STK36 (human, Q9NRP7) — curation notes

Working journal for the GO annotation review.

1. What the protein is

STK36 (Fused homolog, EC 2.7.11.1) is a 1315-residue serine/threonine protein
kinase with an N-terminal kinase domain (UniProt DOMAIN 4..254, catalytic
proton acceptor Asp-125, ATP-binding Lys-33) followed by a long, largely
uncharacterised C-terminal region. It is the vertebrate orthologue of Drosophila
Fused (Fu).

PANTHER classification (authoritative source interpro/panther/panther-members.tsv):

Q9NRP7  PTHR22983:SF6

PTHR22983:SF6 is SERINE_THREONINE-PROTEIN KINASE 36, and Drosophila fu
(P23647) sits in the same subfamily. Sharing a PANTHER subfamily is direct
classification support for the STK36-fu orthology, independent of pathway
reasoning — which matters here, because the vertebrate and fly proteins differ in
how essential they are (section 4). ULK3, the other vertebrate Fused-family
kinase active in Hedgehog signalling, is in a different PANTHER family, which is
why modules/hedgehog_signaling.yaml records two family ids for its
fused_family_kinase annoton (PTHR22983 plus PTHR24348) rather than one. ULK4,
STK36's closest paralogue, is a pseudokinase.

The GOA row GO:0106310 protein serine kinase activity is an ISS transferred
directly from Drosophila fu (UniProtKB:P23647), so the classification and the
annotation record agree on the orthology.

Human disease: homozygous loss of function causes primary ciliary dyskinesia 46
(CILD46), with a central-pair axonemal defect and no laterality abnormality
PMID:28543983.

2. Established catalytic activity and its regulation

STK36 is a genuine, active kinase, and the activity has been measured rather than
inferred. Zhou et al. 2023 assayed immunopurified STK36 on a GST-Gli2N substrate,
identified the activation-loop autophosphorylation sites, and used the
kinase-dead and phospho-mutant forms as controls:

• PMID:38096226
• PMID:38096226
• PMID:38096226
• PMID:38096226
• PMID:38096226

UniProt records both half-reactions with experimental evidence codes — the Ser
reaction (Rhea:RHEA:17989, the same reaction id behind the GO:0106310 Rhea-based
IEA row) from PMID:31279575 and PMID:38096226, and the Thr reaction
(Rhea:RHEA:46608) from PMID:38096226 — and records the autophosphorylation:
[file:human/STK36/STK36-uniprot.txt "Autophosphorylated at Thr-158 and Ser-159, activating the protein"].

Two substrates are established: GLI2 (on Ser230/Ser232, conserved with the
Fu sites on Ci) and ULK4 (on Thr1021/Thr1023).

• PMID:31279575
• PMID:31279575
• PMID:38096226
• [file:human/STK36/STK36-uniprot.txt "Interacts with ULK4; promoting association with GLI2 substrate"]

Because GO:0004674 is directly supported by assay, the InterPro-derived
GO:0004672 protein kinase activity row is one level too general and should be
replaced rather than merely accepted.

There is also a UniProt-recorded non-catalytic requirement, from the original
Murone paper: [file:human/STK36/STK36-uniprot.txt "ULK4 adapter (PubMed:38096226). GLI2 requires an additional function of"]
— continuing "STK36 to become transcriptionally active, but the enzyme does not
need to possess an active kinase catalytic site for this to occur". This is worth
flagging as an open question rather than annotating, since no molecular activity
has been attached to it.

3. Where STK36 acts: ciliary tip, with ULK4

Zhou et al. established the compartment and made it mutually dependent with ULK4:

• PMID:38096226
• PMID:38096226
• PMID:38096226
• PMID:38096226
• [file:human/STK36/STK36-uniprot.txt "dependent for localization to ciliary tip (PubMed:38096226). Low levels"]

This is exactly the arrangement the GO-CAM model records: gocams/index.tsv
lists STK36 in model 696022cd00001645 with GO:0004674 / GO:0045880 /
GO:0097542, and ULK4 in the same model with GO:0030674 /
GO:0045880 / GO:0097542 — the kinase and its bridging adaptor at the same
location. The modules/hedgehog_signaling.yaml fused_family_kinase annoton
carries the same GO:0004674 + GO:0045880 pairing, so nothing here contradicts
the module.

STK36 is additionally cytoplasmic, present at low levels in the nucleus
(both from PMID:10806483, recorded in UniProt), and — in motile cilia — an
axonemal protein: [file:human/STK36/STK36-uniprot.txt "Cytoplasm, cytoskeleton, cilium axoneme"]
and PMID:28543983

4. The evolutionary complication, and how to read the Hedgehog annotations

This is the part of STK36 biology a reviewer has to get right, because the
literature looks contradictory until the redundancy is taken into account.

Mouse knockouts are Hedgehog-normal. Two independent mouse studies found no
embryonic Hedgehog phenotype:

• PMID:16055717
• PMID:16055717
• PMID:16055717
• PMID:19305393
• PMID:19305393

But the biochemistry is conserved, and the reason the knockout looks normal is
paralogue redundancy with ULK3. Han et al. showed Fu-family phosphorylation of
Ci/Gli is the conserved activation mechanism and that in mammals two kinases do
it (PMID:31279575, quoted in section 2); Zhou et al. then showed the genetic
epistasis directly — ULK4 and STK36 act in one linear pathway, ULK3 in parallel:
PMID:38096226. A single-gene knockout of one of two redundant
kinases is expected to be phenotypically silent, so the mouse results are not
evidence against the annotation.

Two further points close the loop. First, the intermediate mouse literature
already saw the cell-context dependence: PMID:18600476 and PMID:18600476 — this is the mouse IMP behind the human ISS rows for GO:0007224 and
GO:0045880. Second, zebrafish Fu does have a non-redundant Hedgehog role and
mouse Fu can supply it: PMID:19305393 and
PMID:19305393 The mammalian protein therefore retains the activity; what changed
is the degree to which the organism depends on it alone.

Conclusion for the review: the GO:0007224 and GO:0045880 rows are correct and
core, and the IBA at node PTN001122621 (donors including Drosophila fu, mouse
Stk36, human STK36 and zebrafish) is sound. The redundancy belongs in the
rationale, not in a downgrade.

5. The second, separable function: motile ciliogenesis

STK36 has a Hedgehog-independent role that is, if anything, the more essential
one in mammals.

• PMID:19305393
• PMID:19305393
• PMID:28543983
• [file:human/STK36/STK36-uniprot.txt "Essential for construction of the central pair"]
• PMID:37584603
• PMID:37584603
• PMID:37584603

Note the interesting symmetry: STK36 partners with a Costal-2-family kinesin in
both of its functions — KIF7 in the vertebrate Hedgehog module (through the
shared GLI-SUFU tip complex) and KIF27, the other mammalian Cos2 orthologue, in
central-pair construction (UniProt: "Interacts with SPAG16 and KIF27 (By
similarity)").

This is why the cilium-assembly annotations should be made motile-specific.
GO:0060271 cilium assembly is generic and covers primary cilia, but STK36-null
cells build primary cilia perfectly well — that is precisely why Hedgehog
signalling still works in them. Every ciliary phenotype reported for STK36 loss
is in motile 9+2 cilia: respiratory epithelium, ependyma, Kupffer's vesicle.
GO:0044458 motile cilium assembly states this correctly and remains a
descendant of GO:0060271, so nothing is lost.

6. The interaction rows

STK36 has ten bare GO:0005515 rows, in three groups.

(a) SUFU, two rows (PMID:10806483 co-immunoprecipitation; PMID:25241761 in
situ proximity ligation on endogenous protein in HeLa cells). The same GOA record
already carries a more informative term for this contact from the same 2000 paper:
GO:0001222 transcription corepressor binding. UniProt confirms the interaction
independently: [file:human/STK36/STK36-uniprot.txt "Q9NRP7; Q9UMX1: SUFU; NbExp=3; IntAct=EBI-863797, EBI-740595;"].
The proximity-ligation row is a genuinely independent, endogenous-protein
confirmation: PMID:25241761

(b) ULK4, two rows (PMID:32707033 kinase-interactome mass spectrometry;
PMID:37584603 yeast two-hybrid plus co-IP). This is not a generic contact: ULK4
is an STK36 substrate and its bridging adaptor.
PMID:37584603
and PMID:37584603
The informative molecular function is the kinase activity STK36 exerts on ULK4.

(c) DMWD, KLF11, GFAP, ATXN3 and SPRED1, five rows, all from PMID:32814053.
A systematic yeast two-hybrid screen of neurodegeneration-related proteins:
PMID:32814053
None has any functional follow-up for STK36, none is a Hedgehog or ciliary
protein, and none supports a more informative molecular function. These are the
removal case.

7. Notes on evidence availability

The cached record for PMID:10806483 (Murone et al. 2000) contains title and
citation metadata only — no abstract (full_text_available: false, and the
Abstract section holds just the citation block). Six GOA rows cite it: the
GO:0001222 IDA, the GO:0004674 and GO:0005524 TAS rows, the GO:0005515
SUFU IPI, and the GO:0005634 and GO:0005737 IDA rows. Rather than leave all
six UNDECIDED, each was adjudicated against the structured UniProt annotations
that cite that same paper with experimental evidence codes — SUBUNIT "Interacts
with GLI1, GLI2 and GLI3 (PubMed:10806483)", SUBCELLULAR LOCATION cytoplasm and
nucleus with ECO:0000269|PubMed:10806483, BINDING 33 ATP with
ECO:0000305|PubMed:10806483, and the reference-block line
RP ... FUNCTION, INTERACTION WITH GLI1; GLI2; GLI3 AND SUFU, TISSUE SPECIFICITY, SUBCELLULAR LOCATION, AND MUTAGENESIS OF LYS-33 — all of which are
deterministic records of what that paper reported. Every one of the six
conclusions is independently corroborated by a later experimental row, so no
action depends on the missing abstract.

8. Decisions taken, and the ones deliberately not taken

Actions: 20 ACCEPT (the kinase activity, ATP binding, cilium/axoneme/ciliary-tip
locations, the smoothened-pathway and positive-regulation rows, axoneme
assembly); 9 MODIFY (GO:0004672 to GO:0004674; the two SUFU protein-binding
rows to GO:0001222; the two ULK4 protein-binding rows to GO:0004674;
GO:0007017 to GO:0035082; all three GO:0060271 rows to GO:0044458);
5 REMOVE (the yeast two-hybrid neurodegeneration interactors); 7
KEEP_AS_NON_CORE (GO:0001222, nucleus x2, cytoplasm x3, cytosol); 2
MARK_AS_OVER_ANNOTATED (both GO:0009791 post-embryonic development rows).
Nothing left UNDECIDED.

Candidate NEW terms considered and rejected:

• GO:0090660 cerebrospinal fluid circulation. Two mouse studies tie STK36
  loss to hydrocephalus and defective CSF flow, and one of them puts the phrase
  in its title. It fails the participation test: STK36 does not move
  cerebrospinal fluid. The ependymal motile cilia do, and STK36's contribution is
  to build them. Annotating the fluid-flow process would be annotating a
  two-step-downstream organismal consequence — the same error as the
  post-embryonic development rows already marked over-annotated. The
  motile-cilium-assembly replacement is where this evidence belongs.
• GO:0003341 cilium movement / GO:0003356 regulation of cilium beat frequency. Same objection, one step closer: the axoneme and its dyneins
  generate the beat; STK36 assembles the central pair that constrains it. The
  comparator check supports this reading — mouse Stk36 carries GO:0003351
  (epithelial cilium movement involved in extracellular fluid movement) from a
  hydrocephalus-survey paper (PMID:21746835) and that is the only movement term
  on the whole family record, with no corresponding human annotation in two
  decades of curation.
• GO:0140693 molecular condensate scaffold activity or an adaptor term for the
  non-catalytic GLI2 requirement. UniProt records that GLI2 needs a function of
  STK36 that does not require an active catalytic site. That is intriguing but no
  molecular activity has been identified for it, and inventing one from a
  kinase-dead rescue result would be exactly the sort of unsupported assertion the
  guidance warns against. Raised in suggested_questions instead.

Terms verified against QuickGO before use (label, aspect, non-obsolete status, and
ancestry where it mattered): GO:0044458 motile cilium assembly (a direct child
of GO:0060271 and a sibling rather than an ancestor of GO:0035082),
GO:0035082 axoneme assembly (a descendant of both GO:0060271 and
GO:0007017), GO:0004674, GO:0106310, GO:0001222.