Neuromuscular junction acetylcholine receptor clustering — neural agrin, LRP4 coreceptor, MuSK kinase activation by DOK7, CRK/CRKL adaptors and rapsyn anchoring; AGRN/LRP4/MUSK/DOK7/CRK/CRKL/RAPSN

At the vertebrate neuromuscular junction, nicotinic acetylcholine receptors (AChRs) are packed into the postsynaptic membrane of the skeletal muscle fiber at about 10,000 per square micrometre, directly opposite the motor nerve terminal. They get there in two ways: subsynaptic nuclei transcribe the receptor genes at a higher rate, and receptors already in the membrane are clustered by tethering to a subsynaptic scaffold. The clustering signal starts with neural agrin, the z+ splice forms of AGRN that motor neurons secrete into the synaptic basal lamina. Agrin binds LRP4, a single-pass LDL-receptor-family protein on the muscle surface. LRP4 also binds the muscle-specific receptor tyrosine kinase MuSK. The arc-shaped LRP4 ectodomain holds agrin and MuSK together in a ternary complex, and this activates MuSK. Inside the cell, the adaptor DOK7 binds the phosphorylated NPXY juxtamembrane motif of MuSK through its PTB domain. Its PH-PTB region dimerises, which drives trans-autophosphorylation of the MuSK activation loop, so DOK7 acts as an intracellular activator of the kinase. MuSK then phosphorylates two tyrosines in the DOK7 C-terminal tail, and these recruit the adaptors CRK and CRKL, which relay the signal towards the actin cytoskeleton. Rapsyn (RAPSN), a 43 kDa peripheral membrane protein on the cytoplasmic face of the postsynaptic membrane, binds the AChR and links it to the MuSK-containing scaffold. Without rapsyn, receptors do not form clusters at all. Before innervation, MuSK, LRP4, rapsyn and AChRs already form aneural clusters in the central region of the muscle (prepatterning); nerve-derived agrin then sharpens and stabilises these clusters at the synapse. Loss-of-function variants in AGRN, LRP4, MUSK, DOK7 or RAPSN cause congenital myasthenic syndromes. Autoantibodies against MuSK or LRP4 cause forms of myasthenia gravis.

Derivation. This module was split out of modules/acetylcholine_signaling.yaml, whose boundary decision (4) excludes the agrin-LRP4-MuSK-DOK7-rapsyn set. The reason is that those proteins build the synapse rather than receiving or transducing the ACh signal. That module names GO:0071340 and the mouse GO-CAM gomodel:64c3044100000089 as the grounding for a separate module. The members come from the RGD Pathway Ontology diagram PW:0000842, which draws Agrn/Lrp4/Musk/Dok7/Rapsn as a cluster acting on the NMJ nAChR. PW:0000842 is listed as an evidence source of this module as well, because this module is derived from that cluster of the diagram. PW:0000841 (the ACh overview) and PW:0000843 (muscarinic) are not cited, because they do not draw this machinery. Boundary decisions. (1) The muscle AChR is the cargo of this process, not a participant in it, and is therefore not a part. The CLAUDE.md test is whether a product does any of the work of the process. Rapsyn does: it binds the receptor and tethers it to the scaffold, and without it no clusters form (PMID:7675108, PMID:10320756). The AChR pentamer is the thing that is gathered. Comparator check (QuickGO, GO:0071340 with is_a/part_of descendants, human/mouse/rat, October 2026): the term and its regulation children are carried by Agrn, Lrp4, Musk, Rapsn, Fnta, Dnaja3, Colq, Dvl1, Etv5 and Rer1 (several as acts_upstream_of_or_within), and by none of Chrna1, Chrnb1, Chrnd, Chrne or Chrng. Human CHRNA1 instead carries channel, signalling and NMJ-location terms (GO:0022848, GO:0095500, GO:0007274, GO:0031594), and the gene reviews used for acetylcholine_signaling record those as its core. The term's parent GO:0043113 receptor clustering is a protein-localisation process (is_a GO:0072657 protein localization to membrane), and such terms name the machinery that moves or holds the protein, not the protein that is localised. The receptor therefore appears only as the target of the rapsyn adaptor function (acetylcholine-gated channel complex GO:0005892, with CHRNA1 as the representative subunit). Its channel function is modelled in acetylcholine_signaling#nachr_muscle and is not repeated here. A module_ref part was deliberately not used, because it would make the receptor a required step of the clustering process. (2) Agrin is modelled as the neural (z+) splice class of AGRN on the canonical accession O00468. The UniProt record states that isoforms 1, 4 and 5 are the neuron-specific z+ forms that activate AChR clustering, and that the z8 insert is needed for LRP4 binding. The z- muscle and endothelial forms lack this activity. The annoton uses the AGRN gene-review core function, transmembrane receptor protein tyrosine kinase activator activity GO:0030297. This is defensible because the agrin LRP4 MuSK cryo-EM structure shows that LRP4 brings agrin into direct contact with MuSK (PMID:37252960). The mouse GO-CAM uses receptor ligand activity GO:0048018 for Agrn; that term is a sibling under signaling receptor activator activity GO:0030546, not a conflict. Agrin's laminin and dystroglycan binding, which anchor it in the basal lamina, are recorded in the AGRN review and are not modelled here. (3) LRP4 coreceptor activity GO:0015026 follows the LRP4 review core function and the GO-CAM. LRP4 also has a second role as a retrograde signal: muscle LRP4 binds the motor axon and promotes presynaptic differentiation. That role is a presynaptic-assembly process and is out of scope, like the LRP4 sclerostin/Wnt role in bone. (4) MuSK activation is one node with two annotons. MUSK carries transmembrane receptor protein tyrosine kinase activity GO:0004714 (GOA IBA plus EXP for GO:0004713). DOK7 carries protein tyrosine kinase activator activity GO:0030296, the first core function in the DOK7 review, supported by the Dok7 PH-PTB dimer structure (PMID:20603078). (5) The DOK7 C-terminal tail to CRK/CRKL relay is a required part. DOK7's second review core function is transmembrane receptor protein tyrosine kinase adaptor activity GO:0005068, and CRK and CRKL carry signaling adaptor activity GO:0035591. CRK has this by IDA; CRKL has it by IBA, and the mouse GO-CAM uses it for both Crk and Crkl. Crk and Crk-L are redundant, so they are one family annoton, and only loss of both in muscle gives severe NMJ defects (PMID:21041412). Further downstream, the GO-CAM continues through FARP1 guanyl-nucleotide exchange factor activity, protein geranylgeranyltransferase I (FNTA), RAC1 and positive regulation of Rac protein signal transduction (GO:0035022) to postsynaptic actin organisation. That Rac/actin effector arm is not modelled, because RGD does not draw it and its members (FARP1, FNTA, RAC1, and the Abl/Src-family kinases that phosphorylate the AChR beta subunit) are not on the RGD gene list. It is a candidate for a later extension. (6) Candidates considered and not added. TID1/DNAJA3 binds MuSK and is required for agrin-induced clustering in mouse myotubes. Its human and rat GOA rows for GO:0071340 are ISS/ISO only, and no paper on it was cached and read for this module, so it is recorded here only. Neuregulin/ErbB-driven synapse-specific transcription, the Wnt coreceptor roles of MuSK and LRP4, ColQ/perlecan anchoring of AChE (which sits in acetylcholine_signaling), and NMJ formation in general (GO:0007528, GO:0051124) are out of scope. ACh reception itself is acetylcholine_signaling. (7) Rapsyn's molecular function is protein-macromolecule adaptor activity GO:0030674, as in the mouse GO-CAM. The only experimental MF row in GOA is acetylcholine receptor binding GO:0033130 (mouse IDA, transferred to human by IBA), which describes the binding but not the bridging role. GO-CAM. gomodel:64c3044100000089 "Skeletal muscle acetylcholine-gated channel clustering 1 (Mouse)" contains every annoton modelled here: Agrn, Lrp4, Musk (two activities), Dok7 (two activities), Crk, Crkl and Rapsn. Its activity links are recorded under gocam_associations. Possible upstream issue: in that model the Rapsn activity (GO:0030674, part_of GO:0071340) has input Chrna9 (MGI:MGI:1202403). The evidence cited for it is PMID:18420419, an inner-ear hair-cell study of rapsyn and the alpha9 receptor. In a skeletal-muscle clustering model the expected input would be the muscle receptor (Chrna1 and partners), so the has_input may have been carried over from the hair-cell paper. PMID:18420419 is not cached and its full text was not read, so this is flagged as a question rather than asserted as an error. No human GO-CAM covers this pathway. Gene reviews exist for AGRN, LRP4 and DOK7, and their core molecular functions were used. MUSK, RAPSN, CRK and CRKL have no review yet; their functions come from GOA and the cited literature. PANTHER family ids and PTNs are deliberately not asserted.