ADCK2: does it retain the UbiB / uPKL motifs of the characterised family members?

Generated by ubib_motif_analysis.py. Do not hand-edit -- rerun the script.

Question

UniProt names ADCK2 an aarF domain-containing protein kinase, gives it EC=2.7.11.- and the keyword Serine/threonine-protein kinase, while its own FUNCTION comment says the function is unknown. The two structurally and biochemically characterised members of the family, COQ8A/ADCK3 and COQ8B/ADCK4, were shown not to carry canonical in-trans protein kinase activity (PMID:25498144, PMID:27499294): the family-specific KxGQ domain occludes the peptide-substrate groove, and the alanines of the UbiB A-rich loop (which replaces the canonical Gly-rich loop) actively suppress phosphotransfer.

So the testable questions for ADCK2 are:

  1. does it retain the catalytic machinery, i.e. is it a plausible phosphotransferase at all; and
  2. does it retain the UbiB-specific suppressors of canonical protein kinase activity?

Method

Clustal Omega MSA (EBI REST) of the UbiB family plus a canonical protein kinase as negative control, then projection of COQ8A's own UniProt-annotated motif and ligand positions through the alignment. Residue identity and "lands on a position the target itself annotates" are reported as separate conditions, never collapsed. The reference positions are cross-checked against the live UniProt feature table at runtime and the script aborts if they have drifted.

Sequences used

accession entry length status % id to COQ8A (global) % id to COQ8A (PKL core)
Q7Z695 ADCK2_HUMAN 626 aa Swiss-Prot 14.9% 14.2%
Q8NI60 COQ8A_HUMAN 647 aa Swiss-Prot - -
Q96D53 COQ8B_HUMAN 544 aa Swiss-Prot 56.3% 58.9%
Q02981 YP109_YEAST 657 aa Swiss-Prot 15.3% 16.3%
Q06567 MCP2_YEAST 569 aa Swiss-Prot 18.9% 21.8%
Q86TW2 ADCK1_HUMAN 530 aa Swiss-Prot 24.2% 24.6%
Q3MIX3 ADCK5_HUMAN 580 aa Swiss-Prot 20.2% 21.6%
P0A6A0 UBIB_ECOLI 546 aa Swiss-Prot 20.7% 21.9%
P17612 KAPCA_HUMAN 351 aa Swiss-Prot 11.9% 11.9%

COQ8A_HUMAN is the projection reference; KAPCA_HUMAN is the negative control (a canonical Ser/Thr protein kinase outside the UbiB family).

These identity figures are properties of this MSA, not of the sequence pairs, and are not comparable across alignments. Clustal Omega's gap placement depends on the whole input set, so adding or removing a single sequence moves every figure in this table. They were computed over the 9 sequences listed above; do not compare them against numbers derived from a different membership, recompute instead.

Alignment register, judged by the negative control

A column counts as in register only if the negative control's aligned position falls inside one of its own UniProt-annotated sites. Register and residue identity are different questions: the A-rich loop sits at the canonical Gly-rich-loop position, so the control is expected to land there while carrying Gly rather than Ala.

site COQ8A context ADCK2 context control residue column in register
KxGQ_K rgaal[K]lgqml gptyi[K]lgqwa - n/a - no equivalent in the control
KxGQ_Q alklg[Q]mlsiq yiklg[Q]wastr N3 n/a - no equivalent in the control
Arich_A1 eerpf[A]aasig nrepv[G]sgcva G51 yes
Arich_A3 rpfaa[A]sigqv epvgs[G]cvaqv G53 yes
Arich_S pfaaa[S]igqvh pvgsg[C]vaqvy S54 yes
beta3_K revam[K]iqypg isvav[K]vlhpg K73 yes
cat_D hfmqt[D]pnwsn nfvha[D]lhpgn D167 yes
cat_N dpnws[N]ffydp dlhpg[N]ilvqg N172 yes
DFG_D kvall[D]fgatr rlvll[D]agiva D185 NOT CONFIRMED

6 of 9 columns are confirmed in register. The control reproduces 4/4 canonical catalytic residues and 0/2 KxGQ positions, so the KxGQ motif is diagnostic for the UbiB family in this alignment and is not an artefact of aligning any protein kinase.

Note where the two conditions come apart: DFG_D is not confirmed by the strict test, because the control's UniProt entry annotates no feature at that position -- yet the control carries the same residue as the reference there. Absence of an annotation in the control is not evidence that the column is out of register, so this reads as unconfirmable rather than wrong.

Projected sites across the family

site what it is in COQ8A ADCK2 COQ8A COQ8B CQD1 CQD2 ADCK1 ADCK5 UBIB KAPCA
KxGQ_K KxGQ motif, invariant Lys; occludes peptide-substrate site K147 K276* K155* K178 K125 K102 K147 K70 -
KxGQ_Q KxGQ motif Gln Q150 Q279* Q158* Q181 Q128 Q105 Q150 Q73 N3
Arich_A1 A-rich loop A337 (Gly-rich-loop equivalent) G207* A337* A216* G239 G185 G162* A207 A131* G51*
Arich_A3 A-rich loop A339; A339G de-represses autophosphorylation G209* A339* A218* G241 A187 A164* A209 A133* G53*
Arich_S A-rich loop Ser; ATP-binding C210* S340* S219* S242 S188 S165* S210 S134* S54*
beta3_K beta3 Lys (VAIK equivalent); ATP-binding K311* K358* K237* K275 K210 K183* K228 K153* K73*
cat_D catalytic Asp, proton acceptor (HRD/HAD equivalent) D445* D488* D367* D412 D344 D315* D360 D288* D167*
cat_N catalytic-loop Asn; Mg2+ / ATP-binding N450 N493* N372* N417 N349 N320 N365 N293 N172*
DFG_D DFG-equivalent Asp; Mg2+ / ATP-binding D493 D507* D386* D477 D372 D338 D382 D310 D185

\* = the position also carries a UniProt feature annotation in that entry.

Findings for ADCK2

1. The catalytic core is intact. ADCK2 matches COQ8A at 4 of 4 canonical catalytic positions (beta3_K, cat_D, cat_N, DFG_D). ADCK2 is therefore not a pseudokinase by loss of catalytic residues; a fold-plus-lost-residues argument for removing a catalytic annotation is not available here.

2. The UbiB KxGQ motif is retained (2/2 positions identical to COQ8A: KxGQ_K, KxGQ_Q). This is the feature shown to occlude the peptide-substrate site in COQ8A, and PMID:27499294 predicts unorthodox-PKL functionality throughout the family on exactly this basis. ADCK2 is a uPKL by that criterion, which argues against a canonical protein-serine/threonine-kinase annotation.

3. The A-rich loop is NOT alanine-rich in ADCK2. Over the aligned span, COQ8A carries AAAS (337-340) and ADCK2 carries GSGC (207-210); all three projected positions fall inside ADCK2's own annotated ATP-binding site. Both COQ8A alanines project onto glycines in ADCK2. This is the one place where ADCK2 looks more like a conventional kinase than COQ8A does, and it must not be over-read: in COQ8A the A339G substitution de-represses cis autophosphorylation, which PMID:27499294 shows is dispensable for function in vivo, and it does not create in-trans peptide kinase activity while the KxGQ domain remains in place.

Reciprocal-orthologue test at the A339-equivalent position

PMID:34362905 pairs yeast Cqd1 with human ADCK2 and yeast Cqd2 with human ADCK1/5, from genetics. The A339-equivalent column tests that pairing from sequence alone, i.e. independently of the genetics -- but NOT independently of PANTHER, whose subfamily assignment is itself derived from sequence. Treat this as a second sequence-based line agreeing with the genetics, not as a third independent line.

The split is exactly the ADCK2/Cqd1 orthologue pair against everything else, including Cqd2 and ADCK1, which carry the suppressor alanine. So a single residue reproduces the published orthology assignment, and it does so reciprocally: the two branches differ at precisely the position whose substitution was shown to change COQ8A's behaviour. This corroborates the pairing; it does not by itself show the two branches differ functionally.

Only this column is branch-diagnostic. At the adjacent Arich_A1 position, Gly is carried by 4 proteins: the same ADCK2/CQD1 pair plus ADCK1 and CQD2. That set therefore cuts across the pairing rather than along it, so the claim is specifically about the A339-equivalent position and not about the A-rich loop as a whole.

Bottom line

ADCK2 has an intact phosphotransfer-competent active site inside a UbiB/uPKL architecture. The evidence supports neither removing nucleotide-related chemistry from the gene nor asserting canonical protein-serine/threonine kinase activity for it. No GO annotation on ADCK2 currently asserts either, so this analysis bears on UniProt's EC=2.7.11.- and Serine/threonine-protein kinase keyword rather than on any row of the GOA file.