csy2 / Cas5f (Q02MM0) — Pseudomonas aeruginosa UCBPP-PA14 — curation notes

Deep research was unavailable in this environment (no provider API keys), so this file
is my own literature journal. Every assertion carries an inline PMID plus a verbatim
quote from the cached record in publications/.

What this protein is

PA14_33320, 327 aa, "CRISPR-associated protein Csy2" = Cas5f. It sits with Cas8f in
the PAM-proximal tail of the surveillance complex:
PMID:32170016, within
PMID:32170016. The PA14 complex was first characterised biochemically as
PMID:21536913. GOA's two
GO:0005515 rows are the IntAct Csy2–Csy1 pair (Q02MM0–Q02ML9, NbExp=6).

Structural work is on the P. aeruginosa complex:
PMID:28985564.

What Cas5f does

Caps and reads the crRNA 5' handle. This is Cas5f's defining job across type I:
PMID:32170016, and critically the contacts are base-specific, not backbone-only:
PMID:32170016.

Contributes a thumb to the target-duplex pocket.
PMID:28985564, with one Cas5f residue in the pocket:
PMID:28985564. Note the "especially": the DNA contacts are
dominated by Cas8f. I therefore do not propose a DNA-binding term for Cas5f — see
below.

Required in vivo for crRNA. Csy2 is in the strict-requirement class:
PMID:21398535. UniProt records this as "Absolutely required for
crRNA production or stability" and, for the disruption mutant, "Loss of production of
crRNA" — a loss, not a decrease, unlike csy1 and csy3. Mechanistically this is
assembly-dependent stabilisation: unprotected crRNA is degraded, and
PMID:22522703.

Adaptation also needs the intact complex:
PMID:26586803.

Curation decisions

• Both GO:0005515 protein binding rows (IPI, partner Csy1) → MODIFY to
  GO:0005198 structural molecule activity. The Csy2–Csy1 IPI samples a contact
  internal to the complex, where Cas5f and Cas8f jointly clamp the crRNA 5' handle and
  their thumb and hook form two of the three walls of the target-duplex pocket.

Aspect correction (follow-up). My first pass proposed
part_of GO:0032991 protein-containing complex as the replacement, reasoning from the
E. coli comparator where all five Cascade subunits carry exactly that. That was wrong
in form: these GOA rows carry qualifier: enables, so they are molecular-function
rows, and a gene product cannot enable a cellular component. However
well-supported complex membership is, it cannot be expressed by swapping a CC term
into an enables MF row. The E. coli reviews were not making this mistake — their
GOA already contained separate part_of GO:0032991 rows, which they MODIFY to
GO:1990904; their enables GO:0005515 rows went to GO:0005198 instead.
Reconciled to the same convention:

• MODIFY the enables rows to GO:0005198 structural molecule activity ("the
  action of a molecule that contributes to the structural integrity of a complex").
  Aspect-correct, and none of its 22 children fits a non-ribosomal ribonucleoprotein,
  so the parent is the right level. Each row is backed by the verbatim structural
  quote naming the contact this subunit actually makes.
• Carry complex membership on its own NEW part_of GO:1990904 ribonucleoprotein complex row. PA14 has no part_of complex row for any Csy subunit, unlike all
  five E. coli Cascade subunits, so this is a real missing annotation rather than a
  reformulation. GO:1990904 rather than the bare GO:0032991 root because the
  60-nt crRNA is an integral component, not a ligand.
• core_functions[].in_complex → GO:1990904 for the same reason.
• Added a proposed_new_terms entry for a "CRISPR RNA-guided surveillance
  complex" CC term under GO:1990904, identical in substance to the one the five
  E. coli Cascade reviews carry, so the two systems request one term rather than two.
  GO's entire CRISPR vocabulary is five BP terms (GO:0099048, GO:0043571,
  GO:0098672 and the obsolete GO:0110132/GO:0110133); ComplexPortal models the
  type I-E complex as CPX-1005, GO does not.

Gaps identified (missing annotations)

• GO:0003723 RNA binding (NEW, IDA, PMID:32170016). Comparator: casD, the E. coli
  Cas5, carries GO:0003723 by IDA twice (PMID:25103409, PMID:18703739) plus an IEA;
  casA, casC and casE carry it too. Here the sequence-specific 5'-handle recognition is
  directly resolved.
• GO:0043571 maintenance of CRISPR repeat elements (NEW, IMP, PMID:21398535;
  involved_in). Comparator: casD carries GO:0043571 (IEA, InterPro IPR021124),
  and the PA14 paralogues cas6f and csy3 already hold it. The term's definition
  explicitly covers "transcription of the CRISPR repeat arrays into RNA and processing",
  which is what a csy2 null abolishes. This is the one place where the PA14
  gene-by-gene data are sharper than the pipelines: Csy2 is in the essential class for
  crRNA, Csy1 and Csy3 are not, and only Csy2 lacks the term.
• GO:0099048 CRISPR-cas system (NEW, IDA). Comparator: all five E. coli Cascade
  subunits carry it by IDA (PMID:18703739); no PA14 Csy subunit does. Cas5f performs
  part of the crRNA-biogenesis and interference stages named in the term's definition
  (it holds the 5' handle that defines guide register). Checked via QuickGO that
  GO:0099048 is not an ancestor or descendant of GO:0043571 or GO:0051607.

Not proposed, and why

• GO:0003677 DNA binding / GO:0003690 double-stranded DNA binding. Only a single
  Cas5f residue (R90) is placed in the duplex pocket and the paper's own emphasis is on
  Cas8f. Comparator: E. coli casD does not carry a DNA-binding term — it carries
  GO:0071667 DNA/RNA hybrid binding instead, which reflects the Cas5 position on the
  5' handle rather than on the duplex. Rather than transfer GO:0071667 on that
  analogy, I leave it off: the type I-F hybrid is pinned by Cas7f thumbs, and I have no
  PA14 Cas5f-specific hybrid-contact quote.
• GO:0051607 defense response to virus. No csy2 allele has been assayed for phage
  resistance, and the PA14 locus-level picture is contested (see cas3-notes.md).
• No molecular-function term exists for crRNA-guided target recognition, the activity
  the assembled complex has; recorded as an ontology gap in
  modules/crispr_cas_adaptive_immunity.yaml and raised in suggested_questions
  rather than papered over with an ill-fitting id.