ACBD3 binds SREBP1 directly and blocks its intracellular maturation - not by disrupting the SREBP1-SCAP-Insig1 complex, but by engaging the regulator itself.

The FUNCTIONAL effect is attributed to the ACB-domain-containing N-terminal region, and only partially: deleting it attenuates ACBD3's suppression of the FASN promoter (40% vs 76%) without abolishing it, and the paper's own wording is "plays an important role", not "maps to". The SREBP1 co-IP itself was done with full-length ACBD3 and carries no domain mapping at all. So the ACB region contributes to a function that is separate from acyl-CoA binding as such, without being shown to be where that function resides.

The lipogenic output, stated with the paper's own timing caveat: de novo palmitate synthesis is unchanged at 48 h of ACBD3 overexpression and reduced only at 72 h and at higher doses. An overexpression phenotype, so IMP-grade at best, and quoted here at the resolution the experiment supports.

The binding assay, for the record, is full-length ACBD3 against the SREBP1 precursor - specific for the precursor over the mature nuclear form, but with no domain resolution.

The N-terminal deletion attenuates rather than abolishes, which is why the domain attribution is reported as partial.

And the residual activity is stated explicitly by the authors.

The loss-of-function direction: nuclear SREBP1 rises when endogenous ACBD3 is depleted, in two cell lines. Together with the overexpression result this makes the perturbation bidirectional.

With the authors' conclusion, and the word they choose - "intrinsically".

The Q-domain switch is this paper's central result, mapped rather than asserted: fine-scale mapping shows the TBC1D22A/B and PI4KB binding determinants within ACBD3 are the same.

And the two interactions are mutually exclusive, which is what makes the domain a switch rather than a shared dock. This is the primary source behind UniProt's mutual-exclusivity statement and behind the MODIFY on the TBC1D22 rows.

The differential effect that gives the paper its title: enteroviral and kobuviral 3A proteins engage ACBD3 with different consequences for TBC1D22A/B and PI4KB recruitment - which is why the nine WITH/FROM accessions on this row span both virus groups and both host partners.

The reference behind ACBD3 acyl-CoA binding, reached via PMID:38134218. Its subject is ACBD6, and the ACBD3 datum survives here only inside a contrastive clause: ligand binding dimerises ACBD1 and ACBD3 but not ACBD6. Deliberately not stated more strongly than that - the record is abstract-only, so whether ACBD3 was assayed in this work or the contrast points at prior literature cannot be determined from what is readable. The weight of the claim is carried instead by PMID:38134218's discussion, where an independent 2024 group asserts of this same reference that ACBD3 has been demonstrated to oligomerise with C18:1-CoA or C16:0-CoA.

The structural basis of ACBD3's core molecular function: an NMR structure of the ACBD3:PI4KB complex. This is structural support for the Q-domain assignment, but not independent of it: the paper follows Greninger's mammalian two-hybrid mapping (PMID:23572552) and describes itself as confirming it, so the constructs were chosen from that result. Methodologically independent, evidentially downstream. UniProt's DOMAIN line cites PubMed:23572552, not this paper.

The construct fixes the domain: the NMR complex is ACBD3 residues 241-308 - the Q domain - with the PI4KB N-terminus, not the GOLD domain. Worth recording explicitly, since an earlier draft of this review mis-assigned the interaction to GOLD on a provider sentence.

The functional consequence, and the basis of GO:0043495 plus the proposed GO:0034067: ACBD3 recruits PI4KB to membranes in vitro and in vivo, recruitment raises the kinase's activity, and the complex is required for normal Golgi function.

The provenance, in the paper's own words - it localises the interaction to the Q domain following Greninger and colleagues.

BioPlex 2.0, a proteome-scale affinity-purification/mass-spectrometry survey rather than a study of ACBD3. Its contribution here is one WITH/FROM accession, TBC1D22B (Q9NU19), corroborating a partner that PMID:23572552 characterised mechanistically. Screen-scale support, and treated as such.

ACBD3 is a FAPP2 interacting partner - the primary source for the FAPP2 arm, fetched after a reviewer noted that PMID:38134218 attributes FAPP2 recruitment to the ACBD domain while this review credited it to the gene with no domain attached.

The mechanism as this paper frames it is INDIRECT, which matters for how the ACB attribution should be read: ACBD3 knockdown fragments the Golgi, and FAPP2 dispersal follows from that. The title says the same - "through maintaining the Golgi integrity". So "the ACBD domain recruits FAPP2" is PMID:38134218's compression of this work, not a domain mapping made here; the cached abstract contains no ACB-domain assignment at all.

With the rescue control that makes the knockdown interpretable.

Read the results paragraph, not the abstract: it opens by assigning the ACB domain TWO functions with primary citations - long-chain acyl-CoA binding (ref 29) and SREBP1 binding with reduced de novo palmitate synthesis (ref 30, = PMID:23166793). The acyl-CoA half is family-level (ACBD1 to ACBD7) and hedged as "suggested", but both resolve to real primary evidence, and together they are why GO:0000062 is now KEEP_AS_NON_CORE rather than MARK_AS_OVER_ANNOTATED. The abstract, which this review originally quoted from, carries neither.

The sufficiency form of the domain result, tighter than "ACB and CAR are dispensable": the Q and GOLD domains together suffice to support enterovirus replication.

And the internal negative control in the same figure: GOLD alone rescues neither replication nor PI4KB recruitment. So it is Q+GOLD, not GOLD - which independently requires the Q domain for PI4KB recruitment, consistent with PMID:27009356's NMR mapping.

Independent corroboration for the shared-surface conclusion, which otherwise rests on PMID:38134218 alone: this paper states that enterovirus and kobuvirus 3A proteins bind the GOLD domain most probably at the same site as giantin. Hedged as "most probably", and recorded as such.

ACBD3 is upstream of PI4KB recruitment in the viral setting: no PI4KB recruitment occurs without ACBD3. Consistent with the reciprocal-knockout direction control in PMID:38134218 and with the proposed GO:0034067 annotation.

The ACBD3-PI4KB interaction itself, not merely ACBD3's presence, is what the virus needs: an ACBD3 mutant unable to bind PI4KB restores 3A localisation but not replication.

Structures of ACBD3 bound to the 3A proteins of four enteroviruses (poliovirus, EV-A71, EV-D68, rhinovirus B14), which is the structural basis for treating the GOLD/UR surface as the viral interface.

The convergence result: enteroviruses and kobuviruses, distinct picornavirus groups, arrive independently at the same ACBD3-based recruitment mechanism. That recurrence is the argument that ACBD3's membrane-adaptor activity, not an incidental surface, is what is being hijacked.

BioPlex 3.0 plus an HCT116 network - again proteome-scale rather than ACBD3-specific. It contributes two WITH/FROM accessions, TBC1D22A (Q8WUA7) and TBC1D22B (Q9NU19), so it corroborates both paralogs where the PMID:28514442 row carries only TBC1D22B.

The GOLD domain of ACBD3 binds PKA regulatory subunit RIIalpha and recruits the PKA holoenzyme to the Golgi; RIalpha is not detected in the same GST pull-down. The authors do not treat that as excluding RI - see the next finding. What the result does contradict is UniProt's (By similarity) assertion that ACBD3 does NOT interact with RII-alpha.

The paper reconciles its own RII-only pull-down with RI binding, so the two records are not in flat conflict: a typical dual-specific AKAP binds RI 10-100 fold more weakly than RII, RIalpha binding may therefore sit below the assay's detection limit, and the authors attribute the organelle difference to RI and RII being enriched at mitochondria and the Golgi respectively.

The compartment split stated explicitly - which is also why GO:0005739 is kept non-core and excluded from the AKAP core function's locations.

ACBD3 makes retrograde transport cargo-dependent: its depletion reduces Golgi RII and renders PKA activation and KDEL-receptor retrograde transport constitutive, independent of cargo influx from the ER. This is the basis of the proposed GO:2000156 annotation.

What the mutagenesis establishes is the alpha helix, not a residue. The scan is a proline scan of the single alpha helix at the unique-region/GOLD boundary: K381P greatly reduces RIIalpha binding, Q379P and I380P have almost no effect, and the authors read this as K381P disrupting the helical structure that carries the interface. So the RIIalpha-binding interface is that helix, spanning roughly 379-383; the scan identifies no individual contact residue.

Why the scan cannot name a contact even in principle: the substitutions are prolines, chosen because proline breaks helices, so a loss of binding reports on secondary structure rather than on a side-chain contact.

Ile380 is the authors' proposed contact within that helix, and it is an inference rather than a measurement: it comes from inspecting the crystal structure of the GOLD domain alone, with no R subunit bound, is hedged as "likely", and is offered to explain why the buried Phe383 does not make ACBD3 RI-selective. Its only direct test in the same paper, I380P, had almost no effect - so I380 is a hypothesis about the contact, not a mapped one.

The structure inspected is apo ACBD3 GOLD (PDB 5LZ1), which is why the Ile380 argument is a solvent-accessibility inference rather than an observed interface.

The GOLD domain carries a single alpha helix, which is why "the helix" and "the RIIalpha-binding interface" are the same object in this paper.

Alanine scanning of the 21-residue unique region (UR) immediately upstream of the GOLD domain identifies MWT374-376 as the Golgi-targeting motif; MWT374-376>AAA relocalises ACBD3 to the cytosol. The paper calls this locus "the UR of the GOLD domain", so "MWT motif" and "GOLD domain region" are two resolutions of one site, not rival claims.

Giantin and golgin-45 recruit ACBD3 redundantly through that motif: single giantin knockout does not displace ACBD3, only the giantin/golgin-45 double knockout does, and the MWT374-376>AAA mutant co-immunoprecipitates neither golgin. This is what reconciles the older giantin model with the observation that giantin loss alone has no effect.

The redundancy is stated directly: loss of both golgins, not either alone, removes ACBD3 from the Golgi.

The paper places UniProt's GOLD-domain/giantin attribution in context: it originates in a yeast two-hybrid interaction (Sohda 2001, PMID:11590181), and giantin knockdown does not abolish ACBD3 Golgi recruitment. So the UniProt DOMAIN line is not contradicted so much as refined - same locus, finer resolution, and a redundant second golgin.

SCFD1 (SLY1), a Sec1/Munc18 protein, is a previously unrecognised and essential upstream recruitment factor: its CRISPR knockout strips ACBD3 from the Golgi, and with it PI4KIIIbeta.

The GOLD domain plus its unique region binds the longin domain of the v-SNARE SEC22B; the SNARE and transmembrane domains do not bind. Basis of the proposed GO:0000149 annotation, and evidence that GOLD is not a viral-only surface.

The two mechanisms are sequential rather than parallel: ACBD3-giantin binding falls in SCFD1-knockout cells while ACBD3-SCFD1 binding is unaffected in the golgin double knockout, putting SCFD1 upstream.

ACBD3 is required for PI4KIIIbeta to reach the Golgi, and the dependency is one-way: PI4KIIIbeta knockout does not move ACBD3. Basis of the proposed GO:0034067 annotation.

The picornaviral 3A peptide binds the SAME locus as the golgins, not a separate one: the paper places 3A recruitment at "the UR of the GOLD domain", its Figure 1D puts the 3A contact residues I380/K381 within the unique region a few residues from MWT374-376, and the study's whole premise is that 3A must outcompete the endogenous Golgi anchor. Combined with PMID:37044218, whose RIIalpha-binding interface is the single alpha helix spanning roughly 379-383, residues 374-381 are a shared hub for the golgins, PKA RII and 3A - overlapping surfaces, not an identified common contact residue - so the hijack is best read as competition with ACBD3's own anchors, while Q-domain PI4KB binding is untouched.

Figure 1D names both picornaviral 3A contact residues explicitly, and both lie inside the alpha helix that PMID:37044218 identifies as the RIIalpha-binding interface - so the PKA and viral sites overlap as surfaces. Whether they share a contact residue is untested: I380 is that paper's proposed RII contact, but its own I380P substitution had almost no effect.

A limit on the upstream step: SCFD1 knockout is clean (ACBD3 becomes cytosolic), but SEC22B knockout in the same screen fragments the Golgi, so SEC22B's specific requirement for ACBD3 recruitment is confounded. The review therefore annotates ACBD3-SEC22B binding but attributes the requirement to SCFD1 only.

Machine-fetched Affinage record synthesising 30+ primary papers on ACBD3. It does not mention acyl-CoA anywhere - which earlier versions of this review wrongly treated as evidence that none had been reported. It is a coverage gap, not a finding: see the reference_review below.