Long-term depression-inducing stimuli promote cleavage of the synaptic adhesion molecule NGL-3 through NMDA receptors, matrix metalloproteinases and presenilin/γ-secretase

Long-term depression (LTD) reduces the functional strength of excitatory synapses through mechanisms that include the removal of AMPA glutamate receptors from the postsynaptic membrane. LTD induction is also known to result in structural changes at excitatory synapses, including the shrinkage of dendritic spines. Synaptic adhesion molecules are thought to contribute to the development, function and plasticity of neuronal synapses largely through their trans-synaptic adhesions. However, little is known about how synaptic adhesion molecules are altered during LTD. We report here that NGL-3 (netrin-G ligand-3), a postsynaptic adhesion molecule that trans-synaptically interacts with the LAR family of receptor tyrosine phosphatases and intracellularly with the postsynaptic scaffolding protein PSD-95, undergoes a proteolytic cleavage process. NGL-3 cleavage is induced by NMDA treatment in cultured neurons and low-frequency stimulation in brain slices and requires the activities of NMDA glutamate receptors, matrix metalloproteinases (MMPs) and presenilin/γ-secretase. These results suggest that NGL-3 is a novel substrate of MMPs and γ-secretase and that NGL-3 cleavage may regulate synaptic adhesion during LTD.

Synaptic adhesion molecules are thought to contribute to synaptic development largely through their trans-synaptic adhesions, but relatively little is known about how synaptic adhesions are structurally weakened and how this contributes to functional weakening of synapses. It has recently been shown that neuroligin-1 is cleaved in an activity-dependent manner through mechanisms requiring the activation of metalloproteinases MMP-9 (matrix metalloproteinase 9) and ADAM-10 (a disintegrin and metalloproteinase 10); this cleavage leads to both structural and functional weakening of synapses [30,31]. However, it remains unclear whether other synaptic adhesion molecules are similarly regulated. It is also not certain whether long-term depression (LTD), which is accompanied by the shrinkage of dendritic spines, loss of F-actin in spines and separation of pre-and postsynaptic structures [32][33][34], leads to proteolytic cleavage of synaptic adhesion molecules.
In this study, we tested whether NGL-3, used as a model adhesion molecule, undergoes a series of proteolytic cleavages in a neuronal activity-dependent manner. We found that LTDinducing chemical and electrical stimuli caused proteolytic cleavage of NGL-3 in a manner requiring the activation of NMDA receptors, MMPs and g-secretase. These results suggest that NGL-3 cleavage may regulate synaptic structure and function during LTD.

(a) NMDA treatment of cultured neurons induces NGL-3 cleavage
To test whether LTD induction leads to the cleavage of NGL-3, we first treated cultured hippocampal neurons with NMDA (20 mM, 3 min), which is known to induce chemical LTD in brain slices [100] and a long-lasting decrease in surface levels of AMPA receptors in cultured neurons [101,102]. Immunoblot analyses showed that NMDA treatment of cultured neurons increased the levels of an NGL-3 fragment (approx. 22 kDa) recognized by an NGL-3 antibody raised against the cytoplasmic region of NGL-3 (amino acids 622-690; #1948) (figure 1a). Three other antibodies raised against the cytoplasmic region of NGL-3 recognized the same 22 kDa band (see electronic supplementary material, figure S1), suggesting the authenticity of the band as NGL-3 C-terminal fragments (hereafter termed NGL-3-CTFs). The small levels of NGL-3-CTFs before NMDA treatment suggest that NGL-3 cleavage occurs under basal conditions. A faint band beneath the NGL-3-CTF band is not likely to be an alternative or subsequent cleavage product, because it is not recognized by other NGL-3 antibodies (see electronic supplementary material, figure S1).
(c) Long-term depression-inducing low-frequency stimulation causes NGL-3 cleavage in brain slices We next tested whether LTD-inducing electrical stimulation in brain slices causes NGL-3 cleavage. To accomplish this, we stimulated the Schaffer collateral pathway in mouse hippocampal slices (three weeks) by low-frequency stimulation (LFS, 1 Hz, 900 pulses), a stimulation paradigm known to cause NMDA receptor-dependent LTD in rat and mouse hippocampal slices [109,110], followed by confirmation of LTD induction by electrophysiological measurements, and immunoblotting analysis of hippocampal lysates. We found that LFS-LTD induced a significant increase in NGL-3 cleavage (       (e) NGL-3 cleavage requires g-secretase activity MMP cleavage of surface membrane proteins is usually followed by subsequent cleavage of their CTFs by g-secretase [82]. To test whether this also occurs for NGL-3, we treated  We obtained a similar result by using L-685,458 (1 mM), another g-secretase inhibitor (figure 3c). These results suggest that NGL-3-CTFs can be further cleaved by g-secretase.

Discussion
Our data indicate that LTD-inducing stimuli promote proteolytic cleavage of NGL-3 in a manner that requires the activation of NMDA receptors, MMPs and presenilin/g-secretase. These results suggest that (i) NGL-3 is a novel substrate of MMPs and g-secretase, (ii) LTD induction promotes NGL-3 processing and (iii) NGL-3 may regulate excitatory synapse structure and function during LTD. What might be the consequences of NGL-3 cleavage by MMPs and g-secretase? One straightforward possibility would be the weakening of NGL-3-mediated trans-synaptic adhesion at excitatory synapses that undergo LTD ( figure 4). In addition, g-secretase-mediated cleavage of NGL-3-CTFs, the step following MMP cleavage, would further remove the C-terminal tail from NGL-3, destabilizing the interaction between NGL-3 and PSD-95 (figure 4).
NGL-3 trans-synaptically interacts with LAR family receptor protein tyrosine phosphatases (LAR, PTPd and PTPs) [20,21]. Therefore, NGL-3 cleavage may cause the removal of LARs from synapses, similar to the removal of presynaptic neurexins observed at the site of neuroligin-1 cleavage [30]. Functionally, LAR is a well-known regulator of presynaptic development and function [41,[114][115][116]. Synaptic removal of LAR induced by NGL-3 cleavage is thus likely to exert a significant influence on presynaptic structure and function. Notably, LAR can be processed by ADAM-17/TACE (an a-secretase) and g-secretase [97,117,118]. It is conceivable that NGL-3 and LAR may undergo simultaneous proteolytic cleavages by specific MMPs/ADAMs during LTD, although we could not test LTD-induced LAR cleavage in this study owing to a lack of suitable LAR antibodies.
Our results are reminiscent of the recently reported activitydependent cleavage of neuroligin-1 mediated by MMP-9/ ADAM-10 and g-secretase [30,31]. Neurexin-1b and neurexin-3b, which interact with neuroligins, have also been shown to be substrates of g-secretase [98,99]. The complex of neuroligins and neurexins has been clearly demonstrated to regulate diverse aspects of synapse development and function [2][3][4]13]. Therefore, these previous observations taken together with the results of this study support the notion that the cleavage of synaptic adhesion molecules by MMPs/ADAMs and g-secretase regulates diverse aspects of synapse structure and function. One rather unique finding of the current study is that LTD-inducing low-frequency stimulation in slices leads to NGL-3 cleavage. It thus may be worth testing whether neuroligin-1 undergoes a similar LTD-induced cleavage and whether neuroligin-1 and NGL-3 act in a synergistic manner during LTD.
g-Secretase action is usually preceded by the MMP-mediated cleavage of membrane proteins into N-and C-terminal fragments (NTFs and CTFs) [82][83][84][85]. CTFs are further processed by g-secretase to generate intracellular domains (ICDs), which are known to regulate intracellular signalling in the cytoplasm or nucleus by, for instance, interacting with transcriptional regulators. Alternatively, ICDs are degraded by the proteasome for protein turnover. In this study, we could not observe detectable levels of NGL-3-ICDs. This suggests that NGL-3-ICDs may be labile and degraded by the proteasome, similar to the case of ICDs from neuroligin-1 [31]. However, this does not exclude the possibility that NGL-3-ICDs have some cytoplasmic functions. Notably, ICDs derived from NGL-3-interacting LAR translocate to the nucleus and regulate b-catenin-dependent gene expression [97].
A member of the LAR family, PTPd (encoded by the PTPRD gene), has been associated with attention deficit/ hyperactivity disorder (ADHD) [137] and restless leg syndrome [138], a disorder often comorbid with ADHD [139], autism spectrum disorder [140] and bipolar disorder [141]. This suggests the possibility that abnormalities in the transsynaptic interactions between LAR family proteins and their postsynaptic partners including NGL-3 may contribute to the development of these disorders.
In summary, our data suggest that induction of LTD in neurons leads to proteolytic cleavage of NGL-3 in a manner requiring the activation of NMDA receptors, MMPs and g-secretase. This cleavage may lead to the weakening of NGL-3-dependent trans-synaptic adhesion at excitatory synapses and contribute to structural and functional weakening of excitatory synapses during LTD.

Material and methods (a) Antibodies and animals
Guinea pig polyclonal NGL-3 antibodies (#2020 and #2021) were raised in this study using synthetic peptides mimicking the last 30 amino acid residues of NGL-3 (CGAKGPGLNSIHEPLLFKSCGS KENVQETQI). Rabbit polyclonal pan-NGL (#1583; against last 15 amino acid residues of NGL-2; CIIQTHTKDKVQETQI) [17] and rabbit polyclonal NGL-3 antibodies (#1948; against GST-NGL-3 amino acids 622-690) [21] have been described. a-Tubulin antibody was purchased from Sigma. Experiments on animals were performed in accordance with the guidelines of the Animal Welfare Committee of KAIST, Korea.

(b) Quantification of immunoblot results
NGL-3-CTF bands were quantified by normalizing the integrated intensities of the 22 kDa bands to those of tubulin bands, and comparing these normalized values from treated cultured neurons or stimulated brain slices with those from untreated control neurons or slices.
Funding statement. This study was supported by the Institute for Basic Science (IBS).