SIL1 (Silaffin-1), Cylindrotheca fusiformis — review notes

UniProt: Q9SE35 (SIL1_CYLFU), Swiss-Prot, 265 aa precursor, PE:1 (evidence at
protein level). NCBITaxon:2853.

Summary of what this protein is

SIL1 encodes the silaffin-1 precursor. It is not an enzyme in the usual
sense: it is a post-translationally hypermodified, highly repetitive polycationic
polypeptide that is proteolytically processed into a set of short peptides
(silaffin-1A1, -1A2, -1B) which precipitate silica from silicic acid solution
within seconds. The protein is a structural and templating component of the
diatom silica cell wall (frustule), not a catalyst with a measurable turnover
number.

Architecture from the UniProt record (Q9SE35 feature table):

• 1–19 signal peptide (ECO:0000255)
• 20–107 acidic propeptide (PRO_0000032597); UniProt explicitly flags that
  it is unknown whether this acidic N-terminal chain is functional or is merely
  a propeptide (CC -!- DOMAIN)
• 108–136 silaffin-1B (PRO_0000032598)
• 141–158 silaffin-1A2 (PRO_0000032600)
• 163–177, 182–196, 201–215, 220–234, 239–253 — five copies of silaffin-1A1
  (PRO_0000032602, _604, _606, _608, _610)
• seven atypical/canonical repeats R1–R7 over 108–257, with a declared
  7 X 19 AA consensus S-S-K-K-S-G-S-Y-S-G-S-K-G-S-K-R-R-[IL]-L
• cleavage occurs on pairs of basic residues (KW Cleavage on pair of basic residues), consistent with the RR/KR motifs flanking each repeat

So one sil1 transcript yields many mineralization peptides from one
precursor — the repeats are the functional units, and the mature peptides share
the Q9SE35 accession. This is the POMC-style case the repo models with
functional_isoforms / CLEAVAGE_PRODUCT rather than separate gene folders.

The silica-precipitating activity

The founding observation PMID:10550045.
The same paper establishes that the activity depends on the lysine modifications,
not on the peptide backbone alone PMID:10550045.

The chemical structure work assigns the activity specifically to the sil1 gene
product PMID:11349130 and confirms the in vitro
activity of the purified peptides PMID:11349130. Note the verbatim typographic quirk "and-1A(2)" in the cached
abstract — quoted as-is.

Native-state extraction showed the peptides are additionally and essentially
phosphorylated PMID:12386330, giving a
zwitterionic peptide that self-assembles PMID:12386330. This matters for how the function should be described: the
active species is a self-assembled polyanionic/polycationic aggregate, and
the earlier HF-extracted ("natSil"/dephosphorylated) preparations were missing a
modification required for the native activity.

The mechanistic reading in the literature is templating/acceleration rather than
catalysis PMID:11106386. Note the hedge ("probably") — this is the authors' inference, not a
demonstrated mechanism.

Localization

Silaffins are cell wall (biosilica) components, and are so tightly held that
they require HF dissolution of the wall to release PMID:11106386. This is direct
evidence for a cell-wall / external encapsulating structure location.

In vivo, silica deposition happens in the silica deposition vesicle (SDV), an
intracellular acidic compartment; UniProt's PTM note refers to it ("may also
control the silica colloid size within the silica deposition vesicle (SDV)").
GO has no term for the SDV (OLS search in go for "silica deposition
vesicle" returns zero classes), so the SDV cannot be annotated — see
proposed_new_terms in the review. I did not annotate an SDV location, because
the cited work demonstrates the mature location (biosilica/cell wall), not a
direct SDV localization experiment for this protein.

GO annotation status — a genuine, total gap

Q9SE35 has zero GO annotations. just fetch-gene CYLFU SIL1 produced a
SIL1-goa.tsv with a header row and no data, and a direct QuickGO query
(.../annotation/search?geneProductId=Q9SE35) returns "numberOfHits":0.
So this review has no existing annotations to adjudicate; its whole content is
NEW rows plus proposed terms. This is unusual, and worth stating plainly:
it is not that curators looked at silaffin-1 and declined to annotate it, it is
that a well-characterized Swiss-Prot entry with PE:1 and four primary papers has
never been annotated at all. There is no PANTHER family for it either
(fetch-gene reported "No PANTHER family found in UniProt data"), which is
consistent — silaffins are lineage-restricted and compositionally biased, so
they sit outside the usual family-propagation machinery.

Ontology gaps that block a faithful annotation

Two terms that this protein needs do not exist, verified via OLS against go:

1. No MF for silica polycondensation / biosilica nucleation. The only
   silica-related MF in GO is GO:0015115 silicate transmembrane transporter
   activity, which is a transporter and plainly wrong here. Searching go for
   "silica" returns only GO:0034021 response to silicon dioxide and
   GO:0071251 cellular response to silicon dioxide — both response terms, also
   wrong. Per CLAUDE.md, rather than forcing an ill-fitting id I left
   molecular_function unset on the core function and used
   proposed_molecular_function against a proposed_new_terms entry.
2. No CC for the silica deposition vesicle, as above.

A third, softer issue: GO:0031214 biomineral tissue development is the obvious
"biomineralization" BP, but it is_a GO:0009888 tissue development and its
ancestors run through multicellular-organism development. A unicellular diatom
has no tissue, so I did not use it. GO:0042546 cell wall biogenesis is the
honest BP here. For the location I used GO:0030312 external encapsulating
structure rather than GO:0005618 cell wall, because GO:0005618 carries an
explicit "Not to be used for manual annotation" instruction directing curators
to a lineage-specific child, and the children it lists (bacterial, plant-type,
fungal-type, S-layer) have no diatom option. The missing diatom child is
recorded as a proposed term.

Widened BP search (review follow-up, PR #5636): OLS searches of go for
"biomineral" and "mineralization" return no lineage-neutral biomineralization
or mineral-deposition process. Every "biomineral" hit is GO:0031214
biomineral tissue development, its regulation terms (GO:0070167-GO:0070169,
under regulation of multicellular organismal development) or a child of it
(e.g. GO:0036340 chitin-based cuticle sclerotization by biomineralization);
the "mineralization" hits are bone, tooth/dentin and otolith mineralization
(all under GO:0031214/tissue development) plus unrelated mineralocorticoid
terms. So there is no non-tissue parent for a unicellular silicifier, and
GO:0042546 cell wall biogenesis remains the most accurate existing BP.

CMMMech cross-check (biosilica / silica precipitation)

CMMMech grounds Q9SE35 in data/records/anatase.yaml, in the
inp_silaffin_templated_anatase_graph causal graph, as node
sil1_mineralization_peptide. The source is Ping et al. 2016 (Chem. Sci. 7:6330–6336,
DOI 10.1039/C6SC02311H), an engineered E. coli surface-display construct: the
InaZ ice-nucleation anchor fused to the synthetic Sil-1 peptide
SSKKSGSYSGSKGSKRRIL, used to template amorphous titanium oxide from TALH,
which becomes anatase only after 600 °C calcination.

Two observations, recorded here rather than as source_defects (per the brief,
source_defects is for GOA/UniProt/InterPro, not for CMMMech's own fields):

1. The CMMMech node description overstates the biology. It describes the
   SIL1 node as "C-terminal silaffin peptide domain catalyzing titanium
   precursor condensation". Nothing in the silaffin literature supports
   catalysis, and the activity demonstrated for silaffin-1 is on silicic
   acid → silica, not on titanium. CMMMech's own evidence block is
   appropriately careful (supports: partial, non-engineered controls absent),
   and the record-level functional_role says "nucleates", which is better.
   The mismatch is confined to the node description field. UniProt's own
   CC -!- FUNCTION line unfortunately uses the same loose word ("Catalyzes the
   polymerization of silica spheres"), so CMMMech may simply have inherited it.
   I did not propagate "catalyzes" into this review.
2. The grounded peptide is a repeat-consensus synthetic peptide, not the
   protein. SSKKSGSYSGSKGSKRRIL is the R3–R7 consensus
   (S-S-K-K-S-G-S-Y-S-G-S-K-G-S-K-R-R-[IL]-L), chemically synthesized and
   therefore carrying none of the polyamine, methylation, hydroxylation or
   phosphorylation modifications that PMID:10550045 and PMID:12386330 show
   are required for the native silica-precipitating activity. The grounding to
   Q9SE35 is correct as a sequence provenance statement, which is all CMMMech's
   annotation_note claims. But any inference from the Ping et al. result back
   to the in vivo function of the diatom protein would be unsound in both
   directions: the construct is unmodified, heterologous, surface-displayed, and
   mineralizing a different element. Nothing from Ping et al. 2016 entered this
   review as evidence.

I did not edit CMMMech.

Review decisions

No existing annotations, so no ACCEPT/MODIFY/REMOVE calls. Three NEW rows,
kept deliberately narrow:

• GO:0030312 external encapsulating structure (CC, IDA, PMID:11106386) — the
  HF-requirement result is direct evidence of integration into the wall. After
  review on PR #5636 the row quotes the paper's own Results (C. fusiformis walls
  isolated and dissolved in anhydrous HF, silaffin pattern recovered), not its
  introduction's restatement of PMID:10550045.
• GO:0042546 cell wall biogenesis (BP, IDA, PMID:10550045) — silaffin-1
  peptides themselves perform the silica precipitation step that builds the
  wall, so the participation test in CLAUDE.md is satisfied: this gene product
  does the work, it is not merely required for or consumed by the process.
• GO:0005199 structural constituent of cell wall (MF, IC, PMID:12386330) —
  a curator inference, not a direct demonstration. Originally entered as IDA on
  the HF-insolubility result; changed after review on PR #5636 because HF
  insolubility is a location fact (it is the evidence for the GO:0030312 row)
  and no experiment has measured a contribution to wall integrity. Now anchored
  on PMID:12386330 PMID:12386330, with the authors'
  "may represent" hedge kept, and HF occlusion as corroboration only. This is the
  structural half of the function; the precipitating half has no GO term
  and is left to proposed_new_terms.

I considered and rejected a NEW row for GO:0031214 biomineral tissue
development (wrong branch: tissue/multicellular, see above) and for any
polyamine-biosynthesis term (the polyamine modification is installed on SIL1 by
other, unidentified enzymes — SIL1 is the substrate of that modification, which
is exactly the non-participation case CLAUDE.md warns about).