Effect of TNF-α on human ARPE-19-secreted proteins

SILAC and TNF-α treatment

ARPE-19 was purchased from the American Type Culture Collection (ATCC; Manassas, VA). The cells were routinely maintained in Dulbecco’s modified Eagle’s medium and Ham’s F12 medium (DMEM/F12) containing 10% FBS (ATCC), 100 U/ml of penicillin, 100 µg/ml of streptomycin (Invitrogen, Gibco, Carlsbad, CA), and 1 mM of sodium pyruvate (Sigma Aldrich, St. Louis, MO). To study the effect of TNF-α on ARPE-19 secreted proteins, we performed differential stable isotope labeling of amino acids in cell culture and used mass spectroscopy to measure ratios of individual secreted proteins in TNF-α treated ARPE-19 and untreated ARPE-19 cells (Figure 1). One set of ARPE-19 cells were subcultured in Dulbecco’s modified Eagle’s medium and Ham’s F12 medium (DMEM/F12; Atlanta Biologicals, Lawrenceville, GA) custom-made medium where Arg and Lys were replaced by ¹³C₆-Arg and ¹⁵N_2,¹³C₆-Lys. After at least 12 cell doublings (roughly four passages), cellular proteins were uniformly labeled with these stable isotope-containing amino acids. In parallel, the same numbers of cells were grown in DMEM/F12 unlabeled medium. Ten days after confluence, both labeled and unlabeled cells were washed 6 times with sterile PBS (Invitrogen, Carlsbad, CA) before incubating them in serum-free medium containing stable isotope-labeled amino acids for labeled cells and regular amino acids for unlabeled cells. The metabolically labeled cells were treated with 10 ng/ml TNF-α (a nonlethal dose), while unlabeled cells were kept untreated. After 24 h incubation, spent media of both treated and untreated cells were collected, mixed at 1:1 ratio and analyzed as described below. To validate the data, we performed a reverse experiment in which unlabeled cells were treated with TNF-α while labeled cells were kept untreated.

Analysis of ARPE-19-secreted proteins

Spent media from TNF-α-treated and untreated ARPE-19 cells were collected and filtered through 0.2 µm polyethersulfone syringe filter (Corning Inc., Corning, NY) to remove floating cells and debris. Filtered media from treated and untreated cells were then mixed at 1:1 ratio (v/v) and concentrated to dryness using centrifugal filtration (Millipore Corporation, Billerica, MA) followed by vacuum centrifugation. The resultant residue was then dissolved in 6 µl of Laemmli buffer containing 50 mM DTT and further fractionated by SDS–PAGE using 10% Tris HCl pre-cast criterion gels (Bio-Rad, Hercules, CA). The gel was then stained with Bio-Safe Coommassie (Bio-Rad) and each lane was excised into 36 to 43 bands. Each band was then in-gel digested using trypsin (Promega, Madison, WI), and the resultant peptides were extracted and analyzed by nano-LC-MS/MS as described in the next section.

Preparation of microsomal fraction

After collection of spent media, treated and untreated cells were washed with PBS, dissociated in 0.05% trypsin-EDTA solution (Invitrogen) and transferred into previously weighed Eppendorf tube. Cells were pelleted by centrifugation at 300x g then washed with 2 volumes of PBS solution and 100 mM NaCl. Treated and untreated cells were then mixed at 1:1 ratio (w/w) and resuspended in 1.5 ml of Tris-EDTA mix buffer that contained 10 mM Tris HCl pH 7.4, 1 mM EDTA, 2.5 M sucrose (Sigma Aldrich), and 1 protease inhibitor tablet (Pierce, Rockford, IL) per 10 ml. The cells were then mechanically lysed by passing them 15 times through a 1 ml syringe with a 23 gauge needle. The suspension was centrifuged for 10 min at 4000x g at 4 °C, and the supernatant containing the microsomal fraction was transferred to a clean 1.5 ml Eppendorf tube and further centrifuged at 13,000x g for 20 min to obtain the microsomal pellet. The microsomal pellet was then resuspended in protein extraction buffer containing 7 M urea,2 M thiourea, 2% CHAPS (w/v), and 50 mM DTT. Protein concentration of each fraction was determined by Bio-Rad protein assay (Bio-Rad). Aliquots of 50 µg were separated on SDS–PAGE. Gel bands were excised, digested with trypsin, and the resultant peptides analyzed by liquid chromatography mass spectrometry as described in the next section.

Nanoflow LC-MS/MS

Nano-LC tandem mass spectrometry was performed on an LC-packing system (Dionex Ultimate Capillary/Nano LC system, Dionex Corp., Sunnyvale, CA) connected to an a linear trap quadrupole (LTQ) instrument (Thermo Fisher Scientific, San Jose, CA). Each sample was injected via an auto-sampler and loaded onto a C18 trap column (5 μm, 300 μm i.d. x 5 mm; Dionex LC-Packings, Sunnyvale, CA) and washed with water containing 0.1% trifluoroacetic acid for 6 min at a flow rate of 10 μl/min. The sample was subsequently separated by a C18 reverse-phase column (3.5 μm, 100 μm x 15 cm, Agilent, Santa Clara, CA) at a flow rate of 200 nl/min. The mobile phases consisted of water with 0.1% formic acid (A) and 90% acetonitrile with 0.1% formic acid (B). A linear gradient from 5 to 65% of solvent B was employed over a period of 100 min to separate peptide mixture. Eluted peptides were introduced into the mass spectrometer via a 10 μm silica tip (New Objective Inc., Ringoes, NJ) adapted to a nano-electrospray source (Thermo Fisher Scientific Inc., Waltham, MA). The spray voltage was set at 1.7 kV and the heated capillary at 180 °C. LTQ was operated in the data-dependent mode in which each cycle consisted of one full-MS survey and subsequently three sequential pairs of intercalated zoom scans and MS/MS events. The targeted ions count in the mass spectrometer trap was 30,000 for full-MS scans, 3,000 for zoom scans, and 10,000 for MS/MS scans. Peptides were fragmented in the linear ion trap using collision-induced dissociation with the collision gas (helium) pressure set at 1.3 ml and the normalized collision energy value set at 35%. The zoom scan events, of higher resolution, were used to determine the charge state of each ion as well as the ratio of labeled to unlabeled peptide pairs using the ZoomQuant software developed by Halligan et al. [30].

Database search

Protein identification was performed by searching the 17,806 human protein sequences in the SwissProt database (UniProtKB/Swiss-Prot release 54.6 of December 2007) using Bioworks 3.3 (Thermo Fisher Scientific, San Jose, CA) and indexed with assumptions for enzymatic tryptic hydrolysis with the possibility of two missed cleavages and the following modifications: 16 Da shift for oxidized Met, 6 and 8 Da shifts for incorporation of stable isotope labeled Arg and Lys, respectively. Stringent filtering criteria were used for database search and peptide identifications: DeltaCn (ΔCn) > 0.1, a variable threshold of Xcorr versus charge state (Xcorr=1.9 for z=1, Xcorr=2.5 for z=2, and Xcorr=3.5 for z=3), peptide probability-based score with a p value <0.001, and more than 2 different peptides per protein. A reverse database search using the same criteria gave a false positive rate of 0%, thus increasing confidence in protein identification:

ZoomQuant analysis

For quantitative analysis we used ZoomQuant Software [30]. In this case proteins were identified using BioWorks 3.1 software that is compatible with ZoomQuant software to generate Sequest output files and extract zoom scans from the mass spectrometry raw data. Each peptide was manually validated by examining zoom scans to verify the labeled and unlabeled pairs. Identified peptides were required to have a matched peptide partner 6 Da greater for Arg terminating peptides, 8 Da greater for Lys containing peptides, and other proper combinations for missed cleavage peptides. Because manually checking zoom scan with pairs of labeled and unlabeled peaks is itself stringent, filtration based on the X corr was lowered in this case to increase proteome coverage (e.g., X corr ≥ 1.8 for doubly charged ions and X corr ≥ 2.5 for triply charged ions).

Statistical analysis

To check the significance of differentially expressed proteins in TNF-α treated cells versus non-treated cells a t-tests for equal variances was performed using the Sidak method [31]. Basically an average ratio was calculated from the total scans of labeled and unlabeled peptide pairs detected for a given protein. These averaged ratios were then compared in the control pairs versus TNF-α treated cells and untreated cells in both the forward and reverse SILAC experiment. The obtained p values were then adjusted for the three way multiple comparisons using the Sidak method.

Western blot analysis

To validate proteome profiling data, we cultured ARPE-19 cells in 10 cm dishes (Corning Inc. Life Sciences, Lowell, MA) to 100% confluence. Cells were washed 6 times with PBS to remove any residual FBS proteins, and treated with various TNF-α concentrations (0.25 to 20 ng/ml) in serum free media for 24 h. Spent media were collected and filtered though 0.2 µm Polyethersulfone syringe filter (Corning Inc.). To check NF-κB activation, we collected cell pellets, washed them with PBS, and lysed with 100 µl of RIPA (Sigma Aldrich, St. Louis, MO) buffer. Protein extracts were separated by SDS–PAGE on a 4 to 15% Tris HCl gradient gel (BioRad) under reducing conditions then transferred to a nitrocellulose membrane (Amersham Biosciences, Piscataway, NJ) for western blot analysis. Nonspecific binding was blocked with 5% milk (BioRad Laboratories Inc., Hercules, CA), and the membrane was incubated overnight with goat polyclonal antibody against human C3 (Complement Technology, Tyler, TX), rabbit polyclonal antibody against human IκB-α (Cell Signaling, Danvers, MA) or mouse monoclonal antibody against human vinculin (Sigma) in 2.5% milk. The next day, the membrane was rinsed with Tris buffer containing salt and Tween (TBST) and incubated with antigoat antibody (Jackson Lab, West Grove, PA) for C3, and antirabbit IgG (BioRad) for IκB-α or antimouse antibody (GE Healthcare, Buckinghamshire, England) for vinculin. All secondary antibodies were conjugated with horseradish peroxidase, and immunoblots were visualized by chemiluminescence (Amersham Biosciences Inc., Piscataway, NJ)

Effect of TNF-α on the viability of ARPE-19 cells

To ensure minimum cell death and minimum release of intracellular proteins into the condition medium of TNF-α-treated ARPE-19 cells, we monitored cell viability after incubating them for 24 h with different concentrations of TNF-α using trypan blue exclusion. Figure 2 shows the percent of viable cells after 24 h incubation with different TNF-α concentrations. TNF-α concentration up to 30 ng/ml produced only a small percent of cell death (e.g., <10% compared to untreated cells). This data are in agreement with a previous study where it has been shown that RPE cells are more resistant to TNF-α-induced apoptosis than other cell type [32]. We have thus chosen 10 ng/ml as the optimal concentration to study the effect of TNF-α on ARPE-19-secreted proteins. This concentration resulted in minimum cell death (<5%) and detectable cell response.

Effect of TNF-α on ARPE-19-secreted proteins

One of the advantages of using the SILAC strategy is that all cellular proteins, including secreted proteins, are uniformly labeled with the stable isotope-labeled amino acids (e.g., ¹³C₆-Arg, ¹³C₆, ¹⁵N₂-Lys). Thus, in a 1:1 mixture of spent media collected from labeled and unlabeled cells, most of the peptides generated from cellular proteins would be detected as pairs of labeled and unlabeled peaks, while peptides from residual bovine serum proteins and contaminant proteins, such as keratin, would be detected as single peaks. In this study, a total of 221 proteins were identified in the spent medium of ARPE-19 cells, but only 146 proteins were of RPE cell origin; these are listed in Appendix 1 with their accession number, peptide count, subcellular localization, and function. The cellular localization of these proteins as well as their function was deduced from their annotations in the UniProt database using the Protein Information Resources (PIR) tool. As shown in Figure 3A, 76 (52%) of these 146 proteins were found to be putatively secreted, 6 (4%) were of microsomal origin, 52 (36%) were of cytosolic origin, and the remaining 12 (8%) were unassigned. While the number of identified cytosolic proteins was high, their actual relative quantities estimated from spectral count were low compared to the secreted proteins (Figure 3B). All proteins in Appendix 1 were identified with at least two different peptides and validated by detection of labeled and unlabeled peptide pairs. Additional proteins that were identified by a single peptide pair in the spent medium of ARPE-19 cells are reported in Appendix 2. Even though each of these proteins was represented by a single peptide sequence, their identity was still validated by detection and sequencing of both labeled and unlabeled peptide pair. Detection of only one representative peptide for each of these proteins suggests that they might be present in very low abundance in ARPE-19 spent medium.

Zoom scan spectra of labeled and unlabeled peptide pairs were used to determine secreted protein ratios in TNF-α-treated cells versus untreated cells. Figure 4 shows example spectra of labeled and unlabeled peptide pairs of complement C3 and syndecan-4 whose secretion was significantly altered by TNF-α treatment. This differential secretion was clearly maintained in both forward and reverse SILAC experiments thus validating the data. Among the 146 proteins reported in Appendix 1, only 6 were found to be altered in their concentrations after TNF-α treatment (Figure 5). Trans-golgi network protein-2 (TGON2) was the only protein whose secretion was decreased (by a factor of 2±0.01) in TNF-α treated cells versus untreated cells while the other 5 proteins, complement C3, fibronectin, sulfhydryl oxidase 1, plasminogen activator inhibitor 1 and syndercan-4 were increased by a factor of 2.03±0.42, 1.70±0.12, 2.04±0.26, 1.92±0.16, 4.35±0.54, respectively.

Because of the importance of the complement pathway in AMD, secretion of C3 was further investigated in ARPE-19 cells treated with different doses of TNF-α. Vinculin, whose abundance remained unchanged between TNF-α treated and untreated cells, was used as a loading control for data normalization while IκB-α degradation was used to monitor the TNF-α induced NF-κB pathway. As shown in Figure 6, the secretion of C3 was increased with TNF-α in dose dependant manner while I-κB significantly decreased, indicating activation of the NF-κB pathway. This experiment was repeated three times, yielding same trend of C3 expression as a function of TNF-α concentration.

Complement C3

The secretion of this protein was significantly induced by TNF-α in ARPE-19 cells. This result is in agreement with previous studies where it has been shown that cytokines regulate C3 production in different cell types including human intestinal epithelial cell line, Caco-2 [38] as well as rat glomerular endothelial cells [39]. The production of C3 in interleukin-1-treated intestinal cells was abolished by NF-κB inhibitors [38], indicating that C3 is a target gene for NF-κB transcription factor. It is most likely that a similar pathway is involved in TNF-α-treated ARPE-19. However, to the best of our knowledge, this is the first time that TNF-α has been shown to induce C3 secretion in human RPE cells. Local secretion of C3 by RPE cells could be detrimental in individuals carrying Y402H CFH variant since CFH is the physiologic inhibitor of C3 activation and protect cells from autologous injuries by complement C3 [40]. Additionally, it has been shown that monocytes, which are a major source of TNF-α, are attracted by sub-RPE deposits [41]. Therefore, in individuals with compromised CFH activity due to the Y402H mutation, this could generate a positive-feedback loop whereby the newly attracted mononcytes generate a chronic exposure of RPE cells to cytokines. In turn, this leads to increased secretion of C3 by these cells, leading to progressive immune complex deposition and autologous complement-mediated RPE injury. This hypothesis remains however to be verified in a model system using primary RPE cells with mutated and wild type CFH.

Fibronectin and syndecan-4

Secretion of fibronectin and syndecan-4 was significantly induced in TNF-α-treated ARPE-19 cells. Syndecan-4, especially, was threefold to fivefold more abundant in the spent medium of TNF-α-treated cells versus untreated cells. These two proteins are major constituents of the extracellular matrix (ECM) and play a crucial role in cell adhesion and connectivity. While fibronectin is an extracellular protein that adheres to the cell surface to maintain their integrity and shape, syndecan-4 is a single pass transmembrane protein composed of an extracellular domain “ectodomain” bearing four heparan sulfate chains, a transmembrane domain, and a short C-terminal cytoplasmic domain.

Even though treatment of ARPE-19 with TNF-α resulted in a significant increase of fibronectin secretion neither the literature nor software prediction support the involvement of the NF-κB pathway in this process, suggesting that this protein might be indirectly upregulated. Indeed, fibronectin was identified in SDS–PAGE around 100 to 250 kDa indicating that the protein was fragmented [42]. Upregulation of fibronectin fragments might be due to proteases involved in extracellular matrix degradation such as the serine protease HtrA1 [41]. Interestingly, mutation in the promoter region of the gene encoding for HTrA1 was recently found to be strongly associated with AMD pathology [9–11]. Nevertheless, the exact mechanism by which this single point mutation leads to drusen formation and macular degeneration remains to be further investigated. Although the molecular mass of syndecan-4 core protein is around 21 kDa, this protein was detected at around 250 kDa in our SDS–PAGE. This abnormal SDS–PAGE migration is most probably due to the O-linked glycosaminglycan chains. This hypothesis is in agreement with previous studies where it has been shown that intact syndecan migrates at higher molecular mass in the gel than does heparinase-treated syndecan [43]. All the peptides identifying this protein in our study belonged to the ectodomain sequence (Figure 7) suggesting that the N-terminal part of the protein is shed into the spent medium of TNF-α treated ARPE-19 cells. Previous study has shown that TNF-α induces the overall expression of syndecan-4 in endothelial cells via NF-κB activation [44]. Other studies have shown that syndecan-4 shedding is mediated by matrix metalloproteinase-9 (MMP9) and is dependent of NF-κB signaling [45]. In our proteomic profiling data we did detect 2 peptides for MMP9 and these were found to be more abundant in TNF-α-treated than untreated cells (data not shown). Unfortunately, due to the inherent low amount of this protein in the spent medium of ARPE-19, we were not able to detect these same peptides in our reverse experiment. What remains to be examined is whether the increased amount of syndecan-4 seen in the spent medium of TNF-α-treated ARPE-19 cells was due to induction of its expression via NF-κB activation or to the increased shedding of its ectodomain. In either case, shedding of this from the surface of ARPE-19 cells exposed to TNF-α could be a protective mechanism against TNF-α cytotoxicity, and this may explain the resistance of these cells to TNF-α-induced apoptosis [32]. Indeed, it has been documented that the TNF-α signaling pathway synergizes with the binding of matricellular protein CCN1/CYR61 to syndecan-4 ectodomain and triggers a cascade leading to apoptosis [46]. Nevertheless, the contribution of this shedding to AMD pathogenesis remains to be carefully examined even though some studies demonstrated that drusen contained some glycosaminoglycan and heparan sulfate-conjugated proteoglycans [47,48]. Also, at present it is uncertain whether CFH, an important factor in AMD pathology, binds to heparan sulfate of syndecans and whether shedding of syndecan will make cell more vulnerable to increased complement C3 secretion. More studies on the interactions between cell surface proteoglycan and CFH (both the Y402H variant and the wild type) are needed.

Plasminogen activator inhibitor-1

Based on the spectral count of total peptides generated per protein, plasminogen activator inhibitor-1 (PAI-1) appears as one of the highly abundant proteins in the spent medium of ARPE-19 cells (29 peptides and 1474 spectral count were detected for this protein even under basal conditions). The secretion of PAI-1 was further increased by twofold in TNF-α-treated ARPE-19 cells versus untreated cells. Previous studies have shown that TNF-α readily induces PAI-1 in different cell types [49–52]. PAI-1 is a physiologic inhibitor of plasminogen activator and plays an important role in the regulation of the extracellular matrix. Its expression is increased in several fibrotic diseases and regulated by various cytokines, growth factors, and chemical/physical agents [53]. RPE cells secrete both urokinase plasminogen activator and its inhibitor PAI-1 [54] and the balance between these two components likely plays a crucial role in maintaining the extracellular environment of RPE cells including regulation of angiogenesis [55–58]. Interestingly, PAI-1 was found to be more abundant in the serum of patients with early or advanced AMD than of healthy individuals, suggesting its potential implication in AMD pathogenesis [59].

Sulfhydryl oxidase 1

A protein also known as quiescin Q6 (QSOX), sulfhydryl oxidase 1 has 2 known isoforms generated by alternative splicing. Isoform 1, a single pass membrane protein, can be found in the endoplasmic reticulum and Golgi apparatus while isoform 2, missing the C-terminal domain of isoform 1, is mainly found in the extracellular space [60]. All the 17 peptides identified for this protein in our study belonged to isoform 2, consistent with a secreted form. In other words, we did not detect any peptide sequence belonging to the C-terminal domain of isoform 1. QSOX protein utilizes a thioredoxin domain and a small flavin adenine dinucleotide (FAD)-binding domain to form disulfide bonds in proteins and peptides especially in proteins destined for secretion [60]. QSOX is typically found in the cell along the secretory machinery and in the extracellular space. However, little is known about the secreted form and its significance. Whether this secreted form has a role in disulfide bond formation of proteins in the extracellular space or is merely carried along with proteins during their secretion remain to be determined. Nevertheless and to the best of our knowledge, this is the first demonstration that TNF-α induces secretion of QSOX in human RPE cells despite the lack of a binding site for activated NF-κB promoter region of its gene. It is believed that this protein may play a crucial role in ECM remodeling [61] which might be relevant in diseases involving ECM deposition such as AMD, Alzheimer and fibrosis.

Trans-golgi network protein 2

TGON2 was the only protein that was found to be decreased in the spent medium of TNF-α-treated versus untreated ARPE-19 cells. While TGON2 was decreased in the extracellular space of treated cells, its amount was actually increased in the microsomal fraction of this same set of cells (Figure 8), indicating that the changes were due to retention of this protein rather than inhibition of its expression. Indeed, TGON2 is involved in regulating membrane traffic to and from the trans-Golgi network. RPE cells express both TNF receptor-associated factor-1 and 2 (TRAF-1 and TRAF-2) at their surface [33]. The decreased amount of TGON2 at the surface of TNF-α-treated cells could be associated with the internalization of TNF-α/ TNF-α receptor (TNFR) complex to modulate TNF-signaling [62,63]. Indeed, it has been shown that TNFR can be localized both at the cell surface and to the Golgi apparatus of endothelial cells, and that TNF-α binding modulates its trafficking [64].

Overall this study showed that TNF-α modulate the RPE secretion of a variety of proteins that play a major role in extracellular matrix remodeling, complement network and angiogenesis which might be relevant to AMD pathogenesis. Although preliminary, this study demonstrated that TNF-α induced secretion of specific proteins in ARPE19 cells. Even though these cells are not an ideal RPE model in respect to primary cultures, they were used as proof of principle to demonstrate that several secreted proteins can be qualitatively and quantitatively monitored in TNF-α-treated versus untreated cells. These proteins set may serve as a catalog for future targeted studies that use human primary RPE cells or human retinal cross-sections.