TRIB3 / TRB3 (Q96RU7) - a pseudokinase that is the ISR's own brake, falls 3.1 log2 under either inhibitor, and has a published physical link to the MLL complex in the wrong tissue
also called TRIB3 · TRB3 · NIPK · SKIP3 · C20orf97 · Tribbles homolog 3 · TRIB3_HUMAN
TRIB3 is a fake kinase: it has the shape of an enzyme but no activity, and its job is to switch off the very stress programme that turns it on. It falls about eight-fold in KMT2A-rearranged leukaemia cells treated with either of two structurally unrelated drugs.
The same thing, for a scientist
TRIB3 (Q96RU7, 358 aa, catalytically inactive protein-kinase domain 68-316) is an ISR regulator induced via the ATF4-CHOP pathway that then **represses CHOP/ATF4 function and its own induction** - a negative-feedback node, not a simple output gene [E2]. In RS4;11 it falls **-1.23 log2 on the best array probe under MI-2-2 (two probes agreeing, -1.18 and -1.23) and -3.10 under VTP-50469 at day 7** [E1]. Two things make it more than another signature member. **First, the feedback architecture inverts the naive reading**: if TRIB3 is the brake on ATF4/CHOP output, losing TRIB3 should *de-repress* that output, so a drug that lowers both ATF4 output and its brake is doing something the linear model does not describe. **Second, there is a published physical link to the MLL complex and it is in the wrong tissue**: PMID 36142452 reports TRIB3 modulating PPARgamma-mediated growth inhibition **by interfering with the MLL complex in breast cancer cells** - the only record PubMed returns for TRIB3 jointly with any KMT2A-fusion operand [E4]. DepMap 26.06: mean gene effect **+0.0937** across 1,258 screens, 0.0% of lines below -0.5, leukaemia mean +0.0908 - **the most positive gene effect in this beat's panel**, so TRIB3 loss is, if anything, mildly favourable to growth [E3].
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 measurement, and the two probes that agree
| dataset | cell | arm | DMSO | log2FC |
|---|---|---|---|---|
| array 218145_at | RS4;11 | MI-2-2 | 8.78 | -1.23 (p 0.0007) |
| array 1555788_a_at | RS4;11 | MI-2-2 | 6.00 | -1.18 (p 0.013) |
| array 218145_at | SEM | MI-2-2 | 7.88 | -1.22 (p 0.024) |
| array 218145_at | REH | MI-2-2 | 10.31 | -0.68 (p 0.073, n.s.) |
| RNA-seq | RS4;11 d7 | VTP-50469 | 7.57 | -3.10 |
| RNA-seq | RS4;11 d7 | EPZ5676 | 7.57 | -3.07 |
| RNA-seq | RS4;11 d2 | VTP-50469 | 5.12 | -0.40 |
| RNA-seq | MOLM13 d7 | VTP-50469 | 6.08 | -0.46 |
Two independent array probes agree to 0.05 log2, and SEM agrees with RS4;11 to 0.01 log2 - this is the most internally consistent array result in the panel, unlike CHAC1 where SEM disagreed.
And the REH control is the WEAKEST in the panel. TRIB3 is the one concordant gene where the KMT2A-wild-type line also moves appreciably (-0.68, the largest REH excursion of any of the 24), and it is more highly expressed in REH (10.31) than in either t(4;11) line. The prior entry already noted this - "only TRIB3 (-0.68) exceeds 0.5" in REH - and it is worth restating as the limit on fusion-selectivity for this particular gene, not for the signature as a whole.
The feedback problem, which is the interesting part
Ohoka 2005 established TRIB3 as a target of CHOP/ATF4 that downregulates its own induction by repressing CHOP/ATF4 function, and showed that TRIB3 knockdown decreases ER-stress-dependent cell death 2. So in the canonical architecture:
- ATF4/CHOP up -> TRIB3 up -> ATF4/CHOP output damped -> TRIB3 induction damped. A classic negative feedback loop.
Under a menin inhibitor, both the output (CHAC1 -5.11, STC2 -4.40) and the brake (TRIB3 -3.10) fall together. A linear "ATF4 output is lost" model predicts that, but it also means the system has lost its damping, and the prediction from that is that the residual ATF4 output should be disproportionately large - which is not what is observed. The honest statement is that the feedback architecture is not reconcilable with the observed pattern without a measurement of ATF4 protein, which does not exist.
A cleaner falsifiable consequence: if TRIB3 loss de-represses CHOP, DDIT3 should fall LESS than the ATF4 targets that have no brake. Measured: DDIT3 -1.50 against CHAC1 -5.11 and STC2 -4.40 at RS4;11 day 7 1. That is the predicted ordering, and it is the only piece of internal mechanistic consistency this beat found. It is also exactly the kind of agreement that can arise from promoter strength and detection floors rather than from feedback, so it is graded moderate and not strong.
Identifier traps: TRB3 collides, and the MLL link is in a breast cancer
Rule 88: no TRIB3, TRB3, NIPK or SKIP3 row exists in data/symbol-collisions.json (157 rows read 2026-09-22). The dispatch brief predicted a TRB3 collision and it is real, measured as ID sets 4:
| token | records | shared with TRIB3 | Jaccard |
|---|---|---|---|
TRIB3[tiab] | 563 | - | - |
TRB3[tiab] | 288 | 11 | 0.0131 |
NIPK[tiab] | 11 | 1 | 0.0017 |
SKIP3[tiab] | 7 | 1 | 0.0018 |
TRB3 reaches 288 records and only 11 of them mention TRIB3. The remainder are largely T-cell receptor beta locus nomenclature (TRBV/TRBD/TRBJ/TRB) and unrelated abbreviations - and in a lymphoid leukaemia project that is the worst possible collision, because T-cell receptor beta rearrangement is on-topic vocabulary here. Fusion-scoped TRB3 returns 0 with a valid querytranslation, so nothing was missed, but a beat that searched TRB3 in an immunology context would be reading a different gene entirely. This is a registry-worthy row.
The one fusion-scoped TRIB3 record is PMID 36142452, "TRIB3 Modulates PPARgamma-Mediated Growth Inhibition by Interfering with the MLL Complex in Breast Cancer Cells" - title read, abstract not obtained. Rule 77 clause: a mechanism involving this graph's central complex does not have to be published in this graph's disease, and here it is not. It is recorded as a lead, not as evidence, precisely because it was not read.
| uniprot | Q96RU7 |
|---|---|
| length AA | 358 |
| entry name | TRIB3_HUMAN |
| protein name | Tribbles homolog 3 |
| ensembl gene ID | ENSG00000101255 |
| PDB note | UniProt Q96RU7 cross-references NO PDB entries. No human TRIB3 structure exists in the PDB. |
| alphafold ID | AF-Q96RU7-F1 |
| accession resolution | gene_exact:TRIB3 AND organism_id:9606 AND reviewed:true returns EXACTLY ONE reviewed row, Q96RU7, TRIB3_HUMAN, 358 aa, synonyms C20orf97, NIPK, SKIP3, TRB3; re-fetched alone with name and length compared back: PASS. |
| localization | NUCLEUS, on a single experimental citation - UniProt Q96RU7, ECO:0000269 from PubMed 22977230. RULE 91 FLAG: that paper is 'Human Tribbles 3 protects nuclear DNA from cytidine deamination by APOBEC3A', an APOBEC3A paper, NOT a TRIB3 localisation study. It supports TRIB3 acting on nuclear DNA and is therefore consistent with a nuclear pool, but it is not the founding or best observation of TRIB3's compartment, and no better coded source is annotated. Nothing places TRIB3 anywhere in a haematopoietic cell. |
| subnuclear localization | NOT ASSIGNED and NEVER LOOKED FOR. UniProt annotates Nucleus with no subnuclear qualifier, from an APOBEC3A paper. No nuclear body, speckle, nucleolar or chromatin-address observation for TRIB3 exists in any cell type. Recorded as a bounded absence; no compartment is invented. |
| cell type | RS4;11 and SEM (t(4;11) KMT2A::AFF1 B-ALL), REH (KMT2A wild-type, and the weakest negative control in the panel) and MOLM13 (KMT2A::MLLT3 AML), measured here [E1]. The mechanism work is in non-haematopoietic culture under tunicamycin [E2]. UniProt records highest expression in liver, pancreas, peripheral blood leukocytes and BONE MARROW - so the tissue is right even though no experiment has been done there. No primary leukaemia sample. |
| cell state | Measured in the day-6/day-7 drug-arrested state. TRIB3 is induced by ER stress LATER than CHOP [E2], so it is a late-state gene, and consistently its VTP-50469 effect is -0.40 at day 2 and -3.10 at day 7 - a state-dependent, not an immediate, response. Neither dataset controlled amino-acid supply, which is the physiological input to the ATF4-CHOP-TRIB3 arm. |
| epigenetic readout | No writer, eraser or reader activity of its own. The one published link to chromatin machinery is PMID 36142452, TRIB3 'interfering with the MLL complex' in BREAST CANCER cells - title-level only, unread, recorded as a lead. In this beat's data TRIB3's fall is reproduced to 0.02 log2 by the DOT1L inhibitor EPZ5676, so its expression tracks H3K79-methylation-dependent fusion output rather than the menin interface. No histone-mark measurement at the TRIB3 locus exists in any KMT2A-rearranged cell. |
| shown for fusion | MEASURED in KMT2A::AFF1 (RS4;11 and SEM, two agreeing array probes) and KMT2A::MLLT3 (MOLM13, only -0.46). Transcript only. The KMT2A wild-type control is WEAK here (-0.68 in REH), so of the 24 concordant genes TRIB3 is the one whose fusion-dependence is least well established. No protein, no ChIP, no functional test in any fusion context. |
| absence class | literature-absence-bounded |
| absence class note | RECLASSIFIED 2026-09-23 (cycle 76). This field held free text, which `scripts/check-absence-class.py` rejects; the prose is preserved here VERBATIM and unsummarised because it is more informative than the token that replaced it. Token chosen: literature-absence-bounded - the two-part prose says MEASURED-AND-<something> for the data arm and BOUNDED-UNPUBLISHED for the prose arm. The data arm is a RESULT, not an absence; the absence is the bounded literature, and the bound's own limits (titles and abstracts only, per rule 80) are stated in the preserved prose. ORIGINAL absence_class, verbatim: «MEASURED-AND-LARGE in transcript; BOUNDED-UNPUBLISHED in prose, with one unread title-level lead in breast cancer (PMID 36142452) and 0 menin-drug-scoped records.» |
PDB IDS
[]
domains
[
{
"name": "Protein kinase (catalytically INACTIVE pseudokinase)",
"start": 68,
"end": 316
}
]the measurement
{
"rs411_array_mi22_lfc_probe_218145_at": -1.233,
"rs411_array_dmso_log2_rma": 8.78,
"rs411_array_mi22_lfc_probe_1555788_a_at": -1.183,
"sem_array_mi22_lfc": -1.221,
"reh_array_mi22_lfc": -0.684,
"reh_array_dmso_log2_rma": 10.31,
"rs411_rnaseq_d7_vtp_lfc": -3.096,
"rs411_rnaseq_d7_epz5676_lfc": -3.072,
"rs411_rnaseq_d2_vtp_lfc": -0.397,
"molm13_rnaseq_d7_vtp_lfc": -0.457,
"menin_specificity": "SHARED WITH DOT1L INHIBITION (difference 0.02 log2)",
"note_on_selectivity": "TRIB3 has the WEAKEST fusion-selectivity of the 24 concordant genes: REH moves -0.68, the largest wild-type-line excursion in the set, and REH expresses TRIB3 more highly than either t(4;11) line."
}essentiality
{
"depmap_release": "26.06 (read back from meta{dataVersion})",
"depmap_screens": 1258,
"chronos_mean": 0.0937,
"pct_lines_below_minus_0_5": 0,
"leukaemia_screens": 57,
"leukaemia_mean_gene_effect": 0.0908,
"reading": "the most positive gene effect in this beat's panel; not a dependency in any lineage"
}identifier traps
{
"registry_status": "NOT in data/symbol-collisions.json as of 2026-09-22. The dispatch brief PREDICTED this collision and the measurement confirms it.",
"TRB3": "288 PubMed records against TRIB3's 563, sharing only 11 (Jaccard 0.0131). The TRB3-only set is largely T-cell receptor beta locus nomenclature - which in a lymphoid leukaemia project is on-topic vocabulary, making this the most dangerous collision in the beat. Fusion-scoped TRB3 returns a valid 0.",
"NIPK": "11 records, 1 shared",
"SKIP3": "7 records, 1 shared",
"uniprot": "clean - exactly one reviewed human record"
}- TRIB3 is the ISR's own brake, and a drug that removes the brake along with the output has an unpredictable net effect on stress-induced death - losing TRIB3 DECREASES ER-stress-dependent cell death in the published knockdown, which is the wrong direction for a therapy.
- TRIB3's DepMap gene effect is the most positive in this panel (+0.094), so if anything its loss favours growth. Any narrative that counts TRIB3 repression as part of the drug's benefit is arguing against the only functional genomics available.
- TRIB3 is the one concordant gene that also moves in the KMT2A-wild-type line, so it is the member most likely to be reporting general drug stress rather than fusion biology - and it is one of the five genes the graph used to infer the ATF4 signature in the first place.
- Does TRIB3 loss de-repress residual CHOP/ATF4 output in drug-treated RS4;11? A TRIB3 knockout in the presence of VTP-50469 would test the feedback architecture directly and is the cheapest way to tell whether TRIB3 is a participant or a passenger.
- What does PMID 36142452 actually show about TRIB3 and the MLL complex? It is the only fusion-scoped record for this gene, it was identified by title only, and reading it would either produce a real interactome lead or close the question.
- Why is REH's TRIB3 response (-0.68) so much larger than every other wild-type-line response in the 24-gene set, and why is TRIB3 more highly expressed in REH than in either t(4;11) line? If TRIB3 is partly fusion-independent, it should be dropped from the fusion-selective signature.
- Is there a TRIB3 protein in KMT2A-rearranged cells at all, and where? The nuclear call rests on one APOBEC3A paper and no haematopoietic experiment exists.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
TRIB3 falls 1.23 and 1.18 log2 on two independent GPL570 probes under MI-2-2 in RS4;11 and 1.22 log2 in SEM, and 3.10 log2 under VTP-50469 in RS4;11 at day 7 (EPZ5676 -3.07); in the KMT2A-wild-type line REH it is more highly expressed (10.31 vs 8.78) and moves -0.68, the largest REH excursion of any of the 24 concordant genes.
What this evidence cannot show
Two probes and two t(4;11) lines agree, which is the strongest internal consistency in this panel - but the REH negative control is the weakest of the 24, so fusion-selectivity is LESS well supported for TRIB3 than for the rest of the signature and the entry says so. RS4;11 array DMSO arm n=2. Day 2 gives only -0.40 against day 7's -3.10, so this is a late effect.
Evidence that argues against this (1)
- The same collapse is produced almost exactly by EPZ5676, a potent DOT1L inhibitor, in the same cells at the same timepoint - so the effect is not a readout of the menin-KMT2A interface. It marks disruption of the fusion's transcriptional output by either route, and a menin-specific interpretation of it is unsupported. NCBI GEO, re-analysed here 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
TRIB3 is induced by ER stress later than CHOP, is a transcriptional target of cooperating ATF4 and CHOP acting at overlapping amino-acid-response elements in its promoter, associates with CHOP to suppress CHOP-dependent transactivation and thereby downregulates its own induction, and TRIB3 knockdown decreases ER-stress-dependent cell death.
TRB3, a novel ER stress-inducible gene, is induced via ATF4-CHOP pathway and is involved in cell death. · Ohoka N, Yoshii S, Hattori T, Onozaki K, et al. · EMBO Journal 2005What this evidence cannot show
Chemical ER stress (tunicamycin) in non-haematopoietic cell culture, with promoter-reporter, dominant-negative and knockdown methods - a mechanism paper, not a disease paper, and nothing in it concerns leukaemia, a fusion or a menin inhibitor. Abstract read; the promoter mapping figures were not. Not previously cited in this graph (checked 2026-09-22).
Evidence that argues against this (1)
- The feedback model predicts that losing the brake should amplify residual ATF4/CHOP output, and that is not what the menin-inhibitor data show - output falls further than the brake. Either the feedback does not operate in this cell at these timescales, or ATF4 protein has fallen enough to make the brake irrelevant, and no ATF4 protein measurement exists to distinguish them. The architecture is therefore cited as a model the data do not fully fit. NCBI GEO, re-analysed here 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
In DepMap 26.06, TRIB3 has a mean CRISPR gene effect of +0.0937 across 1,258 screens with 0.0% of lines below -0.5 and a leukaemia mean of +0.0908 - the most positive gene effect in this beat's panel.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL depMapEssentiality field; dataVersion year 26 month 06, read back from meta{dataVersion} in the same session · DepMap / Open Targets Platform 2026What this evidence cannot show
Release recorded per rule 83; screen count is gene-specific. A positive mean gene effect of +0.09 is small and within the noise band of a genome-wide screen; the defensible statement is 'not a dependency and if anything mildly growth-suppressive', not 'a tumour suppressor'. Leukaemia screens by substring match, ids persisted per rule 90.
Evidence that argues against this (1)
- Like CHAC1, TRIB3 is a stress-induced gene and DepMap screens are run without the stress that induces it, so a null in a rich-medium screen is partly a statement about the assay. TRIB3 is also reported as highly expressed in primary lung, colon and breast tumours and in bone marrow, which is hard to reconcile with a gene whose loss is growth-neutral everywhere. UniProt Knowledgebase 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PubMed returns exactly one record for TRIB3 jointly with any of eight KMT2A-fusion operands - PMID 36142452, on TRIB3 interfering with the MLL complex in breast cancer cells - and zero for TRIB3 with a thirteen-name menin-inhibitor list; the alias TRB3 reaches 288 records of which only 11 also mention TRIB3 (Jaccard 0.0131), and fusion-scoped TRB3 returns a valid zero.
NCBI PubMed esearch, run 2026-09-22, with querytranslation and warninglist read back from the response for every query quoted · NCBI PubMed 2026What this evidence cannot show
The one on-topic record was identified by TITLE ONLY - the abstract was not obtained and the paper was not read - so it is a lead and not evidence, and nothing about its methods, cell lines or the nature of the 'interference' is asserted here. Rule 3 satisfied on the zeros (non-empty querytranslation plus 'No items found.'). Rule 94 form; rule 95 observed. Rule 80: titles and abstracts only.
Evidence that argues against this (1)
- A single title-level hit in a breast cancer paper is extremely weak support for any KMT2A-complex relationship, and the far more likely explanation of the match is that MLL appears in that paper for an unrelated reason. Rule 77 cuts both ways: scoping to the disease can hide the answer, and unscoping it can manufacture one. NCBI PubMed 2026
negative feedback: TRIB3 is induced via ATF4-CHOP and then represses CHOP/ATF4 transactivation and its own induction (PMID 15775988)
ATF4 and CHOP cooperate at overlapping amino-acid-response elements in the TRIB3 promoter (PMID 15775988)
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
-1.23/-1.18 on two array probes in RS4;11 and -1.22 in SEM under MI-2-2, -3.10 under VTP-50469; but REH also moves -0.68, the weakest wild-type control of the 24
- related to →KMT2A / MLL1 — the gene that breaks
UNREAD LEAD: PMID 36142452 reports TRIB3 interfering with the MLL complex in breast cancer cells - title-level only, recorded as a lead not as evidence
CHOP and ATF4 cooperate at overlapping amino-acid-response elements in the TRIB3 promoter (PMID 15775988)