Human Retinal Pigment Epithelium Proteome Changes in Early Diabetes

Abstract

Aims/hypothesis

Diabetic retinopathy (DR) is the most common complication of diabetes and a leading cause of blindness among working-age adults. Anatomical and functional changes occur in the retina and retinal pigment epithelium (RPE) prior to clinical symptoms of diabetic retinopathy. However, the molecular mechanisms responsible for these early changes, particularly in the RPE, remain unclear. We conducted a comparative proteomics study of human donor RPE to begin defining the molecular changes associated with pre-retinopathic diabetes.

Methods

The RPE was dissected from diabetic human donor eyes with no clinically apparent DR (n=6) and age-matched controls (n=17). Soluble proteins were separated based upon their mass and charge using two-dimensional gel electrophoresis. Protein spots were visualized with a fluorescent dye and spot densities were compared between diabetic and control gels. Proteins from spots with significant disease-related changes in density were identified using mass spectrometry.

Results

Analysis of 325 spots on 2-D gels identified 31 spots that were either up- or down-regulated relative to age-matched controls. The protein identity of eighteen spots was determined by mass spectrometry. A majority of altered proteins belonged to two major functional groups, metabolism and chaperones, while other affected categories included protein degradation, synthesis and transport, oxidoreductases, cytoskeletal structure, and retinoid metabolism.

Conclusions/interpretation

We identified changes in the RPE proteome of pre-retinopathic diabetic donor eyes compared to age-matched controls that suggest specific cellular alterations that may contribute to DR. Defining the pre-retinopathic changes affecting the RPE could provide important insight into the molecular events that lead to DR.

Introduction

Diabetes mellitus is a multifactorial metabolic disorder that currently affects over 200 million people worldwide [1]. Models estimate that this number will nearly double by the year 2025 [1]. Diabetic retinopathy (DR) is the most common complication in diabetes and the leading cause of blindness among working-age adults [2]. The clinical manifestations of DR, including retinal microaneurysms, neovascularization, hemorrhages and macular edema eventually lead to visual impairment [3]. Although vascular abnormalities in the retina clearly contribute to vision loss, other anatomical and functional changes are apparent soon after the onset of diabetes [4-6]. Anatomical changes include reduced retinal thickness [4, 6] and morphological changes of several retinal cell types, including the retinal pigment epithelium (RPE)[5, 7]. Functional changes in retinal electrophysiology and contrast sensitivity have been observed in diabetic patients and animal models [6, 8, 9]. These anatomical and functional changes are apparent in the pre-retinopathic stage that precedes clinically evident vascular changes associated with DR. Importantly, the onset of retinopathy occurs after a prolonged interval following disease onset. Klein and colleagues reported that retinopathy occurs in 73% of Type I diabetic patients approximately 14 years after the initial diagnosis [2]. Thus, the pre-retinopathic stage provides a window of opportunity for intervention that would delay or prevent the onset of blindness. To fully maximize this potential, a thorough understanding of the biochemical changes occurring in retinal tissue is required.

The retina is composed of two major components, the neural retina and the RPE. Previous investigations have focused on the neural retina where vascular changes occur. The neural retina is composed of seven different cell types, including the photoreceptors that are responsible for visual signal transduction. The RPE is essential for vision because of its role in maintaining the photoreceptors [10]. Key functions of the RPE include regenerating the rhodopsin chromophore, regulating nutrient transport to the photoreceptors, phagocytosis of spent tips of photoreceptor outer segments [10], and the production of cytokines, including neurotrophic and angiogenic factors [11]. Additionally, the RPE provides 60-80% of retinal glucose via its high capacity transport system, making it a key source for the high glucose needs of the retina [12].

Recently, several studies have examined changes in the proteome of the neural retina in diabetic animal models to characterize early changes leading to diabetic retinopathy [13, 14]. The present study has focused on the RPE because of its role in supporting the retina before and after the onset of diabetes. We applied a proteomic approach using two-dimensional (2-D) gel electrophoresis and matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry to identify proteins exhibiting altered expression in human donor RPE at the pre-retinopathic stage of diabetes.

Subjects, Materials and Methods

Human Tissue Procurement

Donor eyes obtained from the Minnesota Lions Eye Bank were acquired with the consent of the donor or the donor’s family for use in medical research in accordance with the principles outlined in the Declaration of Helsinki and as approved by the University of Minnesota Institutional Review Board. Information provided by the Lions Eye Bank included gender, age, time and cause of death, and a family report of a limited medical history including ocular history and the occurrence of diabetes.

Eyes were maintained in a moist chamber at 4°C until processing for evaluation of ocular pathology as previously outlined [15-17]. Anterior segments were removed for direct visualization and stereoscopic fundus photography of the posterior segment. Methodology for imaging the posterior pole using the Minnesota Grading System (MGS) was employed and has been previously described for studies of age-related macular degeneration (AMD) [18]. These eyes are carefully examined for diabetic retinopathy, an exclusion criteria for AMD studies. The fundus, including the macula and optic nerve, was carefully examined for signs of proliferative diabetic retinopathy, microaneurysms, characteristic dot blot hemorrhages or signs of hard exudation. Some post-mortem hemorrhages are common; therefore, we cannot completely distinguish small intraretinal hemorrhages of early diabetic retinopathy from post-mortem changes. After reviewing eyebank eyes for over seven years, we feel confident in distinguishing such hemorrhages. Post mortem hemorrhages are typically large, asymmetric, intra-, sub-, or pre-retinal and regional (e.g., in one quadrant). Diabetic hemorrhages are usually small, diffuse, symmetric, and specifically intraretinal. Digitized images were taken before and after removal of the neural retina for a simultaneous examination of the RPE cell layer by two ophthalmologists (TWO and XF). Eyes were eliminated from the study if there was clinical evidence of retinal pathology. Eyes were excluded if post-mortem retinal vascular changes suggested that they could be related to diabetes or eyes that had evidence of AMD. Donors with a medical history of diabetes but no signs of diabetic retinopathy were included in the pre-retinopathic group. Three diabetic donors reported type I and two with type II diabetes. The type of diabetes for one donor was not known. The age-matched control group includes donors with no history of diabetes and no evidence of retinal pathology.

Protein Isolation

Total RPE was dissected after collecting fundus images, pelleted by centrifugation at 1100g for 30 minutes and frozen at −80°C until processing [15]. RPE cells were fractioned by two freeze-thaw cycles and homogenized by 6 passes through a 26 gauge needle in a buffer containing 20 mMol/l HEPES, 10 mMol/l KCl, 1.5 mMol/l MgCl₂, 250 mMol/l sucrose, 1mMol/l EDTA, 1 mMol/l EGTA, 1 mMol/l phenylmethylsulfonyl fluoride, and 0.5% NP40. Nuclei and intact cells were pelleted by centrifugation at 600g for 15 minutes at 4°C and the supernatant reserved. After repeating homogenization and centrifugation of the intact cell pellet, the first and second supernatants were combined. The supernatant from a final centrifugation of 13000g for 15 minutes was stored at −80°C until use. Protein concentrations were determined using the bicinchoninic acid protein assay (Pierce Biotechnology, Rockford, IL), with bovine serum albumin as the protein standard. Average protein yield for the control and diabetic eyes was 486.2 ± 34.2 μg and 558.1 ± 64.4 μg (mean ± SE), respectively. The yields were not significantly different between the two groups (p = 0.314).

2-D SDS-PAGE

The proteins extracted from the RPE of each donor were subjected to 2-D SDS PAGE (125μg). The first dimension separation was performed with pH 5 to 8 immobilized linear gradient strips (Bio-Rad, Hercules, CA, USA). Protein samples (125μg) were dissolved in a rehydration solution (9Mol/l urea, 3 Mol/l thiourea, 6% CHAPS, 1% ASB-14, 1% Biolytes pH 3–10 (Bio-Rad), and 50mMol/l DTT and loaded onto 11cm IPG strips. The conditions for strip rehydration, focusing, equilibration and second-dimension separation were as outlined [17].

The 2-D gels for expression analysis were stained with Flamingo fluorescent stain (Bio-Rad) following the manufacturer’s protocol. Gels for mass spectrometry were stained with silver using a mass spectrometry-compatible kit (Silver Stain Plus Kit; Bio-Rad) and imaged with a GS-800 calibrated densitometer (Bio-Rad).

2-D Gel Quantification and Analysis

Flamingo-stained gels were imaged at two different exposure times using a ChemiDoc XRS (Bio-Rad) and a Dark Reader Transilluminator (Clare Chemical Research, Dolores, CO, USA) to optimize the number of spots analyzed. The Dark Reader was used because the light source wavelength (400-500nm) closely matched the excitation maximum (512 nm) of the dye. Exposure times were based on the fluorescence intensity of two standard proteins run on each gel. Protein spot identification, alignment and quantification of intensity were performed using the 2-D gel analysis software PDQuest 7.1.1 (Bio-Rad). One gel with well-resolved protein spots was chosen as the master gel. Background, streaks and other staining artifacts were subtracted. Automatic spot detection and matching were followed by manual inspection and editing. Spot intensities were normalized to the total intensity of resolved spots.

In-Gel Digestion and MALDI-TOF Analysis

Spots were manually excised from 2-D silver stained polyacrylamide gels and in-gel digestion was performed with trypsin as described [17]. Peptides were analyzed by MALDI-TOF using a QSTAR XL quadrupole-TOF mass spectrometer (Applied Biosystems Inc. [ABI], Foster City, CA, USA) located at the University of Minnesota Mass Spectrometry Consortium for the Life Sciences. All mass spectra were externally calibrated with human angiotensin II tryptic peptides (monoisotopic [MH+] 1046.5417) and adrenocorticotropin hormone fragment (monoisotopic [MH+] 2465.1989). Mass values corresponding to known contaminants (e.g., keratin) and to published trypsin autolysis fragments and matrix clusters were removed [19]. Monoisotopic peaks were automatically identified (Bioanalyst; ABI) and verified by manual inspection. Peak lists were submitted to the Mascot search engine (www.matrixscience.com; Matrix Science, Inc., Boston, MA, USA) and searched in the human Swiss-Prot database. Enzyme specificity was set to trypsin with no missed cleavages, a mass tolerance window of 100 ppm, carbamidomethyl as a fixed cysteine modification, and oxidation of methionine as a variable modification. Initial identifications were accepted for spectra with a significant molecular weight search (MOWSE) score. Initial identities were confirmed by peptide sequencing with tandem mass spectrometry (MS/MS) using a QSTAR XL mass spectrometer. The MS/MS spectra were submitted to Mascot set to the human Swiss-Prot database with a fragment tolerance window of 0.8 Da, peptide tolerance of 1.2 Da, and MALDI-QUAD-TOF selected as the instrument. Only MS/MS spectra with a significant Mascot score (p<0.05) were considered acceptable. The proteins were considered positively identified with the criteria of a significant full-scan match and one or more significantly-matched peptide sequences. In some instances, a significant score derived from several peptide sequences was considered as positive verification (Table 2 and 3).

Table 2.

Proteins increased in content in pre-retinopathic RPE.

Average ratio DM/Ctrla	P valueb	Spot No.c	Protein name	Accession number (Swiss-Protd)	Sequence Coverage (%)	Mascot score and [p value]	Theoretical MW (Da) [pI]e	Observed MW (Da) [pI]f	MS/MS Peptides (#)	Function (Swiss-Prot/ PubMedd)	Altered expression in DMg
1.66	0.004	2	Aldehyde dehydrogenase mitochondrial precursor	P05091	29	82 [9.0E-05]	56859 [6.63]	55000 [6.2]	2	Oxidoreductase	45
1.67	0.039	3	Aldehyde dehydrogenase mitochondrial precursor	P05091	42	112 [1.00 E-07]	56859 [6.63]	55000 [6.5]	1	Oxidoreductase	45
1.60	0.049	4	Annexin A4	P09525	48	83 [8.5E-05]	35860 [5.84]	34000 [6.2]	2	Membrane dynamics
1.75	0.001	5	Annexin A7	P20073	27	58 [0.025]	52991 [5.52]	47000 [6.7]	2	Membrane dynamics
5.20	0.0006	1	Beta-actin	P60709	23	62 [0.0088]	41710 [5.29]	41000 [5.4]	2	Structural protein	46
2.31	0.011	6	Cathepsin D precursor	P07339	30	77 [0.00029]	45037 [6.10]	31000 [5.5]	11	Protease (lysosome)	47
1.83	0.001	7	Cellular retinaldehyde binding protein	P12271	45	87 [3.2E-05]	39451 [4.98]	39000 [5.4]	5	Retinoid metabolism
1.62	0.046	8	Dihydrolipoyl dehydrogenase	P09622	18	55 [0.049]	54686 [7.59]	60000 [7.5]	3h	Energy metabolism
2.03	0.003	9	Elongation factor tu, mitochondrial	P49411	44	111 [1.30-07]	49852 [7.26]	45000 [7.2]	1	Chaperone (mitochondrial), Protein Synthesis	48
2.05	0.04	10	GRP75	P38646	27	60 [0.0024]	73635 [5.87]	70000 [5.9]	2	Chaperone (mitochondrial)	14
1.80	0.008	11	Heat shock cognate 71kDa protein	P11142	47	141 [1.30E-07]	71082 [5.37]	70000 [5.5]	2	Chaperone (cytosol)
1.61	0.04	12	Protein disulfide-isomerase A3 precursor	P30101	25	77 [0.00029]	56747 [5.98]	58000 [6.1]	4	Chaperone (endoplasmic reticulum)	33,50
1.35	0.013	13	Protein disulfide-isomerase A3 precursor	P30101	44	140 [1.6E-10]	56747 [5.98]	58000 [6.2]	2	Chaperone (endoplasmic reticulum)	33,50
1.80	0.001	14	Selenium binding protein 1	Q13228	25	77 [0.0003]	52907 [6.13]	55000 [6.5]	3h	Protein transport
1.81	0.016	15	Sterol carrier protein x	P22307	25	58 [0.026]	59640 [6.44]	50000 [6.4]	1	Energy metabolism	31

^aNormalized intensity values averaged for the identified protein spots are calculated as a ratio (Diabetes/Control)

^bResults from Student’s t test comparison of diabetic and control samples

^cSpot numbers correspond to numbers used in Figures 1 and 2

^dDatabases containing protein sequence information and journal citations

^eTheoretical molecular weight (MW) in Daltons (Da) and isoelectric point (pI)

^fObserved approximate molecular weight (MW) in Daltons (Da) and isoelectric point (pI)

^gProteins previously reported to be associated with diabetes in tissues other than RPE.

^hAll sequenced peptides collectively gave a significant match to the protein identified. Number of peptides sequenced are listed.

Table 3.

Proteins decreased in content in pre-retinopathic RPE.

Average ratio DM/Ctrla	P valueb	Spot No.c	Protein name	Accession number (Swiss- Protd)	Sequence Coverage (%)	Mascot score [p value]	Theoretical MW (Da) [pI]e	Observed MW (Da) [pI]f	MS/MS Peptides (#)	Function (Swiss-Prot/ PubMedd)	Altered expression in DMg
0.53	0.006	16	Gamma enolase	P09104	40	106 [3.9E-07]	47239 [4.91]	47000 [5.2]	2	Energy metabolism	22
0.58	0.026	17	Phosphoglycerate mutase 1	P18669	47	82 [9.40E-05]	28900 [6.67]	28000 [7.5]	2	Energy metabolism	26
0.64	0.038	18	Succinyl CoA: 3-ketoacid coenzyme A	P55809	22	56 [0.040]	56122 [7.14]	58000 [7.5]	1	Energy metabolism	49

^aNormalized intensity values averaged for the identified protein spots are calculated as a ratio (Diabetes/Control)

^bResults from Student’s t test comparison of diabetic and control samples

^cSpot numbers correspond to numbers used in Figures 1 and 2

^dDatabases containing protein sequence information and journal citations

^eTheoretical molecular weight (MW) in Daltons (Da) and isoelectric point (pI)

^fObserved approximate molecular weight (MW) in Daltons (Da) and isoelectric point (pI)

^gProteins previously reported to be associated with diabetes in tissues other than RPE.

Statistical analysis

To determine the number of samples necessary to detect statistically significant changes between groups, a power analysis was performed on Flamingo-stained gels (n=17 control and n=6 diabetic) as outlined [15]. Linear regression analysis was performed to compare spot density with time from death to freezing. Analysis was done for both gel exposure times. The p value for linear regression and critical values for correlation coefficient (R) was taken into consideration when verifying the significance of the relationship (Origin Lab 7.5).

The normalized intensity values of individual protein spots were compared between the two groups by Student’s two-tailed t test for unpaired samples. Values outside three interquartile ranges from the 25^th and 75^th percentiles of the data distribution were removed. When the assumption of equal variance (Modified-Levene Equal-Variance Test) was violated, the Aspin-Welch Unequal-Variance Test was used. When normality assumptions were not met, data were either transformed to natural log to obtain a normal distribution or Kolgomarov-Smirnov nonparametric test was used. The results are expressed as mean ± SEM and p< 0.05 was considered statistically significant (NCSS 2001, Kaysville, Utah, USA).

Results

Experimental Design

Demographic and clinical donor information obtained from the Minnesota Lions Eye Bank is summarized in Table 1. Donors with clinically evident eye disease were excluded from the study. Average time from enucleation to freezing (17.7±3.5 hours; mean ±SD) was not significantly different between the two groups (p=0.996).

Table 1.

Donor Demographics

Group	Sample Size	Gendera		Age (yrs)b		TAD (hrs)c	Cause of Death ( # )d
		M	F	Mean	Range
Controls	17	11	7	65± 9	52-86	17.8 ± 3.7	Cancer (5), Sepsis (7), Respiratory (3), CVA (1), Multiorgan failure (1), Cardiomyopathy (1)
DM	6	3	3	61± 10	48-79	17.5 ± 3.3	Cancer (1), Sepsis (1), Renal failure (2), MI (1), Multiorgan failure (1).

^aM=male; F=female

^bAge in years, Mean ± SD and age-range

^cTime after death (TAD) until tissue freezing (mean ± SD).

^dNumber of donors for each cause of death category is shown in parenthesis. CVA = Cerebro-Vascular Accident; MI = Myocardial Infarction.

Analysis included 17 and 6 gels from control and diabetic donors, respectively. A total of 325 spots were analyzed. In some instances, spots were eliminated from analysis if they were not clearly resolved, had artifacts that interfered with density measurements or they were statistically-defined outliers. However, each spot had to be present in a minimum of 5 samples from the diabetic group and 14 samples from the controls group to be included in the analysis. A power analysis based on the average variation in intensity of individual spots from 2-D gels indicated that this number was sufficient to detect at least a 75% difference in intensity with 80% power and α= 0.05.

Linear regression was performed (spot density versus time from death to freezing) to evaluate postmortem stability of proteins in our samples. No time-dependent change in density was noted for 94% of all spots examined, and no change was observed for the spots identified in our study. These results confirm that the protein expression changes observed in this study are not due to postmortem protein degradation, but rather result from altered protein content due to the disease state.

Expression Analysis and Protein Identification

The RPE proteome was analyzed using 2-D gel electrophoresis and mass spectrometry to identify differentially expressed proteins between the diabetic and control groups. A representative Flamingo-stained fluorescent gel is shown in Figure 1. A total of 325 spots were analyzed. Thirty-one protein spots exhibited a significant change in expression with diabetes. Protein identity was determined for 18 spots; as shown in Figure 2, 15 spots were upregulated and 3 spots were downregulated. Initial identification of protein spots was obtained by MALDI-TOF MS peptide mass fingerprinting and confirmed by MS/MS peptide sequencing. Protein identification was based on six or more matching peptides (an average of 11, range of 6-21), statistically significant MOWSE score (p<0.05) and the sequencing of at least 1 peptide. Since multiple proteins can co-migrate in a single spot, we re-examined our MALDI full scans by removing the peptides that matched the identified protein and re-searching for peptides from other proteins. This secondary search did not reveal additional proteins.

FIGURE 1.

RPE proteins resolved by 2-D gel electrophoresis. Representative Flamingo stained-gel (125 μg) indicates identified proteins showing altered expression with diabetes. Boxed spots show increased expression and circled spots show decreased expression. The pI range for separation in the first dimension (pH 5-8) is shown at the top. The position of molecular mass markers is indicated on the left. Numbers correspond with density summaries in Figure 2 and identified proteins listed in Tables 2 and 3.

FIGURE 2.

Summary of protein spot density. Results of densitometry for protein spots demonstrating a significant increase (a, b) or decrease (c) in spot density that were identified by mass spectrometry. Numbers correspond with Figure 1 and identified proteins listed in Tables 2 and 3.

Overall, the experimental molecular mass (MW) and isoelectric point (pI) were similar to the theoretical values for each protein, with the exception of cathepsin D (Table 2). Spot #6, identified as cathepsin D, migrates at an apparent mass of 31 kDa but matched to a theoretical MW of 45kDa. This protein is initially synthesized as an inactive proenzyme (52 kDa) that is subsequently converted into an active intermediate (46 kDa) and finally cleaved in the lysosome to generate the mature 31 kDa form [20]. The migration of spot #6 is consistent with the theoretical migration of mature cathepsin D protein.

Two proteins we identified migrated in multiple spots. Aldehyde dehydrogenase was identified in two spots (# 2 and #3) that migrated at a similar MW but a different pI, as did protein disulfide-isomerase A3 (#12 and #13). These findings could reflect posttranslational modifications that cause an acidic or basic shift in migration for a subset of the protein population.

Classification of the Identified Proteins

The identified proteins were categorized according to their function based on published literature and the Swiss-Prot database. The major functional groups include energy metabolism (29%) and chaperones (23%). Other functional groups include membrane dynamics (12%), structural proteins (6%), protein transport (6%), protein degradation (6%), protein synthesis (6%), retinoid metabolism (6%), and oxidoreductases (6%) (Figure 3).

FIGURE 3.

Summary of the functional groups for proteins with altered expression in pre-retinopathic RPE. The diagram indicates the relative percent of proteins in each functional group. a: Protein degradation, b: Structural, c: Chaperones, d: Membrane dynamics, e: Energy metabolism, f: Protein synthesis , g: Protein transport, h: Retinoid metabolism, i: Oxidoreductases

Our list includes proteins residing in different subcellular compartments. The majority of proteins were cytoplasmic (59%). Other proteins were identified from the mitochondria (aldehyde dehydrogenase, elongation factor Tu, GRP75, and succinyl CoA: 3-ketoacid-coenzyme A transferase 1), the endoplasmic reticulum (protein disulfide isomerase A3), peroxisomes (sterol carrier protein x) and lysosomes (cathepsin D).

Discussion

In the present study we analyzed the RPE proteome in pre-retinopathic, human, diabetic eyebank eyes. Analysis of 325 spots on 2-D gels identified 31 spots that were either up- or down-regulated relative to age-matched controls. The protein identity of eighteen spots was determined by mass spectrometry. A majority of altered proteins belonged to two major functional groups, metabolism and chaperones, while other affected categories included protein degradation, synthesis and transport, oxidoreductases, cytoskeletal structure, and retinoid metabolism (Figure 3). Approximately 62% of these proteins have been previously reported to undergo diabetes-related expression changes in non-ocular tissues (Tables 2 and 3) and thus represent novel findings for the retina. Because of the unique role of the RPE in supporting the photoreceptors and the retinal microenvironment, the pre-clinical changes identified here provide our first insights into the early molecular changes that precede diabetic retinopathy.

Our diabetic sample included donors with either diabetes mellitus type 1 or type 2, based upon Eye Bank records rather than direct clinical records. Ideally, we would analyze type 1 and 2 samples separately. However, this was not feasible because of limited tissue availability. If the proteome of the two groups were significantly different in the pre-retinopathic stage we would have observed high variability within our diabetic group. No outliers were detected in this group suggesting that the proteomes of our type 1 and type 2 donors were comparable in the pre-retinopathic stage.

The 16 proteins identified herein reflect a conservative estimate, due to our limited sample size and technical constraints of 2-D gels. We could detect on average a 75% difference with our sample size, but smaller changes that are physiologically meaningful may have been undetected. Membrane proteins resolve poorly in the first dimension of a 2-D gel and are likely underrepresented in the analysis. Low abundant proteins are often below the detection limit of 2-D gels and would be under represented in this analysis. While the pH range of 5 to 8 was used because it resolved the greatest number of spots, proteins outside that range would not have been detected using this methodology. Finally, due to constraints with peptide recovery in mass spectrometry, we were unable to identify low-mass spots. Despite these caveats, this study identified many novel diabetes-related changes in the RPE proteome.

Vascular endothelial growth factor (VEGF) and pigment epithelium derived factor (PEDF), growth factors produced by RPE, are considerably altered in DR [21]. The differential expression and secretion of these factors lead to non-proliferative and proliferative DR (i.e. leaky vasculature and neovascularization) [21]. No change was detected for either factor in the present study. The absence of these two growth factors in our analysis is likely due to both technical constraints of the 2-D gels and secretion of these growth factors from the RPE into the extracellular space. For example, the isoelectric point of VEGF-A is 8.5, which is outside the pI range of our first dimension (i.e. pI 5-8) and thus would not resolve under our experimental conditions. PEDF was detected in previous 2-D analyses but only in extracellular fractions such as vitreous [22], suggesting that the intracellular content is not sufficient in RPE homogenates for detection in 2-D gels. Consistent with our results, these growth factors were not reported in a previous characterization of the human RPE cellular proteome [23].

To date, biochemical studies of human diabetic retinopathy have been mostly limited to analysis of the vitreous [22, 24-26]. These analyses have revealed several pathways and factors that provided insight into diabetes-related ocular alterations. For example, several studies have found an imbalance in angiogenic (i.e., VEGF-A) and anti-angiogenic (i.e., sol flt-1 receptor, PEDF) factors present in proliferative diabetic retinopathy by measuring their levels in the vitreous [25, 26]. Despite the contributions to understanding diabetic retinopathy, there are several limitations associated with these studies. First, vitreous protein levels are an indirect, secondary measure of intracellular events occurring in the retina. Second, vitreous collection usually occurs during a surgical procedure at the later stages of retinopathy where early biochemical changes leading to pathogenesis may not be evident. Third, the vitreous includes a large amount of albumin and immunoglobulin that may overlap with small spots or less abundant proteins and compromise the utility of 2-D gel analysis. Finally, the cellular origin of many vitreous proteins is unclear and may be non-retinal [27].

There are several advantages associated with the present study. First, we directly analyzed retinal proteins from affected human tissue. Second, based on direct clinical evaluation we excluded donor eyes with characteristic diabetic retinopathy and exclusively studied pre-retinopathic RPE. Third, our proteomic approach enabled simultaneous analysis of over 300 protein spots. Finally, the RPE does not contain disproportionately abundant proteins that otherwise interfere with the resolution of a large number of protein spots.

Twenty-nine percent of the proteins identified (Figure 3) participate in various metabolic pathways including glycolysis (phosphoglycerate mutase and gamma-enolase), the citric acid cycle (dihydrolipoyl dehydrogenase), lipid metabolism (sterol carrier protein x), and ketolysis (Succinyl CoA: 3-ketoacid-coenzyme A transferase 1). In other tissues prone to diabetic complications, such as liver, kidney, skeletal and heart muscle, energy metabolism is one of the key areas affected [28-32]. However, metabolic changes associated with diabetes are tissue-specific. For example, an increase in glycolysis was reported in skeletal muscle from type 1 and type 2 patients [28, 29] while a reduction was demonstrated in diabetic heart [30, 33] and in streptozotocin (STZ)-induced diabetic rat retinas after 3 months of disease [34]. Our results indicate that systemic metabolic changes observed in other tissues affected by diabetes also occur in the pre-retinopathic RPE. These tissue-specific changes may result in loss of RPE functions that subsequently impact retinal health and photoreceptor activity.

Chaperones comprised a second major functional category altered in diabetic RPE (Figure 3, Table 2). Chaperones mediate protein folding, defend against protein damage and aggregation due to misfolding, assist with translocation of proteins across intracellular membranes, and stabilize unstable protein conformers [35]. Chaperones from multiple subcellular compartments were identified, including mitochondria (GRP75 and elongation factor TU), cytosol (heat shock cognate 71) and the endoplasmic reticulum (protein disulfide-isomerase A3). Protein disulfide-isomerase A3 (PDIA3) is a specialized endoplasmic reticulum (ER) protein that assists with folding and formation of native disulfide bonds in nascent and unfolded proteins [36]. Specific upregulation of this ER chaperone suggests elevated unfolded proteins in the ER, which is associated with ER stress [37]. Numerous studies report the presence of ER stress in diabetes. For example, expression of ER stress markers such as BiP, GRP94 and CHOP are elevated in the Akita diabetic mouse model [38] and pancreatic beta cells of type II diabetic patients [39].

Aldehyde dehydrogenase 2 (ALDH2), a mitochondrial isoform of the aldehyde dehydrogenase family, was increased in our study. It is involved in detoxifying reactive aldehydes produced by lipid peroxidation [40]. Increased content of this protein likely reflects increased lipoxidation products in the diabetic RPE. Lipid peroxidation and oxidation of glycated proteins are two major mechanisms of protein damage resulting from oxidative stress in the diabetic retina [41]. Oxidative stress occurs when the production of oxidants, including reactive oxygen species (ROS), exceeds the level of antioxidants in the cell [42]. Numerous studies have shown elevated ROS and impaired antioxidant defense in the retina of diabetic animal models and humans [reviewed in 42]. Taken together, elevated levels of ALDH2 and chaperones may be a compensatory response to oxidative stress in the human pre-retinopathic RPE.

Some of the major protein categories altered in our study, such as energy metabolism, chaperones and cytoskeletal proteins [43, 44], overlap with findings from other tissues affected by diabetes. However, some proteins altered in the present study are completely novel to diabetes, such as proteins involved in retinoid metabolism (CRALBP), membrane dynamics (annexin A4, A7), and protein transport (selenium binding protein). While the consequences of these changes have yet to be determined, they are consistent with the global effect that diabetes has on multiple cellular processes.

In summary, we compared the human RPE proteome in a pre-retinopathic stage of diabetes to age-matched controls using a high throughput proteomic approach. Our study has demonstrated that significant biochemical changes take place in the RPE prior to clinically evident diabetic retinopathy. Our results indicate that changes in metabolic proteins parallel those found in other diabetic tissues. Changes in proteins associated with oxidative and ER stress also suggest protein damage in the pre-retinopathic RPE. Given the importance of the RPE in supporting the retinal microenvironment these alterations may hold significance in the early pathogenesis of diabetic retinopathy.

ABBREVIATIONS

2-D

two-dimensional

DR

Diabetic Retinopathy

DTT

Dithiothreitol

ER

endoplasmic Reticulum

IEF

Isoelectric focusing

MALDI-TOF

Matrix-assisted laser desorption/ionization time-of-flight

MW

molecular weight

MS

Mass spectrometry

pI

Isoelectric point

ROS

reactive oxygen species

RPE

retinal pigment epithelium

STZ

streptozotocin