ADCK1: which protein-kinase catalytic motifs does it actually retain?

Generated by ubib_motif_scan.py. Do not hand-edit — a fresh run overwrites it.

Question

UniProt Q86TW2 asserts EC=2.7.-.-, the keywords Kinase and
Serine/threonine-protein kinase, and ATP-ligand binding sites, all from the
generic protein-kinase ProRule PRU00159 (ECO:0000255) — while the same entry's
FUNCTION comment says it is not known whether ADCK1 has protein kinase activity or
what it would phosphorylate. This measures which canonical motifs are present,
rather than inferring activity from the fold name.

Method and its guards

Every motif column is anchored on a residue that PKA's own UniProt feature table
annotates (BINDING 50..58 ATP, BINDING 73 ATP, ACT_SITE 167, BINDING 169..172),
so no position is hand-assigned. Three passes with different failure modes:

  1. PASS 1, alignment-free. PKA, ADCK1 and ADCK2 each carry a curated 9-residue
    ATP BINDING site over the P-loop; comparing them column-for-column involves no
    alignment, so no alignment artefact can enter.
  2. PASS 2 / 2b, alignment. Sequences are trimmed to each protein's annotated
    "Protein kinase" DOMAIN before MAFFT L-INS-i, so long N-terminal extensions cannot
    displace catalytic columns. The run aborts unless it reproduces curated/published
    anchors, and unless PASS 2b agrees with PASS 1 wherever both apply.
  3. PASS 3, alignment-free re-test. PASS 2 reported the HRD-arginine column as a gap
    in every UbiB protein; a gap can be an artefact, so the two residues before each
    protein's own catalytic Asp are re-read straight from its sequence.

All 9/9 entries are Swiss-Prot reviewed
(tested with entryType.startswith("UniProtKB reviewed"), since "reviewed" in ...
also matches unreviewed). Each accession is additionally verified against the gene
or ORF name it is labelled with, which caught two wrong yeast accessions in the first
draft (Q12296 is MAM3, not Cqd1; Q08058 is COQ10, not Coq8).

Anchor-reproduction gate

protein motif curated/published aligned reproduced
ADCK1 beta3_lys 183 183 True
ADCK1 catalytic_asp 315 315 True
COQ8A beta3_lys 358 358 True
COQ8A catalytic_asp 488 488 True
COQ8A catalytic_asn 493 493 True
COQ8A dfg_asp 507 507 True

Plus: the alignment must place COQ8A A339 — the residue Stefely et al. 2015
(PMID:25498144) mutated to Gly to flip ADP/ATP selectivity — in PKA's second
G-rich-loop glycine column. It does.

Result 1 — the P-loop is Ala-rich, and A164 is the COQ8A A339 position

PKA's GxGxxG glycines sit at P-loop offsets [1, 3, 6].

protein curated range P-loop glycines residues at the three glycine offsets source
PKA-Ca 50..58 LGTGSFGRV 3/3 G51, G53, G56 curated
ADCK1 161..169 LGTASLAQV 1/3 G162, A164, A167 curated
ADCK5 - - 0/3 A207, A209, A212 aligned
ADCK2 206..214 VGSGCVAQV 2/3 G207, G209, A212 curated
COQ8A - - 1/3 A337, A339, G342 aligned
COQ8B - - 1/3 A216, A218, G221 aligned
Cqd2 - - 1/3 G185, A187, A190 aligned
Cqd1 - - 2/3 G239, G241, A244 aligned
Coq8 - - 1/3 A195, A197, G200 aligned

ADCK1 retains 1/3 of the canonical glycines. Its offset-3 residue is A164, which occupies the same column as COQ8A A339 — the ADP/ATP selectivity determinant. ADCK1's BINDING sites are annotated with ATP as ligand purely by ProRule transfer; the one characterised relative sharing this loop is ADP-selective and carries GO:0043531 ADP binding by IDA, not GO:0005524.

Result 2 — the phosphotransfer active site is INTACT

protein β3 Lys catalytic Asp catalytic Asn Mg-binding Asp catalytic loop (ends at the Asp) Arg at D-1
PKA-Ca K73* D167* N172* D185 YRD yes
ADCK1 K183* D315* N320 D338 HCD no
ADCK5 K228 D360 N365 D382 HSD no
ADCK2 K311* D445* N450 D493 HAD no
COQ8A K358* D488* N493* D507* QTD no
COQ8B K237* D367* N372* D386* QTD no
Cqd2 K210 D344 N349 D372 HCD no
Cqd1 K275 D412 N417 D477 HAD no
Coq8 K216* D346* N351* D365* QTD no

* = the transferred column lands on a site that protein's own feature table annotates.

All four core catalytic residues are canonical in 8/8 of the UbiB
proteins examined, ADCK1 included (K183, D315, N320, D338). This is the result that
refuses a "pseudokinase, catalytic residues lost" reading, and it agrees with
Kemmerer et al. 2021, who found the yeast ortholog Cqd2's function requires exactly
these residues (PMID:34362905).

Result 3 — but the HRD arginine is absent family-wide

0/8 UbiB proteins have an arginine
immediately before the catalytic aspartate; the PKA control does. Read alignment-free,
so this is biology rather than a gap artefact. ADCK1 reads
HCD (H313-C314-D315), identical to yeast Cqd2 —
a small independent corroboration of the orthology PANTHER asserts, and distinct from
ADCK2/Cqd1 (HAD) and the COQ8
clade (QTD).

What this supports, and what it does not

Supports: removing Serine/threonine-protein kinase from UniProt's keyword set.
Protein-kinase activity has been demonstrated exactly once in this family — an ancestrally
reconstructed COQ8B phosphorylates COQ3 in trans (PMID:38425362) — and that work did not
determine the residue class, so no UbiB protein has been shown to be a Ser/Thr kinase.
Meanwhile COQ8A and COQ8B both carry explicit GO:0004672 NOT|enables IDA rows, and ADCK1's
own FUNCTION comment states its substrate class is unknown. The specificity claim has no basis.

Does not support: calling ADCK1 catalytically dead. The active site is intact, the
yeast ortholog needs it, and no in vitro assay of purified ADCK1 has ever been published.
The nucleotide site is untested, not refuted, so no molecular-function or
nucleotide-binding term is proposed for ADCK1 in this review.

Reproduce

uv run python ubib_motif_scan.py

Requires mafft on PATH. UniProt responses are cached under cache/; delete it to
refetch. A missing input is a hard error naming the fix, never a silently degraded run.