ARFGEF1 (BIG1, Q9Y6D6) — review notes

Human ARFGEF1 encodes BIG1 (brefeldin A-inhibited guanine nucleotide-exchange
protein 1), a 1849-aa peripheral membrane protein of the large, BFA-sensitive
Arf-GEF family. PE 1: Evidence at protein level; 3D-structure keyword; an
OMIM disease (DEDISB, MIM 619964).

1. What the protein does

Catalysis is real, measured, and domain-mapped. UniProt places a single
SEC7 domain at residues 709..840 (ECO:0000255|PROSITE-ProRule:PRU00189),
and the activity has been measured directly on the human protein:

"A 39-kDa fragment spanning the Sec7 domain catalyzed loading of guanosine
5'-[gamma-thio]triphosphate onto class I ARFs and displayed clear sensitivity
to BFA."
PMID:10393931

The same paper establishes the drug mechanism that names the family:

"BFA did not compete with ARF for interaction with p200 but, rather, acted as
an uncompetitive inhibitor that only targeted the p200-ARF complex with an
inhibition constant of 7 microM."
PMID:10393931

Substrates are the class I ARFs. UniProt FUNCTION: Promotes guanine-nucleotide exchange on ARF1 and ARF3. The 2008 nucleolar paper restates it:

"BIG1, a brefeldin A-inhibited guanine nucleotide-exchange protein, activates
class I ADP-ribosylation factors (ARF1-3) by catalyzing the replacement of
bound GDP by GTP"
PMID:18292223

This is not a name-implied activity of the kind this campaign keeps finding.
The catalytic domain is present, the assay was done on the human protein, and
the drug sensitivity is quantitative.

BIG1 is explicitly not a GAP. GOA carries NOT|enables GO:0005096 by IDA.
The underlying result is that BIG1 inhibits another protein's GAP activity:

"the GAP activity of myosin IXb was significantly inhibited by the addition of
BIG1 with IC(50) of 0.06 microm"
PMID:15644318

and the mechanism is steric competition:

"These results suggest that BIG1 and RhoA compete with each other for the
binding to myosin IXb, thus resulting in the inhibition of the GAP activity by
BIG1."
PMID:15644318

So the GO:0034260 negative regulation of GTPase activity IDA and the
NOT GO:0005096 are two halves of one coherent result, not a contradiction.

A GEF-independent structural role at the Golgi. Depleting BIG1 fragments the
Golgi without blocking cargo export:

"suppression of BIG1 induces the formation of Golgi mini-stacks still
polarized and functional in terms of cargo export"
PMID:20360857

"BIG1 is required to maintain the normal morphology of the Golgi; BIG2 is
important for endosomal compartment integrity and cannot replace the function
of BIG1 in Golgi organization."
PMID:20360857

UniProt records the same non-catalytic reading: Required for the maintenance of Golgi structure; the function may be independent of its GEF activity.

Recruitment. Arl1 recruits BIG1 to the trans-Golgi through the DCB domain,
with a crystal structure and a mutational map:

"We find that Arl1 binds to the dimerization and cyclophilin binding (DCB)
domain in BIG1 and report a crystal structure of human Arl1 bound to this
domain. Residues in the DCB domain that bind Arl1 are required for BIG1 to
locate to the Golgi in vivo."
PMID:27373159

Four UniProt MUTAGEN entries cite this paper — K105D, Y109K, L156D and Q200E —
all recorded as abolishing the ARL1 interaction, and a second, independent
crystal structure exists PMID:27436755. A third paper shows the requirement in
cells PMID:22291037.

Oligomerisation. The same DCB domain drives homodimerisation:

"Our data demonstrate a strong interaction between DCB domains within GBF1,
BIG1, and BIG2 to maintain homodimers and an interaction between DCB and HUS
domains within each homodimer."
PMID:17640864

UniProt SUBUNIT: Homodimer (PubMed:17640864).

Scaffolding, GEF-independently. BIG1 anchors a myosin phosphatase complex:

"Reciprocal coimmunoprecipitation of endogenous HeLa cell BIG1 and BIG2 with
myosin IIA was demonstrably independent of Arf guanine nucleotide-exchange
factor activity"
PMID:23918382

"by anchoring or scaffolding the assembly, organization, and efficient
operation of multimolecular myosin phosphatase complexes that include myosin
IIA, protein phosphatase 1δ, and myosin phosphatase-targeting subunit 1, BIG1
and BIG2 serve to integrate diverse biophysical and biochemical events in
cells."
PMID:23918382

and the related AKAP role is what GO:0034237 protein kinase A regulatory subunit binding records. UniProt hedges it (Proposed to act as A kinase-anchoring protein (AKAP)) but asserts the binding experimentally:
Interacts with PRKAR1A and PRKAR2A (PubMed:12571360).

Regulation. PKA phosphorylation reduces GEF activity and PP1γ restores it:

"GEP activity of each was significantly decreased after incubation with
recombinant PKA plus ATP and restored by incubation with PP1gamma."
PMID:17360629

A separate PKA event, phosphorylation of Ser-883 together with a 711..715 NLS,
drives cAMP-induced nuclear accumulation PMID:16467138 (UniProt MUTAGEN 883,
S->A: Abolishes cAMP-induced nuclear localization.). PDE3A-containing AKAP
complexes keep local cAMP low so the GEF stays active PMID:19332778.

These two are routinely conflated and the conflation is wrong. They act on
different axes: Ser-883 phosphorylation plus the 711..715 NLS controls where the
protein is (cAMP-induced nuclear accumulation), while the PDE3A/AKAP arrangement
controls whether the GEF is active by holding local cAMP down. Treating the first
as evidence about exchange activity reads a localisation control as a catalytic one.

Traffic step. Double knockdown of BIG1 and BIG2 in human cells blocks an
identified retrograde route:

"knockdown of both BIG2 and BIG1 by RNAi causes mislocalization of a subset of
proteins associated with the TGN and recycling endosomes and blocks retrograde
transport of furin from late endosomes to the TGN"
PMID:18417613

Cell line confirmed from the PMC full text ("HeLa cells were cultured in minimal
essential medium…", Materials and Methods), so an IMP on the human gene is
species-correct.

2. Findings that changed the review

2.1 GO:0005086 ARF guanyl-nucleotide exchange factor activity no longer exists

This gene's single most characteristic fact — that it is an ARF-specific GEF
— cannot be stated in GO's molecular-function branch any more.

QuickGO /ontology/go/terms/GO:0005086/complete resolves to
GO:0005085 guanyl-nucleotide exchange factor activity, and GO:0005085
lists as secondaryIds:

GO:0005086 GO:0005087 GO:0005088 GO:0005089 GO:0005090 GO:0008321 GO:0008433
GO:0016219 GO:0016220 GO:0017034 GO:0017112 GO:0017132 GO:0019839 GO:0030676

i.e. the ARF, Rho, Ras, Ran, Rab and Rac substrate-specific GEF terms were all
merged into the generic parent. GO:0005085 now has no substrate-specific
is_a children at all — its only children are GO:1905098 (a regulation
term, negatively_regulates) and GO:0032045 (capable_of). An OLS search for
"ARF guanyl-nucleotide exchange factor activity" returns only GO:0005085.

This is the same merge the campaign already documented on the GAP side
(GO:0008060 ARF GTPase activator activity → GO:0005096). So GO:0005085 is
already maximal: ACCEPT it, do not propose a child, and record the substrate
machine-readably instead (extensions RO:0002233 has_input, and
core_functions[].substrates). Reproduced by
ARFGEF1-bioinformatics/check_terms.py → term_status.json.

Corroborating observation: across the 18 protein donors of ARFGEF1's
GO:0005085 IBA — yeast Sec7/Gea1/Gea2/Syt1/Mon2, Arabidopsis GNOM,
Drosophila Sec71/garz, human/mouse cytohesins, PSD, PSD3, IQSEC2, ARFGEF2 —
every single one holds GO:0005085 and not one holds any descendant
(donor_evidence.tsv). That is what a fully merged branch looks like from the
annotation side.

2.2 GO:0030532 small nuclear ribonucleoprotein complex is the wrong RNA class

GO:0030532 is annotated part_of by IDA from PMID:18292223. That paper is
about U3 small nucleolar RNA, not a small nuclear RNA:

"(32)P labeling of RNAs immunoprecipitated with BIG1 or nucleolin from nuclei
revealed bands of approximately 210 bases that also hybridized with U3 small
nucleolar (sno)RNA-specific oligonucleotides."
PMID:18292223

The term definitions settle it. GO:0030532 requires "at least one RNA of the
small nuclear RNA (snRNA) class"; GO:0005732 sno(s)RNA-containing ribonucleoprotein complex is "A ribonucleoprotein complex that contains an RNA
molecule of the snoRNA family and associated proteins." They are sibling
is_a children of GO:0030529 ribonucleoprotein complex, so this is not a
granularity question — it is the wrong sibling. Almost certainly a
sno → sn transcription slip. MODIFY → GO:0005732.

I did not go further to GO:0031428 box C/D methylation guide snoRNP complex.
U3 is a box C/D snoRNA, but it guides pre-rRNA cleavage, not 2'-O-methylation,
and GO:0031428's definition requires the complex be "capable of
ribose-2'-O-methylation of target RNAs". GO:0005732 is the honest level.

The paper is also careful, and the review should be too: the co-IP was abolished
by RNase A or DNase, so the association is nucleic-acid-dependent, and the
authors write only that the components "may exist together in nuclear
complexes". part_of is retained because that is the curator's call and the
term is being corrected, not the claim.

2.3 A paralog shadow set on MYO9A from PMID:15644318

GO:0017022 myosin binding appears twice on ARFGEF1 from PMID:15644318, once
with UniProtKB:Q13459 (MYO9B) and once with UniProtKB:B2RTY4 (MYO9A).
The paper's title, abstract and every described experiment concern myosin
IXb; myosin IXa is never mentioned in the abstract.

Querying GOA by reference rather than by gene shows the shape of the problem
(reference_annotations.tsv, 18 rows for this PMID):

MYO9A has acquired an exact copy of MYO9B's partner set (RHOA and ARFGEF1) but
none of MYO9B's functional rows (GO:0005096, the NOT GO:0032011) — the
signature of an accession slip rather than a second protein having been assayed.

I did not act on this as a removal. The cached record is abstract-only
(full_text_available: false) and the JBC full text returns HTTP 403, so I
cannot see what the curator saw; the campaign rule is explicit that an
experimental annotation is not overruled from an abstract. Checked and excluded
alternatives: B2RTY4's twelve secondary accessions do not include Q13459
or any of its secondaries, so this is not an accession-identity migration. The
row's term is in any case correct for BIG1 — GO:0017022 is independently
established by the MYO9B row and by the myosin IIA co-IP in PMID:23918382. So
the MYO9A row is UNDECIDED, and the observation is filed as a question for
the assigning curator, naming all four affected annotations across three genes.

2.4 The GOA record is a coverage problem, not an over-annotation problem

literature_coverage.py takes the union of the affinage citation block and
every RX PubMed= line in the UniProt entry (48 PMIDs) and asks QuickGO how
many GO annotations each reference supports anywhere. Restricted to the 26
functional papers affinage returned:

The most striking single entry is PMID:10393931, the paper that measured the
Sec7 fragment loading GTPγS onto class I ARFs and determined the BFA inhibition
constant. It is the best molecular-function experiment ever done on the human
protein and it has produced zero GO annotations. The GO:0005085 IDA on
ARFGEF1 instead cites PMID:15644318, a myosin IXb paper.

Also absent from GO: the entire PKA/PP1γ regulatory switch (16467138, 17360629),
the homodimer (17640864), both Arl1 crystal structures (27373159, 27436755), the
retrograde-transport step (18417613) and the myosin-phosphatase scaffold
(23918382). Four of these are the basis of this review's NEW rows.

2.5 GO:0006887 exocytosis is not supported for BIG1

Two rows: a TAS from PMID:10212200 (assigned by ProtInc in 2003) and an
ARBA IEA. PMID:10212200 is the BIG1/BIG2 cloning paper; its only relevant
statements are background about the family — "ARFs ... play an important role in
intracellular vesicular trafficking" and BFA "blocks protein secretion" — and
its one biochemical assay is on BIG2, not BIG1:

"BIG2, synthesized as a His6 fusion protein in Sf9 cells, accelerated guanosine
5'-3-O-(thio)triphosphate binding by recombinant ARF1, ARF5, and ARF6."
PMID:10212200

GO:0006887's definition ends "and ends when molecules are secreted from the
cell". No experiment on human BIG1 has measured the plasma-membrane fusion step.
Where BIG1 has been implicated in surface delivery it is of integral membrane
cargo — ABCA1 PMID:23220274, GABA-A receptors PMID:24198228 — which is
GO:0006893 Golgi to plasma membrane transport, a sibling that "precedes
exocytosis" by its own definition. Both rows → MARK_AS_OVER_ANNOTATED;
GO:0016192 (already annotated, IBA) is the supported level.

2.6 GO:0090303 positive regulation of wound healing overshoots the assay

GO:0090303 is defined at tissue level: "the series of events that restore
integrity to a damaged tissue, following an injury." What PMID:22084092
measured is a scratch assay on a cell monolayer:

"Treatment of cells with BIG1- or KANK1-specific siRNA interfered significantly
with directed cell migration and initial orientation of Golgi/MTOC toward the
leading edge, which was not mimicked by KIF21A depletion."
PMID:22084092

MODIFY → GO:0030335 positive regulation of cell migration. The companion row
from the same paper, GO:2000114 regulation of establishment of cell polarity,
is accepted unchanged — Golgi/MTOC reorientation is exactly what that term
covers, and it is the half of the phenotype that was measured directly.

3. Checks that came back negative — recorded so they are not re-run

• Partner identity. All 12 GO:0005515 IPI partner accessions resolve to
  reviewed (Swiss-Prot) canonical entries with lengths matching the
  canonical protein; no TrEMBL fragments, no partial ORFeome clones
  (partner_checks.tsv). Two tokens are isoform-suffixed (Q13459-2,
  Q7Z4S6-2) and both resolve to the right gene.
• NbExp inflation. Expanding the IntAct records and counting distinct
  (publication, detection-method) pairs rather than NbExp did not collapse
  any partner to a single screen logged several ways. ARFGEF2 6 experiments /
  4 publications (MI 0.84), MYCBP 6/4 (0.79), KIF21A 6/2 (0.71). The thin ones
  are NCL (1 experiment) and TBC1D22B (1 record, spoke-expanded, MI 0.35);
  NCKAP1L, TBC1D22A and rat Dpy30 have no IntAct record at all.
• Projection from a complex. No reference in ARFGEF1's GOA shows the
  ACTR8-style signature. Every functional reference annotates 1–4 entities
  (reference_scope.tsv). The three large references (BioPlex 9514
  annotations, OpenCell 2876, multimodal cell maps 3026) are paginated, so entity
  counts are reported as unavailable rather than derived from one page; all three
  contribute only GO:0005515.
• Donor evidence for propagated rows. 43 of 44 (donor, term) pairs carry the
  donor's own experimental evidence for the propagated term
  (donor_evidence.tsv). The one exception is Dictyostelium Q86KG9, which
  holds GO:0016192 by IBA only. No propagated row could be argued down on
  "the source only carries the same family-level inference".
• Self-referential IBA. GO:0007030 IBA lists UniProtKB:Q9Y6D6 in its own
  WITH/FROM. That is a PAINT curator judging the function core using ARFGEF1's
  own IMPs, not circularity. root_cause: NO_FAILURE_CORE.
• IBA node heterogeneity. PTHR node PTN008950430 carries both
  GO:0016192 and GO:0005085 and spans the whole Sec7 superfamily — large
  BFA-sensitive GEFs, cytohesins, EFA6/PSD, BRAG/IQSEC, and yeast Mon2. At that
  depth GO:0016192 vesicle-mediated transport is the genuine LCA (the donors'
  own terms are spread over GO:0006888, GO:0006890, GO:0006891,
  GO:0006893, GO:0006895, GO:0042147, GO:0016197, GO:0048193), so
  "too general" would be the wrong criticism. ACCEPT.
• A lipid-affinity screen hit deliberately not imported. IntAct records
  ARFGEF1 against PI(3)P (CHEBI:17283, pull down, PMID:23416715). That study
  reports "the identification of 681 proteins/protein complexes which interact
  with PI(3)P", ARFGEF1 has no FYVE, PX or PH domain in its InterPro content, and
  there is no GOA row. No GO:0032266 proposed.
• Retractions / errata. None of the PMIDs relied on here carries a
  retraction, erratum or expression-of-concern note in its cached record.
• Paralog cross-check. ARFGEF2 is not yet reviewed in this repository, so
  no sibling review exists to reconcile against. GBF1 carries GO:0042147 by
  IMP, which is the comparator that licenses the same term on ARFGEF1.

4. Things a curator should look at

• The GO:0005085 IDA on ARFGEF1 cites PMID:15644318; PMID:10393931 is the
  paper that actually measured the exchange reaction and the BFA inhibition
  constant on the human protein, and it carries no GO annotation at all.
• The MYO9A rows from PMID:15644318 (§2.3).
• interpro2go maps IPR000904 (Sec7 domain) to GO:0032012 regulation of ARF protein signal transduction. Every characterised Sec7-domain protein is an
  activator, and GO:0032014 positive regulation of ARF protein signal transduction exists. Whether the mapping can be sharpened family-wide is a
  question for InterPro, not something to assert from one gene — but the
  per-gene MODIFY is warranted here because the direction is measured
  PMID:10393931.
• UniProtKB:Q8K3E7 (rat Dpy30, 99 aa) is the named partner of an IPI on the
  human gene from PMID:19651892. The row is being removed as uninformative
  anyway, but the species mismatch is worth a curator's eye.
• ARFGEF1 haploinsufficiency causes DEDISB (MIM 619964) with seizures in about
  half of affected individuals PMID:34113008; the mouse model attributes the
  lowered seizure threshold to reduced surface GABA-A receptors PMID:31678406.
  The human-cell counterpart of that trafficking defect has not been shown, and
  PMID:24198228 does not state the species of the neurons used — which is why
  no GABA-A trafficking row was added here.

5. Provenance for the analyses

All scripts are committed under ARFGEF1-bioinformatics/ and run from the repo
root with uv run python. See that folder's README.md for what each produces
and RESULTS.md for the numbers quoted above.