UniProt: O94905 (ERLN2_HUMAN), 339 aa. Synonyms SPFH2, C8orf2. HGNC:1356.
Single-pass type II ER membrane protein with a lumenal SPFH/prohibitin (band 7) domain. Band 7/mec-2
family; associates with lipid-raft-like ER domains. Disease: hereditary spastic paraplegia SPG18A/SPG18B
and recessive intellectual disability/joint-contractures syndrome.
All four cited papers below were PMID-verified via PubMed (DOI-to-PMID); none are in the
publications/ cache, so no verbatim supporting_text was added to the review — references
added with reference_review only, plus the Veronese ref id on the E3-ligase/TMUB1 core_function.
All 59 current imported contexts have reviewed decisions: {'ACCEPT': 21, 'MODIFY': 12, 'KEEP_AS_NON_CORE': 26}. Forty-six old contexts match one-to-one. Four historical binding contexts lack explicit partner identities and remain as replaced historical records while all 12 current partner-specific contexts are appended independently. The additional ERAD IBA is also appended. The absent historical cholesterol-binding IBA remains separately preserved: its retired status records the current source boundary, while its biological assessment is UNDECIDED. Current PAINT places cholesterol binding on an ERLIN1 clade outside the ERLIN2 ancestry; this is not a claim of biological loss. The record contains 64 rows, 59 current and five historical. All 57 inherited quotation paths, texts and reference associations remain fixed. No new quotation or NEW annotation is introduced. Three source-described natural products (339, 152 and 206 residues) remain distinct. The ERLIN2–FGFR1 fusion context is retained in its 12 Reactome location records without assigning fusion kinase function to natural ERLIN2. The inherited description, three cores and historical findings are unchanged here and await separate synthesis.
The three ordered summaries now distinguish the integral erlin scaffold, sterol-pathway restraint, and measured ubiquitin-ligase associations. The 64 annotation/source/review objects, three product records and all 57 inherited quotation paths and associated references remain unchanged. No annotation or publication quotation was added. PMID:19751772 and PMID:21610068 were appended using genuine normal records; the preceding 47 reference identities and order remain intact.
The scaffold assignment describes an integral structural component that supports erlin assembly and partner organization. The ERAD quality-control process has separate positive support from misfolded luminal A1AT-NHK and TTR-D18G turnover after combined erlin depletion in human HEK/HeLa cells (PMID:19751772, model-substrate Methods and Figure 6). This modest, selective effect is not a claim that ERLIN2 is universally required for ERAD. Activated IP3R turnover is a regulatory substrate context, not by itself evidence that the receptor is misfolded. Human mHeLa reciprocal-depletion experiments place ERLIN2-dependent complex recruitment before RNF170 recruitment (PMID:21610068, Figure 5C/D and Priority of binding to activated IP3 receptors); the cloned RNF170 reagent is mouse, and separate rodent experiments retain their species boundaries.
The sterol unit retains negative SREBP regulation and regulation of cholesterol biosynthesis without assigning cholesterol-binding molecular function to ERLIN2. The available binding assays do not resolve ERLIN2-specific binding to native cholesterol: PMID:24217618 uses purified human ERLIN1 and a native mouse mixed complex, while the ERLIN2 probe evidence in PMID:23396283 has the recorded ligand-specificity limits. This is a target/ligand-attribution limit, not a rule that GO binding annotations require purified proteins. SREBP processing, transcripts and lipid pools do not themselves measure isotope-resolved biosynthetic flux. Positive gp78-associated HMGCR results in PMID:21343306 coexist with the negative mHeLa/Rat1 conditions in PMID:19751772, Figure 7.
The ligase-association summary does not duplicate an ERAD process or imply that every association recruits a ligase to every cargo. TMUB1 bridging constrains direct-binary interpretations. PMID:24019521, Figure 4A, recovers endogenous human ERLIN2 with RNF185 and RNF5, whereas the principal CFTR functional perturbations concern the ligases.
Modern findings supersede stronger earlier authored interpretations in these notes. In PMID:38782601, erlin double knockout increases cholesterol esterification and lipid storage; re-expression and SOAT1 inhibition support the cellular phenotype. Adjacent-subunit interfaces are AlphaFold predictions, and simultaneous TMUB1/RNF170 occupancy is unproven. In PMID:38607533, comparable gross high-molecular-weight assemblies in one heterozygous V168M fibroblast line do not exclude cargo, partner-selection or all dominant-negative defects. In PMID:40225166, patient-cell IP3R1 abundance, calcium measurements and RNF213-associated co-immunoprecipitation support a proposed mechanism, not demonstrated RNF213-dependent ubiquitination or degradation. Isogenic correction, RNF213 rescue and substrate half-life tests remain useful experiments. The questions no longer assume an experimentally resolved human structure, a proven ERLIN2 cholesterol-binding site, or unverified variant examples.
Inherited quotations are retained as historical evidence with their original references. The core 0 PMID:19240031 snippets support complex assembly and activated-receptor turnover; the new quality-control support is the separate unquoted PMID:19751772 citation. The core 1 cholesterol-binding snippet refers to the reagent/complex scope above, not resolved ERLIN2-specific binding. The core 2 RNF185/RNF5 snippet concerns the ligase module; the actual ERLIN2 association is supported by Figure 4A. These quotations were not reassigned to different claims or references.
Current GO:0036503 is named “ERAD quality control pathway”; the inherited label “ERAD pathway” is retained for compatibility with the frozen ontology used by validation and is a current broad synonym. Its current definition governs the biological interpretation. GO:0140378 is an unrestricted molecular-function term for an integral complex scaffold; the ER complex term is placed in in_complex, with ER membrane retained as the location. The two sterol processes are not ancestors or descendants of one another in the inspected current ontology. The three product records and the shorter-product boundaries remain unchanged; no canonical-product function was transferred to the 152- or 206-amino-acid products.
Whole-candidate and component schema checks, and strict typed-term/label checks for all three core objects, passed in the private frozen runtime. Final gene-wide status, best-practice validation and rendering remain separate publication checks.
The structural core now records GO:0005198 structural molecule activity for ERLIN2's integral contribution to oligomer integrity, and contributes_to_molecular_function: GO:0140378 for the assembly's scaffold activity. Physical assembly evidence (PMID:19240031) and human complex-dependent partner association (PMID:38782601) support this synthesis without independently resolved ERLIN2 bridging or equal/interchangeable subunits. The two slots express distinct attribution levels within one core. No NEW scaffold annotation is proposed because the synthesis is bounded to structural participation in the complex; an independently enabled ERLIN2 scaffold activity is not established. All 64 annotation/review objects, three products, 49 reference identities and 57 inherited quotation paths remain exact. No new original-access or figure-inspection claim is made by this refinement.
Final canonical validation records DRAFT with 17 advisories. All 59 current source contexts, five retained historical contexts, three product records and 57 inherited quotation paths were checked; all 139 authored narratives appear in the private HTML rendering. The structural-molecule core and complex scaffold contribution remain bounded synthesis without a duplicate NEW assertion.
The twelve Reactome plasma-membrane contexts are removed as native ERLIN2 annotations. Their actual participant chains terminate in valid ERLIN2–FGFR1 disease-fusion entities; the correction concerns the native-product GOA projection, not deletion of those events or changing the legitimate O94905 reference segment. Structured, unreported source defects record that projection and the separate EML4 display-name inconsistency. Independent HPA U2OS membrane/cytosol observations remain positive evidence with antibody/product limits, and are not converted into a new annotation. This supersedes the earlier contextual-retention decision for those twelve sources.
The campaign curation instructions explicitly retain supported generic binding as non-core when no evidence-backed specific replacement is established, and preserve inherited quotations while adding no new publication quotations. Both standing user instructions govern this revision. Thus the twelve supported generic-binding contexts remain non-core; specific ligase-binding refinements remain. The inherited ER-membrane, erlin-complex and CC INTERACTION quotations are unchanged at their original paths and references, but their limited scope is made explicit: they do not prove fusion localization, alpha1D-AR signalosome membership or independent replication of an IntAct interaction. Original assays and datasets provide the corresponding positive evidence. Record-section metadata is derived by the repository backfill tool, not manually assigned.
RNF185 and RNF5 binding is accepted within the existing narrow ligase-association core because Figure 4A directly recovers endogenous ERLIN2 with each ligase in human cells. This is not an inference that ERLIN2 performs or controls the ligases' CFTR reaction. The ARBA ER-complex refinement and the distinct alpha1D-AR affinity-complex context retain different actions because they describe different assemblies. No core terms, new annotations or natural-product equivalence were added.
The published canonical validation recorded DRAFT with 17 advisories: twelve generic-binding policy warnings, one historical cholesterol IBA propagation-metadata warning, two context-dependent action-consistency warnings, one unused provider reference warning and one structural-core term not separately proposed as NEW. These are the prior measured results, not a forecast of the revised validation. DRAFT remains until recomputed by final validation; a warning does not justify an unsupported new assertion. The GO:0036503 imported label and authored core label retain their source/ontology-version boundary.
Revised canonical validation passes with DRAFT status and 16 advisories. All 139 authored narratives are present in the private rendering. Current decisions are 23 ACCEPT, 12 MODIFY, 12 KEEP_AS_NON_CORE and 12 REMOVE across 59 contexts; five historical contexts give 64 total rows. The 57 inherited quotation paths/text/reference associations, three products, three ordered cores and 62 selected normal source objects remain exact.
The remaining 16 advisories are twelve supported generic-binding policy exceptions, one historical IBA record without resolved propagation metadata, one different-assembly action-consistency warning, one available deep-research file not cited by annotations, and one structural core term without a duplicate NEW assertion. The previous ligase-binding action split is resolved. The backfill tool’s 32 derived record-section values required quoted YAML scalars for compatibility with the reference validator; that serialization correction changes no parsed values, source records or quotation content.