cas3 (Cas2-3, Q02ML8) — 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_33340, 1076 aa, UniProt name "CRISPR-associated nuclease/helicase Cas3 subtype
I-F/YPEST". It is a Cas2-Cas3 fusion, not a bare Cas3:
PMID:26586803. The InterPro
complement reflects this (IPR048823 Cas3_I-F_Cas2, IPR006483 Cas3-HD,
IPR027417 P-loop NTPase), and UniProt places an HD Cas3-type domain at 102–318 and a
DEAD-box motif at 576–579.

GOA for Q02ML8 is strikingly thin: only the CACAO IMP rows on
GO:0043571. No IEA at all — no InterPro2GO, no UniRule — even though the orthologue
P38036 (E. coli Cas3) carries nine molecular-function annotations. The most likely
reason is that the fused Cas2-3 architecture does not match the signatures the
pipelines key on. That is a curation gap, not a biological absence, and it is the main
thing this review has to fix.

Role in the pathway

Cas3 is the destruction step: the Csy complex finds and verifies the target, then
recruits Cas3.
PMID:28985564. The docking site is the Csy large subunit:
PMID:32170016.

This recruitment step is the target of the anti-CRISPR AcrIF3:
PMID:26416740.
The mechanism is structurally resolved on this protein:
PMID:27455460 and
PMID:27455460.

Direct evidence for activities, in this protein

DNA nuclease activity (IDA). The cryo-EM/biochemistry study of the type I-F complex
reconstituted Csy + Cas3 + dsDNA in vitro with ATP and Mg(2+), and read out DNA
cleavage that is strictly Cas3-dependent — the no-Cas3 lane is the negative control:
PMID:32170016 and, on relief of inhibition,
PMID:32170016. The Csy complex came from the PA14
expression plasmid and the paper identifies the Cas3 as P. aeruginosa (quote above),
so this is direct evidence on the gene product being reviewed.

ATP binding (IDA). PaCas3 was crystallised with nucleotide bound — "locks PaCas3 in
an ADP-bound form" (quote above) — placing a bound adenine nucleotide in the helicase
domain of this protein.

Both catalytic centres are required in vivo (IMP). Point mutants in each domain
abolish function:
PMID:26586803, and independently
PMID:21398535. The Cas2 moiety is separately required:
PMID:26586803.

Curation decisions

• GO:0043571 maintenance of CRISPR repeat elements, IMP,
  acts_upstream_of_or_within_positive_effect → ACCEPT. Well grounded: the Cas2-moiety
  truncation and both catalytic-domain point mutants abolish spacer acquisition. The
  hedged qualifier is the right one, because in type I-F the adaptation requirement is
  partly carried by the fused Cas2 domain and partly indirect (via the Csy complex and
  crRNA).

Gaps identified (missing annotations)

Comparator check against E. coli Cas3 (P38036, genes/ECOLI/ygcB/ygcB-goa.tsv): it
holds GO:0000014, GO:0008296, GO:0033677, GO:0003690, GO:0000287,
GO:0003724 and GO:0005524, all by IDA/EXP/IBA. PA14 Cas3 holds none. Every type I
system's Cas3 is the degradation nuclease, so a systematic absence here is one-protein
pipeline failure, not an ontology convention. Proposing, conservatively:

• GO:0004536 DNA nuclease activity (NEW, IDA, PMID:32170016). Deliberately the
  moderately specific term, not E. coli's GO:0000014 single-stranded DNA endonuclease activity or GO:0008296 3'-5'-DNA exonuclease activity: the in vitro assay used
  dsDNA presented by the Csy complex and did not resolve exo- versus endonucleolytic
  mode or strand preference for this orthologue. Asserting the E. coli-specific
  children would be transferring a result I cannot cite for PaCas3.
• GO:0004386 helicase activity (NEW, IMP, PMID:26586803 + PMID:21398535, with
  PMID:27455460 for the structure). The DEAD-box point mutant D576N abolishes function
  in vivo and the AcrF3 co-crystal shows a nucleotide-bound helicase domain with a
  DNA-binding tunnel. Again the generic term rather than E. coli's GO:0033677 DNA/RNA helicase activity: no unwinding assay on PaCas3 has established the substrate pairing.
• GO:0005524 ATP binding (NEW, IDA, PMID:27455460).
• GO:0099048 CRISPR-cas system (NEW, IMP, PMID:26586803). The term's definition names
  "target interference" as one of its three stages, which is the step Cas3 performs.
  Checked via QuickGO: not an ancestor or descendant of GO:0043571.

Not proposed, and why

• GO:0051607 defense response to virus. E. coli Cas3 carries it by IMP and IDA, but
  the PA14-specific phenotype runs the other way:
  PMID:21398535. Later work shows the
  locus does restrict other phages
  PMID:23242138, but that is a
  whole-locus (ΔCR/cas) deletion, not a cas3 allele. With a published negative result
  for this gene and no gene-specific positive one, I leave the term off and raise it as
  a question. GO:0099048 carries the immunity role without the contested claim.
• GO:0046872 metal ion binding / GO:0000287 magnesium ion binding. UniProt's Mg(2+)
  sites at 124 and 220 are ECO:0000255 (sequence-model) only, and I have no direct
  metal-binding measurement for PaCas3.
• The biofilm phenotype (DMS3-dependent loss of biofilm formation) is a strain- and
  prophage-specific downstream consequence of target recognition, not a function of
  Cas3. Recorded here, not annotated.