The M1 muscarinic acetylcholine receptor regulates the surface expression of the AMPA receptor subunit GluA2 via PICK1

Muscarinic acetylcholine receptors (mAChRs) have been shown to play significant roles in the regulation of normal cognitive processes in the hippocampus, and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) are also involved in these processes. This study aims to explore the mAChR-mediated regulation of AMPARs GluA2 trafficking and to reveal the key proteins and the signaling cascade involved in this process. Primary hippocampal neurons, as cell models, were treated with agonist 77-LH-28-1 and antagonist VU0255035, Fsc231, and APV. C57BL/6J male mice were stereotactically injected with 77-LH-28-1 and Fsc231 to obtain hippocampal slices. The trafficking of GluA2 was detected by surface biotinylation and immunostaining. Activation of M1 mAChRs promoted endocytosis and decreased the postsynaptic localization of the AMPA receptor subunit GluA2 and that phosphorylation of GluA2 at Ser880 was increased by M1 mAChR activity. Fsc231 blocked the endocytosis and postsynaptic localization of GluA2 induced by 77-LH-28-1 without affecting the phosphorylation of Ser880. PICK1 was required for M1 mAChR-mediated GluA2 endocytosis and downstream of phosphorylation of GluA2-Ser880, and the PICK1-GluA2 interaction was essential for M1 mAChR-mediated postsynaptic expression of GluA2. Taken together, our results show a functional correlation of M1 mAChRs with GluA2 and the role of PICK1 in their interplay. The schematic diagram for the modulation of GluA2 trafficking by M1 mAChRs. Activation of M1 mAChRs induces PKC activation, and the interaction of PICK1-GluA2 determines the endocytosis and postsynaptic localization of GluA2 The schematic diagram for the modulation of GluA2 trafficking by M1 mAChRs. Activation of M1 mAChRs induces PKC activation, and the interaction of PICK1-GluA2 determines the endocytosis and postsynaptic localization of GluA2

Introduction α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) are glutamate-gated ion channels, and the changes in their properties and postsynaptic abundance underlie various forms of synaptic plasticity, including long-term potentiation (LTP), long-term depression (LTD), and homeostatic scaling (Diering and Huganir 2018). Four highly homologous subunits, GluA1-4, combine into homologous dimers and form tetramers, which are functional AMPARs with different subunit combinations. GluA1-3 is mainly expressed in the hippocampus, and tetramers composed of GluA1/2 are the most important AMPA receptor subtypes in hippocampal neurons (Lu, et al. 2009;Traynelis, et al. 2010;Purkey, et al. 2020). GluA2 determines the biophysical properties of natural receptors; notably, receptors lacking GluA2 are Ca2+ permeable, and their activation can induce multiple forms of plasticity (Plant, et al. 2006) that can alter the physiology of neurons and sometimes trigger neurological dysfunction and cell death (Liu and Zukin 2007). Studies have shown that the C-terminal domains in GluA2 are necessary and sufficient to drive NMDA receptor-dependent LTD but do not affect LTP (Zhou, et al. 2018). The intracellular C-terminal domain in GluA2 contains a PDZ (PSD-95/ Dlg/ZO-1)-binding sequence that interacts with proteins, particularly the PDZ proteins GRIP and ABP, to regulate synaptic anchoring of AMPARs (Matsuda, et al. 2000;Osten, et al. 2000). Protein interacting with C-kinase 1 (PICK1), which is expressed mostly abundant in the brain and has a well-characterized PDZ domain and a BAR (Bin/ amphiphysin/Rvs) domain (Wang, et al. 2003), interacts with GluA2 likewise in C-terminal domain. Studies have consistently demonstrated that some forms of LTP and 1 3 multiple distinct forms of LTD were impaired in PICK1knockout mice (Volk, et al. 2010).
There is a wealth of literature showing that muscarinic acetylcholine receptors (mAChRs) are involved in the regulation of normal cognitive processes and play important roles in learning, memory, emotion, and other higher cognitive activities (Bubser, et al. 2012;Scarr 2012). Indeed, the role for muscarinic receptors in mood disorders came from the tricyclic antidepressant drugs which showed appreciable occupancy of these muscarinic receptors (Siafis and Papazisis 2018). In addition, the adrenergic-cholinergic balance hypothesis of affective disorders (Janowsky 2011) brought our insights into the role of cholinergic receptors in depressive disorder. In recent years, the newly approved antidepressant (S)-ketamine, underscoring the priority of glutamate system in depression, required the activation of AMPA receptors to play their roles (Freudenberg, et al. 2015). Besides, synaptic plasticity and other forms of neuroplasticity have been shown to be disrupted in depression (Pittenger and Duman 2008), and the observed reduction in AMPA receptor subunits expression possibly explained the impairments (Player, et al. 2013). Currently, the role of cholinergic (Witkin, et al. 2020) and AMPA receptors, as well as each subunit in depression, is not fully understood, and studies have shown that scopolamine's antidepressant-like effects depend on AMPA receptors and mTOR signaling (Martin, et al. 2017).
In this study, we investigated the trafficking of GluA2 upon M1 mAChR activation and possible mechanisms involved in this process. We discovered that a selective, bioavailable, and brain-penetrant agonist of M1 mAChRs, 77-LH-28-1 (Langmead, et al. 2008), promoted the endocytosis of GluA2 on the neuronal surface and decreased its postsynaptic colocalization with postsynaptic density-95 (PSD-95), which required the participation of PICK1, but inhibition of PDZ domain in PICK1 exerted no influence on the phosphorylation of GluA2 at Ser880.

Hippocampal primary culture
Cultured hippocampal neurons were obtained from newborn Sprague-Dawley rats within 24 h of birth, as previously described (Kam, et al. 2010). The digested hippocampal tissue was centrifuged, suspended, and plated on poly-L-lysine-treated six-well plates at 1.5×10 6 cells/well. For immunofluorescence experiments, 1×10 6 cells were plated onto 35-mm confocal dishes treated with polylysine.
Neurons cultured for 18 to 24 days in vitro (DIV) were used for experiments.

Surface biotinylation
Fsc231, 77-LH-28-1, and VU0255035 were dissolved in DMSO, and APV was dissolved in water. No inhibitors and 77-LH-28-1 were added in vehicle controls. After different durations of 77-LH-28-1 treatment, primary hippocampal neurons were incubated with 1 mg/mL sulfo-NHS-LC-biotin (Thermo Scientific) to label surface proteins according to the manufacturer's instructions. Lysates were centrifuged, and 30 μL of supernatant was obtained to determine total GluA2 input, and the remaining lysate was incubated with streptavidin-agarose beads (Thermo Scientific) with rotation overnight at 4°C to detect surface GluA2 (Ennion, et al. 2000).

Stereotactic injection and drug treatment
Chloral hydrate and a stereotactic frame were used to anesthetize and immobilize, respectively, C57BL/6J male mice. The left ventricle was located 1.0 mm posterior to the anterior fontanelle, 0.5 mm to the left of the midline, and 2.6 mm ventral to the skull surface. Five microliters of 5 μM 77-LH-28-1 was injected slowly into the left ventricle 30 min before removing the hippocampus for tissue slice preparation. Five microliters of 50 μM Fsc231 (Merck Millipore, USA, #529531) was injected one hour before 77-LH-28-1 treatment. Control mice were injected with saline, not Fsc231. Each group had three biological and technical replicates.
The colocalization of GluA2 and PSD95 was analyzed quantitatively with the COLOC function in ImageJ (Version 1.53n, RRID:SCR_003070), which enables the comparison of pixel intensities of two designated channels for use in determining the Pearson correlation coefficient. The Pearson correlation coefficient is positively correlated with the degree of colocalization. A Pearson correlation coefficient >0.6 was considered to indicate strong colocalization. To calculate the Pearson correlation coefficient, green-labeled GluA2 channels (channel A) were identified within a defined region of interest (ROI). Then, the stereo pixel intensities of the ROIs with channel A and PSD95 (channel B, in red) were determined.

Statistical analysis
All data points in bar graphs represent the means± standard deviation (SD). A statistically significant difference between means is represented by *P<0.05. Normality and lognormality tests were used for all data before analysis, and Shapiro-Wilk test was used for small sample data. After passing normality test, subsequent analysis was conducted. We used one-way ANOVA with Dunnett's post hoc test to detect significant differences between three or more groups; these data are presented in Figs. 1 and 4A-D. Unpaired t tests were performed for immunostaining assays and for obtaining quantitative protein levels in tissue specimens. Statistical analyses were carried out by a specialized statistician, who was blinded to the experimental protocol, using GraphPad Prism 8 software (RRID: SCR_002798).

PICK1 is involved in the endocytosis of GluA2 mediated by M1 mAChR activation
Since there is no evidence showing trafficking of GluA2 after M1 mAChR activation, we first treated primary hippocampal neurons with 77-LH-28-1 for different durations. The amount of GluA2 on the surface of neurons was reduced to 58.12±22.01% and 42.34±1.95% of the control group at 15 min and 30 min, respectively, after 1 μM 77-LH-28-1 was administered (F (5,12) =0.76, P<0.001, Fig. 1A and B). In addition, 5 μM VU0255035, which has been shown to be an M1 mAChR antagonist (Weaver, et al. 2009), was used to block the endocytosis of GluA2 on the neuronal membrane (F (5,12) =0.90, P=0.73, Fig. 1E and F). PICK1 has previously been associated with downregulation of GluA2 expression on the cell surface under basal conditions (Jin, et al. 2006;Yao, et al. 2008) and during LTD in the cerebellum and hippocampus (Jin, et al. 2006;Steinberg, et al. 2006;Hu, et al. 2007;Terashima, et al. 2008). Nevertheless, one study showed no PICK1 dependence in hippocampal LTD (Daw, et al. 2000). To determine whether PICK1 is involved in the endocytosis of GluA2 in response to M1 mAChR activation, we used the small-molecule inhibitor Fsc231 (Thorsen, et al. 2010), which binds to the PDZ domain in PICK1 with an affinity similar to that observed for endogenous peptide ligands. As shown in Fig. 1C and D, 50 μM Fsc231 pretreatment blocked the endocytosis of GluA2 induced by 77-LH-28-1 (F (5,12) =1.25, P=0.60).
In addition, it has been reported that, as the physiological basis of synaptic plasticity in the hippocampus, NMDA receptor (NMDAR)-dependent LTD is characterized by a reduction in synaptic strength due to the loss of surface and synaptic GluA2 (Parkinson, et al. 2018). To determine whether NMDAR is involved in the endocytosis of GluA2 triggered by mAChRs, 100 μM APV was preincubated with samples, and the results showed that it did not weakened the effect of 77-LH-28-1 (F (5,12) =2.13, P<0.005, Fig. 1G and H).
Finally, to determine the effect of GluA2 trafficking after M1 mAChR activation, we examined the colocalization of GluA2 and PSD95 in C57BL/6J male mice after stereotactic injection of agonists and inhibitors into lateral ventricles.  A, B) was blocked by the PICK1 inhibitor Fsc231 (C, D). n = 3 independent experiments. E, F Pretreatment with VU0255035 blocked the 77-LH-28-1-induced endocytosis of GluA2. n = 3 independent experiments. G, H Pretreatment with APV, an inhibitor of NMDAR, did not influence the endocytosis of GluA2 by 77-LH-28-1. n = 3 independent experiments. *P < 0.05, **P < 0.01, vs. 0 min (control), post hoc Dunnett's test. The data are shown as the means ± SD As shown in Fig. 3A and B, with mAChR activation, the Pearson correlation coefficient of GluA2 and PSD95 was obviously decreased, to 0.14 compared to that of the control group, which was 0.36 ( Fig. 3A and B, t(4)=5.22, **P<0.01, by unpaired t test). However, a significant decrease in colocalization in hippocampi of Fsc231-injected mice was not evident, as indicated by Pearson correlation coefficients between 0.18 and 0.20 ( Fig. 3A and B, t(4)=0.34, P=0.75).
Thus, the failure of 30 min 77-LH-28-1 treatment to profoundly decrease total and surface postsynaptic GluA2/PSD-95 colocalization in neurons and hippocampi in response to Fsc231 corresponded with the changes in GluA2 protein levels on the cell membrane surface in response to this same treatment, as described above.

Fsc231 preincubation abrogated the M1 mAChR-mediated endocytosis of GluA2 independent of phosphorylation at Ser880
As shown above, Fsc231 abrogated M1 mAChR-mediated GluA2 endocytosis and postsynaptic localization, and previous studies have shown that phosphorylation at Ser880 of GluA2 is pivotal for GluA2 endocytosis during LTD (Seidenman, et al. 2003;Steinberg, et al. 2006); therefore, we further investigated whether Ser880 is involved in the M1 mAChR effects. First, we investigated how 77-LH-28-1 affected the phosphorylation of Ser880 in cultured primary neurons. One micromolar of 77-LH-28-1 significantly increased the phosphorylation at Ser880 in GluA2, to 149.30 ± 13.74% of the control, after 30  Fig. 4A and B). However, without blocking the effect of 77-LH-28-1 with 50 μM Fsc231, the phosphorylation of GluA2 at Ser880 remained 134.40 ± 12.91% and 146.00 ± 8.90% of the control group at 15 min and 30 min, respectively (F (5,12) =0.74,**P<0.01, Fig. 4C and D). Moreover, we explored whether the phosphorylation of GluA2 at Ser880 varied in C57BL/6J mice with different drug treatments. Drugs were injected into the lateral ventricle of mice through a microsyringe and maintained for different durations. Fsc231 was treated one hour before 77-LH-28-1 treatment, which was maintained for 30 min in the lateral ventricle. Then, the mice were sacrificed to obtain the hippocampi for protein detection. Immunoblotting with antibodies recognizing phosphorylated Ser880 of GluA2 showed that the phosphorylation level in the 77-LH-28-1 treatment group was greater, at a 2.29±0.22-fold-change normalized to that of the vehicle (t(4)=9.98, ***P<0.001, Fig. 4E and F). Furthermore, hippocampi of Fsc231-treated mice had previously exhibited a significant increase (3.05±0.46-fold-change normalized to that of the vehicle, **P<0.01 by unpaired t test) in basal Ser880 phosphorylation and showed a further increase (t(4)=8.04, **P<0.01 by unpaired t test) after 30 min of 77-LH-28-1 treatment. In addition, we detected the amount of PSD95 in the hippocampi of mice after 77-LH-28-1 and Fsc231 administration. As shown in Fig. 4, no change in PSD95 expression was found in the hippocampi after 30 min of 77-LH-28-1 treatment, and the amount of PSD95 after treatment with Fsc231 had no statistical significance (t(4)=0.42, P=0.70 between vehicle and 77-LH-28-1, t(4)=1.33, P=0.25 between Fsc231 and Fsc231+77-LH-28-1, Fig. 4G and H, by unpaired t test) after 77-LH-28-1 application. Fig. 3 PICK1 was involved in the decrease of GluA2 and PSD95 colocalization upon M1 mAChR activation in hippocampi. Slices of C57BL/6J mouse hippocampi were immunostained for GluA2 and PSD95. A Representative images displaying total GluA2 (green) and PSD-95 (red) colocalization (yellow) in hippocampi after 30 min of 77-LH-28-1 treatment under control conditions or pretreatment with Fsc231 for 1 h. Scale bars, 5 μm, n=3 mice. B Quantification of Pearson's correlation (R) base on the images shown in A.

Discussion
Our findings confirmed that direct activation of M1 mAChR can promote endocytosis of the AMPA receptor GluA2 subunit on the cell surface and that PICK1 was involved in this process. Consistently, GluA2 postsynaptic localization was decreased, and PICK1 was involved in the regulation of GluA2 trafficking when M1 mAChRs were activated. We also reported that Fsc231 preincubation abrogated M1 mAChR-mediated GluA2 endocytosis without affecting phosphorylation of Ser880. This evidence leads to the conclusion that M1 mAChR modulates the trafficking of the AMPAR GluA2 subunit and the phosphorylation of Ser880 and that not only the phosphorylation of Ser880 but also PICK1-GluA2 interaction is required in GluA2 trafficking.  A, B) triggered by 77-LH-28-1 treatment was not influenced by the PICK1 inhibitor Fsc231 (C, D), n = 3 independent experiments. E, F 77-LH-28-1 increased the phosphorylation of GluA2 at Ser880 in the hippocampus in vehicle control and Fsc231-treated mice, n = 3 independent experiments. G, H 77-LH-28-1 had no effect on the amount of PSD95 in the hippocampus in vehicle control and Fsc231-treated mice. n = 3 independent experiments. The data are presented as the means ± SD mAChR hypersensitivity is hypothesized to mediate depressiogenic effects of overabundant acetylcholine via as yet undiscovered pathways, probably involving AMPAR and neuroplasticity (Hasselmann 2014). Acetylcholine has previously been reported to increase GluA2 insertion into the membrane attributed to M1 mAChRs (Fernández de Sevilla, et al. 2008). Nevertheless, our results revealed that M1 mAChR activation induced by 77-LH-28-1 promoted GluA2 endocytosis, thereby reducing GluA2 levels on the cell surface. Moreover, M1 mAChR has been considered to be critical for this effect because GluA2 endocytosis is blocked by VU0255035. Meanwhile, the non-specific muscarinic agonist acetylcholine, not unexpectedly, acted similarly to 77-LH-28-1 (Supplementary materials). In fact, LTD induced by M1 mAChR has been found, but it was NMDARdependent (Scheiderer, et al. 2006). In our study, the addition of APV failed to block 77-LH-28-1-induced GluA2 internalization, which indicated modulation of GluA2 trafficking by M1 mAChR independent of NMDAR.
Studies have reported that the GRIP/ABP interaction, but not the PICK1/GluA2 interaction, can profoundly regulate synaptic transmission and is necessary for hippocampal LTD (Daw, et al. 2000). However, in the present study, the PICK1-GluA2 interaction was required for M1 mAChR-induced endocytosis and the postsynaptic localization of GluA2, consistent with the previous studies on hippocampal LTD (Kim, et al. 2001;Seidenman, et al. 2003). Some findings help to explain how the PDZ domain in PICK1 binds putative partners, including PKCα and GluA2 (Sheng and Sala 2001;Cheng, et al. 2009), and the inhibitor Fsc231, which we used in the present study, bound the PICK1 PDZ domain with an affinity similar to that observed for endogenous peptide ligands (Ki ~10.1 μM) (Thorsen, et al. 2010). Our data showed that the phosphorylation of GluA2 at Ser880 was not affected by pretreatment with Fsc231 in cultured hippocampal neurons and was significantly increased compared to the control when M1 mAChR was activated in male mice, but GluA2 was no longer endocytosed. Even when the PDZ domain is occupied, PICK1 is thought to bring PKCα close to the tail of GluA2 (Perez, et al. 2001) due to its homomultimer form, thus promoting phosphorylation of GluA2 at Ser880. Notably, endocytosis of GluA2-containing AMPARs from synapses owing to the phosphorylation of Ser880 has emerged as an important contributor to the pathogenesis of neuropathic pain and pain-related depressive-like symptoms in a recently published study (Jiang, et al. 2021). Moreover, specific inhibitors of PDZ interaction have been successfully used in the mouse FST and TST (Doucet, et al. 2013), suggesting a promising approach to target the glutamatergic synapse in the treatment of depression. Our studies revealed that Fsc231 possibly affected the normal binding of PKC to PICK1 without obstructing PKC acting on GluA2 but interfered with the internalization of phosphorylated GluA2.
The PICK1 and GluA2 interact downstream of M1 mAChR activation-induced PKC phosphorylation of GluA2.
Colocalization of GluA2 with PSD95 in the postsynaptic membrane suggests the possibility that it constitutes the core of functional AMPA receptors and mediates synaptic plasticity. Our study revealed that both total and surface colocalization of GluA2 and PSD95 was affected by 30 min of 77-LH-28-1, and pretreatment with Fsc231 blocked the reduction in the postsynaptic membrane localization of GluA2 with PSD95. The decrease in the number of yellow dots on the neuronal surface was a manifestation of GluA2 internalization mediated by M1 mAChR, and the reduction in total GluA2 colocalization may be the best evidence of its departure from the postsynaptic density. Immunofluorescence of hippocampal slices confirmed the results showing GluA2 removal from PSD-like assemblies upon M1 receptor activation (Zeng, et al. 2018); these PSDlike assemblies contain multivalent interaction networks formed by major excitatory postsynaptic density (PSD) scaffold proteins via phase separation. The Pearson's correlation coefficient of GluA2 on the cell surface with PSD95 was lower than that of total GluA2 with PSD95, which might have been a result of the overlapping relationship between intracellular PSD95 and surface GluA2 and may not reflect true colocalization. Moreover, in our study, the phosphorylation level of Ser880 in the Fsc231-treated hippocampi was higher than that in the vehicle-treated hippocampi, and the colocalization level of GluA2 and PSD95 in the Fsc231-treated hippocampal slices was lower than that in the vehicle-treated hippocampal slices. Since endocytic sorting of AMPARs through distinct recycling and retention pathways existed in response to differential activation of postsynaptic glutamate receptors (Ehlers 2000), Fsc231 was reported to accelerate GluA2 recycling after NMDA receptor-induced internalization (Thorsen, et al. 2010). Furthermore, PICK1 was proved to reduce the recycling rates of interaction partners sorted to Rab11dependent slow recycling pathway (Madsen, et al. 2012). Therefore, we speculated that Fsc231 interfered with GluA2 sorted to recycling endosomes, and this sorting and fusion with endosomes could lead to the dephosphorylation of Ser880. Meanwhile, Fsc231 promoted GluA2 recycling without significantly affecting the amplitude of internalization (Thorsen, et al. 2010); in such a scenario, the lower colocalization level of GluA2 and PSD95 indicated that GluA2 reinserted into membrane might not be immediately recruited to the postsynaptic dense area.
Without doubt, there are some limitations to our study. The unique and specific M1 mAChR agonist is not sufficient to reflect all possibilities. The use of antagonist Fsc231 mainly reflects the importance of interaction between GluA2 and the PDZ protein PICK1, and more studies are needed on other AMPAR interacting proteins.