Expression and Interaction Proteomics of GluA1- and GluA3-Subunit-Containing AMPARs Reveal Distinct Protein Composition

The AMPA glutamate receptor (AMPAR) is the major type of synaptic excitatory ionotropic receptor in the brain. AMPARs have four different subunits, GluA1–4 (each encoded by different genes, Gria1, Gria2, Gria3 and Gria4), that can form distinct tetrameric assemblies. The most abundant AMPAR subtypes in the hippocampus are GluA1/2 and GluA2/3 heterotetramers. Each subtype contributes differentially to mechanisms of synaptic plasticity, which may be in part caused by how these receptors are regulated by specific associated proteins. A broad range of AMPAR interacting proteins have been identified, including the well-studied transmembrane AMPA receptor regulatory proteins TARP-γ2 (also known as Stargazin) and TARP-γ8, Cornichon homolog 2 (CNIH-2) and many others. Several interactors were shown to affect biogenesis, AMPAR trafficking, and channel properties, alone or in distinct assemblies, and several revealed preferred binding to specific AMPAR subunits. To date, a systematic specific interactome analysis of the major GluA1/2 and GluA2/3 AMPAR subtypes separately is lacking. To reveal interactors belonging to specific AMPAR subcomplexes, we performed both expression and interaction proteomics on hippocampi of wildtype and Gria1- or Gria3 knock-out mice. Whereas GluA1/2 receptors co-purified TARP-γ8, synapse differentiation-induced protein 4 (SynDIG4, also known as Prrt1) and CNIH-2 with highest abundances, GluA2/3 receptors revealed strongest co-purification of CNIH-2, TARP-γ2, and Noelin1 (or Olfactomedin-1). Further analysis revealed that TARP-γ8-SynDIG4 interact directly and co-assemble into an AMPAR subcomplex especially at synaptic sites. Together, these data provide a framework for further functional analysis into AMPAR subtype specific pathways in health and disease.

Of note, Gria1 KO mice revealed strong reduction in the expression of α-synuclein. Loss of α-synuclein expression has been observed previously in a sub population of C57BL/6J mice without alteration of additional genes or a noticeable phenotype [42]. As reduced α-synuclein in the current study is likely due to cross breeding with this C57BL/6J strain, we removed this protein from further analysis.

Antibodies
Detailed information on the antibodies used is shown in the Supplemental Materials and Methods.

Preparation of Crude Synaptosomal Fractions
Biochemical fractions containing crude synaptosomes and microsomes (P2+M) were prepared as previously described [43] (Supplemental Materials and Methods).

Immuno-Purifications/in-Gel Digestion/Data-Dependent Acquisition Analysis
Proteins were extracted from P2+M using n-Dodecyl β-D-maltoside (DDM) (Thermo Fisher, Waltham, MA, USA) dissolved in sample suspension buffer (25 mM, 150 mM NaCl and protease inhibitor cocktail (Roche, Basel, Switzerland), pH 7.4), at a 1% endconcentration, two times for 1 h at 4 • C. Following each extraction, samples were centrifuged at 20,000× g for 20 min. Next, supernatant was incubated with 10 µg of antibody overnight at 4 • C, followed by incubation with 80 µL of protein A/G PLUS-Agarose beads (Santacruz, Dallas, TX, USA) for 1 h at 4 • C. Samples were centrifuged at 1000× g for 1 min, supernatant was discarded and beads were washed four times with 1 mL washing buffer containing 0.1% DDM, 150 mM NaCl (Sigma-Aldrich, St. Louis, MO, USA), 250 mM HEPES (Sigma-Aldrich, St. Louis, MO, USA), pH 7.4. SDS sample buffer was added to the final pellet, samples were heated at 98 • C and run on a home-made 10% SDS polyacrylamide gel. All reported n-numbers are biological replicates.
Gels were fixed overnight in 50% ethanol and 3% phosphoric acid (Sigma-Aldrich, St. Louis, MO, USA), washed in MilliQ water and stained with Colloidal Coomassie Blue. Each sample lane was cut in 3-5 slices that were subsequently cut into smaller pieces. The gel pieces were transferred to a Multiscreen HV filter Plate (Sigma-Aldrich, St. Louis, MO, USA), washed and destained with a mixture of 50 mM ammonium bicarbonate (Sigma-Aldrich, St. Louis, MO, USA) in acetonitrile (VWR, Radnor, PA, USA). The gel pieces were dried with 100% acetonitrile and incubated overnight at 37 • C with trypsin (Mass Spec Grade, Promega, Madison, WI, USA) dissolved in 50 mM ammoniumbicarbonate. Peptides were extracted twice in 0.1% Trifluoroacetic acid (Protein sequence grade; Applied Biosystems, Warrington, UK) and 50% acetonitrile, followed by extraction in 0.1% Trifluoroacetic acid and 80% acetonitrile. Subsequently the samples were dried in a speed vac (Savant, Thermo Fisher, Waltham, MA, USA) and stored at −20 • C until mass spectrometry analysis.
Peptides were analyzed on an LTQ-Orbitrap discovery (Thermo Fisher, Waltham, MA, USA) mass spectrometer as previously described [43], with some modifications (Supplemental Materials and Methods).

Depletion Immuno-Purifications
Depletion Immuno-purifications (IPs) were performed using a similar protocol as described for the regular IPs, with some modifications (Supplemental Materials and Methods).

Immuno-Purifications/Blue Native-PAGE/Data-Dependent Acquisition Analysis
IPs were performed using the protocol described above, now using 30 mg P2+M, 100 µL antibody and 1000 µL of beads. After purification and washing of the samples, purified protein complexes were eluted twice using 500 µg peptide dissolved in 1 mL washing buffer for 1 h. The samples were then concentrated using a 30 kDa filter (Bio-Rad, Hercules, CA, USA) for 30 min, and mixed with Blue Native (BN)-PAGE loading buffer (Thermo Fisher, Waltham, MA, USA), 0.5 µL molecular weight marker (Thermo Gels were fixed overnight in 50% ethanol, 3% phosphoric acid, washed in MilliQ water and stained with Colloidal Coomassie Blue. Each sample was cut into 70 slices using a grid cutter (Gel Company, San Francisco, CA, USA), and transferred to a Multiscreen HV filter Plate. Cysteines were derivatized using 1 mM tris(2-carboxyethyl)phosphine (TCEP) (Sigma-Aldrich, St. Louis, MO, USA) in 50 mM ammonium bicarbonate for 30 min at 37 • C and incubated with 4 mM methyl methanethiosulfonate (MMTS) (Fluka, Honeywell, Charlotte, NC, USA) in 50 mM ammonium bicarbonate for 15 min at room temperature. Next, samples were washed, destained, dried and digested following the in-gel digestion protocol described above. The samples were dried in a speed vac and stored at −20 • C before analysis on the mass spectrometer.
Each slice was analyzed separately on the Triple TOF 5600 (Sciex, Framingham, MA, USA) in data-dependent acquisition (DDA) mode as described previously [44], with some modifications (Supplemental Materials and Methods).
After 48 h, the HEK293 cells were washed with phosphate-buffered saline resuspended in extraction buffer (1% DDM, 25 mM HEPES, 150 mM NaCl, and protease inhibitor cocktail, pH 7.4), and incubated for 1 h at 4 • C. After two consecutive centrifugation steps at 20,000× g for 15 min. 4 • C, 4 ug of antibody was added to the supernatant, incubated overnight at 4 • C, followed by 1 h incubation with beads, 4 • C. The samples were washed four times with wash buffer (0.1% DDM, 25 mM HEPES, and 150 mM NaCl) in between centrifugation at 1000× g, 4 • C, and the purified proteins were eluted-with 2× sodium dodecyl sulfate (SDS) sample buffer. Input samples were prepared from the supernatant fraction by addition of SDS sample buffer to a 2× final concentration.

BN-PAGE/Immunoblot Analysis
BN-PAGE for immunoblot analysis was performed following the manufacturer's recommendations (Thermo Fisher, Waltham, MA, USA), with some modification (Supplemental Materials and Methods). Immunoblot analysis was conducted following the regular immunoblot protocol described in the Supplemental Materials and Methods.

Quantitative Proteomics by in-Gel Digestion/Data-Independent Acquisition
Wildtype, Gria1and Gria3 KO P2+M samples were run on a home-made 10% SDS polyacrylamide gel. Each sample was cut into small pieces, 100 µL of 50 mM ammoniumbicarbonate and 5 mM TCEP was added and incubated for 30 min, 37 • C. Next, 100 µL of 50 mM ammoniumbicarbonate and 2.5 mM MMTS was incubated for 15 min, room temperature. The proteins were digested using the in-gel digestion protocol described above. All reported n-numbers are biological replicates.

Primary Neuronal Culture
Detailed information on the preparation of dissociated hippocampal neuronal cultures is shown in Supplemental Materials and Methods.

Immunocytochemistry
Detailed information on immunolabeling of hippocampal neurons is shown in Supplemental Materials and Methods.

STED Microscopy and Analysis
Images were acquired on a TCS SP8 gated Stimulated Emission Depletion (STED) 3X Microscope (Leica, Wetzlar, Germany). Fluorophores were excited with a pulsed white light laser at their excitation peak, and a pulsed 775 nm STED laser was used for depletion in the 635 nm (TARP-γ8) and 580 nm (SynDIG4) channel obtaining a lateral resolution of 80 nm. Images in the 488 nm (Homer) channel were taken in confocal mode. Images were obtained with a 100× oil objective (NA = 1.4), a mechanical zoom of 5 and the pinhole set at 1 Airy Units (AU). Signals were detected with a gated hybrid detector (HyD) set in photon counting mode.
The Images were deconvolved with Huygens Software (Scientific Volume Imaging B.V., Hilversum, The Netherlands) using the Good's Roughness Maximum Likelihood Estimation (GMLE) algorithm and analyzed with ImageJ extended in the Fiji framework. Analysis was performed on the maximum projections of the z-stack, and a threshold determined by the default algorithm was applied on all channels. The Manders' coefficients were obtained in the coloc2 application.

Expression Proteomics on Gria1-and Gria3 KO Synapses Reveals Differential Expression of Known AMPAR Interactors
We first performed quantitative proteomics on hippocampal synapse enriched fractions of both Gria1and Gria3 KO mice and their wildtype controls (n = 5-6/condition) ( Figure 1). Per dataset, differential expression analysis (DEA) was performed using highquality peptides detected in at least 75% of the samples in each experimental condition. In addition, ambiguous peptides assigned to multiple protein groups were removed. Both Gria1and Gria3 KO datasets revealed similar numbers of peptides and proteins and Coefficient of Variation (CoV) per sample group ( Figure S1). In the Gria1 KO dataset, filtering left 15,954 peptides that mapped to 3051 unique proteins with a CoV of 12.6% and 12.2% in wildtype and Gria1 KO samples, respectively ( Figure S1a). In the Gria3 KO dataset, 15,867 peptides were retained that mapped to 3048 proteins, and revealed a CoV of 12.2% in wildtype and 14.8% Gria3 KO samples ( Figure S1b). In the Gria1 KO dataset, two unique GluA1 peptides were detected, albeit at a 97% lower expression compared to wildtype ( Figure S1c). Both peptides originated from the N-terminal domain. This is in agreement with a previous report demonstrating low expression of a truncated GluA1 Nterminal fragment in this Gria1 KO line [41]. Gria3 KO mice revealed no expression of GluA3 unique peptides.
Subsequently, selective regulation of GluA1, SynDIG4 and TARP-γ8 in Gria1 KOs was validated by immunoblotting ( Figure 1c). Quantification of Shisa6 revealed a trend of upregulation, without reaching statistical significance (fold-change of 1.22, p-value = 0.37) (Figure 1d, Table S2). Of interest, CNIH-2 was detected with one peptide in wildtypes and Gria3 KO mice. In Gria1 KO mice, this peptide failed the quality criteria for quantitative analysis, suggestive of a down-regulation, which was corroborated by immunoblotting (Figure 1c,d, Table S2). Together, these data revealed a specific subset of AMPA-receptor interactors with robust regulation in Gria1 KO specifically, suggesting selective binding of these interactors to GluA1-containing receptors.
Next, we performed AMPAR immuno-purifications (IPs) on the Gria1and Gria3 KO synapse enriched fraction, to assess the interactomes of GluA1/2 and GluA2/3 receptors in a direct manner.

Validation with Immunoblotting of GluA2/3 IP in the GluA1 Depleted Synapse Extract
To further validate the observations on preferential interactions, independently in wildtype animals, we performed AMPAR IPs on wildtype hippocampus after depletion of GluA1 containing receptors by IP, followed by immunoblotting (Figure 2b). Based on the Gria1 KO IP-MS data (Figure 2a), removal of GluA1-containing receptors is expected to cause major reduction in levels of SynDIG4 and TARP-γ8, whereas CNIH-2 and TARP-γ2 are expected to be less affected. After protein extraction from a synaptic fraction, GluA1containing receptors were removed by IP with 33 µg of GluA1 specific antibody in half of the lysates. After antibody incubation, all lysates were incubated two times with 200 µL A/G PLUS agarose beads, and subsequently used for AMPAR-purification with 10 µg anti-GluA2/3 per experiment. Indeed, anti-GluA2/3 revealed a lack of GluA1 immunoreactivity after GluA1-depletion (Figure 2b). In addition, immunoreactivity of SynDIG4 and TARP-γ8 were absent post depletion of GluA1, whereas immunoreactivity remained present for CNIH-2 and TARP-γ2 (Figure 2b). This suggests that SynDIG4 and TARP-γ8 are major interactors of GluA1-containing receptors, in contrast to the remaining GluA3-containing receptors. Taken together, these data demonstrate GluA1/2 containing receptors have a preferred interaction with CNIH-2, TARP-γ8 and SynDIG4 whereas GluA2/3 containing receptors strongly interact with CNIH-2, TARP-γ2 and Noelin1.

Validation with Immunoblotting of GluA2/3 IP in the GluA1 Depleted Synapse Extract
To further validate the observations on preferential interactions, independent wildtype animals, we performed AMPAR IPs on wildtype hippocampus after depl of GluA1 containing receptors by IP, followed by immunoblotting (Figure 2b). Base   Both TARP-γ8 [48] and SynDIG4 [17] are known to directly bind AMPAR subunits. Similarly, AMPAR interactors FRRS1L and CPT1c bind the AMPAR directly, in addition to binding each other [29]. To test if also TARP-γ8 and SynDIG4 can bind in absence of AMPAR subunits, we purified overexpressed TARP-γ8-myc from HEK293 cells in the presence of SynDIG4-HA (Figure 3b). Indeed, isolation of TARP-γ8-myc revealed co-assembly with SynDIG4 demonstrating these proteins directly interact (Figure 3b).

Combined IP-Blue Native Quantitative Proteomics Demonstrates the Presence of TARP-γ8 and SynDIG4 in an AMPAR Subcomplex
To further scrutinize this TARP-γ8-SynDIG4 assembly as a subcomplex of the AMPAR in the hippocampus, we investigated the migration of TARP-γ8 and SynDIG4 immunopurified native complexes on BN-PAGE followed by mass spectrometry (termed IP-BN-PAGE-MS), as described previously ( Figure S4) [44]. Following IP, native complexes were eluted with an epitope-mimicking peptide, mixed with marker proteins and separated by size on a BN-PAGE gel. The gel was cut into consecutive slices that were separately analyzed by mass spectrometry for protein identification and quantification ( Figure S4). Protein abundance values were normalized to their max intensity across the gel, and gel slices were numbered relative to the 720 kDa spiked-in marker protein.
In the gel, purified TARP-γ8 and SynDIG4 were expected to co-migrate together with GluA1 in the migration range of the AMPAR at~720 kDa and higher if they are indeed part of an AMPAR assembly. Figure 3c reveals the GluA1 and GluA2 immunoreactivity of the synaptic extract fractionated on BN gel followed by immunoblotting analysis (Figure 3c). IP-BN-PAGE-MS of anti-TARP-γ2/8 revealed highest abundance of TARP-γ8 between slice −14 till 1, peaking above the 720 kDa spiked in marker protein (slice −3) (Figure 3d). In the same range also SynDIG4 and GluA1 co-migrated, peaking at slightly higher (slice −5) or lower (slice −2) molecular weight, respectively, with large overlapping migration profiles (Figure 3d). Migration of TARP-γ8 bait protein below the 720 kDa marker may result from disassembly in the BN-gel or represent native AMPAR-independent complexes. Similarly, IP-BN-PAGE-MS of SynDIG4 revealed SynDIG4 migration across a broad range of molecular weights above and below 720 kDa; peak abundance above the 720 kDa marker (slice −4), and large overlapping migration profiles of both TARP-γ8 (peaking at slice −6) and GluA1 (peaking at slice −4) (Figure 3e). Taken together, these data are in line with the presence of a TARP-γ8-SynDIG4 containing AMPAR subcomplex.

Discussion
In the current study, we analyzed the distinct interactomes of GluA1/2 and Glu receptors using wildtype and Gria1 KO or Gria3 KO hippocampi. Interaction proteo revealed TARP-γ8, CNIH-2 and SynDIG4 as highest abundant interactors of the Glu subtype specifically, whereas GluA2/3 IP-MS revealed strongest co-purification of TA γ2, CNIH-2 and Noelin1. Further co-expression analysis revealed that TARP-γ8-SynD directly interact, and STED microscopy showed co-assembly into an AMPAR subcom especially at synaptic sites.
In the past decades, multiple AMPAR interactors have been identified [1 Known AMPAR binding partners vary in their interaction strength and stability [17] IP-MS protocol used in the current study favored the identification of a subset of e
In the past decades, multiple AMPAR interactors have been identified [17,32]. Known AMPAR binding partners vary in their interaction strength and stability [17]. The IP-MS protocol used in the current study favored the identification of a subset of estab-lished interactors. These included the more stable interacting transmembrane proteins, consistently identified by proteomics studies and which are considered 'core' interactors [17,25]. Stabilization of transient interactions by use of a crosslinker before IP-MS could improve coverage of the AMPAR interactome and its analysis in different (KO) conditions in future studies.
Previous IP-MS analysis on the total pool of hippocampal AMPARs revealed TARP-γ8 and CNIH-2 as the most abundant interactors [9]. In the current study, TARP-γ8, CNIH-2 were identified as the highest abundant interactors specifically of the GluA1/2 receptor subtype in addition to SynDIG4. Other interactors, including TARP-γ2 and Noelin1, only revealed a >10 times lower intensity. For the first time, we investigated the interactome of the lower abundant GluA2/3 receptor in isolation. In contrast to the GluA1/2 receptor, GluA2/3 receptors revealed the strongest interaction with TARP-γ2, CNIH-2 and Noelin1. These latter proteins may therefore be of highest interest for functional studies on the GluA2/3 receptor subtype, and GluA3-dependent disease mechanisms, like the induction of Amyloid-β pathology in Alzheimer's disease models [15]. Previous work revealed the requirement of GluA3-containing AMPAR endocytosis for the synaptotoxic and cognitive effects of Amyloid-β [15]. The exact mechanism underlying the Amyloid-β induced pathway remains unknown, and logically may involve major GluA3 interactors. Interestingly, oligomeric Amyloid-β was shown to be able to bind TARP-γ2 and Noelin1 [50]. As TARP-γ2 and Noelin1 are major interactors of GluA2/3 receptors, these proteins may be interesting candidates for further investigation into the Amyloid-β associated pathway. -Of note, removal of GluA3 in our current study did not affect expression of TARP-γ2, CNIH-2 and Noelin1, suggesting that, conversely, GluA2/3 receptors are not their major interactors.
In the current study, we observed an AMPAR subtype containing both GluA1 and 3 subunits, and potentially GluA2, in the hippocampus. The GluA1/(2)/3 receptor subtype has been observed in previous studies [4,5,17], but is often overlooked. We validated its presence in mouse hippocampus by direct purification with GluA1 or GluA3 specific antibodies. In addition, we revealed SynDIG4 as an interactor of the GluA1/(2)/3 receptor subtype. Whereas GluA1 and SynDIG4 co-purified with GluA3 in wildtype samples, they were both absent in GluA2/3 IP-MS performed on Gria1 KO mice. Further experiments are necessary to determine additional interactors of the GluA1/(2)/3 receptor subtype.
In addition, we revealed co-assembly of TARP-γ8 and SynDIG4 in an AMPAR subcomplex by TARP-γ8 and SynDIG4 IP-MS and IP-BN-PAGE-MS. IP-MS revealed co-purification of the AMPAR and SynDIG4 when pulling down TARP-γ8. Conversely, pull down of SynDIG4 revealed co-isolation of the AMPAR and TARP-γ8. However, as co-IP on overexpressed TARP-γ8 and SynDIG4 showed that these two proteins can interact directly, the IP-MS experiments do not necessarily demonstrate their co-assembly in an AMPAR subcomplex. To reveal the presence of a TARP-γ8-SynDIG4 containing AMPAR subcomplex, we separated subcomplexes by size using IP-BN-PAGE-MS. TARP-γ8 (~50 kDa) and SynDIG4 (~37 kDa) were expected to comigrate on the gel together with the AMPAR (>720 kDa), at a molecular weight much higher than that of the two proteins combined (87 kDa). Indeed, IP-BN-PAGE-MS revealed comigration of TARP-γ8, SynDIG4 and GluA1 together above 720 kDa.
In a previous study, both TARP-γ8 and SynDIG4 revealed a similar expression profile across biochemical synaptic subfractions, including de-enrichment at the PSD [8]. Our microscopy analysis revealed colocalization of these proteins largely overlapping with Homer positive synaptic puncta, revealing this subcomplex exists mainly at synaptic sites.
The primary function of SynDIG4 is thought be retaining AMPARs extrasynaptically [40]. Upon stimulation, this block may be released, allowing other interactors to transport the receptor into the synapse [40]. In the current study, we identified TARP-γ8-SynDIG4 as part of an AMPAR co-assembly, a direct interaction between TARP-γ8 and SynDIG4 and their colocalization at synaptic sites. Hence, these two proteins may work together in a mechanism of AMPAR release and synapse insertion. Their strong association with the GluA1/2 receptor subtype may underly the typical activity dependent insertion of GluA1/2 receptors into the synapse during fast excitatory transmission [13]. The exact mechanism of AMPAR regulation by TARP-γ8 and SynDIG4, and the interplay between these two proteins, remains to be established. To test the functionality of this AMPAR subcomplex directly, one could block the interaction between TARP-γ8 and SynDIG4 and measure the effects on AMPAR localization with and without stimulation. For this, identification of the TARP-γ8 binding site to SynDIG4 will be necessary, and can be accomplished, for instance, by crosslink mass spectrometry or a peptide array interaction assay [31].
IP-MS of TARP-γ8 revealed a >2.5 times higher abundance ratio between TARP-γ8 (bait) and SynDIG4 (interactor), than observed between SynDIG4 (bait) and TARP-γ8 (interactor) in the SynDIG4 IP-MS experiments. In agreement with this, TARP-γ8 revealed a higher level of colocalization with SynDIG4, than SynDIG4 with TARP-γ8. This indicates that a larger portion of TARP-γ8 protein is associated with AMPAR receptors decorated with SynDIG4, than the other way around. Possibly a small portion of SynDIG4 protein is associated with AMPAR-TARP-γ8, and is additionally part of other AMPAR-(in)dependent interactions.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells11223648/s1, Figure S1: Proteins and peptides used for differential expression analysis (DEA) in the Mass Spectrometry-Downstream Analysis Pipeline (MS-DAP); Figure S2: Known AMPAR interactors quantified in wildtype and Gria1-or Gria3knock-out (KO) mice; Figure S3: Antibodies tested for specificity by immunoblot; Figure S4: Immunopurification Blue Native Polyacrylamide Gel Electrophoreses (IP/BN-PAGE-MS) explained; Figure  S5: Super-resolution microscopy on hippocampal neurons revealing TARP-γ8 and anti-SynDIG4 colocalization; Table S1: Summary of the mean, s.e.m., n numbers and eBayes non-FDR adjusted p-values of relative protein abundances measured with mass spectrometry, as shown in Figure 1b; Table S2: Summary of the mean, s.e.m., n numbers and statistics of relative protein abundances measured with immunoblot, as shown in Figure 1d; Table S3: Excel file with raw iBAQ values of AMPAR immunopurifications performed on wildtype and Gria1 or Gria3 deficient hippocampus crude synaptosomes; Table S4: Excel file with raw iBAQ values of TARP-γ8 and SynDIG4 immunopurifications performed on hippocampus crude synaptosomes.