1 INTRODUCTION

Synapses are critical for neuronal communication as well as learning and memory. Synaptogenesis in the brain is highly organized and orchestrated by a wide variety of synaptic cellular adhesion molecules (CAMs) that contribute to the formation, stabilization, and diverse structure of synapses (Moreland & Poulain, 2022). These synaptic CAMs include cadherins, neuroligins, neurexins, and amyloid precursor protein (APP), the precursor for the Alzheimer's disease (AD)-related peptide amyloid beta (Aβ).

Classical cadherins play important roles in modulating synaptic specificity and remodeling. N-cadherin (also known as cadherin-2 or Cdh2) is a type I classical cadherin that is expressed pre- and post-synaptically in mature glutamatergic synapses (Basu et al., 2015; Stan et al., 2010). Trans-synaptic adhesion occurs because of homophilic interactions of N-cadherin across the synaptic cleft via five extracellular calcium-binding cadherin repeats, while the cytoplasmic region of N-cadherin is primarily responsible for actin association and downstream signaling by binding to catenins (Seong et al., 2015). Although N-cadherin is known to play an integral role in the construction of synapses, its function in dismantling synapses is not understood. Studies have shown that forced asymmetrical expression of N-cadherin (loss on either pre- or post-synaptic side only) results in defects in synaptic transmission, reduced synapse formation, and increased synapse elimination (Jüngling et al., 2006; Pielarski et al., 2013). These results suggest that the removal of N-cadherin from one side of the synapse could be a physiological mechanism to trigger synapse destabilization and loss.

In addition to synaptogenesis, N-cadherin has been shown to be crucial for synaptic plasticity and is required for long-term potentiation (LTP) (Bozdagi et al., 2000). Moreover, dendritic spines undergo actin-dependent changes in morphology during LTP (Harris, 2020; Hruska et al., 2018), and mutating alpha-catenin, which is required for cadherin's actin-binding, leads to severe impairments in dendritic spine morphology (Togashi et al., 2002). Collectively, these data suggest that cadherins are important for both synaptic development and strengthening, but the role of N-cadherin in synaptic weakening is less clear.

Synapse weakening and loss are prominent features of neurodegenerative diseases, and genome-wide association studies have suggested that synaptic CAMs are impaired in AD (Bao et al., 2015). N-cadherin is widely expressed in the hippocampus, a structure which is important in learning and memory and perturbed in AD. Previous studies have shown a physical association of N-cadherin and APP using co-immunoprecipitation from transfected HEK293 cells and mouse brain lysates (Asada-Utsugi et al., 2011). N-cadherin also promotes the localization of γ-secretase to the cell surface, increasing the probability of encounter with APP, and production of Aβ (Uemura et al., 2008). This enhancement can be blocked in the presence of an N-cadherin antagonist, ADH-1 (Asada-Utsugi et al., 2011). Additionally, Aβ42 synthetic peptides added to primary neuronal cultures result in p38 MAPK-mediated cell death, together with reduced N-cadherin expression, indicating that N-cadherin is downstream of Aβ (Ando et al., 2011). Collectively, these studies suggest that N-cadherin forms a complex with APP and engages in reciprocal regulation with Aβ.

Interestingly, N-cadherin and APP are proteolytically processed by a similar set of enzymes. N-cadherin is initially cleaved by the alpha-secretase ADAM 10, which results in the formation of an N-terminal fragment (NTF) and a C-terminal fragment 1 (CTF-1) (Reiss et al., 2005). CTF-1 is further cleaved by γ-secretase component presenilin (PS1) to form C-terminal fragment 2 (CTF-2), which binds to transcriptional coactivator CREB-binding protein (CBP) and promotes its proteasomal degradation, thus resulting in reduced CREB activity (Marambaud et al., 2003). These similar processing pathways further suggest that APP and N-cadherin may have some functional overlap.

Previously, we have shown that APP is a target for beta secretase processing during homeostatic synaptic plasticity, a class of compensatory mechanisms by which neurons adapt to and counteract excessive changes in network activity. Polo-like kinase 2 (Plk2) is an essential regulator of homeostatic plasticity; its expression is stimulated by neuronal hyperexcitation and functions to down-regulate excitatory synapses (Lee et al., 2011; Seeburg et al., 2008; Seeburg & Sheng, 2008). AD is associated with neuronal hyperactivity (Zott et al., 2018), which also stimulates Aβ production (Cirrito et al., 2008). Plk2 links these events by directly phosphorylating APP during hyperexcitation, resulting in APP cleavage and the generation of Aβ (Lee et al., 2017). This pathway may normally help to restrain excitability through the reduction of AMPARs and synapse loss (Hsieh et al., 2006). We have further shown that Plk2 inhibitors applied during hyperexcitation result in reduced Aβ generation and elevated surface expression of AMPARs (Lee et al., 2019).

Here, we examined the functional relationship between Plk2 and N-cadherin. We found that Plk2 directly phosphorylates N-cadherin, and during neuronal hyperexcitation Plk2 promotes N-cadherin proteolytic processing, degradation, and disruption of complexes with APP. Loss of N-cadherin adhesive function destabilizes excitatory synapses and promotes their structural dismantling as a prerequisite to eventual synapse elimination.

2 MATERIALS AND METHODS

3 RESULTS

We hypothesized that hyperexcitation-induced Plk2 causes down-regulation of synapses by phosphorylating synaptic structural components, leading to their degradation and ultimately synapse destabilization and elimination. To test whether synaptic CAMs are potential targets for Plk2, we performed a kinase-dependent molecular degradation screen of a panel of candidate synaptic CAMs. COS-7 cells were co-transfected with either myc-Plk2 constitutively active (Plk2-CA, T236E) or myc-Plk2 kinase-dead (Plk2-KD, K108M mutant), together with one of the following synaptic CAMs: β3-integrin, SynCAM1, tropomyosin receptor kinase B (TrkB), cadherin 9 (Cdh9), cadherin 8 (Cdh8), cadherin 10 (Cdh10), cadherin 13 (Cdh13), N-cadherin, Ephrin A4 (EphA4), and Ephrin B2 (EphB2) (Figure 1a). The results indicate that Plk2-CA resulted in loss of N-cadherin expression, with little effect on other synaptic CAMs (Figure 1b). Quantification indicated that N-cadherin levels were reduced by ~90% in the presence of Plk2-CA, as assayed by both immunoblotting and immunocytochemistry (Figure 1b,d). N-cadherin was not affected in the presence of an equivalent amount of Plk2-KD, suggesting that Plk2 promotes the elimination of N-cadherin in a kinase-dependent manner.

To test whether Plk2 directly phosphorylates N-cadherin, an in vitro kinase assay was performed using N-cadherin immunopurified from COS-7 cells and purified full-length Plk2 produced in baculovirus (Active Motif) in the presence of ³²P-γ-ATP. The autoradiogram showed that Plk2 indeed can phosphorylate N-cadherin, as well as exhibiting autophosphorylation (Figure 1c). These results demonstrate that N-cadherin is a bona fide and direct phosphorylation substrate of Plk2.

To determine whether Plk2 is capable of eliminating N-cadherin in neurons, we transfected Plk2-KD, Plk2-CA, and Plk2-WT into primary hippocampal neurons and measured the expression of endogenous N-cadherin. Neurons transfected with Plk2-CA exhibited ~60% loss of N-cadherin in secondary dendrites as well as in the soma, as compared to inactive Plk2-KD (Figure 2). Additionally, a significant decrease in N-cadherin somatic expression was observed in Plk2-WT neurons compared to Plk2-KD (Figure 2). No significant difference in Plk2 expression in the soma was observed among any of the constructs (Figure 2). These data further support that N-cadherin loss is kinase dependent and occurs in neuronal dendrites.

Next, we examined if N-cadherin loss occurs physiologically in neurons during hyperexcitation conditions that induce Plk2 expression. For this purpose, we used PTX to block inhibitory GABA_A receptors, which causes global stimulation of excitatory synaptic activity and triggers homeostatic adaptation. Under basal conditions, N-cadherin is distributed in a highly punctate pattern in dendrites. Upon chronic stimulation with PTX for 24 h, N-cadherin was substantially diminished in secondary dendrites and soma in comparison to unstimulated control (Figure 3). PTX had no significant effect on levels of the dendritic protein MAP2 (Figure S1), indicating the specificity of the N-cadherin reduction. To test if PTX-induced N-cadherin loss is mediated by Plk2, we used two different antagonists (pan Plk-family inhibitor BI-6727, or selective Plk2 inhibitor TCS-7005; 50 nM each) for 4 h prior to collection. Both Plk2 inhibitors blocked the removal of N-cadherin during hyperexcitation in secondary dendrites and soma (Figure 3). Thus, hyperexcitation-induced loss of native N-cadherin in neurons requires endogenous Plk2 activity.

Accumulating evidence points to a protein complex of N-cadherin and APP in stabilizing synapses. To examine whether hyperexcitation alters this complex association, we immunostained for N-cadherin and APP. We found that both synaptic CAMs were lost following PTX treatment, which was rescued by BI6727 or TCS-7005 (Figure 4). Additionally, we found that APP co-immunoprecipitated with N-cadherin from primary hippocampal neurons (Figure 5a) and transfected heterologous cells (Figure 5b). When neurons were treated with PTX for 24 h, we found that hyperexcitation abolished the N-cadherin and APP association (Figure 5a). BI-6727 or TC S-7005 inhibition of Plk2 restored the interaction of these CAMs, and indeed rescued to a level higher than baseline. This finding confirms that APP and N-cadherin form a physiological assembly and that during hyperexcitation Plk2 causes the dismantling of this complex.

We further analyzed the association of APP and N-cadherin at synapses using triple immunostaining of these proteins along with PSD-95 as a marker of excitatory synapses. We observed considerable overlap of the three proteins in dendritic spines (seen as white puncta because of triple co-localization) (Figure 5c,d); APP colocalization with N-cadherin decreased (albeit in a non-significant manner) with PTX, but Plk2 inhibitors restored the colocalization to a level significantly different from PTX alone (Figure 5c,d). N-cadherin, however, did not change its colocalization extent with APP under any conditions (Figure 5c,d), suggesting that N-cadherin association with APP is strictly maintained even under dynamically changing protein levels.

To examine the molecular mechanisms that underlie the activity- and Plk2-dependent loss of N-cadherin, we hypothesized that this pathway may involve proteolytic cleavage and degradation. N-cadherin is a substrate of multiple proteases including calpains, matrix metalloproteinases (MMPs), and the proteasome. Pharmacological inhibition was utilized to identify proteolytic pathways involved in N-cadherin loss in hyperexcitation. First, we tested GM6001 (a pan-MMP inhibitor) and GI254023X (a specific inhibitor of ADAM10) and found that these inhibitors rescued the loss of N-cadherin normally observed with chronic PTX (Figure 6). This observation suggests hyperexcitation induces the processing of N-cadherin through the up-regulation of ADAM10 (Figure 6).

Calpains, which are calcium-dependent cysteine proteases (Goll et al., 2003), are also up-regulated by increased intracellular calcium during hyperexcitation (Lynch & Baudry, 1984). Thus, we wanted to examine if calpains are involved in the proteolytic cleavage of N-cadherin. We tested two different concentrations of calpain I and calpain III inhibitors (2 and 50 μM.) We found that 25 μM of Calpain I inhibitor did not rescue N-cadherin loss during hyperexcitation; however, 25 μM of Calpain III inhibitor did prevent the proteolytic cleavage of N-cadherin. With 50 μM, both inhibitors rescued N-cadherin loss in comparison to PTX-treated control neurons (Figure 7). This result suggests that calpains are up-regulated during hyperexcitation and also play a role in the proteolytic cleavage of N-cadherin (Figure 7).

Next, we wanted to determine if these proteolytic cleavage products are ultimately degraded by the proteasome or lysosome. We utilized two pharmacological inhibitors of the proteasome (MG-132 and lactacystin), but neither inhibitor rescued the loss of N-cadherin induced by PTX (1 μM or 10 μM for 5 h), suggesting that N-cadherin is not degraded by the proteasome. We then tested two pharmacological inhibitors of the lysosome (10 nM bafilomycin and 40 μM chloroquine), and found that lysosome inhibition for 4 h resulted in rescue of N-cadherin during hyperexcitation (Figure 8). Collectively, these data suggest that hyperexcitation results in an up-regulation of Plk2, which signals proteolytic cleavage of N-cadherin by ADAM10 and calpains, followed by degradation of C-terminal cleavage products by the lysosome.

To understand the functional role of the loss of N-cadherin on the synapse, we examined the expression of the post-synaptic marker PSD-95 as a readout of synapse stability. Global stimulation of synaptic activity by PTX resulted in the loss of both N-cadherin and PSD-95 after 24 h (Figure S2). To exclude the possibility that the loss of N-cadherin is simply a secondary effect of synapse elimination, neurons were stimulated with PTX over a time course of 4, 8, and 16 h and immunostained for N-cadherin and PSD-95. The results indicate that at 4 h there is no change yet in the expression of N-cadherin and PSD-95 with PTX (Figure S3A). At the 8-h PTX timepoint, there was no significant change in PSD-95, however, unexpectedly, there was increased expression of N-cadherin (Figure S3B). After 16 h of PTX treatment, both PSD95 and N-cadherin were reduced in secondary dendrites and the soma, but to different degrees (Figure S3C). These results support the idea that N-cadherin levels are not simply changing as a result of altered synapse size or density.

Previous findings demonstrated functional and physical interactions among APP, N-cadherin, and GluA2 (Lee et al., 2010; Saglietti et al., 2007). Therefore, we wished to examine changes in AMPAR surface expression during hyperexcitation, and whether such regulation required the cleavage of N-cadherin and APP. During hyperexcitation, there is a reduction of surface GluA2 (sGluA2) receptors in comparison to unstimulated control. In the presence of beta secretase inhibitor (BSI) to block APP beta processing, sGluA2 levels were still reduced with PTX. However, PTX with the addition of GI254023X alone, or GI254023X in combination with BSI, did rescued sGluA2 levels (Figure 9), suggesting that N-cadherin cleavage may be important for the removal of surface AMPARs during homeostatic adaptation to hyperexcitation.

4 DISCUSSION

In this study, we examined molecular mechanisms contributing to the homeostatic loss of synapses that occurs as a compensatory response to chronic neuronal hyperexcitation. During overactivity of neurons, expression of the homeostatic regulator Plk2 is known to be induced, resulting in the phosphorylation of synaptic substrates and down-regulation of excitatory synapses. By screening several synaptic CAMs, here we identified N-cadherin as a novel and selective Plk2 target. Plk2 caused the dramatic loss of N-cadherin expression, both in heterologous cells and in hippocampal neurons. This effect was dependent on Plk2 kinase activity, and indeed, N-cadherin could be directly phosphorylated by Plk2 in a purified system. Moreover, hyperexcitation induced the removal of N-cadherin puncta from excitatory synapses in a manner requiring Plk2 kinase activity. Examination of MAP2 and previous studies of other synaptic proteins such as liprin (Lee et al., 2011) further underscored the selectivity of the effect on N-cadherin. Together, these findings indicate that this mechanism constitutes a physiological pathway of specific CAM disassembly during homeostatic synaptic plasticity.

Using various pharmacological inhibitors, we identified proteases responsible for the loss of N-cadherin in hyperexcitation. MMPs have been shown to cleave N-cadherin resulting in an NTF and CTF-1 that is further cleaved by PS1 and ultimately degraded (Marambaud et al., 2003; Reiss et al., 2005). This cleavage is activity dependent, as it is enhanced by bicuculline, a GABAA receptor inhibitor similar to PTX (Malinverno et al., 2010). Our findings confirmed that the selective ADAM10 inhibitor GI254023X rescued N-cadherin loss during hyperexcitation. ADAM10 is known to localize to excitatory synapses, and inhibition of ADAM10 cleavage of N-cadherin leads to increased dendritic spine size (Malinverno et al., 2010) as well as improvement in spatial working memory (Asada-Utsugi et al., 2021). Our data build upon this idea that proteolytic cleavage of N-cadherin results in reductions in synaptic efficacy, expanding such function to govern homeostatic weakening and loss of synapses in hyperexcitation.

In addition to MMPs, we found that calpains also contribute to N-cadherin degradation during neuronal overactivity. In the CNS, calpains are widely expressed and play roles in cell death, proliferation, and synaptic plasticity (Baudry et al., 2013), while several calpain substrates are important in neurodegeneration affecting Aβ production, tau phosphorylation, and disrupted neuronal structure (Atherton et al., 2014; Jin et al., 2015; Lynch & Baudry, 1987). Inhibiting calpain I and calpain II through gene knockout caused decreased dendritic spine number, reduced expression of NMDAR and AMPAR subunits, lowered levels of PSD-95, and impairment in LTPs and learning and memory (Amini et al., 2013). Furthermore, a reduction of calpains in vitro (Bednarski et al., 1995) or in vivo ameliorates cell death in glutamate- or NMDA-induced excitotoxicity (Amini et al., 2013). Conversely, activation of calpain results in the degradation of several synaptic scaffolding proteins such as GRIP-1 (Jourdi et al., 2005; Lu et al., 2000, 2001). GRIP-1 is responsible for trafficking of sGluA2 receptors to control synaptic scaling (Liang et al., 2017; Tan et al., 2015), and GRIP-1 also links N-cadherin to AMPARs (Heisler et al., 2014). Degradation of GRIP-1 in response to hyperexcitation results in the reduction of sGluA2 and internalization of receptors (Tan et al., 2015). Thus, both inhibition and activation of calpains appear to perturb synapse structure and function, and the precise balance of calpain levels may be needed to preserve N-cadherin integrity and synapse stability.

To address the functional role of N-cadherin on the synapse in response to hyperexcitation, we analyzed the distribution and levels of PSD-95 as a marker of excitatory synapses. PSD-95 is a scaffold protein that creates slots for stabilization of AMPARs (Malinow & Malenka, 2002; Newpher & Ehlers, 2008; Schnell et al., 2002), and as such, the amount of PSD-95 at synapses is an index of AMPAR content. During homeostatic synaptic scaling, hyperexcitation reduces synaptic markers such as PSD-95 and PSD-93 (Sun & Turrigiano, 2011). N-cadherin knockdown in vitro (Aiga et al., 2011) or knockout in vivo (Stan et al., 2010) also decreases PSD-95. Our results showed that a selective inhibitor of Plk2 rescues both PSD-95 and N-cadherin in response to hyperexcitation.

To test whether N-cadherin degradation led to loss of PSD-95, or whether loss of excitatory synapses led to the disappearance of N-cadherin as a secondary consequence, we performed a time course from 4 to 24 h. N-cadherin and PSD-95 generally displayed similar changes at all time points tested except at 8 h, at which there was no change in PSD-95 intensity but an unexpected increase in the expression of N-cadherin with PTX stimulation. These findings demonstrate that N-cadherin dynamics does not simply follow PSD-95 levels passively. However, after the 8 h time point Plk2 is known to be induced, which leads to a dramatic decrease in both N-cadherin and PSD-95. It may be that at early time points, a relatively brief period of heightened synaptic activity up-regulates N-cadherin expression to increase adhesive properties and maintain synapse number. In contrast, more prolonged overactivity prompts a more severe response necessitating synapse elimination.

Lastly, we and others have shown functional and physical interactions between N-cadherin, APP, and AMPARs (Okuda et al., 2007; Saglietti et al., 2007; Tai et al., 2008; Xie et al., 2008). Our colocalization data in particular argue that APP and N-cadherin are tightly coregulated during plasticity, maintaining relatively consistent colocalization at synapses while undergoing dramatic changes in synaptic protein levels. To examine the relative importance of the degradation of these synaptic CAMs to changes in surface AMPAR subunit GluA2, we blocked the cleavage of both N-cadherin and APP using a combination of ADAM10 and BSIs. During hyperexcitation, sGluA2 levels were reduced but BSI treatment did not block this AMPAR down-regulation. However, ADAM10 inhibitor rescued sGluA2 levels, and a combination of both drugs further increased sGluA2, consistent with a synergistic effect between N-cadherin and APP on determining AMPAR levels.

Collectively, these studies suggest that Plk2 initiates a signaling cascade beginning with N-cadherin and APP processing, which triggers downstream removal of PSD-95 that, in conjunction with other destabilizing events such as GRIP-1 degradation, culminates in the synaptic scaling of AMPARs. Further understanding the molecular underpinnings of the synapse provided by synaptic CAMs such as N-cadherin and APP and their remodeling in synaptic plasticity is essential for dissecting abnormalities in synapse function that underlie neurodegenerative diseases and other cognitive disorders.

AUTHOR CONTRIBUTIONS

Mai Abdel-Ghani designed the experiments, conducted the experiments, and wrote the manuscript. Yeunkum Lee designed and conducted experiments. Lyna Ait Akli, Marielena Moran, Amanda Schneeweis, Sarra Djemil, Rebecca El Choueiry, and Ruqaya Murtadha conducted experiments. Daniel T. S. Pak designed the experiments, wrote, and edited the manuscript.

FUNDING INFORMATION

This work was supported by NIH grant RF1 AG056603 (to DTSP).

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to declare.