Membrane Interactome of a Recombinant Fragment of Human Surfactant Protein D Reveals GRP78 as a Novel Binding Partner in PC3, a Metastatic Prostate Cancer Cell Line

Surfactant protein-D (SP-D), a member of the collectin family has been shown to induce apoptosis in cancer cells. SP-D is composed of an N-terminal collagen-like domain and a calcium-dependent carbohydrate recognition domain (CRD). Recently, we reported that a recombinant fragment of human SP-D (rfhSP-D), composed of homotrimeric CRD region, induced intrinsic apoptotic pathway in prostate cancer cells. Here, we analyzed the membrane interactome of rfhSP-D in an androgen-independent prostate cancer cell line, PC3, by high resolution mass spectrometry and identified 347 proteins. Computational analysis of PPI network of this interactome in the context of prostate cancer metastasis and apoptosis revealed Glucose Regulated Protein of 78 kDa (GRP78) as an important binding partner of rfhSP-D. Docking studies suggested that rfhSP-D (CRD) bound to the substrate-binding domain of glycosylated GRP78. This was further supported by the observations that human recombinant GRP78 interfered with the binding of rfhSP-D to anti-SP-D polyclonal antibodies; GRP78 also significantly inhibited the binding of recombinant full-length human SP-D with a monoclonal antibody specific to the CRD in a dose-dependent manner. We conclude that the interaction with rfhSP-D is likely to interfere with the pro-survival signaling of GRP78.


INTRODUCTION
Surfactant protein D (SP-D) recognizes an array of carbohydrate moieties present on the microbial surfaces (1,2). SP-D is a hydrophilic glycoprotein; its primary structure contains Nterminal region with cysteine residues, a triple-helical collagen domain, an a-helical coiled-coil neck domain, and a C-terminal C-type (calcium-dependent) lectin or carbohydrate recognition domain (CRD). Four SP-D trimeric subunits assemble as a cruciform dodecamer, and their further oligomers appear as fuzzy balls under electron microscope (3)(4)(5). Pathogens bound to SP-D get aggregated and opsonized leading to enhanced phagocytosis and oxidative damage (6). The CRD of SP-D interacts with various selfligands as well as receptors such as CD14, Signal Inhibitory Regulator protein (SIRP)-a, Toll-like receptor (TLR)-2, and TLR-4 to bring about immune regulation (7,8).
An important role of SP-D in allergy was noted when a recombinant fragment of human SP-D (rfhSP-D) composed of neck and CRD region induced apoptosis in activated eosinophils of allergic patients (9); this involved p53 upregulation as demonstrated in rfhSP-D treated AML14.3D10 cells, an eosinophilic leukemic cell line (10). Subsequently, SP-D was shown to bind the Epidermal Growth Factor Receptor (EGFR) on A549 lung cancer cells, inducing cell death (11). Kaur et al. recently reported that rfhSP-D induced apoptosis in pancreatic cancer cell lines via TNF-a/Fas pathway irrespective of the p53 status (12). Further, inhibition of TGF-b expression in the pancreatic cancer cells suppressed their epithelial-tomesenchymal transition (EMT) and ability to invade (13). A significantly reduced expression of SP-D transcripts has been reported in lung, gastric, and breast cancers, whereas ovarian cancer tissues express more SP-D. The presence of SP-D predicted a favorable outcome in lung cancer. Conversely, in gastric, breast, and ovarian cancers, SP-D expression suggested a poor prognosis (14).
Reduced expression of SP-D has been observed in the glandular cells of human prostate cancer tissues. SP-D levels negatively correlated with Gleason score, a scoring system based on the cancer cell types and patterns on the tissue sections as well as tumor volume (15). We recently reported that androgens regulate SP-D expression in the androgen-sensitive LNCaP prostate adenocarcinoma cells (16). Importantly, rfhSP-D specifically induced apoptosis in the tissue explants, primary tumor cells of patients with metastasized prostate cancer, and the prostate cancer cell lines (LNCaP, PC3). The rfhSP-D treatment did not affect the normal prostate epithelial cells (16).
The broad spectrum anti-cancer activity of rfhSP-D is possibly due to the simultaneous targeting of multiple growth factors, kinases, transcription factors, and apoptotic pathways (9)(10)(11)(12)(13)16). Here, we set out to examine the interaction of rfhSP-D with the membrane proteins of the prostate cancer cell line, PC3, using Mass Spectrometry-based LC-MS/MS analysis. Based on the protein−protein interaction (PPI) network analysis of the interactome proteins, Glucose Regulated Protein of 78 kDa (GRP78) was found to be an important interactor and was thus selected for further analysis.
GRP78 is an endoplasmic reticulum (ER) resident chaperone and is also known as HSPA5. GRP78 acts as an ER stress sensor and is upregulated under stress conditions, for maintaining ER homeostasis and cell survival. For cellular homeostasis, GRP78 interacts with three ER stress sensor proteins: protein kinase RNAlike ER kinase (PERK), inositol-requiring kinase 1 (IRE1a), and activating transcription factor 6 (ATF6). However, under ER stress conditions, unfolded proteins titrate away GRP78 to activate the unfolded protein response (UPR) and alleviate cellular stress. Activated UPR decreases protein influx into the ER and induces the synthesis of components involved in protein folding to support cell survival (17). In addition, GRP78 localizes to cell surface (csGRP78) and operates as a receptor, interacting with various ligands to induce pro-proliferative, pro-survival, and proapoptotic signalling (18). GRP78 is significantly upregulated in various cancers due to over stressful microenvironment (17). The upregulated expression of GRP78 in patients with castrationresistant prostate cancer has been associated with resistance to chemotherapy (19,20). Here, using mass spectrometry, proteinprotein interactome analysis and molecular modelling, we report GRP78 as a putative receptor for the CRD region of human SP-D. The interaction between rfhSP-D and GRP78 may interfere with pro-tumorigenic role of GRP78 in prostate cancer and could be a plausible mechanism for rfhSP-D-mediated apoptosis of prostate cancer cells.

Isolation of PC3 Membrane Proteins
Interacting With rfhSP-D for 2 h at 37°C in the CO 2 incubator (PC3 cells were preincubated with rfhSP-D to enable isolation of complexes of rfhSP-D and membrane proteins). Following incubation, the unbound rfhSP-D was removed along with the medium followed by a wash with sterile PBS. Then, PC3 cells were harvested for membrane protein isolation by scrapping in the wash solution supplied with Mem-PER ™ Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific; #89842).
For the co-immunoprecipitation and pull-down experiments, 10 µg of membrane protein extract, along with bound rfhSP-D, was combined with 10 µg of polyclonal antibody against human SP-D (Santacruz) and incubated overnight at 4°C with mixing. First, Pierce Protein A/G Magnetic beads (Thermo Fisher Scientific; #88802) were washed three times with TBST wash buffer. The antigen and antibody mixture was added to 1.5 ml tube containing pre-washed protein A/G magnetic beads (250 µg) and incubated at room temperature for 2 h on a rotary shaker. Beads were then separated on a magnetic stand, and the flow-through was saved for analysis. Beads were washed three times with the wash buffer and a final wash was given with Milli Q water. Beads were then resuspended in the elution buffer (0.1 M glycine, pH 2.0) and incubated at room temperature for 10 min with mixing. Beads were separated magnetically and supernatant containing the target antigen was saved for further analysis. Low pH was neutralized by adding neutralizing buffer (1M Tris-HCl, pH 7.5) to the eluate. PC3-derived membrane protein fraction used for LC-MS/MS analysis was pooled from five independent pull-down experiments with three technical replicates.

Western Blot Analysis of Various Fractions
PC3 cells (5 × 10 6 ) in serum-free RPMI medium were plated in a T75 tissue culture flask and pre-incubated with rfhSP-D (20 mg/ ml) for 2 h, as described above. PC3-membrane as well as cytoplasmic proteins were isolated, as described above, and analyzed by Western blotting. Lysate proteins (30 mg) were subjected to 12% v/v SDS-PAGE and electro-transferred onto PVDF membranes (Pall Corporation, NY, USA). To confirm the presence of rfhSP-D binding proteins in the two protein fractions, one of the blot was blocked (3% skimmed milk powder in TBS, 45 min on a shaker at room temperature), incubated with rfhSP-D in the presence of CaCl 2 or EDTA (2 h at 37°C), and probed with primary rabbit anti-human SP-D polyclonal conjugate (overnight at 4°C) and secondary goat anti-rabbit Horseradish Peroxidase (HRP) conjugate (1 h at room temperature on a shaker). To confirm the purity of the two protein fractions, another blot was probed with primary antibodies against human Prostate-Specific Membrane Antigen (specific to the membrane protein fraction of prostate cells) (rabbit monoclonal PSMA antibody; Cell Signalling Technology, # D7I8E, 1:500) overnight at 4°C, followed by HRP-conjugated secondary goat anti-rabbit IgG antibodies (diluted 1:6000) (1 h at room temperature on a shaker). The blots were developed using the enhanced chemiluminescent (ECL) Kit (Millipore, USA). The blot images were captured using the Syngene (Chem Genius).

Immune-Depletion of Proteins From Membrane Fraction
Membrane proteins in the eluate were concentrated using 3 K filters (Millipore, MA, USA) and then subjected to IgG depletion using Multiple Affinity Removal System Spin Cartridge, HSA/ IgG (Agilent, CA, USA), following the manufacturer's instruction. A buffer exchange followed using Amicon Ultra-0.5 (3K) device in PBS. Total protein was estimated by Bradford assay, followed by SDS-PAGE analysis to assess the profile of cell lysate, and membrane fraction with rfhSP-D treatment after immune-depletion.
To confirm the presence of rfhSP-D in the two protein fractions (20 mg), the supernatant of pull-down fraction (5 mg), flow-through of pull-down fraction (5 mg), eluate of pull-down (5 mg), and flow-through containing IgG after immune-depletion (5 mg) were separated via 12% v/v SDS-PAGE and electrotransferred to PVDF membranes. Then, the membrane was blocked and incubated with the rabbit anti-human SP-D polyclonal antibodies (overnight at 4°C) and secondary goat anti-rabbit HRP conjugate (1 h at room temperature). The blot was developed using the enhanced chemiluminescent (ECL) Kit (Millipore, USA). Image was acquired by Syngene (Chem Genius).

Lys-C/Trypsin Digestion
Eluates of the five individual pull-down experiments were diluted in ammonium bicarbonate buffer (TEABC; Sigma) to a final volume of 100 ml, followed by reduction using 10 mM dithiothreitol (DTT) for 20 min at 60°C. Once the proteins came to room temperature, they were alkylated (20 mM Iodoacetamide; Merck) for 10-15 min at room temperature in the dark. The proteins were digested with 1 mg of Lys-C for 3-4 h at room temperature, followed by digestion with 2 mg of trypsin overnight at 37°C. Enzyme activity was terminated with 1% v/v formic acid. Peptides were fractionated via C-18 columns and eluted with a buffer comprising acetonitrile (40%) and formic acid (0.1%). Fractionated peptides were lyophilized on a SpeedVac (30 min at 45°C) and then solubilized in 0.1% formic acid prior to LC-MS/MS analysis.

Mass Spectrometry
Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific), connected with an Easy-nLC II nanoflow liquid chromatography system (Thermo Fisher Scientific), was used for peptide analysis. A trap column (75 µm × 2 cm, Magic-C18-AQ material 5 µm, 100 Å) was used to enrich peptides. The peptides were separated at a flow rate of 30 ml/min on a 20 cm long column of 5 µm Magic-C18-AQ (Michrom Bioresources, Inc., Auburn, CA, USA) using a gradient of 8-30% solvent B (90% acetonitrile in 0.1% formic acid) over 103 min for a 120 min run. Mass spectrometry data were collected at a resolution of 120,000 in a range of 350-1,600 m/z. The highly intense ions with charge state >2 were isolated in 3 s cycle and subjected to HCD fragmentation with 32% normalized collision energy. These fragmented ions were sensed at a resolution of 15,000 at 200 m/z. The limit of dynamic exclusion was fixed at 40 s with a 10-ppm mass window. The maximum ion injection times were 50 ms for MS and 75 ms for MS/MS. The automatic gain control targets were 4 × 10 5 for MS and 1 × 10 5 for MS/MS.

Identification of Peptides and Proteins From Database
The uninterpreted MS/MS data from the complete LC-MS/MS run was subjected to a search in the database (Human RefSeq protein database). The search algorithm used was SEQUEST, and the platform used was Proteome Discoverer (version 2.1, Thermo Scientific). A maximum of two missed cleavages, carbamidomethylation at cysteine as fixed and oxidation of methionine as variable modifications, were included as the search parameters. Mass tolerance of monoisotopic peptide was set at 10 ppm and the MS/ MS tolerance was limited to 0.02 Da. At the PSM as well as the protein level, the false discovery rate of 1% was set. MS data has been submitted to the ProteomeXchange Consortium (http:// www.proteomexchange.org) via the PRIDE partner repository (dataset identifier PXD008098).

Ingenuity Pathway Analysis
Proteins identified by LC-MS/MS analysis were further analyzed for their interactions and molecular pathways using Ingenuity Pathways Analysis (IPA) software (http://www.ingenuity. com) (21).

Network Analysis
Protein-protein interactions (PPIs) of the rfhSP-D interactome (347 proteins) were downloaded from STRING v11 (22). Proteins in the network were scored based on network topological properties, such as degree of connectivity, clustering coefficient, betweenness centrality, closeness centrality, and shortest path using R package igraph (23). The importance of each protein in the network was assessed based on perturbation and disruption scores. Perturbation score for each protein reflected the difference in centrality of network, calculated based on average of clustering coefficient, betweenness centrality, and closeness centrality after deletion of a protein (24). Disruption score of each protein was calculated based on average increase in shortest path length of protein pairs caused by deletion of the protein (25). Proteins were ranked individually based on hub (degree of connectivity), perturbation and disruption analysis; the average rank obtained from the three methods was used for creating the final ranked list. The top 5% proteins of the rank list were shortlisted and screened for their expression in prostate glandular cells as per protein atlas data (26); prostate cancer (C0376358) and metastasis of prostate cancer (C1282496) using data available in DisGeNET (27); and role in cancer cell apoptosis and survival as per information in ApocanD database (28) and IPA.

Protein Docking
The crystal structure of human GRP78 complexed with ADP (produced in Escherichia coli BL21) (PDB ID: 5E84 Chain A) was downloaded from PDB (29). O-glycosylation of GRP78 is critical for its stability, subcellular localization, and anti-apoptotic function (30). Putative O-glycosylation sites have been identified in GRP78 (Thr85, Thr151, Thr166, Thr184, and Thr203) based on in vitro immunoprecipitation assays (30). The crystal structure of GRP78 was glycosylated by addition of five O-linked N-acetylgalactosamine (GalNAc) at the putative sites. The glycosylated structure was generated using CHARMM-GUI web server and optimized using CHARMm force field (31). Monomeric form of GRP78 is present in stress induced cells (32); hence, the monomeric glycosylated structure of GRP78 was blind docked with crystal structure of CRD domain of active recombinant fragment of human lung surfactant protein SP-D (PDB ID:1PW9) using ZDOCK (Biovia, Discovery studio Version 2017) (33,34). The input parameters for docking were set as default options. The top 2,000 docked poses generated by ZDock were re-ranked utilizing detailed electrostatic, van der Waals and solvation forces by ZRank (35,36). The top ranked pose based on ZRank was further analyzed for intermolecular interactions.

Interaction Studies between rfhSP-D/ rFLhSP-D and GRP78 by ELISA
To analyze the rfhSP-D/rFLhSP-D (recombinant full-length human SP-D; R & D) interaction with GRP78, the study utilized non-glycosylated version of GRP78 (purified from E. coli; Cayman, catalog no. 22730) since the O-linked glycosylated GRP78 specific to cancer cells is a minor fraction of the total GRP78 (30). Using a direct ELISA, GRP78 (1 µg/ml in PBS) was coated on microtitre wells overnight at room temperature. Postwashing (three times with PBST, PBS + 0.05% Tween 20), the wells were incubated with 1% w/v BSA in PBS for 1 h at room temperature to block the additional sites. After extensive washing, rfhSP-D (1 µg/ml in PBS buffer) was added to the wells with and without 5 mM CaCl 2 , or EDTA (10 mM). Bound rfhSP-D was probed with primary rabbit anti-human SP-D polyclonal antibodies (1 h at 37°C) and secondary goat antirabbit HRP conjugate (1 h at 37°C), or with biotinylated Human SP-D detection antibody (500 ng/ml; Duoset SP-D ELISA kit, R & D Systems, USA) (1 h at room temperature with shaking) and Streptavidin-HRP conjugate (1:200; 45 min at room temperature with shaking).
The ability of GRP78 to bind rfhSP-D and block its binding to rabbit anti-SP-D polyclonal antibodies was analyzed by coating rfhSP-D (1 µg/ml in PBS) in duplicates, followed by blocking and incubation with GRP78 (1 µg/ml in PBS). The wells were then probed with primary rabbit anti-human SP-D polyclonal antibodies and secondary goat anti-rabbit HRP conjugate.
The ability of GRP78 to bind rFLhSP-D and block its binding to anti-SP-D monoclonal antibody was analyzed by a sandwich ELISA (Human SP-D Duoset ELISA kit). Human SP-D capture antibody was coated (2 µg/ml) overnight at room temperature. After washing, the wells were incubated with 1% w/v BSA in PBS for 1 h at room temperature to block the unoccupied sites. The wells were then incubated with 5 ng/ml of rFLhSP-D (116 × 10 −3 nM) and increasing concentrations of human recombinant GRP78 from 5 ng/ml (70 × 10 −3 nM) up to 20 ng/ml (280 × 10 −3 nM) (1 h at 37°C), and probed with Human SP-D detection antibody (500 ng/ml) (

Statistical Analysis
The data obtained for various experiments were analyzed for statistical significance and graphical representation using GraphPad PRISM ver6.0 (GraphPad Software Inc., San Diego, CA). To compare the untreated and treated groups, the unpaired t-test was used. Data have been represented as mean ± SD. The p values less than 0.05 showed that the findings were statistically significant.

Isolation of Membrane Proteins From rfhSP-D Pre-Treated PC3 Cells
To investigate the rfhSP-D interactome in PC3 cells, we performed co-immunoprecipitation and pull-down assays using polyclonal antibodies to rFLhSP-D, followed by immunedepletion of abundant proteins (HSA/IgG) and LC-MS/MS analysis (workflow scheme shown in Figure 1).
The protein profile of the membrane as well as cytosolic fractions isolated after rfhSP-D pull-down assay using rfhSP-Dtreated PC3 cells was distinct, as revealed by SDS-PAGE ( Figure  2A). Figures 2B(i) and C(i) represent protein bands electrotransferred to PVDF membrane, stained with Ponceau dye and used as loading control. Ligand blotting of the membrane and cytosolic protein fractions with rfhSP-D showed presence of rfhSP-D binding proteins in both the fractions [ Figure 2B(ii)]. Western blot analysis showed significant enrichment of Prostatespecific Membrane Antigen (PSMA) in the membrane fraction in comparison to cytosolic fraction [ Figure 2B(iii)]. Figure 2C (ii) shows the Western blot of isolated PC3 cell membrane proteins after 2 h incubation with rfhSP-D and subsequent fractions, using anti-human SP-D polyclonal antibody. Following 2 h incubation, the unbound rfhSP-D was removed along with the medium, followed by a wash with sterile PBS. Hence, only the ligand-bound rfhSP-D is expected to be present in the protein fractions analyzed. Native SP-D protein levels in the cell culture supernatants of PC3 cells (1 × 10 6 /ml-3.2 µg of total protein/µl), as evaluated by ELISA, were found to be 220 ± 21 pg/ml, indicating approximately 68.75 pg of SP-D/mg of protein (16). Expected molecular weight of rfhSP-D is approximately 20 kDa and that of full-length SP-D is~43 kDa , and A2ML1 (a2macroglobulin-like protein 1), were also present in the identified interactome (Supplementary File S1) (2, 6). Identification of these putative SP-D receptors validated the interactome analysis. As discussed above, the proteins identified from the LC-MS/MS analysis did not include SP-D. Therefore, SP-D was added to the list of 347 proteins for subsequent bioinformatics analysis in order to decipher the networks and protein-protein interactions.

Pathway Analysis of the rfhSP-D Interactome
The rfhSP-D interactome [348 proteins, 347 identified proteins + SP-D (P35247)] was analyzed by ingenuity pathway analysis (IPA) for associated biological functions, canonical pathways, and networks. Our network analysis revealed novel signaling proteins that are likely to interact (directly/indirectly) with rfhSP-D and/or regulate its associated networks ( Figure 4A).

Network Analysis of rfhSP-D Interactome
The PPI network of 347 proteins was analyzed for network parameters (Materials and Methods) to shortlist critical proteins of the rfhSP-D interactome. 20 proteins (GAPDH, HSPA8, HSP90AA1, HSPA4, TPI1, ENO1, VCP, CCT2, HSPA5, LCN2, EEF2, ACTG1, HSPD1, PDIA6, CCT8, S100A7, HSPA9, LCN1, HSP90AB1, CCT5) were selected and ranked based on connectivity, perturbation, and disruption scores ( Table 1; Supplementary File S3). 15 of these 20 proteins are known to  be associated with prostate cancer, 13 with prostate cancer as well as apoptosis, and only two (HSP90AA1, HSPA5) are known to be associated with metastasis of prostate cancer ( Figure 5; Table  2; Supplementary File S2). Of the two, HSPA5 is known to be expressed in prostate glandular cells as per the protein atlas; therefore, it was selected for further in silico and in vitro interaction studies. A summary of the network analysis of rfhSP-D interactome is represented as a Venn diagram ( Figure  5). HSPA5 or GRP78, a chaperone expressed on cell surface of cancer cells only, is known to be associated with malignancy, development of castration-resistant prostate cancer, and resistance to chemotherapy. GRP78 seems to act upstream of PI3K/Akt; monoclonal antibody against GRP78 suppresses pAkt expression, suggesting promotion of apoptosis (44).

In Silico Validation of the Interaction of SP-D With GRP78
From the proteomic analysis, GRP78 was the top-ranked protein of the rfhSP-D-PC3 membrane interactome and hypothesized to be involved in the apoptosis signaling by binding to rfhSP-D molecule.
In silico molecular docking was performed to validate this hypothesis. The top ranked pose (Zrank = −131.3) from blind docking of rfhSP-D and GRP78 revealed that CRD of SP-D can bind with GRP78 via the substrate binding domain of GRP78 ( Figure 6; Table 3).

In Vitro Verification of Interaction Between rfhSP-D/rFLhSP-D and GRP78
To examine GRP78 binding to rfhSP-D, we carried out a direct ELISA with recombinant GRP78 and rfhSP-D that was probed using the polyclonal and monoclonal antibodies to SP-D ( Figure  7A). Polyclonal antibodies could detect rfhSP-D (1 µg/ml, 0.232 nM) bound to coated GRP78 (1 µg/ml, 0.280 nM) in the presence of Ca 2+ . However, EDTA did not significantly inhibit the binding, indicating a protein-protein interaction. Monoclonal anti-human SP-D antibodies recognizing peptides NEAAFLSMTDSK (positions 308-319) and SAAENAALQQLVVAK (positions 293-307) that are located in the CRD region of SP-D (45), could not detect the rfhSP-D bound to coated GRP78, suggesting involvement of the CRD region of rfhSP-D in binding to GRP78.
To evaluate if GRP78 binding can interfere with the interaction of rfhSP-D and polyclonal antibodies, a direct ELISA was carried out using rfhSP-D coated on the microtitre wells, followed by GRP78 as the ligand and probed with polyclonal antibodies. Minimal binding of polyclonal anti-  human SP-D antibody to rfhSP-D (1 µg/ml i.e. 0.232 nM) was observed due to the blocking/masking of antibody-recognized domains by GRP78 (1 µg/ml i.e. 0.280 nM).
To examine if GRP78 binding to rFLhSP-D (recombinant full-length human SP-D) is via CRD domain, we carried out a competitive sandwich ELISA (R&D ELISA kit) using the monoclonal anti-human SP-D antibody. The assay detected rFLhSP-D from 0.115 to 10 ng/ml in a linear standard curve ( Figure 7B). rFLhSP-D bound GRP78 in a dose-dependent manner leading to dose dependent inhibition of binding of monoclonal anti-SP-D detection antibody ( Figure 7C). GRP78 (20 ng/ml i.e. 281 × 10 −3 nM) on binding to rfhSP-D (5 ng/ml i.e. 116 × 10 −3 nM) showed approximately 57.15% inhibition of binding to monoclonal anti-human SP-D antibodies. Thus, GRP78 interfered with the interaction between rFLhSP-D and monoclonal anti-human SP-D antibody, suggesting that CRD of rFLhSP-D is likely to be involved in its binding to GRP78.

DISCUSSION
To delineate the underlying mechanisms of the anti-prostate cancer activity of rfhSP-D, we used a pull-down strategy to unravel the rfhSP-D interactome from the membrane fraction of the metastatic prostate cancer cell line, PC3. A high-throughput proteomic work flow led to the identification of 347 membrane proteins with ≥2 unique peptides, each with ≥2 PSMs. Based on the PPi network properties such as connectivity, perturbation, and disruption scores, 20 proteins were selected, including GRP78 that probably interacted directly or indirectly with rfhSP-D. The interaction between rfhSP-D and GRP78 was validated by in silico and in vitro experimental analysis.
We recently reported a differentially increased binding of rfhSP-D to the metastatic prostate cancer cells and induction of the apoptosis (16). The membrane interactome of PC3 cells, treated with rfhSP-D, showed involvement of several heat-shock protein family members such as GRP78 (HSPA5), HSP90AB1, HSPA8, HSP90AA1, HSPA1B, HSP90B1, and HSPD1, as previously reported (46). Elevated levels of HSPs have been FIGURE 5 | Network analysis of rfhSP-D binding proteins of PC3 cells. Network representation of 20 shortlisted proteins from network analysis; pie of node represents the analysis (hub, perturbation, disruption). Overlaying Venn represents proteins known to be associated with prostate cancer, prostate cancer metastasis, apoptosis, and expressed in prostate cells.
a As per data available in protein atlas; b as per data available in DisGeNET; C as per data available in ApocanD and IPA. *Association with apoptosis pathway.
reported in many cancers, including prostate cancer. Both GRP78 and HSP90AA1 were present in the databases of prostate cancer (C0376358), metastasis of prostate cancer (C1282496) using data available in DisGeNET, and have a significant role in cancer cell apoptosis and survival as per information in ApocanD database and IPA. However, HSP90AA1 is not expressed in the prostate glandular cells as per protein atlas data (26). In addition, it does not relocate to the cell membrane like GRP78 in prostate cancer and is rather secreted extracellularly (47). Interestingly, the 'Cellular and Molecular Pathway' analysis of rfhSP-D interactome listed GRP78 and SP-D together in the top ranked 'Cell Death and Survival' category. Hsp90AB1 induces angiogenesis in the hepatocellular carcinomas, promotes EMT in gastric cancer, and its upregulated expression is implicated in metastasis and differentiation of lung cancer (48)(49)(50). HSPA1B and HSPA8 have shown involvement in the apoptotic signaling. Proteomics analysis of endometrial carcinoma tissue identified HSPA8 as the most upregulated candidate, and siRNA-mediated inhibition of HSPA8 significantly downregulated cell proliferation and promoted cell apoptosis in RL-95-2 and HEC-1B, two endometrial cancer cell lines (51). siRNA targeting HSPA1B gene inhibited proliferation of HeLa, MCF-7, PC-3, HuH-7, and gastric cancer SGC-7901 cells, while the non-tumorigenic HBL-100 mammary cells were not affected adversely (52). HSP90B1, a stress-inducible chaperone protein, significantly reduced the cell proliferation and survival of the malignant cells (53). In comparison with the benign prostate samples, HSPD1/E1 Complex is over-expressed in prostate cancer lesions and in carcinomas (54).
GRP78, also called heat-shock protein 5 (HSPA5), is a member of the HSP70 superfamily, and has a critical role in the regulation of the unfolded protein response (UPR) via appropriate protein folding, inhibiting aggregation of newly-made proteins, and regulating the stimulation of transmembrane sensors in the endoplasmic reticulum (ER) (55). Expression of GRP78 is significantly enhanced in various cancers, and is linked with the prostate cancer malignancy, metastasis and acquisition of resistance to chemotherapy (56). Prostate tumorigenesis was potently arrested in bi-allelic conditional knockout mice for both GRP78 and PTEN (57). Environment of the fast-growing solid tumors is marked with increased hypoxia, and reduced nutrients and acidosis, leading to UPR and increased GRP78 expression. GRP78 directly interacted with apoptotic pathway intermediates to block caspase activation, and eventually led to increased cell survival (58)(59)(60).
GRP78 has been primarily localized to the ER owing to the presence of the KDEL-retention motif, although a minor proportion of GRP78 evades the ER-retention mechanism and reaches the cell surface to promote cell survival (61). Proteomics of the membranes of various tumor cells unravelled a number of heat-shock chaperones and glucose-regulated proteins, including GRP78 (62). These reports are in coherence with our observation of GRP78 to be in the membrane proteome of PC3 cells. Additionally, Thapsigargin, a mediator of ER stress, is involved in surface localization of GRP78 in 293T, HeLa, and MCF-7 cell lines (63). Furthermore, an increased GRP78 expression can lead to its translocation to the membrane even when there is no ER stress (64). Membrane GRP78 may bind to a2-macroglobulin (a2-M), tumor differentiation factor, and vaspin via its substrate binding domain and induce AKT/PI3K pro-survival pathway (65)(66)(67)(68)(69)(70)(71). Peptides targeting membrane GRP78 induce selective tumor cell death. Antibody ligation to cell-surface GRP78 slowed growth rate in prostate cancer cells and blocked PI3K/AKT signaling (72). The predicted involvement of the substrate binding domain of GRP78 in the interaction with rfhSP-D suggested that SP-D may interfere with these pro-survival mechanisms. Thus, the rfhSP-D induced inhibition of PI3K/Akt pathway leading to apoptosis of prostate cancer cells could be plausibly mediated by GRP78 (16).
Case-control studies of lung cancer patients have revealed that the circulating SP-D levels may predict susceptibility to lung cancer (73,74). The collagen and CRD regions of SP-D play different but vital roles in the immune surveillance. The CRD region mediates the pattern recognition function, while the collagen region remains important for the signaling interaction via Calreticulin-CD91 complex. However, studies using rfhSP-D have revealed that the homotrimeric neck and CRD region are endowed with virtual selfsufficiency in many aspects (6). Several lines of evidence suggest the importance of CRD region of SP-D in its biological activity, e.g. CRD region interfered with the EGF and EGF receptor (EGFR) interaction, causing downregulation of the EGF induced signaling in A549 cell line (11). EGF-EGFR interaction results in increased epithelial tumor cell proliferation, angiogenic differentiation, and invasive capability, leading to increased probability of metastasis. Recently, it was demonstrated that SP-D interacted with the EGFR mutant and interfered with its dimerization that was independent of  its interaction with EGF (75). The interaction of SP-D and rfhSP-D with eosinophilic leukemic AML cells was also mediated by CRD (9,10). Kaur et al. proposed that the CRD region of SP-D is involved in binding to target ligand on the pancreatic cancer cell surface via protein-protein interaction (13). Our results demonstrate that the interaction of rfhSP-D with GRP78 also involved the CRD domain, which could be relevant in the induction of apoptosis in cancer cells. O-glycosylation of GRP78 is critical for its anti-apoptotic function (76). Blind docking of the crystal structures of active rfhSP-D and monomeric glycosylated structure of GRP78 revealed that the CRD of rfhSP-D can bind GRP78 via the substrate binding domain of GRP78. Thus, our in silico and in vitro analyses validate the interaction between rFLhSP-D/rfhSP-D and GRP78 via CRD region, substantiating GRP78 as binding partner of SP-D and may potentially be a novel intermediary in SP-D mediated innate immune surveillance against prostate cancer. The therapeutic relevance of SP-D-GRP78 interaction may be further explored using in vivo studies with knockout mice models bearing prostate cancer.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repositories and accession number(s) can be found in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
Conception, co-ordination, and design of the study were accomplished by GT and TM. GT performed and analyzed the experiments and prepared the first draft of the manuscript. SA and UK cloned, recombinantly expressed, and characterized the purified rfhSP-D used in the study. RS, PG, and GS designed, performed, and analyzed the LC-MS/MS data. SI-T and IK designed, executed, and analyzed network and in silico molecular docking studies. PG and BB performed and analyzed LC-MS/MS data by IPA software. SI-T, BB, PG, RS, and UK reviewed the first draft and provided valuable suggestions. TM defended the proposal to acquire the Institutional grant and IEC approval, analyzed the data, and edited the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
Intra-mural grant allocated to TM by the Institute ICMR-NIRRH (Accession no. 921) was used for all the consumables used for experiments. ICMR-NIRRH-JRF and ICMR-SRF supported GT's salary.