TCF7L2 - the protein IntAct's 'DOT1L-TCF4' row is actually about, and the Wnt effector that DOT1L and AF10 coactivate

One database on one day, REST endpoint (brief rule 93). The inference that the bait was mis-assigned rests on the paper's own vocabulary and on the internal oddity of Tcf7l2 appearing as a prey of Tcf4; IntAct has not been contacted and the curation record was not inspected beyond the REST fields. The reading is stated as an adjudication, not as a confirmed database error.

• An alternative reading is that the mouse crypt pulldown genuinely recovered both ITF2 and TCF7L2 and the curator assigned each correctly. That cannot be excluded from the REST record alone, but it does not explain the human arm, where TCF7L2 is absent entirely from a paper whose subject is the TCF4/beta-catenin complex in colorectal cancer. EMBL-EBI IntAct 2026

Mouse intestine, zebrafish, colorectal cancer lines and HEK293T - no haematopoietic system, no KMT2A fusion (brief rule 73). The 'Tcf4' of this paper is the Wnt effector TCF7L2, and the entry is written on that reading; the identifier adjudication is in ana-the-dot1l-tcf4-partner-is-tcf7l2-not-tcf4. The interaction rows themselves are spoke-expanded, so the binary contact between TCF7L2 and DOT1L is inferred, not observed - the strong grade here is on the functional epistasis, not on the pairwise binding. New to this graph (checked with scripts/check-citation-reuse.py).

• The Wnt frame does not exclude the other TCF4, because ITF2 is itself a Wnt/beta-catenin target gene. Zhai et al. quantified six candidate beta-catenin/TCF-regulated genes across 44 primary ovarian endometrioid adenocarcinomas and found ITF2 - which is P15884, the gene whose CURRENT official symbol is TCF4 - significantly elevated in the tumours with deregulated beta-catenin, alongside MMP-7, CCND1, CX43 and PPAR-delta, while c-myc was not. So 'the paper is about Wnt in intestinal crypts, therefore its Tcf4 is the Wnt effector' is a strong argument and not a closed one: the other TCF4 is a documented member of the same pathway's output in the same tissue type. The adjudication in this entry should be read as well-supported rather than as settled. Zhai Y et al. 2002
• The collision is live inside this project's own disease, and nothing in this entry resolves it there. Kang et al. analysed 100 T-ALL patients and identified 330 probe sets discriminating KMT2A-rearranged (n=12) and MLLT10-rearranged (n=9) cases, naming RUNX2, TCF4 and MYO6 among the novel targets - a KMT2A-rearranged leukaemia paper reporting 'TCF4' as a transcriptional target with no accession, and with both candidate genes plausible (the Wnt effector in a DOT1L/MLLT10 context, the bHLH E2-2 in a lymphoid one). Importing this entry's intestinal adjudication to that record would be exactly the move brief rule 66 forbids: the source document is ambiguous and no lookup resolves it. Kang H et al. 2018

A statement about identifiers, verified by a single TSV query returning all four records in one response; the primary gene name was read back for each (brief rules 62 and 65). It is not a claim that any specific entry in this graph is wrong.

• UniProt understates the relationship and HGNC names it precisely: TCF4 is not a synonym of TCF7L2, it is TCF7L2's PREVIOUS OFFICIAL SYMBOL. Fetched 2026-09-21, HGNC:11641 TCF7L2 carries prev_symbol ['TCF4'] and alias_symbol ['TCF-4'], while HGNC:11634 TCF4 carries no previous symbol at all and the aliases SEF2-1B, ITF2, bHLHb19 and E2-2. That is brief rule 69's class - a previous official symbol is not an alias - and it is worse than a synonym clash for two reasons: pre-rename literature used TCF4 as the CORRECT name of TCF7L2, so the ambiguity is chronological rather than stylistic; and alias-expanding resolvers that deliberately exclude previous symbols (which is the safe default after the CD15/FUT4 case) will fail to connect the older Wnt literature to Q9NQB0 at all. HGNC 2026

One database on one day, REST endpoint (brief rule 93). The row is spoke-expanded, so the binary pair was inferred by a curator from an n-ary deposition and was never observed as a binary interaction. Recorded so that TCF7L2's own standing in this interactome is on the record and not conflated with the Wnt result, which comes from a different publication.

• A single spoke-expanded row from a proteome-scale network is the weakest class of evidence this interactome contains - 70.6% of all its partner pairs rest on nothing else - so this row supports the existence of a curated record and nothing about a contact. EMBL-EBI IntAct 2026

Abstract-level. A proteome-scale interaction network in a single cell background; no haematopoietic system, no KMT2A fusion (brief rule 73). Cited here to establish what kind of study the row came from, which is the part that determines how much the row is worth.

• 'Exactly one row' is one pair out of a hundred on a single interaction record. Re-enumerated at the endpoint on 2026-09-21, PMID 30833792 contributes 3,055 IntAct rows carrying only 102 distinct interaction ACs - 3,022 spoke-expanded against 33 non-spoke - and EBI-21265942, the AC that carries the TCF7L2-PSIP1 pair, is one AP-MS deposition split into 100 pairs, almost all at intactMiscore 0.35. So this is not 'one weak row' in the sense of one weak experiment; it is one prey identified in one pulldown alongside ninety-nine others, and the same AC also produces TCF7L2 against CTNNB1 (P35222, mi 0.97), CTNNB1-adjacent cadherins and the entire mitochondrial ribosome, which is the profile of a bait whose prey list is dominated by abundance. EMBL-EBI IntAct 2026

Database comment fetched on one day. The PML-body line rests on SUMOylation papers, so the compartment is a modification-dependent pool and not a resting address. Nothing haematopoietic in the citation set.

• 'PML body' is probably not a defining address for this protein but one of several sequestration sinks, which weakens its value as a localisation record. The same protein is independently reported in a DIFFERENT body class - HIC1 bodies - by a different mechanism, and UniProt's own note concedes the dominant pattern is diffuse. A protein found in whichever nuclear body its partner builds is being relocated by partners rather than residing anywhere, and the PML assignment should not be quoted as though it were TCF7L2's compartment. EMBO J 2006

Non-haematopoietic cells; read at abstract level. The paper's TCF-4 is the Wnt effector TCF7L2 and not ITF2/P15884 - resolved by subject, because the symbol alone cannot decide it (collision registry, TCF4 row). Sequestration into bodies is shown as a mechanism of repression, not as a constitutive localisation.

• The mechanism may be absent exactly where this graph would want it. HIC1 is epigenetically inactivated by hypermethylation in cancer - the paper's own opening premise - so in most transformed cells there is too little HIC1 protein to build HIC1 bodies, and the sequestration switch cannot operate. What would survive HIC1 loss is the SUMO/PIAS4-driven PML-body pool, a different compartment with a different trigger. Any attempt to import 'body sequestration silences a cofactor' into a leukaemic nucleus has to say which body, and the HIC1 route is the one least likely to be available. EMBO J 2003

Antibody immunofluorescence in non-haematopoietic lines; HPA's Nuclear bodies class does not identify WHICH bodies, so its agreement with the PML/HIC1 literature is suggestive rather than confirmatory. Blood Atlas values are RNA.

• The monocyte nTPM values are RNA and probably overstate a protein pool that may not exist. Protein-level work reports TCF-4 as RESTRICTED and high-level specifically in intestinal and mammary gland epithelium, which is an explicit statement that high TCF-4 protein is a tissue-restricted phenomenon. A 61.8 nTPM transcript in non-classical monocytes is not evidence of a functional monocyte pool, and this entry's cell-type answer should be read as the weakest in the tranche until somebody blots for the protein in sorted monocytes. Am J Pathol 1999