Affinity proteomics identifies novel functional modules related to adhesion GPCRs

Adhesion G protein–coupled receptors (ADGRs) have recently become a target of intense research. Their unique protein structure, which consists of a G protein–coupled receptor combined with long adhesive extracellular domains, suggests a dual role in cell signaling and adhesion. Despite considerable progress in the understanding of ADGR signaling over the past years, the knowledge about ADGR protein networks is still limited. For most receptors, only a few interaction partners are known thus far. We aimed to identify novel ADGR‐interacting partners to shed light on cellular protein networks that rely on ADGR function. For this, we applied affinity proteomics, utilizing tandem affinity purifications combined with mass spectrometry. Analysis of the acquired proteomics data provides evidence that ADGRs not only have functional roles at synapses but also at intracellular membranes, namely at the endoplasmic reticulum, the Golgi apparatus, mitochondria, and mitochondria‐associated membranes (MAMs). Specifically, we found an association of ADGRs with several scaffold proteins of the membrane‐associated guanylate kinases family, elementary units of the γ‐secretase complex, the outer/inner mitochondrial membrane, MAMs, and regulators of the Wnt signaling pathways. Furthermore, the nuclear localization of ADGR domains together with their physical interaction with nuclear proteins and several transcription factors suggests a role of ADGRs in gene regulation.


Introduction
Over the last two decades, the adhesion G proteincoupled receptor (aGPCR or ADGR) protein family and their functional roles in cell adhesion and signaling have received increasing attention. 1 The ADGR protein family comprises 33 members, characterized by a unique protein structure composed of a large extracellular domain with adhesive function and a seven transmembrane (7TM) moiety that resembles G protein-coupled receptors (GPCRs) of the secretin family. According to the molecular signature of their 7TM domains, the 33 human AGPRs are categorized into nine distinct subfamilies (I-IX). 2 A characteristic feature of ADGRs is their ability to undergo autocleavage at the G protein-coupled receptor proteolytic site (GPS), which is located near the first transmembrane domain and embedded in the GPCR autoproteolysis-inducing (GAIN) domain (Fig. 1A). 3 ADGR autocleavage results in an N-terminal fragment (NTF) and a C-terminal fragment (CTF) that usually stay associated as a dimer. 4 However, there is increasing evidence that the NTF and CTF can act independently. 5,6 Many CTFs show increased activation compared with the full-length receptor when overexpressed in cellular systems [7][8][9][10] and it has been presumed that the signaling function of ADGRs is regulated by their NTFs. 3 recently, it has been demonstrated that the first ∼5-10 amino acids of the CTF, called Stachel, mediate the activation of ADGRs upon NTF removal or conformational change. 13,14 However, there is also evidence for Stachel-independent signaling that requires alternative, more complex activation. 15,16 Despite considerable progress in ADGR research, many functions of these unique receptors and their underlying molecular mechanisms remain largely elusive. The elucidation of ADGR function is hampered by the lack of known ligands and intracellular interaction partners. 17  Identification of such interacting partners should provide a more comprehensive view of ADGR functions in defined cellular modules. In the present study, we used a powerful affinity proteomics approach to identify physically interacting partners and protein complexes of nine ADGRs from five different ADGR groups, namely LPHN2 (ADGRL2, group I), CD97 (ADGRE5, group II), GPR123 (ADGRA1, group III), GPR124 (ADGRA2, group III), GPR125 (ADGRA3, group III), BAI1 (ADGRB1, group VI), BAI2 (ADGRB2, group VI), BAI3 (ADGRB3, group VI), and VLGR1 (ADGRV1, group IX). To identify proteins and protein complexes interacting with the cellular parts of the ADGR, we tagged the intracellular domain (ICD) or CTF of the ADGRs with the Strep II/FLAG tandem affinity purification (SF-TAP) tag and expressed the SF-tagged proteins in HEK 293T cells. We subsequently performed tandem affinity purification (TAP) to isolate ADGR-associated complexes close to their native functional state from cell lysates. 18 Subsequently, the content of purified protein complexes was analyzed by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). Analysis of the identified complex components revealed novel intracellular ADGR interaction partners, including synaptic scaffold proteins, nuclear proteins, and regulatory phosphatases. In addition, we determined for the first time the association of ADGRs with intracellular membrane networks associated with the endoplasmic reticulum (ER), the Golgi apparatus, and mitochondria. Specifically, we identified the subunits of the γ-secretase complex as a common interaction partner for VLGR1, CD97, LPHN2, and GPR124. A large overlap in the interactomes of VLGR1, CD97, and LPHN2 indicates a common joint protein interactome. Moreover, their nuclear localization paired with their physical interaction with nuclear proteins and several transcription factors suggest an unexpected role of ADGRs in gene regulation.

Cell culture
We used HEK 293T cells, which are commonly used as human cell models for the TAP analyses, 19 including studies on GPCRs. 20 With the exception of BAI3 and GPR123, all analyzed ADGRs are expressed in HEK 293T cells 21 (VLGR1, own unpublished data). For the present experiments, HEK 293T cells and HeLa cells were cultured in Dulbecco's modified Eagle's medium containing 10% heat-inactivated fetal calf serum. Cells were transfected with GeneJuice R (Merck Millipore, Darmstadt, Germany) according to manufacturer's instructions.
Tandem affinity purification TAP was performed as described previously, 22 allowing the isolation of protein complexes under mild conditions. For this, we tagged the N-or C-terminus of the ICDs and CTFs of ADGRs with the SF-TAP ( Fig. 1A-C). SF-tagged proteins were expressed in HEK 293T cells for 48 hours. Since we observed apoptosis in the cells expressing CD97_CTF, BAI2_CTF, and LPHN2_CTF, we shortened the time of expression to 18 hours. Mocktreated HEK 293T cells were used as controls. The cells were lysed and the lysate was cleared by centrifugation. The supernatant was then subjected to a two-step purification on Strep-Tactin R Superflow R beads (IBA, Göttingen, Germany) and anti-FLAG M2 agarose beads (Sigma-Aldrich, Hamburg, Germany). Competitive elution was achieved using desbiothin (IBA) in the first step and FLAG R peptide (Sigma-Aldrich) in the second step. The eluate was precipitated by methanol-chloroform and then subjected to mass spectrometry analysis. Eluted affinity 146 Ann purified complexes were subsequently analyzed by LC-MS/MS.

Mass spectrometry
LC-MS/MS was performed as previously described. 19 In brief, SF-TAP-purified protein complexes were solubilized before subjected to trypsin cleavage. The resulting peptides were desalted and purified using stage tips before separation on a Dionex TM RSLC system. Eluted peptides were directly ionized by nanospray ionization and detected by an LTQ Orbitrap TM Velos mass spectrometer (Thermo Fisher Scientific). We search the raw spectra against the human SwissProt database using Mascot and verified the results by Scaffold (version Scaffold 4.02.01, Proteome Software Inc.) to validate MS/MS-based peptide and protein identifications.

Data processing
Mass spectrometry data of all TAPs were compared with the corresponding data for mock-transfected cells. Proteins that occurred in the mock dataset were not considered for subsequent analysis. We also compared our datasets with a total of 140 TAPs of the protein RAF1, commonly used as an unrelated control TAP analysis. 18,19 Furthermore, we compared hits with the data listed in the Contaminant Repository for Affinity Purifications (CRAPome) database. 23 The CRAPome is a collection of common contaminants in affinity proteomic MS data and contains data for control experiments from an increasing number of affinity purifications. We further analyzed only those with an occurrence below 5%. The gene names of ADGR preys were used as input for the Cytoscape plugins STRING and ClueGO and the STRAP software. The parameter confidence (score) cutout was set to 0.4 and the parameter maximum number of interactors was set to 0 for STRING analysis. ClueGO v2.3.3 was used for Gene Ontology (GO) term enrichment analysis. Network specificity was set to default (medium).

Antibodies
Mouse anti-FLAG M2 (Sigma-Aldrich) and mouse anti-SIGMAR1 (sc-166392, Santa Cruz Biotechnology) were used as the primary antibodies for immunocytochemistry. Secondary antibody conjugated to Alexa568 was purchased from Molecular Probes TM (Life Technologies, Darmstadt, Ger-many). Nuclear DNA was stained with DAPI (1 mg/mL) (Sigma-Aldrich).

Immmunocytochemistry
Cells were fixed and permeabilized in ice-cold methanol for 10 min and washed with phosphatebuffered saline. After washing, the cells were covered with blocking solution and incubated overnight with the primary antibody at 4°C. Cells were washed and then incubated with the secondary antibody in blocking solution containing DAPI for 1.5 h at room temperature. After washing, sections were mounted in Mowiol R (Roth). Specimens were analyzed on a Leica DM6000B microscope and images were processed with Leica imaging software and Adobe Photoshop CS (intensity adjustment). For the analysis of the colocalization of VLGR1 and SIGMAR1, the Leica DMi8 system in combination with the Thunder software was used.

Colocalization analysis
The Pearson correlation coefficient (R) was used to determinate the degree of colocalization between VLGR1_CTF_HA and SIGMAR1 in HeLa cells. Calculation of the Pearson correlation coefficient is a mathematical method to measure the strength of a linear association between two variables. 24 The correlation value has a range from + 1 to -1. A value of 0 indicates no association, greater than 0 indicates a positive association, and less than 0 indicates a negative association between the two variables. The stronger the positive association of the two variables, the closer R is to + 1. The Pearson coefficient was calculated using the Coloc 2 plugin of ImageJ (https://imagej.nih.gov/ij/).

γ-Secretase inhibitor assays
Both potent γ-secretase inhibitors LY411575 and DAPT (N-[N-(3,5-difluorophenacetyl)-l-alanyl]-Sphenylglycine t-butyl ester) were purchased from Sigma Aldrich and were added to culture media at a final concentration of 100 nM in DMSO after transfection of HEK 293T cells. As control, the same volume DMSO was added to the cells after transfection.

Results
We aimed to identify interacting partners in protein complexes associated with ADGRs by affinity proteomics in HEK 293T cells, applying TAPs in combination with mass spectrometric determination of the protein content of eluates. 18,19 We analyzed the proteomic datasets by STRAP 25 and grouped proteins according to their assigned GO terms (Fig. 1D). We used ClueGO, 26 a Cytoscape (http: //www.cytoscape.org/) plugin, for GO term enrichment analysis (e.g., Figs. S1 and S2, online only) and evaluated the specificity of the identified prey proteins by comparing our data with the CRAPome dataset. We excluded from further analyses the molecules identified in our TAPs that were in the controls and the ones that are present in the CRAPome in 5% of the datasets in order to secure a high probability of specific interaction with ADGRs.
In TAPs for BAI3, we also found the scaffold protein SNTB2, and for GPR124 and GPR125, we found the scaffold protein SCRIB, which was also present in TAPs of LPHN2. Both SNTB2 and SCRIB do not belong to the MAGUK family but do contain PDZ domains. In TAPs with CD97 constructs, we did not find any scaffold proteins.
Notably, we also did not identify any of the abovementioned scaffold proteins in TAPs with the CTFs of VLGR1, BAI2, GPR124, and LPHN2, potentially due to the C-terminal SF-tag, which may block binding to the ADGR PDZ-binding motif (PBM). However, the TAP data that we acquired with ADGR_CTFs indicated synaptic localization and functions of the receptors: they contain various proteins involved in vesicle fusion and numerous interactors that are associated with synapse-related GO terms, such as synaptic signaling and synaptic vesicle cycle (Fig. 2B).

TAP data indicate binding of phosphatases and kinases to ADGRs
In TAPs of all ADGRs except GPR123, we identified diverse phosphatases and kinases (Table 1). For most ADGRs, we identified catalytic and regulatory subunits of protein phosphatases 2 (PP2) and 6 (PP6). BAI proteins also interacted with subunits of protein phosphatases 1 and 4. Interestingly, the ICDs (except in the case of VLGR1) were sufficient to interact with these phosphatases (Table 1), indicating that phosphatases preferentially bind to the ICD of ADGRs. In TAPs with CTFs of GPR124, BAI2, CD97, and VLGR1, the protein phosphatase Mg 2+ /Mn 2+ -dependent 1B was also found. Furthermore, our TAPs revealed casein kinase 2 (CK2), calcium/calmodulin-dependent protein kinase II (CAMK2), cyclin-dependent kinase 4 (CDK4), and mitogen-activated protein kinase 1 or 3 (MAPK1/MAPK3) as possible candidates that phosphorylate ADGRs. In silico analysis of ADGR amino acid sequences using the kinasephos 2.0 tool (http://kinasephos2.mbc.nctu.edu.tw/predict. php) predicts phosphorylation sites in all ADGRs studied, defining them as potential targets for phosphorylation. All ADGRs contain potential CK2 and CDK phosphorylation motifs in their ICD. A potential MAPK phosphorylation site was predicted for all ADGRs except CD97. Potential CAMK target sites were predicted for LPHN2, BAI2, and BAI3.

ADGR presence in protein complexes of the ER, the Golgi apparatus, and mitochondria
Comparing the TAP datasets of ADGR_CTFs, we identified 116 molecules present in all TAPs of all five ADGRs (Table S1, online only). We grouped these proteins according to their assigned GO terms with STRAP ( Fig. 1D) and analyzed them for enriched GO terms with ClueGO ( Fig. S1, Table S1, online only). Our analysis demonstrated that the most significantly enriched GO terms in the category Cellular Component were related to the ER, the Golgi apparatus, mitochondria, and the nucleus (Fig. S1A, online only). In the GO term category Biological Process, we found an enrichment in the terms mitochondrial ATP synthesis coupled proton transport, membrane lipid biosynthetic process, protein import into nucleus, and response to ER stress (Fig. S1B, online only). Notably, many of the ADGR-associated proteins were amino acid and zinc transporters of the solute carrier (SLC) family and ATPases that are coupled to H + and Ca 2+ transmembrane transport (Table S1, online only). Several of our TAP preys were enriched in mitochondria-associated membranes (MAMs) ( Table 2). MAMs are contact sites of the ER to the outer mitochondrial membrane. 27 The association with MAM proteins was most prevalent for VLGR1 and the

Protein phosphatases and kinases identified in ADGR-TAPs
Mitogen-activated protein kinase 1

MAPK3
Note: Listed are the phosphatase subunits and kinases that were found by TAP for each ADGR. Most subunits were identified with the ICD; those marked with * were identified with the CTF. Only those that show a low occurrence in the CRAPome (less than 5%) are listed.
least prevalent for BAI2. Immunocytochemical double labeling of VLGR1 and the MAM core protein sigma1R (SIGMAR1) in human HeLa cells transfected with VLGR1_CTF_HA revealed partial colocalization of SIGMAR1 and VLGR1 (Fig. 3C).
We also identified proteins that are located at the outer and inner mitochondrial membrane as putative interactors for all ADGR_CTFs (Table 3). These include proteins that are mainly involved in transmembrane transport and membrane folding. We found that all ADGR_CTFs associate with components of the respiratory chain, namely subunits for NADH-ubiquinone oxidoreductase (complex I), the cytochrome bc1 complex (complex III), cytochrome c oxidase (complex IV), and the F 1 F 0 ATP synthase (complex V) ( Fig. 3 and Table 3). In addition to components of the respiratory chain complex, we identified proteins that are involved in the assembly of mitochondrial complexes. These include COA3, OXA1L, and TIMM21, which are necessary for complex IV assembly, [28][29][30] as well as NDUFAF-1, -3, and -4, and TMEM126B, which participate in complex I assembly. 31 Apoptosis, mitochondria morphology 106 - Targets IP3Rs for degradation 112 Catalyzes the degradation of heme 116 - ER stress, chaperone 117 - Binds VDAC 76 - Cardiolipin acyl chain remodeling 118 ER protein retention 121 - Phospholipid metabolism 123 ER-mitochondria tethering 126 - Fatty acid transport 127 Lipid transport and calcium signaling 117 Fatty acid transport 128 Ion exchange 76 Note: Only identified molecules that show a low occurrence in the CRAPome (less than 5%) are listed.
we further identified BCS1L, which is necessary for complex III assembly. 34 All ADGR_CTFs were associated with translocases of the inner mitochondrial membrane (TIMs) and to a lesser extent with translocases of the outer mitochondrial membrane (TOMs) ( Table 3). Moreover, all ADGR_CTFs bound to components of the MICOS complex that is crucial for the formation and maintenance of the mitochondrial cristae structure. 35 In addition, we found YME1L1 in TAPs for GPR124, BAI2, and LPHN2, as well as PARL for VLGR1. Both YME1L1 and PARL also maintain cristae morphology and have an antiapoptotic effect. 36,37 CTFs of ADGR physically interact with components of the γ-secretase complex Strikingly, we found LPHN2 as a prey for CD97_CTF and VLGR1_CTF, and CD97 as a prey for VLGR1_CTF. This indicates that these three ADGRs may be part of the same protein complexes. This prompted us to check for preys shared by CD97, LPHN2, and VLGR1 TAPs, and we found that 196 proteins were common for all three of these ADGRs (Fig. S2A, online only). After filtering out proteins that occur in more than 5% of the negative controls in the CRAPome, we performed a GO term enrichment analysis with the remaining 89 interacting proteins. Our analyses   Translocase of  the inner  membrane   TIMM17A  TIMM21  TIMM23B  TIMM50   TIMM17B  TIMM23  TIMM50   TIMM17A  TIMM21  TIMM23B  TIMM50  TIMMDC1   TIMM17A  TIMM21  TIMM23B  TIMM50  TIMMDC1   TIMM17B  TIMM23  TIMM50  TIMMDC1   Translocase of  the outer  membrane   TOMM22  TOMM22  TOMM70A   TOMM22  -TOMM22  TOMM40 Mitochondrial SLC transporters Note: Only identified molecules that show a low occurrence in the CRAPome (less than 5%) are listed. revealed that the GO terms were mostly related to ER and mitochondria localization and function ( Fig. S2 and Table S3, online only). We observed that all three TAPs contained subunits of the γ-secretase complex ( Fig. 4A and B). Interestingly, the γ-secretase complex has recently been reported to be present in MAMs. [38][39][40] In TAPs of VLGR1 and CD97, we identified nicastrin (NCSTN), presenilins 1 and 2 (PSEN1 and PSEN2), as well as aph-1 homolog A (APH1A), but not presenilin enhancer 2 (PEN2), which is important for endoproteolysis of presenilins, activating the enzyme. In TAPs of LPHN2, the proteins NCSTN, PSEN2, and APH1A were present (Fig. 4B). Notably, we also found that GPR124_CTF interacts with NCSTN. In contrast, we did not identify any γ-secretase subunits in the BAI2_CTF TAP dataset. Besides the γ-secretase complex itself, we also found several γ-secretase modulators in the ADGR TAPs. In all ADGR_CTF TAPs, the regulatory subunit basigin (BSG) was present. 41 Furthermore, we identified the presenilin cleavage proteins histocompatibility minor 13 (HM13) and the transmembrane p24 trafficking protein 10 (TMED10) in all ADGR_CTF TAPs, with the exception of BAI2 TAP. 42,43 We further aimed to determine whether ADGRs may be substrates for the γ-secretase complex

(B) Table of γ-secretase subunits that bound to ADGR_CTFs in TAP. (C) Amino acid sequence alignments of positively charged residues (colored) at the junction of the transmembrane helix 7 only of ADGRs and other receptors. Positively charged residues in Notch and APP (three to four residues after the cleavage sites) are the primary determinants for substrate binding. 44 (D) Schematic representation of ADGR cleavage by γ-secretase. (E) VLGR1_CTF-HA is cleaved, releasing a smaller band of ∼26 kDa. * The intensity of the lower part was increased relative to the upper part of the blot. (F) VLGR1-CTF and smaller VLGR1 fragments at ∼26 kDa are detected in PC12, IMCD3, and RPE1 cell lysates. Panels E and F each show an exemplary western blot from three independent experiments.
( Fig. 4C-F). Amino acid sequence alignments revealed conserved, positively charged residues at the junction of transmembrane helix 7 only in the ADGRs for which we found interactions with γ-secretase components (Fig. 4C). Charged residues located three to four residues after the cleavage sites were recently described for NOTCH and the amyloid precursor protein (APP) as the primary determinants for substrate binding. 44 Since most γ-secretase subunits were found with VLGR1 and yet overexpression of VLGR1_CTF did not induce apoptotic effects, we chose VLGR1 for further investigation. Upon recombinant expression of C-terminally tagged VLGR1_CTF-HA in HEK 293T cells, we observed that, in addition to the full-length CTF with a molecular weight of ∼50 kDa, two smaller bands at ∼26 kDa occurred (Fig. 4E). This size corresponds to the molecular size of the ICD of VLGR1. In addition, corresponding protein fragments were detected in cell lysates of PC12, IMCD3, and RPE1 cells (Fig. 4F).
In conclusion, these data suggested that like other γ-secretase substrates, VLGR1 is proteolytically cleaved in the transmembrane helix 7 by the γ-secretase. However, in several preliminary experiments applying well-established γ-secretase inhibitors, namely LY-411575 and DAPT, we demonstrated the inhibition of APP cleavage but did not observe any effect on ICD release from ADGRs (data not shown).

ADGRs interact with nuclear proteins and localize to the nucleus
We observed that all ADGRs studied here associated with resident nuclear proteins, which are involved in nuclear-specific functions, such as gene regulation, RNA splicing, and transcription. This was the case for both the ADGR_CTF and ADGR_ICD baits and was most prominent for GPR124, BAI1, and BAI2.
In TAPs of the GPR124_ICD and GPCR124_ CTF, we found an enrichment of proteins that were assigned to the GO term nuclear speckles, the chromatin-free nuclear compartment of RNA splicing. For GPR124_CTF and BAI2_CTF, we observed an enrichment in proteins assigned with the GO term nucleolus, the site of ribosomal RNA transcription. Furthermore, the ICDs of BAI1 and GPR124 interacted with the PAF1 complex, which is involved in RNA polymerase II transcription elongation and transcription-coupled histone modifications. 45 GO terms related to histone modification, splicing, and DNA unwinding were enriched for all ADGRs; however, the numbers greatly varied.
The association with nuclear import proteins and the nuclear localization of some of the ADGR_ICDs led us to check whether ADGR_ICDs contain nuclear localization sequences (NLSs). For this, we applied the NLS prediction tool cNLS Mapper. 46 We found high scores for BAI1 and BAI2 and medium scores for BAI3 and GPR123. GPR124, GPR125, and LPHN2 gave only low scores, whereas no NLSs were predicted for VLGR1 and CD97 ICDs (Fig. 5B).
The results from our TAPs, together with our immunocytochemical localization analysis and the presence of NLS in some ADGRs, indicate that the ICDs of ADGRs may be cleaved and potentially act as transcription factors in the nucleus. We therefore checked our TAP data for proteins that act as transcriptional regulators. Indeed, we found numerous proteins that are related to transcriptional regulation. The molecules involved in transcriptional regulation identified in our ADGR TAPs are included in Table 4.

Identification of modulators of the Wnt signaling pathway in ADGR TAPs
The GO term analyses of our TAP data also revealed an enrichment for the GO term Wnt signaling in the BAI2_CTF, GPR124_CTF, and VLGR1_CTF datasets. Furthermore, single proteins involved in Wnt pathways were also found in TAPs of CD97_CTF and LPHN2_CTF ( Table 5). Most of these molecules modulate Wnt signaling by targeting β-catenin. Some prey proteins have functions in the nucleus. SLC30A9 is part of the β-catenin transcription complex and participates in the regulation of Wnt downstream genes. 47

Discussion
In the present study, we identified previously described binding partners and numerous novel putative interactors of ADGRs by TAPs in combination with subsequent mass spectrometry. Nevertheless, our affinity capture approach bears limitations, which have to be considered for the interpretation of our datasets. In our study, we used HEK 293T cells as established human cellular models for the ADGR TAPs. 19,20 Therefore, we were not able to capture tissue-or cell-specific interacting partners that are Regulation of transcription factors 144 Transcriptional activation of genes (e.g., collagenase and insulin) Regulates epigenetic gene silencing 146 Modulates various viral and cellular promoters 147 --+ -+ -Note: Only identified molecules that show a low occurrence in the CRAPome (less than 5%) are listed. not expressed in HEK 293T cells. Since we cultured the cells under standard conditions, we also missed proteins that require specific physiological conditions for binding to ADGRs, such as mechanical or chemical ligand stimulation. Although TAP protein complexes are purified in close-to-native conditions, the membrane protein complexes lose their membrane context during the purification steps and therefore are difficult to obtain in general. 20 In addition, overexpression of ADGR domains may lead to changes in the stoichiometry of purified protein complex compositions. In any case, as in other screens, it is necessary to validate all newly identified putative binary interactions by applying independent, complementary experiments from the molecular to organismic level to sort out false positive hits. Nevertheless, the obtained datasets highlight the usefulness of our affinity capture approach even for membrane proteins. The identified ADGR-interacting proteins support previously discussed functions of ADGRs, but more importantly define novel physiological functions related to ADGRs.

ADGR subfamilies interact with scaffold proteins
The TAP data of our study reveal numerous interactions of ADGRs and scaffold proteins. All identified scaffold proteins possess PDZ domains and all ADGRs investigated in the present study contain a PBM at their C-terminal end (Fig. 1) that is predetermined to bind to PDZ domains. Therefore, their mutual interaction most probably occurs via the binding of ADGRs' PBM to PDZ domains. This is also confirmed by previous experimental data.
We have previously shown that VLGR1 (USH2D) directly binds the two other Usher syndrome proteins, harmonin (USH1C) and whirlin (USH2D), both of which serve as cytoplasmic anchors in Functions as an adaptor between Wnt receptors and V-ATPase Represses canonical Wnt signaling 150 Inhibitor for β-catenin 151 - Modulates Wnt signaling activity via LRP6 Mediates degradation of β-catenin Participates in transcriptional activation of Wnt-responsive genes Ubiquitinates and upregulates β-catenin Regulates axin stability 156 -- Regulates Wnt protein sorting and secretion 157 + + + + + Note: Only identified molecules that show a low occurrence in the CRAPome (less than 5%) are listed.
membrane adhesion complexes of the inner ear hair cells and retinal photoreceptor cells. [48][49][50] It has also been reported previously that BAI1 and GPR124 interact with DLG1 (SAP97). 10,51 Interestingly, although the direct interaction of DLG1 with the C-terminal PBM of CD97 has been shown recently, 52 we did not find DLG1 or any other PDZcontaining protein in our TAPs of CD97. This is most probably due to the fact that the interaction of DLG1 with CD97 is induced by phosphorylation of the PBM of CD97. However, this phosphorylation is induced by mechanical stimulation, a condition not present in our TAPs. Nevertheless, our TAP data revealed the interaction of ADGRs of group I (latrophilins), group III (GPR123, GPR124, and GPR125), and group VII (BAIs) with the same set of scaffold proteins (Fig. 2B). The question of whether the ADGRs are integrated in common protein networks and signaling hubs organized by scaffold proteins in cells and tissues will be reserved for future studies.

ADGR subfamilies are part of protein networks at synapses
Although the identified scaffold proteins are ubiquitously expressed, they are essential components of the protein networks of synapses, particularly of the postsynaptic density of neurons. 53 These scaf-fold proteins are mainly members of the MAGUK family, which are involved in the establishment and maintenance of cell polarity and the dynamic arrangement of receptors and channels at synaptic membranes. 10,54,55 In addition, we identified several proteins involved in the targeting and fusion of synaptic vesicles as putative interaction partners of ADGRs (Fig. 2B), suggesting an involvement of ADGRs in these processes too. For example, SNAP23, which interacts with all ADGR_CTFs, is known to be involved in the exocytosis of glutamate receptors to the postsynaptic membrane. 56 Our findings are also in line with the enrichment of ADGRs, namely VLGR1, latrophilins, and BAIs, at the postsynapse of dendrites 50,57-59 and the function of latrophilins and BAIs in synaptic remodeling. 55,[60][61][62] Although the ADGRs CD97, GPR123, GPR124, and GPR125 have not been mapped to synaptic subdomains, a role in synaptic function cannot be ruled out yet. relationship with this intramembranous complex. The γ-secretase complex is well known for the generation of the β-amyloid peptide protein by the sequential proteolysis of the APP, which is a crucial step in the development of Alzheimer's disease. 63 However, in addition to APP, the γ-secretase also proteolytically cleaves a variety of integral membrane proteins that initiate downstream pathways involved in transciptional regulation in the nucleus. 64 Indeed, we also observed the release from several ADGRs of a fragment with the size of the ICD in cultures of diverse cell lines (Fig. 4). In addition, we demonstrated the localization of ADGR fragments in the nucleus, which is further supported by the presence of NLS sequences in the ICDs of diverse ADGRs (Fig. 5). Together, our findings suggest a novel signaling mechanism of ADGRs that comprises the release of an ICD upon γ-secretase cleavage and subsequent translocation of the ICD into the nucleus for gene regulation. This scenario is similar to the described signaling pathway related to the γ-secretase substrate polycystin 1 (PKD1). 65 PKD1 shows high structural similarity to ADGRs: it is also a multispanning transmembrane protein that undergoes autocleavage at a GPS, analogous to ADGRs. 66,67 Upon γ-secretase cleavage, a small (∼30 kDa) PKD1_ICD fragment is released and directed into the nucleus. 68,69 The release of PKD1_ ICD is most probably induced by mechanical stimuli, 68 an activation mechanism that has also been discussed for ADGRs. [70][71][72] Although our data favor a noncanonical signaling pathway triggered by γ-secretase cleavage, our approach of applying established γ-secretase inhibitors did not alter the release of ICDs from the ADGR. ADGRs may alternatively play a role in γ-secretase positioning in specific membrane domains (e.g., lipid rafts) or in its regulation. This is supported by the fact that, besides γ-secretase subunits, we also identified γ-secretase regulators in our TAPs, such as BSG, TMED10, and HM13. In particular, BSG was found in the TAP datasets for LPHN2, CD97, and VLGR1 CTFs and was previously described as a potential additional γ-secretase regulatory subunit. 41 TMED10 was present in all ADGR_CTF TAPs except for BAI2. It is part of presenilin complexes and regulates γ-secretase cleavage activity. 73 Finally, HM13 was identified as a TAP prey for all ADGR_CTFs. It is an activa-tor for PSEN1 and thereby promotes γ-secretase cleavage. 43 In any case, our TAP data revealing the interaction of ADGR with the γ-secretase complex pave the way for further investigations to identify the role of this interplay. Interestingly, the γ-secretase complex is not only present in the plasma membrane but is even more prominently associated with internal cell membranes, such as the ER or MAMs. [38][39][40] Evidence for association of ADGRs with MAMs and participation at biogenesis of mitochondria In the present TAPs, we identified many proteins known to locate mainly to specific intracellular compartments, such as the ER and mitochondria. The high number of molecules related to intracellular compartments suggests novel roles of ADGRs in the cell that have not been described so far. In particular, we found various proteins located at contact sites between the ER and mitochondria, namely the MAMs. Interestingly, the γ-secretase subunits PSEN2 and PSEN1 are enriched in this compartment. Besides γ-secretase subunits, we identified the MAM protein SIGMAR1 as an interactor for all ADGRs analyzed. The present immunocytochemistry demonstrates partial colocalization of this MAM core protein with VLGR1. SIGMAR1 regulates ITP3R-dependent calcium efflux at the ER and the biogenesis of lipids. 74 The spatial regulation of calcium homeostasis and lipid biogenesis are both major processes associated with MAMs. 27,75 In addition, our analyses of TAPs indicate that all ADGRs interact with the voltagedependent anion channels 1 and 2 (VDAC1 or VDAC2) that localize to the outer mitochondrial membrane and allow the exchange of small hydrophilic molecules. 76,77 It is conceivable that ADGRs assist in the function of the highly dynamic MAMs. 78 Surpisingly, we identified various ADGR interactors that localize to the mitochondrial inner membrane. The majority of the identified mitochondrial proteins have very low occurrence in the CRAPome, indicating the specificity of their interaction with ADGRs. These TAP hits are subunits of the respiratory chain complexes I, III, and IV, as well as subunits of the F 0 F 1 ATP synthase, all localizing to mitochondrial cristae (Fig. 4). 79 addition, we identified components of the MICOS complex, which is essential for cristae formation, 35 and proteins that are involved in complex assembly of the respiratory chain. Furthermore, we found inner and outer membrane translocases (TIMs and TOMs), which are essential for transmembrane translocations.
ADGRs have not been found in mitoproteomes so far. 81,82 However, a previous report indicated a substantial role of GPR126, a group VII ADGR, in the biogenesis of mitochondria. 6 GPR126 deficiency results in mitochondrial defects in the developing heart of mice. These defects manifest in defective differentiation of mitochondrial cristae. The present identification of putative ADGR interactors that are localized to cristae or support the formation of cristae suggests that ADGRs may regulate the biogenesis of mitochondria in general. Patra et al. 6 also observed the accumulation of lipids in Gpr126 −/− cardiomyocytes, indicating defective lipid metabolism. This is in line with a function of ADGRs in MAMs, which are very important for lipid biogenesis and represent sites of constant lipid exchange.
It is noteworthy that the respiratory chain subunits that we identified are all transcribed in the nucleus and have to be imported into mitochondria. There is growing evidence that MAMs play an essential role in this translocation processes. Based on our findings, we hypothesize that ADGRs are part of the dynamic MAMs and participate in the control of the delivery of mitochondrial components into mitochondria during mitochondrial biogenesis. For this, they may interact only transiently with mitochondrial proteins, which may also explain why ADGRs are not present in the existing mitoproteomes.

TAP data indicate regulation of ADGRs by phosphorylation
The identification of diverse kinases and phosphatases in the present TAPs (Table 1) indicates that both kinases and phosphatases bind to ADGRs. ADGRs may serve as scaffolds for these enzymes. Alternatively, they may regulate the ADGR function by the yin and yang of protein phosphorylation and dephosphorylation. The latter hypothesis is supported by previous data demonstrating that extracellular domain phosphorylation may facilitate ADGR signaling. 12 However, most putative phos-phorylation sites have been identified in the ICD of ADGRs. 12 This is also in line with the predicted phosphorylation sites of all kinases identified by the present TAPs. The phosphorylation of residues in the PBMs of the ICD may regulate the binding to PDZ domains of diverse scaffold proteins also identified as potential interactors in the present study (see above). Such regulation has recently been described for the binding of CD97 to one of the PDZ domains of the DLG1 scaffold protein triggered by the phosphorylation of the C-terminal PBM of CD97. 52 This mechanism might be a general mechanism for the regulation of PBM and PDZ domain interaction.
It is notable that we did not identify any member of the GPCR kinase (GRK) family by TAPs. GRKs phosphorylate activated canonical GPCRs, which promotes the binding of arrestins, precluding further G protein coupling. 83 GRKs phosphorylate GPCRs usually at serine, threonine, or tyrosine residues present in the third intracellular loop of the 7TM domain and the ICD. However, GRKs for ADGR phosphorylation have not been identified so far. Therefore, the absence of GRKs in our TAPs can be due to the fact that we did not stimulate the ADGRs in our TAPs or that indeed no GRKs exist for ADGRs.
The identified kinases are part of signaling pathways, which may be related to ADGR. For example, MAPK1 and MAPK3 (also known as extracellular signal-regulated kinases (ERK1/2)) are central kinases of the ERK pathway that plays an important role in integrating external signals into signaling events promoting cell growth and proliferation in many mammalian cell types. 84 One of the targets of the MAPK/ERK kinase cascade is another identified kinase, CDK4, which controls the G 1 -S phase in the cell cycle. The elucidation of the role of ADGRs and their putative interacting kinases in signaling pathways has to be reserved for future experimental investigations.

ADGRs interact with Wnt/planar cell polarity signaling proteins
Our TAP data provide several lines of evidence that ADGRs are modulators of Wnt signaling. This is in line with previous reports describing crosstalk between the Wnt and ADGR pathways, which focused on GPR124 and GPR125. [85][86][87][88][89] In particular for GPR124, it has been shown that this interplay facilates angiogenesis in the central nervous system. 86,[89][90][91] Since ADGRs and their interacting partners (e.g., synaptic scaffold proteins, see above) are localized at both the pre-and postsynapse, it is conceivable that ADGRs act together with WLS in the transsynaptic translocation of the Wnt1 ligand. 92 Other Wnt-related interactors identified in TAPs differ between the ADGRs analyzed. However, all Wnt-related interactors indicate a modulating role in Wnt signaling. The most Wnt-related preys were identified in TAPs with VLGR1 constructs. These include molecules previously linked to Wnt signaling, such as ATP6AP2 and PTK7. As a planar cell polarity (PCP) core protein, ATP6AP2 also interacts with another ADGR, namely CELSR1. [93][94][95] PTK7 is a regulator of both canonical and noncanonical (PCP) Wnt signaling, and like VLGR1 it is essential for hair cell development in the cochlea. [96][97][98] For GPR124_CTF and VLGR1_CTF, we also identified SLC30A9, which directly interacts with β-catenin and is part of a complex that activates Wnt-responsive gene transcription. 99,100 Importantly, the transcriptional regulator FHL2, which we found in the VLGR1_ICD dataset, is part of the same complex. Moreover, GPR124_CTF and BAI2_CTF interacted with AMER1, which negatively regulates Wnt signaling by promoting β-catenin degradation. 101,102 Two additional proteins identified in our TAP datasets, namely NONO and SFPQ, further support the relevance of this finding. The transcriptional activator NONO, which colocalizes with AMER1 at nuclear speckles, 99 was found in TAPs of BAI2_CTF. SFPQ, which forms heterodimers with NONO, 103,104 was found in both the GPR124_CTF and BAI2_CTF datasets.
In summary, our data support the prominent role of GPR124 in the regulation of the Wnt signaling pathway. 86,[89][90][91] In addition, our findings indicate that the crosstalk of ADGRs and Wnt signaling is not only restricted to GPR124 or GPR125 85 but may also be mediated by other ADGRs, namely VLGR1 and BAI2.

Concluding remarks
In conclusion, we identified protein networks related to ADGRs by an affinity capture approach. Our data not only support previous findings but also reveal novel molecular relationships, which suggest novel cellular functions for ADGRs. Our data demonstrate that ADGRs of groups I, II, III, VII, and IX are involved in synaptic processes and are modulators of Wnt signaling. In addition, we found evidence for an association of ADGRs with the γ-secretase complex and cleavage of their ICDs, which may act in transcriptional regulation in the nucleus. Whether ADGRs are substrates of the γ-secretase or other proteases remains to be determined. Moreover, our data indicate that ADGRs may have novel roles associated with the intracellular membranes of the ER and mitochondria, and in particular with the joint protein complexes of the MAMs. In any case, we regard the outcome of the present study as springboard for future investigations that should be carried out in order to understand the complex function of ADGRs in health and disease. In addition, we thank Dr. Helen May-Simera for helpful discussions and language editing.

Author contributions
B.K. conducted most of the experiments, analyzed data, and prepared most of the figures for publication. J.R. performed sets of tandem affinity purifications. J.K. contributed to experiments analyzing γ-secretase and MAM complexes. K.B., N.H., and M.U. carried out mass spectrometry analysis and analyzed data. B.K. and U.W. designed the studies and wrote the manuscript. All authors read, contributed to, and approved the final manuscript.

Supporting information
Additional supporting information may be found in the online version of this article.