AIGR Gene Hypothesis Deep Research — Final Report

Target: Drosophila melanogaster ttv (UniProt D5SHU8) — glycosyltransferase activity (GO:0016757)

Target gene: Drosophila melanogaster ttv (FlyBase FBgn0265974 / CG10117), UniProt D5SHU8 (299 aa, TrEMBL/unreviewed).
Term under test: glycosyltransferase activity — GO:0016757 (Molecular Function).
Seed hypothesis: "The Drosophila melanogaster protein D5SHU8 has glycosyltransferase activity."
Investigation: 3 iterations completed; 3 confirmed findings; 5 primary/structural literature sources reviewed.

Summary

The seed hypothesis is partially supported. Three independent lines of sequence and homology evidence establish beyond reasonable doubt that D5SHU8 is a genuine, deeply conserved fragment of the Drosophila EXT1 ortholog tout-velu (ttv). The 299-aa sequence aligns co-linearly and at 100% identity to residues 462–760 of the canonical, reviewed ttv protein Q9V730 (760 aa), and is 55.7% identical (Smith-Waterman, 166/298 aa) to the C-terminal catalytic region of human EXT1 (Q16394[466–744]). This region corresponds precisely to the isolated C-terminal GT64 (α1,4-GlcNAc-transferase) catalytic domain (Pfam PF09258; InterPro IPR015338/IPR004263), and it retains the active-site aspartate (canonical Asp670 → D5SHU8 Asp209), the DED binding motif, and 8 of 9 known catalytic/substrate-binding residues. On these grounds, a broad glycosyltransferase-activity term (GO:0016757) is a defensible, homology-based annotation for this sequence — it is not a spurious match, a divergent paralog, or a mis-mapped fragment.

However, the hypothesis as literally stated — that this exact truncated sequence is a catalytically competent glycosyltransferase — is not directly demonstrated, and several structural facts argue for caution. D5SHU8 is a truncated short isoform beginning at internal Met462; it lacks the N-terminal type-II transmembrane signal-anchor (canonical residues ~7–25) that targets ttv to the Golgi, and it lacks the entire N-terminal GT47/exostosin glucuronyltransferase (GlcA-transferase) domain. The only direct enzymatic evidence for the ttv/EXT1 protein comes from assays on the full-length protein, which further showed that robust dual glycosyltransferase activity is partner-dependent — markedly augmented by co-expression with SOTV/EXT2. No published study reports catalytic activity for an isolated EXT1/ttv GT64 domain, nor for this specific 299-aa isoform.

Bottom line for the curator: GO:0016757 should be retained for D5SHU8 only as an inferred, generic, isoform-caveated annotation with an evidence code appropriate to sequence/structural homology (e.g., ISS or IEA), not upgraded to a directly-assayed, more specific molecular-function term (such as a specific GlcNAc-transferase EC-level term) and not paired with a confident Golgi-membrane cellular-component assertion for this exact sequence. The most important caveats are the missing GT47 domain, the missing membrane anchor, the partner-dependence of the demonstrated activity, and the complete absence of any isolated-domain enzymatic assay.

Identity and Sequence Provenance

The 299-aa length and exact residue string match the frozen FASTA, and the 100% co-linear mapping to Q9V730 fixes the identity unambiguously. The record is an N-terminally truncated cDNA product (internal Met462 start) representing the isolated C-terminal GT64 domain.

Key Findings

Finding 1 — D5SHU8 is the isolated C-terminal GT64 domain of ttv/EXT1 (a truncated short isoform)

The 299-residue D5SHU8 sequence aligns co-linearly and with 100% identity (299/299) to residues 462–760 of the canonical reviewed ttv protein Q9V730 (760 aa), beginning at canonical offset 461. It thus corresponds precisely to the C-terminal 299 residues of the full-length protein with no substitutions. This C-terminal region carries a GT64 glycosyltransferase domain (Pfam PF09258; InterPro IPR015338 "GT64_dom"; associated with IPR004263 Exostosin), spanning approximately D5SHU8 residues 16–280.

Crucially, the truncation removes two functionally essential N-terminal features of full-length ttv/EXT1: (1) the type-II transmembrane signal-anchor (canonical residues ~7–25), the sole membrane-targeting determinant that localizes ttv to the Golgi apparatus where heparan sulfate is polymerized; and (2) the entire N-terminal GT47/exostosin domain, which encodes the glucuronyltransferase (GlcA-transferase, GlcAT-II) half of the bifunctional EXT1 co-polymerase.

Of the nine canonical catalytic/substrate-binding residues, 8 are retained in D5SHU8, including the active-site Asp670 → Asp209 and the DED binding motif (canonical 581–583 → D5SHU8 120–122). The single missing binding residue, Arg437, lies in the deleted N-terminal GT47 region. Thus D5SHU8 preserves the GlcNAc-transferase (GlcNAcT-II) active-site machinery but discards the glucuronyltransferase machinery — it is a single-domain, half-enzyme fragment relative to the bifunctional full-length protein.

Finding 2 — Direct enzyme evidence exists for full-length TTV, not for the isolated GT64 isoform

The decisive biochemical evidence in the literature concerns the full-length protein. Izumikawa et al. 2006 (PMID: 16303756) demonstrated in vitro that TTV/DEXT1 and SOTV/DEXT2 have GlcNAcT-II and GlcAT-II activities required for biosynthesis of the repeating disaccharide units of the heparan sulfate (HS) backbone, and — critically — that co-expression of TTV with SOTV markedly augmented both glycosyltransferase activities compared with TTV or SOTV expressed alone. The TTV–SOTV complex lacked GlcNAcT-I (chain initiation), which requires BOTV/EXTL3. This establishes two facts that qualify the seed hypothesis: (a) the demonstrated activity is a property of the full-length protein, and (b) full activity is partner-dependent, not a robust autonomous property.

Complementary in vivo evidence from Dasgupta et al. 2007 (PMID: 17610078) showed that the human EXT1 gene functionally complements the ttv mutation in Drosophila — rescuing a ttv-null mutant to adulthood and restoring glycosaminoglycan (GAG) biosynthesis — firmly establishing ttv as the functional Drosophila EXT1 ortholog with a bona fide glycosyltransferase role in vivo. The structural work on the EXT1–EXT2 co-polymerase (PMID: 36593275) confirms that the HS backbone is synthesized by an obligate EXT1–EXT2 co-polymerase complex. None of these studies reports catalytic activity for an isolated EXT1/ttv GT64 domain or for this specific 299-aa isoform. The activity is therefore transferred to D5SHU8 by homology, not measured on it.

Verified snippet (PMID 16303756): "we demonstrated that TTV/DEXT1 and SOTV/DEXT2 had GlcNAcT-II and GlcAT-II activities required for the biosynthesis of repeating disaccharide units of the HS backbone, and the coexpression of TTV with SOTV markedly augmented both glycosyltransferase activities when compared with the expression of TTV or SOTV alone."

Verified snippet (PMID 17610078): "the human EXT1 (hEXT1) gene is conserved through species and can functionally complement the ttv mutation in Drosophila. The hEXT1 gene was able to rescue a ttv null mutant to adulthood and restore GAG biosynthesis."

Finding 3 — D5SHU8 GT64 domain is 55.7% identical to human EXT1 C-terminal catalytic region

An independent Smith-Waterman local alignment (BLOSUM62 substitution matrix) of the full 299-aa D5SHU8 against human EXT1 (Q16394, 746 aa) yields a single high-scoring block: D5SHU8[2–295] vs EXT1[466–744], 166/298 = 55.7% identity, alignment score 838. This block is precisely the C-terminal GT64 (GlcNAc-transferase) region of EXT1. Combined with the 100% co-linear identity to canonical Drosophila ttv Q9V730[462–760], this cross-species conservation confirms that D5SHU8 is a bona fide, deeply conserved EXT1/ttv GT64 catalytic domain — not a spurious annotation, a divergent paralog, or a mis-mapped fragment. High sequence identity retained across ~500 million years of fly–human divergence is consistent with a functionally constrained catalytic module. (An earlier naïve same-offset comparison had given ~35%; the rigorous local alignment corrects this to 55.7% and decisively confirms the conserved fold.)

Mechanistic Model / Interpretation

The full-length ttv/EXT1 protein is a bifunctional, Golgi-resident type-II membrane glycosyltransferase that, together with its partner SOTV/EXT2, forms the HS co-polymerase that builds the (-GlcAβ1,4-GlcNAcα1,4-)n heparan sulfate backbone. Its two catalytic activities are spatially segregated into two domains:

Full-length ttv/EXT1 (canonical Q9V730, 760 aa)
┌───────┬───────────────────────────┬───────────────────────────────┐
│ TM    │  GT47 / exostosin domain  │   GT64 domain (PF09258)        │
│ anchor│  GlcA-transferase (GlcAT) │   GlcNAc-transferase (GlcNAcT) │
│ ~7-25 │  binding residue Arg437   │   active-site Asp670; DED motif│
└───────┴───────────────────────────┴───────────────────────────────┘
   Golgi         [ REMOVED in D5SHU8 ]        [ = D5SHU8, res 462-760 ]
 targeting

D5SHU8 (299 aa) = C-terminal GT64 domain ONLY
  ├─ retains active-site Asp209 + DED motif (120-122)
  ├─ retains 8/9 catalytic/binding residues
  ├─ LACKS TM signal-anchor  → no Golgi targeting
  └─ LACKS GT47/GlcA-transferase → half-enzyme

Interpretation. D5SHU8 is a structurally plausible catalytic module — it preserves the GlcNAc-transferase active site and the substrate-binding DED motif of a well-characterized enzyme family. This is why automated pipelines (ARBA/UniRule/InterPro2GO) legitimately propagate GO:0016757 to it: the GT64 domain signature maps to the glycosyltransferase parent term. However, three mechanistic gaps separate "structural plausibility" from "demonstrated function of this sequence":

1. Domain amputation. The demonstrated in vitro TTV activity is bifunctional (both GlcNAcT-II and GlcAT-II). D5SHU8 has only the GT64 half; the GlcAT-II activity depends on the deleted GT47 domain. At best D5SHU8 could carry only the GlcNAcT-II half-activity, and even that has never been tested in isolation.
2. Partner dependence. Izumikawa et al. showed activity is markedly augmented by SOTV/EXT2 co-expression, and the structural study shows an obligate co-polymerase complex. Whether the isolated GT64 domain can fold and catalyze without the rest of the protein and without its partner is unknown.
3. Loss of targeting. Without the transmembrane anchor, D5SHU8 would not be retained in the Golgi in vivo, so even if intrinsically active it could not participate in the physiological glycan-biosynthesis pathway in its native compartment.

The seed's "decisive question" — intrinsic isolated-domain catalysis vs. an in-vivo glycan-biosynthesis role — resolves as: neither has been experimentally demonstrated for this exact sequence. The glycosyltransferase assignment rests on homology to a domain that is part of a demonstrated enzyme, which is a reasonable but not decisive basis.

Evidence Base / Evidence Matrix

GO Curation Implications

Lead requiring curator verification.

GO Decision Table

Net: the current prediction "F:glycosyltransferase activity" is correct in direction but should not be strengthened; treat it as a non-core, homology-inferred, isoform-caveated annotation. Avoid "protein binding" as a summary; the informative call here is the GT64 catalytic-domain MF.

Mechanistic Scope

The immediate molecular function under test is catalysis of glycosyl transfer — specifically, in the ttv/EXT1 context, transfer of N-acetylglucosamine (GlcNAcT-II) during heparan sulfate chain elongation. This is distinct from, and must be separated from:

• Downstream pathway consequences — HS/HSPG biosynthesis (a BP), the aggregate output of the EXT1–EXT2 co-polymerase, not a direct readout of the isolated GT64 domain.
• Developmental/phenotypic outcomes — morphogen (Hh, Dpp, Wg) gradient formation and wing patterning, which are pleiotropic loss-of-function phenotypes of ttv mutation (PMID: 14645127, PMID: 14998928), several steps removed from the enzymatic act. They support the gene's role in glycan biosynthesis but do not demonstrate that the D5SHU8 fragment catalyzes anything.
• Disease manifestation — human EXT1 loss causes hereditary multiple exostoses; relevant to the family's importance but not evidence about this isoform's activity.

The seed hypothesis is a function-assignment claim about direct gene-product activity. The evidence supports the direct activity of the full-length gene product and the homologous plausibility of the D5SHU8 domain, but not a measured direct activity of D5SHU8 itself.

Conflicts and Alternatives

1. Isoform truncation (primary conflict). The record is an N-terminally truncated cDNA product (internal Met462 start) — the C-terminal GT64 half only. Automated annotation transfers "glycosyltransferase activity" from the domain, but the deleted GT47 domain and TM anchor mean the isoform is not a faithful stand-in for the full enzyme.
2. Partner-dependence. Izumikawa et al. show activity is markedly augmented by SOTV/EXT2 and the structural work shows an obligate complex. An isolated, anchor-less single domain may not be autonomously catalytic.
3. "Closely related to glycosyltransferase activities" hedging. The EXT-family literature (PMID: 14645127) itself uses cautious phrasing, reflecting that domain presence ≠ proven catalysis for every member.
4. Database carry-over risk. The GO:0016757 assignment likely originates from InterPro2GO/ARBA mapping of PF09258 and from the full ortholog; auto-annotated CC terms ("ER/Golgi membrane") partly contradict the truncated sequence (no TM anchor) — a source of over-annotation if applied uncritically.
5. No paralog confusion detected. The 100% identity to ttv Q9V730 and 55.7% to human EXT1 (not EXTL1/2/3) rules out mis-assignment to a different EXT paralog. Identity is secure; only isoform completeness and measured activity are in question.

Limitations and Knowledge Gaps

Additional caveats: domain-boundary coordinates for GT47/GT64 are approximate; catalytic-residue calls are UniProt feature transfers (inference, not assay).

Discriminating Tests

1. In vitro glycosyltransferase assay on recombinant D5SHU8 (residues 462–760), measuring GlcNAcT-II activity with UDP-GlcNAc donor and a heparosan/oligosaccharide acceptor, with and without co-expressed SOTV/EXT2 — the single most decisive experiment for intrinsic vs. partner-dependent catalysis.
2. Folding/stability check (CD, thermal shift, or limited proteolysis) on the isolated GT64 domain to determine whether it forms a stable fold without the GT47 domain.
3. Domain-swap/complementation: does D5SHU8 rescue any ttv-null phenotype? (Prediction: no, due to missing anchor + GT47.)
4. Expression/topology check: GFP-tag localization — predict soluble/cytosolic, not Golgi-anchored, unlike full-length.
5. FlyBase/RNA-seq isoform verification to confirm whether the 299-aa form is a genuinely expressed transcript or a partial-cDNA artifact.
6. Structural comparison / modeling of D5SHU8 against the EXT1–EXT2 cryo-EM structure (PMID: 36593275) — used as a hypothesis-generator only, not as validation.

Proposed Follow-up Experiments / Curation Leads

All labeled as leads requiring curator verification.

• Curation action (immediate): Keep GO:0016757 on D5SHU8 with a homology/electronic evidence code and add an isoform caveat noting (a) truncation to the GT64 domain, (b) absence of the GT47/GlcAT-II domain and TM anchor, and (c) that direct activity is demonstrated only for full-length TTV and is partner-dependent. Do not add specific child MF terms or a confident Golgi CC term for this record.
• Reference to attach (verify snippet): PMID: 16303756 — "TTV/DEXT1 and SOTV/DEXT2 had GlcNAcT-II and GlcAT-II activities … coexpression of TTV with SOTV markedly augmented both glycosyltransferase activities…" — label as support for the full-length gene product, qualifying for the isoform.
• Reference to attach (verify snippet): PMID: 17610078 — "the human EXT1 (hEXT1) gene … can functionally complement the ttv mutation … restore GAG biosynthesis." — supports ortholog identity/in-vivo function at the gene level.
• Experiment to prioritize: the in vitro isolated-domain GlcNAcT-II assay (Discriminating Test 1), which would move the annotation from homology-inferred to directly-tested for this sequence — resolving the core of the "decisive question."
• Suggested curator question: Is the 299-aa D5SHU8 a curated biological isoform or a partial cDNA? The answer determines whether any localization/process annotations should be propagated to this record at all.

Provenance Notes

• Sequence identity established computationally: D5SHU8 (299 aa) vs canonical ttv Q9V730 (760 aa): co-linear, offset 461, 299/299 = 100% identity → residues 462–760. D5SHU8[2–295] vs human EXT1 Q16394[466–744]: Smith-Waterman (BLOSUM62), 166/298 = 55.7% identity, score 838.
• Domain assignment: GT64 (Pfam PF09258; InterPro IPR015338; Exostosin IPR004263), D5SHU8 residues ~16–280. Catalytic Asp670→Asp209; DED motif 581–583→120–122; 8/9 catalytic/binding residues retained; Arg437 (GT47 region) absent.
• Identity cross-references: gene ttv/CG10117/FBgn0265974; cDNA clone MIP20656p / GenBank ADG57806.1; RefSeq NP_001260972.1.
• Frozen input: 299 residues, SHA-256 76d919802eaf321c50ecd217e0e45d5f74f2ec9e6fb7a3806aa74f3433760493 (frozen 2026-09-08). SHA-256 recomputation was blocked in the sandbox; length and exact residue string match the frozen FASTA.
• Artifacts saved during the run: /tmp/catalytic_residue_mapping.csv, /tmp/ttv_domain_map.png, /tmp/GO_decision_table.csv.

Report compiled from 3 completed investigation iterations, 3 confirmed sequence/homology findings, and 5 primary/structural literature sources. No direct enzymatic assay on the D5SHU8 isoform exists in the accessed literature; the glycosyltransferase assignment is homology-based and should be curated accordingly.

Artifacts

• OpenScientist final report
• OpenScientist final report