Characterization of Usher Syndrome Type 2-Associated Proteins in the Retina via Affinity Purification-Mass Spectrometry

TL;DR

This study identifies proteins interacting with Usher syndrome type 2 proteins in the retina, revealing their roles in cell structure and signaling.

Contribution

Systematic identification of USH2 protein interactors in the retina using affinity purification and mass spectrometry.

Findings

USH2 proteins connect the extracellular matrix to the actin network and signal via Gαi/Gαq.

ADGRV1 interacts with complexes involved in cell projections and TGFβ signaling.

Usherin and ADGRV1 interact with proteins related to ECM remodeling and ciliary function.

Abstract

Usher syndrome is the leading cause of inherited deaf-blindness, with type 2 (Usher syndrome type 2, USH2) being the most common form. USH2A, ADGRV1, and WHRN are the three known USH2 causative genes, which are also linked to isolated retinal degeneration and hearing loss. These genes encode usherin, ADGRV1, and whirlin, respectively, collectively called USH2 proteins. These proteins form a multiprotein complex (USH2 complex) at the periciliary membrane in retinal photoreceptors and at the stereociliary ankle link in inner ear hair cells. The molecular function of the USH2 complex and its disease mechanisms are poorly understood. Currently, there is no cure for diseases caused by mutations in the three USH2 genes. In this study, we employed multiple affinity purification methods combined with mass spectrometry to systematically identify the interaction partners of USH2 proteins in the retina. The ADGRV1 intracellular bait pulled down proteins involved in actin-based cell projections, the chaperone-containing TCP-1 complex, and the Bardet-Biedl syndrome complex. The extracellular domains of ADGRV1 and usherin pulled down proteins related to peptidase regulation, collagen biosynthesis and modification, and elastic fiber formation. The EAR/EPTP repeats of ADGRV1 specifically pulled down TGFβ signaling proteins. Further immunoprecipitation experiments identified, with high confidence, Gαi and Gαq as ADGRV1-interacting proteins, and retinal degeneration and ciliary proteins as interaction partners of USH2 proteins. We also demonstrated that the usherin extracellular domains interact with each other and with ADGRV1. Overall, these findings suggest that the USH2 complex connects the extracellular matrix (ECM) to the intracellular actin network, signals through Gαi and Gαq, and participates in ECM remodeling, TGFβ signaling, cell adhesion, and ciliary function in photoreceptors.

•Affinity purification approaches to systematically explore USH2-associated proteins.•The USH2 complex links to collagen/elastic fiber homeostasis and TGFβ activation.•ADGRV1 interacts with BBS/CCT complex components and actin-associated proteins.•ADGRV1 may signal via Gαi3/Gα11 and PI3K/PKB/mTOR pathways in the retina.•ADGRV1 interacts with usherin and EMILIN3; usherin interacts with EPHA3 and MMP19.

Affinity purification approaches to systematically explore USH2-associated proteins.

The USH2 complex links to collagen/elastic fiber homeostasis and TGFβ activation.

ADGRV1 interacts with BBS/CCT complex components and actin-associated proteins.

ADGRV1 may signal via Gαi3/Gα11 and PI3K/PKB/mTOR pathways in the retina.

ADGRV1 interacts with usherin and EMILIN3; usherin interacts with EPHA3 and MMP19.

Usher syndrome type 2 (USH2) is the leading cause of inherited deaf-blindness and is incurable. The mechanism underlying USH2, especially retinal degeneration, is unclear. To address this, we employed pull-down and immunoprecipitation approaches, coupled with LC-MS/MS, to explore the interactome of USH2 proteins. Our findings suggest that the USH2 protein complex connects extracellular matrix (ECM) constituents with the intracellular actin network, transduces signals via Gαi/Gαq, and participates in ECM remodeling, TGFβ signaling, cell adhesion, and ciliary function in photoreceptors.

Figures

Full text

Retinitis pigmentosa (RP) is an inherited retinal degenerative disease that is mainly caused by photoreceptor cell death, affecting about one in 4000 people worldwide (1). RP patients experience early night and peripheral vision loss and eventual legal blindness. Usher syndrome (USH) is an RP condition combined with hearing loss, which is the leading cause of inherited deaf-blindness and occurs in one in 23,000 people globally (1, 2, 3). Among the three USH clinical types, type 2 (USH2) is the most common, accounting for up to 70% of all USH cases (1). Currently, there is no cure for RP or USH.

USH2A, ADGRV1 (also known as GPR98, VLGR1, and MASS1), and WHRN are the three known causative genes for USH2 (4, 5, 6). Some of their mutations also cause nonsyndromic autosomal recessive RP (arRP) and hearing loss (7, 8, 9, 10, 11). Specifically, USH2A mutations account for 50 to 70% of all USH2 cases and are the major cause (∼8%) of arRP (1, 12). Usherin, encoded by USH2A, is a very long single-pass transmembrane protein with 5202 amino acids (aa) in humans (Fig. 1A). Approximately 97% of the protein is extracellular and consists of one laminin G-like (LGL) domain, one laminin N-terminal (LN) domain, 10 EGF laminin (LE) domains, two laminin G (LG) domains, and 32 fibronectin III (F) repeats (13, 14). Adhesion G protein–coupled receptor V1 (ADGRV1) is the largest adhesion G protein–coupled receptor (GPCR) with 6306 aa in humans (15). ADGRV1 also has an extremely large extracellular region, which comprises ∼95% of the entire protein and contains 21 tandem calcium exchange β repeats (Calxβ), six epilepsy-associated repeats (EARs), one epilepsy-associated/epitemptin-like domain (EPTP), one LG domain, and one GPCR autoproteolysis-inducing domain (GAIN) (Fig. 1A). Previous in vitro studies suggest that ADGRV1 may signal through coupling with Gα proteins. This activity may be regulated by ADGRV1 autoproteolysis at the GPCR proteolytic site (GPS) within GAIN (15, 16, 17, 18, 19). Whirlin, encoded by WHRN, is an intracellular protein with PDZ, harmonin-N like (HN), and proline-rich domains (Fig. 1A). Its PDZ domains bind to the intracellular C-terminal PDZ-binding motif (PBM) of both usherin and ADGRV1, which is thought to be important for regulating the WDSUB1-mediated ubiquitylation and degradation of usherin (13, 20, 21, 22). The three USH2-associated proteins (hereafter referred to as USH2 proteins) are localized at the periciliary membrane complex (PMC) in photoreceptor inner segments (ISs) (20, 21, 23, 24, 25) and at the ankle link complex (ALC) in hair cell stereocilia (26, 27), with ADGRV1 being the most critical protein for ALC integrity and function (28). Mouse genetic studies have shown that loss of one of the three proteins disrupts the PMC and the ALC, leading to progressive retinal degeneration and congenital hearing loss, respectively (21, 23, 24, 25, 29, 30).Fig. 1**Domain distribution of USH2 proteins and workflow diagram of this study.**A, functional domains in usherin, ADGRV1, and whirlin. Amino acid numbers refer to human proteins. Red lines show antigen regions of USH2 antibodies. Cyan lines indicate the usherin and ADGRV1 fragments used in this study. B, workflow for pull-down, immunoprecipitation, data analysis, and verification. Legends are in the bottom-left corner. ADGRV1, adhesion G protein–coupled receptor V1; USH2, Usher syndrome type 2.

Further studies have demonstrated that the ALC is critical for maintaining the staircase organization and integrity of hair cell stereocilia in the cochlea (28, 31, 32), while the function of the photoreceptor PMC in the retina remains elusive and may differ from the ALC’s role in hair cells, because the actin bundle-based hair cell stereocilia and the cilium-based photoreceptor outer segment are entirely different cellular structures. The PMC is thought to participate in trafficking rhodopsin and cone opsins from the photoreceptor IS to the OS, given its location between the two subcellular compartments (21). This idea has been supported by the mislocalization of rhodopsin and cone opsins observed in the retina of ush2a^u507^ and ush2a^rmc^ mutant zebrafish (33, 34), as well as Ush2a^−/−^ and Ush2a^delG^ mice (30, 35). In addition, usherin and ADGRV1 have been reported to associate with RAB8 (33, 36), a protein involved in rhodopsin transport in lower vertebrate retinas (37). However, the late onset of visual pigment mislocalization in the above-mentioned ush2a/Ush2a mutant zebrafish and mouse retinas, the normal distribution of visual pigments in ush2a^−/−^ zebrafish retinas (33), and the dispensable role of RAB8 in visual pigment transport in mouse retinas (38) indicate that the PMC is not essential for visual pigment transport in photoreceptors. ADGRV1 is proposed to transduce signals related to cell adhesion (15). An interactome analysis using various ADGRV1 C-terminal fragments as baits suggests that many focal adhesion components interact with ADGRV1 in hTERT-RPE1 cells (39), where ADGRV1 is a component of focal adhesions and can regulate cell spreading and migration (39). Another interactome analysis using the same group of ADGRV1 baits, however, did not identify focal adhesion components among the prey in HEK293T cells (40). Based on findings from these two studies in immortalized cell lines and primary cells, ADGRV1 is suggested to function in calcium homeostasis at the mitochondria-endoplasmic reticulum interface (41) and in autophagy (42). However, these cultured cell models express little endogenous USH2 proteins and differ markedly from photoreceptors in their morphology, structure, and function. In addition, the ADGRV1-interacting proteins identified in these cells have not been validated using orthogonal approaches. Therefore, the physiological relevance of these observations to the PMC's role in healthy and diseased photoreceptors remains uncertain.

In this study, we aimed to gain new insights into the retinal function of the PMC and the mechanisms underlying PMC-associated retinal degeneration by systematically screening proteins that interact with usherin, ADGRV1, and whirlin in bovine and mouse retinas (Fig. 1B). Because usherin and ADGRV1 are membrane-anchored proteins that are predominantly extracellular, we focused on identifying interacting proteins using functionally important extracellular fragments as baits for affinity purification. We also performed immunoprecipitation with antibodies against usherin, ADGRV1, and whirlin. We compared our results with previous reports on the ADGRV1 interactome from cultured cells (39, 40) and tested some of the identified proteins for orthogonal verification. We further conducted functional enrichment analyses of prey proteins captured from different affinity purifications. Our results demonstrate that usherin can self-interact through its laminin and F domains and that the extracellular regions of usherin and ADGRV1 directly interact. Our data further support the notion that the extracellular portion of the PMC associates with collagen fibers, participates in collagen and elastic fiber dynamics, and engages in transforming growth factor β (TGFβ) signaling; its intracellular portion contributes to cell adhesion and cilium-related cellular processes; and ADGRV1 transduces signals primarily through the Gαi and Gαq pathways at the plasma membrane. These findings provide, for the first time since the PMC was discovered nearly 2 decades ago, novel insights into its function in photoreceptors.

Experimental Procedures

Animals

Whrn knockout (Whrn^−/−^, also known as Whrn^neo/neo^ and Whrn^tm1Tili^) and Ush2a knockout (Ush2a^−/−^, also known as Ush2a^tm1Tili^) mice were described previously (21, 23). Adgrv1 knockout (Adgrv1^−/−^) mice were generated using a CRISPR/Cas9 strategy with an sgRNA targeting exon 82 (ATACACAGACATGTGTGAAC). Sanger sequencing confirmed a 191-base-pair deletion, consisting of 106 bp in exon 82 and 85 bp in intron 82, in this mouse model (Supplemental Fig. S1). All experiments involving animals were approved by the Institutional Animal Care and Use Committee at the University of Utah (Protocol number 00001537).

Plasmids

Full-length ADGRV1 cDNA was cloned into a pcDNA3.1(−) plasmid by multiple steps of restriction digestion and ligation of six ADGRV1 cDNA fragments that were synthesized by reverse transcription and polymerase chain reaction from mouse retinal total RNA (12–2883 bp, 2862–8845 bp, 8822–9897 bp, 9873–12104 bp, 12,081–14020 bp, 13,999–19062 bp, NM_054053). Usherin LGLNLE and F17 to 21 fragments (145–1035 aa and 3441–3854 aa, respectively, NP_067383) and ADGRV1 V2 fragment (2229–3635 aa, NP_473394) were cloned into pCEP-Pu plasmids and fused in-frame with an N-terminal BM40 (osteonectin, also known as SPARC) signal peptide and a C-terminal human IgG2 Fc fragment (hFc) (43). ADGRV1 V1, V1m, V2, V3, and V4 fragments (29–1661 aa, 1724–2319 aa, 2229–3636 aa, 3933–5049 aa, and 5632–5907 aa, respectively) were cloned into pCEP-Pu plasmids and fused in-frame with an N-terminal BM40 signal peptide and a C-terminal Flag tag. Usherin LNLE, F19 to 21, and F11 to 32 fragments (310–1035 aa, 3582–3854 aa, and 2521–4923 aa, respectively) were cloned into pSec-Tag2A-mFc-biotin plasmids and fused in-frame with an N-terminal mouse Igκ signal peptide and C-terminal mouse IgG2b Fc (mFc) and biotinylation signal (44). All the above DNA plasmids were confirmed by Sanger DNA sequencing. Flag- and HA-tagged mouse elastin microfibril interfacer 3 (EMILIN3), mouse USH2A full-length, and GST-tagged ADGRV1-C plasmids were described previously (14, 20, 45). Human collagen type VI alpha 1 chain (COL6A1) (82–3165 bp, NM_001848.3, OHu17965D), EPHA3 (129–3077 bp, NM_005233.6, OHu19238D), and MMP19 (154–1677 bp, NM_002429.5, OHu27465D) complementary DNA (cDNA) were synthesized and cloned in-frame with a C-terminal Flag tag in pcDNA3.1+/C-DYK plasmids by GenScript. The pDisplay plasmid with an N-terminal Igκ signal peptide, platelet-derived growth factor receptor (PDGFR) transmembrane fragment, c-Myc, and HA tag was purchased from Thermo Fisher Scientific. The γPCDH-A3-GFP plasmid (#197343) was obtained from Addgene.

Cells and Cell Transfection

HEK293 (American Type Culture Collection (ATCC): CRL1573; CVCL_0045) and COS-7 (ATCC: CRL1651; CVCL_0224) cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) fetal bovine serum, 100 units/ml penicillin, and 100 μg/ml streptomycin (Thermo Fisher Scientific). The 293-EBNA cells (ATCC: CRL-10852; CVCL_6974) were maintained in DMEM F12 medium supplemented with 10% (v/v) fetal bovine serum, 100 units/ml penicillin, 100 μg/ml streptomycin, and 250 μg/ml geneticin (Thermo Fisher Scientific). All cells were transfected with DNA plasmids using Lipofectamine 2000 (Thermo Fisher Scientific). The 293-EBNA cells transfected with pCEP-Pu plasmids were enriched by culturing in medium supplemented with 1 μg/ml puromycin (Thermo Fisher Scientific).

Antibodies

Rabbit and guinea pig polyclonal EMILIN3 antibodies, rabbit polyclonal usherin-N, usherin-C, ADGRV1-N, whirlin-N, and whirlin-C antibodies, and chicken polyclonal rootletin antibody were described in our previous reports (14, 28, 31, 45, 46). Rabbit polyclonal collagen VI antibody (ab6588) was purchased from Abcam Limited. Rabbit polyclonal HEXIM1 (15676-1-AP), SIPA1 (PA5-28122), and PTPN23 (PA5-100047) antibodies were obtained from Thermo Fisher Scientific. Rabbit polyclonal SIPA1 antibody (26793-1-AP) was acquired from Proteintech. Mouse monoclonal c-Myc (9E10) was brought from Takara Bio, and mouse monoclonal HA (HA-7) and Flag M2 antibodies were ordered from Millipore Sigma. Rabbit and chicken polyclonal ADGRV1-C antibodies were custom-made using a recombinant mouse antigen (6198–6298 aa) generated from BL21 cells, as described previously (28).

Affinity Purification Experiments

All the affinity purification experiments were conducted at 4 °C. Details of the experiments coupled with mass spectrometry are provided in Supplemental Table S1. ADGRV1 and usherin extracellular baits were expressed in HEK293 and 293-EBNA cells, two mammalian cell lines allowing proper posttranslational modifications, folding, and secretion of extracellular proteins. Because we did not observe any noticeable difference in extracellular protein expression in HEK293 and 293-EBNA cells, we selected the cell line based on availability at the time of each experiment. HEK293 cells were transfected with usherin LNLE-mFc, F19-21-mFc, and empty pSec-Tag2A-mFc-biotin plasmids. The conditioned culture media from cells transfected with F19-21-mFc and empty plasmids were cleared by gentle centrifugation twice. The transfected LNLE-mFc cells were lysed in lysis buffer (PBS, 1 mM DTT, 1% Triton X-100, and EDTA-free protease inhibitor cocktail). The cleared culture media and cell lysate were then incubated with protein G Sepharose 4FF (Thermo Fisher Scientific) for 2 h, washed four times with lysis buffer, and incubated overnight with adult bovine retinal lysates. The protein G Sepharose was subsequently washed four times with lysis buffer and eluted in 2X Laemmli protein loading buffer by vortexing and boiling. After verification of successful bait pull-down by immunoblotting (Fig. 9 in (14)), the eluted protein samples were subjected to SDS-PAGE and submitted for liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) analysis.

Subsequently, 293-EBNA cells were transfected with usherin F17-21-hFc, LGLNLE-hFc, ADGRV1 V2-hFc, V1-Flag, V1m-Flag, or V4-Flag plasmid. The conditioned culture media from the transfected cells was adjusted to contain 50 mM Tris, 150 mM NaCl, 0.5% Triton X-100, and 1X protease inhibitor cocktail. For the hFc-tagged proteins, the conditioned media were incubated with protein G Sepharose by gently shaking overnight. Human γ-globulin (Jackson ImmunoResearch Laboratories Inc) was included as a negative control. The protein G beads were washed four times, incubated with bovine retinal lysates by end-over-end mixing overnight, washed four times, and eluted by vortexing and boiling in Laemmli protein loading buffer. For the Flag-tagged proteins, the conditioned media were incubated with anti-Flag affinity resin (Pierce, Thermo Fisher Scientific) by gently shaking. 3X Flag peptide (Pierce, Thermo Fisher Scientific) in the same culture medium was included as a negative control. Other procedures were similar to the Fc pull-down experiments.

The GST-tagged ADGRV1 intracellular C fragment was expressed in and purified from One Shot BL21 Star (DE3) cells (Thermo Fisher Scientific), a cost-effective and high-yield bacterial expression system. The purified protein was incubated with bovine retinal lysate and Pierce Glutathione Superflow agarose (Thermo Fisher Scientific). GST protein expressed from the empty pGEX4t-1 plasmid in the same BL21 cells was used as a negative control. The precipitated proteins were electrophoresed on 10% SDS polyacrylamide gels. The gels were cut into pieces and submitted for LC-MS/MS analysis. More details can be found in (47).

To verify potential protein interactions, pull-down assays were performed 24 h posttransfection in HEK293 or 293-EBNA cells. Depending on protein localization, conditioned culture media were collected, or cells were lysed in lysis buffer. The culture media and cell lysates were cleared by centrifugation at 13,500 g for 10 min at 4 °C. The subsequent procedures were similar to those described above for mFc, hFc, and Flag pulldowns.

Immunoprecipitation

Mouse and bovine retinas were homogenized in lysis buffer and subjected to immunoprecipitation using rabbit ADGRV1-N, ADGRV1-C, usherin-C, whirlin-N, and whirlin-C antibodies as described previously (27, 31, 47). Nonimmune immunoglobin G (IgG), Whrn^−/−^, Ush2a^−/−^, and Adgrv1^−/−^ retinas were used as negative controls.

Immunoblotting, Immunostaining, and Colocalization Analysis

Immunoblotting and immunostaining procedures for human and mouse retinas and cultured cells were performed as described in our previous publications (27, 47). Human retinal sections were obtained from a tissue bank, fully de-identified, and exempt from Institutional Review Board review. Immunofluorescence images were captured on a Leica SP8 confocal microscope with a HC PL APO 63X 1.40 OIL CS objective. Colocalization of proteins double-transfected in COS-7 and HEK293 cells was quantified on single-scan two-dimensional (2D) confocal images using the FIJI plugin, BIOP JACoP (48). The whole-cell region was defined as the region of interest (ROI), unless otherwise specified. Colocalization analysis on z-stacked three-dimensional (3D) confocal images was performed using Imaris (10.2.0). A 3D surface of the entire single cell was created and defined as an ROI. The threshold for each channel within the ROI was determined using the automatic thresholding algorithm (Costes). Pearson’s correlation coefficient was calculated within the ROI area for 2D images and within the ROI volume for 3D images. Because no significant difference was observed in Pearson’s correlation coefficient calculated from single-scan and z-stacked images, single-scan images were used for most colocalization analysis in this study.

Sample Preparation and LC-MS/MS Analysis

The procedures were performed by the Taplin Mass Spectrometry Facility at Harvard Medical School. Excised gel bands were cut into approximately 1 mm^3^ pieces. Gel pieces were processed using a modified in-gel trypsin digestion protocol (49), washed, dehydrated with acetonitrile for 10 min, followed by acetonitrile removal, and thorough drying in a SpeedVac. Rehydration was performed with 50 mM ammonium bicarbonate containing 12.5 ng/μl of modified sequencing-grade trypsin (Promega) at 4 °C. After 45 min, the excess trypsin solution was removed and replaced with 50 mM ammonium bicarbonate to cover the gel pieces. The samples were then incubated at 37 °C overnight. Peptides were extracted by removing the ammonium bicarbonate solution, followed by one wash with a solution containing 50% acetonitrile and 1% formic acid. The extracts were dried in a SpeedVac (∼1 h) and stored at 4 °C until analysis.

On the day of analysis, the samples were reconstituted in 5 to 10 μl of HPLC solvent A (2.5% acetonitrile and 0.1% formic acid). A nanoscale reverse-phase HPLC capillary column was prepared by packing 2.6 μm C18 spherical silica beads into a fused silica capillary column (100 μm inner diameter, approximately 30 cm in length) with a flame-drawn tip (50). After equilibrating the column, each sample was loaded onto the column using a Famos autosampler (LC Packings). A gradient was generated, and peptides were eluted with increasing concentrations of solvent B (97.5% acetonitrile and 0.1% formic acid). The sample was loaded for 12 min at 100% solvent A, after which the gradient began at 5% solvent B and increased to 30% over 60 min. This was followed by a 3-min column wash in 95% solvent B and a 7-min re-equilibration in 100% solvent A.

As peptides eluted, they were subjected to electrospray ionization and then introduced into a Velos Orbitrap Pro ion-trap mass spectrometer (Thermo Fisher Scientific). Peptides were detected, isolated, and fragmented to generate a tandem mass spectrum of specific fragment ions. The full scan in positive mode was set to a resolution of 60,000 and collected in the Orbitrap. Data were acquired in centroid mode with a scan range from 360 to 1250. Twenty-two data-dependent MS2 scans were collected in rapid-scan mode using the ion trap. The top 22 peaks from the full scan were selected with a minimum threshold of 250 counts. Dynamic exclusion was enabled with a repeat count of 1, a duration of 15 s, and an exclusion size of 100.

Peptide sequences (and thus protein identity) were determined by matching the acquired fragmentation pattern to protein databases using the software Sequest (version 28, revision 13, Thermo Fisher Scientific) (51). The mouse protein database contained 48,494 total entries (24,247 forward and 24,247 reverse; https://www.uniprot.org/taxonomy/10090 - downloaded on 4/20/2015). The bovine protein database contained 167,290 total entries (83,645 forward and 83,645 reverse; https://www.uniprot.org/taxonomy/9913, downloaded on 11/9/2015). The data were filtered to achieve peptide and protein false discovery rates of 1%-2%. The maximum number of allowed missed internal cleavages was set to two. The mass tolerance for precursor ions and fragment ions was set to 100 ppm and 1.0 Da, respectively. The fixed and variable modifications were 57.0215 on cysteine (iodoacetamide) and 15.9949 on methionine (oxidation), respectively.

Identification of High-Confidence Interacting Proteins and Functional Enrichment Analysis

Prey proteins pulled down by baits and identified by LC-MS/MS were compared with nonspecific proteins pulled down in negative controls. Only proteins absent in the negative control pools were selected for further analysis. The physical interactions among the prey proteins were identified based on the STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) knowledgebase in Cytoscape (version 3.10.3). The interacting network of prey proteins captured by ADGRV1 C bait was further clustered using the Leiden network clustering algorithm with a 0.5 resolution parameter, 10 iterations, and default parameters in ClusterMaker2 (52). These clusters were then ranked based on the total peptide counts identified for proteins within each cluster using the Multiple Attribute Additive Method (MAA) algorithm with basic normalization. For ADGRV1 and usherin extracellular baits, we focused on candidate interacting proteins localized to the extracellular space. These proteins were identified using functional annotations from DAVID (Database for Annotation, Visualization, and Integrated Discovery) (53), applying keywords such as “extracellular” (excluding extracellular exosome), “transmembrane”, “plasma membrane”, and “cell surface”, as well as by referencing the matrisome gene set (54).

Protein intensities detected by LC-MS/MS in both experimental and negative control groups after immunoprecipitation were analyzed using Perseus v2.0.7.0 (55). Total protein intensity was first calculated for each biological sample. The means of these total protein intensities across different batches of immunoprecipitation experiments were used to normalize batch effects. Missing values were imputed as half the minimum intensity observed across all normalized data. In Perseus, log_2_-transformed data were used to generate volcano plots for ADGRV1-, usherin-, and whirlin-interacting proteins using the Hawaii plot function, with the corresponding control groups as negative controls. High-confidence interacting proteins were defined as proteins with a p value less than 0.05 in a post hoc Tukey’s Honestly significant difference (HSD) test following one-way ANOVA across the ADGRV1, usherin, whirlin, and control groups. Note that several ADGRV1 peptides were detected in the immunoprecipitate from Adgrv1^−/−^ retinas, indicating that the Adgrv1^−/−^ mouse may not be Adgrv1-null. We thus excluded the Adgrv1^−/−^ sample as a negative control in the Perseus analysis.

Functional enrichment analyses were performed with either the entire human genome or all immunoprecipitated proteins as the reference background, using the STRING Enrichment App in Cytoscape 3.10.3 (56), DAVID tools (53), and ClusterProfiler 4.14.4 (57). These tools and background sets produced highly similar enrichment results. Consequently, ClusterProfiler was used to visualize the results in figures. To reduce redundancy from pathways or terms driven by substantially overlapping proteins, a single representative enriched pathway or term was manually selected from each clade in the clustering plots (Supplemental Figs. S2 and S3), based on relevance to photoreceptor biology. The pathway enrichment comparison across different baits was also performed using ClusterProfiler, with a minimum cutoff of two proteins per pathway/term.

Experimental Design and Statistical Rationale

For each bait, antibody, or species, a single biological sample and a single technical run were performed (Fig. 1B). Experiments using closely related baits and antibodies were treated as biological replicates, such as usherin LGLNLE and LNLE baits, usherin F19 to 21 and F17 to 21 baits, N- and C-terminal ADGRV1 antibodies, N- and C-terminal whirlin antibodies, and the same antibodies applied to mouse and bovine retinas. Affinity purification experiments were performed once for each ADGRV1 bait (C, V1, V1m, V2, and V4) and once for each tag-only negative control (GST, 3X Flag, and human γ-globulin) (Fig. 1B). The V1, V1m, and V2 baits shared Calxβ repeats (Fig. 1A), allowing identification of Calxβ-binding proteins. The usherin LNLE and LGLNLE baits, as well as the F19 to 21 and F17 to 21 baits, contained large overlapping regions (Fig. 1A), so these baits were tested once using mFc from the pSec-Tag2A-mFc-biotin plasmid and human γ-globulin as negative controls. Selected prey proteins were validated in cultured cells using pull-down assays with at least two biological replicates and colocalization analysis in approximately 10 cells, including corresponding negative controls. Pearson correlation coefficients from colocalization assays were compared using one-way ANOVA followed by Dunnett’s or Tukey’s multiple comparison test using GraphPad Prism (version 10.6.1, www.graphpad.com).

Immunoprecipitation experiments were performed using whirlin antibodies twice in mouse retinas and once in bovine retinas; ADGRV1 antibodies once in mouse retinas and twice in bovine retinas; and the usherin-C antibody once each in mouse and bovine retinas (Fig. 1B). These experiments were performed by three laboratory members. Negative controls included rabbit γ-globulin applied to mouse retinas three times, whirlin antibodies applied to Whrn^−/−^ mouse retinas twice, and the usherin-C antibody applied to Ush2a^−/−^ mouse retinas once. To identify proteins coimmunoprecipitated with specific USH2 proteins, one-way ANOVA followed by Tukey’s HSD post hoc test was conducted using Perseus v2.0.7.0 (55). Several high-confidence interacting protein candidates were validated in cultured cells using pull-down assays and colocalization analysis, as mentioned above, or in mouse retinas using immunoprecipitation (two biological replicates) and immunostaining assays.

Results

Identification of Intracellular ADGRV1-Interacting Proteins

ADGRV1 intracellular C-terminal region has a PDZ-binding motif (Fig. 1A) and binds PDZ proteins, whirlin, harmonin (USH1C), and PDZD7 (20, 21, 58, 59, 60). To identify more proteins interacting with this region, we performed a pull-down experiment from bovine retinal lysate using a GST-fused ADGRV1 C fragment as bait (Fig. 1). LC-MS/MS identified 135 proteins as potential intracellular ADGRV1-binding partners after subtraction of the nonspecific proteins pulled down by negative control GST protein (Supplemental Table S2). Whirlin and harmonin were among the identified 16 PDZ proteins (Supplemental Table S2), indicating the success of our experiment. Among the identified proteins, 73 had known direct interactions with each other according to the STRING human database (61). After breaking down the interaction network of these proteins into small groups by network clustering (52), actin-based cell projection proteins, EZR, SLC9A3R, RDX, and MPP1, were ranked the strongest binding partners considering their detected peptide numbers (Fig. 2A). In addition, proteins in the chaperone-containing TCP-1 (CCT) complex, Bardet-Biedl syndrome complex (BBSome), dystrophin-associated glycoprotein complex, dynein complex, heat shock 70 regulation, homer protein-binding, and septin complex were identified (Fig. 2A).Fig. 2**Proteins pulled down from bovine retinas by the ADGRV1 C intracellular bait.**A, clusters of prey proteins with known physical interactions based on the STRING database. B, prey proteins shared with previously reported ADGRV1 interactomes (39, 40). Left, Venn diagram of prey proteins identified in this study and prey proteins identified previously by the ICD bait in HEK293T cells (HEK ICD) (40), all baits other than ICD in HEK293T cells (HEK others) (40), and all baits other than ICD in hTERT-RPE1 cells (RPE others) (39). Right, the list of prey proteins shared between this and previous studies, with their known physical interactions. C, GO enrichment (left) and heatmap (right) results of the prey proteins shared with previously reported ADGRV1 interactomes. ADGRV1, adhesion G protein–coupled receptor V1; GO, gene ontology; ICD, intracellular domain; STRING, search tool for the retrieval of interacting genes/proteins.

During our study, candidate ADGRV1-interacting proteins were identified in HEK293 and hTERT-RPE1 cells using tandem affinity purification with Strep II and Flag tags and mass spectrometry-based proteomics (39, 40). In the reports, the intracellular domain (ICD) bait was an ADGRV1 intracellular fragment, similar to our ADGRV1 C bait. Other baits contained the C-terminal fragment (CTF) generated after ADGRV1 autocleavage at the GPS and a relatively longer CTF that extended into the extracellular region (Fig. 1A). Compared with the published proteins pulled down by the ICD bait (40), six proteins (SNX27, TXN, HSPA6, NHERF2, HSPA1L, CAD, and RUVBL1) were shared, besides whirlin and harmonin (Fig. 2B and Supplemental Table S2). Another 28 proteins were shared when considering the proteins pulled down by other ADGRV1 baits (Fig. 2B and Supplemental Table S2). Gene ontology (GO) enrichment analysis on the total shared 37 proteins demonstrated that the CCT complex, proteasome complex, regulation of protein catabolic process, septin ring, and cadherin-binding proteins were overrepresented (Fig. 2C). This result was generally consistent with the above observation from all 135 ADGRV1 prey proteins identified in this study (Fig. 2A). Taken together, our data suggest that ADGRV1 functions in cadherin and actin-mediated cell adhesion, dynein and BBSome-mediated intracellular trafficking, as well as chaperone and proteasome-involved protein folding and homeostasis.

Identification of Extracellular ADGRV1-Interacting Proteins

The ADGRV1 intracellular region accounts for only a small portion of the protein (∼5%). It is thus essential to identify extracellular ADGRV1-interacting partners to understand the protein’s function and mechanism of action. The ADGRV1 extracellular region contains numerous repetitive calcium-binding Calxβ domains interspersed with one LG domain, EAR/EPTP repeats, and one GAIN domain (Fig. 1A). The missense mutations identified in patients were distributed throughout the entire ADGRV1 extracellular region (Human Gene Mutation Database). We therefore cloned the baits, Flag-tagged V1, V1m, and V4 fragments, and hFc-tagged V2 fragment, to cover the Calxβ, LG, EAR/EPTP, and GAIN domains (Fig. 1A). These extracellular baits were expressed in 293-EBNA cells to maintain their proper posttranslational modification, folding, and secretion. The baits were then purified from the conditioned culture media through Flag and protein G pull-down and incubated with bovine retinal lysates. 3X Flag peptide and human gamma immunoglobulin were included as negative controls (Fig. 1B). After LC-MS/MS, 23, 48, 67, and 31 extracellular and plasma membrane proteins, based on the DAVID bioinformatics knowledgebase (53) and the matrisome gene set (54), were identified for the V1, V1m, V2, and V4 baits, respectively (Fig. 3A and Supplemental Table S3). In total, 120 proteins were pulled down by the four baits. Among them, 32 proteins were pulled down by more than two baits. Specifically, PLOD1 (Procollagen-lysine, 2-oxoglutarate 5-dioxygenase 1), PLOD2, and PCOLCE2 (Procollagen C-endopeptidase enhancer 2) were shared among all four baits (Fig. 3A and Supplemental Table S3).Fig. 3Proteins pulled down from bovine retinas by ADGRV1 extracellular baits. A, Venn diagram of ECM and plasma membrane proteins pulled down by the four ADGRV1 extracellular baits. The total number of prey proteins from each bait is shown next to the bait name, and the number of prey proteins unique to each bait is indicated within the diagram. Proteins shown in red are shared with previous ADGRV1 interactome studies (39, 40). Prey proteins detected by more than one bait, as well as those detected by only one bait but shared with previous reports, are listed. B, Venn diagram of ECM and plasma membrane prey proteins shared between this and previous studies (39, 40). C, GO and Reactome pathways/terms enriched in the ECM and plasma membrane proteins pulled down by all four ADGRV1 extracellular baits. D, prey proteins contributing to the enrichment of the most significant and biologically relevant GO and Reactome pathways/terms. The color intensity reflects the confidence level of each potential ADGRV1-interacting protein, based on replication with previous interactome studies, consistency among ADGRV1 extracellular baits, and the number of peptides detected in our study (darker shading indicates higher confidence; details are in Supplemental Table S3). The blue border of the nodes indicates that the prey proteins have been previously reported (39, 40). Normalized peptide counts for each protein across the four ADGRV1 baits are shown as a bar chart inside the node. The legend is at the bottom center. Gray lines connecting nodes indicate known physical interactions from the STRING knowledgebase. ADGRV1, adhesion G protein–coupled receptor V1; ECM, extracellular matrix; GO, gene ontology; STRING, search tool for the retrieval of interacting genes/proteins.

Pathway enrichment analysis of the 120 prey proteins using the human genome as a reference background showed that extracellular matrix (ECM) organization was the top enriched term in both GO and Reactome pathways, as expected (Fig. 3C). In addition, peptidase regulation, elastic fiber formation, collagen fibril enzymes, and retinoid metabolism and transport were among the significantly enriched terms (Fig. 3C and Supplemental Table S5). These terms were from different clades of the GO and Reactome enrichment clustering plots (Supplemental Fig. S2), indicating that the prey proteins contributing to these terms did not substantially overlap. PLOD1, PLOD2, PLOD3, and PCOLCE2 were collagen biosynthesis and modifying enzymes (Fig. 3D). Eighteen prey proteins belonged to the regulation of peptidase activity pathway (Fig. 3D), suggesting that ADGRV1 may play a role in ECM remodeling and signal molecule processing. Eleven prey proteins were known to participate in the formation of the crucial ECM elastic fibers, which provide tissue elasticity and resilience (Fig. 3D). Finally, six prey proteins were associated with retinoid metabolism and transport (Fig. 3D), although their biological implication in the retinal cycle between photoreceptors and retinal pigment epithelium is uncertain.

In the above-mentioned ADGRV1 interactome studies (39, 40), several baits containing extracellular portions were utilized. We thus compared our 120 prey proteins with those reported using these baits. Thirty-one proteins discovered in our experiment were also reported (Fig. 3B and Supplemental Table S3). Pathway enrichment analysis of these proteins identified GO and Reactome pathways similar to those obtained from all 120 proteins, including regulation of peptidase activity, collagen biosynthesis/modifying enzymes, and elastic fiber formation. The shared proteins are labeled in these pathways in Figure 3D (blue border). Notably, PLOD1, PLOD2, and PLOD3 proteins were not only shared among our ADGRV1 extracellular baits but also with the reported baits (Fig. 3B) (39, 40).

Identification of Extracellular Usherin-Interacting Proteins

To efficiently identify functionally important extracellular usherin-interacting proteins, we focused on the regions where pathogenic usherin missense mutations were concentrated. Besides the usherin LE5 domain, which hosts the most common RP mutation p.C759F, we previously showed that the usherin LN and F18 domains harbor the most diverse types of pathogenic missense mutations in patients (14). We thus expressed and purified the hFc-tagged usherin LGLNLE and F17 to 21 fragments from 293-EBNA cells, as well as the mFc-tagged LNLE and F19 to 21 fragments from HEK293 cells, which are similar to 293-EBNA cells (Fig. 1B). After Fc pulldown and LC-MS/MS experiments, we identified 16, 15, 18, and 7 extracellular and plasma membrane proteins from bovine retinal lysates for the four respective usherin baits (Fig. 4A and Supplemental Table S4). mFc and hFc fragments were included as negative controls (Fig. 1B). Interestingly, EPHA3 (ephrin type-A receptor 3) and MMP19 (matrix metalloprotease 19) were pulled down by both LGLNLE and LNLE baits, while ADGRV1 was isolated using both F17 to 21 and F19 to 21 baits (Fig. 4A), indicating that these prey proteins bind uniquely to the LNLE and FN3 regions of usherin, respectively.Fig. 4Proteins pulled down from bovine retinas by usherin extracellular baits. A, Venn diagram of ECM and plasma membrane proteins pulled down by the four usherin extracellular baits. The total number of prey proteins from each bait is listed in the bracket after the bait name. B, GO and Reactome pathways/terms enriched in the ECM and plasma membrane proteins identified by all four usherin baits. C, prey proteins contributing to the enrichment of the most significant and biologically relevant GO and Reactome pathways/terms. The color intensity reflects the confidence level in the potential usherin-interacting proteins, based on consistency among the usherin extracellular baits and the number of peptides detected in our study (darker shading indicates higher confidence; details are in Supplemental Table S4). Normalized peptide counts are displayed as a bar chart within the prey protein nodes for the four usherin baits. The legend is at the top right. Gray lines connecting nodes indicate known physical interactions from the STRING knowledgebase. ECM, extracellular matrix; GO, gene ontology; STRING, search tool for the retrieval of interacting genes/proteins.

We then conducted pathway enrichment analysis on all 47 identified extracellular and plasma membrane proteins potentially associated with usherin, using the human genome as a reference background. As expected, ECM organization was the most significantly enriched term in both GO and Reactome pathway analyses (Fig. 4B). In addition, response to mechanical stimulus, PI3K/PKB signal transduction, cell-substrate adhesion, collagen metabolic process, and calcium-dependent protein binding were identified as nonredundant, functionally related enriched terms in the GO and Reactome clustering plots (Figs. 4B, S3, and Supplemental Table S5). The proteins contributing to some of the enriched terms were shown in Figure 4C. ADGRV1, which was pulled down by usherin FN3 baits, was annotated to the pathway of response to mechanical stimulus. EPHA3 and MMP19, which potentially interact with usherin LNLE, were associated with cell-substrate adhesion and collagen metabolic process, respectively.

Comparison of Enriched Pathways Across ADGRV1 and Usherin Extracellular Regions

As components of the same protein complex in photoreceptors and hair cells, the extracellular regions of ADGRV1 and usherin are likely to participate in similar cellular pathways to support the overall function of the complex. However, the distinct extracellular domains of ADGRV1 and usherin indicate that the proteins contribute to the overall function in different ways. To explore the potential shared and unique roles of ADGRV1 and usherin extracellular domains, we conducted GO and Reactome enrichment analyses on the proteins pulled down by individual ADGRV1 and usherin baits (Fig. 5A and Supplemental Table S6). Because of the significant overlapping regions of the usherin baits (Fig. 1A), we combined the prey proteins from the LGLNLE and LNLE baits (L baits) and the prey proteins from the F17 to 21 and F19 to 21 baits (F baits). We discovered that all ADGRV1 and usherin baits bound glycosaminoglycans and sulfur compounds, which are characteristic of ECM proteins. The prey proteins from all four ADGRV1 baits were linked to peptidase regulator/inhibitor activity, collagen biosynthesis/modifying enzymes, and elastic fiber formation. Given that three of the ADGRV1 baits contained calxβ repeats, this finding suggests that calxβ repeats mediate these cellular pathways. The usherin F baits shared more enriched terms with the four ADGRV1 baits than the usherin L baits, indicating that usherin FN3 repeats work in conjunction with ADGRV1 in the extracellular environment.Fig. 5**Comparison of GO and Reactome pathways/terms associated with different ADGRV1 and usherin extracellular baits.**A, the enrichment of GO and Reactome pathways/terms in prey proteins found by different ADGRV1 and usherin extracellular baits. B, the prey proteins contributing to the GO and Reactome pathways/terms that are unique to specific ADGRV1 and usherin baits (headings). Bar charts of normalized peptide counts across different baits are shown inside the node of prey proteins. The bait legend is on the right. Gray lines connecting nodes indicate known physical interactions from the STRING knowledgebase. ADGRV1, adhesion G protein–coupled receptor V1; GO, gene ontology; STRING, search tool for the retrieval of interacting genes/proteins.

Interestingly, both GO and Reactome analyses revealed that the ADGRV1 V2 bait was uniquely associated with cytokine activity, particularly the TGFβ signaling (Fig. 5A). The contributing proteins were TGFB1, TGFB2, LTBP1, LTBP4, BMP4, BMP7, BMP8A, WNT5B, WNT11, and others (Fig. 5B). The ADGRV1 V2 bait was also linked to serine-type endopeptidase activity (Fig. 5, A and B). These bait-specific enriched pathways were probably attributed to the unique EAR/EPTP repeats in the ADGRV1 V2 bait. The ADGRV1 V4 bait exhibited exclusive enrichment for terms related to cysteine-type endopeptidase inhibitor activity, insulin receptor signaling, and basigin interactions (Fig. 5, A and B), suggesting that the ADGRV1 GAIN domain may play a role in these cellular pathways. We did not identify unique enriched pathways for the ADGRV1 V1 bait, likely because the folding of the V1 bait masked the LG domain. Usherin L baits displayed enriched terms of cell adhesion mediator activity, S100 protein binding, and phosphatidylinositol bisphosphate binding, which were not observed for ADGRV1 or usherin F baits (Fig. 5A). Notably, 12 out of the 29 prey proteins from usherin L baits were associated with cell adhesion (GO:0007155, Fig. 5B).

Identification of ADGRV1-, Usherin-, and Whirlin-Interacting Proteins by Immunoprecipitation and Mass Spectrometry

Considering that immunoprecipitation and tag pull-down approaches can identify complementary and overlapping groups of interacting proteins, we immunoprecipitated endogenous ADGRV1, usherin, and whirlin proteins, along with their interacting proteins, using ADGRV1-N, ADGRV1-C, usherin-C, whirlin-N, and whirlin-C antibodies (Figs. 1 and 6A). The specificity of these antibodies has been verified in mutant mice in our previous studies (28, 31). We performed immunoprecipitation using bovine and mouse retinal lysates, as the bovine retina is large in size, and the mouse retina has Adgrv1^−/−^, Ush2a^−/−^, and Whrn^−/−^ retinas as stringent negative controls. In addition, nonimmune rabbit immunoglobulin was included as another negative control (Figs. 1B and 6A).Fig. 6Proteins coimmunoprecipitated with USH2 proteins from retinas. A, diagram of the experimental design. Immunoprecipitation was performed on mouse (m) and bovine (b) retinas using whirlin-N (Wn), whirlin-C (Wc), usherin-C (U), ADGRV1-N (An), and ADGRV1-C (Ac) antibodies, as well as nonimmunoglobulin (IgG), by three different people. W/H, wild-type and USH2 heterozygous mutant retinas. W, wild-type bovine retinas. B, volcano plots of proteins coimmunoprecipitated with ADGRV1 (left), usherin (middle), and whirlin (right). Proteins with a p value less than 0.05 and Log_2_FC (fold change) greater than 2 are shown as yellow dots (t test). Proteins marked with red dots are considered significant based on one-way ANOVA and post hoc Tukey’s HSD test (p < 0.05) (Supplemental Table S7). RP proteins, Gα proteins, and proteins involved in nuclear function are labeled in red, purple, and green, respectively. Proteins that were further tested by orthogonal approaches are labeled in cyan. C, GO enrichment results from the high-confidence ADGRV1-, usherin-, and whirlin-interacting proteins, as indicated by red dots in (B). D, heatmaps of proteins contributing to the top five enriched GO pathways/terms for ADGRV1, usherin, and whirlin. ADGRV1, adhesion G protein–coupled receptor V1; GO, gene ontology; RP, retinitis pigmentosa; USH2, Usher syndrome type 2.

After LC-MS/MS analysis, 4065 proteins were detected in the immunoprecipitates (Supplemental Table S7). As expected, ADGRV1, usherin, and whirlin were successfully immunoprecipitated by their corresponding antibodies (Fig. 6B). However, these proteins did not coimmunoprecipitate with each other, suggesting that the USH2 complex may assemble dynamically through relatively weak interactions, consistent with the signaling role of ADGRV1 (22). One-way ANOVA followed by Tukey’s HSD multiple-comparison tests identified 124, 26, and 7 proteins as high-confidence ADGRV1-, usherin-, and whirlin-interacting proteins, respectively, compared with negative controls (Fig. 6B and Supplemental Table S7). Among the 124 high-confidence ADGRV1-interacting proteins, 10 were shared with the 135 prey proteins captured by the ADGRV1 C bait, and an additional 10 were shared with the 120 proteins captured by all ADGRV1 V baits (Supplemental Table S7). Notably, the ECM protein EMILIN3 was pulled down by both ADGRV1-N and ADGRV1-C antibodies as well as the ADGRV1 V1m bait from bovine retinas (Figs. 3A and 6B). EMILIN3 was also present in the ADGRV1 interactome previously reported in HEK293 cells (Fig. 3B) (40). Several proteins implicated in inherited retinal degenerative diseases, i.e., BBS8 (62), KIAA1549 (63), and TUB (64), were each coimmunoprecipitated with ADGRV1, usherin, and whirlin, respectively (Fig. 6B), suggesting that the USH2 pathological mechanism may cross-talk with other known RP disease pathways. In addition, 2 Gα proteins were specifically coimmunoprecipitated with ADGRV1 from bovine (GNA11 and GNAI3) and mouse (GNA11) retinas, indicating these Gα proteins may mediate ADGRV1 signaling.

Further analysis using all the 4065 detected proteins as a reference background revealed the following enriched GO pathways/terms among the 124 high-confidence ADGRV1-interacting proteins: endoplasmic reticulum membrane, regulation of small GTPase signaling, negative regulation of mTORC1 signaling, and 7SK snRNA binding (Fig. 6, C and D, and Supplemental Table S7), suggesting these cellular events may be downstream of ADGRV1 signaling. Among the 26 high-confidence usherin-interacting proteins, the following enriched GO pathways/terms were identified: early endosome, negative regulation of cell cycle, cellular response to amino acid stimulus, RNA catabolic process, and photoreceptor connecting cilium (Fig. 6, C and D, and Supplemental Table S7). Finally, cilium and sensory perception of sound were the enriched GO pathways/terms among the seven high-confidence whirlin-interacting proteins (Fig. 6, C and D, and Supplemental Table S7).

Usherin Extracellular Self-Interaction and Interactions With ECM Proteins

Our previous homology modeling study on usherin extracellular structures predicts that usherin FN3 domains may interact with usherin LN and LE domains (14). This self-interaction could be a mechanism to regulate usherin interactions with other partners. We thus tested this self-interaction by double-transfecting usherin F11 to 32-mFc and LGLNLE-Flag plasmids into HEK293 cells. We found that usherin LGLNLE-Flag was specifically pulled down by usherin F11 to 32-mFc from the cell lysate (Fig. 7A), indicating that, besides the known self-interaction of the usherin intracellular region (20), the usherin N- and C-terminal extracellular regions can bind each other intramolecularly and/or intermolecularly.Fig. 7Interactions of usherin with itself, EPHA3, and MMP19. A, usherin F11 to 32-mFc in conditioned medium pulled down usherin LGLNLE-Flag from the lysate of their double-transfected HEK293 cells. B, EPHA3-Flag and MMP19-Flag were pulled down by usherin LGLNLE-hFc from the lysates of their respective double-transfected HEK293 cells. C, EPHA3-Flag partially, and MMP19-Flag nearly completely, colocalized with usherin LGLNLE-hFc in their double-transfected cells. A PDGFR transmembrane fragment (PDGFRf) in the pDisplay plasmid was used as a negative control. The scale bars represent 10 μm. D, Pearson’s correlation coefficients measuring colocalization of usherin LGLNLE with EPHA3, MMP19, and PDGFRf. One-way ANOVA followed by Tukey’s multiple comparisons test. ∗∗, adjusted p < 0.01; ∗, adjusted p = 0.012. EPHA3, ephrin type-A receptor 3; hFc, human IgG2 Fc fragment; mFc, mouse IgG2b Fc; MMP19, matrix metalloprotease 19; PDGFR, platelet-derived growth factor receptor.

We further verified the interactions of usherin LGLNLE with EPHA3 and MMP19, the two proteins pulled down by both usherin LNLE and LGLNLE baits in our bovine affinity purification experiments (Fig. 4A). Supportively, MMP19 was also immunoprecipitated by both ADGRV1-N and ADGRV1-C antibodies from bovine retinas (Fig. 6B and Supplemental Table S7). In HEK293 cells transfected with usherin LGLNLE and either EPHA3 or MMP19, we found usherin LGLNLE pulled down both EPHA3 and MMP19 (Fig. 7B). In addition, in HEK293 and COS-7 cells double-transfected with the respective constructs, MMP19 showed robust colocalization with usherin LGLNLE, whereas EPHA3 exhibited partial colocalization (Fig. 7, C and D). This result was consistently supported by the quantification of Pearson’s correlation coefficient on single-scan confocal images of HEK293 cells and z-stacked confocal images of COS-7 cells (Fig. 7D). Although we were unable to assess the colocalization of EPHA3 and MMP19 with usherin in mouse photoreceptors due to the lack of high-quality antibodies, our cell-based findings suggest that both EPHA3 and MMP19 interact extracellularly with usherin in photoreceptors.

Interactions of ADGRV1 With Usherin and EMILIN3

We identified ADGRV1 as a candidate interactor of usherin F19 to 21 and F17 to 21 baits in the bovine affinity purification experiment (Fig. 4A). Because of the known colocalization of ADGRV1 and usherin in the USH2 complex in photoreceptors and hair cells (3) and the absence of direct interaction between their intracellular regions (20), we examined whether ADGRV1 and usherin extracellular regions interacted and which domains were involved. We found that usherin F17 to 21 was able to pull down and colocalize with ADGRV1 V1, V1m, V2, V3, and V4 fragments in their double-transfected HEK293 cells (Fig. 8). Usherin F19 to 21 fragment consistently pulled down ADGRV1 V2 and V4 fragments from their double-transfected 293-EBNA cells (Supplemental Fig. S4, A and B). These results indicate that usherin and ADGRV1 can bind directly via their extracellular FN3 domains and Calxβ motifs, respectively, supporting our finding of shared associated pathways between these two specific regions (Fig. 5A).Fig. 8Interaction between usherin and ADGRV1 extracellular fragments. A, usherin F17-21-hFc in conditioned medium pulled down Flag-tagged ADGRV1 V1, V1m, V2, V3, and V4 fragments from the lysates of their respective double-transfected HEK293 cells. B, colocalization of usherin F17-21-hFc with the various Flag-tagged ADGRV1 extracellular fragments in their respective double-transfected HEK293 cells. A PDGFR transmembrane fragment (PDGFRf) expressed from the pDisplay plasmid was used as a negative control. The scale bars represent 10 μm. C, Pearson’s correlation coefficients measuring colocalization of usherin F17 to 21 with the various ADGRV1 V fragments and PDGFRf. One-way ANOVA followed by Dunnett’s multiple comparisons test. ∗∗, adjusted p < 0.01. ADGRV1, adhesion G protein–coupled receptor V1; hFc, human IgG2 Fc fragment.

EMILIN3 was identified as an ADGRV1-interacting candidate protein by ADGRV1 V1m bait and by both ADGRV1-N and -C antibodies from bovine retinas (Figs. 3A, 6B, and Supplemental Table S7). It was also reported previously in the ADGRV1 interactome from HEK293 cells (Fig. 3B), although its interaction with ADGRV1 has not been experimentally validated (40). We thus tested whether EMILIN3 interacted with ADGRV1 in double-transfected cultured cells. We found that EMILIN3 colocalized with ADGRV1 and usherin, but not with the two negative controls γPCDH-A3 transmembrane protein or PDGFR transmembrane fragment (PDGFRf), in the cytoplasm of COS-7 cells (Fig. 9, A and B). ADGRV1 V1m and V2 fragments pulled down EMILIN3 from the transfected HEK293 cell lysates (Fig. 9C). Using a rabbit EMILIN3 antibody, we found that EMILIN3 was at the USH2 complex position on the top of the ciliary rootlet (rootletin signal) in mouse photoreceptors (Fig. 9D). Consistently, we detected colocalization of EMILIN3 with both ADGRV1 and usherin in mouse and human photoreceptors using a second EMILIN3 antibody generated from guinea pig (Fig. 9D). Together, these findings strongly indicate that EMILIN3 interacts with ADGRV1 at the ECM around the periciliary membrane of photoreceptors.Fig. 9Interaction between ADGRV1 and EMILIN3. A, EMILIN3 colocalized with ADGRV1 and usherin but not with γPCDH-A3 transmembrane protein or PDGFR transmembrane fragment (PDGFRf) in COS-7 cells. The scale bars represent 10 μm. B, Pearson’s correlation coefficients measuring the colocalization of EMILIN3 with ADGRV1, usherin, and γPCDH-A3 in the cytoplasm of COS-7 cells. One-way ANOVA followed by Dunnett’s multiple comparisons test. ∗∗, adjusted p < 0.01. C, ADGRV1 V1m and V2 fragments pulled down EMILIN3 from the lysates of their respective double-transfected HEK293 cells. D, EMILIN3 colocalized with ADGRV1 and usherin at the top of the ciliary rootlet in mouse and human photoreceptors. EMILIN3 rabbit (rb) and guinea pig (gp) antibodies were used. The scale bars represent 5 μm. ADGRV1, adhesion G protein–coupled receptor V1; EMILIN3, elastin microfibril interfacer 3.

Verification and Localization of High-Confidence Interacting Proteins in Mouse Photoreceptors

To verify additional USH2-interacting proteins in photoreceptors beyond EMILIN3, we performed immunostaining and coimmunoprecipitation experiments with other candidate proteins for which suitable antibodies were available. COL6A1 was identified in the prey proteins from ADGRV1 V1m and V2 baits (Fig. 3A) and usherin F17 to 21 bait (Fig. 4A). However, we could not pull down COL6A1 using these ADGRV1 or usherin baits in their double-transfected HEK293 cells. Collagen VI is known to be assembled into a triple helix from three different alpha chains (65, 66). Therefore, our results suggest that USH2 proteins do not interact with individual unassembled COL6A1. We then examined COL6A1 distribution in mouse retinas by immunostaining. COL6A1 was colocalized with ADGRV1 at the periciliary region in photoreceptors (Fig. 10A). Since the ADGRV1 V1m bait pulled down both COL6A1 and COL6A2 (Fig. 3D) and collagen VI always consists of these two and a third variable alpha chains (65), our findings suggest that ADGRV1 interacts and colocalizes with the assembled native collagen VI.Fig. 10Localization and coimmunoprecipitation of high-confidence USH2-interacting proteins in mouse photoreceptors. A–D, localization of COL6A1 (A), HEXIM1 (B), SIPA1 (C), and PTPN23 (D) in mouse photoreceptors relative to ADGRV1. For each panel, the right side shows single-channel images corresponding to the boxed region in the merged images. For COL6A1 and HEXIM1, intensity profile plots were generated along a line intersecting the ADGRV1 signal at the three labeled positions. OS, outer segment; CC, connecting cilium; IS, inner segment; ONL, outer nuclear layer. The scale bars represent 5 μm. E, representative immunoblots showing that SIPA1 coimmunoprecipitates with usherin and that EIF2B4 coimmunoprecipitates with ADGRV1. ADGRV1, adhesion G protein–coupled receptor V1; COL6A1, collagen type VI alpha 1 chain; HEXIM1, hexamethylene bis-acetamide-inducible protein 1; PTPN23, protein tyrosine phosphatase non-receptor type 23; SIPA1, signal-induced proliferation-associated 1.

Hexamethylene bis-acetamide-inducible protein 1 (HEXIM1) regulates RNA polymerase II activity by binding to and inactivating the p-TEFb complex, which consists of CDK9 and CCNT1, in the nucleus (67). HEXIM1, CDK9, and CCNT1 were all coimmunoprecipitated with ADGRV1 from bovine and mouse retinas (Fig. 6B). We therefore examined the distribution of HEXIM1 in mouse photoreceptors using immunostaining. HEXIM1 was detected in the photoreceptor cell body and within the euchromatin region of the nucleus (Fig. 10B). A weak HEXIM1 signal was also observed at the periciliary region, where it colocalized with ADGRV1 (Fig. 10B). However, we could not detect coimmunoprecipitation between HEXIM1 and ADGRV1 by immunoblotting. These findings suggest that a small fraction of HEXIM1 interacts weakly with ADGRV1 at the USH2 complex in vivo.

We next examined signal-induced proliferation-associated 1 (SIPA1), EIF2B4 (translation initiation factor eIF2B subunit delta), and PTPN23 (Protein tyrosine phosphatase nonreceptor type 23). Mass spectrometry identified PTPN23 in complexes coimmunoprecipitated with ADGRV1 and usherin, whereas EIF2B4 and SIPA1 were coimmunoprecipitated with ADGRV1 and usherin, respectively, from both bovine and mouse retinas (Fig. 6B and Supplemental Table S7). In photoreceptors, SIPA1 immunostaining signal appeared as doublets resembling the basal body and daughter centriole (Fig. 10C). This SIPA1 signal was positioned adjacent to the ADGRV1 signal. SIPA1 was further confirmed in usherin immunoprecipitates from mouse retinas by immunoblotting (Fig. 10E; two independent experiments with different SIPA1 antibodies). PTPN23 showed a distribution throughout the photoreceptor IS as well as doublets at the IS distal end. Although its signal did not directly overlap with ADGRV1, the two proteins were closely apposed (Fig. 10D). Coimmunoprecipitation could not be performed for PTPN23 due to limitations of the available antibody. However, EIF2B4 did coimmunoprecipitate with ADGRV1 in mouse retinas, as demonstrated by immunoblotting (Fig. 10E, two independent experiments), although the EIF2B4 antibody produced no immunostaining signal in mouse photoreceptors. Taken together, our results support the conclusion that collagen VI, HEXIM1, SIPA1, EIF2B4, and PTPN23 are strong candidate proteins interacting with usherin and ADGRV1 in vivo.

Discussion

In this study, we systematically defined the interaction partners of the USH2 complex using affinity purification-mass spectrometry. We used USH2 antibodies for immunoprecipitations to identify partners interacting with native USH2 proteins. In addition, we used USH2 protein fragments for pulldowns to find domain-specific partners that full-length folded proteins may mask. Unlike previous work, which focused solely on ADGRV1 in immortalized cultured cell lines (39, 40), our study targeted all three USH2 proteins directly in bovine and mouse retinas, enabling the identification of physiologically relevant, particularly extracellular, interaction partners. As expected, we recovered previously reported ADGRV1-associated molecular modules, including the intracellular actin network, CCT complex, and BBSome, validating our approach (39, 40, 68). Beyond these known components, our extracellular ADGRV1 and usherin baits revealed a substantially expanded set of ECM-related interaction partners, including proteins participating in collagen biosynthesis and modification, elastic fiber formation, and ECM proteolysis, none of which had been previously observed or validated. We also identified extracellular usherin self-interaction, direct usherin-ADGRV1 binding, and domain-specific associations. Two of these domain-specific associations are the ADGRV1 EAR/EPTP region with TGFβ signaling and the usherin laminin domains with cell adhesion. Together, these findings support a model in which the USH2 complex bridges ECM constituents with the intracellular actin network, mediates bidirectional mechanical signals, and contributes to ECM remodeling, TGFβ signaling, cell adhesion, and ciliary function in photoreceptors.

In the ECM, pro-TGFβ proteins exist as a latent TGFβ complex in which the propeptide dimer shields the mature TGFβ dimer from activating its receptors. LTBP1 and LTBP4 bind these propeptides at elastic fiber microfibrils and regulate TGFβ release via mechanical tension or proteolytic cleavage (69, 70). Once activated (70), TGFβ signaling modulates ECM turnover by suppressing metalloproteinases and serine proteases or enhancing protease inhibitors (71). EMILIN3, an elastin-microfibril interface protein, binds pro-TGFβ proteins and inhibits TGFβ signaling (45). In this study, we demonstrated that EMILIN3 interacts with ADGRV1 at the periciliary region of mouse and human photoreceptors, where TGFβ cytokines are expressed (71). Consistent with this, our ADGRV1 EAR/EPTP bait captured TGFβ1 propeptide, TGFβ2 propeptide, LTBP1, and LTBP4. Notably, a prior ADGRV1 interactome study reported enrichment of SMAD-related pathways, which are known downstream of TGFβ signaling (40). Therefore, these findings strongly suggest that ADGRV1 contributes to the regulation of TGFβ sequestration and release in the periciliary ECM of photoreceptors.

ADGRV1 and usherin also pulled down collagen proteins, including beaded filament-forming collagen VI (COL6A1 and COL6A2), fibrillar collagen V (COL5A3), and network-forming collagen IV (COL4A4) (72). We showed that COL6A1 colocalizes with ADGRV1 in photoreceptors, but we could not pull down overexpressed COL6A1 with ADGRV1 or usherin baits, suggesting that USH2 proteins interact with native assembled collagen VI rather than its individual α chains, similar to integrins α1β1 and α2β1 (73, 74). Such interactions may also occur through assistance from other proteins, such as integrins and CSPG4 (66). ADGRV1 baits further captured the collagen biosynthesis and modifying enzymes, PLOD1-3 and PCOLCE2. Although PCOLCE2 enhances C-terminal propeptide cleavage during procollagen I extracellular processing (75), PLOD proteins are procollagen lysyl hydroxylases essential for collagen cross-linking, glycosylation, and assembly (76, 77). PLOD3 is a glucosyltransferase as well and is crucial for the assembly of highly glycosylated collagen IV and VI (76). MMP19, an enzyme capable of hydrolyzing collagen IV (78), was pulled down by usherin laminin baits and ADGRV1 antibodies. Furthermore, PLOD1 and PLOD2 are regulated by TGFβ signaling (77), supporting a role of the USH2 complex in TGFβ signaling. Collectively, these findings support the notion that the USH2 complex binds to ECM collagens and participates in their assembly and homeostasis, potentially acting downstream of TGFβ signaling.

Studies in HEK293 cells have reported different downstream G-protein pathways for ADGRV1, including Gαs, Gαi, and Gαq (16, 17, 40). ADGRV1 CTF can bind exogenous Gαi proteins and inhibit adenylate cyclase (16), while a synthetic ADGRV1 fragment binds exogenous Gαs and Gαq and activates calcium-dependent PKA and PKC pathways (17). Recently, ADGRV1a and CTF were shown to activate Gαs and Gαi3, respectively, and a Stachel peptide released from ADGRV1 CTF after autocleavage activated Gαq, raising the possibility of a shift from Gαs-to Gαi-mediated signaling following Gαq activation by Stachel peptide (40). In retinal ADGRV1 immunoprecipitates, we identified endogenous Gαi3 (GNAI3) and Gα11 (GNA11), a Gαq family member (79), indicating that Gαi- and Gαq-mediated pathways predominate in vivo. mTOR-related signaling was also enriched in prey proteins from ADGRV1a, CTF, and ICD baits (40) and in differentially expressed genes in Adgrv1^del7TM^ retinas (42), consistent with our observation that mTORC1 and PI3K/PKB pathways appear enriched among high-confidence ADGRV1 and usherin binding partners. Given the established links between ADGRV1 and several epilepsy forms (80, 81, 82, 83) and the involvement of PI3K/PKB/mTOR signaling in epilepsy (84, 85, 86, 87), we propose that mTOR and PI3K/PKB represent downstream pathways of Gαi- and Gαq-mediated ADGRV1 signaling in the retina.

EPHA3 is a receptor tyrosine kinase that binds ephrin-A ligands and regulates cell detachment and adhesion through bidirectional responses and small GTPases-mediated actin reorganization (88). We identified a direct interaction between EPHA3 and usherin, consistent with the enrichment of ephrin signaling in previously reported ADGRV1 interactomes (40). We also found actin-associated proteins pulled down by ADGRV1 C bait and small GTPase-regulating proteins (ARHGAP21, ARHGAP32, ARHGEF17, and RASGRF2) in ADGRV1 immunoprecipitates, in line with previous reports linking USH2 proteins to the actin cytoskeleton and cell adhesion (39, 40, 89). In photoreceptors, the USH2 complex resides at the periciliary plasma membrane, facing the connecting cilium through the ECM. We observed abnormal membrane fusion at this interface in Whrn mutant photoreceptors (21), suggesting impaired adhesion. Because EPHA3 contains two FN3 domains and a PDZ-binding motif (90), it may associate with usherin and whirlin through these regions and help maintain membrane integrity at the periciliary collar in photoreceptors. Further work is needed to elucidate the exact role of EPHA3-usherin interaction, EPHA3 ligand(s), and their association with actin- and small GTPase-mediated pathways in photoreceptors.

BBS8, TUB, and KIAA1549, three proteins coimmunoprecipitated with USH2 proteins, are all linked to inherited retinal degeneration (62, 63, 64). BBS8 is a core component of the BBSome (91), which mediates membrane protein trafficking at the photoreceptor connecting cilium (92, 93, 94, 95, 96). Consistent with this, our ADGRV1 C bait interactome also contained additional BBSome core components and CCT complex components required for BBSome assembly (97), supporting a recently reported association between ADGRV1, the BBSome, and the CCT complex (68). TUB localizes to the photoreceptor cilium and nuclear euchromatin (64), and Gαq/Gα11 signaling can drive its translocation to the nucleus (98). Loss of TUB causes mislocalization of rhodopsin and cone opsin (99), suggesting a role of TUB in ciliary trafficking. KIAA1549 is present at the photoreceptor connecting cilium (63, 100); however, its function remains unknown. Given all three RP proteins localize to the photoreceptor cilium and ADGRV1 is present at the base of primary cilia and spindle poles in RPE1 cells (40), our findings suggest that a small fraction of ADGRV1, likely below the detection threshold of immunostaining, exists at the basal body and daughter centriole in photoreceptors, where cell adhesion protein SIPA1 (101) and cilium growth protein PTPN23 (102, 103) are localized, making these two proteins strong candidates for USH2 interactions. ADGRV1 ICD has been proposed to undergo γ-secretase cleavage and nuclear translocation (36, 40), which may explain the presence of the transcription repressor HEXIM1 in our ADGRV1 interactome and its weak colocalization with ADGRV1 in photoreceptors. ADGRV1 activation may influence HEXIM1 movement from the USH2 complex to the nucleus, where HEXIM1 binds to the p-TEFb complex for transcription regulation (67). Finally, EIF2B4 is the regulatory δ subunit of the eIF2B complex that activates eIF2 by switching its GDP binding to GTP binding (104). In response to integrated stress, phosphorylated eIF2 interacts with EIF2B4 (105, 106) and inhibits EIF2B4 activity and protein synthesis (107). Therefore, the identification of EIF2B4 suggests that ADGRV1 may regulate protein synthesis under stress.

The affinity purification-mass spectrometry approach used in this study is a powerful tool for identifying interaction partners, but it has inherent limitations and risks of false-positive results. First, copurification may reflect nonspecific binding or indirect associations rather than direct physical interactions. Second, because we used short USH2 protein fragments as baits, our interactomes may underrepresent interactions that depend on intact full-length proteins, native membrane context, or photoreceptor-specific posttranslational modifications; likewise, bait production and some verification steps performed in heterologous cell systems may not fully recapitulate photoreceptor biology. Third, our extracellular affinity purifications were performed once per bait, limiting our ability to assess biological variability. Fourth, the use of whole bovine and mouse retinal lysates sacrifices single-cell resolution, meaning some candidates may originate from nonphotoreceptor retinal cells. In addition, cross-species comparisons may be affected by sequence divergence, potentially missing interaction partners unique to a given species. Finally, because it is impractical to validate every candidate and many proteins lack reliable antibodies, we confirmed only a subset of interactions and photoreceptor localizations using orthogonal experimental approaches.

Despite these limitations, several lines of evidence strengthen confidence in our interactomes, including their consistency with previous reports, the strong agreement between pull-down and coimmunoprecipitation results across multiple baits, antibodies, and species, the use of stringent negative controls, the integration of both qualitative and quantitative analyses, and the validation of several interaction partners. Moving forward, expanded orthogonal validation and functional studies in photoreceptor models and USH2 mutant backgrounds will be essential to confirm direct interactions, define interaction interfaces, and determine the implicated pathways contributing to retinal degeneration. Overall, our findings point to a broader signaling role for the USH2 complex, integrating mechanical cues from the ECM with intracellular pathways governed by Gαi3 and Gα11. The involved downstream pathways likely include ECM remodeling, TGFβ regulation, cell adhesion, and ciliary function, suggesting that multiple mechanistic defects may lead to USH2-associated photoreceptor degeneration. Therefore, our study establishes a molecular framework that will guide future efforts to elucidate disease mechanisms.

Data Availability

The MS interactome data and annotated mass spectra for proteins identified on the basis of one unique peptide have been deposited to the ProteomeXchange Consortium (http://www.proteomexchange.org) via the MASSIVE repository, with the dataset identifier MassIVE MSV000098342.

Supplemental Data

This article contains supplemental data.

Conflict of Interest

The authors declare no competing interests.