AADACL3 is an uncharacterised human member of the arylacetamide deacetylase (AADAC) family of 'GDXG' lipolytic enzymes. The characterised members of that family are serine hydrolases that cleave ester and amide bonds in both lipid and xenobiotic substrates: AADAC itself is an endoplasmic reticulum type II membrane glycoprotein that deacetylates arylacetamide drugs such as flutamide, phenacetin and rifamycin and hydrolyses diacylglycerol, while NCEH1/AADACL1 hydrolyses acetyl monoalkylglycerol ether, cholesteryl esters and retinyl esters. AADACL3 retains the full catalytic apparatus of the family - an alpha/beta-hydrolase fold-3 core carrying a Ser193-Asp347-His377 triad in a canonical GDSFGG nucleophile elbow, together with an HGG oxyanion loop - and UniProt assigns it the carboxylic-ester hydrolase class EC 3.1.1.-, but no catalytic activity, substrate or physiological role has been measured for it, and the protein itself has not yet been detected. Its N-terminal region is predicted to be membrane-embedded rather than a cleaved signal peptide, which would leave the catalytic domain facing the lumenal or extracellular side of a membrane as in AADAC and NCEH1; the comparably related AADACL2, by contrast, is secreted. AADACL3 lies in a tandem pair with AADACL4 at chromosome 1p36.21, separate from the AADAC/AADACL2 pair at 3q25.1, and its nearest characterised relative shares only about a third of its sequence. Transcript expression is enhanced in placenta and skin, in contrast to the liver and intestinal expression of AADAC.
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0016787 hydrolase activity | IBA GO_REF:0000033 | MODIFY | Summary: Phylogenetic transfer of the generic term hydrolase activity from a deliberately wide clade. Resolving all seventeen WITH/FROM entries explains why the term is so general: the set mixes carboxylesterases (E. coli Aes, three M. tuberculosis esterases), arylacetamide deacetylase and neutral cholesterol ester hydrolase 1 from mouse, rat and human, two formylkynurenine formamidases that are amidohydrolases rather than esterases (mouse Afmid and yeast BNA7), four Arabidopsis loci, and soybean 2-hydroxyisoflavanone dehydratase, which UniProt gives two EC numbers, 3.1.1.1 carboxylesterase and 4.2.1.105 dehydratase. Hydrolase activity is close to the most specific term that survives across esterases and amidohydrolases together, so as a statement about that clade node the annotation is defensible, and it is certainly not wrong for AADACL3. It is nonetheless the least informative of the three molecular-function terms GOA carries for this gene, and it is a proper ancestor of GO:0052689 carboxylic ester hydrolase activity, which the same GOA record already asserts from the arylacetamide-deacetylase subfamily signature IPR017157. The weakest donor in the supporting set is soybean HIDH (Q5NUF3). UniProt assigns it both EC 3.1.1.1 carboxylesterase and EC 4.2.1.105 2-hydroxyisoflavanone dehydratase, so it does formally license a hydrolase term, but only weakly: the dehydration is the physiologically important reaction and the esterase activity is described as slight, toward p-nitrophenyl butyrate. Reading the donors' own active sites settles whether a more specific term could have been used at this node, and the answer is no. Of the seventeen WITH/FROM tokens, fourteen resolve to a protein whose nucleophile can be read, and thirteen of those have a serine nucleophile - including the formamidase Afmid, so the amidohydrolase members are serine hydrolases and are no obstacle. The obstacle is HIDH, whose nucleophile is Thr164. That single member is enough to keep GO:0017171 serine hydrolase activity off this node, and the two formamidases hydrolyse an amide rather than an ester bond, which keeps GO:0052689 off it as well. GO:0016787 really is the lowest common ancestor of this donor set, so the term is correct curation rather than a lazy or over-general choice, and this row is not an over-annotation or a term-scoping error. What remains is redundancy. AADACL3's GOA already carries GO:0052689 carboxylic ester hydrolase activity from the arylacetamide-deacetylase subfamily signature IPR017157, which is stronger, more direct evidence for this particular gene than a transfer from a clade whose breadth forces the term up to a parent. AADACL3's own assignment is much tighter than the clade it was annotated from: PIRSF037251 and IPR017157 both match residues 19-403 (PIRSF at 5.8e-82), UniProt gives the RecName EC number 3.1.1.-, and the catalytic apparatus is verifiably intact - Ser193, Asp347 and His377 read out of the sequence as serine, aspartate and histidine in a canonical GDSFGG nucleophile elbow with an HGG oxyanion loop at 119-121, all three landing on curated active sites of the same residue identity in AADAC, NCEH1, mouse Nceh1 and the mouse ortholog. That verified triad does support GO:0017171 for the gene, whose definition it satisfies literally - serine nucleophile activated through an acidic and a basic residue. But it cannot be reached by transfer from this node, for the reason above, so it is raised as a node-placement recommendation for PAINT in suggested_questions rather than proposed as a replacement here. Reason: The essence is right but the granularity is not. Redundant with, and subsumed by, GO:0052689 already present on the same gene from a subfamily-specific signature, which is supported independently by the intact Ser-Asp-His triad and by UniProt's own EC 3.1.1.- assignment. This is a redundancy judgement, not a claim that the term is too general for its node. The donor audit confirms GO:0016787 is the true lowest common ancestor here: HIDH's threonine nucleophile blocks a serine-hydrolase term and the two formamidases' amide chemistry blocks an ester term, so PAINT's choice is correct curation. What makes the row replaceable is that the same GOA record already carries GO:0052689 from a subfamily-specific signature, which is stronger and more direct evidence for this gene than a transfer from a clade that broad. No failure mode is recorded. Read literally, GRANULARITY_MISMATCH (a parent term that is true but uninformative) does describe how much this term tells you about this gene; but the slot is the biological shape of a propagation issue, and this propagation has none - the term is uninformative because the donor set is heterogeneous, not because the transfer could have been more specific. That is the same encoding now used for AADACL2 and AADACL4, which carry the identical row, so the three paralogs no longer disagree about it. The row is still described honestly on both readings: correct where it was asserted, uninformative where it landed. Propagation Review Root cause: EVIDENCE CIRCULAR OR REDUNDANT Sources checked: UniProtKB:P22760 · AADAC, human arylacetamide deacetylase SUPPORTS TRANSFER Characterised deacetylase and diacylglycerol lipase, 33.3% identical to AADACL3, and all three of AADACL3's active-site positions align onto its curated active sites Ser189, Asp343 and His373. Supports a more specific ester-hydrolase term, not merely hydrolase. MGI:MGI:2448704 · Afmid, mouse arylformamidase SUPPORTS SOURCE BUT NOT TARGET An amidohydrolase (EC 3.5.1.9) rather than an esterase, so it blocks an ester-hydrolase term at this node; together with yeast BNA7 it is what forces the clade term up to the generic hydrolase level. Note it does not block a mechanism term: its own nucleophile is Ser162, so the formamidases are serine hydrolases too. SGD:S000002836 · BNA7, yeast formylkynurenine formamidase SUPPORTS SOURCE BUT NOT TARGET Also an amidohydrolase of the kynurenine pathway; same effect on the clade term as Afmid. This gene's own audit could not resolve it through its Alliance record and reported the nucleophile as unresolved; the shared node audit resolves it directly through xref:sgd-S000002836 to Q04066, whose ACT_SITE 110 UniProt labels the nucleophile and which reads as Ser. So BNA7 is a serine hydrolase too, and the node-wide tally is 15 of 16 donors with a serine nucleophile rather than 13 of 14 - which strengthens rather than weakens the conclusion, since the sole non-serine donor remains HIDH. UniProtKB:Q5NUF3 · HIDH, soybean 2-hydroxyisoflavanone dehydratase SUPPORTS TRANSFER Carries two EC numbers in UniProt, 3.1.1.1 carboxylesterase and 4.2.1.105 2-hydroxyisoflavanone dehydratase, and its carboxylesterase activity is not merely formal: GO:0106435 carboxylesterase activity is held by IDA, alongside the GO:0033987 dehydratase IDA. So it genuinely supports the hydrolase parent - the dehydration is the physiologically important reaction and the esterase activity is slight, but slight is not inferred. It is also the ECO:0000250 source for AADACL3's oxyanion-hole motif, and in the alignment only two of AADACL3's three triad positions map onto its annotated sites, and just one of those is the same residue. It is also the member that keeps a serine-hydrolase term off this node: its nucleophile is Thr164, not serine, so it is the single reason GO:0016787 rather than GO:0017171 is the correct lowest common ancestor here. PANTHER:PTN009058710 · PANTHER ancestral node used for the transfer SUPPORTS TRANSFER Not AADACL3 itself, so this is not a self-referential IBA. A different node (PTN009058713) was used for the membrane annotation. Proposed replacements: carboxylic ester hydrolase activity Supporting Evidence: file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | 193 | S193 | S | yes | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | AAAD_HUMAN (P22760) | 33.3 | S193→S189/D347→D343/H377→H373 | 3/3 | 3/3 | `HGG` | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_HUMAN (Q5VUY0) | 19-403 | 19-403 (5.8e-82) [representative] | 115-245; 310-380 | 74-385 | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md Audited: GO:0016787 (IBA, GO_REF:0000033). 17 WITH/FROM tokens, 14 resolved to a protein, 14 with a readable nucleophile, of which **13 are serine**. Non-serine: HIDH_SOYBN T164. file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | `UniProtKB:Q5NUF3` | HIDH_SOYBN | T164 | **no** | UniProt accession given directly | file:human/AADACL2/AADACL2-bioinformatics/NODE_PTN009058710.md No donor refutes `GO:0016787`, and every more specific candidate is refuted by at least one donor. file:human/AADACL2/AADACL2-bioinformatics/NODE_PTN009058710.md GO:0016787 hydrolase activity: TRUE 16, FALSE 0, UNDETERMINED 0 file:human/AADACL2/AADACL2-bioinformatics/NODE_PTN009058710.md So HIDH is a genuine, if physiologically minor, carboxylic-ester hydrolase; it refutes the *serine* term only. |
| GO:0016020 membrane | IBA GO_REF:0000033 | ACCEPT | Summary: Unlike the hydrolase row, this transfer used a tight and topologically coherent donor set: mouse Aadac, mouse Nceh1 and human AADAC, all three curated as single-pass type II membrane proteins whose catalytic domain faces the endoplasmic reticulum lumen or the extracellular space. The secreted family member AADACL2 is not among the donors, so the transfer is discriminating rather than indiscriminate, and that matters here because the family gives conflicting topology at roughly equal sequence distance: AADAC and NCEH1 are membrane-anchored at about 33% identity to AADACL3 while AADACL2 is secreted at about 34%. The inference is independently corroborated. Phobius places transmembrane segments at 6-29 and 41-60 in AADACL3 and TMHMM places helices at 4-26 and 43-60, while neither SignalP nor Phobius reports a cleaved signal peptide; both methods assign residues 61-407, the whole catalytic domain, to the non-cytoplasmic side. The discriminating control is AADACL2, where SignalP does fire and UniProt's curated location is Secreted, so the absence of a signal-peptide call for AADACL3 is informative rather than a default. Worth recording rather than smoothing over: both predictors call two helices in AADACL3, not one, and the mouse ortholog settles the point. UniProt curates Aadacl3 (A2A7Z8), 61.9% identical to the human protein and carrying the same triad, as `Membrane; Multi-pass membrane protein` with three helices at 2-22, 46-66 and 109-129. So the family's single-pass type II description is an under-call for this branch, not just for the human protein. Since either topology leaves the catalytic domain non-cytoplasmic, the biological reading is unchanged, and because the donors split between endoplasmic reticulum membrane (AADAC) and plasma membrane (NCEH1), the generic term membrane is the correct level. That is worth stating positively rather than as a hedge: the donors' curated locations resolve to four distinct strings between them - endoplasmic reticulum membrane and microsome membrane for AADAC, cell membrane and microsome for Nceh1 - so refining the term would mean picking one donor over another arbitrarily. GO:0016020 is the lowest common ancestor of the donor locations just as GO:0016787 is of the donor activities, and in both cases the general term is the ontology working correctly rather than curation stopping short. Note also that GO:0016020 has absorbed GO:0016021 integral component of membrane as a secondary identifier, so a transmembrane-helix prediction now supports this exact term. One qualifier nuance is worth acknowledging rather than silently accepting. The row asserts is_active_in, which locates the activity in the bilayer, while the topology reading places the catalytic domain on the non-cytoplasmic side of it - so on a strict reading the hydrolysis happens at the membrane surface rather than within the lipid phase. The alternative would be to name the compartment the catalytic domain faces, and that cannot be done for AADACL3 without knowing which membrane it sits in, so the qualifier is left as GOA has it. Reason: Correct at the right level of generality, from an appropriately restricted donor set, and corroborated by concordant Phobius and TMHMM predictions on the AADACL3 sequence itself with a secreted paralog as a negative control. Kept generic deliberately: the evidence supports membrane integration but cannot identify which membrane. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:P22760 · AADAC, human arylacetamide deacetylase SUPPORTS TRANSFER Curated as an endoplasmic reticulum and microsome membrane single-pass type II membrane protein with TRANSMEM 6-23, and reported in the family review as a type II membrane glycoprotein facing the ER lumen with its active side. MGI:MGI:2443191 · Nceh1, mouse neutral cholesterol ester hydrolase 1 SUPPORTS TRANSFER Curated as a cell-membrane single-pass type II membrane protein with TRANSMEM 5-25. It is also the ECO:0000250 donor for AADACL3's three active-site positions. MGI:MGI:1915008 · Aadac, mouse arylacetamide deacetylase SUPPORTS TRANSFER Mouse ortholog of the family prototype; same type II membrane topology. PANTHER:PTN009058713 · PANTHER ancestral node used for the transfer SUPPORTS TRANSFER A different, more restricted node than the one behind the hydrolase-activity row, and not AADACL3 itself, so this is not a self-referential IBA. Supporting Evidence: file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_HUMAN (Q5VUY0) | - | - | - | [(6, 29), (41, 60)] | - | [(4, 26), (43, 60)] | - | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL2_HUMAN (Q6P093) | - | [(1, 18)] | Secreted | - | [(1, 19)] | - | [(1, 18), (1, 19)] | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_MOUSE (A2A7Z8) | [(2, 22), (46, 66), (109, 129)] | - | Membrane; Multi-pass membrane protein | [(42, 61), (111, 129)] | [(1, 26)] | [(2, 24)] | - | PMID:35736449 The arylacetamide deacetylase (AADAC) protein family name giving protein AADAC is a type II membrane glycoprotein, facing with its active side to the lumen of the endoplasmic reticulum (ER) [1]. |
| GO:0016020 membrane | IEA GO_REF:0000120 | ACCEPT | Summary: The same conclusion as the phylogenetic membrane annotation, reached independently through UniProt's automatic pipelines from an ARBA rule and from the arylacetamide-deacetylase subfamily signature IPR017157, which matches AADACL3 across residues 19-403. It is retained on the same evidence and for the same reason. One point about this row deserves emphasis because it inverts the usual direction of concern. AADACL3's UniProt entry contains no TRANSMEM feature, no signal-peptide feature, no membrane or transmembrane keyword and no SUBCELLULAR LOCATION comment at all, so this GO annotation looks at first sight to be unsupported by the underlying record. It is not: its tandem paralog AADACL4 is the same length, carries the same numbered active sites and oxyanion motif, receives identical Phobius transmembrane calls, and is annotated by UniProt with TRANSMEM 5-25, both topological domains and Single-pass type II membrane protein. The difference between the two entries is coverage of UniProt's automatic annotation, not biology. The mouse ortholog closes the argument: UniProt curates Aadacl3 (A2A7Z8) as `Membrane; Multi-pass membrane protein` with three transmembrane helices, so the same gene in mouse carries a curated membrane location while the human entry carries none at all. The GO record is therefore ahead of the UniProt feature table here, and the gap is an error to report upstream rather than a reason to weaken the annotation. Reason: Independently derived, biologically supported, and at the appropriate level of generality. The apparent absence of supporting UniProt features is a coverage gap in UniProt's automatic annotation of this entry rather than evidence against membrane localisation, as shown by the identical predictions on the tandem paralog AADACL4, which did receive the features. Propagation Review Root cause: NO FAILURE CORE Sources checked: InterPro:IPR017157 · Arylacetamide deacetylase (subfamily signature) SUPPORTS TRANSFER Matches AADACL3 at residues 19-403; the PIRSF member signature PIRSF037251 scores 5.8e-82 over the same span, so this is a subfamily-level rather than fold-level assignment. ARBA:ARBA00028763 · ARBA automatic annotation rule SUPPORTS TRANSFER Rule-based transfer reaching the same conclusion as PAINT from a different route; the conclusion is corroborated by Phobius and TMHMM calls on the AADACL3 sequence itself. Supporting Evidence: file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL4_HUMAN (Q5VUY2) | [(5, 25)] | - | Membrane; Single-pass type II membrane protein | [(6, 29), (41, 60)] | - | [(2, 24)] | - | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_MOUSE (A2A7Z8) | [(2, 22), (46, 66), (109, 129)] | - | Membrane; Multi-pass membrane protein | [(42, 61), (111, 129)] | [(1, 26)] | [(2, 24)] | - | |
| GO:0016787 hydrolase activity | IEA GO_REF:0000002 | MODIFY | Summary: This is the row where the recurring pattern of a domain name propagating into GO as an activity is actually visible in the evidence. The mapping source is IPR013094 Alpha/beta hydrolase fold-3, which is a fold signature: it constrains architecture, not which reaction the architecture performs. The signature comparison makes the difference concrete. IPR013094 also matches soybean HIDH, at residues 74-298, even though HIDH falls outside the arylacetamide-deacetylase subfamily entirely - it matches neither IPR017157 nor PIRSF037251 nor even the family-level IPR050300 - and its physiologically important reaction is a dehydration rather than a hydrolysis. So fold-3 membership reaches beyond the subfamily whose chemistry is being asserted, and within the panel the same scaffold supports amide hydrolysis (AADAC's arylacetamide deacetylation), ester hydrolysis and a dehydration. In AADACL3 the fold match is also split into two segments, 115-245 and 310-380, and that first segment is the shortest of any family member examined. The mapping nevertheless reaches a defensible conclusion here, and the reason is worth stating because it is the check that distinguishes a fold-name artefact from a real inference: the catalytic residues are present. Ser193, Asp347 and His377 read out of the sequence as the expected serine, aspartate and histidine, they sit in a canonical GDSFGG nucleophile elbow, the oxyanion loop at 119-121 is HGG, and all three align onto curated active sites in AADAC, NCEH1, mouse Nceh1, AADACL2 and AADACL4. A near-miss trap should be recorded alongside that. PROSITE PS01174, the GDXG-family putative serine active site pattern, matches AADAC, NCEH1 and AADACL2 but does not match AADACL3, which taken at face value looks exactly like the loss of catalytic machinery one would use to argue an esterase annotation away. Walking the pattern position by position shows both mismatches are flanking positions - Cys190 where the pattern wants serine or alanine, Phe194 where it wants cysteine, alanine or serine - while pattern positions 5 to 7, the G-D-S core containing the nucleophile, are intact. The right reading is a possible shift in the acyl-binding pocket and hence in substrate preference, not loss of activity. As with the phylogenetic row, the term is not wrong for its source - `hydrolase activity` is a fair statement about a hydrolase fold - and it is not proposed for removal. It is replaced because it is a proper ancestor of GO:0052689, which GOA already carries for this gene from a subfamily signature rather than from the fold, and which is therefore both more specific and better provenanced. GO:0017171 serine hydrolase activity is also true of AADACL3 on the verified triad, but a fold that includes a threonine-nucleophile member cannot deliver it, so it is raised as a PAINT node-placement recommendation in suggested_questions rather than proposed here. Reason: Correct for its source but uninformative for this gene, and derived from a fold-level signature that reaches outside the arylacetamide-deacetylase subfamily whose chemistry it is being used to assert, so it is weaker in provenance than the specific annotation it duplicates. Replaced as redundant, not as over-general for the fold it describes. GO:0052689 carboxylic ester hydrolase activity, already present on this gene from the arylacetamide-deacetylase subfamily signature IPR017157 and consistent with UniProt's EC 3.1.1.- assignment, is the better term and is supported by the verified catalytic triad. Propagation Review Root cause: EVIDENCE CIRCULAR OR REDUNDANT Sources checked: InterPro:IPR013094 · Alpha/beta hydrolase fold-3 (fold-level domain) CIRCULAR OR REDUNDANT A fold signature, not an activity. It reaches beyond the subfamily whose chemistry is being asserted: it also matches soybean HIDH at 74-298, which matches neither IPR017157 nor PIRSF037251 nor the family-level IPR050300, and whose physiologically important reaction is a dehydration rather than a hydrolysis. In AADACL3 the match is also split, 115-245 and 310-380. The activity conclusion here is carried by the intact Ser-Asp-His triad and by the subfamily signature IPR017157, not by this fold match. Proposed replacements: carboxylic ester hydrolase activity Supporting Evidence: file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md Pattern (fetched from PROSITE): `[LIVM]-x-[LIVMF]-[SA]-G-D-S-[CAS]-G-[GA]-x-[LI]-[CAVT]` file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_HUMAN (Q5VUY0) | no | - | pos 4 wants [SA], has C190, pos 8 wants [CAS], has F194 | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_HUMAN (Q5VUY0) | [(115, 245), (310, 380)] | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | HIDH_SOYBN (Q5NUF3) | - | - | 74-298 | - | |
| GO:0052689 carboxylic ester hydrolase activity | IEA GO_REF:0000002 | ACCEPT | Summary: The best-supported statement in AADACL3's GO record, and the term the two generic hydrolase-activity rows should collapse onto. Three independent lines converge on it. First the signature is subfamily-specific rather than fold-level: IPR017157 Arylacetamide deacetylase matches residues 19-403, and its PIRSF member signature PIRSF037251 scores 5.8e-82 over the same span and is InterPro's representative match for the protein. Second UniProt's own RecName carries EC 3.1.1.-, the carboxylic-ester hydrolase class that GO:0052689 corresponds to. Third, and decisively for distinguishing a family-name artefact from a real prediction, the catalytic machinery is present and correctly positioned: Ser193 in a GDSFGG nucleophile elbow, Asp347, His377, and an HGG oxyanion loop at 119-121, with all three triad residues aligning onto curated active sites in every characterised relative examined. Two honest limits on how far this should be taken. It remains a prediction: no assay of AADACL3 protein exists, UniProt records only transcript-level evidence, and Pharos classes the protein as Tdark, so nothing here demonstrates turnover of any substrate. And the family's own history warns against inferring a substrate from membership - AADAC and NCEH1 diverge enough that NCEH1 cannot rescue the sterol-acetate phenotype of the yeast AADAC ortholog - so the annotation should stay at the substrate-agnostic ester-hydrolase level and must not be sharpened to a diacylglycerol, cholesteryl-ester or arylamide term by family analogy. Reason: A well-grounded prediction rather than a fold-name artefact: subfamily-specific signature across almost the entire chain, an EC class assigned by UniProt, and a demonstrably intact Ser-Asp-His triad with the family oxyanion loop. Correctly substrate-agnostic, which is the right level for a protein with no functional data. Supporting Evidence: file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | 193 | S193 | S | yes | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md Oxyanion-hole motif 119-121 reads `HGG`; the nucleophile elbow 191-196 reads `GDSFGG`. file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | AAAD_HUMAN (P22760) | 33.3 | S193→S189/D347→D343/H377→H373 | 3/3 | 3/3 | `HGG` | file:human/AADACL3/AADACL3-bioinformatics/RESULTS.md | ADCL3_HUMAN (Q5VUY0) | 19-403 | 19-403 (5.8e-82) [representative] | 115-245; 310-380 | 74-385 | file:human/AADACL3/AADACL3-uniprot.txt DE EC=3.1.1.-; PMID:35736449 This hypothesis has been tested in yeast, where KIAA1363 (AADACL1) was unable to rescue the cholesterol acetate accumulation phenotype of the AADAC yeast ortholog Say1Δ-mutant [6], suggesting that the two enzymes have different substrate specificity, with the latter lacking sterol acetate hydrolase activity. |
Loading supporting content…
Download this section (compressed HTML)Q: Could PAINT place GO:0017171 serine hydrolase activity at a node that reaches the AADACL paralogs? The mechanism is the one property that travels well in this family, and human AADAC and NCEH1 already receive the term at their own ortholog nodes PTN002745055 and PTN002745068, but AADACL2, AADACL3 and AADACL4 receive nothing on that axis. Reading the nucleophile of every source on AADACL3's own hydrolase-activity IBA gives thirteen serines out of the fourteen that resolve to a protein, the formamidase Afmid included; the single exception is soybean HIDH with Thr164. That one member is what makes GO:0016787 rather than GO:0017171 the correct lowest common ancestor of node PTN009058710, so the term cannot simply be moved there as-is. Three routes would work. PAINT's own device for a divergent descendant is to annotate the ancestral node and mark HIDH with a NOT, which is the most direct answer and leaves the paralogs annotated; alternatively the tighter node PTN009058713, already used for the membrane annotation, contains no threonine-nucleophile member; or a sub-node of PTN009058710 excluding HIDH would serve. Two things to be clear about, since both are easy to overstate. First, this is not a claim that the existing generic terms are wrong: the donor set for the activity spans ester and amide chemistry, and the donor locations for GO:0016020 resolve to four distinct strings, so in both cases the general term is the correct ancestor of a genuinely heterogeneous set, and this review replaces the activity rows for redundancy with a subfamily-derived term rather than for over-generality. Second, AADACL3 and AADACL4 do inherit a molecular-function term already, and so does AADACL2: all three carry GO:0016787 from PTN009058710, and on that row the WITH/FROM sets are identical token-for-token across the three - so the gap is an uninformative parent rather than nothing at all. All three also already carry GO:0052689 from InterPro:IPR017157, which is what makes the remedy portable: modifying the generic row toward GO:0052689 merges two rows that already exist on each record, whereas modifying it toward GO:0017171 would introduce a term that appears on none of them and that one donor of the cited node refutes. All three paralogs now resolve that row the same way - MODIFY toward GO:0052689 on redundancy grounds, with GO:0017171 pursued as the node move asked about here rather than asserted on the row - so the question for PAINT is only where the mechanism term should sit, not which of three readings of the row is right.
Q: Why does UniProt make three different N-terminal calls across three paralogs with identical catalytic registers, and can they be reconciled? AADACL4 is annotated TRANSMEM 5-25 with a type II signal anchor and Single-pass type II membrane protein; AADACL2 is annotated SIGNAL 1-18 with location Secreted; AADACL3 is annotated with no N-terminal feature at all, no membrane or transmembrane keyword, and no SUBCELLULAR LOCATION comment, despite a comparably hydrophobic N-terminal segment and concordant Phobius and TMHMM transmembrane calls. AADACL3 is the extreme case, and the position it leaves a curator in is not that UniProt disagrees with the GO annotation but that UniProt has not made a call at all. The mouse ortholog makes the inconsistency harder to explain as biology: UniProt curates Aadacl3 (A2A7Z8), 61.9% identical with the same triad, as Membrane; Multi-pass membrane protein with three helices. The asymmetry is visible in GOA as well - AADACL4's membrane IEA cites UniProtKB-SubCell:SL-0162 and AADACL3's does not, because that mapping can only fire for an entry that already has a curated location.
Q: Is AADACL3's N-terminal hydrophobic region an uncleaved signal anchor, making it a type II membrane protein with a lumenal or extracellular catalytic domain like AADAC, NCEH1 and AADACL4, or is it a cleaved signal peptide, making AADACL3 a secreted enzyme like AADACL2? Phobius and TMHMM favour membrane integration and SignalP gives no cleavage call, but the family contains both outcomes at comparable sequence distance, so this cannot be settled from sequence alone.
Q: AADACL3 does not match PROSITE PS01174, the GDXG-family serine active-site pattern, because of Cys190 and Phe194 in positions flanking the intact G-D-S core. Do those two substitutions reshape the acyl-binding pocket and shift substrate preference relative to AADAC, or are they neutral?
Q: AADACL3 transcript is enhanced in placenta and skin, and its 3q25.1 relative AADACL2 is recorded by the Human Protein Atlas as skin-enriched, whereas the characterised members of the family are expressed in liver and intestine (AADAC) or brain, macrophage and platelet (NCEH1). Is there a skin or placental ester-hydrolase role for this branch of the family, for instance in barrier lipid processing?
Q: AADACL3 and AADACL4 form a tandem pair at 1p36.21 and are 55% identical to each other but only about a third identical to any characterised family member. Do they act on a shared substrate, and is either of them under selection in humans, given that both remain undetected at the protein level?
Experiment: Express epitope-tagged AADACL3 in a cell line that does not express it endogenously, and determine topology directly: test whether the N-terminal segment is cleaved by comparing the apparent mass of N- and C-terminally tagged constructs, test membrane integration by alkaline extraction and carbonate flotation of microsomes, test the orientation of the catalytic domain by protease protection of sealed microsomes with and without detergent, and test lumenal exposure by endoglycosidase H sensitivity of the N-glycans. Assay conditioned medium for the tagged protein in parallel, using AADACL2 as a secreted positive control.
Hypothesis: AADACL3 is a membrane-integral protein whose catalytic domain faces the endoplasmic reticulum lumen or the extracellular space, and is not secreted.
Type: cell biology
Experiment: Purify recombinant AADACL3 lacking the N-terminal transmembrane region alongside a S193A mutant and screen for hydrolase activity against a substrate panel spanning the family's known chemistry: p-nitrophenyl esters of increasing acyl-chain length, mono- and diacylglycerols, triacylglycerol, cholesteryl oleate, retinyl palmitate, 1-O-hexadecyl-2-acetyl-sn-glycerol, and arylacetamide drug substrates including flutamide and phenacetin. Any activity lost in S193A is attributable to the predicted nucleophile, which would convert the GO:0052689 prediction into an experimentally supported annotation.
Hypothesis: AADACL3 is a catalytically active serine hydrolase, and its activity depends on Ser193.
Type: enzymology
Experiment: Apply activity-based protein profiling with fluorophosphonate probes, the approach that established AADACL1/KIAA1363 as an active serine hydrolase, to lysates of human skin and placenta, and identify AADACL3 in the labelled proteome by mass spectrometry. This would also supply the first protein-level detection of AADACL3, whose protein existence is currently transcript-level only. Competition with an unlabelled probe and comparison with AADACL3-depleted material would confirm the assignment.
Hypothesis: Endogenous AADACL3 is an active serine hydrolase in the tissues where its transcript is enriched.
Type: chemoproteomics
Experiment: Compare wild-type AADACL3 with C190S and F194A single mutants, and with the double mutant that restores the PS01174-conforming pattern, across an acyl-chain-length series of p-nitrophenyl esters. A shift in chain-length optimum without loss of activity would show the substitutions tune the acyl-binding pocket; loss of activity would instead indicate they are structurally required.
Hypothesis: The Cys190 and Phe194 substitutions that break the PROSITE PS01174 match alter substrate preference rather than abolishing catalysis.
Type: enzymology
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)