Abstract: Retinal vascular hyperpermeability causes macular edema, leading to visual deterioration in retinal diseases such as diabetic retinopathy and retinal vascular occlusion. Dysregulation of junction integrity between endothelial cells by vascular endothelial growth factor (VEGF) was shown to cause retinal vascular hyperpermeability. Accordingly, anti-VEGF agents have been used to treat retinal vascular hyperpermeability. However, they can confer potential toxicity through their deleterious effects on maintenance and survival of neuronal and endothelial cells in the retina. Thus, it is important to identify novel therapeutic targets for retinal vascular hyperpermeability other than VEGF. Here, we prepared murine retinas showing VEGF-induced vascular leakage from superficial retinal vascular plexus and prevention of VEGF-induced leakage by anti-VEGF antibody treatment. We then performed comprehensive proteome profiling of these samples and identified retinal proteins for which abundances were differentially expressed by VEGF, but such alterations were inhibited by anti-VEGF antibody. Functional enrichment and network analyses of these proteins revealed the β2 integrin pathway, which can prevent dysregulation of junction integrity between endothelial cells through cytoskeletal rearrangement, as a potential therapeutic target for retinal vascular hyperpermeability. Finally, we experimentally demonstrated that inhibition of the β2 integrin pathway salvaged VEGF-induced retinal vascular hyperpermeability, supporting its validity as an alternative therapeutic target to anti-VEGF agents. Retinal vascular hyperpermeability causes macular edema, leading to visual deterioration in retinal diseases such as diabetic retinopathy and retinal vascular occlusion. Dysregulation of junction integrity between endothelial cells by vascular endothelial growth factor (VEGF) was shown to cause retinal vascular hyperpermeability. Accordingly, anti-VEGF agents have been used to treat retinal vascular hyperpermeability. However, they can confer potential toxicity through their deleterious effects on maintenance and survival of neuronal and endothelial cells in the retina. Thus, it is important to identify novel therapeutic targets for retinal vascular hyperpermeability other than VEGF. Here, we prepared murine retinas showing VEGF-induced vascular leakage from superficial retinal vascular plexus and prevention of VEGF-induced leakage by anti-VEGF antibody treatment. We then performed comprehensive proteome profiling of these samples and identified retinal proteins for which abundances were differentially expressed by VEGF, but such alterations were inhibited by anti-VEGF antibody. Functional enrichment and network analyses of these proteins revealed the β2 integrin pathway, which can prevent dysregulation of junction integrity between endothelial cells through cytoskeletal rearrangement, as a potential therapeutic target for retinal vascular hyperpermeability. Finally, we experimentally demonstrated that inhibition of the β2 integrin pathway salvaged VEGF-induced retinal vascular hyperpermeability, supporting its validity as an alternative therapeutic target to anti-VEGF agents. Macular edema resulting from increased vascular permeability leads to visual deterioration in a broad spectrum of retinal diseases, including diabetic retinopathy and retinal vein occlusion (1Campochiaro P.A. Wykoff C.C. Singer M. Johnson R. Marcus D. Yau L. Sternberg G. Monthly versus as-needed ranibizumab injections in patients with retinal vein occlusion: The SHORE study.Ophthalmology. 2014; 121: 2432-2442Abstract Full Text Full Text PDF PubMed Scopus (62) Google Scholar, 2Campochiaro P.A. Wykoff C.C. Shapiro H. Rubio R.G. Ehrlich J.S. Neutralization of vascular endothelial growth factor slows progression of retinal nonperfusion in patients with diabetic macular edema.Ophthalmology. 2014; 121: 1783-1789Abstract Full Text Full Text PDF PubMed Scopus (141) Google Scholar). In vitreous samples from patients with macular edema, the concentrations of vascular endothelial growth factor (VEGF) were reported to be higher than those of normal controls (3Ehlken C. Rennel E.S. Michels D. Grundel B. Pielen A. Junker B. Stahl A. Hansen L.L. Feltgen N. Agostini H.T. Martin G. Levels of VEGF but not VEGF(165b) are increased in the vitreous of patients with retinal vein occlusion.Am. J. Ophthalmol. 2011; 152: 298-303Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar, 4Shimada H. Akaza E. Yuzawa M. Kawashima M. Concentration gradient of vascular endothelial growth factor in the vitreous of eyes with diabetic macular edema.Invest. Ophthalmol. Vis. Sci. 2009; 50: 2953-2955Crossref PubMed Scopus (41) Google Scholar). A number of in vitro and in vivo studies have shown that VEGF, also known as vascular permeability factor, caused loss of tightness between endothelial cells and resultant vascular leakage (5Tolentino M.J. Miller J.W. Gragoudas E.S. Jakobiec F.A. Flynn E. Chatzistefanou K. Ferrara N. Adamis A.P. Intravitreous injections of vascular endothelial growth factor produce retinal ischemia and microangiopathy in an adult primate.Ophthalmology. 1996; 103: 1820-1828Abstract Full Text PDF PubMed Scopus (449) Google Scholar, 6Kim J.H. Kim J.H. Lee Y.M. Ahn E.M. Kim K.W. Yu Y.S. Decursin inhibits VEGF-mediated inner blood-retinal barrier breakdown by suppression of VEGFR-2 activation.J. Cereb. Blood Flow Metab. 2009; 29: 1559-1567Crossref PubMed Scopus (31) Google Scholar, 7Maeng Y.S. Maharjan S. Kim J.H. Park J.H. Suk Yu Y. Kim Y.M. Kwon Y.G. Rk1, a ginsenoside, is a new blocker of vascular leakage acting through actin structure remodeling.PLoS ONE. 2013; 8: e68659Crossref PubMed Scopus (27) Google Scholar). In monkeys, intravitreal injection of bioactive VEGF led to fluorescein leakage from retinal vessels (5Tolentino M.J. Miller J.W. Gragoudas E.S. Jakobiec F.A. Flynn E. Chatzistefanou K. Ferrara N. Adamis A.P. Intravitreous injections of vascular endothelial growth factor produce retinal ischemia and microangiopathy in an adult primate.Ophthalmology. 1996; 103: 1820-1828Abstract Full Text PDF PubMed Scopus (449) Google Scholar). Furthermore, VEGF was shown to affect the integrity of tight junction complexes along endothelial cell-cell interfaces (6Kim J.H. Kim J.H. Lee Y.M. Ahn E.M. Kim K.W. Yu Y.S. Decursin inhibits VEGF-mediated inner blood-retinal barrier breakdown by suppression of VEGFR-2 activation.J. Cereb. Blood Flow Metab. 2009; 29: 1559-1567Crossref PubMed Scopus (31) Google Scholar, 7Maeng Y.S. Maharjan S. Kim J.H. Park J.H. Suk Yu Y. Kim Y.M. Kwon Y.G. Rk1, a ginsenoside, is a new blocker of vascular leakage acting through actin structure remodeling.PLoS ONE. 2013; 8: e68659Crossref PubMed Scopus (27) Google Scholar). These data indicate that VEGF plays a major role in the pathogenesis of retinal vascular hyperpermeability by dysregulating tight junction integrity between endothelial cells in the retinal vascular system (8Jo D.H. Kim J.H. Kim J.H. How to overcome diabetic retinopathy: focusing on blood-retinal barrier.Immunol. Endocr. Metab. Agents Med. Chem. 2012; 12: 110-117Crossref Scopus (8) Google Scholar, 9Miller J.W. Le Couter J. Strauss E.C. Ferrara N. Vascular endothelial growth factor A in intraocular vascular disease.Ophthalmology. 2013; 120: 106-114Abstract Full Text Full Text PDF PubMed Scopus (305) Google Scholar). It has been observed in the clinic that anti-VEGF agents have therapeutic potential to treat retinal vascular hyperpermeability induced by VEGF (8Jo D.H. Kim J.H. Kim J.H. How to overcome diabetic retinopathy: focusing on blood-retinal barrier.Immunol. Endocr. Metab. Agents Med. Chem. 2012; 12: 110-117Crossref Scopus (8) Google Scholar, 9Miller J.W. Le Couter J. Strauss E.C. Ferrara N. Vascular endothelial growth factor A in intraocular vascular disease.Ophthalmology. 2013; 120: 106-114Abstract Full Text Full Text PDF PubMed Scopus (305) Google Scholar). However, VEGF not only promotes vascular permeability, but also plays an important role in the survival of normal endothelial and neuronal cells in the retina (10Heo J.W. Kim J.H. Cho C.S. Jun H.O. Kim D.H. Yu Y.S. Kim J.H. Inhibitory activity of bevacizumab to differentiation of retinoblastoma cells.PLoS ONE. 2012; 7: e33456Crossref PubMed Scopus (26) Google Scholar, 11Kurihara T. Westenskow P.D. Bravo S. Aguilar E. Friedlander M. Targeted deletion of Vegfa in adult mice induces vision loss.J. Clin. Invest. 2012; 122: 4213-4217Crossref PubMed Scopus (239) Google Scholar). Furthermore, VEGF affects various downstream intracellular molecular pathways that are associated with the maintenance of retinal endothelial cells (9Miller J.W. Le Couter J. Strauss E.C. Ferrara N. Vascular endothelial growth factor A in intraocular vascular disease.Ophthalmology. 2013; 120: 106-114Abstract Full Text Full Text PDF PubMed Scopus (305) Google Scholar, 12Chen X.L. Nam J.O. Jean C. Lawson C. Walsh C.T. Goka E. Lim S.T. Tomar A. Tancioni I. Uryu S. Guan J.L. Acevedo L.M. Weis S.M. Cheresh D.A. Schlaepfer D.D. VEGF-induced vascular permeability is mediated by FAK.Dev. Cell. 2012; 22: 146-157Abstract Full Text Full Text PDF PubMed Scopus (249) Google Scholar). These data collectively suggest potential toxicity of anti-VEGF agents when they were used to treat retinal vascular hyperpermeability. In this context, it is important to identify novel therapeutic targets other than VEGF to treat VEGF-induced retinal vascular hyperpermeability. Here, in search of new therapeutic targets, we first prepared mouse retina that exhibited VEGF-induced vascular leakage from the superficial vascular plexus and effective phenotypical prevention of vascular leakage by anti-VEGF antibody, as described previously (13Jones C.A. London N.R. Chen H. Park K.W. Sauvaget D. Stockton R.A. Wythe J.D. Suh W. Larrieu-Lahargue F. Mukouyama Y.S. Lindblom P. Seth P. Frias A. Nishiya N. Ginsberg M.H. et al.Robo4 stabilizes the vascular network by inhibiting pathologic angiogenesis and endothelial hyperpermeability.Nat. Med. 2008; 14: 448-453Crossref PubMed Scopus (313) Google Scholar, 14Scheppke L. Aguilar E. Gariano R.F. Jacobson R. Hood J. Doukas J. Cao J. Noronha G. Yee S. Weis S. Martin M.B. Soll R. Cheresh D.A. Friedlander M. Retinal vascular permeability suppression by topical application of a novel VEGFR2/Src kinase inhibitor in mice and rabbits.J. Clin. Invest. 2008; 118: 2337-2346PubMed Google Scholar). We then performed comprehensive proteome profiling of the retinas treated with VEGF and with VEGF plus anti-VEGF antibody by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. Using the isobaric tag for relative and absolute quantitation (iTRAQ) 1The abbreviations used are:iTRAQisobaric tag for relative and absolute quantitationPSMpeptide-spectrum matchFDRfalse discovery rateDEPdifferentially expressed proteinGOBPgene ontology biological processHRMEChuman retinal microvascular endothelial cellFAKfocal adhesion kinasemRP fractionationmid-pH reverse-phase liquid chromatography fractionationFN1fibronectin 1ITGB2β2 integrinF2coagulation factor IICD14CD14 antigenGSNgelsolinMYL2myosin light chain 2MyYLPFmyosin light chain, phosphorylatable, fast skeletal muscleBRBblood-retinal barrier. data obtained from LC-MS/MS analyses, we identified retinal proteins whose abundances were altered by treatment of VEGF, but the alterations were inhibited by cotreatment of anti-VEGF antibody with VEGF. Functional enrichment and network analyses of these proteins suggested the β2 integrin pathway regulating cytoskeletal rearrangement of endothelial cells as a potential therapeutic target for VEGF-induced retinal vascular hyperpermeability. Both in vitro immunoassay and in vivo experiments confirmed that the modulation of the β2 integrin pathway inhibited VEGF-induced retinal vascular hyperpermeability, suggesting its validity as a therapeutic target other than VEGF. isobaric tag for relative and absolute quantitation peptide-spectrum match false discovery rate differentially expressed protein gene ontology biological process human retinal microvascular endothelial cell focal adhesion kinase mid-pH reverse-phase liquid chromatography fractionation fibronectin 1 β2 integrin coagulation factor II CD14 antigen gelsolin myosin light chain 2 myosin light chain, phosphorylatable, fast skeletal muscle blood-retinal barrier. 6-Week-old male C57BL/6 mice (Central Laboratory Animal, Seoul, Republic of Korea) were used in this study. All animal procedures were approved by Institutional Animal Care and Use Committee of Seoul National University and conducted in agreement with the Association for Research in Vision and Ophthalmology statement for the use of animals in ophthalmic and vision research. To induce retinal vascular hyperpermeability, we injected recombinant mouse VEGF164 (100 ng/1.5 μl; catalog no. 493-MV-005, R&D Systems, Minneapolis, MN) into the vitreous cavity of the right eyes of mice (VEGF condition). For control samples, phosphate-buffered saline (PBS) (1.5 μl) was injected (control condition). To show the effects of VEGF scavenging, we injected both recombinant mouse VEGF164 (100 ng) and affinity-purified polyclonal antibody against mouse VEGF164 (1 μg; catalog no. AF-493-NA, R&D Systems; source, goat IgG) in 1.5 μl of PBS (anti-VEGF condition). To analyze therapeutic effects of β2 integrin antagonism, we injected both recombinant mouse VEGF164 (100 ng) and monoclonal antibody against mouse β2 integrin with neutralizing effects of β2 integrin activity (1 μg; catalog no. MA1-10122, Thermo, Waltham, MA; source, rat IgG) in 1.5 μl of PBS. To evaluate the effect of the injection of VEGF, VEGF plus anti-VEGF antibody, or VEGF plus anti-β2 integrin antibody on retinal vascular hyperpermeability, and the levels of proteins measured, the retinas were prepared at 24 h after the injection. At 24 h after the injection, enucleation was performed after deep anesthesia and subsequent sacrifice with CO2 inhalation. To demonstrate vascular integrity, intracardiac injection of fluorescein isothiocyanate-dextran (FITC-dextran, catalog no. 46945, Sigma, St. Louis, MO) was performed 1 h prior to sacrifice with CO2 inhalation. Retinal flat mounts were observed under the fluorescence microscope (Eclipse 80i, Nikon, Tokyo, Japan). Quantitative analyses of mean green fluorescence intensity were performed using four randomly selected images of paracentral areas of retinas from ×200 magnification photographs per each eye with ImageJ software (National Institutes of Health, Bethesda, MD) (n = 3). In each of the three conditions (control, VEGF, and anti-VEGF), we obtained four samples (n = 4). In VEGF or anti-VEGF condition, we obtained four samples after treatment of VEGF or VEGF plus anti-VEGF antibody, respectively. For each sample in the three conditions, we collected the retinas from six mice (six retinas per sample) into the microcentrifuge tube to minimize the effect of interindividual variability in mice. From each of the six mice, enucleation and subsequent sacrifice were performed as described above. Enucleated eyes were immediately put into cold PBS. Under the stereomicroscope (Leica, Wetzlar, Germany), the cornea was dissected, and the lens was removed to facilitate retinal preparation. By gentle pressure on the sclera with forceps, the retinas were isolated and put into clean microcentrifuge tubes, which were prepared in ice. In total, the 12 samples (n = 4 per condition) were stored at −80 °C before further preparation. 20 mg of each sample in the three conditions (control, VEGF, and anti-VEGF) was cryo-pulverized by Covaris CP02 Prep (Covaris, Woburn, MA). Briefly, the retinas were transferred to a Covaris tissue bag (Covaris, TT1, 520007), and the tissue bag was placed in liquid nitrogen for 30 s. Immediately after the freezing, the tissue bag was placed in the impact chamber of the cryo-pulverizer and pulverized by using impact level 2 (tissue weight <50 mg, impact level 2). The cryo-pulverized retinas were transferred into a siliconized low-retention microcentrifuge tube (Thermo) and mixed with lysis buffer (4% SDS in 0.1 m Tris-HCl, pH 7.6, containing a Complete Mini Protease inhibitor tablet (Roche Applied Science, Basel, Switzerland)), followed by tissue lysis using a handheld sonicator (Q55, QSONICA, Newtown, CT) for 30 s (at 30 watts) on ice. The tissue lysate was centrifuged at 16,000 × g and 20 °C for 10 min, and the supernatant was then transferred to a new tube. Protein concentration of the supernatant was determined by BCA assay (Pierce, Waltham, MA). For each sample in the three conditions, the retinal proteins were divided into 500-μg protein units, and each 500-μg protein unit was digested using a slightly modified version of filter-aided sample preparation digestion method (15Wiśniewski J.R. Zougman A. Nagaraj N. Mann M. Universal sample preparation method for proteome analysis.Nat. Methods. 2009; 6: 359-362Crossref PubMed Scopus (5041) Google Scholar). Briefly, the proteins were reduced with SDT buffer (4% SDS in 0.1 m Tris-HCl, pH 7.6, and 0.1 m DTT) for 45 min at 37 °C and then boiled for 10 min at 95 °C. Subsequently, protein samples were sonicated for 10 min in a bath sonicator (Power Sonic 505, Hwashin Technology, Seoul, Republic of Korea) and centrifuged at 16,000 × g for 5 min. The protein sample was transferred to a membrane filter device (YM-30, Millipore, Billerica, MA) and mixed with 200 μl of 8 m urea in 0.1 m Tris-HCl, pH 8.5. The device was centrifuged at 14,000 × g at 20 °C for 60 min to remove the SDS. This step was repeated three times. Subsequently, proteins were alkylated with 100 μl of 50 mm iodoacetamide in 8 m urea for 25 min at room temperature in the dark and followed by centrifugation at 14,000 × g for 30 min. The filter was washed with 200 μl of 8 m urea four times and then washed with 100 μl of 50 mm NH4HCO3 twice for buffer exchange. Trypsin (Promega, Madison, WI) was added to the proteins at an enzyme to protein ratio of 1:50 (w/w), and the filter device was placed in a thermomixer (Eppendorf, Hamburg, Germany) and incubated at 37 °C overnight. After the first digestion, the second digestion was carried out with additional trypsin (1:100 enzyme to protein ratio) at 37 °C for 6 h. After digestion, the tryptic peptides were eluted by centrifugation at 14,000 × g at 20 °C for 30 min. After collecting the tryptic peptides, the filter were rinsed with 60 μl of 50 mm NH4HCO3 and centrifuged at 14,000 × g at 20 °C for 20 min, and the eluent was combined with the first eluent. The combined eluent was dried by vacuum centrifugation, and the peptide concentration was determined by BCA assay. The peptide sample was divided into 100-μg units in Eppendorf tubes and kept in −80 °C until the subsequent iTRAQ labeling. We performed three 4-plex iTRAQ labeling experiments (Fig. 1A). For each experiment, we used one sample from the three conditions (control, VEGF, and anti-VEGF) for 114, 115, and 116 channels, respectively, and an additional sample from one of the three conditions for the 117 channel (i.e. experimental sets 1–3 with additional samples from control, VEGF, anti-VEGF conditions, respectively). A total of 800 μg of peptides (200 μg of peptides per channel) from each experimental set were labeled with two units of 4-plex iTRAQ reagent (AB Sciex, Framingham, MA) according to the manufacturer's instructions. After the iTRAQ labeling, all iTRAQ peptides were pooled and immediately subjected to a mid-pH reverse-phase fractionation, as described previously (16Park J.M. Park J.H. Mun D.G. Bae J. Jung J.H. Back S. Lee H. Kim H. Jung H.J. Kim H.K. Lee H. Kim K.P. Hwang D. Lee S.W. Integrated analysis of global proteome, phosphoproteome, and glycoproteome enables complementary interpretation of disease-related protein networks.Sci. Rep. 2015; 5: 18189Crossref PubMed Scopus (25) Google Scholar). 800 μg of iTRAQ-labeled tryptic peptides was separated using Agilent 1260 Infinity HPLC system (Agilent, Santa Clara, CA) equipped with an Xbridge C18 guard column (4.6 × 20 mm, 130 Å, 5 μm) and an analytical column (4.6 × 250 mm, 130 Å, 5 μm). Mid-pH reverse-phase liquid chromatography was performed at a flow rate of 0.5 ml/min during the 130-min gradient using solvent A (10 mm triethylammonium bicarbonate (TEAB) in water, pH 7.5) and solvent B (10 mm TEAB in 90% ACN, pH 7.5). The gradient used is as follows: 0% solvent B for 10 min, 0–5% solvent B in 10 min, 5–35% in 60 min, 35–70% in 15 min, 70% for 10 min, 70–0% in 10 min, and finally held at 0% over 15 min. 96 fractions were collected every minute from 15 to 110 min and were non-contiguously concatenated into 24 fractions by pooling four fractions from each of the early section (fractions 1–24), the first mid-section (fractions 25–48), the second mid-section (fractions 49–72), and the late section (fractions 73–96) of fractions (Fig. 1A). The 24 fractions were dried in a vacuum centrifuge concentrator and stored at −80 °C until LC-MS/MS experiments. 10 μg of iTRAQ-labeled peptides from each of 24 fractions was individually analyzed by Q Exactive mass spectrometer (Thermo), which was coupled to a dual on-line LC system (17Lee H. Lee J.H. Kim H. Kim S.J. Bae J. Kim H.K. Lee S.W. A fully automated dual-online multifunctional ultrahigh pressure liquid chromatography system for high-throughput proteomics analysis.J. Chromatogr. A. 2014; 1329: 83-89Crossref PubMed Scopus (14) Google Scholar). The dual on-line LC system was equipped with two capillary columns (75-μm inner diameter × 360-μm outer diameter, 100 cm) and two solid-phase extraction columns (150-μm inner diameter × 360-μm outer diameter, 3 cm) that were prepared by slurry packing frit-ended fused silica capillaries with C18 resin (3 μm diameter, 300-Å pore size, Jupiter). A 180-min linear gradient (1–40% solvent B over 160 min, 40–80% over 5 min, 80% for 10 min and holding at 1% for 5 min) was used. Solvent A and B were 0.1% formic acid in water and 0.1% formic acid in ACN, respectively. The column flow rate was 300 nl/min. The eluting peptides were ionized at the electric potential of 2.4 kV and the desolvation capillary temperature of 250 °C. MS precursor scans (m/z 400–2,000 thomson) were acquired at the resolution of 70,000 with an automated gain control target value of 1.0 × 106 and a maximum ion injection time of 20 ms. The MS/MS data for up to the 10 most abundant ions were acquired in a data-dependent mode using higher energy collisional dissociation at a normalized collision energy of 30 with fixed first mass of 100 thomson at the resolution of 17,500 with automated gain control target value of 1.0 × 106 and a maximum injection time of 60 ms. Each of the three iTRAQ datasets includes 24 MS/MS datasets for 24 fractions. For each MS/MS dataset, post-experiment monoisotopic mass refinement method was used to process the MS/MS data, which was previously demonstrated to accurately assign precursor mass to the tandem mass spectrometric data (18Shin B. Jung H.J. Hyung S.W. Kim H. Lee D. Lee C. Yu M.H. Lee S.W. Postexperiment monoisotopic mass filtering and refinement (PE-MMR) of tandem mass spectrometric data increases accuracy of peptide identification in LC/MS/MS.Mol. Cell. Proteomics. 2008; 7: 1124-1134Abstract Full Text Full Text PDF PubMed Scopus (46) Google Scholar). The resultant MS/MS data from post-experiment monoisotopic mass refinement process (i.e. mgf files) were subjected to database search using the MS-GF+ search engine (version 9387) (19Kim S. Mischerikow N. Bandeira N. Navarro J.D. Wich L. Mohammed S. Heck A.J. Pevzner P.A. The generating function of CID, ETD, and CID/ETD pairs of tandem mass spectra: applications to database search.Mol. Cell. Proteomics. 2010; 9: 2840-2852Abstract Full Text Full Text PDF PubMed Scopus (196) Google Scholar) against Swiss-Prot mouse reference database (released September, 2013; 24,544 entries). Search parameters were set to the following: precursor mass tolerance of 10 ppm, semi-tryptic, static modifications of carbamidomethylation (+57.0214 Da) to cysteine and iTRAQ (+144.102063 Da) to N termini and lysine, and variable modification of oxidation (+15.994915 Da) to methionine. The search results from the 24 MS/MS datasets were combined. The peptide-spectrum matches (PSMs) at the false discovery rate (FDR) of 1% were obtained by the target-decoy method built in the MS-GF+ search engine (19Kim S. Mischerikow N. Bandeira N. Navarro J.D. Wich L. Mohammed S. Heck A.J. Pevzner P.A. The generating function of CID, ETD, and CID/ETD pairs of tandem mass spectra: applications to database search.Mol. Cell. Proteomics. 2010; 9: 2840-2852Abstract Full Text Full Text PDF PubMed Scopus (196) Google Scholar). The identified peptides with PSM-level FDR <0.01 from the three iTRAQ datasets were combined to generate an alignment table in which the identified peptides from different iTRAQ datasets with the same sequences were matched. The identified peptides in the alignment table were used to infer protein groups by a bipartite graph analysis (21Zhang B. Chambers M.C. Tabb D.L. Proteomic parsimony through bipartite graph analysis improves accuracy and transparency.J. Proteome Res. 2007; 6: 3549-3557Crossref PubMed Scopus (264) Google Scholar) using an in-house software (22Kim S.J. Chae S. Kim H. Mun D.G. Back S. Choi H.Y. Park K.S. Hwang D. Choi S.H. Lee S.W. A protein profile of visceral adipose tissues linked to early pathogenesis of type 2 diabetes mellitus.Mol. Cell. Proteomics. 2014; 13: 811-822Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar). Finally, we selected protein groups that have protein-level FDR ≤1% and more than two detected non-redundant peptides. Global protein-level FDR was estimated in the target-decoy setting as described previously (20Tang W.H. Shilov I.V. Seymour S.L. Nonlinear fitting method for determining local false discovery rates from decoy database searches.J. Proteome Res. 2008; 7: 3661-3667Crossref PubMed Scopus (273) Google Scholar). The representative protein of protein group was selected as the protein with the highest number of peptides. When more of two proteins in one group had the same number of peptides, the protein with the higher sequence coverage was selected as the representative protein, and the protein containing a unique peptide also was selected as the representative protein. Protein groups of two or more peptide hits were used for further analyses. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (23Vizcaíno J.A. Csordas A. Del-Toro N. Dianes J.A. Griss J. Lavidas I. Mayer G. Perez-Riverol Y. Reisinger F. Ternent T. Xu Q.W. Wang R. Hermjakob H. 2016 update of the PRIDE database and related tools.Nucleic Acids Res. 2016; 44: D447-D456Crossref PubMed Scopus (2775) Google Scholar) partner repository with the dataset identifier PXD003656 and 10.6019/PXD003656. Of the identified peptides, we first selected the unique peptides that were detected in two or more of the three iTRAQ experiments. The reporter ion intensities of these selected unique peptides in the alignment table were normalized by the medians of control samples in the corresponding iTRAQ experiment to remove the batch effects in iTRAQ experiments. Using the normalized intensities, the previously reported integrative statistical method (24Chae S. Ahn B.Y. Byun K. Cho Y.M. Yu M.H. Lee B. Hwang D. Park K.S. A systems approach for decoding mitochondrial retrograde signaling pathways.Sci. Signal. 2013; 6: rs4Crossref PubMed Scopus (136) Google Scholar) was performed to identify differentially expressed peptides in the two comparisons: 1) VEGF versus control (VEGF/control) and 2) anti-VEGF versus VEGF (anti-VEGF/VEGF). Briefly, two sample t tests (e.g. four samples in VEGF versus four samples in control) and log2-median ratio test were applied to calculate T values and log2-median ratios for the selected peptides. To compute p values of selected peptides for the individual tests, empirical distributions of T values and log2-median ratios for the null hypothesis (i.e. a peptide is not differentially expressed) were estimated by performing all possible random permutations of the samples and then by applying the Gaussian kernel density estimation method to T values, and log2-median ratios resulted from the random permutations (25Bowman A.W. Azzalini A. Applied Smoothing Techniques for Data Analysis: The Kernel Approach with S-Plus Illustrations. Oxford University Press, Oxford, NY1997: 25-47Google Scholar). For each peptide, the adjusted p values from the two tests were computed by the two-sided test using the corresponding empirical null distributions and then combined into an overall p value using Stouffer's method (26Hwang D. Rust A.G. Ramsey S. Smith J.J. Leslie D.M. Weston A.D. de Atauri P. Aitchison J.D. Hood L. Siegel A.F. Bolouri H. A data integration methodology for systems biology.Proc. Natl. Acad. Sci. U.S.A. 2005; 102: 17296-17301Crossref PubMed Scopus (257) Google Scholar). Next, FDRs were estimated for the overall p values using the Storey method (27Storey J.D. Tibshirani R. Statistical significance for genome-wide studies.Proc. Natl. Acad. Sci. U.S.A. 2003; 100: 9440-9445Crossref PubMed Scopus (7083) Google Scholar). The differentially expressed peptides were selected as the ones with FDR ≤0.05 and absolute log2 fold-changes ≥0.58 (1.5-fold). Finally, we selected a set of DEPs that have at least two differentially expressed peptides. Functional enrichment analysis was performed using DAVID software to identify the GOBPs represented by the DEPs in the two major clusters (clusters 1 and 2 in Fig. 2B) (28Huang da W. Sherman B.T. Lempicki R.A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources.Nat. Protoc. 2009; 4: 44-57Crossref PubMed Scopus (25331