Bacillus phage phi29 gene 15 (gp15 / p15, endolysin) — curation notes

UniProt: P11187 (ENLYS_BPPH2), 258 aa, EC 3.2.1.17. Host: Bacillus subtilis
(NCBI:txid1423). Reviewed for the CACAO project (batch 3, screen 1).

Deep research was unavailable for this review. No deep-research API keys
exist in this container, so just deep-research was not run and no
-deep-research-<provider>.md file exists. Everything below was gathered by
reading the cached publications, by PubMed E-utilities / web search, and by
live queries to UniProt, InterPro and QuickGO. Both papers cited by the two
CACAO rows are abstract-only in the cache (full_text_available: false),
and neither is in the PMC open-access subset (EuropePMC fullTextXML and the
NCBI BioC endpoint both return nothing for PMC304338 and PMC193017), so no
figure-level reading was possible.

GONUTS is gone, not offline. Batches 1 and 2 recorded gowiki.tamu.edu as
"temporarily offline while we perform upgrades". Queried again on 2026-11-01 it
now returns HTTP 410 Gone with the page "GONUTS, the GO wiki, has been
retired" — "The GO wiki that was at gowiki.tamu.edu is no longer online" —
and a per-protein page (/wiki/index.php/UniProtKB:P11187) also returns 410.
The Internet Archive has no snapshot of the phi29 pages. The project's standing
fallback for abstract-only CACAO rows is therefore permanently unavailable, not
pending; this should be recorded in projects/CACAO.md.

What the protein is

Gene 15 is the second gene of the phi29 lysis cassette; gene 14, immediately
upstream, is the holin. The pair was sequenced together
PMID:3027653, and gp15 is a phage-type
lysozyme: PMID:3027653.

The activity was established by cloning and purification, not by mutant
analysis: PMID:3469652 and PMID:3469652. The same paper notes
PMID:3469652 — the unusual feature being the two
C-terminal LysM repeats appended to the GH24 catalytic domain
(file:BPPH2/15/15-uniprot.txt "LysM domains are thought to be involved in
peptidoglycan binding.", two PF01476/PS51782 matches in the record).

UniProt curates the mechanism as the canonical holin/endolysin sequence:
[file:BPPH2/15/15-uniprot.txt "Endolysin with lysozyme activity that degrades
host peptidoglycans and participates with the holin and spanin proteins in the
sequential events which lead to the programmed host cell lysis releasing the
mature viral particles."] and [file:BPPH2/15/15-uniprot.txt "The endolysin is
cytoplasmic, but can reach the periplasmic space with the help of the holins
which disrupt the host cell membrane."]. Family placement:
[file:BPPH2/15/15-uniprot.txt "Belongs to the glycosyl hydrolase 24 family."],
[file:BPPH2/15/15-uniprot.txt "Pfam; PF00959; Phage_lysozyme; 1."].

The holin partner is independently established: PMID:8432697 and PMID:8432697. The division of labour is stated flatly
in PMID:7831803.

One caution worth noting but not acting on: the entry records
[file:BPPH2/15/15-uniprot.txt "Lacks the conserved Asp active site."]. This
is a HAMAP sequence observation; it does not survive as a doubt about activity,
because the purified protein was assayed and is a lysozyme (PMID:3469652).
Only Glu15 is annotated as proton donor/acceptor.

Screen 1 (adjacent-locus confusion): CONFIRMED on the GO:0044659 CACAO row

The CACAO row GO:0044659 viral release from host cell by cytolysis, IMP,
acts_upstream_of_or_within_positive_effect, PMID:820555 (2025-02-17) is the
screen's target. PMID:820555 is Carrascosa et al. 1976, a classical
characterization of phi29 gene products by sus mutants. Its abstract offers one
sentence about lysis: PMID:820555 and PMID:820555. The cistron is not named in the
abstract.

Five independent lines say that cistron is gene 14, not gene 15:

1. The sus mutant that exists in this region of the phi29 collection is a
   gene-14 mutant: PMID:3027653 There is no sus15.
2. Gene 14's product is the one whose loss impairs lysis: PMID:8432697
3. PMID:7831803 A holin mutant gives exactly the delayed-lysis-with-normal-burst
   phenotype the 1976 abstract describes, because the endolysin accumulates in
   the cytoplasm but cannot reach the murein.
4. gp15 was not characterized by mutant phenotype at all; it was cloned,
   overproduced and assayed eleven years later (PMID:3469652). A 1976
   mutant-phenotype paper cannot be the source of an IMP on gene 15.
5. The decisive internal evidence: the same CACAO curator made two rows from
   that one sentence. projects/CACAO/data/cacao-quickgo-2026-10-28.tsv shows
   GO:0044659 IMP PMID:820555 on P03682 (gene 4, the late-genes transcription
   activator p4, 2025-02-20) as well as on P11187 (gene 15, 2025-02-17). One
   cistron in the paper, two genes annotated, and the actual lysis cistron
   (gene 14, UniProtKB:P11188, HOLIN_BPPH2) carries no row at all. At least
   one of the two is misattributed, and neither is gene 14.

A cross-check requested by the batch coordinator, who is also reviewing gene 17
and gene 4: the choice is not between gene 15 and gene 4. Both are wrong, because
the cistron is gene 14. The later phi29 literature attributes the 1976 abstract's
exact phenotype — delayed lysis under restrictive conditions with a normal or
larger burst after artificial lysis — to the named alleles sus14(1241) and
sus14(1242), and Steiner et al. report that phi29-induced lysis in B. subtilis
is only delayed when protein 14 is absent. So the gene-4 row and the gene-15 row
are both misattributions of one sentence, and the gene that earned it has no
CACAO row. Confidence is recorded as PROBABLE rather than CONFIRMED only because
the 1976 full text itself could not be read (Eur J Biochem 1976 is not in PMC).

On the qualifier: all ten of phi29's 2025-dated CACAO rows use
acts_upstream_of_or_within_positive_effect, so the hedge is a campaign habit
and has to be judged per gene rather than condemned wholesale — on gene 17 it
reportedly fits the evidence better than involved_in. It does not fit here:
gp15's lysozyme activity was measured directly on purified protein, and that
activity is the cytolytic step of GO:0044659, so involved_in is right.

So the shape predicted by screen 1 is present. It behaves as it did on T4
gp5.4: the term is independently true of gene 15 — gp15 is the endolysin that
performs the peptidoglycan hydrolysis releasing virions, and the UniRule IEA row
in the same GOA file carries GO:0044659 on sound grounds — while the stated
evidence belongs to the neighbouring gene. Under the curator-deference rule the
row is kept (ACCEPT, with the defect filed), not removed: the full text was not
readable, and removing it would drop a true claim.

The same row is a second, independent hit on screen 3: the qualifier
acts_upstream_of_or_within_positive_effect is wrong for gene 15, which
executes the murein-hydrolysis step of lysis rather than acting upstream of it.
(The qualifier fits the evidence — a mutant delaying a downstream outcome —
which is itself a sign that the row was written from a gene-14 phenotype.)

The other CACAO row, GO:0003796 lysozyme activity IMP PMID:3469652
(2012-11-16), is about the right gene and the right paper; only the evidence
code is wrong. PMID:3469652 assays a purified protein, with no mutant anywhere
in it, so the code should be IDA. Same shape as the SWP7 and T5 D15 rows in
batch 2.

Endolysin own-process shape: PRESENT, in three places and from two routes

This replicates the PRD1 P15 (genes/BPPRD/XV) and phage K LysK
(genes/BPPGK/lysK) findings, and extends them.

• GO:0009253 peptidoglycan catabolic process, IEA, GO_REF:0000120, from
  UniRule:UR000433564 + InterPro:IPR002196/IPR034690. Ancestry checked in
  QuickGO: GO:0009253 sits wholly under GO:0009987 cellular process via
  peptidoglycan/aminoglycan/macromolecule catabolism, with no interspecies
  parent. phi29 has no peptidoglycan of its own.
• GO:0016998 cell wall macromolecule catabolic process, IEA, GO_REF:0000002,
  from InterPro:IPR002196. Ancestry checked in QuickGO: GO:0044036 →
  GO:0071554 cell wall organization or biogenesis → GO:0009987 cellular process. Same defect, different term — this one was not seen on XV or lysK.
• GO:0042742 defense response to bacterium, IEA, GO_REF:0000043, from
  UniProtKB-KW:KW-0081. Third instance of the inverted-direction keyword
  mapping after XV and lysK.

Both own-process terms take GO:0052009 symbiont-mediated disruption of host cell wall (QuickGO: under GO:0098933 symbiont-mediated disruption of host cell envelope → GO:0044419), the same replacement the two comparator reviews used.

Two things are new relative to the comparators:

• The rule route, not just InterPro2GO. UniRule:UR000433564 is HAMAP
  MF_04110, whose name is ENDOLYSIN_T4 (queried live at
  rest.uniprot.org/unirule): a rule whose entire scope is phage endolysins.
  Its GO set is GO:0003796, GO:0044659 and GO:0009253. Every protein
  the rule fires on is a phage endolysin acting on another organism's wall, so
  the own-process term is wrong for 100% of its targets — a stronger case than
  lysK's, where IPR002502 Amidase_2 is genuinely shared with bacterial
  autolysins whose substrate is their own sacculus.

• IPR002196 is a phage-lysozyme family. InterPro's live record for
  IPR002196 ("Glycoside hydrolase, family 24", a family entry) maps it to
  GO:0003796, GO:0009253 and GO:0016998; 1,570 of its 8,193 UniProt
  members are viral (InterPro API, taxonomy 10239). The two process terms
  mis-direct all 1,570.

• A keyword-name correction for the earlier reviews. The XV and lysK reviews
  both describe KW-0081 as "Antimicrobial". The live UniProt keyword record
  gives KW-0081 = Bacteriolytic enzyme ("Enzyme, e.g. lysozyme or
  endopeptidase, essential for lysis of bacterial cell walls"), a child of
  KW-0929 Antimicrobial. Its GO mapping is GO:0003824, GO:0008152,
  GO:0031640 and GO:0042742. This sharpens the upstream report rather than
  weakening it: the offending keyword is not the broad antimicrobial one that
  also covers host defence peptides, but the narrow one that means
  "bacteriolytic enzyme", for which defense response to bacterium is never
  right, and whose mapping to GO:0003824 catalytic activity and
  GO:0008152 metabolic process is uninformative root-level noise.

CACAO row summary

Open questions

• GO has no "host periplasmic space" cellular-component term (QuickGO search),
  so the compartment where this enzyme actually works cannot be expressed; the
  only CC available is GO:0030430 host cell cytoplasm, which is where the
  enzyme sits before the holin acts.
• What the two LysM repeats contribute is unknown. No binding or deletion
  experiment on phi29 gp15 has been published; the UniProt DOMAIN line is
  inference from the domain's general properties.

Addendum (2026-11-04, on review of PR #5956)

Two corrections to the body above, both caught on review.

The qualifier count is twelve, not ten. Line 122 says "all ten of phi29's
2025-dated CACAO rows". Counted from the seven phi29 GOA files: gene 1 once
(20250210), gene 2 three times (20250208, 20250210, 20250217), gene 3 four
times (20250210 ×2, 20250224, 20250324), and genes 4, 6, 15 and 17 once each —
twelve, all acts_upstream_of_or_within_positive_effect. The only pre-2025
CACAO rows on those genes are gene 15's 20121116 enables and gene 6's four
2017 rows using enables, part_of and involved_in, so the claim the
sentence makes is right and in fact stronger: the flip is total across the
2025 cohort. 15-ai-review.yaml carries the corrected figure, which matters
because just export-source-defects ships that explanation verbatim to
CACAO.

PMID:820555's full text was obtained within this batch. Line 120 states
flatly that it could not be read. That is true of the attempts made for this
gene but not of the batch's record: genes/BPPH2/1/1-notes.md:124-133 records
an OCR copy via a third-party aggregator and quotes its Discussion. No
lysis-cistron sentence was recorded from that copy and it could not be
re-obtained — Wiley returns HTTP 403 on every URL including the one Unpaywall
lists as free, and the DIGITAL.CSIC copy sits behind a bot check that was not
circumvented. So the PROBABLE rating on the WRONG_GENE_PRODUCT defect
stands, but for that reason rather than the one line 120 gives; the YAML
explanation states it correctly.