Proteolytic Processing of the Extracellular Scaffolding Protein LEV-9 Is Required for Clustering Acetylcholine Receptors*

Background: LEV-9 is an extracellular scaffolding protein required for synaptic clustering of acetylcholine receptors in C. elegans. Results: C-terminal cleavage of LEV-9 is necessary for its synaptic function. Conclusion: LEV-9 is activated by proteolytic cleavage following the final tandem complement control protein (CCP) modules. Significance: Many LEV-9-related factors are likely cleaved after CCP domains, suggesting that cleavage location can be conserved in distant architecturally related proteins. Correct positioning of neurotransmitter-gated receptors at postsynapses is essential for synaptic transmission. At Caenorhabditis elegans neuromuscular junctions, clustering of levamisole-sensitive acetylcholine receptors (L-AChRs) requires the muscle-secreted scaffolding protein LEV-9, a multidomain factor containing complement control protein (CCP) modules. Here we show that LEV-9 needs to be cleaved at its C terminus to exert its function. LEV-9 cleavage is not required for trafficking nor secretion but directly controls scaffolding activity. The cleavage site is evolutionarily conserved, and post-translational cleavage ensures the structural and functional decoupling between different isoforms encoded by the lev-9 gene. Data mining indicates that most human CCP-containing factors are likely cleaved C-terminally from CCP tandems, suggesting that not only domain architectures but also cleavage location can be conserved in distant architecturally related proteins.

Scaffold-mediated receptor clustering at postsynapses is a key element of synaptic transmission. The activity of scaffold proteins can be regulated by several post-translational mechanisms including phosphorylation and degradation (1,2), but also by limited proteolysis. For example, the rapid insertion of excitatory glutamate-gated AMPA receptors is regulated by cleaving the scaffolding protein GRIP (glutamate receptor-interacting protein) by Ca 2ϩ -dependent calpains (3). Calpains are also thought to regulate glutamategated NMDA receptor levels by acting upon the canonical cytoplasmic scaffolding protein PSD-95 (4). Extracellular proteolysis of scaffolding factors can also regulate synaptic receptor quantities. For instance, cleavage of the neuronal pentraxin receptor NRP by the extracellular metalloprotease TNF-converting enzyme (TACE) accelerates AMPA receptor endocytosis, resulting in a decreased content of AMPA receptors at synapses, a process necessary for mGluR1/5-dependent long-term depression in hippocampal and cerebellar synapses (5).
At excitatory neuromuscular junctions (NMJs) 3 in Caenorhabditis elegans, an extracellular scaffold containing the LEV-9 and LEV-10 proteins is required for L-AChR clustering (6,7). LEV-9 is a muscle-secreted factor composed by a WAP domain (whey acidic protein) followed by a tandem of eight consecutive CCP modules (also known as short consensus repeats or SUSHI repeats) (see Fig. 1A). It is architecturally related to several plasma proteins in vertebrates, including regulators of the complement system (8). LEV-10 is composed chiefly by CUB ectodomains (for complement C1r/C1s, Uegf, Bmp1) anchored to the muscle membrane via a transmembrane domain. LEV-10 belongs to a superfamily of ancillary proteins that regulate glutamate-gated receptors in vertebrates and invertebrates (8). Interaction between LEV-9, LEV-10, and L-AChRs is stabilized by OIG-4, a small secreted protein made of one immunoglobulin domain (9). In the absence of any of its components, including the L-AChRs themselves, the LEV-9-LEV-10 complex disassembles entirely. In lev-9 or lev-10 mutants, L-AChRs are declustered, but the total amount of functional receptors in the muscle membrane remains unchanged. However, these mutants survive chronic exposure to concentrations of the L-AChR agonist levamisole that are lethal to wild-type animals (6,7,9). Therefore, the L-AChR-associated complex is an attractive model for synaptic scaffolding by virtue of its domain architecture, constitutive activity, and involvement in downstream regulation of synaptic transmission.
In the present work, we show that LEV-9 is cleaved both in a heterologous expression system and in vivo. The cleaving site maps to a highly conserved set of residues present at the C-terminal end of LEV-9 that are predicted to be recognized by proprotein convertases. LEV-9 cleavage is required for its clustering activity but not for secretion nor traffic. We also show that the lev-9 locus encodes multiple isoforms, including a long polypeptide containing 17 CCP domains that is found in other species. We suggest that the post-translational cleavage of LEV-9 might serve to ensure its functional and structural decoupling from any C-terminal polypeptide attachments, as we provide some evidence suggesting that other CCP-containing proteins found in humans might undergo similar post-translational processing.
Cell Culture-S2 Drosophila cells were maintained following the supplier recommendations (Invitrogen, catalog number R690-07) as described previously (7) and transiently transfected using calcium phosphate. Recombinant protein expression was driven by a metallothionein promoter and induced by copper sulfate.
Fluorescent Microscopy and Quantification-Young adult worms were immobilized on agarose pads containing M9 medium/ sodium azide. Images were acquired using a Leica DM5000B microscope, a CSU10 (Yokogawa) spinning disk head, a CoolSNAP HQ2 (Photometrics) CCD camera, and the software MetaMorph (v7.6.1.0, Molecular Devices). A cross-profile was obtained from z-projection of the most anterior 30 m of the dorsal nerve cord using ImageJ. A background line was interpolated using the "tail" values, and the corresponding background values were subtracted from each point of the curve. The total intensity was obtained by summing the latter normalized values.
Bioinformatics-LEV-9 orthologues were identified with PSI-Blast, using C. elegans and Caenorhabditis briggsae orthologues as probes. The National Center for Biotechnology Information (NCBI) (ncbi.nlm.nih.gov), Joint Genome Institute (JGI), ENSEMBL, and Wellcome Trust Sanger Institute databases were analyzed. A subset of sequences was selected considering phylogenic distribution; alternative splice variants and redundant sequences were removed. Multiple sequence alignment was done using M-Coffee (10) and curated using BioEdit (v7.1.3, Ibis Biosciences). The Eukaryotic Linear Motif (ELM) resource (11) and ProP 1.0 Server (12) were used to identify protease sites. European Molecular Biology Laboratory (EMBL) SMART (13) was used to identify protein domains. WebLogo was used to create sequence logos. Secondary structure and disorder regions were predicted with Jpred3 (14) and PrDOS (15), respectively.
Behavioral and Pharmacological Assays-Chronic (Ͼ16 h) adaptation and acute response tests to levamisole (Sigma-Aldrich, catalog number L9756) were done as described previously (6). To analyze thrashing defects, single worms were transferred to wells of 48-well plates containing M9 medium and incubated there for 1 min; head swings were counted three times per animal for 15 s. Results were reported as swings per second.
Statistics-Levamisole adaptation tests were analyzed using a nonparametric Kolmogorov-Smirnov test. Receptor clustering quantification, which fitted a normal distribution (p Ͻ 0.05, normal distribution Kolmogorov-Smirnov test), was analyzed using an unpaired non-equal variance twosample t test.

RESULTS AND DISCUSSION
The C-terminal Region of LEV-9 Is Required for AChR Clustering-LEV-9 orthologs are readily identified in ecdysozoans including nematodes and insects. Amino acid conservation is high in the region coding the WAP and the eight CCP domains but suddenly drops 33 residues after the last CCP domain. No protein domains are predicted to be fully contained within this 66-residue C-terminal (Ct) region of LEV-9. However, the Eukaryotic Linear Motif (ELM) on-line resource (11) predicted two adjacent recognition sites for convertases: one motif for a dibasic nardilysin convertase (residues Asp-571/ Arg-572/Lys-573) and another motif for subtilisin/kexin2p-like proprotein convertases (PC) (residues Lys-574/Arg-575/Ser-576) (Fig. 1A). A predicted PC cleavage signal at the same site was confirmed by the PC-specific prediction software ProP (score 0.819; false-negative prediction Ͻ0.9% (12)) (Fig. 1A). Nardilysins (NdK) are zinc proteases of the M16 peptidase family that have been recently involved in axon maturation and  myelination (16). PCs are related to subtilisin and yeast kexin2p protease families that play a major role in the proteolytic processing of both neuropeptides and peptide hormone precursors (17,18). Both motifs were found present in most, if not all, identified LEV-9 orthologs (Fig. 1, B and C).
To test whether the Ct region was required for LEV-9 function, we expressed in muscle a LEV-9 version truncated after the eighth CCP domain and tagged with GFP. Like LEV-9-GFP, L9⌬Ct-GFP was detected in coelomocytes, six scavenger cells that filter the fluid of the C. elegans pseudocoelomic cavity, arguing for correct folding and secretion into the extracellular medium (Fig. 1D). L-AChR clustering was assessed and quantified using a knock-in strain engineered by homologous recombination in which the subunit UNC-29 is tagged with tagRFPt (19) (Fig. 1E). As opposed to full-length LEV-9, L9⌬Ct-GFP did not rescue L-AChR clustering defects of lev-9 null mutant. Surprisingly, worms expressing L9⌬Ct-GFP showed a wild-type like sensitivity to 1 mM levamisole (p Ͼ 0.5, Kolmogorov-Smirnov test) despite considerable and significant (p Ͻ 0.001, t test) lack of L-AChRs at NMJs when compared with the wild type (Fig. 1, E and F).
These data indicated for the first time that it was possible to dissociate the levamisole effect, which kills the worms and requires the N-terminal part of the LEV-9 protein, and the clustering of the levamisole receptor, which requires the C-terminal region of LEV-9. There are two main models to account for this result. First, the N-terminal part of LEV-9 might couple L-AChRs to an actuator such as a channel or a kinase that would amplify L-AChR signaling and make levamisole lethal. Second, the N-terminal part of LEV-9 might stabilize L-AChRs in the plasma membrane independently from synaptic determinants. In fact, prolonged exposure to levamisole causes partial removal of L-AChRs from the muscle cell surface. 4 The absence of LEV-9 in lev-9 mutants might enhance L-AChR removal, thus decreasing levamisole toxicity. In this model, L9⌬Ct would be able to stabilize the levamisole receptors in the membrane and rescue levamisole sensitivity independently from L-AChR synaptic clustering.
A LEV-9 Predicted Cleavage Site Is Biochemically Active-To test whether the cleavage sites predicted in the Ct region are functional, we used S2 cells to express a LEV-9 construct tagged with a T7 epitope at its N terminus and with a MYC tag at its C terminus (T7-L9-MYC). The S2 (Schneider-2) cell line is derived from Drosophila embryonic cells and was successfully used in the past to express C. elegans proteins (7,9). In cell extracts, T7-L9-MYC was detected by Western blot using anti-T7 or anti-MYC antibodies, whereas it was only detected by anti-T7 antibodies in cell medium, suggesting that the C-ter-minal MYC tag was cleaved off the secreted protein (Fig. 2, A  and B). Similar results were obtained after removing the CCP-7 or the CCP-8 domain (T7-⌬CCP7-MYC and T7-⌬CCP8-MYC, respectively, in Fig. 2, A and B). However, the MYC tag remained detected in the medium of cells expressing a tagged version with a deletion of the Ct region, suggesting that the cleavage site was contained within this region.
To identify which of the predicted cleavage sites might be processed, we synthesized constructs lacking either the nardilysin (T7-⌬NdK-MYC) or the proprotein convertase (T7-⌬PC-MYC) sites (Fig. 2, C-E). In both ⌬NdK and ⌬PC, intracellular and secreted LEV-9 molecules could be detected through their N-terminal T7 epitope, but the apparent size of wild-type and ⌬NdK variants decreased after secretion (Fig. 2D). No size shift could be found with ⌬PC constructs (Fig. 2D). When the samples were run in a high percentage acrylamide gel (15%) and probed with the C-terminal MYC epitope, small proteolytic fragments could be found in the extracellular medium of wildtype and ⌬NdK transfected cells, but not in ⌬PC (Fig. 2E). The sizes of these fragments corresponded to the size expected if the constructs were cut in the Ct region. These results showed that in S2 cells LEV-9 was proteolytically processed at the predicted PC site, but processing of LEV-9 was not required for its secretion.
To verify that a similar post-translational maturation was also occurring in C. elegans, we expressed GFP-tagged LEV-9 with wild-type sequence or deletion of the protease predicted sites (L9⌬NdK-GFP and L9⌬PC-GFP in Fig. 2F). Whole protein extracts were analyzed by Western blot using anti-GFP antibodies (Fig. 2F). Partial proteolytic cleavage of the GFP was observed when LEV-9 was tagged at its C terminus, and the observed fragment corresponded to the size expected if the constructs would have been cut approximately at their Ct region. N-terminally GFP-tagged LEV-9 is also very likely cleaved at its C terminus; however, the little size difference (5 kDa) between the cleaved and uncleaved forms could not be resolved. Clipping off the C-end GFP tag was suppressed by removing either the whole Ct region or the predicted PC site, but not by removing the predicted NdK site (Fig. 2F and data not shown). In the construct where the whole Ct region was removed, an additional band of ϳ50 kDa was observed; one possible explanation would be the presence of a cryptic cleaving site that could be favored when the Ct region is removed. Prevention of the cleavage did not impair secretion based on GFP detection in coelomocytes of all constructs (Fig. 2G).
LEV-9 Cleavage Is Required for L-AChR Synaptic Clustering-To test the functional significance of LEV-9 cleavage at the PC site, we tested the activity of L9⌬PC-GFP and demonstrated that it was no longer rescuing L-AChR clustering, whereas it restored levamisole sensitivity, similar to the truncated LEV-9  version lacking the entire Ct region (Figs. 1A and 2H and data not shown). Worms expressing LEV-9 lacking the nonfunctional NdK site (residues Asp-571/Arg-572/Lys-573 removed) partially rescued L-AChR clustering, but not as efficiently as wild-type LEV-9. This suggested that the last residues present just before the C terminus generated by cleavage at position Arg-575 were important for LEV-9 clustering function. Interestingly, a CCP7-CCP8-Ct construct only partially rescues AChR clustering, suggesting that the Ct functionally interacts with other LEV-9 regions. 5 Alternatively, the short C terminus peptide release by cleavage might also have some biological activity. Expression of a truncated LEV-9 version ending at position Arg-575 (L9⌬ProP) was as efficient as WT LEV-9 (Fig.  2H). Further removal of 5 residues reduced the clustering activity. Adding a GFP at the end of this truncated version completely abolished the activity of the protein. These data suggested that the C-terminal propeptide had no intrinsic activity and that unmasking the C-terminal residues before position Arg-575 is required to activate L-AChR clustering activity.
Because LEV-9 is part of multimolecular scaffold, we tested whether the processing-resistant LEV-9 variant would behave as a dominant-negative. In fact, overexpression of the processing-resistant LEV-9 variant in a wild-type background completely abrogated the clustering of L-AChRs (p Ͻ 0.001, t test) (Fig. 2I). Overexpression of wild-type LEV-9 partially reduced the number of L-AChR at NMJs (p Ͻ 0.001, t test), suggesting that LEV-9 expression or activity is tightly regulated to achieve proper L-AChR clustering.
LEV-9 Cleavage Is Required for Proper Function of the Neuromuscular Junction-Because LEV-9 cleavage is required for proper L-AChR positioning but can be inhibited without affecting the sensitivity to levamisole, we wondered whether LEV-9 processing modulated synaptic activity. We therefore quantified the coordinated movement of worms in liquid, commonly referred to as thrashing. Animals lacking L-AChRs, such as unc-29 mutants, or lacking LEV-9 have a significant thrashing frequency decrease as compared with wild-type worms (Fig. 2J). Thrashing rescue of lev-9 mutants was significantly more efficient with wild-type LEV-9 than with a cleavage-resistant LEV-9 version, indicating that LEV-9 processing is required for proper function of the neuromuscular junction. Altogether, these results suggest that cleaving LEV-9 at its C-end region is a limiting factor for L-AChR clustering and synaptic function.
Proteolytic Processing Generates a Constant LEV-9 Product out of Multiple lev-9 Transcripts-We previously demonstrated that lev-9 is encoded by the T07H6.5 gene because this genomic region rescues all synaptic defects of lev-9 mutants (7). A second gene, T07H6.4, located 300 bp downstream to T07H6.5, is predicted to code for a CCP-containing polypeptide. However, it lacks an identifiable signal peptide in contrast to LEV-9, and most CCP-containing proteins are usually extracellular. Disruption of T07H6.4 by a Tc1 transposon insertion causes no synaptic phenotype (7). Recent data from the mod-ENCODE consortium annotated the genomic region between T07H6.5 and T07H6.4 as the intron of a putative T07H6.5-.4 joined transcript expressed in embryos, mid-L1 and dauer larvae (Fig. 3A) (modENCODE Project website) (20). Using RT-PCR on a mixed staged cDNA library, we were able to amplify a low abundance fragment containing both T07H6.5 last exon and T07H6.4 first exon (Fig. 3, A and B). Sequencing this RT-PCR product revealed that two distinct splice acceptors are used in T07H6.4 (Fig. 3C), generating two novel lev-9 transcripts. lev-9-B is a predicted full-length T07H6.5-T07H6.4 inframe transcript, whereas in isoform C, a reading frameshift causes a nonsense mutation 18 bps downstream from the acceptor site. Interestingly, the lev-9B mRNA region covering the junction between T07H6.5 and T07H6.4 encodes a conserved cryptic CCP domain (CCP-9 in Fig. 3D). Such organization of the lev-9 locus is evolutionarily conserved. In Drosophila melanogaster and C. briggsae for example, a first region homologous to T07H6.5 encodes a predicted signal peptide, an N-terminal WAP domain, and eight CCP domains and is followed by a region homologous to T07H6.4 encoding nine CCP domains but no signal peptide (Fig. 3D). The proteolytic cleavage site after the eight CCP domain is readily detectable in all species. These results suggest that at least three distinct propeptides are encoded by the lev-9 locus and are able to generate the active LEV-9 scaffolding protein by cleavage at position Arg-575. The function of the C-terminal peptide of LEV-9B containing nine CCP domains is yet to be determined (Fig. 3E).
Given our current results, we wondered to what extent the C-terminal cleavage of LEV-9 might be extended to other CCPcontaining secreted factors. Interestingly, for several proteins containing tandems of CCP domains, it has been experimentally determined that cleavage is required. Examples include abundant factors secreted in human serum such as the 2-glycoprotein I or the complement factor C2, in addition to membrane-bound factors such as L-selectins and membrane cofactor protein CD46 (21-24) (Fig. 3F and supplemental Table 1). Cleavage serves either to activate function or to decouple CCP regulatory domains from catalytic domains (21)(22)(23)(24). Interestingly, it occurs mostly C-terminally from tandems of CCP domains, as is the case for LEV-9 in C. elegans. Furthermore, we found conserved predicted PC sites C-terminally from CCP tandems in additional membrane-bound and secreted human factors ( Fig. 3G and supplemental Table 1). This analysis therefore suggests that cleaving C-terminally from CCP domains to either control activity or structurally decouple CCP domains from C-terminal attachments is an architectural principle that has been conserved in proteins of unrelated function.