RFT1 (Q96AA3) — Cavity-vs-Portal Structural Hypothesis: Curation Report OpenScientist openscientist-autonomous 6 artifacts 2026-08-24T15:47:28.714624

OpenScientist final report (text/html)

OpenScientist final report (application/pdf)

OpenScientist rft1 cavity portal (application/json)

OpenScientist rft1 cavity portal (image/png)

OpenScientist rft1 evidence matrix (text/csv)

OpenScientist rft1 residue classification (text/csv)

RFT1 (Q96AA3) — Cavity-vs-Portal Structural Hypothesis: Curation Report

Gene: RFT1 (human, Homo sapiens, NCBITaxon:9606) · UniProt: Q96AA3
Hypothesis slug: scramblase-vs-binding-cavity · Term in context: GO:0140303 intramembrane lipid transporter activity

Seed hypothesis. Human RFT1 is an alternating-access MOP-superfamily transporter, and the three RFT1-CDG residues (p.R67C, p.K152E, p.E298K) line the central substrate-binding cavity that coordinates the anionic pyrophosphate headgroup of Man5GlcNAc2-PP-dolichol (M5-DLO), not the lateral membrane portal for the dolichol tail.

1. Executive Judgment

Verdict: PARTIALLY SUPPORTED (with a discriminated correction), and the two mechanistic models cannot be separated by structure alone.

Fold assignment (part 1): supported qualitatively. The high-confidence AlphaFold model of human RFT1 (AF-Q96AA3-F1 v6; mean pLDDT 90.4; 93 % of residues >70) is a polytopic ~12–14-TM, two-lobed helical bundle enclosing a discrete central pore axis that is geometrically separable from the lipid-facing periphery. A detectable — though modest — internal inverted-topology repeat (40 Cα pairs at 3.07 Å RMSD after a 180° rotation about the membrane normal) is consistent with the MOP/MATE/MurJ two-domain architecture independently modelled for Rft1 (PMID 42417535). This is inference from a computational model, not an experimental structure.
Residue mapping (part 2): partially supported. R67 is convincingly central-cavity-lining (buried, near the pore axis, side chain opens inward; an invariant Arg across all orthologs). K152 is a plausible inner-vestibule/cavity residue (inward-facing, conserved basic K/R). E298 refutes the seed claim for that residue: it is solvent-exposed, outward-facing, and a conserved acidic residue — inconsistent with lining a cationic anion-binding cavity, and it is not in the lipid portal either. No disease residue lines the lateral dolichol portal.
Model A vs Model B: unresolved by structure. Because all three disease residues implicate substrate binding/recognition (headgroup region) rather than the dolichol-tail translocation portal, the structural evidence is fully compatible with both Model A (RFT1 is the translocase; direct support from reconstitution, PMID 38886340) and Model B (RFT1 binds/routes; leaning of PMID 42417535). Headgroup binding is required for transport under either model, so cavity-localised disease mutations do not discriminate them.

Most important caveats: apo human model without docked M5-DLO; the "cationic central cavity" is evident in substrate-docked yeast models but the innermost lining of the apo human model is not net-cationic (net −2); membrane axis and cavity membership are geometric estimates and borderline residues (K152, E298) are axis-sensitive.

2. Evidence Matrix

Citation	Type	Stance	Claim tested	Key finding	Context	Confidence / limits
PMID 42417535 (Chiduza & Menon 2026)	structural/computational + mutant phenotype	qualifies / competing	RFT1 is a MOP alternating-access transporter with cationic central cavity + lateral dolichol portal	AF3/Chai-1 yeast Rft1–M5-DLO models show alternating access; cationic cavity binds anionic headgroup, dolichol tail exits a lateral portal; 2/26 cavity mutants grew poorly; portal-blocking mutant predicted to lack scramblase grew robustly	yeast Rft1; Tet-off reporter; in silico + in vivo	High for architecture; suggests scrambling may be a moonlighting function (leans Model B); yeast, not human
PMID 41427416 (Chiduza et al. 2025)	structural/computational	supports (architecture)	Mechanism of Rft1-mediated M5-DLO scrambling	Cavity/portal scrambling mechanism for the anionic glycolipid	yeast Rft1 model	model-level; mechanism proposed, not proven
PMID 38886340 (Chen et al. 2024)	direct assay (reconstitution)	supports Model A	Purified Rft1 is itself the M5-DLO translocase (GO:0140303)	Fully reconstituted assay: purified Rft1 catalyses transbilayer translocation of M5GN2-PP-Dol with substrate selectivity	yeast Rft1; proteoliposomes, in vitro	Strong direct MF evidence; in vitro only; does not localise disease residues
PMID 19701946 (Vleugels et al. 2009)	mutant phenotype	supports (residue identity)	R67C/K152E/E298K cause RFT1-CDG	Three unrelated patients homozygous for R67C, K152E, E298K; M5GlcNAc2-PP-Dol accumulates; rescued by WT RFT1	human fibroblasts	Establishes disease residues; loss-of-function, not translocation mechanism
PMID 19267216 (Clayton & Grünewald 2009)	clinical / mutant phenotype	orientation	RFT1-deficiency phenotype (CDG-In)	First patient: severe multisystem CDG; DolPP-GlcNAc2Man5 accumulation	human patient	Clinical; flippase role stated as hypothesis
This analysis	computational (structure geometry)	partially supports / qualifies	CDG residues line central headgroup cavity, not lipid portal	R67 buried, near-axis, inward (cavity, high conf); K152 inner vestibule (moderate); E298 exposed, outward-facing (neither)	AF-Q96AA3-F1 v6 human apo model; Shrake-Rupley SASA + pore-axis geometry	Apo model; geometric axis; borderline membership axis-sensitive
This analysis	structural/evolutionary	supports R67/K152; refutes E298-cavity	Conservation of cavity charge at disease positions	R67 invariant Arg 9/9; K152 conserved basic (K/R) 7/9; E298 conserved acidic (E/D) 9/9	NW alignment, human vs 9 orthologs (human→fungi→Dictyostelium)	Small ortholog set; pairwise NW, not full MSA
This analysis	structural/evolutionary	qualifies	MOP/MATE inverted-topology fold present	~12–14 TM two-lobed bundle; internal C2 repeat 40 Cα @ 3.07 Å; apo axis lining net −2	AF-Q96AA3-F1 v6 human	Repeat modest (17 % of N-half); no experimental structure

Per-residue provenance (computed) — see `rft1_residue_classification.csv`

Residue	rSASA (burial)	radial from pore axis (Å)	side-chain exposure-radial-out	z vs membrane centre (Å)	conservation (9 orthologs)	Classification
R67	0.13 (buried)	9.1 (central)	−0.21 (opens inward)	−14.9	Arg invariant 9/9	Central-cavity-lining (high)
K152	0.26	12.6	−0.09 (inward/neutral)	−10.8	basic K/R 7/9	Cavity / inner vestibule (moderate)
E298	0.46 (exposed)	13.8	+0.44 (opens outward)	+4.9	acidic E/D 9/9	Neither cavity nor portal (refutes)

Reference clouds: axis-facing TM residues median exposure-radial-out = −0.07; lipid/portal-facing = +0.75. None of the three disease residues sits in the lipid portal zone (radial >18 Å, exposure ≈ +0.75). Figure: rft1_cavity_portal.png.

3. GO Curation Implications (leads — require curator verification)

GO:0140303 intramembrane lipid transporter activity (MF): The structural analysis is consistent with a transporter fold and a discrete substrate-binding cavity, but is computational and cannot establish translocase activity on its own. The primary direct support for this MF term is the reconstitution assay in PMID 38886340 (purified Rft1 translocates M5-DLO) — recommend retain GO:0140303 with experimental (IDA) evidence anchored to PMID 38886340, noting the assay is in vitro and in yeast. The structural/evolutionary evidence here would be ISS/IEA-level support only.
Do not upgrade the residue-level mechanism as fully settled. If the review cites the cavity-vs-portal mechanism, record it as partially supported: R67 (and K152) support the "cavity coordinates headgroup" claim; E298 does not and should not be described as headgroup-cavity-lining.
BP: dolichol-linked oligosaccharide biosynthetic process / protein N-linked glycosylation — supported by disease genetics (PMID 19701946); CC: endoplasmic reticulum membrane (GO:0005789) — supported.
Avoid "protein binding" as a terminal annotation — more informative MF (GO:0140303) is available.
Paralog/isoform caution: a divergent mammalian ~450-aa set (e.g., human Q9NWF4, mouse Q9D8F3; 26 % identity to canonical, no aligned residue at position 67) should not receive transitive annotation from the 541-aa canonical RFT1 without review.

4. Mechanistic Scope

Direct molecular function under test: transbilayer movement / binding of the anionic glycolipid M5-DLO at the ER membrane, and the structural sub-question of where the disease residues act (headgroup cavity vs dolichol portal). Directly addressed: residue localisation on the fold (cavity for R67/K152; exposed for E298). Downstream / not direct: M5-DLO accumulation, hypoglycosylation, multisystem CDG phenotype and lethality are loss-of-function consequences, not evidence about the translocation step itself. Whether RFT1 performs the translocation (Model A) or only binds/routes it (Model B) is a mechanistic question the structure cannot resolve.

5. Conflicts and Alternatives

Direct conflict in the literature: PMID 38886340 (purified Rft1 is the flippase → Model A) vs PMID 42417535 (portal-blocking, scramblase-dead mutant still supports growth; scrambling may be moonlighting → Model B). Both are credible; the debate is >20 years old and remains open.
Apo vs holo: the "cationic central cavity" is a property of substrate-docked yeast models; the apo human model's innermost lining is net-acidic (−2). The cationic pocket may be substrate-induced or dependent on exact axis choice.
Species/numbering: the published cavity/portal residues are yeast; human numbering (R67/K152/E298) was mapped here directly on the human model, avoiding cross-species residue transfer errors.
Isoform/paralog carry-over: the ~450-aa mammalian cluster is a genuine annotation-transfer risk.
E298 reinterpretation: a conserved, exposed, acidic residue is more consistent with a surface/interface or conformational role (e.g., a salt bridge or a gating/interaction determinant) than with headgroup coordination.

6. Knowledge Gaps

No experimental RFT1 structure (checked AlphaFold DB + PubMed). Matters because fold/portal assignment rests on models. Resolve: cryo-EM of human/yeast RFT1 ± M5-DLO.
Apo vs substrate-bound cavity electrostatics — checked apo geometry only (net −2). A docked M5-DLO complex (AF3/experimental) is needed to confirm a cationic headgroup pocket and to test whether R67/K152 directly contact the pyrophosphate.
E298's actual role — not defined here beyond "exposed, conserved acidic." Needs contact/interface analysis (partner protein? intramolecular salt bridge?).
Portal helix identity in human RFT1 — not explicitly assigned; needed to confirm no disease residue is portal-lining across conformational states.
Formal conservation used a 9-ortholog pairwise NW, not a deep MSA/ConSurf; a full MSA would sharpen per-position scores.
Model A vs B — the decisive gap; structure cannot resolve it.

7. Discriminating Tests

Cryo-EM of RFT1 ± M5-DLO in inward- and outward-open states → confirm alternating access, portal helices, and whether R67/K152 coordinate the pyrophosphate.
Charge-reversal/neutralisation at R67 and K152 (e.g., R67E, K152E already disease) with reconstituted transport (per PMID 38886340) → test headgroup coordination directly in vitro.
Portal-blocking salt-bridge mutant in human RFT1 with in-vitro translocation + in-vivo rescue (mirroring PMID 42417535) → the sharpest Model A vs B discriminator: if a scramblase-dead portal mutant still rescues, favours Model B.
E298 interface mapping (crosslinking/AP-MS, or E298K vs E298Q in transport vs stability assays) → define its non-cavity role.
Deep MSA/ConSurf across the full RFT1 family → rigorous per-residue conservation.

8. Curation Leads (require curator verification)

Retain GO:0140303 (MF) anchored to PMID 38886340 (direct reconstitution; IDA, in vitro, yeast ortholog). Flag the ongoing Model A/B debate (PMID 42417535).
Candidate references to verify (exact snippets):
PMID 42417535: "an alternating access mechanism, typical of Multidrug/Oligosaccharidyl-lipid/Polysaccharide (MOP) superfamily transporters, in which a cationic central cavity coordinates the anionic headgroup of M5-DLO, while the dolichol tail of the lipid is accommodated through a lateral portal formed by two transmembrane helices."
PMID 38886340: "purified Rft1 catalyzes the translocation of M5GN2-PP-Dol across the lipid bilayer … confirm the molecular identity of Rft1 as the M5GN2-PP-Dol ER flippase."
PMID 19701946: "homozygous for the missense mutation c.454A>G (p.K152E) and c.892G>A (p.E298 K)" (and R67C).
Record the residue mechanism as PARTIALLY supported: R67/K152 = headgroup-cavity residues; E298 is not a cavity residue (mark the seed's E298 assignment as not supported by structure/conservation).
Add an isoform/paralog note to prevent annotation transfer to the divergent ~450-aa entries (human Q9NWF4, mouse Q9D8F3).
Suggested curator question: does the review intend GO:0140303 to assert RFT1 as the translocase (Model A) or as a substrate-binding component (Model B)? The evidence supports the MF term but not, on its own, the stronger translocase claim.

9. Provenance / Artifacts

rft1_cavity_portal.png — computed cavity-vs-portal orientation and burial of R67/K152/E298 among RFT1 TM residues (iteration 1).
rft1_residue_classification.csv — per-residue computed metrics and classification.
rft1_evidence_matrix.csv — machine-readable evidence matrix.
All structural metrics computed from AlphaFold DB model AF-Q96AA3-F1 v6; SASA via Shrake–Rupley (192 points, 1.4 Å probe); pore axis from SVD of the buried TM core; conservation via Needleman–Wunsch (BLOSUM62) vs 9 orthologs. No values were fabricated; where a resource was unavailable (experimental structure; full MSA) it is stated.