GRIN1 (human, Q05586) review notes

2026-11-10: first review

Deep research status

• just deep-research-falcon human GRIN1 --fallback perplexity-lite failed: neither the falcon nor the
  perplexity provider is configured in this environment ("Provider 'falcon' not available. Available:
  openscientist, claude_code").
• just deep-research-openscientist human GRIN1 was run as a substitute and finished after the verdicts
  were drafted (GRIN1-deep-research-openscientist.md). It was read as a cross-check. It agrees with the
  review on every point compared and changed no verdict. It supplied one extra citation for the glutamate
  binding removal PMID:12805203. Its quote attributed to PMID:9089692 is not in
  that paper's abstract, so it was not used; its other citations were not checked.
• No provider deep-research file was written by hand.

What the gene does

• Human NMDAR1 expressed in oocytes has the properties of the NMDA receptor PMID:7679115.
• Currents need glycine PMID:7685113.
• GluN1 carries the glycine site, GluN2 the glutamate site [UniProt Q05586 DOMAIN, "The ligand-binding
  domain (LBD) binds to glycine (GluN1 and GluN3 subunits) and glutamate (GluN2 subunits) and control
  opening of the channel gate"].
• Receptors are heterotetramers PMID:36309015.
• With GluN3, GluN1 forms glycine-gated channels [PMID:38598639 "Uniquely, NMDARs composed of GluN1 and
  GluN3 are activated exclusively by glycine"; PMID:11823786 "NR3A or NR3B co-assembles with NR1 to form
  excitatory glycine receptors that are unaffected by glutamate or NMDA"].
• GluN1 is needed for GluN2 ER exit and depends on GluN2 for synaptic accumulation [PMID:15317856
  "GluRepsilon/NR2 subunits strictly require the GluRzeta1/NR1 subunit to exit from endoplasmic reticulum
  (ER) to the cell surface in vitro and to the postsynapse in vivo"; "the GluRzeta1 subunit, by itself, is
  virtually unable to accumulate at postsynaptic sites, regardless of C-terminal forms"].
• CA1-specific deletion in mouse removes NMDA currents, LTP and spatial memory PMID:8980238.

Decisions worth a second look

1. glutamate binding (GO:0016595, IDA, PMID:7685113): REMOVE. The full text was read on 2026-11-10 from
   a PDF supplied by the user (the cache has the abstract only). All experiments are voltage-clamp recordings
   from oocytes injected with hNR1 alone; no binding assay. The glutamate dose-response of current is fitted
   to a binding isotherm (EC50 3.7 uM), currents are 4 to 60 nA, and the authors note that NR1 plus NR2 gives
   about 100-fold larger currents. PMID:12805203, the UniProt DOMAIN comment (evidence ECO:0000269) and the
   structural literature place glutamate on GluN2 and glycine on GluN1. So the row is removed because the
   function is contradicted and the paper holds no direct binding evidence. The glycine binding row from the
   same paper is also functional (glycine EC50 0.45 uM, block by 7-chlorokynurenate) but is kept, because
   glycine binding by GluN1 is confirmed structurally.
   An OpenScientist hypothesis run (2026-11-11, GRIN1-hypotheses/homomeric-glun1-oocyte-currents-need-xenopus-glun2/)
   confirmed the removal with direct evidence [PMID:16641235 "binding of glutamate to NR1 was not detectable
   even at a 300 mm concentration"; PMID:16214956 "In HEK293 cells, we did not get receptor response unless
   an NR2 subunit was coexpressed"]. It also refuted my first-draft explanation that the GluN1-alone oocyte
   currents come from an endogenous Xenopus GluN2 partner: the named candidate XenU1 was experimentally
   excluded PMID:11809857 and is a kainate-binding protein, not a GluN2. The reason text was corrected to
   drop that mechanism; the origin of those currents is left as unresolved.
2. propylene metabolic process (GO:0018964, ISS from mouse P35438): REMOVE, with a source defect for
   ARUK-UCL/BHF-UCL. The mouse IDA cites PMID:21795692 (Shank3 and mGluR5); the full text does not contain
   the word propylene.
3. NOT cellular response to amyloid-beta (IGI, PMID:21883149): MARK_AS_OVER_ANNOTATED. The negative
   result is real PMID:21883149 but
   comes from one oligomer preparation on recombinant receptors in oocytes.
4. amyloid-beta binding (ISS from rat): MARK_AS_OVER_ANNOTATED PMID:17308309.
5. calcium ion binding (ISS from mouse): UNDECIDED. Donor PMID:15663482 is abstract-only and shows a
   Ca2+/calmodulin dependence, not Ca2+ binding by NR1.

OpenScientist hypothesis runs (2026-11-11)

Two focused hypothesis jobs were run (just gene-hypothesis-research openscientist human GRIN1 ...),
in GRIN1-hypotheses/:

• glun1-glycine-site-inhibits-glun1-glun3-current: SUPPORTED. In GluN1/GluN3 receptors glycine at
  GluN3 opens the channel and glycine at GluN1 auto-inhibits; GluN1 glycine-site antagonists potentiate
  10-30x (PMID:17214961, PMID:30425244 native neurons, PMID:19726695). Confirms the correction already made
  to the module's negative GluN1->GluN1/GluN3 connection and the two GluN3 GO-CAM defect queries. Citations
  added to core function 3; no verdict changed.
• homomeric-glun1-oocyte-currents-need-xenopus-glun2: PARTIALLY SUPPORTED (see item 1 above). The
  glutamate-binding REMOVE is confirmed by direct evidence; the Xenopus-partner mechanism is refuted.

Key primary citations from both reports were verified against PubMed before use. The reports themselves are
cited as file: references with reference_review, and most of their other citations are unchecked.
6. "regulation of" terms changed to the process itself (GO:2000463 to GO:0060079, GO:1904062 to
GO:0098655, GO:0010524 to GO:0097553): the channel carries the current or flux, it does not regulate it.
7. General terms changed to specific descendants already on the gene (relationships checked in
QuickGO on 2026-11-10): ligand-gated/monoatomic ion channel activity, signaling receptor activity and
transmitter-gated channel activity to NMDA glutamate receptor activity and/or glycine-gated cation
channel activity; calcium channel activity to glutamate-gated calcium ion channel activity; membrane,
synapse, synaptic membrane to plasma/postsynaptic membrane; neuron projection to dendrite and dendritic
spine; chemical synaptic transmission to excitatory chemical synaptic transmission; ion and cation
transport to monoatomic cation transmembrane transport.
8. ER membrane (Reactome TAS x4): MARK_AS_OVER_ANNOTATED (changed from KEEP_AS_NON_CORE on 2026-11-10
after review by the user). GluN1 does not function in the ER; the rows describe assembly and transit.
No checked source shows endogenous GluN1 residing in neuronal ER, and unassembled GluN1 did not
accumulate there in GluN2-deficient cerebellum PMID:15317856. Not REMOVE,
because the location is true during biosynthesis and the Reactome statements are not wrong
(projects/ORGANELLES.md rule 7). A primary paper showing an endogenous ER pool would justify
KEEP_AS_NON_CORE.
9. synaptic vesicle (ISS from mouse): MARK_AS_OVER_ANNOTATED. Donor PMID:10846156 describes KIF17
transport vesicles in dendrites.
10. protein binding (4 IPI rows): REMOVE under the protein-binding policy. Partners GRIN2B and GRIN3A
are covered by the complex term; APP and FUS are yeast two-hybrid hits.

Things not done

• No NEW annotations were proposed. Candidates considered and left out: D-serine binding (not checked
  against a cached primary paper), long-term synaptic potentiation (covered by the accepted plasticity
  terms).
• PAINT losses: just paint-losses human GRIN1 reports no IRD or IKR on the lineage.
• The seven GO-CAM models containing GRIN1 (6690711d00000596, 66c7d41500000672, 66c7d41500001120,
  66c7d41500001953, 66c7d41500003371, 66c7d41500003458, 66e382fb00000049) were reviewed after the gene
  review; see gocams/<model>/<model>-review.yaml. Their GRIN1 activities match the core functions here.
  The two GluN3 models are queried on the sign of the GRIN1 edge and on calcium as a transported ion.
• The module modules/nmda_receptor_gating.yaml covers gating of GluN1-containing receptors. The gene
  review schema has no slot for a back-link to a module (module_associations exists only inside
  decomposed_into), so the link is recorded here.
• Row count: the GOA file has 185 lines and the review 181 rows; the four extra lines are exact duplicates
  that differ only by date, and just validate human GRIN1 passes its completeness check.
• Several early papers are cached as abstracts only (PMID:7679115, 7681588, 7685113, 9489750, 21376300),
  so their rows were accepted on the abstract plus the curator's reading.