CASK and FARP localize two classes of post-synaptic ACh receptors thereby promoting cholinergic transmission

Changes in neurotransmitter receptor abundance at post-synaptic elements play a pivotal role in regulating synaptic strength. For this reason, there is significant interest in identifying and characterizing the scaffolds required for receptor localization at different synapses. Here we analyze the role of two C. elegans post-synaptic scaffolding proteins (LIN-2/CASK and FRM-3/FARP) at cholinergic neuromuscular junctions. Constitutive knockouts or muscle specific inactivation of lin-2 and frm-3 dramatically reduced spontaneous and evoked post-synaptic currents. These synaptic defects resulted from the decreased abundance of two classes of post-synaptic ionotropic acetylcholine receptors (ACR-16/CHRNA7 and levamisole-activated AChRs). LIN-2’s AChR scaffolding function is mediated by its SH3 and PDZ domains, which interact with AChRs and FRM-3/FARP, respectively. Thus, our findings show that post-synaptic LIN-2/FRM-3 complexes promote cholinergic synaptic transmission by recruiting AChRs to post-synaptic elements.


Introduction
Synaptic connections mediate fast intercellular communication in the nervous system.
Synapses are elaborate structures comprising hundreds of pre-and post-synaptic proteins, each highly concentrated at the sites of intercellular contact. Although synaptic proteins are highly enriched at these contacts, they nonetheless must retain the ability to undergo dynamic processes (e.g. insertion, removal, intermolecular contacts, and post-translational modifications). The abundance of a protein at a synapse often dictates signaling properties of that synapse. For example, the location and abundance of the synaptic vesicle priming protein (UNC-13) controls the probability and kinetics of release (Hu et al., 2013;Liu et al., 2019).
Similarly, the abundance of post-synaptic receptors dictates the size of synaptic currents (Kauer et al., 1988;Luscher et al., 1999). Beyond recruiting rate limiting synaptic components, synaptic transmission also relies upon the precise coordination of pre-and postsynaptic specializations. For example, the site of synaptic vesicle (SV) fusion in presynaptic cells (termed Active Zones) must be precisely aligned with clustered receptors in the contacting post-synaptic cell. Slight shifts in trans-synaptic alignment results in significant alterations in the ongoing signaling at that synapse (Biederer et al., 2017;Chen et al., 2018;Haas et al., 2018). For these reasons, there is significant interest in determining how rate limiting synaptic components are localized and coordinated trans-synaptically.
CASK (a MAGUK family scaffold protein) regulates synaptic function (Dimitratos et al., 1997;Butz et al., 1998;Hsueh et al., 1998;Chen and Featherstone, 2011). Apart from the PDZ, SH3, and GK domains found in all MAGUK proteins, CASK also contains an Nterminal Ca 2+ /calmodulin-dependent protein kinase (CaM) domain ( Figure 1A). CASK orthologs are found in invertebrates and vertebrates, suggesting that CASK's synaptic function is strongly conserved across phylogeny. In mammals, CASK is primarily expressed in the brain and localizes in synaptic regions (Hata et al., 1996;Hsueh et al., 1998;Stevenson et al., 2000). Similar to other scaffolds, CASK participates in multiple interactions in different signalling processes, and has been implicated in synaptic protein targeting, synaptic organization, and transcriptional regulation (Hata et al., 1996;Hsueh et al., 2000). Human CASK has also been linked to neurological disorders such as autism and brain malformation (Dimitratos et al., 1998;Najm et al., 2008). In mouse cortical neurons and Drosophila NMJ, glutamatergic synaptic transmission is impaired by the knockout of CASK (Atasoy et al., 2007;Chen and Featherstone, 2011). In C. elegans, LIN-2/CASK together with its binding partner FRM-3 (ortholog of the mammalian scaffold FARP), promote GABAergic synaptic 5 transmission by immobilizing GABAA receptors at the NMJ (Tong et al., 2015;Zhou et al., 2020). A recent study showed that LIN-2/CASK and FRM-3/FARP are also required for stabilizing ACR-16/CHNRA7 receptors at the C. elegans NMJ (Zhou et al., 2021). Despite these advances in both vertebrates and invertebrates, key questions remain to be addressed regarding the function of CASK in cholinergic synapses. How do CASK and its binding partners regulate AChRs? Does CASK function by similar mechanisms at cholinergic and GABAergic synapses?
Here we show that LIN-2/CASK and FRM-3/FARP promote cholinergic transmission by recruiting two classes of ionotropic AChRs, ACR-16/CHRNA7 and levamisole-activated AChRs (L-AChRs), to post-synaptic elements. The SH3 domain in LIN-2/CASK plays important roles by directly interacting with ionotropic AChRs. Our results provide a detailed picture of how LIN-2/FRM-3 complexes shape the function of cholinergic synapses.

Loss of function of lin-2 or frm-3 decreases locomotion speed
To assess LIN-2 and FRM-3's impact on behavior, we measured locomotion speed in lin-2 and frm-3 mutants. The lin-2 gene encodes two isoforms, which share the PDZ, SH3, and GK domains while only LIN-2A contains the CaMK homology domain (Hoskins et al., 1996). The lin-2(e1309) mutation deletes an N-terminal region of LIN-2A but has no effect on LIN-2B ( Figure 1A). The frm-3 gene encodes three isoforms, two of which (FRM-3A and B) possess a FERM domain. The gk585 allele is a deletion disrupting both frm-3a and b ( Figure 1C). Locomotion speed was significantly decreased in lin-2(e1309) mutants and frm-3(gk585) mutants ( Figure 1E, F), suggesting that LIN-2A and FRM-3 are required for the function of the locomotion circuit. No additional decrease in 6 locomotion speed was observed in lin-2(e1309); frm-3(gk585) double mutants, indicating that LIN-2 and FRM-3 act together to promote locomotion.
Compared to lin-2(e1309) mutants, lin-2 null mutants exhibited more severe defects in cholinergic synaptic transmission, including larger decreases in mEPSC frequency and amplitude, and evoked EPSC amplitude and charge transfer (Figure 2A, B, E, F). The mEPSC frequency in frm-3 null mutants was also significantly lower than in frm-3(gk585) mutants. Moreover, mIPSC frequencies, which were unaltered in lin-2(e1309) and frm-3(gk585) mutants, were significantly decreased in both lin-2 and frm-3 null mutants ( Figure   2C, D). These results demonstrate that synaptic transmission was more severely impaired in the lin-2 and frm-3 null mutants, further establishing that these scaffolds play a pivotal role in the function of cholinergic synapses. It should be noted that LIN-2's impact on cholinergic transmission does not require its other binding partners LIN-7/Velis and LIN-10/Mint (Butz et al., 1998;Kaech et al., 1998), as the lin-7 and lin-10 mutants exhibited normal mEPSCs and evoked EPSCs ( Figure S1).
The decreased mEPSC amplitude suggests that postsynaptic ionotropic AChRs are disrupted in lin-2 and frm-3 mutants (e.g., decrease in synaptic abundance). It should be noted that smaller mEPSC amplitudes may cause some mEPSC events to become undetectable, thereby reducing mEPSC frequency. This is supported by the fact that mEPSC frequency is decreased by 55% in acr-16 mutants, which lack nicotine-sensitive receptors that are orthologous to mammalian CHRNA7 receptors ( Figure S2A, B). Thus, the decreased mEPSC rates observed in lin-2 and frm-3 mutants could be caused by the decreased mEPSC amplitude. Moreover, the severe defects in mEPSCs but normal locomotion speed in the acr-8 16 mutants ( Figure S2C) also explained the moderate decrease in locomotion in lin-2 and frm-3 mutants. Taken together, our results revealed that LIN-2/CASK and FRM-3/FARP play essential roles in cholinergic synaptic transmission, likely by promoting post-synaptic responses to ACh.
Using CRISPR, we introduced a stop cassette into lin-2 and frm-3 introns (in the opposite orientation). For both genes, the STOP cassette was inserted into an intron shared by both isoforms ( Figure 3A). The STOP cassette is bounded by FLEX recombination signals mediating CRE-induced inversions. In this manner, CRE expression inverts the stop cassette thereby blocking expression of LIN-2(A and B) and FRM-3(A and B). Neuron-specific knockouts (KO) were made by expressing Psbt-1::Cre (termed Neuron Cre ) while body wall muscle KOs were made by expressing Pmyo-3::Cre (termed Muscle Cre ).
We next examined synaptic transmission following lin-2 and frm-3 muscle and neuron KO. Compared to control lin-2(nu743) and frm-3(nu751) mutants (lacking Cre expression), the frequency and amplitude of mEPSCs, and the amplitude and charge transfer of evoked EPSCs, were all severely reduced by Muscle Cre , but were not changed by Neuron Cre (Figure 3B-I), demonstrating that LIN-2/CASK and FRM-3/FARP both promote cholinergic synaptic transmission by acting in body wall muscles. Similarly, Muscle Cre also significantly decreased the mIPSC amplitude in lin-2(nu743) and frm-3(nu751) mutants ( Figure 3B, C, F, G), indicating that LIN-2 and FRM-3 also act in muscle to regulate 9 postsynaptic GABAA receptors, consistent with prior studies (Tong et al., 2015;Zhou et al., 2020). However, unlike cholinergic synapses, Muscle Cre decreased the mIPSC frequency by only 50% in lin-2(nu743) mutants (versus 90% in lin-2 null mutants), and did not change mIPSC frequency in frm-3(nu751) mutants. Collectively, these results indicate that LIN-2 and FRM-3 control synaptic transmission at both cholinergic and GABAergic synapses but that the detailed mechanisms likely differ at these two synapses.
We next analyzed the locomotion speed in lin-2 and frm-3 conditional knockout mutants. Our results showed that the speed is significantly decreased in lin-2 Muscle Cre mutants ( Figure S3), consistent with that in lin-2 null mutants ( Figure 1F). However, the speed is not altered in either the Muscle Cre or Neuron Cre mutants of frm-3. One possibility is that the efficiency of the Cre recombinase in frm-3(nu751) is not as strong as that in lin-2(nu743). This is likely because the mIPSC frequency is not decreased in frm-3 Muscle Cre mutants, unlike frm-3 null mutants ( Figure 1F and Figure 2D). The unchanged synaptic transmission in the Neuron Cre mutants of lin-2 and frm-3 may arise from a failure of the Cre recombinase to inactivate lin-2 and frm-3 in ACh neurons. To address this concern, we asked if Cre recombinase expression could inactivate an essential pre-synaptic gene, unc-2. The unc-2 gene encodes the P/Q-type Ca 2+ channel in C. elegans.
Loss of unc-2 function leads to strong decrease in both locomotion rate and Ca 2+ -dependent neurotransmitter release (Tong et al., 2017;Liu et al., 2018). The unc-2(nu657 FLEX) allele contains an inverted STOP cassette in intron 17. The presence of CRE is expected to inactivate unc-2, leading to behaviour and transmission defects. As shown in Figure S4, Neuron Cre in unc-2 (nu657) significantly decreased locomotion speed, mEPSC frequency, and evoked EPSC amplitude and charge transfer. These results are similar to those observed in unc-2 mutants, demonstrating that the Cre recombinase is highly efficient in cholinergic motor neurons.

AChRs
The dramatic decrease in mEPSC amplitude in lin-2 and frm-3 null mutants suggests that the ionotropic AChRs are impaired in postsynaptic body wall muscles. There are two classes of ionotropic AChRs expressed on body wall muscles in C. elegans, nicotinesensitive homo-pentameric receptors (nAChRs) consisting of five ACR-16/CHRNA7 subunits, and levamisole-sensitive hetero-pentameric receptors (L-AChRs) containing alternative a subunits (UNC-38, UNC-63, and LEV-8) and non-a subunits (UNC-29 and LEV-1) (Richmond and Jorgensen, 1999;Brown et al., 2006). These two types of AChRs can be distinguished by their differential desensitization in response to long-time exposure to acetylcholine (Liu et al., 2009). Moreover, because of the difference in channel conductance, activation of nAChRs elicits large and rapidly decaying responses, and L-AChRs produce small synaptic responses with slow decay when activated (Zhang et al., 1996;Ullian et al., 1997;Babu et al., 2011).
To estimate the synaptic abundance of nAChRs, we imaged ACR-16::RFP (expressed by the myo-3 promoter in a single copy transgene). ACR-16::RFP exhibits a punctate distribution in the nerve cords, consistent with previous observation (Babu et al., 2011).
Compared to wild-type animals, the lin-2 and frm-3 null mutants exhibited significantly reduced ACR-16::RFP puncta fluorescence in the dorsal nerve cord ( Figure 4A, B), consistent with results reported in a recent study (Zhou et al., 2021). These observations are consistent with the decreased mEPSC and evoked EPSC amplitude ( Figure 2). We next asked if loss of LIN-2 and FRM-3 alters the synaptic abundance of endogenously expressed L-AChRs, by analyzing the puncta intensity produced by the unc-29(kr208) allele, which contains an RFP tag (Tu et al., 2015). Our data showed that the UNC-29::RFP puncta fluorescence was also remarkably decreased in both lin-2 and frm-3 null mutants ( Figure 4C, D). There were no additional decreases in ACR-16::RFP and UNC-29::RFP in the lin-2;frm-3 double mutants ( Figure 4B, D), suggesting that LIN-2 and FRM-3 function together to regulate both ACR-16/CHRNA7 and L-AChR synaptic abundance.
LIN-2 and FRM-3 regulation of post-synaptic AChRs is also supported by the altered kinetics of mEPSCs. In lin-2 and frm-3 null mutants, as well as their Muscle Cre mutants, mEPSCs exhibited significantly slower decay kinetics ( Figure S5), which likely results from a greater contribution of L-AChRs to the synaptic current. These observations further suggest a defect in postsynaptic AChRs in lin-2 and frm-3 mutants. Together, our results demonstrate that LIN-2 and FRM-3 act in the same genetic pathway regulating the synaptic abundance of two classes of ionotropic AChRs in body wall muscles.
To examine the function of endogenous nAChRs and L-AChRs on the muscle surface, we measured acetylcholine-and levamisole-activated synaptic responses. A pulse application of ACh (0.5mM, 100ms) onto body wall muscle elicited a large and robust current in wild-type animals. In lin-2 and frm-3 null mutants, ACh-activated currents were indistinguishable from that in wild type ( Figure 4E, F). Puffing levamisole (100µM, 100ms) produced a small but long-lasting current in wild-type animals, and the levamisole-evoked currents were also unaltered in lin-2 and frm-3 null mutants ( Figure 4G, H). These results indicate that inactivating LIN-2 and FRM-3 did not alter the total expression level or surface delivery of nAChRs and L-AChRs in muscles. Instead, inactivating LIN-2 and FRM-3 prevent nAChRs and L-AChRs clustering or destabilize receptor clusters at synapses, resulting in decreased ACR-16::RFP and UNC-29::RFP puncta fluorescence and decreased mEPSC and evoked EPSC amplitudes.
We found that GABA-activated currents were significantly reduced (by 50%) in lin-2 null mutants but were unchanged in frm-3 null mutants ( Figure 4I, J). These results suggest that LIN-2 also controls total surface GABAAR levels in muscles.
To further confirm the regulation of LIN-2 and FRM-3 on the surface levels of AChRs and GABAARs, we also measured ACh-, levamisole-, and GABA-activated currents in the muscle cKO mutants of lin-2 and frm-3. Overall, the results were similar to those observed in the null mutants ( Figure S6). We only found a significant decrease in GABAactivated current in lin-2 Muscle Cre mutants, providing further support for a previously unknown role of LIN-2 in the trafficking of GABAARs to the cell surface.

Synaptic morphology is normal in lin-2 and frm-3 mutants
Thus far, our results suggest that LIN-2 and FRM-3 are required for post-synaptic function; however, it remains possible that these scaffolds also have important pre-synaptic functions. To address this possibility, we examined the synaptic ultrastructure of cholinergic motor neurons. Morphometric analyses showed that the terminal size and total number of SVs in lin-2 and frm-3 mutants were indistinguishable from that in wild-type control ( Figure   5A, B). Like the cholinergic synapses, the GABAergic synapses also exhibited normal 13 terminal size and number of SVs (data not shown). These results indicate that lin-2 and frm-3 mutations do not have dramatic effects on the morphology of presynaptic terminals.

SV abundance is not impaired in lin-2 and frm-3 mutants
To further examine the potential effects of LIN-2 and FRM-3 on presynaptic terminals, we examined SV abundance in lin-2 and frm-3 mutants. SV abundance was analyzed by imaging synaptic vesicle markers UNC-57::mCherry (expressed by the unc-129 promoter in a single copy transgene). UNC-57::mCherry fluorescence displays punctate distribution in dorsal nerve cord in wild-type animals ( Figure 6A). Our data showed that UNC-57::mCherry in lin-2 and frm-3 mutants have comparable puncta fluorescence and density compared to wild-type controls ( Figure 6B), indicating that presynaptic SV abundance is normal in these two mutants. These results are consistent with the observation in EM in which the total number of SVs is unaltered in lin-2 and frm-3 mutants ( Figure 5B).
Together, our results indicate that SV abundance is unchanged in lin-2 and frm-3 mutants, and consequently that the cholinergic transmission defects observed in these mutants are likely a consequence of post-synaptic defects.

The SH3 domain in LIN-2/CASK binds to both ACR-16 and UNC-29
The decreased nAChRs and L-AChRs in lin-2 and frm-3 mutants indicates that LIN-2/CASK and FRM-3/FARP may physically bind to these receptors and regulate their synaptic abundance. To test this, we performed yeast two-hybrid (Y2H) assays. Our data showed that LIN-2A could bind to the second cytoplasmic loop of both ACR-16 and UNC-29 ( Figure 7A-C), with stronger interaction with the ACR-16 loop than the UNC-29 loop. These results are consistent with the findings that lin-2 or frm-3 knockout led to a decrease in puncta fluorescence of AChRs ( Figure 4). However, no direct interactions were detected between 14 FRM-3 and the cytoplasmic loop of ACR-16 or UNC-29. By contrast, a recent study reported that ACR-16 binds both LIN-2 and FRM-3 (Zhou et al., 2021). This discrepancy could result from the use of different binding assays (i.e. Y2H versus GST pull-down). Nevertheless, since FRM-3 physically binds to LIN-2, FRM-3's impact on AChRs is likely mediated by LIN-2/FRM-3/ACR-16 and LIN-2/FRM-3/UNC-29 ternary complexes.
LIN-2 contains PDZ and SH3 domains that mediate protein-protein interactions. To examine which domain(s) in LIN-2 mediates the interaction with ACR-16 and UNC-29, we performed a Y2H assay to test the interaction of LIN-2's PDZ and SH3 domains with ACR-16. Our results showed that the SH3 but not the PDZ domain directly binds to a cytoplasmic loop of ACR-16 ( Figure 7A

The PDZ domain in LIN-2/CASK and the FERM domain in FRM-3/FARP are required for LIN-2/FRM-3 interaction
To understand how the LIN-2/FRM-3 complex is formed, we examined what domains in LIN-2 and FRM-3 mediate their interaction. In LIN-2, we focused on the central PDZ and SH3 domains. Consistent with our previous findings, strong interaction was observed 15 between full-length LIN-2 and FRM-3 ( Figure 7A, F) (Tong et al., 2015). No interaction was detected between the SH3 domain and FRM-3, indicating that SH3 does not mediate the LIN-2/FRM-3 interaction ( Figure 7A, F). This was confirmed by the LIN-2DSH3, which still exhibited strong binding to FRM-3 ( Figure 7A, F). Like the SH3 domain, the isolated PDZ domain also did not exhibit binding to FRM-3. However, the LIN-2DPDZ failed to bind to FRM-3 ( Figure 7A, F). This indicates that the PDZ domain is required for LIN-2/FRM-3 interaction, although it alone is not sufficient to mediate this interaction. In FRM-3, the FERM domain exhibited strong binding to both the full-length LIN-2 and the LIN-2DSH3, but not the SH3, PDZ, and LIN-2DPDZ ( Figure 7A, G), similar to the binding pattern seen with full-length FRM-3. Taken together, our results demonstrate that LIN-2 binds to the FERM domain in FRM-3, and this interaction requires the PDZ domain in LIN-2.

2/CASK
The interaction between LIN-2 SH3 and AChRs prompted us to investigate the functional importance of this domain. We deleted SH3 by CRISPR and analyzed synaptic transmission ( Figure 8A; syb1921, DSH3). Our data showed that cholinergic transmission was dramatically decreased in DSH3 mutants, including mEPSC frequency and amplitude, evoked EPSC amplitude and charge transfer, and mEPSC charge ( Figure 8B-F). These results indicate that the SH3 domain is indispensable for LIN-2's function, and the cholinergic synaptic transmission is determined by its interaction with AChRs. As the PDZ domain is required for LIN-2/FRM-3 interaction, we also tested whether it participates in the regulation of cholinergic synaptic transmission. Interestingly, deleting the PDZ domain (syb1937, DPDZ; Figure 8A) caused similarly stronger decrease in cholinergic transmission, demonstrating that the PDZ domain also plays essential roles in LIN-2's synaptic function.
The functional importance of the SH3 and PDZ domains was also evaluated in GABAergic synapses. The mIPSC amplitude in DSH3 and DPDZ mutants was decreased to a similar extent to that in lin-2 null mutants ( Figure 8G, H). However, the mIPSC frequency was not changed in the DSH3 and DPDZ mutants, unlike the lin-2 null mutants ( Figure 2).
These results suggest that the mutant LIN-2 proteins are expressed and retain some function.
Collectively, our findings provide novel and interesting models regarding the functions of scaffolding proteins LIN-2/CASK and FRM-3/FARP in C. elegans NMJ. We demonstrate that LIN-2 and FRM-3 regulate the clustering and/or stabilization of nAChRs and L-AChRs in body wall muscles, without affecting their overall expression or delivery to the cell surface. These functions require the participation of the SH3 and PDZ domains of LIN-2, which mediate interaction with AChRs and FRM-3. At the GABAergic synapses, postsynaptic LIN-2 and FRM-3 are also involved in the regulation of GABAARs. Moreover, the delivery of GABAARs to the cell surface requires LIN-2 but not FRM-3.

Discussion
Different from the NMJs in mouse and fly which use ACh or glutamate as their primary neurotransmitters, the C. elegans NMJs comprise both cholinergic and GABAergic synapses (Richmond and Jorgensen, 1999). This allows us to use the worm NMJ to investigate three different neurotransmitter receptors, nAChRs, L-AChRs, and GABAARs.
Moreover, the major known synaptic organizers found in vertebrates, such as PSD-95, rapsyn, MuSK, and neuroligin, also exist in worm NMJ, suggesting that the regulatory mechanisms of the AChRs and GABAARs in C. elegans are similar to those in other species.
In this study, we investigated the mechanisms by which LIN-2/CASK and FRM-3/FARP regulate cholinergic synaptic transmission at the C. elegans NMJ. Our results lead to several previously unknown findings. First, the lin-2 and frm-3 null mutants exhibit severe defects in cholinergic synaptic transmission and locomotion behavior. Second, the cholinergic Below we discuss the significance of these findings.

CASK and FARP regulate cholinergic synaptic transmission
Since its discovery as a neurexin-binding protein (Hata et al., 1996), CASK has been found to be broadly involved in the regulation of cell-cell junction organization. The high homology of the C-terminal domains in CASK (e.g., PDZ, SH3, and GK) with other MAGUK proteins such as PSD95 has suggested that CASK may function similarly by acting as a scaffolding protein, while the presence of the unique N-terminal CaM kinase domain endows CASK additional functions in various cellular pathways (Lu et al., 2003). In addition to neurexin, CASK also binds to many other membrane proteins, such as Mint-1, Velis, Syndecan, FARP, glycophorins, and SynCaM through its central PDZ domain (Hata et al., 1998;Martinez-Estrada et al., 2001;Hsueh, 2006;Mukherjee et al., 2008). The CASK homolog in C. elegans LIN-2 is required for vulva development and proper localization of the EGF receptor LET-23 (Horvitz and Sulston, 1980;Hoskins et al., 1996).
Despite its functions in membrane organization, the physiological importance of CASK in synapses has been controversial. Knockout of CASK in mice causes lethal phenotype but does not affect the development of neurons. Cortical neurons lacking CASK display normal ultrastructural morphology and evoked EPSCs, whereas spontaneous glutamate release is increased but spontaneous GABA release is decreased (Atasoy et al., 2007). In Drosophila, loss of CAKI, the CASK ortholog, leads to increased frequency of MEPPs without changes in MEPP amplitude (Zordan et al., 2005). However, a later study in Drosophila found reduced spontaneous and evoked neurotransmitter release, as well as decreased postsynaptic glutamate receptors in CASK knockout mutants (Chen and Featherstone, 2011).
FARP was first recognized as a synaptic component that is required for dendrite growth in spinal motor neurons (Zhuang et al., 2009). Later studies revealed that FARP forms a synaptic complex with SynCAM1 through its FERM domain and functions as a synaptic organizer in cultured dissociated hippocampal neurons (Cheadle and Biederer, 2012). It is enriched at postsynaptic sites and increases synapse number and modulates spine morphology. At C. elegans NMJs, FRM-3/FARP and its binding protein LIN-2/CASK promote post-synaptic localization of GABAAR, most likely by immobilizing receptors at post-synaptic specializations (Tong et al., 2015;Zhou et al., 2020). Here we describe LIN-2/CASK and FRM-3/FARP function at cholinergic synapses, finding that they regulate postsynaptic abundance of ionotropic AChRs.
Several lines of evidence suggest that LIN-2/CASK and FRM-3/FARP form a scaffold complex regulating synaptic transmission at the C. elegans NMJ. First, these two scaffolds physically bind each other (Tong et al., 2015;Zhou et al., 2020). Second, synaptic transmission defects in lin-2 and frm-3 mutants are comparable, and no additional decrease in the lin-2; frm-3 double mutants (Figure 1), demonstrating that they act in the same genetic pathway. Third, both scaffolds function in postsynaptic body wall muscle regulating the synaptic abundance of two classes of AChRs (Figure 4) (Zhou et al., 2021). Although postsynaptic currents are dramatically reduced, locomotion rates were only modestly reduced in lin-2 and frm-3 null mutants. This persistent locomotion behavior likely reflects the residual function of post-synaptic L-AChRs in these mutants. Prior studies of the Drosophila NMJ have proposed that neurexin may also be involved in the CASK and protein 4.1 complex (CASK/neurexin/4.1), which regulates glutamate receptor clustering (Chen and Featherstone, 2011). However, it seems that the LIN-2/FRM-3 complex does not require neurexin for clustering and stabilizing ionotropic AChRs at the worm NMJ, as the worm mutants lacking nrx-1/neurexin exhibit normal mEPSCs and evoked EPSCs (Hu et al., 2012;Tong et al., 2015). A recent study showed that the loss of g-neurexin causes a moderate decrease in evoked EPSCs at the NMJ; however, the mEPSCs, including frequency and amplitude, are unchanged, indicating that the ionotropic AChRs are functionally normal (Kurshan et al., 2018).
Despite their expression in motor neurons, our data show that LIN-2 and FRM-3 are not required for SV fusion at the presynaptic terminals, evidenced by the normal mEPSCs and evoked EPSCs in lin-2 and frm-3 neuron cKO mutants (Figure 3). It remains possible that LIN-2 and FRM-3 may play important roles in other neuronal functions, e.g. promoting the localization of presynaptic inotropic receptors.

Functional roles of LIN-2A and LIN-2B isoforms in synaptic transmission
The C. elegans lin-2 locus expresses two isoforms: a long isoform (LIN-2A) that includes the N-terminal CaM kinase domain and a short isoform (LIN-2B) lacking this domain ( Figure 1) (Hoskins et al., 1996). It is believed that the lin-2 e1309 mutants express only the LIN-2B isoform, as the deletion removes the whole promoter region for LIN-2A and 20 part of the CaM kinase domain ( Figure 1A). Our data show that LIN-2A plays a more dominant role in cholinergic transmission, accounting for 70% of the evoked neurotransmitter release. The more severe decrease in cholinergic synaptic transmission including both mEPSCs and evoked EPSCs in the lin-2 null mutants demonstrates that the LIN-2B isoform also plays important roles in synaptic function. In GABAergic synapses, it seems the LIN-2B isoform plays similarly important role as LIN-2A, as the mIPSC amplitude was decreased by 20% in the e1309 mutants and was decreased by a further 20% in the null mutants, different from the observations in mEPSC amplitude (30% decrease in e1309 and 40% decrease in null). Moreover, the mIPSC frequency was severely decreased in the null mutants but was not changed in the e1309 mutants. These results suggest that LIN-2A and LIN-2B may function redundantly in GABAergic synapses ( Figure 2D synaptic receptor clustering could be mediated (at least in part) by syndecan. This is supported by the recent findings that sdn-1/syndecan regulates the synaptic clustering of AChRs by recruiting LIN-2/CASK and FRM-3/FARP at the C. elegans cholinergic NMJs (Zhou et al., 2021). Future studies will investigate these possibilities, as well as other potential LIN-2/FRM-3 binding partners in organizing post-synaptic receptor clustering.
Several results suggest that the deficits in synaptic transmission observed in lin-2 and frm-3 mutants arise primarily from failure to cluster post-synaptic ACh and GABA receptors.
Muscle specific lin-2 and frm-3 knockouts recapitulate the ACh and GABA NMJ electrophysiological and post-synaptic receptor clustering defects observed in the corresponding null mutants, whereas neuron specific knockouts had no effects. The ultrastructure of ACh and GABA pre-synaptic terminals and the localization of several presynaptic markers was unaffected in lin-2 and frm-3 null mutants. Collectively, these results suggest that pre-synaptic structure and function are largely unaffected by inactivating LIN-2 and FRM-3. Our results do not exclude the possibility that further analysis will reveal presynaptic functions for these scaffolding proteins, as has been described for mammalian synapses (Atasoy et al., 2007).

Strains
Animals were cultivated at room temperature on nematode growth medium (NGM) agar plates seeded with OP50 bacteria. On the day before experiments L4 larval stage animals were transferred to fresh plates seeded with OP50 bacteria for all the electrophysiological, imaging, and behavioural experiments. The following strains were used: 23 EN208 kr208  TXJ0569 lin-2(syb1019); kr208  TXJ0816 frm-3(gk585); kr208  KP9809 nu586 ; nuSi250

CRISPR alleles
CRISPR alleles were isolated as described (Arribere et al., 2014). Briefly, unc-58 was used as a co-CRISPR selection to identify edited animals. Animals were injected with two guide RNAs (gRNAs) and two repair templates, one introducing an unc-58 gain of function mutation and a second modifying a gene of interest. Progeny exhibiting the unc-58(gf) uncoordinated phenotype were screened for successful editing of the second locus by PCR.
Split GFP and split sfCherry constructs are described in (Feng et al., 2017).
Tissue specific frm-3 and lin-2 knockout was performed by introducing a stop cassette into introns in the ON configuration (i.e. in the opposite orientation of the target gene) using CRISPR, creating the frm-3(nu751, flex ON) and lin-2(nu743, flex ON) alleles. In frm-3(nu751, flex ON), the stop cassette was inserted into an intron shared by frm-3a and b (intron 3 of frm-3a). Similarly, the stop cassette in lin-2(nu743, flex ON) was inserted into an 25 intron shared by both lin-2 isoforms (intron 3 of lin-2b). The stop cassette consists of a synthetic exon (containing a consensus splice acceptor sequence and stop codons in all reading frames) followed by a 3' UTR and transcriptional terminator taken from the flp-28 gene (the 564bp sequence just 3' to the flp-28 stop codon). The stop cassette is flanked by FLEX sites (which are modified loxP sites that mediate CRE induced inversions) (Schnutgen and Ghyselinck, 2007). In this manner, orientation of the stop cassette within the intron is controlled by CRE expression. Expression of the targeted gene is reduced when the stop cassette is in the OFF configuration (i.e. the same orientation as the targeted gene) but is unaffected in the ON configuration (opposite orientation). The endogenous flp-28 gene is located in an intron of W07E11.1 (in the opposite orientation). Consequently, we reasoned that the flp-28 transcriptional terminator would interfere with frm-3 and lin-2 expression in an orientation selective manner. A similar strategy was previously described for conditional gene knockouts in Drosophila (Fisher et al., 2017).

Locomotion and behavioral assays
Young adult animals were washed with a drop of PBS and then transferred to fresh NGM plates with no bacterial lawn (30 worms per plate). Worm movement recordings (under room temperature 22°C) were started 10 min after the worms were transferred. A 2 min digital video of each plate was captured at 3.75 Hz frame rate by WormLab System (MBF Bioscience). Average speed and tracks were generated for each animal using WormLab software.

Fluorescence imaging
Worms were immobilized by 30 g/l 2,3-Butanedione monoxime (Sigma) and mounted on 2% agar on glass slides. Fluorescence images were captured by 100x (NA=1.4) objective on an Olympus microscope (BX53). The mean fluorescence intensities of reference FluoSphere microspheres (Thermo Fisher Scientific) were measured during each experiment, and were used to control for illumination intensities changes. Multidimensional data were reconstructed as maximum intensity projections using Metamorph software (Molecular Devices). Line scans were analyzed in Igor Pro (WaveMetrics) using a custom script. The intensity of peak fluorescence of each marker was analyzed.

Electron microscopy
Samples were prepared using high-pressure freeze fixation (Weimer et al., 2006). ~30 young adult hermaphrodites were placed in each specimen chamber containing E. coli and were 27 frozen at -180°C under high pressure (Leica SPF HPM 100). Frozen specimens then underwent freeze substitution (Leica Reichert AFS) during which the samples were held at −90°C for 107 h in 0.1% tannic acid and 2% OsO4 in anhydrous acetone. The temperature was then increased from 5°C/h to −20°C, kept at −20°C for 14 h, and increased by 10°C/h to 20°C. After fixation, samples were infiltrated with 50% Epon/acetone for 4 h, 90% Epon/acetone for 18 h, and 100% Epon for 5 hr. Finally, samples were embedded in Epon and incubated for 48 h at 65°C. Ultra-thin serial sections (50 nm) were cut and glued to a wafer, and counterstained in 0.08 mol/L citrate for 10 minutes. Images were acquired using a GeminiSEM460 scanning electron microscope operating at 5 kV. Images were collected from the ventral and dorsal nerve cord region anterior to the vulva for all genotypes. Cholinergic synapses were identified on the basis of their typical morphology and anatomical features (White et al., 1986). A synapse was defined as a series of sections (profiles) containing a dense projection as well as two flanking sections on both sides without dense projections.
Image acquisition and analysis using NIH ImageJ/Fiji software were performed blinded for genotype.

Yeast-two-hybrid
The yeast two-hybrid assay was performed with the Matchmaker Gold Yeast Two-Hybrid The plasmid pairs were simultaneously transformed into the Y2HGold yeast cells.
Transformants selected from the SD-Leu-Trp plates were restreaked onto SD-Leu-Trp-His-Ade plates to test interactions. Meanwhile, the pGBKT7-53 and pGADT7-T plasmids pair were used as positive controls, and pGBKT7-Lam and pGADT7-T plasmids pair were used as negative controls. False-positive from autoactivation was ruled out by co-transformation of pGBKT7-LIN-2A construct with pGADT7 empty vector alone.

Data acquisition and statistical analysis
All electrophysiological data were obtained using a HEKA EPC10 double amplifier (HEKA Elektronik) filtered at 2 kHz, and analyzed with open-source scripts developed by Eugene Mosharov (http://sulzerlab.org/Quanta_Analysis_8_20.ipf) in Igor Pro 7 (Wavemetrics). All imaging data were analyzed in ImageJ software. Each set of data represents the mean ± SEM of an indicated number (n) of animals. To analyze mEPSCs and mIPSCs, a 3.5pA peak threshold was preset, above which release events are clearly distinguished from background noise. The analyzed results were re-checked by eye to ensure that the release events were accurately selected.

29
All data were statistically analyzed in Prism 8 software. Normality distribution of the data was determined by the D'Agostino-Pearson normality test. When the data followed a normal distribution, an unpaired student's t-test (two-tailed) or one-way ANOVA was used to evaluate the statistical significance. In other cases, a Mann-Whitney test or one-way ANOVA following Kruskal-Wallis test was used.

Data Availability
Data generated or analysed during this study are included in this published article or available upon request.

Acknowledgement
We thank the C. elegans Genetics Stock Center for strains and reagents. This work was Research Project from the Science and Technology Commission of Shanghai Municipality (19JC1414100 to X-J.T. and 21ZR1481000 to X-J.T.).

Author contributions
L Li: conceptualization, formal analysis, investigation, and writing -review and editing.
H Liu: conceptualization, formal analysis, investigation, and writing -review and editing.
KY Qian: conceptualization, formal analysis, investigation, and writing -review and editing.
S Nurrish: conceptualization, formal analysis, investigation, and writing -review and editing.

Conflict of Interest Statement
The authors declare no competing interests. Data are mean ± SEM (**, p < 0.01, ***, p < 0.001 when compared to wild type; n.s., nonsignificant; one-way ANOVA). The number of worms analyzed for each genotype is indicated in the bar.