ADCK2 (aarF domain-containing kinase 2) is a mitochondrial protein of the UbiB family, an ancient group of atypical protein-kinase-like enzymes that also contains COQ8A/ADCK3, COQ8B/ADCK4, ADCK1 and ADCK5. It is required to maintain the mitochondrial pool of coenzyme Q (ubiquinone) in skeletal muscle. Human haploinsufficiency for ADCK2, and heterozygous loss of Adck2 in mouse, lower muscle CoQ, impair mitochondrial fatty acid beta-oxidation and produce a lipid-storage mitochondrial myopathy with hepatic steatosis that is partly corrected by CoQ10 supplementation; the beta-oxidation defect follows from the CoQ deficit rather than from a separate lipid-handling activity. Isotope labelling shows that total cellular synthesis of isoprenoids and CoQ is normal in ADCK2-deficient cells and that only the mitochondrial pool is depleted, so ADCK2 acts on the delivery of CoQ precursors to the organelle rather than catalysing a step of the biosynthetic pathway; consistent with this, its budding-yeast orthologue Cqd1 (YPL109C) controls the partitioning of CoQ between mitochondria and the rest of the cell, and human ADCK2 restores CoQ synthesis to yeast lacking YPL109C. The protein is confined to mitochondria, where protease protection places it in the matrix or bound to the inner membrane, and it carries a predicted single-pass transmembrane helix in the position occupied by the membrane anchor of COQ8A. ADCK2 retains an intact protein-kinase-like catalytic site together with the UbiB-specific KxGQ motif that, in COQ8A, occludes the peptide-substrate groove; no catalytic activity, substrate or nucleotide preference has been measured for ADCK2 itself.
Definition: The directed movement and partitioning of coenzyme Q (ubiquinone) between the mitochondrial inner membrane, where it is synthesised, and other cellular membranes, resulting in the establishment or maintenance of the relative sizes of the mitochondrial and extramitochondrial coenzyme Q pools.
Justification: GO can currently express only the synthesis and breakdown of ubiquinone. The absence was established by two complementary sweeps that could each have returned a hit, in ADCK2-bioinformatics/coq_transport_term_check.py. A label sweep retrieves all 102 GO terms whose label mentions ubiquinone, coenzyme Q, quinone or quinol. Every transport-flavoured term among them is set aside on one of two grounds, both printed by the script with the candidate they apply to. Most are electron-transport terms, where the electron is the cargo and the quinone merely the acceptor. One, GO:1903222 quinolinic acid transmembrane transport, is a genuine transport process and is excluded instead because it matched only through a substring accident - quinol is contained in quinolinic - and quinolinic acid is a tryptophan-pathway NAD precursor unrelated to coenzyme Q. A branch sweep enumerates the descendants of GO:0006869 lipid transport, GO:0010876 lipid localization, GO:0032365 intracellular lipid transport, GO:0120009 intermembrane lipid transfer and GO:0006810 transport - 2584 distinct terms, the roots being nested - and finds no ubiquinone-specific child. So no GO term describes the movement or partitioning of coenzyme Q itself. Yet that is precisely the characterised activity of a whole UbiB sub-branch: Cqd1 and Cqd2 reciprocally control where CoQ ends up in the cell, and loss of Cqd1 shifts it away from mitochondria. Curators of those two yeast genes therefore have no term that states what was measured, and the same gap is why the human ADCK2 evidence has to be recorded as regulation of a biosynthetic process even though the paper's own conclusion is a trafficking defect. The immediate holders of this term would be yeast Cqd1 and Cqd2, not ADCK2 - human ADCK2 has not been shown to distribute CoQ and is proposed only as a candidate once the orthology is tested functionally.
Parent term: intracellular lipid transport
Supporting Evidence:
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0005739 mitochondrion | IBA GO_REF:0000033 | ACCEPT | Summary: Correct and well supported by the donors' own experimental evidence. The node is clean, the term is the right level for a heterogeneous donor set, and the row records a PAINT curator judging mitochondrial localisation core to this orthologue group. Reason: Every WITH/FROM token was resolved and queried for its own evidence for the propagated term. Both non-self donors carry experimental support: mouse Adck2 (Q6NSR3, Swiss-Prot) has HDA from PMID:18614015 plus IDA from PMID:31480808, and yeast YPL109C/Cqd1 (Q02981, Swiss-Prot) has three independent HDA annotations. So the "the sources only carry the same family-level inference" objection is factually false here. The ACRV1 test for a propagation landing above its donor was run and does not apply: the donor set is heterogeneous in specificity, because Cqd1 additionally holds IDA for GO:0005743 mitochondrial inner membrane and GO:0044289 from PMID:37073556 while mouse Adck2 holds only mitochondrion. With donors that disagree, GO:0005739 is the correct least common ancestor, and a downward MODIFY would mean arbitrarily preferring the yeast donor over the mouse one. The node itself was queried directly and is coherent: PTN000059786 propagates 49 annotations over 49 distinct gene products to this single term, its reach is the ADCK2/Cqd1 orthologue group, and yeast Mcp2/Cqd2, which is the donor for ADCK1 under a different PANTHER family, is absent from it. Propagation Review Root cause: NO FAILURE CORE Sources checked: MGI:MGI:1889336 · mouse Adck2 (Q6NSR3, Swiss-Prot, 617 aa) SUPPORTS TRANSFER True one-to-one orthologue. QuickGO shows it carries its own HDA for GO:0005739 from PMID:18614015 (mouse mitochondrial protein compendium) and IDA from PMID:31480808, so the transfer is not circular with the target's own evidence even though the two share one reference. The MGI cross-reference is ambiguous, also resolving to four TrEMBL entries; only the Swiss-Prot entry carries annotations. SGD:S000006030 · S. cerevisiae YPL109C / Cqd1 (Q02981, Swiss-Prot, 657 aa) SUPPORTS TRANSFER The yeast orthologue, sharing PANTHER subfamily PTHR45890:SF1 with ADCK2, and the strain used for the complementation experiment in PMID:31480808. Carries three independent HDA annotations for GO:0005739 and, more specifically, IDA for GO:0005743 and GO:0044289 from PMID:37073556. Holding a more specific term than the one propagated is why this donor set is heterogeneous rather than agreeing. PANTHER:PTN000059786 · PANTHER internal tree node NOT RELEVANT Not a protein but the ancestral node PAINT annotated. Queried directly rather than dismissed: it propagates exactly one term to 49 distinct gene products, all orthologues of ADCK2/Cqd1. UniProtKB:Q7Z695 · ADCK2 itself (self-referential) SUPPORTS TRANSFER The target appears in its own WITH/FROM list. This is a valid PAINT construct recording that the curator judged the term core to the gene, not circular evidence. Supporting Evidence: PMID:31480808 The presence of ADCK2 was not detected in the endoplasmic reticulum and cytosolic fraction. PMID:34362905 we identify two highly conserved but poorly characterized mitochondrial proteins, Ypl109c (Cqd1) and Ylr253w (Cqd2), that reciprocally affect this process |
| GO:0005739 mitochondrion | IEA GO_REF:0000044 | ACCEPT | Summary: Correct. Automatic mapping of UniProt's SUBCELLULAR LOCATION line, which itself rests on experimental evidence; redundant with the experimental rows but not wrong. Reason: Derived from UniProt-SubCell SL-0173, which maps the entry line "SUBCELLULAR LOCATION: Mitochondrion {ECO:0000269|PubMed:33988507}". The underlying assertion is therefore experimentally grounded rather than predicted, and it is independently confirmed by subcellular fractionation of endogenous protein and by the MitoCoP high-confidence proteome. It adds no information over the IDA rows, but there is no defect to correct. Supporting Evidence: PMID:33988507 456 epitope-tagged kinases, representing 85% of the human kinome, were expressed in HeLa cells and imaged by immunofluorescent microscopy |
| GO:0016020 membrane | IEA GO_REF:0000044 | KEEP AS NON CORE | Summary: Retained but non-core. The term rests on a pure sequence prediction, and although the predicted helix sits where COQ8A's experimentally supported anchor sits, no measurement on human ADCK2 establishes membrane integration. Reason: This is the weakest row in the set. It maps UniProt-SubCell SL-0162, whose only basis is the entry line "Membrane {ECO:0000255}; Single-pass membrane protein {ECO:0000255}" - a sequence prediction from TRANSMEM 103..123, with no experimental tag anywhere. Two considerations keep it from being marked over-annotated. First, the predicted helix is well placed rather than arbitrary: the alignment in ADCK2-bioinformatics/ shows ADCK2 preserves the transmembrane-helix then KxGQ then kinase-domain arrangement of COQ8A, whose own membrane anchor is annotated at the corresponding position, and the yeast orthologue Cqd1 has been shown experimentally to be an inner membrane protein. Second, nothing contradicts it: the protease-protection experiment on endogenous human ADCK2 is explicitly indecisive between the matrix and the inner membrane, so it neither confirms nor refutes membrane association. The honest position is that GO:0016020 is plausible and unmeasured, so it is kept but excluded from the core function statement, and resolving the sub-mitochondrial location is filed as a suggested experiment. No more specific term is proposed, because doing so would mean transferring the yeast IDA onto the human protein. Supporting Evidence: PMID:31480808 consistent with its presence in the mitochondrial matrix or bound to the inner mitochondrial membrane PMID:37073556 we identified a novel mitochondrial contact site in Saccharomyces cerevisiae that is formed by the inner membrane protein Cqd1 and the outer membrane proteins Por1 and Om14 |
| GO:0005515 protein binding | IPI PMID:27499296 Mitochondrial Protein Interaction Mapping Identifies Regulat... | MARK AS OVER ANNOTATED | Summary: An uninformative term resting on a single unreplicated, spoke-expanded affinity-capture run against one of the most abundant and promiscuous proteins of the inner membrane. Kept in the record but marked over-annotated; there is no informative molecular function to replace it with. Reason: The partner is UniProtKB:P05141, SLC25A5/ANT2, resolved as the reviewed canonical Swiss-Prot entry with no length discrepancy, so this is not an ORFeome-fragment substitution. The evidence was then expanded rather than counted. UniProt states NbExp=4, but all four IntAct evidences come from one publication, PMID:27499296, by one method (anti tag coip), with ADCK2 as bait and SLC25A5 as prey, at MI-score 0.35, and three of the four are spoke expansions of a co-complex pulldown rather than demonstrated binary interactions. NbExp=4 is therefore one experiment counted four times, the same pattern already seen on ACRV1 and ADAMTSL5. Promiscuity compounds it: SLC25A5 carries 385 IntAct interaction evidences, and ADCK2 itself has 83 distinct partners across 7 publications. The topological objection that carried the ACRV1 case is explicitly absent here and is recorded as a negative result - both proteins face the matrix side of the inner membrane, so the interaction is physically possible - but possibility is not support. The cached full text of the source paper never mentions ADCK2, so no author interpretation exists to weigh. Per the curation guidelines, bare protein binding should be replaced by an informative molecular function where the partner is real; here the partner is a plausible but unreplicated screen hit and no informative term is warranted, so the row is marked over-annotated rather than modified or removed. Supporting Evidence: PMID:27499296 we assessed condition-specific protein-protein interactions for 50 select MXPs using affinity enrichment mass spectrometry |
| GO:0005739 mitochondrion | HTP PMID:34800366 Quantitative high-confidence human mitochondrial proteome an... | ACCEPT | Summary: Correct. Inclusion in the MitoCoP high-confidence mitochondrial proteome, appropriately coded HTP. Reason: MitoCoP is a filtered high-confidence set derived from classifying over 8,000 proteins in mitochondrial preparations, not a raw detection list, so membership is a meaningful call and HTP is the right evidence code for it. It agrees with the two IDA rows and with the IBA. The reference-projection test is inconclusive for this paper because the QuickGO result is paginated at 1235 annotations and an entity count cannot be derived from one page; that is recorded rather than substituted with the annotation total, and it does not affect the verdict, since a proteome survey is expected to annotate many entities and the evidence code already says so. Supporting Evidence: PMID:34800366 We classified >8,000 proteins in mitochondrial preparations of human cells and defined a mitochondrial high-confidence proteome of >1,100 proteins (MitoCoP). |
| GO:0005739 mitochondrion | IDA PMID:31480808 ADCK2 Haploinsufficiency Reduces Mitochondrial Lipid Oxidati... | ACCEPT | Summary: Correct, and the strongest localisation row in the set: subcellular fractionation of endogenous ADCK2 plus protease protection, in a paper about this gene. Reason: Unlike the other localisation rows this one is on endogenous, untagged protein. ADCK2 co-fractionates with TOM20 and MFN2, is absent from ER and cytosolic fractions, and is protected from proteinase K in intact mitochondria while being digested after detergent permeabilisation. The last result also constrains topology usefully: whatever membrane association ADCK2 has, its bulk is not exposed to the cytosol. The paper declines to discriminate matrix from inner membrane, so no more specific term is warranted from it and none is proposed. Supporting Evidence: PMID:31480808 Immunoblotting in various subcellular fractions of HEK293 cells showed a distribution of ADCK2 similar to the mitochondrial TOM20 and Mfn2 PMID:31480808 consistent with its presence in the mitochondrial matrix or bound to the inner mitochondrial membrane |
| GO:0005739 mitochondrion | IDA PMID:33988507 A subcellular map of the human kinome. | ACCEPT | Summary: Correct conclusion, corroborating rather than load-bearing: the image is of an epitope-tagged, ectopically expressed construct, a caveat the authors state themselves. Reason: Kinome Atlas expressed 456 epitope-tagged kinases in HeLa cells and imaged them, assigning ADCK2 to mitochondria. Because the protein was overexpressed and tagged, this is weaker than the endogenous fractionation in PMID:31480808, and the authors flag possible mislocalisation from exactly that cause. It is kept as ACCEPT rather than downgraded because the conclusion is independently confirmed twice over, by endogenous fractionation and by the MitoCoP proteome, so the caveat does not bear on whether the term is true. Worth noting that this screen is nonetheless the sole ECO:0000269 citation behind UniProt's SUBCELLULAR LOCATION line, which is a thinner foundation there than the GO record makes it look. Supporting Evidence: PMID:33988507 456 epitope-tagged kinases, representing 85% of the human kinome, were expressed in HeLa cells and imaged by immunofluorescent microscopy PMID:33988507 Although limited by possible mislocalization due to overexpression or epitope tagging |
| GO:0010795 regulation of ubiquinone biosynthetic process | IDA PMID:31480808 ADCK2 Haploinsufficiency Reduces Mitochondrial Lipid Oxidati... | ACCEPT | Summary: Correct, and correct at the right level. The curator's choice of the regulation term over the biosynthetic-process term is confirmed by the paper's own isotope-labelling experiment, which shows total synthesis is unaffected and only the mitochondrial pool is depleted. This is the gene's core biological role. Reason: Three independent systems support involvement in maintaining the CoQ pool: a human nonsense allele (R333*, matching UniProt VARIANT 333..626 Missing) that lowers CoQ10 in patient fibroblasts and is rescued by re-expressing wild-type ADCK2; Adck2 heterozygous mice with reduced CoQ9 and CoQ10 in skeletal muscle; and heterologous complementation in which human ADCK2 restores CoQ6 synthesis to yeast lacking YPL109C. The choice of GO:0010795 over GO:0006744 ubiquinone biosynthetic process, which COQ8A and COQ8B correctly hold, was tested rather than assumed and is right: 3H-mevalonate labelling found total cellular CoQ and cholesterol synthesis unchanged in Adck2-deficient cells, with only the mitochondrial fraction depleted, so ADCK2 modulates the rate and extent of mitochondrial CoQ accumulation rather than catalysing a pathway step. That reading is reinforced by the orthologue, since loss of yeast Cqd1 shifts CoQ away from mitochondria. One observation for the record rather than a proposed change: the assays behind this row are loss-of-function and complementation rather than a direct assay of purified ADCK2, so the underlying experiments are IMP-shaped, but the term itself is right and the evidence code is the assigning curator's call on the full text. Supporting Evidence: PMID:31480808 which indicated a defect in intracellular trafficking of isoprenoid and cholesterol from the cytoplasm PMID:31480808 yeast strain with wild type YPL109c or human ADCK2 construct rescued CoQ6 biosynthesis PMID:31480808 The mutation produced a termination codon that led to a significant decrease in ADCK2 mRNA and protein levels in dermal fibroblasts PMID:34362905 Loss of Cqd1 skews cellular CoQ distribution away from mitochondria |
Loading supporting content…
Download this section (compressed HTML)Q: Should UniProt withdraw EC=2.7.11.- and the Serine/threonine-protein kinase keyword from Q7Z695? The entry currently asserts protein-serine/threonine kinase activity in three machine-readable fields - the EC number in the RecName, the keyword set, and the keyword-derived DR GO line for GO:0004674 - while its own FUNCTION comment states that it is not known whether the protein has protein kinase activity or what residue class it would phosphorylate. EC 2.7.11.- specifically means protein-serine/threonine kinase, so the machine-readable layer is more committed than any evidence supports, and more committed than the curated prose in the same entry. The family context makes this sharper: GOA records NOT|enables GO:0004672 for both characterised members on the basis of PMID:27499294. Note that GOA itself no longer carries the keyword-derived rows, so this defect is invisible from the GO annotation set and can only be fixed upstream. The ask is deliberately scoped to the protein-kinase claims and does not extend to the ATP-binding and Nucleotide-binding keywords or their GO:0005524 cross-reference: this review's own alignment finds the beta3 lysine, the catalytic aspartate and asparagine and the DFG-equivalent aspartate all intact, so nucleotide binding is the best-supported part of the keyword set. It is still not demonstrated for ADCK2, and the family gives a concrete reason for caution about the ligand's identity rather than its existence, since COQ8A binds ADP in preference to ATP and carries GO:0043531 rather than GO:0005524. So GO:0005524 should be re-examined on ligand identity, not withdrawn as unfounded.
Suggested experts: UniProt curation, GO Consortium
Q: Can the GO record for GO:0004672 on COQ8A and COQ8B be reconciled? Both genes simultaneously hold a NOT|enables IDA from PMID:27499294 and a positive assertion: COQ8B an enables IDA from PMID:38425362, which reports in vitro phosphorylation of COQ3, and COQ8A an enables ISS propagated from COQ8B under GO_REF:0000024. A gene asserting both that it does and does not enable the same molecular function is a contradiction that downstream consumers cannot resolve, and it directly affects how confidently anything can be said about kinase activity elsewhere in the family, including ADCK2.
Suggested experts: GO Consortium, UniProt curation
Q: Is human ADCK2 the functional orthologue of yeast Cqd1 for coenzyme Q distribution, and is ADCK1 the orthologue of Cqd2? PANTHER places ADCK2 with YPL109C/Cqd1 in PTHR45890:SF1 and ADCK1 with YLR253W/Mcp2/Cqd2 in PTHR43173:SF19, matching the assignment proposed by the authors who characterised the two yeast genes. A residue-level check agrees: in the alignment in ADCK2-bioinformatics/, the position equivalent to COQ8A A339 carries glycine in exactly two of the eight UbiB proteins examined, ADCK2 and Cqd1, while all six others including Cqd2 and ADCK1 carry the suppressor alanine - so a single residue reproduces the pairing. This is a second sequence-based line agreeing with the genetics rather than a third independent line, because PANTHER's subfamily assignment is itself sequence-derived; and it is specific to that column, since the adjacent A-rich position carries glycine in four proteins and cuts across the pairing. Because Cqd1 and Cqd2 have reciprocal effects on CoQ partitioning, this predicts that ADCK2 and ADCK1 should act in opposite directions in human cells - a prediction that has not been tested and that matters for whether the two genes should ever receive shared annotations.
Suggested experts: Coenzyme Q biology, PAINT curation
Q: Where in the mitochondrion does human ADCK2 sit? The only human measurement is protease protection, which is explicitly indecisive between the matrix and the inner membrane, while yeast Cqd1 has been localised to the inner membrane at an inner-outer contact site with Por1 and Om14, and the authors expect that arrangement to be conserved to human. Resolving this would decide whether GO:0016020 should become GO:0005743 or GO:0044289 and whether ADCK2 belongs to a human counterpart of that contact site.
Suggested experts: Mitochondrial cell biology
Experiment: Purify recombinant human ADCK2 lacking its N-terminal membrane segment, as was done for COQ8A, and measure nucleotide binding, ATPase activity, and phosphotransfer to generic peptide substrates. Compare wild type with the catalytic-site mutants identified by this review's alignment - K311 in the beta3 strand, D445 the proton acceptor, and D493 in the DFG-equivalent motif - to establish that any activity observed is ADCK2's own rather than a contaminating kinase. This is the single missing measurement behind almost every open question on this gene, and it is the assay that resolved the equivalent question for COQ8A.
Hypothesis: ADCK2 is an unorthodox protein-kinase-like enzyme that hydrolyses ATP but does not phosphorylate protein substrates in trans, as shown for COQ8A and COQ8B.
Type: in vitro enzymology
Experiment: On purified ADCK2, compare autophosphorylation and peptide phosphorylation for wild type against a G207A/G209A variant that installs the COQ8A-like suppressor alanines, and against a KxGQ mutant at K147. The prediction that discriminates the two readings is that the glycine-to-alanine swap should reduce autophosphorylation while leaving in-trans activity absent in every variant that retains the KxGQ motif. Stated explicitly as a hypothesis: the sequence divergence found here is suggestive and has not been tested.
Hypothesis: ADCK2's A-rich loop, which carries glycine where COQ8A carries the alanines that suppress phosphotransfer, makes the ADCK2 branch of the family more permissive for autophosphorylation than the COQ8 branch, without conferring in-trans protein kinase activity while the KxGQ motif remains intact.
Type: structure-function mutagenesis
Experiment: Repeat the 3H-mevalonate partitioning experiment in isogenic ADCK2 and ADCK1 knockouts in the same human cell line, measuring labelled CoQ and cholesterol separately in whole-cell and mitochondrial fractions, and quantify the mitochondrial versus extramitochondrial CoQ pools directly by lipidomics. The reciprocal-orthologue model predicts opposite shifts in the mitochondrial CoQ fraction for the two knockouts; a same-direction result would falsify the Cqd1/Cqd2 mapping in human cells.
Hypothesis: Human ADCK2 acts on the delivery of CoQ precursors into mitochondria rather than on the biosynthetic reactions, and does so in the direction opposite to ADCK1.
Type: isotope labelling and subcellular lipidomics
Experiment: Determine ADCK2's sub-mitochondrial location and topology in human cells by carbonate extraction and by protease protection of mitoplasts rather than intact mitochondria, which is the step that would discriminate matrix from inner membrane where the published experiment could not, and test by proximity labelling whether VDAC family proteins, the human counterparts of Por1, are enriched around ADCK2.
Hypothesis: ADCK2 occupies an inner-membrane contact site analogous to the yeast Cqd1-Por1-Om14 complex.
Type: subcellular fractionation and proximity labelling
Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)