Oncofusion

start / graph / gene-cpt1a

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q96-sarcoma-mbm-partners

CPT1A - the outer-mitochondrial-membrane fatty-acid gatekeeper that is also a lysine succinyltransferase, fused to KMT2A exon 3 in the one case that is tether WITHOUT address

also called CPT1A · CPT1 · CPT1-L · carnitine palmitoyltransferase 1A · carnitine O-palmitoyltransferase 1 liver isoform · P50416

CPT1A sits on the outside of the mitochondrion and decides which fats get in to be burned for energy; it also, surprisingly, tags other proteins with a chemical group called succinyl. In one patient's malignant peripheral nerve sheath tumour KMT2A broke very early, at exon 3, keeping the piece that grips menin but losing the piece that finds CpG islands, and joined the back quarter of CPT1A.

The same thing, for a scientist

CPT1A (P50416, 773 aa) is the rate-controlling enzyme of long-chain fatty-acid oxidation: it transfers the acyl group of long-chain acyl-CoA onto carnitine so the fatty acid can enter the mitochondrion, and it is the target of malonyl-CoA inhibition, the switch that couples fat oxidation to fat synthesis [E1]. It carries **two transmembrane helices (48-73, 103-122)**, a catalytic **His473** proton acceptor and acyl/carnitine binding sites at 555-567, 589 and 602; a conformational switch in the **first 42 residues** sets malonyl-CoA sensitivity [E1, E2]. **It is also a lysine succinyltransferase, and that is the epigenetics answer for this gene.** CPT1A succinylates lysines in vivo and in vitro - 101 proteins gained succinylation in CPT1A-expressing cells, with **enolase 1 (ENO1)** among the most heavily modified and its enzymatic activity reduced - and the **G710E** substitution selectively inactivates the carnitine-acyltransferase activity while leaving the succinyltransferase activity intact [E3]. Lysine succinylation is a mark class that also occurs on histones, but **no histone substrate of CPT1A has been reported**: `CPT1A[tiab] AND "histone succinylation"[tiab]` returns **0** [E3]. So this is the `gene-igf2bp3` case - a real acyl-transfer activity on a **non-histone** substrate, which must not be filed next to the chromatin writers. **In case S12 the fusion is `KMT2A exon 3 :: CPT1A exon 15`** [E4]. KMT2A supplies only aa 1-1052: **MBM (6-25), IBM1, IBM2 and all three AT-hooks, but NOT the CXXC domain**, which lies at 1147-1195 in exons 5-6 [E5]. This is the graph's **tether without address** case - the mirror of YAP1-KMT2A-YAP1, which is address without tether - and the only case in this cohort that separates the two functions. CPT1A contributes aa 581-773 in frame: the C-terminal quarter, **without either transmembrane helix, without the malonyl-CoA-responsive N-terminus and without the catalytic His473** [E6]. Whatever that fragment is, it is not a functional acyltransferase and it is not membrane-anchored.

What it looks like
CPT1A773 aa
TranTransmembrane helix 1 · residues 48–73Transmembrane helix 2 · residues 103–122Active site (His473, proton acceptor) · residues 473–473Acyl/carnitine binding site · residues 555–567MaloMalonyl-CoA-responsive N-terminal switch · residues 1–420200400600

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.

Go deeper
What CPT1A normally does

CPT1A is the gatekeeper of mitochondrial long-chain fatty-acid oxidation. Anchored in the outer mitochondrial membrane by two transmembrane helices, it converts long-chain acyl-CoA plus carnitine to acylcarnitine plus CoA, the committed step for beta-oxidation 1. Its regulation is the textbook reciprocal switch: malonyl-CoA, the first intermediate of fatty-acid synthesis, inhibits it, and the sensitivity is set by an environment-dependent structural switch in the first 42 residues 2.

Beyond that it (i) recruits ER-localised ZDHHC4 to palmitoylate MAVS at Cys79, stabilising MAVS and sustaining a type-I interferon response 7; (ii) is reported to matter for inducible regulatory T-cell differentiation and hepatic triglyceride metabolism, both ECO:0000250 by similarity 1; and (iii) is the gene of CPT1A deficiency, a recessive disorder of hypoketotic hypoglycaemia with risk of sudden death 18.

A UniProt detail worth not copying blindly: the recommended name is carnitine O-palmitoyltransferase 1, LIVER isoform, and the tissue-specificity comment says strong expression in kidney and heart, and lower in liver and skeletal muscle - which contradicts the name, and carries no evidence code at all 1. This corpus already knows that uncoded UniProt statements are not measurements; here the uncoded statement and the protein's own name disagree.

What the fusion joins - tether without address, on both sides
sidegeneexonsresiduesmodules supplied
5-primeKMT2A1-3aa 1-1052MBM 6-25, IBM1 123-134, IBM2 147-152, AT-hooks 169-309. NO CXXC
3-primeCPT1A15-19aa 581-773 (predicted, in frame)no transmembrane helix, no catalytic His473, no malonyl-CoA-responsive N-terminus; binding sites 589 and 602 retained, 555-567 lost

This is the case the king log singled out. KMT2A exon 3 ends at aa 1052, and CXXC sits at 1147-1195 in exons 5-6, so case S12 keeps the menin tether and loses the CpG address 5. The recurrent YAP1-KMT2A-YAP1 arm does the opposite - it inserts KMT2A exons 4-5/6, which is address without tether. The pair separates the two functions within one cohort, which no sarcoma arm examined before cycle 52 did.

And the partner is truncated in the same spirit. CPT1A exon 15 codes aa 581-625 and begins exactly 1,740 nt after the ATG, so the junction is predicted in frame and the chimera would carry CPT1A aa 581-773 6. That fragment has lost both transmembrane helices (48-73, 103-122), so it cannot be anchored in the outer mitochondrial membrane; it has lost the first 42 residues, so it has no malonyl-CoA switch; and it has lost the catalytic His473 proton acceptor, so it cannot run the carnitine acyltransferase reaction. It does retain Gly710 - the residue whose substitution to Glu abolishes the acyltransferase activity while sparing the succinyltransferase activity 3 - but the structural basis of the succinyltransferase activity has never been mapped, so whether the retained fragment could succinylate anything is an open question and not a claim.

Epigenetics (PI dimension 3) - a real acyltransferase, and its substrate is not a histone

CPT1A writes an acyl-lysine mark. It is not a histone mark, and the distinction is the whole point of this section. CPT1A has lysine succinyltransferase activity in cells and in vitro; SILAC proteomics found 101 proteins more succinylated in CPT1A-expressing cells; enolase 1 was among the most heavily succinylated and its enzymatic activity fell; and the G710E mutant separates the two activities, inactivating the acyltransferase but not the succinyltransferase 3.

Lysine succinylation is a histone mark class in general. But no histone substrate of CPT1A has been reported: CPT1A[tiab] AND "histone succinylation"[tiab] = 0, measured 2026-09-22 3. CPT1A has no SET, JmjC, HAT, HDAC, bromodomain, chromodomain, Tudor, PHD or CXXC module and is an outer-mitochondrial-membrane protein. So the correct filing is the gene-igf2bp3 pattern: a genuine writer of a lysine acyl mark on a non-histone substrate, which must not sit next to the chromatin writers in this graph, because a reader skimming "CPT1A is a lysine succinyltransferase" beside SET and JmjC entries would read it as a chromatin protein.

Inbound, and there are two independent routes, both in cancer:

  1. A histone methyltransferase sits upstream of CPT1A in leukaemia. SMYD3 is overexpressed in CML leukaemia stem cells and stimulates fatty-acid beta-oxidation by activating the FABP5/PPARD/CPT1A axis in a methyltransferase-activity-dependent manner; blocking CPT1A-mediated FAO reduced human CML LSC function in vitro and depleted LSCs in vivo 9. This is CML, not KMT2A-rearranged leukaemia - but it is a histone lysine methyltransferase controlling this gene in leukaemia stem cells.
  2. CPT1A is induced by epigenetic drugs and that induction is the effect. CPT1A induction following epigenetic perturbation sustains a double-stranded-RNA-activated type-I interferon response through ZDHHC4-mediated MAVS Cys79 palmitoylation, and CPT1A chemical INDUCERS augment combined epigenetic therapy plus PD-1 blockade in refractory tumours 7. Note the direction: more CPT1A is the therapeutic goal there, which is the opposite of the leukaemia-stem-cell result.

Bounded 2026-09-22: CPT1A[tiab] AND histone[tiab] = 51 and CPT1A[tiab] AND succinylation[tiab] = 34 - this is the one gene of the seven with a large epigenetics-adjacent literature (CPT1A[tiab] alone = 1,939 records, more than twenty times any other partner here), and those counts were not exhaustively title-checked. They are reported as a floor on the reading queue, not as a result.

Localization, cell type and cell state (PI dimensions 1 and 2)

Subcellular: outer mitochondrial membrane - and the coded citations are a rule-91 failure. UniProt gives Mitochondrion outer membrane with two ECO:0000269s, PMIDs 11350182 and 14517221. Neither is a localisation study. The first is a founder-mutation and enzymology paper on a Hutterite kindred whose relevant sentence is a negative control - the G710E mutation "alters neither mitochondrial targeting nor stability" 8. The second analyses five missense mutations against a homology model built from mouse carnitine acetyltransferase and reports no localisation experiment at all 10. The topology line, multi-pass membrane protein, is ECO:0000255 - a sequence-analysis rule, not an observation.

The outer-membrane residency of CPT1 is not in doubt; it is classical biochemistry. But the two papers UniProt files under an experimental evidence code are about enzyme kinetics and about a structural model, which is precisely the failure mode this graph records for MEN1, KMT2A, AFF1 and AFF4. The founding observation is not cited.

Subnuclear: not applicable, and for once that is the right answer - CPT1A is a mitochondrial outer-membrane protein with no reported nuclear pool. That is worth stating because it is the sharpest contrast in this cohort: case S12 fuses a nuclear-addressed KMT2A N-terminus to a fragment of a mitochondrial enzyme.

Cell type: contested inside the database. The recommended name says liver isoform; the tissue comment says strong in kidney and heart, lower in liver and skeletal muscle, and carries no evidence code 1. Cell state: fatty-acid-oxidising. Expression is up-regulated by fatty acids through an intronic PPAR-binding sequence 11, and the two best-characterised states in cancer are the CML leukaemia stem cell (SMYD3-driven FAO) 9 and epigenetic-drug-treated tumour cells mounting an interferon response 7. Nothing in CPT1A's literature is mesenchymal: CPT1A[tiab] AND sarcoma[tiab] returns 1 record and it is an ALS skeletal-muscle lipid-metabolism paper.

Analogous biology in other fusion families (PI dimension 5) - and a token trap

CPT1A is not a fusion partner in any tumour, and the query that looks like it says otherwise is a trap worth recording. CPT1A[tiab] AND fusion[tiab] returns 19 records - a healthy-looking count. All nineteen are about membrane or organelle fusion, not gene fusion: mitochondrial fusion and fission, MFN2, precursor-osteoclast fusion in rheumatoid arthritis, mitochondrial fusion in glioblastoma stem cells, Drosophila cachexia. CPT1A[tiab] AND "gene fusion"[tiab] returns 0, and CPT1A[tiab] AND translocation[tiab] returns 29 with no rearrangement among the titles - translocation in this literature means protein translocation across a membrane 12.

So for a metabolic or membrane protein the words fusion and translocation are contaminated tokens in exactly the way SEPT*, condensate and bivalent are in this project's collision registry. A beat that counted 19 and 29 and concluded CPT1A has a rearrangement literature would have been confidently wrong, and the count is the kind that does not invite checking (rule: read querytranslation on positive counts too). The same trap fires harder for TMPRSS13 in this beat, where all nine fusion hits are about cleaving viral fusion proteins.

The closest real analogy is not a fusion at all: TFE3-rearranged renal cell carcinoma, where TFE3 fusion proteins drive progression via PGC-1alpha-mediated fatty-acid oxidation, appears in the same query set 12. That is a fusion oncoprotein reprogramming FAO, with CPT1A downstream - not CPT1A as a partner.

Status: what this case is and is not
  • One patient. Case S12, malignant peripheral nerve sheath tumour - the only case in this cohort that loses CXXC while keeping the MBM 45.
  • The partner fragment cannot be a functional CPT1A: no membrane anchor, no malonyl-CoA switch, no catalytic His473 6.
  • No fusion protein, no transcript sequence, no reciprocal product, no functional assay.
  • DepMap: nothing. Difference +0.008, arm means -0.116 and -0.124, pan-essentiality -0.165 over 1,208 lines, selectivity rank 10,198 of 17,728 13. Leukaemia lines and the wild-type locus.
  • Not a treatment rationale. See risk-a-menin-inhibitor-prescribed-off-a-kmt2a-gene-name-in-a-solid-tumour.
Facts and figures
uniprotP50416
gene symbolCPT1A
length AA773
accession resolutiongene_exact:CPT1A AND organism_id:9606 AND reviewed:true returned exactly one row, primary gene name CPT1A, entry CPT1A_HUMAN, length 773 aa, synonym CPT1. Primary name read back and length asserted. Paralogue hazard: CPT1B and CPT1C are close family members and much CPT1 pharmacology (etomoxir) is not paralogue-selective, so a CPT1 result is not automatically a CPT1A result.
refseq transcriptNM_001876.4 (19 exons, CDS 156..2477 beginning in exon 2, 773 aa; exon 15 = nt 1896-2030 = aa 581-625)
epigenetic readoutWRITES A LYSINE ACYL MARK, AND ITS SUBSTRATE IS NOT A HISTONE - this is the gene-igf2bp3 case and the distinction is the point. CPT1A has LYSINE SUCCINYLTRANSFERASE activity in vivo and in vitro: 101 proteins gained succinylation in CPT1A-expressing cells, ENOLASE 1 was among the most heavily succinylated and its activity fell, and the G710E substitution selectively inactivates the carnitine-acyltransferase activity while sparing the succinyltransferase activity (PMID 29425493). Lysine succinylation IS a histone mark class in general, but NO HISTONE SUBSTRATE OF CPT1A HAS BEEN REPORTED: CPT1A[tiab] AND 'histone succinylation'[tiab] = 0, measured 2026-09-22, against CPT1A[tiab] AND succinylation[tiab] = 34 through the same shape. CPT1A has no SET, JmjC, HAT, HDAC, bromodomain, chromodomain, Tudor, PHD or CXXC module and is an outer-mitochondrial-membrane protein, so an entry filing it next to the chromatin writers would be read as a chromatin protein by anyone skimming. INBOUND, TWO INDEPENDENT ROUTES, BOTH IN CANCER AND POINTING IN OPPOSITE THERAPEUTIC DIRECTIONS: (1) the histone methyltransferase SMYD3 is overexpressed in CML leukaemia stem cells and stimulates fatty-acid beta-oxidation by activating the FABP5/PPARD/CPT1A axis in a METHYLTRANSFERASE-ACTIVITY-DEPENDENT manner, and blocking CPT1A-mediated FAO depletes CML LSCs in vivo (PMID 40080525 - CML, not KMT2A-rearranged); (2) CPT1A INDUCTION following epigenetic-drug perturbation sustains a dsRNA-activated type-I interferon response via ZDHHC4-mediated MAVS Cys79 palmitoylation, and CPT1A chemical INDUCERS augment combined epigenetic therapy plus PD-1 blockade (PMID 38016475). NOTE: route 1 says block CPT1A and route 2 says induce it, so 'target CPT1A' has no determinate sign. BOUNDED 2026-09-22: CPT1A[tiab] AND histone[tiab] = 51 and AND succinylation[tiab] = 34 - this is the ONE gene of the seven with a large epigenetics-adjacent literature (CPT1A[tiab] alone = 1,939 records, more than twenty times any other partner here) and those counts were NOT exhaustively title-checked, so they are a reading-queue floor rather than a result. AND THE DECISIVE QUALIFICATION FOR THIS BEAT: the S12 fusion retains a CPT1A fragment with no catalytic His473, so none of this is expected to reach the chimera.
mark writtenN6-succinyl-L-lysine, on NON-HISTONE substrates (ENO1 named). No histone substrate reported. CPT1A is not a chromatin writer.
non histone substrateENO1 (enolase 1) - succinylated by CPT1A with a measured reduction in enolase enzymatic activity, from a 101-protein SILAC succinylome. Separately CPT1A recruits ZDHHC4 to palmitoylate MAVS Cys79, which is a non-histone acylation it enables rather than catalyses.
subnuclear localizationNOT APPLICABLE, and for once that is the correct answer rather than a gap: CPT1A is an outer-mitochondrial-membrane protein with two transmembrane helices and no reported nuclear pool of any kind. It is worth stating because it is the sharpest contrast in this cohort - case S12 fuses a NUCLEAR-ADDRESSED KMT2A N-terminus (AT-hooks, and in the other six cases a CXXC CpG-island reader) to a fragment of a MITOCHONDRIAL enzyme, and the predicted fragment has lost both membrane anchors, so where such a chimera would go is unknown and unmeasured.
cell typeCONTESTED INSIDE THE DATABASE. The UniProt recommended name is 'carnitine O-palmitoyltransferase 1, LIVER isoform' while the tissue-specificity comment says 'strong expression in kidney and heart, and lower in liver and skeletal muscle' - and that comment carries NO EVIDENCE CODE. This corpus already records that uncoded UniProt statements are not measurements; here the uncoded statement and the protein's own name disagree, which is worth more than either alone. Nothing in CPT1A's literature is mesenchymal: CPT1A[tiab] AND sarcoma[tiab] = 1 and it is an ALS skeletal-muscle lipid-metabolism paper.
cell stateFATTY-ACID-OXIDISING. Expression is induced by fatty acids through an intronic PPAR-binding sequence (PMID 16271724), and the two best-characterised cancer states are the CML LEUKAEMIA STEM CELL, where SMYD3-driven FAO through FABP5/PPARD/CPT1A supports self-renewal (PMID 40080525), and the EPIGENETIC-DRUG-TREATED TUMOUR CELL, where CPT1A induction sustains a type-I interferon response (PMID 38016475). Also reported, by similarity, in inducible regulatory T-cell differentiation.
localization shown for fusionNO - 'nobody has looked in any fusion' (rule 86). CPT1A[tiab] AND KMT2A[tiab] = 0; CPT1A[tiab] AND MLL[tiab] = 2 and both are off topic (PMID 39703843, an ARMH1 paediatric-AML paper where CPT1A is metabolic, and PMID 38159113 on essential oils in geese). No sarcoma KMT2A fusion protein has been visualised by any method.
fusion partner in other tumoursNONE REPORTED - AND THE QUERY IS A TOKEN TRAP WORTH A REGISTRY ROW. CPT1A[tiab] AND fusion[tiab] returns 19, which looks like a literature; ALL NINETEEN are about mitochondrial, membrane or cell fusion (osteoclast fusion in rheumatoid arthritis, mitochondrial fusion in glioblastoma stem cells, MFN2, mitochondrial fusion/fission in HCC, a Drosophila cachexia model). CPT1A[tiab] AND 'gene fusion'[tiab] = 0. CPT1A[tiab] AND translocation[tiab] = 29 with no rearrangement among the titles, because in this literature 'translocation' means protein translocation across a membrane. For a metabolic or membrane protein, 'fusion' and 'translocation' are contaminated tokens in the same family as SEPT*, condensate and bivalent - and the count looks healthy rather than suspicious, which is what makes it dangerous. The nearest real analogy runs the other way: TFE3-rearranged renal cell carcinoma, where the FUSION protein drives progression via PGC-1alpha-mediated fatty-acid oxidation with CPT1A downstream (PMID 41716419, in the same query set).
absence classnever-measured
domains
Malonyl-CoA-responsive N-…1.000Malonyl-CoA-responsive N-terminal switch: 1Transmembrane helix 148.0Transmembrane helix 1: 48Transmembrane helix 2103.0Transmembrane helix 2: 103Active site (His473, prot…473.0Active site (His473, proton acceptor): 473Acyl/carnitine binding si…555.0Acyl/carnitine binding site: 555
aliases and retired symbols searched
[
  "CPT1A",
  "CPT1",
  "CPT1-L",
  "carnitine palmitoyltransferase 1A"
]
domains — the numbers behind the chart
[
  {
    "name": "Transmembrane helix 1",
    "start": 48,
    "end": 73
  },
  {
    "name": "Transmembrane helix 2",
    "start": 103,
    "end": 122
  },
  {
    "name": "Active site (His473, proton acceptor)",
    "start": 473,
    "end": 473
  },
  {
    "name": "Acyl/carnitine binding site",
    "start": 555,
    "end": 567
  },
  {
    "name": "Malonyl-CoA-responsive N-terminal switch",
    "start": 1,
    "end": 42
  }
]
functions
[
  "carnitine O-palmitoyltransferase - committed step of mitochondrial long-chain fatty-acid oxidation",
  "malonyl-CoA-inhibited gatekeeper coupling fat oxidation to fat synthesis",
  "lysine succinyltransferase acting on non-histone substrates such as ENO1",
  "recruits ZDHHC4 to palmitoylate MAVS Cys79, stabilising MAVS and sustaining a type-I interferon response",
  "hepatic triglyceride metabolism (by similarity)",
  "inducible regulatory T-cell differentiation (by similarity)"
]
sarcoma case
{
  "patient": "S12",
  "diagnosis": "malignant peripheral nerve sheath tumour",
  "fusion": "KMT2A exon 3 :: CPT1A exon 15 (NM_001876)",
  "kmt2a_polarity": "5-prime - the LEUKAEMIC orientation, opposite to the recurrent sarcoma arms",
  "kmt2a_exons_retained": "1-3",
  "kmt2a_residues_retained": "aa 1-1052",
  "kmt2a_modules_retained": [
    "MBM 6-25",
    "IBM1 123-134",
    "IBM2 147-152",
    "AT-hooks 169-309"
  ],
  "kmt2a_modules_lost": [
    "CXXC 1147-1195 (exons 5-6)"
  ],
  "mbm_retained": true,
  "cxxc_retained": false,
  "in_leukaemic_breakpoint_cluster_region": false,
  "why_this_case_matters": "TETHER WITHOUT ADDRESS. This is the only case in the cohort that keeps the menin tether and loses the CpG-island address. The recurrent YAP1-KMT2A-YAP1 arm is the opposite - address without tether - so the pair separates the two functions within ONE cohort, which no sarcoma arm examined before cycle 52 did.",
  "partner_residues_predicted": "aa 581-773 - the C-terminal quarter, 193 residues",
  "partner_modules_predicted_retained": [
    "acyl/carnitine binding residues 589 and 602",
    "Gly710"
  ],
  "partner_modules_lost": [
    "transmembrane helix 48-73",
    "transmembrane helix 103-122",
    "malonyl-CoA-responsive N-terminal 42 residues",
    "catalytic His473 proton acceptor",
    "binding site 555-567"
  ],
  "predicted_frame": "IN FRAME (CPT1A exon 15 starts 1,740 nt after the ATG, 1740 % 3 == 0; KMT2A exon 3 ends at a complete codon)",
  "frame_derivation": "computed 2026-09-22 from NM_001197104.2 and NM_001876.4 exon coordinates - a derivation, not an observation",
  "functional_consequence_for_the_partner": "the retained fragment cannot run the carnitine acyltransferase reaction (no His473), cannot be anchored in the outer mitochondrial membrane (no transmembrane helix) and has no malonyl-CoA switch. It does retain Gly710, whose substitution separates the acyltransferase from the succinyltransferase activity, but the structural basis of the succinyltransferase activity has never been mapped, so whether the fragment could succinylate anything is an OPEN QUESTION and not a claim.",
  "source": "Supplemental Table 2 of PMID 32461620",
  "fusion_protein_observed": false,
  "reciprocal_transcript_reported": false,
  "functional_assay_done": false,
  "authors_own_verdict": "of unclear significance"
}
localization
{
  "CONSTITUTIVE": "MITOCHONDRION OUTER MEMBRANE. RULE 91 FAILS ON BOTH CODED CITATIONS: UniProt gives two ECO:0000269s, PMID 11350182 and PMID 14517221, and neither is a localisation study. The first is a Hutterite founder-mutation and enzymology paper whose only relevant sentence is a NEGATIVE CONTROL - G710E 'alters neither mitochondrial targeting nor stability'. The second maps five missense mutations onto a HOMOLOGY MODEL built from mouse carnitine acetyltransferase and reports no localisation experiment at all. The topology line, 'multi-pass membrane protein', is ECO:0000255, a sequence-analysis rule. The outer-membrane residency of CPT1 is classical biochemistry and is almost certainly correct - the defect is in the database's pointers, and the founding observation is not cited.",
  "systems": "patient fibroblasts, heterologous expression in Saccharomyces cerevisiae, and a homology model. No haematopoietic and no mesenchymal or sarcoma cell."
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.1162,
  "chronos_mean_other_leukaemia": -0.1243,
  "difference": 0.0081,
  "pan_essentiality_all_lines": -0.165,
  "n_all_lines": 1208,
  "selectivity_rank": 10198,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-22 from analyses/depmap-kmt2a-selective-scan-full/results.json (generated 2026-09-19T10:12:06Z), this project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies, including that the 11 KMT2A-rearranged lines are drawn from nine individuals and that the arm is myeloid-enriched. The scan records a generation timestamp and NOT a DepMap RELEASE, which is rule 83 exposure carried forward rather than fixed here.",
  "how_to_read": "A rank is not an effect size and a difference is not a dependency (this graph's DepMap rules). Check pan_essentiality_all_lines before reading anything as selective, and note the anti-selective half is 50% of any gene set by construction, with this graph's own base rate at 45.3% - so a tally of genes in that half is the null (rule 84). The decisive limitation here is lineage: the screen is leukaemia lines, and for a sarcoma fusion that is the wrong tissue. Cellosaurus holds 592 KMT2A-commented cell lines and ZERO are sarcoma or any mesenchymal tumour, so no cell-line screen can address these cases at all. And a specific limitation for this gene: fatty-acid-oxidation dependency is a metabolic-context phenotype, and DepMap lines are grown in fixed rich medium, so this screen cannot detect a dependency that appears only under nutrient limitation - which is the condition under which the CML leukaemia-stem-cell result was obtained."
}
queries run 2026 09 22
[
  "CPT1A[tiab] -> 1939 (the largest literature of the seven by more than twentyfold)",
  "CPT1A[tiab] AND fusion[tiab] -> 19, ALL title-checked and ALL off topic (mitochondrial/membrane/cell fusion)",
  "CPT1A[tiab] AND 'gene fusion'[tiab] -> 0",
  "CPT1A[tiab] AND translocation[tiab] -> 29, no rearrangement in the titles (protein translocation)",
  "CPT1A[tiab] AND KMT2A[tiab] -> 0",
  "CPT1A[tiab] AND MLL[tiab] -> 2 (PMID 39703843 ARMH1/AML, CPT1A metabolic; PMID 38159113 geese essential oils)",
  "CPT1A[tiab] AND histone[tiab] -> 51 (not exhaustively title-checked)",
  "CPT1A[tiab] AND succinylation[tiab] -> 34 (not exhaustively title-checked)",
  "CPT1A[tiab] AND 'histone succinylation'[tiab] -> 0",
  "CPT1A[tiab] AND leukemia[tiab] -> 20",
  "CPT1A[tiab] AND sarcoma[tiab] -> 1 (ALS muscle paper, off topic)",
  "controls: EWSR1[tiab] AND fusion[tiab] -> 1208; EWSR1[tiab] AND translocation[tiab] -> 462; AFF1[tiab] AND KMT2A[tiab] -> 104; MLLT3[tiab] AND MLL[tiab] -> 84; KMT2A[tiab] AND histone[tiab] -> 218"
]
graph neighbours already present
[
  "gene-kmt2a",
  "gene-igf2bp3",
  "disease-kmt2a-r-sarcoma",
  "concept-kmt2a-fusion-polarity-is-inverted-in-sarcoma"
]
How this could bite us back
  • This entry must not be read as a menin-inhibitor rationale. Menin is a TUMOUR SUPPRESSOR in mesenchyme, and the only non-haematologic menin trial with a result, NCT05731947, was terminated by the sponsor after phase 1b for not meeting pre-defined efficacy signals. An MBM-retaining KMT2A fusion in a sarcoma is a test of a hypothesis, not an indication.
  • If menin inhibition were tried here off the gene name alone, the predicted direction of effect is not obviously anti-tumour: displacing menin from chromatin in a mesenchymal cell removes a tumour suppressor from its targets. See risk-a-menin-inhibitor-prescribed-off-a-kmt2a-gene-name-in-a-solid-tumour.
  • Every quantitative statement about the fusion in this entry is a coordinate derivation from transcript records. If the panel's exon numbering differs from RefSeq's, the frame prediction, the retained-residue count and the MBM claim all move together - one systematic error, not several independent ones.
  • There is no coherent CPT1A intervention to propose here, and the reason is instructive rather than incidental: the two cancer results point in opposite directions. Blocking CPT1A-mediated fatty-acid oxidation depletes CML leukaemia stem cells; inducing CPT1A potentiates antitumour immunity under epigenetic therapy. An agent chosen off the gene name could do either.
  • CPT1A loss of function in humans causes hypoketotic hypoglycaemia with a risk of sudden death and neurological sequelae, so systemic inhibition has a known severe phenotype - and the classical CPT1 inhibitor etomoxir is not paralogue-selective and has documented hepatotoxicity.
  • The fusion is predicted to encode a catalytically dead CPT1A fragment with no membrane anchor, so even if CPT1A biology mattered in this tumour, the chimera is not the route to it - a therapeutic argument built on the partner's normal function would be targeting a protein the fusion does not contain.
  • Case S12 loses the CXXC domain, so any reasoning that imports the leukaemic mechanism wholesale is wrong for this case in particular: the chimera has the menin tether without the CpG-island address, a combination for which this graph has no measured phenotype at all.
Still open
  1. Could the retained CPT1A aa 581-773 fragment succinylate anything? It keeps Gly710, the residue whose substitution separates the succinyltransferase from the acyltransferase activity, but loses the catalytic His473 - and the structural basis of the succinyltransferase activity has never been mapped, so this is answerable only by mapping that active site.
  2. Does the tether-without-address chimera bind menin, and does it go anywhere? Case S12 is the graph's cleanest natural separation of the menin tether from the CpG address, and the retained CPT1A fragment has lost both membrane anchors, so the chimera's localisation is entirely unpredicted.
  3. Is there any histone substrate for CPT1A's succinyltransferase activity? The specific query returns zero, and lysine succinylation is a documented histone mark class, so the question is whether nobody has looked or whether it has been looked for and not found - a difference the literature cannot currently resolve.
  4. Does the SMYD3-FABP5-PPARD-CPT1A axis operate in KMT2A-rearranged leukaemia as it does in CML? SMYD3 has no entry in this graph, and a histone methyltransferase controlling fatty-acid oxidation in leukaemia stem cells is directly on this project's beat.
  5. Which direction would a CPT1A agent go in a KMT2A-rearranged tumour? The two cancer results point opposite ways - block CPT1A to deplete CML leukaemia stem cells, induce CPT1A to potentiate antitumour immunity - and nothing resolves the sign for this disease.
Every source behind this entry (13)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    CPT1A (P50416, 773 aa) catalyses transfer of the acyl group of long-chain acyl-CoA onto carnitine - the committed step for mitochondrial fatty-acid uptake and beta-oxidation - from the outer mitochondrial membrane, where it is held by two transmembrane helices at 48-73 and 103-122, with a catalytic His473 proton acceptor and acyl/carnitine binding sites at 555-567, 589 and 602; UniProt additionally annotates it as Succinyltransferase CPT1A.

    What this evidence cannot show

    Accession resolved this cycle by gene_exact query with gene_primary read back: exactly one reviewed human record, primary symbol CPT1A, entry CPT1A_HUMAN, length 773 aa (rules 62, 65, 66; accession AND length asserted). Two things not to copy blindly. (1) The RECOMMENDED NAME says 'liver isoform' while the TISSUE SPECIFICITY comment says 'strong expression in kidney and heart, and lower in liver and skeletal muscle' - and that comment carries NO EVIDENCE CODE, so the database's own name and its own uncoded statement disagree. (2) The topology annotation 'multi-pass membrane protein' is ECO:0000255, a sequence-analysis rule rather than an observation. Feature boundaries are UniProt calls; there is no experimental structure of human CPT1A, and the structural work cited for its active site is a homology model built from mouse carnitine acetyltransferase.

    Evidence that argues against this (1)
    • CPT1 is a three-paralogue family - CPT1A, CPT1B (muscle) and CPT1C (brain) - and the isoform names encode a tissue attribution that this record's own tissue comment contradicts. Any claim about which CPT1 is acting in a given tissue should rest on a measurement in that tissue rather than on the isoform name, and much of the CPT1 pharmacology (etomoxir) is not paralogue-selective. Rao JN et al. 2011
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Malonyl-CoA sensitivity of CPT1A is set by an environment-dependent structural switch in the N-terminal region spanning the first 42 residues.

    An environment-dependent structural switch underlies the regulation of carnitine palmitoyltransferase 1A. · Rao JN, Warren GZL, Estolt-Povedano S et al · Journal of Biological Chemistry 2011
    PMID 21990363DOI
    What this evidence cannot show

    In-vitro structural and biophysical work on the isolated N-terminal region and on the enzyme in different membrane environments, not a measurement in a tumour cell. Cited here because it identifies WHICH residues carry the regulatory function, and therefore what the S12 fusion loses - the fusion retains aa 581-773 and so contains none of the first 42. Abstract-level reading; this is also the paper UniProt cites for its DOMAIN comment, and rule 91 passes for that annotation.

    Evidence that argues against this (1)
    • A regulatory switch mapped on an isolated N-terminal fragment in a defined membrane environment need not behave the same way in the intact protein in a living mitochondrion, and malonyl-CoA sensitivity of CPT1 is known to vary with membrane lipid composition and physiological state - so 'the first 42 residues set sensitivity' is a mechanism, not a constant. UniProt Consortium 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CPT1A has lysine succinyltransferase activity in vivo and in vitro: 101 proteins were more succinylated in cells expressing wild-type CPT1A, enolase 1 was among the most heavily succinylated and its enzymatic activity was reduced, and the G710E substitution selectively inactivates the carnitine-acyltransferase activity while leaving the succinyltransferase activity intact - and no histone substrate of CPT1A has been reported.

    Carnitine Palmitoyltransferase 1A Has a Lysine Succinyltransferase Activity. · Kurmi K, Hitosugi S, Wiese EK et al · Cell Reports 2018
    PMID 29425493DOI
    What this evidence cannot show

    Human cell lines plus in-vitro assays; not a haematopoietic, mesenchymal or KMT2A-rearranged system, and the in-cell experiments are OVEREXPRESSION of wild-type CPT1A against vector control, which bounds how much of the 101-protein succinylome is physiological. THE CLASSIFICATION POINT: lysine succinylation is a mark class that also occurs on histones, but CPT1A's reported substrate is the glycolytic enzyme ENO1, CPT1A has no chromatin-binding module, and it is an outer-mitochondrial-membrane protein - so this is a NON-HISTONE acyl mark, the gene-igf2bp3 situation. Bounded 2026-09-22: CPT1A[tiab] AND 'histone succinylation'[tiab] = 0, against CPT1A[tiab] AND succinylation[tiab] = 34 and CPT1A[tiab] AND histone[tiab] = 51 through the same shape, so the instrument returns ones and the specific zero is real in titles and abstracts. Those 51 and 34 records were NOT exhaustively title-checked - CPT1A[tiab] alone returns 1,939 - so they are a reading-queue floor, not a result.

    Evidence that argues against this (1)
    • Non-enzymatic lysine succinylation from ambient succinyl-CoA is well documented, and mitochondrial succinyl-CoA concentrations are high, so an overexpression experiment that raises a succinyl-CoA-handling enzyme can increase succinylation without the enzyme acting as a transferase. The G710E separation-of-function result is the strongest argument against that reading, and it is a single-residue argument from one study. UniProt Consortium 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In case S12 of a pan-sarcoma KMT2A-rearrangement series, a malignant peripheral nerve sheath tumour carried KMT2A exon 3 fused to CPT1A exon 15 (NM_001876) - the earliest KMT2A breakpoint among the seven 5-prime cases, and the only one that retains the menin-binding motif while losing the CXXC domain.

    PMID 32461620DOI
    What this evidence cannot show

    This is ONE patient (case S12, malignant peripheral nerve sheath tumour) in a commercial targeted RNA-panel series enriched for advanced disease. The source authors call the non-recurrent KMT2A partners of UNCLEAR SIGNIFICANCE and several may be passengers. No fusion protein has been observed for this or any other sarcoma KMT2A fusion by any method, no reciprocal transcript was reported, no functional experiment has been done, and the panel targets specific KMT2A exons, which biases which breakpoints are detectable at all. Supplemental Table 2 reports EXON NUMBERS from the panel's own annotation - the junction nucleotide, the transcript sequence and the reading frame are not given - and it misspells the index gene as KTM2A in two of eighteen rows, so a reader grepping for KMT2A silently loses cases. This graph's parse is at analyses/sarcoma-nonrecurrent-partners/results.json. Naming this case makes it auditable; it does not promote the partner to a driver and it is not a treatment rationale.

    Evidence that argues against this (1)
    • The cohort's recurrent, morphology- and methylation-defined KMT2A sarcoma arms are YAP1-KMT2A-YAP1 and VIM::KMT2A, and those are what the paper's title and conclusions are about. The seven 5-prime non-recurrent cases are explicitly set aside by the authors as of unclear significance, and a targeted panel run in advanced disease will also detect passenger rearrangements. A single MBM-retaining case is a lead, not an entity. Massoth LR et al. 2020
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    KMT2A exons 1-3 encode aa 1-1052, so this fusion retains the menin-binding motif (6-25), LEDGF IBM1 (123-134) and IBM2 (147-152) and all three AT-hooks but NOT the CXXC domain, which lies at 1147-1195 within exons 5-6 - making case S12 tether without address, the mirror of the address-without-tether YAP1-KMT2A-YAP1 arm.

    What this evidence cannot show

    KMT2A exon-to-residue map, re-derived this cycle from RefSeq NM_001197104.2 (CDS 22..11940, 36 exons, NP_001184033.1) and independently reproducing this graph's own ana-kmt2a-exon-to-domain-map: exon 1 = aa 1-144 (MBM 6-25, LEDGF IBM1 123-134), exon 2 = aa 145-168 (IBM2 147-152), exon 3 = aa 168-1052 (all three AT-hooks), exon 4 = 1053-1112, exons 5-6 = 1112-1212 (CXXC 1147-1195). Exon 3 ends at aa 1052, exon 8 at 1362, exon 9 at 1406, exon 15 at 1668, exon 20 at 1888, exon 22 at 1987. All six KMT2A donor exons in this cohort end at a COMPLETE CODON (phase 0), which is why KMT2A can fuse in frame to so many acceptors. Note NM_001197104.2 encodes 3,972 aa while UniProt Q03164 is 3,969 aa; the three-residue difference lies downstream of every coordinate used here. This is a COORDINATE DERIVATION from a transcript record, not a measurement of a fusion protein: no sarcoma KMT2A fusion protein has ever been observed by any method and no menin-binding assay has been run on one. The exon-5/6 residue span 1112-1212 was reproduced exactly from NM_001197104.2 this cycle, which is what makes the CXXC-loss call for an exon-3 breakpoint safe: 1052 < 1147.

    Evidence that argues against this (1)
    • Retaining a motif is not the same as presenting it. The chimera's folding, stability and nuclear import are all unmeasured, and a fusion protein can retain a binding motif in primary sequence while failing to engage its partner. This graph's queued test exp-menin-pulldown-on-yky-versus-kmt2a-aff1 exists precisely because retention has never been shown to imply binding for any sarcoma arm. Massoth LR et al. 2020
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    CPT1A exon 15 (NM_001876.4, nt 1896-2030) begins exactly 1,740 nucleotides after the ATG and codes aa 581-625, so a junction from a phase-0 KMT2A donor exon is predicted in frame and the chimera would carry CPT1A aa 581-773 - a fragment lacking both transmembrane helices, the malonyl-CoA-responsive first 42 residues and the catalytic His473 proton acceptor, and therefore incapable of the carnitine acyltransferase reaction or of outer-membrane anchoring.

    What this evidence cannot show

    A derivation computed this cycle from two RefSeq records, not an observation: the panel's exon numbering need not match RefSeq's, the junction nucleotide is unknown, and an in-frame prediction says nothing about whether the transcript is made or translated. Note CPT1A's CDS begins in exon 2, so its exon numbering and its codon numbering are offset - a place where a careless derivation would go wrong. The fragment retains Gly710, the residue whose substitution separates the acyltransferase from the succinyltransferase activity, but the STRUCTURAL BASIS OF THE SUCCINYLTRANSFERASE ACTIVITY HAS NEVER BEEN MAPPED, so whether this fragment could succinylate anything is explicitly an open question here and NOT a claim.

    Evidence that argues against this (1)
    • An in-frame prediction is not a demonstration of a product, and the same calculation returns the opposite answer for other members of this cohort: C2CD3 exon 24 and RALGAPA2 exon 2 both begin at phase 1 and are predicted FRAMESHIFTED, and AHDC1 exon 2 is entirely 5'-untranslated. That the derivation discriminates between cases is a reason to trust the method, not any one result. Massoth LR et al. 2020
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CPT1A induction following epigenetic perturbation sustains a double-stranded-RNA-activated type-I interferon response by recruiting ER-localised ZDHHC4 to palmitoylate MAVS at Cys79, which stabilises and activates MAVS, and CPT1A chemical inducers augment the effect of combined epigenetic treatment with PD-1 blockade in refractory tumours.

    PMID 38016475DOI
    What this evidence cannot show

    Mouse and human tumour models with epigenetic-drug treatment and anti-PD-1; not leukaemia, not a KMT2A-rearranged system and not a sarcoma. The 'epigenetic perturbation' is pharmacological treatment with epigenetic-regulator inhibitors rather than a defined mark change, so this is an INBOUND epigenetic link at the level of drug class rather than of a named histone modification. Abstract read; full text not read. Cited here for two reasons: it is the mechanism behind UniProt's ZDHHC4/MAVS function annotation, and it fixes the DIRECTION of therapeutic interest in CPT1A as INDUCTION, which is opposite to the leukaemia-stem-cell result at E9.

    Evidence that argues against this (1)
    • The two cancer results about CPT1A point in opposite therapeutic directions - inducing CPT1A potentiates antitumour immunity here, while blocking CPT1A-mediated fatty-acid oxidation depletes CML leukaemia stem cells - so 'target CPT1A' has no determinate sign. Context decides, and neither context is a KMT2A-rearranged sarcoma. Zhou M et al. 2025
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The CPT1A G710E substitution found homozygously in a Hutterite kindred with CPT1A deficiency abolishes catalytic function while altering neither mitochondrial targeting nor protein stability.

    Molecular and enzymatic characterization of a unique carnitine palmitoyltransferase 1A mutation in the Hutterite community. · Prip-Buus C, Thuillier L, Abadi N et al · Molecular Genetics and Metabolism 2001
    PMID 11350182DOI
    What this evidence cannot show

    Patient fibroblasts plus heterologous expression in yeast - a clinical-genetics and enzymology study in a founder population. RULE 91 FLAG: this is one of the two ECO:0000269 citations UniProt uses for CPT1A's 'Mitochondrion outer membrane' localisation comment, and it is NOT a localisation study. Its only relevant sentence is a NEGATIVE CONTROL - that G710E 'alters neither mitochondrial targeting nor stability' - which establishes that a mutant is targeted normally, not where the wild-type protein resides. The founding outer-membrane observations for CPT1 are not cited on the record. Abstract read; full text not read.

    Evidence that argues against this (1)
    • A negative result about mutant targeting cannot establish the wild-type compartment, and the other coded citation on the same comment is a homology-modelling paper with no localisation experiment either. CPT1's outer-membrane residency is classical biochemistry and is almost certainly correct - the defect is in the database's pointers, not in the fact, which is exactly the rule-91 shape. Gobin S et al. 2003
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The histone methyltransferase SMYD3 is overexpressed in chronic myeloid leukaemia stem cells and stimulates fatty-acid beta-oxidation by activating the FABP5/PPARD/CPT1A axis in a methyltransferase-activity-dependent manner, and blocking CPT1A-mediated fatty-acid oxidation reduced human CML leukaemia-stem-cell function in vitro and depleted leukaemia stem cells in vivo.

    PMID 40080525DOI
    What this evidence cannot show

    In CHRONIC MYELOID LEUKAEMIA - human primary CML CD34+ cells and mouse models - not KMT2A-rearranged leukaemia and not a sarcoma (rule 73: the system is in the claim). The link from SMYD3 to CPT1A runs through FABP5 and PPARD, so CPT1A is two steps downstream and no SMYD3 occupancy at the CPT1A locus is claimed. Abstract read; full text not read. Cited as the clearest example of a histone lysine methyltransferase controlling this gene in a leukaemic stem cell, which is the closest this gene's epigenetics gets to the index disease.

    Evidence that argues against this (1)
    • A methyltransferase-dependent effect two nodes upstream is not evidence that the CPT1A locus carries a histone mark, and the therapeutic direction here (block CPT1A) is the reverse of the antitumour-immunity result (induce CPT1A). Neither bears on the S12 sarcoma, where the fusion retains only a catalytically dead CPT1A fragment. Zhang G et al. 2023
  10. E10moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The functional and structural basis of CPT1A deficiency was analysed by mapping five patient missense mutations onto a three-dimensional homology model of human CPT1A built from the crystal structure of mouse carnitine acetyltransferase, dividing them into functional determinants at the active site and structural determinants more than 20 angstroms away.

    Functional and structural basis of carnitine palmitoyltransferase 1A deficiency. · Gobin S, Thuillier L, Jogl G et al · Journal of Biological Chemistry 2003
    PMID 14517221DOI
    What this evidence cannot show

    A HOMOLOGY MODEL, not an experimental structure of human CPT1A, plus heterologous expression in Saccharomyces cerevisiae to validate the mutations. RULE 91 FLAG: this is the second ECO:0000269 citation behind UniProt's 'Mitochondrion outer membrane' comment and its abstract describes NO localisation experiment whatever. There is still no experimental structure of full-length human CPT1A, which matters for any residue-level claim about this protein - including the active-site His473 assignment this entry relies on.

    Evidence that argues against this (1)
    • Residue-level claims from a homology model built on a different enzyme carry the offset risk this graph documents for PDB-derived numbering: a model can be right about the fold and wrong about an individual residue's role. The His473 proton-acceptor assignment should be treated as annotation rather than as structure. UniProt Consortium 2026
  11. E11moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Fatty acids induce human CPT1A expression through an intronic peroxisome-proliferator-activated-receptor-binding sequence.

    PMID 16271724DOI
    What this evidence cannot show

    Reporter and promoter/intron analysis in cell lines, 2005; this is the paper behind UniProt's INDUCTION comment and rule 91 passes for it. It establishes a cis-regulatory element and a nuclear receptor, not a chromatin mark, so it is a transcriptional rather than an epigenetic result - recorded here to keep the two categories apart. Abstract read; full text not read.

    Evidence that argues against this (1)
    • An intronic PPAR-response element is a regulatory feature of the wild-type CPT1A locus and is irrelevant to the S12 fusion, whose CPT1A portion begins at exon 15 and is transcribed from the KMT2A promoter - so none of CPT1A's own regulation applies to the chimeric allele. Massoth LR et al. 2020
  12. E12moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CPT1A is not a reported gene-fusion partner in any tumour, and the queries that appear to say otherwise are token artefacts: all 19 records for CPT1A with 'fusion' concern mitochondrial, membrane or cell fusion rather than gene fusion, and 'CPT1A AND gene fusion' returns zero.

    PMID 32461620DOI
    What this evidence cannot show

    Bounded 2026-09-22 with one [tiab] per term and querytranslation read back (rules 94, 95): CPT1A[tiab] AND fusion[tiab] = 19, ALL title-checked and all off topic (osteoclast fusion in rheumatoid arthritis, mitochondrial fusion in glioblastoma stem cells, MFN2, mitochondrial fusion/fission in hepatocellular carcinoma, a Drosophila cachexia model, and others); CPT1A[tiab] AND 'gene fusion'[tiab] = 0; CPT1A[tiab] AND translocation[tiab] = 29 with no rearrangement among the titles, because 'translocation' in this literature means protein translocation across a membrane; CPT1A[tiab] AND KMT2A[tiab] = 0; CPT1A[tiab] AND MLL[tiab] = 2, both off topic (PMID 39703843, an ARMH1 paediatric-AML marker paper in which CPT1A is metabolic, and PMID 38159113, a study of essential oils in geese). Positive controls through the IDENTICAL one-tag-per-term query shape, run 2026-09-22 (rule 67, rule 94): EWSR1[tiab] AND fusion[tiab] = 1,208; EWSR1[tiab] AND translocation[tiab] = 462; AFF1[tiab] AND KMT2A[tiab] = 104; MLLT3[tiab] AND MLL[tiab] = 84; KMT2A[tiab] AND histone[tiab] = 218. Every query was written with one [tiab] per term and its querytranslation read back from the esearch response, so no parenthesised group silently expanded to [All Fields]. Bounded by TITLES AND ABSTRACTS, not by papers (rule 80): supporting figures were not searched. IDENTIFIER TRAP WORTH A REGISTRY ROW: for a metabolic or membrane protein, 'fusion' and 'translocation' are contaminated tokens in the same family as SEPT*, condensate and bivalent, and the count looks healthy rather than suspicious.

    Evidence that argues against this (1)
    • A literature absence cannot see an unreported rearrangement, and CPT1A is a 19-exon gene at 11q13 in a region with recurrent copy-number change in several tumours. The supportable claim is bibliographic. The nearest real analogy in the same query set runs the other way round: in TFE3-rearranged renal cell carcinoma the FUSION protein drives progression via PGC-1alpha-mediated fatty-acid oxidation, with CPT1A downstream rather than as a partner. Zhou M et al. 2025
  13. E13weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset

    CPT1A shows no KMT2A-selective dependency in this project's own DepMap scan: mean Chronos -0.116 in 11 KMT2A-rearranged lines against -0.124 in 43 other leukaemia lines, difference +0.008, at a pan-essentiality of -0.165 across 1,208 lines and selectivity rank 10,198 of 17,728.

    DepMap CRISPR (Chronos) gene-effect scan of 11 KMT2A-rearranged versus 43 other leukaemia lines (this project's own scan) · Broad Institute DepMap; joined locally from analyses/depmap-kmt2a-selective-scan-full · analyses/depmap-kmt2a-selective-scan-full 2026
    What this evidence cannot show

    Joined 2026-09-22 from analyses/depmap-kmt2a-selective-scan-full/results.json (generated 2026-09-19T10:12:06Z), this project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies, including that the 11 KMT2A-rearranged lines are drawn from nine individuals and that the arm is myeloid-enriched. The scan records a generation timestamp and NOT a DepMap RELEASE, which is rule 83 exposure carried forward rather than fixed here. A rank is not an effect size and a difference is not a dependency (this graph's DepMap rules). Check pan_essentiality_all_lines before reading anything as selective, and note the anti-selective half is 50% of any gene set by construction, with this graph's own base rate at 45.3% - so a tally of genes in that half is the null (rule 84). The decisive limitation here is lineage: the screen is leukaemia lines, and for a sarcoma fusion that is the wrong tissue. Cellosaurus holds 592 KMT2A-commented cell lines and ZERO are sarcoma or any mesenchymal tumour, so no cell-line screen can address these cases at all.

    Evidence that argues against this (1)
    • A difference of +0.008 is indistinguishable from zero and a mid-table rank is the null (rule 84). More to the point, fatty-acid-oxidation dependency is a metabolic-context phenotype: DepMap lines are grown in a fixed rich medium, so a screen there cannot detect a dependency that only appears under nutrient limitation - which is the condition under which the CML leukaemia-stem-cell result was obtained. Zhou M et al. 2025
Connected to