AIGR Focused Review — Drosophila melanogaster Hsp22 (P02515): conditional nuclear localization

Focus type: function_assignment · Hypothesis slug: conditional-nuclear-localization
Target GO term under scrutiny: GO:0005634 nucleus (CC), currently annotated IBA for P02515.

Executive Judgment

Verdict: OVER-ANNOTATED (nucleus IBA) / core location = mitochondrial matrix (strongly supported).

The proposition that Hsp22 has a demonstrated nuclear localization is not supported by any Hsp22-specific
experimental evidence. The nucleus (GO:0005634) annotation on P02515 rests solely on a phylogenetic IBA
(is_active_in, GO_REF:0000033, propagated from PANTHER sHSP tree PTHR45640). The only experimentally supported
compartment is the mitochondrial matrix (GO:0005759), which carries a direct IDA annotation from primary
work (PMID:10896659) and is corroborated by ≥4 additional independent sources. Sequence analysis is consistent:
P02515 has a positively charged, amphipathic N-terminal segment typical of a mitochondrial targeting sequence and
no canonical NLS motif.

The seed hypothesis correctly warns against conflating (i) family-level IBD inference, (ii) ambiguous historical
nuclear-fraction bands, (iii) Hsp23 paralog immunofluorescence, and (iv) same-nickname mammalian HSPB8 / plant
Hsp22. When those confounders are removed, no residual direct evidence for Hsp22 nuclear import remains.

Calibrated conclusion: Absence of evidence, not demonstrated presence. The nucleus IBA should be treated as
paralog/family over-annotation. This is a lead for curator action (remove or do-not-accept the nucleus CC),
while retaining mitochondrial matrix as the core CC. Caveat: IBA is a legitimate GO_Central pipeline call; removal
requires either a curator NOT qualifier / annotation-review, or an upstream PANTHER IBD reassessment.

Evidence Matrix

Provenance: GO annotations retrieved live from QuickGO (geneProductId=P02515, aspect cellular_component,
3 hits). Sequence pulled from UniProt REST (P02515.json). Both executed in-run; console output is in the
iteration log. CSV export to disk was blocked by the sandbox (read-only), so tables are embedded here.

Computed GO decision table (from QuickGO, iteration 2)

GO Curation Implications

• GO:0005759 mitochondrial matrix (CC) — RETAIN as core. Direct IDA, multiply corroborated. This is the
  primary, gene-specific location.
• GO:0005634 nucleus (CC) — candidate REMOVE / do-not-accept (lead). Purely IBA; conflicts with a strong
  experimental matrix localization plus an MTS-bearing, NLS-less sequence. There is no Hsp22-specific assay of
  nuclear import. For a mitochondrially-targeted, derived family member, the family IBD is a poor guide.
• GO:0005737 cytoplasm (CC) — REVIEW (weak). Also IBA-only. A transient/cytosolic pre-import pool is
  biologically plausible for any nucleus-encoded mitochondrial protein, but it is not experimentally documented
  for Hsp22 specifically; treat as non-core at best.
• MF/BP note (orientation, not the focus): the well-supported activities are unfolded protein binding /
  holdase chaperone (MF) and roles in the mitochondrial unfolded protein response and proteostasis (BP). Per the
  seed instruction, these established holdase/longevity annotations are not re-litigated here.
• Recommendation avoids "protein binding": the informative supported CC term is mitochondrial matrix.

Mechanistic Scope

The immediate, direct molecular activity of Hsp22 is ATP-independent holdase chaperone action (α-crystallin
domain, residues 44–154) that binds unfolding substrates and retards their aggregation. Its direct cellular
context is the mitochondrial matrix, where it participates in mitochondrial proteostasis and the UPR^mt.
"Nuclear localization" would be a distinct cell-biological claim about protein trafficking — not a downstream
phenotype but a direct localization statement — and it is precisely that direct claim that lacks Hsp22-specific
evidence. Longevity, oxidative-stress resistance, and UPR^mt effects are downstream physiological consequences
and are not evidence for a nuclear pool.

Conflicts and Alternatives

1. Paralog confusion (primary alternative). Nuclear/nucleolar sHSP signals in Drosophila are documented for
   Hsp23, Hsp26, Hsp27 — not Hsp22. The 1986 Hsp23 study explicitly could not exclude Hsp26/Hsp27 antibody
   cross-reaction. IBD propagation across PTHR45640 lets these paralogs' inferred nuclear activity flow onto Hsp22.
2. Historical band ambiguity. The 1980 nuclear-fraction data pre-date individual sHSP cloning; the 22/23/26/27
   kDa cluster co-migrates and was not resolved with an Hsp22-specific reagent, so a "22 kDa nuclear band" is not
   equivalent to "Hsp22 in the nucleus."
3. Same-nickname trap. Mammalian HSPB8 (human "Hsp22", cytoplasmic/nuclear in some contexts) and plant
   Hsp22 (ER/organellar) are non-orthologous to fly Hsp22; their localizations must not transfer.
4. Fractionation ≠ nucleus. PMID:3109982-type whole-insect soluble/particulate fractionation does not resolve
   a nuclear compartment.
5. Sequence signal points the other way. MTS-like N-terminus + no NLS argues against constitutive nuclear
   import.

No evidence found that positively supports a real, reproducible Hsp22 nuclear pool.

Knowledge Gaps

Discriminating Tests

1. Endogenous GFP-tag or validated antibody confocal in S2 cells and adult tissues, with MitoTracker + DAPI,
   across a heat-shock/recovery time course — the single most decisive test for a real vs. artifactual nuclear pool.
2. Digitonin fractionation / super-resolution to separate matrix from any nuclear signal; protease protection
   already supports matrix.
3. NLS/MTS swap constructs: delete N-terminal MTS and test for nuclear redistribution to probe latent import
   capacity (addresses the seed's "import-mutant mapping does not prove universal nuclear exclusion" caveat).
4. Isoform/paralog-clean reagents: ensure antibodies do not cross-react with Hsp23/26/27 (the historical
   failure mode).
5. PANTHER tree audit: confirm whether the nucleus IBD donor annotations come exclusively from cytosolic/
   nuclear sHSP paralogs, justifying a taxon/branch-specific NOT for the mitochondrial clade.

Curation Leads (require curator verification)

• Action lead: For P02515, do not accept / consider removing GO:0005634 (nucleus) as an over-propagated
  IBA; review GO:0005737 (cytoplasm) as weak/non-core; retain GO:0005759 (mitochondrial matrix) as the
  core experimental CC.
• Candidate anchor reference (retain/strengthen mito): PMID:10896659 — verify snippet on S2 mitochondrial
  colocalization + matrix fractionation/protease protection.
• Corroborating references (mito): PMID:19948727 ("localized in the mitochondrial matrix"); PMID:19420297
  ("localizes to the mitochondrial matrix"); PMID:15491684 ("localizes to the mitochondrial matrix").
• Competing/into-notes references (paralog nuclear signal, do NOT attribute to Hsp22): DOI:10.1139/g86-152 and
  the Hsp23 salivary-gland IF; DOI:10.1016/0012-1606(80)90320-6 historical nuclear-fraction bands.
• Suggested curator question: "Is any donor annotation under PANTHER PTHR45640 nucleus IBD (PTN000897708)
  based on an experimentally nuclear mitochondrial sHSP, or exclusively cytosolic/nuclear paralogs?" If the
  latter, a branch-specific NOT for the mitochondrial clade is justified.
• Suggested experiment: endogenous-level GFP-Hsp22 confocal time course with mito+nuclear co-stains.

Limitations

• No wet-lab data generated; assessment is literature + public-annotation + sequence-heuristic based.
• PANTHER family membership could not be pulled programmatically (endpoint 404); the IBD provenance statement is
  taken from the seed and the QuickGO evidence codes, not an independent tree parse.
• The 1980 Dev Biol full text was not retrieved in-run; its treatment relies on the publisher abstract / seed
  description. NLS detection used regex motif heuristics, not a trained predictor.
• "Absence of evidence" for nuclear Hsp22 is not proof of absence; a rare/conditional nuclear pool cannot be
  formally excluded, which is why the recommendation is framed as a curator lead.

Artifacts

• OpenScientist final report
• OpenScientist final report