DNAAF3 (HGNC:30492; UniProt Q8N9W5), formerly C19orf51, supports cytoplasmic maturation of outer and inner axonemal dynein arms. The eight reviewed source assertions represent nine raw QuickGO records: two DynAP ISS records collapse to one identical projected annotation in the ordinary pipeline. All eight imported source dictionaries are preserved. Three localization and two motile-cilium assembly assertions are accepted; three axonemal-dynein assembly assertions receive specific replacements. One assembly core is retained, with no additional annotation or assigned catalytic molecular function.
The 2012 primary study reports loss of DNAH5, DNAH9, DNAI2 and DNALI1 from affected human respiratory cilia, with electron microscopy supporting outer and inner arm defects (Results, Figures 3–4). Zebrafish morphants retain cilia but lose arms and motility (Figure 5). Chlamydomonas genomic rescue and Myc-tagged PF22 restore assembly; biochemical fractionation places that tagged ortholog in the cytoplasm, not the heavily loaded axoneme fraction (Figure 1). This is not direct human DNAAF3 imaging.
The experiments resolve an assembly-competence defect, not a complete absence of protein association. In Figure 7, HA-IC2 still co-precipitates normal amounts of HCα/β and reduced HCγ from pf22 cytoplasm. Abnormal complexes form but do not complete productive assembly. Inner-arm dependence also varies: Chlamydomonas b/c are affected while f is retained. Neither total loss of every dynein subtype nor an autonomous folding activity is inferred. The original Methods sections on gene cloning, epitope tags, immunoprecipitation and human immunofluorescence were recovered separately because the normal full-text cache omits them.
The replacements for GO:0070286 are outer dynein arm assembly and inner dynein arm assembly, both verified children of that term. Their checked is_a/part_of ancestor paths do not contain motile cilium assembly, so the two original motile-cilium assertions are retained separately as ACCEPT. DNAAF3 performs constituent dynein-arm maturation needed to assemble the functional motile structure; persistent ciliary projections do not negate this work. The three processes belong to one integrated assembly core. These conclusions use the 2012 human arm defects and ortholog rescue/complex experiments, not an assertion that DNAAF3 makes every ciliary component.
The 2018 DynAP study provides the direct donor evidence for the human localization transfer. Its Methods specify DNAAF3 cloned from a Xenopus cDNA library. The actual Figure 2C labels Dnaaf3, Ktu and their merged image, showing the fluorescent fusion concentrated in cytoplasmic DynAPs. The original Results, imaging Methods and controls for nonspecific aggregation were read. Human supplemental immunostaining instead examines LRRC6 and DNAI1; it does not directly localize endogenous human DNAAF3.
The literal donor Q32NQ7 is Xenopus laevis dnaaf3 and carries direct IDA to GO:0120293 from that paper. Its separate assembly-process transfer from the human gene does not make this independently measured localization circular. Official SL-0086 and SL-0542 mappings reproduce cytoplasm and DynAP, respectively. Cytoplasm records the preassembly compartment, while DynAP records its observed concentration in multiciliated cells. Neither location is claimed to be exclusive or demonstrated for every human product.
The 2024 primary study combines structural predictions with Chlamydomonas wild-type and pf22-null dynein mass spectrometry. No arginine-methylated peptides were identified in the mutant preparations, while unmodified counterparts were detected; limited lysine methylation remained. This supports an arginine-methylation role for DNAAF3/PF22. It does not measure purified human DNAAF3 turnover. The Rossmann-like fold and conserved cofactor-pocket features support the authors' methyltransferase model; the CLEAN prediction and mass shifts do not establish a precise type-II reaction or symmetric dimethylarginine chemistry.
The 2025 comparative study finds methylated dyneins across many lineages, but not convincingly in D. melanogaster sperm, whose assembly still requires Dnaaf3. D. willistoni provides a contrasting positive case. This comparison is not active-site mutagenesis. The authors explicitly leave catalytic versus noncatalytic assembly contributions unresolved and propose enzyme-dead/truncation experiments. Its gel-filtration result is compatible with the abnormal/unstable intermediates reported in 2012. The review preserves the supported methyltransferase model in prose and a biological knowledge gap without manufacturing a catalytic annotation. Raw spectra, supplementary datasets and figure pixels from these two methylation papers were not independently reanalysed.
The actual PTHR22118 tree places human Q8N9W5 at leaf PTN002502984 below the stated PTN001858482 root. Both source processes are present. The reused official IBD snapshot dated 2026-08-28 contains fly and human seeds; current raw QuickGO additionally lists zebrafish. This version distinction is preserved. A short donor list or the human target's inclusion as experimental grounding is not treated as a propagation failure. IPR039304 is the verified DNAAF3 family with a canonical target match over residues 1–470 and an explicit GO:0070286 mapping.
The official human record describes five products, independently reconstructed against official FASTAs: Q8N9W5-1 (541 aa), -5 (95 aa; named isoform 2), -3 (608 aa), -6 (588 aa; named isoform 4) and -7 (487 aa; named isoform 5). The 95-residue product has an official predicted nonsense-mediated decay caution; stable expression and equivalent activity are not established. Product -6 is a 47-residue N-terminal extension, consistent with the historical 588-residue reference and Leu108 versus canonical Leu61 numbering. This correspondence does not convert gene-level experiments into an accession-specific natural-product assay.
All four cited PMID caches are genuine normal pipeline outputs marked as full text from XML. The 2012 cache nevertheless omits Methods; separate original XML supplied those sections. The DynAP cache omits target table/image content, so the verified original XML and Figure 2C were inspected separately. Initial rate limits for the newer caches were retained; later ordinary retries succeeded. Falcon authentication and fallback quota failures produced no research report, and no provider text was fabricated.
Each short exact excerpt occurs once. For the 2012 assembly source, a nine-word patient-arm excerpt is at the motile-cilium IMP decision, an eleven-word concluding assembly excerpt is at the axonemal-dynein IMP decision, and a two-word mechanistic excerpt is in the biological notes: 22 words total. The DynAP source retains eight words, the 2024 methylation observation five, and the 2025 mechanistic question nine. Other support uses source references and the precise public locators above. The number of quotations is not the number of evidence observations.