HHAT (human, Q5VTY9) — curation notes

Working journal for the GO annotation review. PMIDs marked [not cached] are
not in this repository's publications/ cache; for those I record only what a
machine-readable source (UniProt entry, QuickGO annotation row, PANTHER PAINT
slice, cached Reactome entry) states, and I do not quote them.

1. What the protein is

HHAT (Hedgehog acyltransferase, UniProt Q5VTY9, HGNC:18023, 493 aa, chromosome
1) is a polytopic endoplasmic-reticulum membrane enzyme of the membrane-bound
O-acyltransferase (MBOAT) superfamily. UniProt's recommended name is
Protein-cysteine N-palmitoyltransferase HHAT, and UniProt assigns it to the
"membrane-bound acyltransferase family, HHAT subfamily"
[file:human/HHAT/HHAT-uniprot.txt "Belongs to the membrane-bound acyltransferase family. HHAT"].
PANTHER places it in family PTHR13285 (official name ACYLTRANSFERASE),
subfamily PTHR13285:SF20 PROTEIN-CYSTEINE N-PALMITOYLTRANSFERASE HHAT
(interpro/panther/panther.obo,
interpro/panther/PTHR13285/PTHR13285-entries.csv). The same family holds
Drosophila Rasp (Q9VZU2, subfamily SF18), the HHAT paralog HHATL (Q9HCP6,
SF19), the yeast MBOATs GUP1/GUP2, and the bacterial DltB/AlgI acyltransferases.

Topology: ten transmembrane domains plus two reentrant loops, with the catalytic
His and Asp on opposite faces of the ER membrane
PMID:25505265, and the protein is itself palmitoylated on
several cytosolic cysteines. Mutating the conserved catalytic His abolishes
activity PMID:25505265. A bound heme b is required for function
PMID:34890564.

Seven isoforms are annotated (Q5VTY9-1 … Q5VTY9-7); no GOA row is
isoform-specific, so no isoform field is set.

2. The reaction, and why the GO molecular-function term is chemically off

HHAT transfers palmitate from palmitoyl-CoA onto the alpha-amino group of the
N-terminal cysteine of Hedgehog proteins, forming an amide (N-linked) bond -
not a thioester on the cysteine sulfur. This was established biochemically
PMID:18534984 and then visualised directly in the product complex
PMID:34112694. UniProt encodes it as RHEA:59528,
N-terminal L-cysteinyl-[protein] + hexadecanoyl-CoA = N-terminal N-hexadecanoyl-L-cysteinyl-[protein] + CoA + H(+), and a second reaction
(RHEA:59580) for the cholesterylated substrate.

The problem. GOA annotates this with GO:0019706
protein-cysteine S-palmitoyltransferase activity, whose definition is explicit
about the acceptor atom: "Catalysis of the transfer of a palmitoyl (systematic
name, hexadecanoyl) group to a sulfur atom on the cysteine of a protein
molecule" (OLS). That is the wrong linkage for HHAT.

Why I did not modify the term. There is no existing GO molecular-function
term for N-terminal-cysteine N-palmitoylation. Searching GO for
N-palmitoyltransferase returns only the peptidyl-lysine N6 terms (GO:0018031,
GO:0140771, GO:0140772), and GO:0140186 protein N-acyltransferase activity is
defined on lysine N6 as well. The nearest correct ancestors are GO:0016409
palmitoyltransferase activity and GO:0019707 protein-cysteine S-acyltransferase
activity - and GO:0019707 has the same sulfur problem. Five independent GOA rows
(four IDA from four different papers, plus one Ensembl-Compara IEA) use
GO:0019706, and those curators had no better option. So GO:0019706 is accepted
as the closest available term, and the gap is recorded in proposed_new_terms
as protein-cysteine N-palmitoyltransferase activity - which is, verbatim, both
UniProt's recommended protein name and PANTHER's subfamily name for HHAT.

Notably GO does make the distinction on the biological-process side:
GO:0018009 N-terminal peptidyl-L-cysteine N-palmitoylation ("The covalent
attachment of a palmitoyl group to a nitrogen (N) atom in an N-terminal cysteine
residue to form N-palmitoyl-L-cysteine") sits beside
GO:0018230 peptidyl-L-cysteine S-palmitoylation. See section 5.

3. Where it acts

The ER membrane is the site of catalysis. Every structural and biochemical study
agrees:
PMID:34112694;
PMID:31875564;
PMID:25505265.
The active site is membrane-embedded, reached by palmitoyl-CoA from the cytosolic
leaflet and by the Hedgehog peptide from the lumen
PMID:34890564.

Golgi localisation is a weaker, secondary claim. The abstract of the primary
paper says only that the modification happens en route through the secretory
system PMID:18534984, and UniProt records the Golgi assignment as a curator
inference rather than a direct observation
(Golgi apparatus membrane {ECO:0000305|PubMed:18534984}, and
Note=Co-localizes with SHH in the ER and Golgi membrane. {ECO:0000305|PubMed:18534984}). The cache holds only the abstract of
PMID:18534984, so I have not tried to adjudicate the imaging; the two Golgi rows
(GO:0005794 IDA, GO:0000139 IEA) are kept and marked non-core, because the ER
membrane is where all the characterised chemistry happens.

4. The palmitoyl-CoA conduit — a second activity, deliberately not annotated

Asciolla and Resh reported that HHAT also moves its own acyl donor across the ER
membrane, and that this is separable from catalysis:
PMID:31875564
and PMID:31875564 UniProt states it as a FUNCTION with ECO:0000269 from three papers:
Promotes the transfer of palmitoyl-CoA from the cytoplasmic to the luminal side of the endoplasmic reticulum membrane, where hedgehog palmitoylation occurs (PubMed:31875564, PubMed:34112694, PubMed:34890564).

I considered proposing a transport annotation and decided against it, on two
grounds.

1. Comparator check. No MBOAT in this family carries a transport term
   anywhere. QuickGO for mouse Hhat (Q8BMT9) and Drosophila rasp (Q9VZU2)
   returns only acyltransferase, localisation, palmitoylation and Hedgehog
   pathway terms - no transporter activity, no lipid transport process. A
   systematic absence across the family is a convention, not an oversight.
2. The 2021 structures reinterpret the phenomenon as part of catalysis. They
   show palmitoyl-CoA entering an "archway" and a membrane-embedded reaction
   chamber within the enzyme, i.e. substrate access to the active site rather
   than net transport of a cargo released on the far side
   PMID:34112694 is the ligand-side
   claim; the mechanism claim is
   PMID:34890564

Also, the only GO MF term close to the claim, GO:0015607 ABC-type fatty-acyl-CoA transporter activity, is wrong by construction (HHAT is not an
ABC transporter). I raise the question in suggested_questions instead of
asserting an annotation.

5. The one gap worth filling: GO:0018009

GO:0018009 N-terminal peptidyl-L-cysteine N-palmitoylation is the biological
process HHAT performs, named at exactly the chemistry HHAT does. Human HHAT does
not carry it. Both of its closest relatives do:

The mouse row is the decisive detail: GO_Central transferred GO:0018009 to
mouse Hhat from human HHAT by ISO. A curator therefore already judged that
human HHAT has this function, and the human row is missing from the GOA
snapshot rather than deliberately withheld. (Source: QuickGO
downloadSearch?goId=GO:0018009.)

Applying the two NEW tests from CLAUDE.md:

• Participation. HHAT catalyses the step. The term's ancestors are
  GO:0018345 protein palmitoylation, GO:0006498 N-terminal protein lipidation,
  GO:0006497 protein lipidation and GO:0043543 protein acylation - all
  attachment processes, which name the enzyme that does the attaching, not the
  substrate. This is the mirror image of the Drosophila hh case, where the
  ligand is the substrate and the acyltransferase activity is not the ligand's
  function: here the acyltransferase is the actor.
• Comparator. rasp carries the term with three experimental rows, and mouse
  Hhat carries it by transfer from human. This is a gap, not a convention.
• Non-redundancy. GO:0018009 is neither an ancestor nor a descendant of
  GO:0051604 protein maturation (protein maturation is not among its ancestors),
  nor of GO:0045880, nor of the molecular-function terms on the record.

So GO:0018009 is added as the single NEW row, with IDA on PMID:34112694 (the
product complex resolving the amide bond) and PMID:18534984 / PMID:31875564 as
additional references.

6. Pathway role, and the substrate-versus-actor line

HHAT's product is what licenses Hedgehog reception. The palmitate is not
decoration: it is structurally read by Patched
PMID:34112694, and palmitoylation
is required for signalling at both ranges
PMID:18534984 and
PMID:31875564. Reactome makes the loss-of-function
side of the same point [Reactome:R-HSA-5358343 "Mutation or depletion of the HHAT
enzyme and mutation of the palmitoyl acceptor cysteine in Hh itself abrogates
palmitoylation of the ligand and reduces Hh signaling"].

GOA expresses this as GO:0045880 positive regulation of smoothened signaling
pathway (three human IDA rows plus an Ensembl-Compara IEA; mouse Hhat has two
IDA rows for the same term). I accepted it, and the reasoning matters because it
is where the substrate/actor distinction bites:

• HHAT is not a component of the smoothened signaling pathway - it never
  touches the receptor, and it acts before the ligand is even released. A bare
  GO:0007224 smoothened signaling pathway row on HHAT would be an
  over-annotation. (Mouse Hhat does carry GO:0007224 by IMP, and Drosophila rasp
  carries GO:0007225 patched signaling pathway by IMP; the human record, which
  uses the regulation term instead, has this better.)
• But HHAT does perform the act that raises pathway output: it covalently
  modifies the ligand into its high-activity form. "Positive regulation of"
  is precisely GO's way of saying that something acts on a pathway from outside
  it, and the work of that regulation is the palmitoyl transfer HHAT catalyses.

The narrower, mechanistically direct processes are GO:0018009 (the chemistry) and
GO:0051604 protein maturation (the ligand-maturation step the GO-CAM uses). All
three are treated as core; nothing further downstream is.

Where the GO-CAM sits. gocams/index.tsv gives HHAT exactly one activity:
model 696022cd00000016, "Protein maturation of SHH by autocleavage,
autocholesterylation and palmitoylation by HHAT (Human)", activity
696022cd00000016/696022cd00000024 = GO:0019706 part_of GO:0051604 occurs_in
GO:0005789. My core function reproduces that triple and adds GO:0018009 and
GO:0045880 as the processes GOA and the ortholog records also carry.

Module consistency. modules/hedgehog_signaling.yaml models HHAT as annoton
hh_n_palmitoylation of the ligand-biogenesis part, with the family selector
PANTHER:PTHR13285 (label ACYLTRANSFERASE, which matches
interpro/panther/panther.obo verbatim), representative members HHAT (human) and
rasp (Drosophila), function GO:0019706, substrates cholesterylated Hedgehog
signaling domain and palmitoyl-CoA, location GO:0005789. My review agrees on all
of these. The module's role_description - "the palmitate is the dominant
contact in the Patched interface, so this step licenses reception rather than
merely decorating the ligand" - is the same judgement I use to accept GO:0045880.
The one thing my review adds that the module does not carry is GO:0018009; that
is a candidate addition to the annoton's processes, not a disagreement.

7. Phylogenetic propagation (IBA)

interpro/panther/PTHR13285/PTHR13285-paint.tsv resolves both IBA rows:

Both placements read correctly. ER residence is a property of the whole
eukaryotic MBOAT group including the yeast GUPs, so a Eukaryota-scoped node is
right; palmitoyltransferase activity is restricted to the Hedgehog-acylating
metazoan clade, so a Eumetazoa-scoped node is right. Human HHAT appearing in its
own WITH/FROM for both is expected, not circular: its experimental rows are
among the descendant evidences the PAINT curator used, and the IBA then adds the
claim that the property is inherited.

Note that the same slice contains PTN002665109 GO:0016409 IRD (negated) and
PTN002665109 GO:0016746 IRD at taxon:117571 - the curator explicitly recorded
loss of acyltransferase activity in the HHATL branch. That is a useful
control: it shows the curator distinguished HHAT from its catalytically
compromised paralog rather than blanketing the family, which strengthens the
GO:0016409 node placement on HHAT.

GO:0016409 is the direct parent of GO:0019706 and, unlike its child, says
nothing about the linkage atom - so for HHAT it is the more chemically accurate
of the two, which is one more reason to accept rather than deepen it.

8. The two Reactome TAS rows for GO:0016747

GO:0016747 acyltransferase activity, transferring groups other than amino-acyl groups is three levels above GO:0019706 (GO:0019706 -> GO:0016409 -> GO:0016747)
and says nothing specific about HHAT. Both rows come from cached Reactome
reactions whose own text is far more specific than the term they export:
[Reactome:R-HSA-5358343 "the N-terminal fragment of Hh (Hh-Np) is also
palmitoylated by the O-acyltransferase HHAT"] and the disease-mutant reaction
[Reactome:R-HSA-5483229 "the mutant protein is unable to palmitoylate SHH or DHH
in an in vitro assay"]. Both are modified to GO:0016409 palmitoyltransferase
activity - the most specific existing term that is chemically correct for HHAT.

9. The one GO:0005515 row

A single IPI row from HuRI with partner SDCBP (O00560, syntenin-1). HuRI is a
systematic yeast two-hybrid screen
PMID:32296183 whose authors state that
PMID:32296183. Syntenin-1 is a PDZ-domain adaptor of the syndecan/exosome axis and
appears in no described HHAT mechanism; there is no follow-up and no more
informative molecular function to modify to. Removed as uninformative, which is
not a claim that the interaction is false.

10. Disease

Biallelic HHAT variants cause Nivelon-Nivelon-Mabille syndrome
(NNMS, MIM:600092): progressive microcephaly, cerebellar vermis hypoplasia and
skeletal dysplasia, with variable early seizures, growth retardation,
chondrodysplasia and micromelia, and 46,XY gonadal dysgenesis
(file:human/HHAT/HHAT-uniprot.txt, DISEASE block; PubMed:24784881,
PubMed:30912300, PubMed:34890564, PubMed:35045414, PubMed:36303863,
PubMed:40326711 - only PMID:34890564 is cached). The G287V allele is the one
Reactome models
[Reactome:R-HSA-5483229 "A G287V loss-of-function mutation in HHAT was
identified in a rare case of Syndromic 46, XY Disorder of Sex Development, which
results in testis dysgenesis."]. HHAT is also a cancer drug target; the inhibitor
IMP-1575 occludes the active site
PMID:34890564. Disease terms are not GO annotations and no action follows.

11. Substrate specificity

Worth recording because it bounds the annotation: HHAT acylates Hedgehog proteins
and apparently nothing else.
PMID:18534984 and
PMID:34112694. The acceptor consensus is the
Hedgehog N-terminal CGPGR motif. This is why the MF and BP terms can be tied
to the Hedgehog pathway without the enzyme being a pathway component.