UniProt Q02936 · FlyBase FBgn0004644 · CG4637 · 471 aa precursor.
PANTHER family PTHR11889, subfamily PTHR11889:SF31 (protostome HEDGEHOG).
These are my working notes for the GO annotation review in hh-ai-review.yaml.
Assertions carry the citation they rest on; quoted text is verbatim from the cached
publication in publications/.
One gene, one ligand. Where vertebrates have SHH, IHH and DHH, the fly has a single
Hedgehog, which makes it the anchor for every pan-metazoan claim about the family — the
PANTHER family review (interpro/panther/PTHR11889/PTHR11889-review.yaml) says as much,
and the PAINT node PTN001726121 that carries patched binding, smoothened signaling
pathway and extracellular region is seeded by FB:FBgn0004644 alongside the vertebrate
genes.
The protein is made as a precursor with two functionally unrelated halves. UniProt's
feature table gives the anatomy: SIGNAL 1..?, PROPEP ?..84, CHAIN 85..471 (Protein
hedgehog), CHAIN 85..257 (Protein hedgehog N-product), SITE 257..258 "Cleavage; by
autolysis", LIPID 85 "N-palmitoyl cysteine", LIPID 257 "Cholesterol glycine ester".
The 25 kDa C-terminal domain performs an intramolecular cleavage coupled to cholesterol
transfer:
That paper is a crystal structure of the Drosophila domain, with active-site residues
identified and probed by mutagenesis PMID:9335337. UniProt records the outcome on the fly protein: SITE 303 "Involved in
cholesterol transfer" with MUTAGEN D303A "No cholesterol transfer"; SITE 326 "Involved in
auto-cleavage" with T326A "Greatly reduced autoprocessing activity"; SITE 329 "Essential
for auto-cleavage" with H329A "No autoprocessing activity". So the fly protein has its
own direct structural and mutational evidence for both halves of the reaction — the
ISS rows for GO:0140853 and GO:0016540 (transferred from mouse Q62226) are not the only
support.
The N-terminal product is the active species PMID:7885476, and the
cleavage site itself was mapped in the same paper.
Intein homology is not splicing. The same structure paper reports that "Aspects of
sequence, structure, and reaction mechanism are conserved between Hh-C17 and the
self-splicing regions of inteins" PMID:9335337. An InterPro2GO mapping turns this into
GO:0016539 intein-mediated protein splicing. Checked the term definition against
QuickGO: it requires excision of the intein and joining of the N- and C-terminal
exteins by a normal peptide bond. Hedgehog ligates nothing — the two products separate,
and one leaves with cholesterol on it. REMOVE. (The human SHH review made the same call.)
Rasp / Skinny hedgehog is the acyltransferase PMID:11486055. Useful for the
participation test: palmitoylation of Hh is performed by Rasp, so it is not an hh
activity, and there is no GOA row asking me to pretend otherwise.
PTHR11889-review.yaml marks the PTN001726121 calcium ion binding row UNRESOLVED
because its three seeds are human paralogs while the node sits at Eumetazoa, and asks
whether the coordinating residues are conserved outside vertebrates. Two observations
from the cached UniProt record settle it in the affirmative, at least for the fly:
ECO:0000250|UniProtKB:Q15465 —This is consistent with the IBA rather than contradicting it, so the row is kept (as
non-core: the metal is a structural feature of the signalling domain, not an activity).
I have not added residue_sites anywhere, and I am not claiming the sites are
experimentally demonstrated in the fly — only that the sequence question the family
review asked has an answer consistent with the annotation.
Vertebrate SHH binds PTCH1 directly. In the fly, Patched alone was never shown to bind
Hh; the receptor is a two-component affair:
Ihog binds Hh through the first of its two fibronectin type III repeats PMID:16630821, heparin-dependently
PMID:17077139. The structure of
the complex is in that paper.
This matters for the five GO:0005515 protein binding rows. Four of them record Ihog
(FB:FBgn0031872 = Q9VM64, confirmed via the UniProt cross-reference query, since the
FlyBase API was down). Per repo policy, a bare protein-binding row gets MODIFY when the
paper supports something more informative, otherwise REMOVE. GO:0005113 patched binding
would be wrong — the recorded partner is Ihog, not Ptc — so I proposed GO:0005102
signaling receptor binding (label verified against QuickGO), which says the partner is
a component of the receptor that transduces the signal, and nothing more. The fifth row
records Shifted (Q9W3W5), a secreted WIF-family modulator of Hh diffusion PMID:15691765 —
no receptor, no catalysis, no evidence-backed specific term, so REMOVE.
The consequence of Hh–Ihog binding is also annotated as GO:0034111 negative regulation
of homotypic cell-cell adhesion PMID:31209108. Unusual in
that Hh really does perform this directly — by competition — but it is the same binding
event GO:0005113 already records, and the adhesive activity belongs to Ihog, so non-core.
GOA carries GO:0005634 (EXP) and GO:0005737 (EXP) from PMID:1280560, plus IEA rows that
inherit the same statement through the UniProt SUBCELLULAR LOCATION lines ("Nuclear up to
embryonic stage 10 and then at stage 11 shifts to the cytoplasm"). The paper's own
abstract says what was observed:
That is the hh transcript, not the Hh protein; the paper is a molecular-organisation and
in-situ expression study and reports no anti-Hh antibody data. A protein with a cleaved
signal sequence that autoprocesses in the ER lumen and is released from the cell has no
route to the nucleoplasm. This is not a case of second-guessing an assay I cannot see —
the abstract states the observation — so REMOVE all four rows, and flag PMID:1280560's
reference_review as MISCITED for this use.
Three GO:0005829 rows trace to cached Drosophila Reactome events. One is literally titled
"N-HH moves from the cytosol to the extracellular region" (R-DME-209271); the others are
"Autocleavage of HH" (R-DME-209313) and "N-HH is palmitoylated by RASP" (R-DME-209357).
All three steps happen in the secretory pathway. GO:0005829 is defined (QuickGO) as the
part of the cytoplasm that does not contain organelles, so this is wrong rather than
merely vague. REMOVE. The three extracellular region rows from the same Reactome model
(R-DME-209178/209271/209272) are fine and kept as core.
Note this differs from the human SHH review, which kept its cytosol TAS rows as non-core
on the reasoning that they concern retrotranslocated fragments destined for degradation.
The fly Reactome events are different events with a different justification, hence the
different call.
The term GO:0016015 asks for action over a concentration gradient. Fly Hh is the textbook
case, and the cited rows give threshold-specific responses at three levels:
Transport is cytoneme-borne as well as diffusive PMID:28825565, and
range is gated by the lipid anchors themselves PMID:29522397.
The DHH caveat recorded in the family review (juxtacrine, poorly autoprocessed) does not
apply to the fly gene; if anything the fly is the cleanest morphogen in the family.
GOA carries both sides, which is the right thing for it to do:
How I resolved it. The repo's consistency rule wants one action per GO term, and the four
GO:0008354 rows (three positive plus the negation) therefore share KEEP_AS_NON_CORE: hh
mutants do show migration defects, but the likely route is the gonadal mesoderm, which is
annotated separately from the same screen PMID:9435287. The NOT row is retained and its reason says plainly that it
asserts absence and that the negation is well supported.
GO:0035232 germ cell attraction is the one I removed. It is the strongest form of the
claim — Hh as an attractant — it is exactly what PMID:19389345 tested and rejected, and its
only support is a traceable author statement in a review published before the refutation.
Sixty-eight rows are KEEP_AS_NON_CORE, and almost all are of one kind: a fly developmental
phenotype. Segment polarity (the phenotype the gene is named for), morphogenetic furrow
progression PMID:8252628, wing/leg/eye/labial/genital disc
patterning, tracheal branching, hindgut, heart (via wg — PMID:8660881), glial migration, ovarian stem cell maintenance
PMID:11279500, and the
adult gut immune response PMID:25639794.
All are sound and worth keeping for retrieval. None is a function of the protein: one
ligand, one biochemical output, many tissues. Core status is reserved for autoproteolysis
with cholesterol transfer, Patched binding, and the smoothened signalling pathway.
A second family of non-core rows is the downstream-consequence terms: inhibition of Ci
proteolysis PMID:9215627, Smo surface accumulation PMID:19088085, regulation of bnl/FGF
PMID:24651658, and the IBA regulation of gene expression. In each the work is done by
something else — the proteasome, Smo trafficking machinery, Ci/Gli — and Hh sets the rate.
Both cited records have no abstract in the local cache, and an efetch query to PubMed
returns title and identifiers only for both. The standing rule is UNDECIDED when the
relevant publication cannot be accessed. Both claims are plausible on general grounds, but
plausibility is not evidence, and no other reference in this set covers either claim. Their
reference_review entries are marked UNVERIFIED with the reason.
(PMID:10625531 is also cited for GO:0048100 wing disc anterior/posterior pattern
formation, which I did not mark undecided: that term carries three further citations,
including PMID:15104233, so the term is not in doubt even
though one of its four anchors is unreadable.)
modules/hedgehog_signaling.yaml types its hh_autoprocessing annoton withNo NEW rows. Two candidates were considered and rejected on the participation test:
GO:0016015 morphogen activity, GO:0005113 patched binding, GO:0140853 and
GO:0007224 were all already present, so the three core functions are drawn from accepted
rows and nothing needed to be manufactured.