acrIF3 / JBD5_035 (L7P7R7, Pseudomonas phage JBD5) — curation notes

Identity

Unreviewed TrEMBL entry, 139 aa, named only RecName: Full=Phage protein {ECO:0008006|Google:ProtNLM} — a machine-generated placeholder. PE 1: Evidence at protein level. Seeded taxon was wrong (NCBITaxon:10663); UniProt gives
OX NCBI_TaxID=1223261 for Pseudomonas phage JBD5 and the review is corrected to that.

Despite the uninformative name, this is the assayed AcrF3. The entry carries three
PDB cross-references, all with chains spanning residues 1-139 of this accession:

and the genome reference is the anti-CRISPR discovery paper itself
(RX PubMed=23242138). InterPro IPR049085 AcrF3-like, Pfam PF21401 AcrF3, CDD
cd22238 AcrIF3 all agree. So two independent groups solved the AcrF3–Cas3 complex using
this sequence, and experimental evidence codes are appropriate.

This is a textbook instance of the second knowledge gap in
modules/anti_crispr_suppression.yaml: InterPro has a mechanism-bearing family name
("Cas3 inhibitor AcrF3-like") while UniProt still calls the protein "Phage protein". The
entry should be nominated for Swiss-Prot review.

Mechanism: the module's Cas3-sequestration assignment holds

The module (annoton acrif3_cas3_sequestration) asserts GO:0140311 protein sequestering activity, with the target being host type I-F Cas3. The assignment holds and is well
supported by two independent structures.

Wang X et al.
PMID:27455460,
PMID:27455460.

Wang J et al. independently
PMID:27585537,
and add a consequence the first paper did not report — inhibition also blocks primed
spacer acquisition
PMID:27585537.
That second point is biologically significant: AcrF3 does not merely spare the current
infection, it prevents the host from updating its CRISPR memory against the phage.

Bondy-Denomy et al. established the mechanism class biochemically before either structure
PMID:26416740,
and noted the striking downstream effect that the inhibited system is converted into a
DNA-binding-only module
PMID:26416740.

GO:0140311 protein sequestering activity is defined as "Binding to a protein to prevent
it from interacting with other partners or to inhibit its localization to the area of the
cell or complex where it is active." Both clauses apply literally: AcrF3 masks the linker
and CTD that Cascade uses to recruit Cas3, so Cas3 cannot interact with its partner and
cannot localise to the R-loop where it is active. This is the one anti-CRISPR mechanism
in the module for which GO already has a precise molecular-function term, and no new
term is proposed here.

Note that in type I-F Cas2 and Cas3 are a single fusion protein
PMID:27585537,
so the sequestered entity is Cas2/3; the term is unaffected.

GOA state: one annotation, and it is an over-annotation

GOA holds exactly one row: enables GO:0046872 metal ion binding, IEA from
GO_REF:0000043, keyword KW-0479 (Metal-binding). The keyword chain is traceable and
the chain is the problem:

PDB 5GNF → UniProt FT BINDING records at residues 2, 2, 3, 65, 66, 70 for three Ca(2+)
ions, all with ECO:0007829|PDB:5GNF (i.e. automatic annotation from PDB ligand records)
→ keywords KW-0106 Calcium and KW-0479 Metal-binding, both with the same PDB-derived
evidence → GO:0046872.

Reasons to treat this as an over-annotation rather than a function:

  1. The paper that produced 5GNF never ascribes any role to calcium. Searching
    PMID:27585537 for "calcium", "Ca2+" or "ion" returns no statement about a metal
    requirement, metal dependence, or a metal site; the mechanism it reports is entirely
    protein-protein (dimerisation plus masking of the Cas3 HD, Linker and CTD domains).
  2. The sites look crystallographic. Two of the three Ca(2+) are coordinated by
    residues 2 and 3, i.e. the extreme N-terminus, and the third by the surface stretch
    65/66/70. Surface and terminal Ca(2+) at a 1.50 Å resolution are the classic signature
    of ions picked up from the crystallisation liquor.
  3. The independent structure disagrees. The 2.60 Å complex 5B7I (PMID:27455460)
    contains no calcium, and its abstract describes the nucleotide state of Cas3 (ADP) as
    the functionally relevant ligand, not a metal on AcrF3.
  4. The mechanism needs no metal. Sequestration by steric masking is not a chemistry
    that requires a cofactor.

Action: MARK_AS_OVER_ANNOTATED rather than REMOVE. The Ca(2+) density is a real
observation in a real deposited structure, so saying "metal ion binding is false" would
overstate the case; what is unsupported is the inference that metal binding is a function
of this protein. This is an IEA keyword mapping, not an experimental annotation, so
challenging it does not run against the rule about deferring to curators who read full
texts — and the relevant full text is cached and has been read.

Not asserted

No Aca repressor is assigned to the JBD5 locus in the evidence reviewed here, so the
module's part 4 is not annotated for this gene.

2026-10-25 comparative anti-CRISPR audit

PMIDs:26416740,27455460,27585537 establish Cas2/3 sequestration, preserved target recognition and loss of recruitment/primed acquisition. PDB 5B7I/5GNF/5GQH map this 139-aa accession. The mechanism is established for the tested type I-F machinery; an assertion that protein sequestration is possible only in class 1 systems was removed. Crystallographic calcium contacts remain over-annotated as a functional metal-binding claim, not declared nonexistent.

Existing GOA assertions were preserved. Earlier reviewer-authored NEW proposals were reassessed independently. Primary references were obtained through the normal publication cache. A normal Falcon attempt with perplexity-lite fallback failed; logs are in .cache/module-audit-logs/anti-crispr-deep-9CAUD-acrIF3.log. These are manual research notes, not provider-generated research.