that matters most in exactly the paediatric population this project serves.
- CDK9 inhibition is the shared node in the analogy above and it is not a selective intervention: P-TEFb is required for the elongation of most genes, and this graph already records its pan-essentiality as the reason ELL's dependency is probably not about the fusion.
- Is brachyury expressed at all in KMT2A-rearranged blasts? HPA records 2.4 nTPM across leukaemia cell lines, which is a bulk cell-line summary and not a measurement in the index disease. One query against an existing infant-ALL expression dataset settles whether this entry describes anything that happens in a patient.
- Is DOT1L's 2-of-102 recovery a fact about DOT1L or about the TF bait panel? The discriminating measurement is whether DOT1L is also under-recovered relative to size and abundance in the RBP/K562 panel - it is recovered by 4 of 50 there against a prediction of 7.9, which points the same way from an unrelated bait set, and that comparison deserves a proper test rather than the eyeball it gets here.
- Chordoma and KMT2A-rearranged leukaemia are both reported sensitive to CDK9 inhibition, one through brachyury super-enhancer addiction and one through the super elongation complex. Is the shared node P-TEFb availability, or is it the generic sensitivity of any highly expressed enhancer-driven transcription factor? The chordoma paper's own framing favours the second, and the graph should not let the coincidence harden.
- Does DOT1L occupy the TBXT super-enhancer in chordoma? H3K79me2 marks actively transcribed gene bodies, the 1.5-Mb T locus is the most highly expressed super-enhancer-associated region in that disease, and no published chordoma dataset has been asked this question.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Göös et al. identified 6,703 BioID and 1,536 AP-MS protein-protein interactions for 109 human transcription factors in HEK293 Flp-In cells; in the resulting IntAct deposition IM-28767, TBXT is one of only two baits whose BioID neighbourhood contained DOT1L.
Human transcription factor protein interaction networks. · Göös H, Kinnunen M, Salokas K, Tan Z, Liu X, Yadav L, Zhang Q, Wei GH, Varjosalo M · Nat Commun 2022What this evidence cannot show
Abstract read; full text and Supplementary Table 1 not read. Every bait is an over-expressed tagged construct in embryonic kidney cells. The paper describes 109 baits; the BioID arm of the deposition resolves to 102 distinct bait symbols in this beat's parse and that discrepancy was not reconciled.
Evidence that argues against this (1)
- A screen of transcription factors in a kidney cell is not a survey of DOT1L's partners. DOT1L appearing in only two neighbourhoods may reflect that the panel is unsuited to reaching it rather than that these two are special. Go CD et al. 2021
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
IntAct serves the TBXT-DOT1L record as PSI-MI MI:1314 proximity-dependent biotin identification, MI:1060 spoke expansion, TBXT experimental role MI:0496 bait and DOT1L MI:0498 prey, MI-score 0.27, host taxid:9606 Homo sapiens HEK293flp-in embryonic kidney cell, IMEx IM-28767, publication PMID 35140242.
What this evidence cannot show
Retrieved live 2026-09-23 by paginating the PSICQUIC stream to exhaustion; 8,331 rows fetched against an endpoint-declared count of 8,331. Rule 93: PSICQUIC is a floor relative to IntAct REST, so a REST query on O15178/Q8TEK3 could return additional rows or a different method, and that query was not run for this pair.
Evidence that argues against this (1)
- MI-score 0.27 is the deposition-wide floor value applied to every row in this arm. It carries no evidence that this pair is better supported than any other row in the same file. source 2026
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Re-derived from the deposition: the BioID arm of IM-28767 resolves to 102 bait symbols and 946 human preys; DOT1L is recovered by 2 baits and KMT2A by 11; a negative-binomial model of bait count on log10 length and log10 abundance predicts 11.5 baits for DOT1L, and against 42 preys matched to DOT1L on both covariates (pool median 6 baits) DOT1L sits at the 21st percentile. Under a null holding both the bait's prey count and the prey's bait count, TBXT (111 preys) recovering DOT1L has p = 0.039, the second smallest of 62 bait-to-core pairs across both depositions, behind GATA3-DOT1L at 0.032.
What this evidence cannot show
A Bonferroni threshold over 62 tests is 0.0008 and NOTHING in either deposition survives it; the p-value is used as an ordering, not a test. The null assumes preys are drawn independently given their degrees, which is false when baits share a complex, making it anti-conservative. 20,000 draws, seed 20260923, exact weighted sampling without replacement, analytic Chung-Lu value printed beside each simulated one.
Evidence that argues against this (1)
- The same null ranks TBXT ahead of every other partner precisely because DOT1L is rarely recovered. If DOT1L is rarely recovered because the TF panel is bad at reaching it - a property of the instrument rather than of DOT1L - then the ranking rewards the instrument's blind spot rather than biology, and this analysis cannot distinguish those. source 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
UniProt O15178 is 435 aa with a single T-box DNA-binding domain at residues 51-219 and a disordered region at 279-308, and assigns subcellular location Nucleus with ECO:0000269 experimental evidence citing PMID 22611028 and PMID 24253444; the latter reports that a homozygous T-box missense mutation p.H171R in three consanguineous families causes sacral agenesis, a persistent notochordal canal and abnormal vertebral ossification, with diminished DNA binding.
Mutations in the T (brachyury) gene cause a novel syndrome consisting of sacral agenesis, abnormal ossification of the vertebral bodies and a persistent notochordal canal. · Postma AV, Alders M, Sylva M, et al. · J Med Genet 2013What this evidence cannot show
Abstract read; full text not read. UniProt record fetched live 2026-09-23 from rest.uniprot.org/uniprotkb/O15178.json. Rule 91: an ECO:0000269 code means a curator saw an experiment in that paper, not that the paper is about brachyury's subcellular localisation - neither cited paper is a localisation study, and the founding observation for nuclear brachyury is not identified here.
Evidence that argues against this (1)
- A recessive human syndrome establishes that brachyury is required for axial development; it says nothing about a role in haematopoiesis, and no phenotype in these pedigrees is haematological. source 2026
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Genome-scale CRISPR-Cas9 screening and small-molecule sensitivity profiling identify T (brachyury; TBXT) as the top selectively essential gene in chordoma; T is the most highly expressed super-enhancer-associated transcription factor in that disease and lies in a 1.5-Mb super-enhancer region; transcriptional CDK inhibitors targeting CDK7/12/13 and CDK9 potently suppress chordoma cell proliferation, preferentially and dose-dependently deplete brachyury protein in every model tested, and substantially reduce tumour growth in vivo.
Small-molecule targeting of brachyury transcription factor addiction in chordoma. · Sharifnia T, Wawer MJ, Chen T, Huang QY, Weir BA, Sizemore A, et al. · Nat Med 2019What this evidence cannot show
Abstract read; full text not read. Chordoma, a bone sarcoma of notochordal remnants, is an entirely different disease from KMT2A-rearranged acute leukaemia and shares no lesion with it. The CDK9 sensitivity is reported alongside CDK7/12/13 and the abstract does not separate their contributions, so 'chordoma is CDK9-dependent' is not established by this citation on its own.
Evidence that argues against this (1)
- Transcriptional CDK inhibitors deplete many highly expressed enhancer-associated transcription factors, which the authors state explicitly. Sensitivity to CDK9 inhibition is therefore a general property of super-enhancer-driven transcription factor addiction and is not specific evidence of a brachyury-elongation-machinery relationship. source 2019
- E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PubMed returns zero records for (TBXT OR brachyury) AND KMT2A in titles and abstracts, and zero for the same pair of names against MLL, and zero for (brachyury OR TBXT) against DOT1L, H3K79, KMT2A or MLL; the instrument returns 270 records for brachyury against DOT1L, H3K79, chordoma or super-enhancer as a positive control.
What this evidence cannot show
Both zeros carry a non-empty querytranslation and a warninglist reading 'No items found.', which is the signature of a genuine zero rather than an instrument failure. Bounded by TITLES AND ABSTRACTS ONLY: a measurement appearing in a supporting figure is invisible to this query, and no chordoma full text was searched. Name list tried: TBXT, brachyury, T (the previous HGNC symbol, not queried alone because it is unusable as a search token).
Evidence that argues against this (1)
- The previous approved symbol for this gene was the single letter T, which cannot be searched. A pre-2020 paper naming the gene only as T would be unreachable by any of these queries, and that is precisely the historical literature the KMT2A/MLL naming rule exists to protect against. source 2026
- E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
The Human Protein Atlas assigns TBXT a main subcellular location of nucleoplasm with reliability 'Supported' and no additional location, assigns DOT1L the same main location of nucleoplasm with reliability 'Enhanced', and classes TBXT's cell-line RNA expression as group enriched with 2.4 nTPM in leukaemia, 6.1 in lymphoma and 8.2 in lung cancer cell lines.
What this evidence cannot show
HPA subcellular calls are antibody-based immunofluorescence in a small panel of cell lines, not in leukaemic blasts, and 'Supported' is the middle reliability tier. The cell-line nTPM figures are the HPA summary field for enriched lines only, not a full per-line table - the per-cell-line download no longer exists, which is recorded in the parent analysis.
Evidence that argues against this (1)
- A shared nucleoplasm call for bait and prey is exactly what makes an uninformative proximity row look informative. Two proteins in the same compartment will be within labelling radius of each other some of the time regardless of whether they ever associate. source 2026
TBXT bait / DOT1L prey, one MI:1314 spoke-expanded BioID row, IM-28767, HEK293 Flp-In; one of only 2 of 102 TF baits reaching DOT1L
the only partner of 22 to clear the pre-registered PROMOTED threshold, p = 0.039 of 62 pairs
- co-occurs with →GATA3 - one of only two transcription-factor baits in a 109-bait BioID screen that recover DOT1L
the other of the two TF baits recovering DOT1L in IM-28767; GATA3-DOT1L p = 0.032, TBXT-DOT1L p = 0.039, the two smallest of 62
one of the 102 BioID baits in IM-28767
chordoma's brachyury addiction is suppressed by CDK7/12/13 and CDK9 inhibitors - a disease analogy to P-TEFb dependence, not evidence for the proximity row
- depends on →A BioID hit is a neighbour, not a complex member - and the two methods are known to disagree
TBXT and DOT1L share an HPA nucleoplasm call, which is what makes an uninformative proximity row look informative