AIGR Deep Research Report — A0A2U1PS28 (Artemisia annua GUF1/LepA homolog)

Hypothesis (slug: prediction-organellar-localization): ProtNLM2 predicts A0A2U1PS28 localizes to the chloroplast (GO:0009507), whereas UniProt UniRule/HAMAP (MF_03137) annotates it as mitochondrial (inner membrane / matrix). Which localization does the sequence best support?

Summary

A0A2U1PS28 is a nucleus-encoded Artemisia annua member of the highly conserved GUF1/LepA (EF-4) family of translational GTPases (ribosomal back-translocases). The curation question is purely about which organelle the protein is imported into: the ProtNLM2 computational prediction says chloroplast (GO:0009507), while the incumbent UniProt UniRule/HAMAP annotation (rule MF_03137, "Translation factor GUF1 homolog, mitochondrial") places it in the mitochondrion. This report independently resolves that conflict from sequence.

Two independent, mutually consistent computational lines of evidence favor the chloroplast. First and most decisively, land plants carry two distinct nucleus-encoded GUF1/LepA paralogs — a chloroplastic cpLEPA and a mitochondrial GUF1 — and A0A2U1PS28 is an ortholog of the chloroplastic one: it is 77.8% identical to the experimentally plastid-localized Arabidopsis protein Q9FNM5 (At5g08650) but only 47.2% to the Arabidopsis mitochondrial paralog Q9FLE4 (At5g39900), a ~30-point gap that is diagnostic of clade membership. Second, its N-terminus carries the classic chloroplast transit peptide compositional signature (Ser/Thr-rich, acidic-free, Arg-poor), matching the annotated transit peptide of Q9FNM5 and unlike a canonical Arg-rich mitochondrial presequence.

The verdict is SUPPORTED: the ProtNLM2 chloroplast prediction is the better-supported localization, and the incumbent mitochondrial annotation is best explained as paralog over-annotation — HAMAP-Rule MF_03137 (written for the mitochondrial GUF1 subfamily) misapplied to a chloroplastic paralog. The principal caveat is that the assignment rests on orthology plus composition; no dedicated targeting predictor (TargetP/DeepLoc) could be run in this environment and there is no direct experimental localization of the Artemisia protein itself. Curators should therefore treat GO:0009507 as a strong lead requiring verification.

Executive Judgment

Verdict: SUPPORTED — the ProtNLM2 chloroplast prediction is correct; the existing UniProt mitochondrial annotation is an automated misapplication (paralog/rule carry-over).

Two independent lines of computational evidence converge:

1. Orthology (decisive). By global alignment, A0A2U1PS28 (661 aa) is 77.8% identical (80.6% over min length) to the Arabidopsis chloroplastic paralog Q9FNM5 ("Translation factor GUF1 homolog, chloroplastic", At5g08650, EF4/cpLEPA), but only 47.2% (49.3%) to the Arabidopsis mitochondrial paralog Q9FLE4 (At5g39900). The ~78% value is ortholog-level identity between two eudicots; ~47% is paralog-level. This unambiguously places A0A2U1PS28 in the chloroplast (cpLEPA/EF4) clade, not the mitochondrial GUF1 clade.

2. N-terminal targeting signal. The ~62-residue N-terminus upstream of the GTPase G-domain is Ser/Thr-rich (22.6%; 35% in residues 1–20), completely devoid of acidic residues (Asp+Glu = 0%), and lacks Arg in the first 30 residues — the canonical uncharged, hydroxylated chloroplast transit peptide signature. It aligns to the experimentally supported chloroplast transit peptide (1–51) of Q9FNM5. It does not resemble a canonical amphipathic, Arg-rich mitochondrial presequence.

The chloroplast ortholog Q9FNM5 is experimentally localized to the plastid (UniProt ECO:0000269; GO:0009507 evidence HDA:TAIR), giving the assignment a strong experimental anchor by transitivity.

Caveat: Localization was not confirmed with a dedicated targeting predictor (TargetP-2.0/DeepLoc were unavailable in this environment) or by experimental localization of the Artemisia protein itself. The conclusion rests on orthology + composition, both of which are strong and mutually consistent.

Evidence Matrix

GO Curation Implications (leads — require curator verification)

• CC — retain/adopt GO:0009507 chloroplast (or more specific GO:0009570 chloroplast stroma / GO:0009534 chloroplast thylakoid via plastid ribosome association). Evidence class: ISO (from experimentally localized ortholog Q9FNM5) + sequence/orthology. ProtNLM2's chloroplast prediction should be accepted as a curation lead.
• CC — remove/replace the mitochondrial terms GO:0005743 (mitochondrial inner membrane) and GO:0005759 (mitochondrial matrix). These derive solely from HAMAP-Rule MF_03137 (the mitochondrial GUF1 subfamily rule) applied to a protein that is actually the chloroplast paralog. This is a rule-misapplication / paralog-overannotation, not biological evidence.
• MF/BP — retain GO:0003924 (GTPase activity), GO:0005525 (GTP binding), GO:0043022 (ribosome binding), and GO:0006412 (translation) / GO:0045727 (positive regulation of translation). These are family-conserved and remain valid, but the relevant compartment is the plastid ribosome (recommend re-anchoring the process to chloroplast translation). "Ribosomal back-translocase" (EF4/LepA fidelity) function is conserved.

Mechanistic Scope

Immediate molecular function: a TRAFAC-class, LepA/EF-4-subfamily translational GTPase (ribosomal back-translocase) that binds the organellar (70S-type) ribosome in a GTP-dependent manner and catalyzes back-translocation to improve translation fidelity/efficiency. The direct activity is GTP hydrolysis coupled to ribosome binding; the compartment tested here is where that ribosome resides. Sequence evidence localizes this activity to the chloroplast plastid ribosome, not the mitochondrion. Downstream phenotypes (photosynthetic efficiency, chloroplast protein steady-state levels — seen for Arabidopsis cpLEPA) are consequences, not the primary molecular function.

Conflicts and Alternatives

• Competing annotation: HAMAP MF_03137 mitochondrial call. Root cause is paralog confusion — plants encode two nuclear LepA/GUF1 genes (mito GUF1 and chloroplast cpLEPA); automated pipelines that carry a single "mitochondrial GUF1" rule can mislabel the chloroplast paralog. All UniProt subcellular evidence for A0A2U1PS28 is ECO:0000256 (automatic), none experimental.
• Dual targeting? N-terminal features favor single (chloroplast) targeting; there is no strong Arg-rich amphipathic mitochondrial motif and no positive experimental evidence for dual localization. Dual targeting cannot be formally excluded without an ambiguous-transit-peptide predictor (e.g., TargetP ambiguous class) or experimental GFP fusion.
• Draft-genome caveat: A0A2U1PS28 is from a WGS entry flagged "preliminary data" (UniProt CAUTION); the N-terminus/gene model could carry assembly/prediction error, though the transit peptide region aligns cleanly to Q9FNM5.

Knowledge Gaps

1. No experimental localization of the Artemisia protein. Checked: UniProt evidence codes (all automatic). Matters because the call is transitive. Resolve with GFP/mCherry fusion or organellar proteomics in A. annua.
2. Trained targeting predictor not run. Checked: composition heuristic only (TargetP/DeepLoc unavailable here). Resolve by running TargetP-2.0 / DeepLoc-2 / Predotar on the full sequence (expect high chloroplast probability).
3. Cleavage site / mature protein boundary not experimentally defined. Resolve with N-terminal proteomics.
4. Confirm At5g08650 = cpLEPA of 23166764. Checked: locus and description consistent; not independently verified in this run.

Discriminating Tests (most efficient first)

1. TargetP-2.0 / DeepLoc-2 on the full sequence — cheapest confirmation; predicts cTP vs mTP class directly.
2. Phylogenetic tree of plant LepA/GUF1 (target + Arabidopsis Q9FNM5/Q9FLE4 + rice/other eudicot pairs + E. coli LepA) — should place target sister to Q9FNM5 with high support.
3. In-vivo localization: N-terminal transit-peptide–GFP fusion transient expression → plastid vs mitochondrial signal.
4. Reciprocal-best-hit / synteny against Arabidopsis to formalize orthology to At5g08650.

Curation Leads (verify before applying)

• Action: Change CC from mitochondrion → chloroplast for A0A2U1PS28; treat mitochondrial CC terms as non-supported automated carry-over.
• Candidate term: GO:0009507 chloroplast (lead); consider GO:0009570/GO:0009534 if plastid ribosome sub-compartment is desired. Evidence code ISO from Q9FNM5, or IEA/computational from sequence.
• Candidate reference + snippet to verify: PMID 23166764, snippet: "LEPA is one of the most conserved translation factors and is found from bacteria to higher plants." (supports the existence and chloroplast function of the plant LepA clade the target belongs to).
• Suggested question for curator: Should the UniProt HAMAP MF_03137 rule assignment be reported as a paralog-misapplication for this and other plant chloroplast GUF1 homologs?
• Suggested experiment: transit-peptide–GFP fusion + TargetP-2.0 to confirm chloroplast import.

Cross-kingdom identity matrix (Iteration 2 provenance)

Global NW pairwise % identity (guf1_identity_matrix.png):

Interpretation: the target's only ortholog-level match (77.8%) is the chloroplast paralog; all mitochondrial/bacterial references sit at 43–47%. Independently, the two mitochondrial GUF1s (Arabidopsis↔human) share 55.0% cross-kingdom orthology and form a clade the target does not belong to. UPGMA pairs target with ARATH_chloro. This confirms plastid-clade membership and excludes the mitochondrial assignment as biological.

N-terminal alignment note (Iteration 3)

Smith-Waterman local alignment of the target N-terminal 70 residues against the N-termini of the Arabidopsis chloroplast (Q9FNM5) and mitochondrial (Q9FLE4) paralogs returns only trivial 5–6 residue matches in both cases (e.g. LSSKPP/LSS-PP). This is the expected result: organellar transit peptides diverge rapidly and are not conserved at the primary-sequence level even between true orthologs — only their amino-acid composition/physicochemical character is conserved. Therefore the transit-peptide question is best answered by (a) composition (Iteration 1: chloroplast-like cTP signature) and (b) mature-domain orthology (77.8% to the chloroplast paralog), both of which are decisive; sequence alignment of the transit peptide itself is uninformative and neither supports nor refutes on its own.

Provenance (computed values)

• Sequence length 661; G-domain P-loop HIDHGKST at res 88–95; transit-peptide window 1–62.
• N-term composition: Ser+Thr 22.6% (1–20: 35.0%), Asp+Glu 0.0%, Arg 0% in 1–30, Ala 1.6%.
• Global alignment identity: vs Q9FNM5 (chloroplastic) 77.8% (533/685; 80.6% min-len); vs Q9FLE4 (mitochondrial) 47.2% (326/691; 49.3% min-len).
• Reference orthologs: Q9FNM5 = At5g08650 (chloroplast, transit peptide 1–51, ECO:0000269, GO:0009507 HDA); Q9FLE4 = At5g39900 (mitochondrion, ECO:0000255).