Provenance note. No deep-research provider was available in this container (no API
keys), so just deep-research was not run and no -deep-research-<provider>.md file
exists for this gene. The notes below were assembled by reading the cached publications
in publications/, the UniProt flat file (D15-uniprot.txt), and PubMed/PMC via the
PubMed MCP tools. Of the eleven cited papers only PMID:27273516 has
full_text_available: true; the rest are abstract-only in the cache. PMC holds scanned
images only for the three oldest papers (PMC427268, PMC515921, PMC15089 return empty
full_text), so for those the abstract is the whole readable record.
The GONUTS wiki (gowiki.tamu.edu), which would record the CACAO students' evidence,
was queried on 2026-10-29 for PMID:6248541 and returned only
"This site is temporarily offline while we perform upgrades", matching the status
recorded in projects/CACAO.md.
A 291-residue early phage protein, N-terminally processed by the host
(FT INIT_MET 1 /note="Removed; by host", ECO:0000269|PubMed:2211703). It is the
founder-type member of the FEN / 5'-nuclease family (Pfam PF01367 5_3_exonuc +
PF02739 5_3_exonuc_N, HAMAP MF_04140 FEN_T5, InterPro IPR043666 FEN_D15-like),
homologous to the 5'-3' exonuclease domain of E. coli DNA polymerase I:
Structurally it is a two-lobed PIN-like nuclease with a helical arch (residues
82–116) forming a hole over the active site and an H3TH DNA-binding motif (188–224).
Crystal structures of substrate and product complexes show single-stranded DNA
threaded through that hole:
The same paper confirms the gene identity of the protein it crystallised:
Two activities in one active site, distinguishable by mutation, pH and metal occupancy.
The two are mechanistically separable:
Metals are obligatory cofactors and their stoichiometry selects the reaction:
UniProt records the same, with the caveat that the endo/exo switching by metal
concentration is an in-vitro observation:
[file:BPT5/D15/D15-uniprot.txt "In vitro low metal concentrations selectively stimulate the endonuclease reaction."]
PMID:9380501
[file:BPT5/D15/D15-uniprot.txt "K->A: No exonuclease activity, retains full endonuclease activity on a flap structure. Binds DNA pseudo-Y substrates with a dissociation constant of 200 nM."]
This is why the bare GO:0003677 DNA binding IEA understates what is known: the
informative claims are Y-form (pseudo-Y) and 5'-flap-structured DNA binding.
D15 amber mutants alsoPMID:4357874
PMID:4357874
The late-gene defect is therefore a downstream consequence of a DNA-processing
defect, on the authors' own reading: the nuclease makes the template competent;
transcription itself is carried out by host RNA polymerase with the phage gpC2
controlling element. D15 performs no step of transcription.
PMID:6248541
The authors label it "putative" and it was isolated as a ~5 MDa species from infected
cells; there is no reconstitution or stoichiometry, and no later confirmation that I
could find. It is the only cellular-component evidence for this protein.
Moyer & Rothe reported that purified D15 nicks duplex T5 DNA provided a single-strand
interruption is already present:
whereas Sayers & Eckstein, with homogeneous overexpressed enzyme, found none on either
closed-circular or nicked duplex:
Both are consistent with a strictly structure-specific enzyme that needs a free,
single-stranded 5' end (a flap, a 5' overhang or a gapped/pseudo-Y junction) rather than
a nick as such — which is how UniProt states it:
[file:BPT5/D15/D15-uniprot.txt "The substrates require a free, single-stranded 5' end, with endonucleolytic hydrolysis occurring at the junction of double- and single-stranded DNA"]
So a bare "double-stranded DNA endonuclease" claim (the CACAO GO:1990238 row) is the
weakest form of the real finding and is partly contradicted by the 1990 study;
GO:0017108 5'-flap endonuclease activity is what the body of work supports.
T5 exonuclease is a widely sold laboratory reagent (seamless-cloning/assembly mixes,
removal of linear or nicked DNA), and the FEN family's best-known member in
biotechnology is the Taq polymerase 5'-nuclease domain:
Nothing in the GOA set is sourced from a reagent datasheet, but two reagent-adjacent
framings do reach the annotations and were treated carefully in the review:
GO:0033567 DNA replication, Okazaki fragment processing, which arrives by HAMAP rule
and is hedged even in UniProt
([file:BPT5/D15/D15-uniprot.txt "This function may be used for example to trim such branched molecules generated by Okazaki fragments synthesis during replication."]),
and the habit of describing the enzyme generically as a "dsDNA exonuclease/endonuclease"
as the datasheets do, which loses the structure specificity that the mutagenesis and
structural work establish.
All 38 GOA rows reviewed: 16 ACCEPT (the six GO:0017108 rows, the eight GO:0051908
rows, GO:0035312 and GO:0039693), 13 MODIFY, 6 TOO_GENERAL (the catalytic activity
/ nuclease / endonuclease / exonuclease / hydrolase /
catalytic activity, acting on DNA ancestors), 2 MARK_AS_OVER_ANNOTATED
(GO:0019086, GO:0033567), 1 KEEP_AS_NON_CORE (GO:0019034). No REMOVE and no NEW
rows: nothing in the set is false, and nothing the literature supports is missing from it
once the generic terms are refined.
| Term | Evidence | PMID | Action | One-line reading |
|---|---|---|---|---|
| GO:0019034 viral replication complex | IMP | 6248541 | KEEP_AS_NON_CORE | One crude co-purification the authors themselves call "putative"; claim plausible, peripheral, and the evidence is co-isolation (IDA/IPI-shaped), not a mutant phenotype |
| GO:0019086 late viral transcription | IMP | 4357874 | MARK_AS_OVER_ANNOTATED | Phenotype read as process: amber mutant makes no late proteins, but the paper's own model is that D15 nicks the template; host RNAP transcribes |
| GO:0035312 5'-3' DNA exonuclease activity | IMP | 9874768 | ACCEPT | Exactly what the mutagenesis shows (K83A kills exonucleolysis); and it adds the ssDNA substrate range that the dsDNA-specific GO:0051908 rows do not cover |
| GO:1990238 double-stranded DNA endonuclease activity | IDA | 904023 | MODIFY → GO:0017108 | Right paper, right protein, but the term drops the structure specificity and the bare claim is contradicted by PMID:2211703 |
Two of the four are sound as claims (one accepted outright); the other two are the
project's known shapes — phenotype read as process and a non-specific term for a
structure-specific activity (a granularity failure, not a false claim). Note also an
evidence-code oddity in the GO:0019034 row: IMP is used for what the cited paper
reports as a biochemical co-isolation; recorded as a source_defects entry rather than
argued in prose.