GALNT3 is a Golgi membrane enzyme that transfers GalNAc from UDP-GalNAc onto protein serine/threonine residues. Human enzyme and cell experiments support this catalytic core and its role in O-GalNAc glycosylation. FGF23 is an important substrate, but the enzyme also acts on mucin-sequence acceptors. The core therefore describes the chemistry without restricting every GALNT3 activity to FGF23.
The 1996 cloning/expression study tested a human GALNT3 construct lacking its membrane anchor, expressed in Sf9 cells. Selected original Methods, Results and Table I were read in the author-uploaded article. The assay used culture supernatants and radiolabeled UDP-GalNAc incorporation into synthetic Muc2, Muc5c and HIV-V3 acceptors with controls. It supports enzyme activity, not physiological HIV infection or activity of a naturally short splice product.
The 1997 purified-enzyme study characterized soluble human GALNT1/2/3 produced in insect cells. Selected original enzyme-preparation, peptide-assay and metal-substitution sections were read in the author-uploaded article; source figure images were not independently inspected. Its manganese-replacement experiment used 5, 10 or 20 mM calcium or magnesium with Muc2 substrate and reported no detectable activity with either substitute. This supports manganese-dependent catalysis. It does not demonstrate that calcium cannot bind, so the calcium-binding annotation remains UNDECIDED. Manganese binding is accepted as part of the catalytic core, without adding a second core function. Human residue features at 277/279/415 are ECO:0000250|UniProtKB:H0ZAB5 transfers from zebra finch, distinct from the direct human functional cofactor experiment. Missing calcium features, the lectin fold and a paralog annotation comparison do not convert absent turnover into a negative binding measurement.
The 2006 FGF23 study is available as an authentic abstract and selected indexed original human transfection/peptide sections; complete original figure inspection was not achieved. Its title's secretion wording is interpreted with the later mechanistic study. The 2007 HHS study was read at abstract level and in selected author-uploaded original Methods/Results and captions: human HOS-TE85 cells, GALNT3 siRNAs and FGF23 constructs connect GALNT3 depletion to altered FGF23 processing. CHO-produced substrate preparations and human-cell silencing are distinct experimental systems. No source images or raw MS measurements were independently reanalyzed.
The 2020 mechanistic study has complete native cached text. Selected Methods, Results and discussion support human enzyme assays, CHO-cell experiments, AFM/SAXS and cooperation between catalytic and lectin domains. The complex crystal structures instead use zebra finch GALNT3; the human crystallization attempts failed. The human 92–633 and finch 91–631 soluble proteins are engineered constructs. They are not the natural short human product. The short pulse-labeling experiment did not show a marked increase in FGF23 secretion, and the later excess of intact extracellular hormone was attributed mainly to stability after secretion. This distinction replaces the inherited unconditional secretion-rate interpretation.
The human HeLa localization study is currently accessible here as an authentic complete abstract, not full original Methods/images. It explicitly reports endogenous and tagged human GALNT1/2/3 localization by fractionation, immunofluorescence and immunoelectron microscopy. GALNT3 occurs in Golgi stacks, preferentially in their medial portion. This supports the Golgi/cisterna location without exclusivity. The current HPA GALNT3 page lists supported Golgi apparatus localization; individual images were not independently scored.
The ocular epithelial study used isoenzyme-specific antibodies including GALNT3 and reports positive target distribution in human corneal and conjunctival epithelia. The perinuclear location is retained as non-core by combining that target-specific imaging with independently established Golgi residence and deferring to the curator. Complete original GALNT3 localization panels remain inaccessible. The abstract's explicit supranuclear statement is not reassigned to GALNT3, and no paralog error is alleged.
For PMID:18570454(https://doi.org/10.1021/pr800225z), the normal repository cache is abstract-only, but an authentic author-hosted original PDF was already retained. Methods/Results were read and Table 2 on printed page 3478 was visually inspected. The portable target-row facts are:
| Table locator | Historical identifier | Reported matched peptides | Fraction |
|---|---|---|---|
| Table 2, UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferase entry | GI1617312 | 1 | 3 |
The official NCBI protein record CAA63371.1 resolves GI1617312 to human GALNT3/Q14435. The original PDF is 1,005,190 bytes, SHA256 7d3b945b98a6696ca6146956da71378304b2de9f40867ee86729248a395372fc. HB1.F3 conditioned medium underwent flow field-flow fractionation, TEM and LC-MS/MS. Table 1 reports fraction-3 particle diameters of 34.0 ± 1.40 nm (16 particles). Results on printed page 3480 require Mascot scores above 50 and repeated detection across three shotgun analyses for single-peptide identifications. The reported target peptide is therefore more than an unvalidated one-spectrum observation, although its raw spectra were not independently inspected.
This supports the non-core vesicle-fraction association and neural-stem-cell context. Technical repetition is not independent biological replication, and co-isolated material remains possible. Neither membrane orientation nor extracellular catalytic activity or vesicle biogenesis is asserted. Those unestablished properties are not prerequisites for retaining the observed fraction membership.
The immutable repository summaries were retained. Successful official ContentService records were read privately to resolve the direct event/catalyst/member chains; no primary-paper assay is inferred merely from the model.
| Source record | Verified GALNT3 chain and scope |
|---|---|
| 8851619 | Catalyst 8851625 → entity 913620 → human Q14435, at Golgi membrane; UDP-GalNAc-dependent FGF23 modification. The substrate and enzyme compartments are distinguished. |
| 913675 | Catalyst 913697 → set 913644 → GALNT3 913620 as an explicit hasMember, supporting the normal mucin GalNAc-transfer reaction without identical specificity across all GALNTs. |
| 913709 | Mucin O-glycosylation pathway directly includes normal reaction 913675. |
| 5096537 | FailedReaction with GALNT3 C574G, Q592, R162 and T272K variants; normalReaction is 913675. Mutant catalytic failure is not positive wild-type activity evidence. |
| 5083625 | Disease pathway contains failed reaction 5096537. The successful external official record was read; its native disease-pathway summary is now also cached. |
| 190372 | Includes upstream GALNT3/FGF23 glycosylation 8851619 before secretion, receptor binding and receptor phosphorylation. Target membership is genuine. |
The last pathway membership does not establish execution of the receptor-triggered molecular signals defined by GO:0008543. GALNT3's upstream chemistry and regulatory relevance are preserved while the signaling annotation is marked over-annotated. No replacement or NEW process assertion is added.
The official Q14435 record, direct product FASTAs and variation reconstruction agree on the 633-residue canonical and 192-residue short products. The short product changes canonical residues 173–192 and lacks 193–633, removing most of the catalytic region and the lectin domain. Its function is unresolved. Predicted topology and zebra-finch-derived human site features are identified by their actual evidence classes rather than called direct human structural observations.
The actual PANTHER19 tree, alignment and target attributes, together with retained IBD records, place human GALNT3 at AN993/PTN002481705, with exact canonical sequence identity. The two catalytic/process assertions originate at AN1/PTN004287905; the Golgi assertion originates at AN11/PTN000187024. All are on the target lineage, with no target-path negative placement. Incoming and literal retained IBD donor sets agree. Target self-evidence is valid descendant grounding. No family label or node was inferred from donor counts.
The six refinements use verified relationships: GO:0016266 is_a GO:0006493 (four process contexts); GO:0000139 is_a GO:0098588 is_a GO:0031090 is_a GO:0016020 (membrane); and GO:0005794 part_of GO:0005737 (cytoplasm). Cytoplasm includes organelles and is not cytosol. The informative Golgi organelle annotations remain ACCEPT. The Golgi cisterna claim is independently supported; the retained query did not establish a typed GO:0032580 → GO:0000139 path, and no such path is invented.
The current review has 35 contexts, two products, one catalytic core and 23 assessed references: 25 ACCEPT, 6 MODIFY, 2 KEEP_AS_NON_CORE, 1 UNDECIDED and 1 MARK_AS_OVER_ANNOTATED. Calcium binding remains unresolved. No NEW annotation is proposed. The two FGF23-input atoms and neural-stem-cell atom preserve the literal original contexts.
The compact source evidence retains current machine fields, sequences and source identities, plus immutable main-history locators for the old review, journal, provider artifacts and all 48 old quotation associations. Those old associations are historical; they are not silently repaired or presented as newly read evidence. Earlier journal entries remain dated records and are superseded by the current assessment where explicitly stated.
The six existing excerpts remain exact. Five additional short anchors now cover the UniProt catalytic description and manganese cofactor, the 2006 FGF23 glycosylation result, the ocular study's target-specific distribution and the original vesicle-table row. Every annotation has an explicit supported_by link to its relevant source; reference-only links reuse evidence without repeating quotations. The disease pathway R-HSA-5083625 is cited on the failed-reaction contexts as disease provenance, not as evidence of wild-type catalysis.
The portable original-source excerpt gives the exact Table 2 target line, accession mapping, public original URLs and selected Methods/Results locators. Its text is also cited from the EV review. All occurrence totals, including this repetition, stay within 25 words per source: PMID:9295285 10, PMID:31932717 18, PMID:17129170 8, PMID:8663203 10, PMID:9394011 11, PMID:12506059 9, PMID:16638743 17, PMID:18570454 12, and the UniProt record 8. Reference notes distinguish abstract-only caches from selected original sections; this portable excerpt does not turn the PMID cache into full text.