JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

Available online at www.sciencedirect.com

ScienceDirect www.elsevier.com/locate/jprot

Comparative proteomic study in serum of patients with primary open-angle glaucoma and pseudoexfoliation glaucoma☆ Héctor González-Iglesiasa,1 , Lydia Álvareza,1 , Montserrat Garcíaa , Julio Escribanob , Pedro Pablo Rodríguez-Calvoa , Luis Fernández-Vegaa , Miguel Coca-Pradosa,c,⁎ a

Fundación de Investigación Oftalmológica, Instituto Oftalmológico Fernandez-Vega, Avenida Doctores Fernández-Vega, 34, Oviedo 33012, Spain Laboratorio de Genética Molecular Humana, Facultad de Medicina/Instituto de Investigación en Discapacidades Neurológicas (IDINE), Universidad de Castilla-La Mancha, Albacete, 02006, Spain c Department of Ophthalmology and Visual Science, Yale University School of Medicine, New Haven, CT. 06510, USA b

AR TIC LE I N FO

ABS TR ACT

Article history:

Alterations in the sera proteins between patients with Primary Open-Angle Glaucoma (POAG),

Received 5 November 2013

Pseudoexfoliation Glaucoma (PEXG), and healthy controls were identified through a proven

Accepted 9 December 2013

approach utilizing equalization of high-abundance serum proteins with ProteoMiner™,

Available online 16 December 2013

two-dimensional fluorescent difference gel electrophoresis (2D-DIGE), MALDI-TOF/TOF, and nanoLC-MS-MS. Quantitative immunoassays of the 17 most-differentially-altered proteins

Keywords:

identified in this analysis confirmed that they were also over expressed in the intact serum of

Glaucoma

newly recruited glaucoma patients. Overall, this report identifies a panel of candidates for

Serum

glaucoma biomarkers and supports their further validation in large population studies.

Equalization

Additionally, functional pathway analysis of these candidate proteins suggested that they are

Biomarkers

part of a network linked to regulating immune and inflammatory-related processes. The data

2D-DIGE

have been deposited to the ProteomeXchange with identifier PXD000198.

Networks analysis Biological significance POAG and PEXG are major causes of age-related blindness in the world; however, treatment can be very effective if they are identified early on in the progression. Genetic linkage studies can only explain a limited number of cases, suggesting that these forms of glaucoma are multigenic in nature. Other important factors, such as modifier genes, epigenetic influences, environ-

Abbreviations: CA, correct assignment; CyDye, cyanine dyes; IPA, ingenuity pathway analysis; IOFV, Institute of Ophthalmology Fernandez-Vega; IOP, intraocular pressure; PEXG, pseudoexfoliation glaucoma; POAG, primary open-angle glaucoma; ROC, receiver operating characteristic; SNPs, single nucleotide polymorphisms. ☆ Financial disclosure: This study has been supported in part by a CENIT-CeyeC research grant CEN-20091021 from the Spanish Ministry of Innovation and Development, Fundación de Investigación Oftalmológica Fernández-Vega (http://fio.fernandez-vega.com), Fundación Ma Cristina Masaveu Peterson (http://www.fundacioncristinamasaveu.com), Fundación Rafael del Pino (http://www.frdelpino.es), and Cooperative Research Network on prevention, diagnosis and treatment of prevalent, degenerative and chronic eye diseases, Instituto de Salud Carlos III (RD07/0062/0014 and RD12/0034; http://www.retics.net). Miguel Coca-Prados is “Catedrático Rafael del Pino en Oftalmología” in the “Fundación de Investigación Oftalmológica, Instituto Oftalmológico Fernández-Vega” Oviedo, Spain. The authors declare that the results reported in this manuscript may be of commercial interest to Instituto Oftalmológico Fernández-Vega, Oviedo, Spain. ⁎ Corresponding author at: Department of Ophthalmology and Visual Science, Yale University School of Medicine, 300, George St, R8100A, New Haven, CT. 06510, USA. Tel.: +1 203 785 2742; fax: + 1 203 785 7401. E-mail address: [email protected] (M. Coca-Prados). 1 These authors contributed equally to this work. 1874-3919/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jprot.2013.12.006

66

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

mental and dietary agents, and inflammatory and oxidative effects are also believed to affect the development of these diseases. The characterization of metabolic and/or proteins changes, for example in bodily fluids, before the clinical manifestation of glaucoma is of considerable relevance for its early diagnosis. In the present work, identification of over-expressed proteins in serum of glaucoma patients (POAG and PEXG) linked to immune and inflammatory processes supports the finding that changes in these pathways also manifest systemically in patients with these pathologies. This study provides a new basis to validate the identified proteins as biomarkers of glaucoma in a large-scale-multiplexed screening in sera. © 2013 Elsevier B.V. All rights reserved.

1.

Introduction

A leading cause of blindness worldwide, glaucoma encompasses a complex group of neurodegenerative disorders that are multigenic and multifactorial in origin, but are all characterized by progressive degeneration of the optic nerve, retinal ganglion cell death and the loss of the visual field. Primary Open-Angle Glaucoma (POAG) and Pseudoexfoliation Glaucoma (PEXG) are among the most prevalent types of glaucoma in developed countries. Whereas an abnormal elevation in the intraocular pressure (IOP) is the best known risk factor associated with the pathogenesis and progression of POAG, the presence of pseudoexfoliation syndrome is the most common identifiable cause of PEXG, a secondary form of glaucoma. The number of people worldwide with glaucoma has been estimated to reach 80 million in 2020 [1]. In certain regions of the world, including the northwest of Spain (i.e., Asturias and Galicia) and some provinces of Saudi Arabia, the prevalence of PEXG among patients with pseudoexfoliation syndrome can reach up to 30% [2]. The increased IOP observed among POAG subjects is usually associated with a dysfunction/obstruction of the normal exit of the aqueous humor fluid through the outflow system (i.e., trabecular meshwork). In PEXG there is an excessive production and progressive accumulation of fibrillar aggregates from all tissues of the anterior segment of the eye (i.e., corneal endothelium, iris, lens capsule, ciliary epithelium) and deposition of it on the anterior chamber structures (i.e., trabecular meshwork) leads to reduced outflow and thus to elevated IOP [3,4]. Additional clinical differences between POAG and PEXG include: (i) a more elevated IOP in PEXG patients than those with POAG; (ii) a lack of an IOP response to steroids among PEXG patients when compared to those with POAG; and (iii) quantitative histological differences of cross-sections of the optic nerve [5–7]. The frequency of onset and rate of optic degeneration in both POAG and PEXG increase with age; however, the prognosis of blindness is greater among PEXG. Among POAG patients, usually both eyes are affected, but two-thirds of PEXG patients present unilateral disease, and the chance of developing glaucoma in the fellow eye is 50% in 15 years [8]. In most cases of glaucoma, degeneration of the optic nerve head precedes detectable field loss, and it has been estimated that by the time early visual field defects are found, 25% to 35% of retinal ganglion cells may already be irrevocably lost [9]. Therefore, successfully treating glaucoma will require methods for earlier detection, thereby preventing disease progression. In spite of the large number of linkage studies, candidate gene reports, and genome-wide association investigations

conducted on POAG and PEXG populations worldwide, only genes with limited population effects have been discovered. These studies have shown, for instance, the limited involvement in POAG of MYOC (myocilin) [10–12] and rare variants of CYP1B1 (Cytochrome P450 1B1) [13,14], and that the discovery of two non-synonymous single-nucleotide polymorphisms (SNPs) in the LOXL1 (lysyl oxidase-like 1) gene are not always associated with PEXG [15]. On the other hand, investigations conducted in the field of differential proteomics in search for molecular biomarkers of glaucoma have been limited. The search for protein biomarkers in readily available biological fluids (those that do not require invasive surgery to obtain) may be beneficial in the diagnosis and management of glaucoma [16]. In the present work, we carried out a comparative differential proteomic analysis of blood serum from patients with POAG, PEXG, and healthy controls. The aim was to identify, in “equalized” serum, proteins that are quantitatively altered in glaucoma patients, and determine whether said proteins are significantly discriminatory between glaucoma patients and healthy subjects with reproducible results in “non-equalized” serum samples from newly recruited glaucoma patients. We employed many techniques in a comprehensive proteomic workflow including equalization of high-abundance serum proteins with ProteoMiner™, two-dimensional difference gel electrophoresis (2D-DIGE) analysis, and protein identification by matrix assisted laser desorption/ionization time-of-flight/time of flight (MALDI-TOF/TOF) and nano-liquid-chromatography tandem mass spectrometry (nLC-MS/MS). This workflow yielded the identification of a panel of 35 proteins that were detected at different levels in POAG, PEXG and healthy subjects. Alterations in the top-17-ranked proteins of the 35-protein panel were verified by quantitative immunoassays such as enzyme-linked immunosorbent assay (ELISA). Lastly, bioinformatic analysis of the 35-protein panel revealed that they are linked to a network significantly enriched in immune- and inflammatory-related pathways. The present findings create an approach to validate the 35-protein panel as potential serum biomarkers for the clinical prediction, prognosis, diagnosis and monitoring of POAG and PEXG cases at large scale.

2.

Materials and methods

2.1.

Study design

The procedures used in the present study are grouped into two steps, outlined here, described in detail below, and summarized in Fig. 1.

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

potential candidate markers identified by “Differential expression analysis” described in Step 1, commercially available ELISAs of individual proteins indicated in the text were performed with whole serum samples (i.e., no equalization with ProteMiner™) following the instructions described by each of the manufacturers. Serum samples were collected from n = 53 newly recruited participants (POAG, PEXG and control). Concentrations of specific proteins, determined by ELISA, were expressed with respect to serum volume (mg/dL); and ix) Statistical analysis: Differences in concentrations of candidate biomarkers were analyzed using the nonparametric Kruskal–Wallis test (Dunn's test for Multiple Comparisons). A p-value less than 0.05 was considered statistically significant. Receiver Operator Characteristic (ROC) curve analysis was performed individually with each of the markers. In addition, different machine learning approaches were also tested, selecting the Naïve Bayes algorithm to assess the correct classification of samples for each candidate marker. Statistical analyses were carried out using Orange Canvas software v2.6 (http:// orange.biolab.si) [17].

Study population

Sample collection/ Serum preparation

Step 1 Differential expression analysis

2D-DIGE

Altered Proteins Identification

Functional Pathways Analysis

Potential Candidate Markers Verification on new recruited patients

Step 2 Screening ELISA

67

ELISA Quantification

Statistical analysis

2.2.

Ethics statement

Candidate markers to test

Fig. 1 – Study design. The work flow consisted of two steps: Step 1 (Differential expression analysis) including patients selection, sample collection, serum sample preparation, 2D-DIGE analysis, altered protein identification and functional pathways analysis. Step 2 (ELISA screening) including potential candidate markers validation by ELISA on new recruited patients and selection of candidate markers to test. Step 1 Differential expression analysis, included: i) Study population. POAG, PEXG, and control subjects (n = 149) were screened and selected through the Institute of Ophthalmology Fernandez-Vega (IOFV), and classified according to their age and gender; ii) Sample collection. Blood samples were drawn from each participant, and the sera were separated from clotting factors and blood cells by centrifugation; iii) Serum sample preparation. Serum samples were subjected to protein equalization using ProteoMiner™ to reduce the dynamic range of the highly abundant proteins; iv) 2D-DIGE. Labeling of serum proteins was carried out in vitro with cyanine dyes (CyDye) (Cy2, Cy3, Cy5), and the proteins were separated by 2D-DIGE; v) Image Acquisition and Data analysis. Gels were scanned and image analysis was performed with Progenesis SameSpots Software; vi) Protein Identification. Protein spots were isolated and their identification carried out by MALDI TOF/TOF and nLC-MS/MS; and vii) Functional pathways analysis: A network analysis of identified proteins was performed using Ingenuity Pathway Analysis. Step 2 ELISA screening, included: viii) ELISA assays: To quantitatively verify the discriminatory nature of the

The study adheres to the tenets of the Declaration of Helsinki on Biomedical Research Involving Human Subjects, and full ethical approval was obtained from the Clinical Research Ethics Committee at the Hospital Universitario Central de Asturias (Oviedo, Spain). All subjects included in this study were recruited, signed an informed consent, and had complete ophthalmologic examinations at the IOFV.

2.3.

Study population

A total of 202 subjects of Western European descent were enrolled in the present study. In Step 1 of the “Differential expression analysis” (see Fig. 1), a total of 149 subjects (53 POAG, 45 PEXG and 51 control) were recruited. In Step 2 of our study design, the ELISA screening was performed on a second group consisting of a total of 53 newly recruited participants (20 POAG, 14 PEXG and 19 controls). The diagnostic criteria for glaucoma, both primary open-angle and pseudoexfoliative, was the presence of characteristic optic-disc damage (e.g., vertical cup-to-disc ratio > 0.3, thin or notched neuroretinal rim, or disc hemorrhage) with the corresponding characteristic changes in the visual field and the presence of open anterior chamber angle (Shaffer grade III or IV). POAG patients presented without secondary causes of optic neuropathy, while the subjects with PEXG exhibited characteristic exfoliative material on the anterior lens surface and/or iris during slit-lamp examination, in one or both eyes. Control subjects were selected from patients undergoing cataract surgery that did not show indications of glaucoma. Of note, most glaucoma patients also had cataracts (90.4% and 94.9% of POAG and PEXG subjects, respectively). No subjects involved in this study presented with other relevant ocular pathologies such as clinically detectable inflammation, infection, retinopathies, or maculopathies.

68

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

Table 1a – Clinical characteristics of control, POAG and PEXG subjects involved in the differential expression analysis (n = 149). Pathological status Variable Age Mean ± SD a Range p-Value b, vs. Control p-Value b, vs. POAG Sex Men, n (mean age ± SD) a Women, n (mean age ± SD) a IOP c (mean ± SD) a C/D d (mean ± SD) a Visual field e (mean ± SD) a

POAG (n = 53)

PEXG (n = 45)

Control (n = 51)

67.46 ± 11.27 35–89 0.3054 –

73.11 ± 8.43 55–92 0.0001 0.0049

64.93 ± 11.34 43–87 – –

29 (65.00 ± 11.17) 24 (69.92 ± 11.03) 14.9 ± 4.0 0.80 ± 0.15 −9.59 ± 9.58

22 (73.09 ± 8.16) 23 (73.13 ± 8.87) 14.1 ± 5.6 0.69 ± 0.25 −9.67 ± 10.62

24 (64.67 ± 9.33) 27 (65.19 ± 13.05) – – –

a

Data are shown as average ± standard deviation. p-Values are calculated between each pair of groups using the Unpaired t test with Welch correction. c IOP (intraocular pressure) measured under medical treatment. d C/D cup-disc ratio of optic nerve. e Visual Field measured using the Humphrey field analyzer (HFA), Program 24-2, SITA-Standard, and values expressed as mean deviation (MD) where subjects, who are able to see dimmer stimuli than others of similar age and race will have positive values for their MD, while subjects who require brighter stimuli will have negative MD values. b

Detailed clinical characteristics of the population groups included in Step 1 and Step 2 are the following: age, sex, median IOP, visual field parameters, and optic disc excavation. These are summarized in Tables 1a and 1b for Step 1 and Step 2, respectively. Most of the glaucoma patients included in this study (97%) were undergoing ocular treatment at the time of recruitment, summarized in Table 2. More briefly, 36.99% of POAG and 35.59% of PEXG patients were treated with prostaglandin analogs, 5.48% of POAG and 10.17% of PEXG patients with beta-blockers, 52.05% of POAG and 47.46% of PEXG subjects

with both, and finally, 4.11% of the POAG and 5.08% of the PEXG patients were untreated.

2.4.

Sample collection

Blood was collected in 5 mL Z Serum Sep Clot Activator tubes coated with microscopic silica particles, which activate the coagulation process (Vacuette, Madrid, Spain). Tubes were centrifuged at 1800 g for 18 min at 4 ºC, and the supernatant (serum) was stored at −80 ºC until use. Blood from glaucoma and control groups was collected and processed in an identical manner.

Table 1b – Clinical characteristics of newly recruited control, POAG and PEXG subjects involved in the ELISA screening (n = 53). Variable

Age Mean ± SD a Range p-Value b, vs. control p-Value b, vs POAG Sex Men, n (mean age ± SD) a Women, n (mean age ± SD) a IOP c (mean ± SD) a C/D d (mean ± SD) a Visual Field e (mean ± SD) a a

Pathological status POAG (n = 20)

PEXG (n = 14)

Control (n = 19)

70.95 ± 8.54 62–84 0.1379 –

72.86 ± 4.13 66–81 0.0360 0.3942

65.21 ± 14.13 48–92 – –

12 (69.92 ± 8.65) 8 (72.50 ± 8.70) 14 ± 5.9 0.78 ± 0.17 −9 ± 8.93

6 (73.50 ± 4.04) 8 (72.38 ± 4.41) 15.21 ± 8.98 0.69 ± 0.26 −9.24 ± 9.35

11 (65.91 ± 15.08) 8 (64.25 ± 13.67) – – –

Data are shown as average ± standard deviation. p-Values are calculated between each pair of groups using the Unpaired t test with Welch correction. c IOP (intraocular pressure) measured under medical treatment. d C/D cup-disc ratio of optic nerve. e Visual field measured using the Humphrey field analyzer (HFA), Program 24-2, SITA-Standard, and values expressed as mean deviation (MD) where subjects, who are able to see dimmer stimuli than others of similar age and race will have positive values for their MD, while subjects who require brighter stimuli will have negative MD values. b

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

Table 2 – Glaucoma treatment of POAG and PEXG patients enrolled in the study (n = 132).

69

Since the protein concentration in sera samples is very high (60–80 μg/μL), and approximately 95% of all the protein content is represented by fourteen proteins (a high dynamic range), we introduced a preliminary step before 2D-DIGE known as “equalization”. Equalization involved using the ProteoMiner™ Small-Capacity Kit (Catalog #163-3006, BioRad Laboratories, Hercules, CA, USA) to increase the relative concentration of the proteins within the medium and low dynamic ranges [18–20]. Briefly, 300 μL of serum was centrifuged at 10,000 g for 10 min to clarify and ensure the samples to be free of precipitate. After column preparation (see the manufacturer's protocol), 200 μL of serum supernatant was incubated with the hexa-bead columns for 2 h at room temperature on a rotator. In these conditions, a total protein load of 20 mg was added to the column for each individual sample, ensuring saturation of the column. The columns were then centrifuged to remove the liquid, and the beads washed 3 times with wash buffer to remove unbound proteins. Bound proteins were eluted by incubating beads with 50 μL of elution buffer for 15 min at room temperature on a rotator, with this last step repeated three times. Upon equalization, serum proteins were precipitated to remove lipids with CleanUp kit (GE Healthcare, Sunnyvale, CA, USA), solubilized in 30 μl of DIGE compatible focusing solution (30 mM Tris–HCl, pH 8.5, 2 M ThioUrea, 7 M Urea and 4% CHAPS) and stored at −80 ºC until use. The protein concentration in each serum sample was determined according to the Bradford method [21] before its fractionation through ProteoMiner™ and upon equalization by the EZQ Protein Quantification Kit (Invitrogen Dynal AS, Oslo, Norway).

or Cy5. This approach reduced by approximately half the total number of 2D gels needed when compared to conventional gel electrophoresis. A total of 75 gels (i.e., 2D-DIGE) were run and analyzed. The internal standard was generated by mixing equimolar amounts of each of the equalized sera samples included in this first stage (n = 149). The pool of proteins in this internal standard was minimally labeled with Cy2 using CyDye, according to the manufacturer's protocol (GE Healthcare, Sunnyvale, CA, USA). The proteins in the other two samples (equalized sera) were labeled with either Cy3 or Cy5 at a pH of 8–9. The internal standard was used to normalize for differences in protein loading and gel-to-gel variations, as well as to aid in the alignment of scanned images. Under these conditions, the abundance of each protein in every sample could be normalized relative to the internal standard. Because the spectral properties of Cy2, Cy3, and Cy5 are not identical, we introduced a Compensation Flip Step, so that Cy3- and Cy5-labeled proteins from samples of each group under study (i.e., Cy3-POAG and Cy5-PEXG; Cy3-PEXG and Cy5-POAG; Cy3-POAG and Cy5-Control; Cy3-Control and Cy5-POAG; Cy3-PEXG and Cy5-Control; Cy3-Control and Cy5-PEXG) would be represented in approximately equal numbers. Labeling of the proteins was performed by the addition of 400 pmol of the required CyDye in 1 μl of anhydrous N, N-dimethylformamide per 50 μg of protein. The labeling reaction was allowed to proceed for 30 min on ice in the dark, after which 1 μl of 10 mM lysine (Sigma Aldrich, St. Louis, MO, USA) was added to terminate the reaction over 10 min. Before 2D gel separation, 25 μg of protein from each of the three samples labeled with Cy2, Cy3 and Cy5 was pooled and mixed with rehydration buffer [30 mM Tris–HCl, pH 8.5, 2 M ThioUrea, 7 M Urea and 4% CHAPS, 20 mM dithiothreitol (DTT) and 0.25% pH 4–7 carrier ampholyte (Bio-lyte; Bio-Rad, Hercules, CA, USA)]. The samples were subsequently applied to immobilized pH gradient strips (24 cm, pH 4–7) and focused in a Protean Isoelectrofocusing Cell (BioRad, Malvern, Pennsylvania, USA) following instructions by the supplier. The focused strips were equilibrated for 15 min with buffer 1 (6 M urea, 0.375 M Tris– HCl, 2% SDS, 20% glycerol), supplemented with 2% (w/v) DTT, then subsequently incubated for another 15 min with buffer 1 supplemented with 2.5% (w/v) iodoacetamide. The second-dimension separation was carried out by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) in large 12.5% polyacrylamide gels using the DALT-SIX electrophoresis system (GE Healthcare, Sunnyvale, CA, USA) according to standard protocols. All electrophoresis procedures (labeling, 1stand 2nd-dimensional separations) were performed in the dark.

2.6.

2.7.

Anti-glaucoma drugs

Pathological Status POAG (n = 73)

Prostaglandin analogs a Prostaglandin analogs + beta-blockers a Beta-blockers a Others Untreated

27 38 4 1 3

(36.99%) (52.05%) (5.48%) (1.37%) (4.11%)

PEXG (n = 59) 21 28 6 1 3

(35.59%) (47.46%) (10.17%) (1.70%) (5.08%)

a

In many cases in combination with others active ingredients as carbonic anhydrase inhibitors and/or alpha-2 adrenergic agonists.

2.5. Serum sample preparation and total protein quantification

2D-DIGE for candidate biomarker discovery

In the present study we used 2D-DIGE to identify changes in the concentrations of proteins between equalized serum samples from POAG, PEXG, and control subjects. Three different samples were run together in one gel. Proteins in each sample were labeled with distinct fluorescent dyes: Cy2, Cy3 and Cy5. The Cy2 labeled proteins corresponded to an internal standard (see below), which was included in each 2D-DIGE gel run. The other two samples in each gel corresponded to equalized serum from POAG, PEXG or control subjects, and they were labeled with Cy3

Image acquisition and data analysis

Images of Cy2-, Cy3- and Cy5-labeled proteins were taken after separation in each gel (i.e., 75 gels 2D-DIGE) upon excitation at the wavelengths of 488, 532, and 633 nm while capturing the emission signal at 530, 605 and 695 nm, respectively, using a VersaDoc Imaging System (BioRad, Malvern, Pennsylvania, USA). Gels were scanned for 180 s of light exposure for each of the channels. Image analysis of Cy2-, Cy3-, and Cy5-labeled proteins in each gel was performed with the Progenesis SameSpots

70

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

Software, version 4.0 (NonLinear Dynamics Limited, Newcastle, UK). Variation in abundance of proteins (spots) between groups (i.e., POAG, PEXG, and controls) was determined by the Significance Analysis of Microarrays (SAM) method [22], using normalized “spot” volumes (obtained from Progenesis SameSpots Software) with a cutoff False-Discovery-Rate value of 0.01.

tolerance, and up to two missed cleavage points. Calibration was performed externally with Pepmix (Bruker Daltonics), and internally with trypsin peptides, when possible.

2.8.2.

Protein identification by nLC-MS/MS

In order to identify protein spots by mass spectrometry (MS), we first had to isolate them. Separation was performed with a 12.5% SDS preparative gel, onto which 400 μg of protein from a pooled sample, which included an equimolar amount of all samples analyzed in the present study, was loaded and stained with SYPRO Ruby (BioRad, Malvern, Pennsylvania, USA). Protein spots with significant statistical power were robotically excised with an Ettan Spot Picker (GE Healthcare) from the preparative gel and subjected to in-gel trypsin digestion according to Shevchenko [23], with the following minor modifications. The gel pieces were reduced with 30 μl of 10 mM DTT at 56 °C for 20 min, followed by alkylation in 30 μl of 50 mM iodoacetamide at room temperature for 20 min in darkness. Next, spots were swollen in an ice bath with a digestion buffer containing 50 mM ammonium bicarbonate and 12.5 ng/μl of trypsin (Roche Diagnostics, recombinant, proteomics grade trypsin, Penzberg, Germany). After 30 min, the supernatant was removed and discarded, 20 μl of 50 mM ammonium bicarbonate was added to the gel piece, and the digestion was allowed to proceed at 37 °C overnight. After trypsinization, the supernatant was transferred to an eppendorf and acidified to 0.1% trifluoroacetic acid. Protein identifications were carried out at CIC bioGUNE Proteomics Platform (Bizkaia, Spain), a member of the Spanish ProteoRed-ISCIII Network and CIBERehd.

Tryptic peptides were preconcentrated and desalted using a Symmetry C18 precolumn (180 μm × 20 mm, 5 μm particle size, Waters) followed by elution on a BEH C130 column (75 μm × 200 mm, 1.7 μm particle size, Waters) using two solvents: A, 0.1% formic acid, and B, 0.1% formic acid in acetonitrile, with a linear gradient of solvent B from 3% to 50% in 50 min. All high-performance liquid chromatography runs were performed using a Waters nanoACQUITY UPLC system (Waters) under a constant flow rate of 300 nL/min. The eluting peptides were scanned and fragmented with a LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific) equipped with a Proxeon nano-electrospray source. An electrospray voltage of 1.6 kV and a capillary voltage of 40 V at 280 ºC were used. Survey scans ranging from 400 to 2000 mass-to-charge ratio were performed in the Orbitrap analyzer (30,000 FWHM). MS/MS fragmentation and measurement of the six most intense ions were performed using collision-induced-dissociation in the LTQ linear ion trap. Normalized collision energy was set to 35%. Searches were performed using Proteome Discoverer 1.3 software (Thermo Fisher Scientific) and Mascot search engine. A tolerance of 5 ppm was allowed for the precursor search, whereas fragments were analyzed with a tolerance of 0.5 Da. Cystein carbamidomethylation was considered a fixed modification and methionine oxidation was considered a variable modification. Two missed cleavages were allowed for tryptic digestion. Spectra were searched against all entries of the UniProtKB/Swiss-Prot database, and proteins with at least two identified peptides (p < 0.05) were considered significant hits.

2.8.1.

2.9.

2.8.

Mass spectrometry analysis

Protein identification by MALDI-TOF/TOF

MALDI-MS analysis was performed with a MALDI-LIFT-TOF AUTOFLEX III SmartBeam (Bruker Daltonics). Each digested sample was loaded (1 μl) onto a target (Bruker 384 ground steel) with 1 μl of α-cyano-4-hydroxycinnamic acid. Datadependent MS acquisitions were performed with a charge state of 1 over a survey mass-to-charge ratio range of 500–4000. Ionization was performed with a solid-state laser with wavelength 360 nm and frequency 200 Hz. Laser intensity energies were varied depending on the analysis required. For MS, 30–50% of intensity was used, while around 90% for MS/MS. Resolution was always over 7500 along all mass-window ranges for MS analysis. Data acquisition was performed manually. Routinely, 1400 scans were collected for Peptide Mass Fingerprint (PMF), whereas the most intense peaks were selected for MS/MS (400 scans for parent selection, and 1600 scans for fragments). Obtained spectra were processed using Flex analysis 3.0 and Biotools 3.2 (Bruker Daltonics). A database search was performed using MASCOT 2.2 (Matrixscience, London, UK) against UniProtKB/Swiss-Prot database (version SwissProt 2011_08; 531,473 sequences entries). The following parameters were adopted for protein identification: carbamidomethylation of cysteines as fixed modification, oxidation of methionines as variable modification, 50 ppm of peptide mass tolerance, 0.7 Da fragment mass

Ingenuity Pathway Analysis (IPA)

The software program IPA (Ingenuity® Systems, www. ingenuity.com) was used to evaluate the significant canonical pathways and networks associated to the 35 proteins identified by MS analysis among POAG, PEXG and control cases as the most discriminatory between the three groups. This is done by assigning a significance score to each network equal to the negative logarithm of a calculated p-value. This score represents the likelihood that the assembly of a set of genes or proteins is part of significant canonical pathways or networks.

3.

Results

3.1.

Subjects characteristics

A total of 149 participants were included in Step 1 of the “Differential expression analysis” (see Fig. 1). The participants' mean age [±standard deviation (SD)] in the POAG (n = 53) and control (n = 51) groups were comparable (67.46 ± 11.27 years for POAG patients compared with 64.93 ± 11.34 years for controls, respectively; p > 0.05). The mean age in the PEXG group (n = 45; 73.11 ± 8.43 years) was slightly higher than that of the POAG (p < 0.05) and control (p < 0.001) groups (Table 1a).

71

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

pI 4 MW 150-

7 55

4538-

60

176

455451 450

8070-

30

36

31

110

14

113 100 90

226 274 288 235 276 281 284 287 346 564 571 431 283 192 297 373 557 397 372 528 566 561 291 358 423 536 605 422 533 447 658 651 740 749 351 594 296 360 514 661 433 606 627 739 593 436 699 652 506 745 544 565 438 446 462 467 507 617 734 952 1007967 959 964 666 609 559 645 941 636 1086 1127 851 830 953 746 3544 997 938 958 1055 1145 1045 1109 960 1153 968 1039 966 1184 1282 1128 1219 1214 1211 1379 1468 1329 1483 1212 1375 1308 1316

13191371 1397

1390

1461 1848

143515371382 1840

1611

301906

1911 1917

1920

1907

3560

2708

1406 1359

1372

3556

1908 2167

1846

3559 2760

2768

153550

2812

3552

kDa Fig. 2 – Representative 2D-PAGE gel of human serum proteome. A total of 400 μg of depleted serum protein from a pooled sample from cataract (n = 51), POAG (n = 53) and PEXG (n = 45) patients, which included an equimolar amount of all samples analyzed in the present study, was loaded and stained with SYPRO Ruby. The statistically most significant (q < 0.01; Gain Ratio ≥ 0.05) protein spots are marked on 2D-gel with their corresponding spot number.

A total of 53 newly-recruited subjects were included in Step 2 “ELISA screening” (see Fig. 1). The participants' mean age (±SD) of POAG (n = 20; 70.95 ± 8.54) patients was comparable with mean age of PEXG patients (n = 14; 72.86 ± 4.13 years; p > 0.05) and control group (n = 19; 65.21 ± 14.13 years; p > 0.05). The mean age in the PEXG group was slightly higher than that of the control group (p < 0.05) (Table 1b). The mean values of IOP, cup-disc ratio of optic nerve and, by gender, the number of men and women in each group were quite similar (see Tables 1a and 1b). Finally the use of anti-glaucoma drugs were similar between POAG and PEXG groups (Table 2).

3.2.

2D-DIGE analysis of serum proteome

A total of 75 gels (2D-DIGE) were needed to analyze 149 equalized serum samples (53 POAG, 45 PEXG, 51 Control) as described in Experimental Procedures. Proteins were separated in the first dimension along an isoelectric point (pI) range of 4 to 7, and in the second dimension between molecular masses of 10 kDa and 150 kDa. To avoid labeling bias of proteins in each sample with Cy3 or Cy5, a Dye-Flip design was applied. Following the successful warping of the gels and importing their images into the Progenesis SameSpots program, an equalized serum protein map was obtained for every sample, where up to 10,000 protein spots were detected per gel. Selective filtering of spots was carried out based on spot properties including area, average normalized volume, and

location. After this, a total of 823 protein spots were detected using the VersaDoc Imaging System Software. After normalization of the spot volumes (area x intensity), the data were expressed in matrix notation and filtered by means of a False-Discovery-Rate cutoff value of 0.01 (q-value < 0.01), which resulted in 214 significant spots. Protein spots that displayed significant fold-change alterations in expression between POAG or PEXG and control cases were ranked from highest to lowest discriminating power, based upon their ReliefF, Information gain and Gain Ratio values. A total of 149 protein spots displaying significant changes in abundance were selected for identification purposes by specifying a threshold of Gain Ratio ≥ 0.05, and subsequently classified according to their Spot ID, RefiefF, Information Gain, and Gain Ratio values (Supplemental Table 1). Fig. 2 shows a representative 2D-PAGE gel of a human serum proteome, used for protein identification purpose only. In this gel, the most statistically significant protein spots found among POAG, PEXG, and control cases are labeled with their corresponding spot number (Spot ID).

3.3.

Identification of the most-differentially-altered proteins

The 149 protein spots listed on Supplemental Tables 1 to 3 were excised from a preparative 2D-PAGE gel, as shown in Fig. 2, and they were subjected to in-gel trypsin digestion and MALDI-TOF/TOF (see Supplemental Table 1) or nLC-MS/MS (see Supplemental Table 2) analysis for identification, as

72

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

Table 3 – List of the 35 differentially expressed proteins in patients suffering glaucoma compared to controls, sorted in descending order of significance. Discrimination ranking

UniProt Accession number

1 2 3 4

P08603 P23142 P01024 Q14624

5

Protein name

Gene symbol

No of spots (spot numbers)

CFH FBLN1 C3 ITIH4

3 1 5 5

P01008

Complement factor H Fibulin-1 Complement C3 Inter-alpha-trypsin inhibitor heavy chain H4 Antithrombin-III

SERPINC1

6 7

P02787 P02768

Serotransferrin Serum albumin

TF ALB

8

P04004

Vitronectin

VTN

9 10 11 12 13 14 15 16 17 18 19 20

P01859 P02743 P01009 P06727 O14791 P0C0L4 O75636 P02647 P02766 P27169 Q15485 Q03591

IGHG2 APCS SERPINA1 APOA4 APOL1 C4A FCN3 APOA1 TTR PON1 FCN2 CFHR1

21 22 23 24 25 26 27 28

P02748 P22352 P01860 Q9BXR6 Q6UX53 P02774 P06681 P19823

C9 GPX3 IGHG3 CFHR5 METTL7B GC C2 ITIH2

4 1 1 1 1 1 1 1

29

P00734

Ig gamma-2 chain C region Serum amyloid P-component Alpha-1 antitrypsin Apolipoprotein A-IV Apolipoprotein L1 Complement C4-A Ficolin-3 Apolipoprotein A-I Transthyretin Serum paraoxonase/arylesterase 1 Ficolin-2 Complement factor H-related protein 1 Complement component C9 Glutathione peroxidase 3 Ig gamma-3 chain C region Complement factor H-related protein 5 Methyltransferase-like protein 7B Vitamin D-binding protein Complement C2 Inter-alpha-trypsin inhibitor heavy chain H2 Prothrombin

10 (297, 281, 287, 507, 565, 745, 1907, 739, 423, 422) 4 (274, 373, 360, 346) 10 (462, 467, 446, 438, 436, 447, 2812, 433, 358, 291) 11 (564, 561, 651, 1086, 3550, 1109, 557, 536, 652, 661, 658) 1 (627) 4 (746, 1611, 1908, 1840) 3 (749, 734, 740)) 1 (959) 4 (1045, 964, 941, 960) 2 (1308, 1375) 3 (1359, 1372, 3556) 4 (1911, 1906, 1917, 1920) 1 (2768) 3 (952, 967, 968) 2 (1219, 1184) 2 (958, 953)

F2

30 31 32

P02649 P02741 P10909

Apolipoprotein E C-reactive protein Clusterin

APOE CRP CLU

33 34 35

P01871 P01834 O00187

Ig mu chain C region Ig kappa chain C region Mannan-binding lectin serine protease 2

IGHM IGKC MASP2

7 (966, 1039, 1382, 1406, 276, 288, 296) 5 (1379, 1435, 1468, 1483, 1537) 1 (1848) 11 (451, 1211, 1319, 455, 450, 1212, 1316, 1282, 1371, 1390, 1397) 1 (351) 1 (1846) 1 (3560)

described in Material and methods. We established the identity of 118 spots out of the total of 149, resulting in the identification of 35 distinct proteins (see Supplemental Table 3). Fourteen of these proteins were present in single spots, while the rest were found in two or more spots. These latter spots could be the result of posttranslational modifications including proteolytic cleavages of specific proteins. Ten protein spots remain unidentified, most likely due to their low concentrations. Twenty-one protein spots were identified as human keratins, which are non-seric proteins, and they were not taken into consideration or included for further analysis because they most likely represent contaminants from gel handling. Table 3 shows the list of the 35 distinct proteins differentially expressed between the three sample groups and ranked by statistical significance from highest to lowest.

3.4.

(30, 1127, 31, 36) (55) (100, 1153, 113, 1128, 1145) (192, 593, 549, 617, 506)

(528, 533, 283, 514) (2167) (636) (397) (645) (666) (699) (830)

Functional pathway analysis

Functional pathway analysis (IPA software) covers three domains: networks, canonical pathways and diseases. We first explored the network characteristics of the 35 serum proteins identified and differentially expressed, using the Ingenuity software package as described in Experimental Procedures. IPA analysis predicted a top-ranked network with a score of 52, in which 22 nodes out of 35 (59% of the total nodes) were enriched and had functions associated with antigen presentation, humoral immune response, and inflammatory response (see Fig. 3). Next, we explored the statistically enriched canonical pathways (p-value ≤ 0.05, Fisher's exact test implemented in the IPA, corrected with Bejamini-Hochberg method), and identified

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

73

Fig. 3 – Pathway analysis based on the Ingenuity Pathway Knowledge Base. The highest scoring network (Antigen presentation, humoral immune response, inflammatory response; score 52) had been obtained from the 35 proteins identified. A black solid edge denotes a direct relationship between two nodes (molecules: proteins, genes). A black dotted edge denotes an indirect relationship between two nodes (molecules: protein, genes). The shaded nodes are genes from among the 35 proteins identified. The various shapes of nodes denote the different functions (legend).

pathways, including: i) liver X receptor/retinoid X receptor activation (p = 9.77E-26); ii) acute phase response signaling (p = 1.36E-19); iii) complement system (p = 6.90E-11); and iv) atherosclerosis signaling (p = 1.23E-10). Lastly, we searched for links to diseases, and found that the 35 serum proteins were connected to: i) inflammatory response disorders including complement activation (p = 5.74E-15); ii) neurological conditions including Alzheimer's disease (p = 2.91E-03); iii) ophthalmic maladies including age-related macular degeneration (p = 1.74E-13); and iv) cardiovascular diseases including myocardial infarction and atherosclerosis (p = 9.99E-11 and p = 1.12E-07, respectively).

3.5. ELISA screening: confirmation of alterations in individual protein levels and statistical assays The differentially-altered serum proteins identified by 2D-DIGE proteomics were quantified by ELISA on sera samples (non-equalized) from a new population comprised of n = 53 newly recruited subjects (20 POAG, 14 PEXG and 19 control, see Table 1b) for each of the 17-top-ranked proteins shown in Table 3. The objective of this step was twofold: on one hand to verify that the concentration range of the top-ranked altered proteins, and more discriminatory among POAG, PEXG and healthy

74 Table 4 – Concentrations of 17 differentially expressed (in milligrams) protein with respect to serum volume (in deciliters) obtained by ELISA analysis in n = 53 newly recruited patients. p-values were obtained by the Kruskal–Wallis test. Gene name

C3 TTR TF VTN FBLN1 APOA1 SERPINA1 CFH APOL1 FCN3 IGHG2 ITIH4

ALB SERPINC1 C4A APCS a

Apolipoprotein A-IV Complement C3 Transthyretin Serotransferrin Vitronectin Fibulin-1 Apolipoprotein A1 Alpha-1 antitrypsin Complement factor H Apolipoprotein L1 Ficolin-3 Ig gamma-2 chain C region Inter-alpha-trypsin inhibitor heavy chain H4 Serum albumin Antithrombin-III Complement C4-A Serum amyloid P-component

Dunn's test for Multiple Comparisons.

POAG group mg/dL

9.14 ± 3.01 422.27 152.25 445.86 103.89 26.61 234.04 289.65 121.45

± ± ± ± ± ± ± ±

62.40 47.75 181.71 40.52 7.89 48.58 65.72 27.81

PEXG group mg/dL

5.18 ± 1.55 311.28 98.32 289.62 74.93 21.26 199.99 226.98 102.39

± ± ± ± ± ± ± ±

36.95 40.59 56.89 23.50 6.76 28.36 59.32 27.62

Control group mg/dL

3.44 ± 0.85 276.40 82.81 263.44 47.22 14.10 183.56 192.30 93.60

± ± ± ± ± ± ± ±

36.07 25.60 57.89 12.06 3.88 37.04 58.78 33.54

FOLD CHANGE

Kruskal–Wallis Test (Non-parametric ANOVA) a

POAG vs Control

PEXG vs Control

POAG vs PEXG

POAG vs Control

PEXG vs Control

POAG vs PEXG

2.7

1.5

1.8

p < 0.001

p > 0.05

p < 0.01

1.5 1.8 1.7 2.2 1.9 1.3 1.5 1.3

1.1 1.2 1.1 1.6 1.5 1.1 1.2 1.1

1.4 1.5 1.5 1.4 1.3 1.2 1.3 1.2

p p p p p p p p

p p p p p p p p

p p p p p p p p

< < < < < < <
> > < < > > >

0.05 0.05 0.05 0.01 0.05 0.05 0.05 0.05

< < < > > > < >

0.001 0.01 0. 01 0.05 0.05 0.05 0.05 0.05

8.47 ± 3.05 114.74 ± 26.64 826.84 ± 265.30

7.87 ± 3.14 88.79 ± 19.15 800.59 ± 223.03

5.90 ± 2.28 85.30 ± 21.66 1144.26 ± 402.83

1.4 1.3 0.7

1.3 1.0 0.7

1.1 1.3 1.0

p < 0.05 p < 0.01 p < 0.05

p > 0.05 p > 0.05 p < 0.05

p > 0.05 p < 0.05 p > 0.05

543.34 ± 275.33

362.38 ± 189.81

372.13 ± 222.88

1.5

1.0

1.5

p > 0.05

p > 0.05

p > 0.05

1.1 0.9 0.8 0.8

1.2 1.0 0.8 1.1

0.9 1.0 1.0 0.8

p p p p

p p p p

p p p p

3695.73 76.41 0.15 15.25

± ± ± ±

1066.51 10.80 0.05 9.00

4092.33 80.39 0.15 19.43

± ± ± ±

954.84 13.95 0.05 13.71

3364.20 81.36 0.19 18.42

± ± ± ±

792.07 13.06 0.03 14.19

> > < >

0.05 0.05 0.05 0.05

> > > >

0.05 0.05 0.05 0.05

> > > >

0.05 0.05 0.05 0.05

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

APOA4

Protein name

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

controls, was reproducible in another distinct population group, and on the other hand to verify that those differences were present in non-equalized serum samples with ProteoMiner™. The ELISA kits for the following proteins were obtained from USCN Life Science Inc. (Wuhan, China): CFH, C3, FBLN1, ALB, ITIH4, TF, APOA1, VTN, APOA4, IGHG2, APCS, APOL1, FCN3, SERPINA1, and TTR. An ELISA kit for C4A was purchased from BD Biosciences (San Jose, CA, USA), and an ELISA kit for SERPINC1 was obtained from AssayPro LLC (St. Charles, MO, USA). The concentrations (mg protein/dL serum, expressed as mean value ± SD) for each of the proteins estimated by ELISA in POAG, PEXG and control groups are shown in Table 4, as are the fold-change and the Kruskal–Wallis Test (non-parametric ANOVA) for each of the proteins between POAG vs control, PEXG vs control, and POAG vs PEXG. Based on the Kruskal–Wallis test, the five proteins ITIH4, ALB, SERPINC1, C4A and APCS, displayed no significant differences among the three sample groups (i.e., POAG, PEXG and control). The rest of the proteins investigated were found in higher concentrations in glaucoma (POAG and/or PEXG) cases compared to control, except in the case of IGHG2 protein, which exhibited a lower concentration in the two glaucoma groups compared to controls. Two proteins: VTN and FBLN1, were found in higher concentrations (1.5- to 2.2-fold) in both POAG and PEXG relative to control cases (i.e., POAG vs control, p < 0.001, PEXG vs control: VTN, p < 0.01; and FBLN1, p < 0.05). None of these proteins showed significant differences between POAG and PEXG groups (p > 0.05). The proteins APOA4, C3, TTR, TF, SERPINA1 and FCN3, were found in higher concentrations (1.3- to 2.7-fold) among POAG relative to control cases (i.e., APOA4, C3, TTR, TF and SERPINA1, p < 0.001; FCN3, p < 0.01). These concentrations were also significantly elevated (1.3- to 1.8-fold) among POAG when compared to PEXG cases (i.e., C3, p < 0.001; APOA4, TTR and TF, p < 0.01; SERPINA1 and FCN3, p < 0.05), but no differences were found between PEXG and control (p > 0.05). The proteins CFH and APOL1 were found in higher concentrations in POAG cases compared to control (i.e., CFH, p < 0.01; APOL1, p < 0.05), but there was no difference between PEXG and control groups (p > 0.05) and between POAG and PEXG cases (p > 0.05). The protein IGHG2 was found in lower concentrations in POAG and PEXG (0.7 fold for both) cases when compared with controls (p < 0.05). No significant differences was found between POAG and PEXG groups (p > 0.05). These results confirmed that the 17-top-ranked proteins were overexpressed among glaucoma patients when compared to healthy controls, and they suggest that the “equalization” of serum samples or differential expression design did not introduce bias in our design analysis. A second statistical assay based on ROC curve analysis by means of Naïve Bayes algorithm was performed with the ELISA data for each of the 17-top-ranked proteins. Table 5 shows the values of the area under the curve (AUC), the sensitivity (Sens.), the specificity (Spec.), and the correct assignment (CA) for each protein under the different compared groups. When POAG vs PEXG vs control groups were compared, the AUC values were greater than 0.8 for the proteins APOA4, C3, TF and FBLN1, between 0.7 and 0.8 for the proteins VTN, TTR, SERPINA1 and FCN3, and between 0.5 and 0.6 for the proteins

75

APOA1, CFH, APOL1, ALB and C4a. The CA is a useful parameter to evaluate the potential of each of the proteins as a candidate biomarker. It represents the percentage of samples correctly classified by using the biomarker concentrations with respect to clinical diagnosis (clinical accuracy assessment). Based on the CA values, APOA4 was able to correctly classify 81% of the cases, meaning that it was the most significant candidate biomarker in the discrimination of the three (POAG vs PEXG vs Control). On the other hand, C3, TF, VTN, SERPINA1, FBLN1 and CFH properly classified at least 60% of patients among the three groups. When comparing POAG and control groups, the proteins APOA4, C3, TF, VTN, TTR, SERPINA1, FBLN1, APOA1, FCN3 and CFH classified properly at least 70% of patients. These same proteins, with the exception of SERPINA1, classified suitably at least 70% of patients in PEXG vs control groups. Finally, when POAG and PEXG groups were compared, correct classification of the cases was only greater than 60% for the proteins APOA4, C3, TF, SERPINA1, FBLN1 and FCN3.

4.

Discussion

In this study we present a comparative proteomic analysis by 2D-DIGE and molecular MS of “equalized” sera from patients with two related but distinct glaucoma pathologies, as well as non-glaucoma controls. The patients belonged to the two most prevalent forms of glaucoma in industrialized countries: POAG and PEXG. These forms of glaucoma are clinically and genetically distinct. POAG is primarily the result of a complex pattern of heritage, whereas PEXG is a secondary form of glaucoma in many patients with exfoliation syndrome. The equalization of serum proteins through chromatographic separation with ProteoMiner ™ has emerged in recent years as a valuable method to “equalize” the protein content in serum through the simultaneous dilution of high-abundance proteins and concentration of those most dilute [24–26]. By implementing the step of “equalization” before 2D-DIGE, we increased the resolution of less abundant proteins and the likelihood of detecting them among glaucoma and control samples. It is important to emphasize that the present comparative analysis by 2D-DIGE of “equalized” serum samples from POAG, PEXG and control subjects, was possible because of the labeling of proteins from each group separately with distinct fluorescent dyes (i.e., Cy2, Cy3 and Cy5). This facilitated the alignment of images of different gels so that quantitative differences among the three groups (i.e., POAG, PEXG, and control) could be determined. In this analysis, the inclusion of a labeled internal standard (i.e., Cy2) in each gel allowed the relative densitometric quantification of each labeled protein spot (i.e., Cy3, Cy5). This methodology permitted analysis in approximately half the number of gels when compared to conventional 2D-SDS-PAGE. We optimized the resolution and separation of serum proteins by 2D-DIGE by comparing the separation of proteins in one dimension through a wide pI range (3–10), followed by a 2nd-dimensional separation with 12.5% SDS-PAGE. Staining of this gel with Sypro Ruby revealed that the maximum resolution and separation of the spot proteins were achieved in the narrow pI range of 4–7, and the lowest resolution on a pI range between 6 and 10 and a relative molecular mass between 10 kDa and 50 kDa.

76

Table 5 – Results of individual biomarker power discrimination obtained with Naïve Bayes algorithm in the classification of POAG, PEXG and control serum samples with newly recruited patients (n = 53). Gene name

POAG vs PEXG vs control CA

a b c d

0.8113 0.6415 0.6792 0.6038 0.5849 0.6038 0.6981 0.4717 0.5849 0.6226 0.3396 0.4906 0.2453 0.3019 0.2453 0.3585 0.3208

CA, correct assignment. Sens., sensitivity. Spec., specificity. AUC, area under the curve.

Sens.

b

1.0000 0.8421 0.8947 1.0000 0.8421 0.8421 0.8421 0.7895 0.6842 0.7895 0.4737 0.6842 0.3158 0.3158 0.2632 0.5789 0.2632

Spec.

c

0.9118 0.8824 0.8235 0.8529 0.7941 0.7059 0.9118 0.7647 0.7059 0.8824 0.4412 0.7059 0.5882 0.6765 0.5882 0.6471 0.5294

POAG vs control AUC

d

0.9233 0.8607 0.8283 0.7786 0.7635 0.7203 0.8380 0.6641 0.7559 0.6933 0.4590 0.5972 0.5778 0.4816 0.2894 0.5346 0.3164

PEXG vs control

POAG vs PEXG

CA

Sens.

Spec.

AUC

CA

Sens.

Spec.

AUC

CA

Sens.

Spec.

AUC

0.9744 0.8718 0.8718 0.9231 0.8718 0.8462 0.8293 0.7692 0.7692 0.7949 0.4872 0.6923 0.3846 0.4359 0.3333 0.4103 0.4359

1.0000 0.8421 0.9474 0.9474 0.8421 0.8947 0.8421 0.7895 0.8421 0.6842 0.4737 0.6842 0.3158 0.3158 0.2632 0.4737 0.3158

0.9500 0.9000 0.8000 0.9000 0.9000 0.8000 0.9091 0.7500 0.7000 0.9000 0.5000 0.7000 0.4500 0.5500 0.4000 0.3500 0.5500

1.0000 0.9763 0.9526 0.9474 0.9342 0.8684 0.8886 0.8158 0.8553 0.8079 0.6158 0.7553 0.5263 0.3842 0.4000 0.3947 0.3368

0.8788 0.7879 0.8485 0.8182 0.7576 0.6667 0.9091 0.7576 0.6970 0.7576 0.5758 0.5455 0.7576 0.6667 0.2727 0.7273 0.5152

0.9474 0.7895 0.9474 0.8421 0.8947 0.7368 0.8947 0.7895 0.7368 0.7895 0.7368 0.6842 0.8421 0.7368 0.4737 0.7895 0.7368

0.7857 0.7857 0.7143 0.7857 0.5714 0.5714 0.9286 0.7143 0.6429 0.7143 0.3571 0.3571 0.6429 0.5714 – 0.6429 0.2143

0.8609 0.8910 0.8120 0.9060 0.6617 0.7556 0.8797 0.7857 0.7293 0.7368 0.3647 0.6955 0.7895 0.6466 0.2068 0.7444 0.3722

0.7941 0.6765 0.6765 0.4706 0.5588 0.6176 0.6471 0.4412 0.6176 0.5588 0.5294 0.3824 0.4706 0.4706 0.3235 0.4706 0.6471

0.8000 0.6500 0.6500 0.6000 0.7000 0.8000 0.8000 0.6500 0.6000 0.7000 0.8000 0.6500 0.6000 0.6500 0.5500 0.6000 0.9000

0.7857 0.7143 0.7143 0.2857 0.3571 0.3571 0.4286 0.1429 0.6429 0.3571 0.1429 – 0.2857 0.2143 – 0.2857 0.2857

0.8143 0.6786 0.6571 0.4464 0.6750 0.4821 0.5821 0.3643 0.6786 0.4893 0.3679 0.2607 0.5357 0.4643 0.2214 0.5500 0.2786

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

APOA4 C3 TF VTN TTR SERPINA1 FBLN1 APOA1 FCN3 CFH ITIH4 APOL1 ALB SERPINC1 IGHG2 C4A APCS

a

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 6 5–7 8

Although it is possible that additional proteins were not resolved under the selected conditions, we think that the identification of a panel of 35 proteins may represent the most differentially expressed proteins between POAG, PEXG and control serum samples. The variation in abundance (i.e., overexpression) of these proteins between groups detected by imaging analysis was confirmed by a standard and independent quantification method such as ELISA on the top-17-ranked proteins of the 35 protein panel. As described in Materials and methods, we applied the ELISA kit assays to non-equalized, instead of “equalized”, serum samples from an independent group of glaucoma and healthy controls. These assays confirmed that the top-17-ranked proteins, out of the 35 protein panel identified, were overexpressed in both glaucoma groups when compared to healthy controls (see Table 4). Although the authors are aware that the verification of these candidate markers by ELISA was performed on a relatively small number of new participants (n = 53), the preliminary results are encouraging and warrant further validation in a large-scale investigation by an independent method, for example, multiplexed technology such as LC-MS/MS in the Multiple Reaction Monitoring mode. The significant discriminative power of each of the proteins for predicting the glaucoma pathologies separately from control cases (Table 5) supports the case for glaucoma biomarker candidates among the top-17-ranked proteins. APOA4 yielded the best performance, correctly classifying all the POAG cases from healthy cases, exhibiting 86% efficacy distinguishing between PEXG and healthy controls, and 81% efficacy discriminating POAG from PEXG cases. The rest of the proteins, including C3, TF, VTN, SERPINA1, FBLN1 and CFH, also classified POAG, PEXG and control groups, but with lower discriminatory power than APOA4. Overall, these results suggest that among the 17-top-ranked proteins, there are candidate biomarkers capable of classifying the glaucoma cases from healthy controls. To learn more about the nature of the panel of 35 altered proteins identified among glaucoma and healthy case groups, we conducted bioinformatic analysis (i.e., IPA), based on the networks and canonical pathways, diseases, and functional annotations that could be linked to the identified serum proteins. The large number of nodes identified (59.45%) among the 35-protein panel linked to immunological and inflammatory networks, suggested that many of the differentially altered proteins are involved in inflammatory pathways. This observation is in agreement with views on the role of inflammation as a risk factor in secondary glaucoma in humans [27] and in a mouse model of glaucoma, suggesting that early events in the pathology of glaucoma are associated to the activation of the complement system [28]. Proteomic analysis conducted in tears of glaucoma patients by Pieragostino et al. also suggested immune-inflammatory risk factors associated with glaucoma [26]. In this latest study, four of the proteins identified, TF, APOA1, IGHM, and IGHG2, were also among the 35-protein panel identified in serum of POAG/PEXG cases reported in our study. Other differential proteomics studies conducted by Duan et al., and Grus et al. [29,30], with aqueous humor from POAG patients, identified a protein (i.e., TTR) that figured in our analysis of serum from POAG and PEXG patients. Interestingly, in this case the relative fold-change difference found for TTR in aqueous humor of POAG patients relative to

77

control was similar to the one found in the present study in serum. Other proteins including APOA4, ALB, SERPINC1, and TTR, which were represented in the 35-top-ranked serum proteins among POAG and PEXG patients in our study, have also been identified as potential biomarkers in a study by Bouhenni et al., which used the aqueous humor from primary congenital glaucoma patients [31]. However, this later study did not yield information on their relative protein concentrations. Thus, although several of the potential biomarkers proposed in earlier studies in tears and aqueous humor of glaucoma patients were also identified in the present study, we found these biomarkers in significantly higher levels in the sera of POAG and PEXG patients than in control cases. These findings open the possibility of screening candidate markers of glaucoma directly from the sera of patients. Future work will allow determination of the behavior of these candidate biomarkers during the progression of the disease and their utility in the diagnosis of sub-forms of glaucoma.

5.

Conclusions

In summary, we applied a new proteomic workflow to identify potential markers in the serum of glaucoma patients that could help in the clinical diagnosis and classification of two of the most frequent subtypes of the disease: POAG and PEXG. This workflow identified a panel of 35 serum proteins that were found in altered concentrations in patients with glaucoma relative to healthy controls. The signaling network of these proteins correlated to an immunological and inflammatory pathway. The top-17-ranked proteins of this panel were determined by ELISA to be present in glaucoma sera in higher concentrations than in serum from the healthy control group. Further work based on a large-scale multiplexed screening with the candidate markers identified here is warranted. The data from this report offer new perspectives in the discovery of glaucoma biomarkers in the serum of POAG and PEXG patients. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral. proteomexchange.org) via the PRIDE partner repository [32] with the dataset identifier PXD000198. This material is available free of charge via the Internet at http://pubs.acs.org. Supplementary data related to this article can be found online at http://dx.doi.org/ 10.1016/j.jprot.2013.12.006.

Acknowledgments The authors wish to thank Javier Soria, Félix Elortza and Ibon Iloro for their excellent technical help, discussion, and suggestions. Carson Petrash for proofreading the manuscript and useful discussions. We also wanted to thank Esther Vázquez and Cristina Díez for their help in blood sampling collection. This research would not have been possible without the cooperation of the patients and the staff of the Institute of Ophthalmology Fernández-Vega. Special thanks to Dr. Amhaz Hussein, Lucia Fernández and Javier Lozano for

78

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 6 5 –78

their inestimable help. Finally we wish to thank the PRIDE Team for their invaluable help in raw data submission.

[17]

REFERENCES [18] [1] Quigley HA, Broman AT. The number of people with glaucoma worldwide in 2010 and 2020. Br J Ophthalmol 2006;90(3):262–7. [2] Viso E, Rodríguez-Ares MT, Gude F. Prevalence of pseudoexfoliation syndrome among adult Spanish in the Salnés eye Study. Ophthalmic Epidemiol 2010;17(2):118–24. [3] Gabelt BT, Kaufman PL. Changes in aqueous humor dynamics with age and glaucoma. Prog Retin Eye Res 2005;24(5):612–37. [4] Schlötzer-Schrehardt U, Naumann GOH. Trabecular meshwork in pseudoexfoliation syndrome with and without open-angle glaucoma. Invest Ophthalmol Vis Sci 1995;36:1750–64. [5] Konstas AGP, Mantziris DA, Stewart WC. Diurnal intraocular pressure in untreated exfo-liation and primary open-angle glaucoma. Arch Ophthalmol 1997;115(2):182–5. [6] Angelilli A, Ritch R. Directed therapy for exfoliation syndrome. Open Ophthalmol J 2009;3:70–4. [7] Gottanka J, Kuhlmann A, Scholz M, Johnson DH, Lütjen-Drecoll E. Pathophysiologic changes in the optic nerves of eyes with primary open angle and pseudoexfoliation glaucoma. Invest Ophthalmol Vis Sci 2005;46(11):4170–81. [8] Vesti E, Kivelä T. Exfoliation syndrome and exfoliation glaucoma. Prog Retin Eye Res 2000;19(3):345–68. [9] Kerrigan-Baumrind LA, Quigley HA, Pease ME, Kerrigan DF, Mitchell RS. Number of ganglion cells in glaucoma eyes compared with threshold visual field tests in the same persons. Invest Ophthalmol Vis Sci 2000;41(3):741–8. [10] Stone EM, Fingert JH, Alward WL, Nguyen TD, Polansky JR, Sunden SL, et al. Identification of a gene that causes primary open angle glaucoma. Science 1997;275(5300):68–70. [11] Fingert JH, Héon E, Liebmann JM, Yamamoto T, Craig JE, Rait J, et al. Analysis of myocilin mutations in 1703 glaucoma patients from five different populations. Hum Mol Genet 1999;8(5):899–905. [12] Lopez-Martinez F, Lopez-Garrido MP, Sanchez-Sanchez F, Campos-Mollo E, Coca-Prados M, Escribano J. Role of MYOC and OPTN sequence variations in Spanish patients with primary open-angle glaucoma. Mol Vis 2007;13:862–72. [13] Melki R, Colomb E, Lefort N, Brézin AP, Garchon HJ. CYP1B1 mutations in French patients with early-onset primary open-angle glaucoma. J Med Genet 2004;41(9):47–51. [14] López-Garrido MP, Blanco-Marchite C, Sánchez-Sánchez F, López-Sánchez E, Chaqués-Alepuz V, Campos-Mollo E, et al. Functional analysis of CYP1B1 mutations and association of heterozygous hypomorphic alleles with primary open-angle glaucoma. Clin Genet 2010;77(1):70–8. [15] Thorleifsson G, Magnusson KP, Sulem P, Walters GB, Gudbjartsson DF, Stefansson H, et al. Common sequence variants in the LOXL1 gene confer susceptibility to exfoliation glaucoma. Science 2007;317(5843):1397–400. [16] Pieragostino D, Bucci S, Agnifili L, Fasanella V, D'Aguanno S, Mastropasqua A, et al. Differential protein expression in tears

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27] [28]

[29]

[30]

[31]

[32]

of patients with primary open angle and pseudoexfoliative glaucoma. Mol Biosyst 2012;8(4):1017–28. Curk T, Demsar J, Xu Q, Leban G, Petrovic U, Bratko I, et al. Microarray data mining with visual programming. Bioinformatics 2005;21(3):396–8. Guerrier L, Thulasiraman V, Castagna A, Fortis F, Lin S, Lomas L, et al. Reducing protein concentration range of biological samples using solid-phase ligand libraries. J Chromatogr B Analyt Technol Biomed Life Sci 2006;833(1):33–40. Righetti PG, Boschetti E. The proteominer and the fortyniners: searching for gold nuggets in the proteomic arena. Mass Spectrom Rev 2008;27:596–608. Boschetti E, Righetti PG. The art of observing rare protein species in proteomes with peptide ligand libraries. Proteomics 2009;9:1492–510. Bradford MM. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 1976;72:248–54. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001;98(9):5116–21. Shevchenko A, Tomas H, Havlis J, Olsen JV, Mann M. In-gel digestion for mass spectrometric characterization of proteins and proteomes. Nat Protoc 2007;1(6):2856–60. Boschetti E, Righetti PG. The ProteoMiner in the proteomic arena: a non-depleting tool for discovering low-abundance species. J Proteomics 2008;71(3):255–64. Meng R, Gormley M, Bhat VB, Rosenberg A, Quong AA. Low abundance protein enrichment for discovery of candidate plasma protein biomarkers for early detection of breast cancer. J Proteomics 2011;75(2):366–74. Selvaraju S, El Rassi Z. Reduction of protein concentration range difference followed by multicolumn fractionation prior to 2-DE and LC-MS/MS profiling of serum proteins. Electrophoresis 2011;32(6–7):674–85. Hall AJ. Secondary glaucoma. Clin Exp Optom 2000;83(3):190–4. Howell GR, Macalinao DG, Sousa GL, Walden M, Soto I, Kneeland SC, et al. Molecular clustering identifies complement and endothelin induction as early events in a mouse model of glaucoma. J Clin Invest 2011;121(4):1429–44. Duan X, Xue P, Wang N, Dong Z, Lu Q, Yang F. Proteomic analysis of aqueous humor from patients with primary open angle glaucoma. Mol Vis 2010;16:2839–46. Grus FH, Joachim SC, Sandmann S, Thiel U, Bruns K, Lackner KJ, et al. Transthyretin and complex protein pattern in aqueous humor of patients with primary open-angle glaucoma. Mol Vis 2008;14(1):1437–45. Bouhenni RA, Al Shahwan S, Morales J, Wakim BT, Chomyk AM, Alkuraya FS, et al. Identification of differentially expressed proteins in the aqueous humor of primary congenital glaucoma. Exp Eye Res 2011;92(1):67–75. Vizcaino JA, Côté RG, Csordas A, Dianes JA, Fabregat A, Foster JM, et al. The Proteomics Identifications (PRIDE) database and associated tools: status in 2013. Nucleic Acids Res 2013;41(D1):D1063–9.

Comparative proteomic study in serum of patients with primary open-angle glaucoma and pseudoexfoliation glaucoma.

Alterations in the sera proteins between patients with Primary Open-Angle Glaucoma (POAG), Pseudoexfoliation Glaucoma (PEXG), and healthy controls wer...
803KB Sizes 0 Downloads 0 Views