Expression of Acute-Phase Response Proteins in Retinal Müller Cells in Diabetes

Abstract

purpose. To characterize the whole spectrum of gene expression changes induced by diabetes in retinal Müller glial cells.

methods. Müller cells were isolated from the retina of streptozotocin-diabetic and age-matched control rats by gradient centrifugation and immediately processed for RNA isolation. The gene expression profile of Müller cells was studied with the GeneChip Rat Genome oligonucleotide array (Affymetrix, Santa Clara, CA). The upregulation of acute-phase proteins in the retina of diabetic rats was confirmed by Northern and Western blot analyses. Real-time-RT-PCR was used to study the retinal expression of inflammatory cytokines.

results. Gene expression profiling identified 78 genes as differentially expressed in diabetic Müller cells. One third of these genes were associated with inflammation, including a large cluster (18% of the differentially expressed genes) of acute-phase response proteins: α₂-macroglobulin, ceruloplasmin, complement components, lipocalin-2, metallothionein, serine protease inhibitor-2, transferrin, tissue inhibitor of metalloproteases-1, transthyretin, and the transcription factor C/EBPδ. Northern and Western blot analyses confirmed the upregulation of α₂-macroglobulin and ceruloplasmin in the diabetic retina, but not in the cerebral cortex and liver of the same animals. The acute-phase response of Müller cells in diabetes was associated with upregulation of interleukin (IL)-1β in the retina.

conclusions. Müller cells acquire a complex and specific reactive phenotype in diabetes characterized by the induction of acute-phase response proteins and other inflammation-related genes. The concomitant upregulation of IL-1β in the retina of diabetic rats points to this cytokine as a possible mediator of the acute-phase response mounted by Müller cells in diabetes.

Diabetic retinopathy is the most frequent complication of diabetes and the leading cause of visual loss and blindness in the adult population of the United States. ¹ Although diabetic retinopathy has been traditionally viewed as a disorder of the retinal vasculature, recent evidence indicates that it also affects the glial and neural cells of the retina. ^{2 3}  

The principal glial cell of the retina is the Müller cell, a specialized radial glial cell spanning the entire depth of the retina. Through the extensive arborization of their processes, Müller cells constitute an anatomic and functional link between neurons and vessels. ⁴ Müller cells play a central role in retinal glucose metabolism, ⁵ regulation of retinal blood flow, ⁴ and the formation and maintenance of the blood–retinal barrier. ⁶ These characteristics make Müller cells both a target of diabetes and potential key players in the vascular alterations in diabetic retinopathy, such as hemodynamic abnormalities and increased vascular permeability. 

Experimental diabetes in the rat has been associated with functional changes in Müller cells, such as decreased activity of glutamine synthetase ⁷ and the glutamate/aspartate transporter GLAST ⁸ and accumulation of γ-aminobutyric acid. ⁹ In addition, we and others have shown that retinal Müller cells manifest increased expression of glial fibrillar acidic protein (GFAP) in both human ² and experimental diabetes. ^{7 10 11} Increased expression of GFAP is a well-known marker of glial cell reactivity, and this has suggested that Müller cells acquire a reactive phenotype in diabetes. 

Many in vivo and in vitro studies indicate that glial cells react to injuring factors by changing the expression of several groups of genes. ¹² Some of these changes have beneficial effects, fostering tissue repair and neuroprotection; others are potentially harmful and may contribute to further tissue damage. The net effect of these changes is often dictated by the type, duration, and site of injury. ¹³ We thus sought to characterize the whole spectrum of gene expression changes induced by diabetes in Müller cells. To this end, we compared the gene expression profile of Müller cells isolated from the retina of streptozotocin-diabetic rats to that of age-matched control rats. Our findings indicate that Müller cells mount an acute-phase response in diabetes, which is associated with retinal upregulation of the proinflammatory cytokine interleukin (IL)-1β. 

Methods

Animals

All experiments conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Use and Care Committee of the Schepens Eye Research Institute. Sprague-Dawley male rats (5 weeks of age; Taconic Farms, Germantown, NY) were randomly assigned to a diabetic or a control group. Diabetes was induced with a single intravenous injection of streptozotocin (55 mg/kg body weight), as previously described. ¹⁴ The diabetic and control rats had full access to food and water. Diabetic animals were treated with insulin (NPH, 2–4 U three times per week) throughout the study to prevent weight loss. Glycohemoglobin was measured at the time of killing (Glyc-Affin GHb assay; PerkinElmer, Norton, OH). To maximize the chances of detecting fully established changes of Müller cells that precede and may contribute to the demise of vascular cells, the gene expression profile of Müller cells was studied after 6 months of diabetes. At this time point, upregulation of GFAP by Müller cells is already established, ^{7 10 11} whereas apoptosis of vascular cells is not detectable until 8 months of diabetes. ^{15 16} Rats were killed by CO₂ inhalation, the eyes immediately removed, and the retinas dissected and processed for Müller cell or total retina RNA isolation. Rats assigned to the extraction of retinal proteins were perfused through the left ventricle with phosphate-buffered saline under deep anesthesia to remove plasma proteins from the retinal vessels. Brain, liver, and plasma samples were also collected and stored at −80°C, before processing for RNA or protein analysis. 

Isolation of Rat Retinal Müller Cells

Müller cells were isolated from rat retinas by density gradient centrifugation as described by Guidry, ¹⁷ with some modifications. A pool of 12 retinas (six rats) was used for each of the gradient purification. The retinas were digested with papain/DNase (Papain Dissociation System; Worthington Biochemicals, Lakewood, NJ) for 40 minutes at 37°C, mechanically dissociated, and the single-cell suspension was layered on a continuous 0% to 50% Percoll (Amersham Biosciences, Piscataway, NJ) gradient in Dulbecco’s modified Eagle’s medium (DMEM). After centrifugation at 800g for 20 minutes, 11 1-mL fractions were collected, starting from the top of the gradient. The Percoll was removed by dilution with DMEM followed by centrifugation, and the pelleted cells were resuspended in fresh medium. A drop of each gradient-fraction cell suspension was spread on glass microscope slides and examined by phase-contrast microscopy to identify the Müller-enriched fractions. Müller cells were easily recognized on the basis of their characteristic bipolar shape and were found to concentrate in fractions 3 and 4 (1.05–1.06 g/L density). The Müller-enriched fractions were then immediately pooled for RNA isolation. 

Immunocytochemistry

To confirm the enrichment and purity of the isolated Müller cells, cells recovered from each gradient fraction were analyzed by indirect immunofluorescence. Immunostaining was performed as described, ¹⁸ with the following primary antibodies: mouse anti-pig vimentin (1:100 dilution; clone V9; Sigma-Aldrich, St. Louis, MO) and mouse anti-sheep glutamine synthetase (GS; 1 μg/mL; Chemicon International, Temecula, CA) for Müller cells; mouse anti-rat Mac1 (10 μg/mL; clone OX-42; Chemicon), mouse anti-phosphotyrosine (PY; 4 μg/mL; clone 4G10, Upstate Biotechnology, Lake Placid, NY), and isolectin B4 (20 μg/mL; Vector Laboratories, Burlingame, CA) for microglia; rabbit anti-cow neuron-specific enolase (NSE; 1:100; Chemicon) for neurons; rabbit anti-human von Willebrand factor (1:1000; Dako, Carpinteria, CA) for endothelial cells; and mouse anti-human α-smooth muscle actin (1 μg/mL; Dako) for pericytes and smooth muscle cells. Rabbit anti-cow GFAP antibodies (1:3000; Dako) were used to identify astrocytes. However, because in the diabetic retina Müller cells become strongly immunoreactive for GFAP, double labeling for GFAP and vimentin was used to discriminate between astrocytes (GFAP⁺/vimentin⁻) and Müller cells (GFAP⁺/vimentin⁺). Slides were mounted in anti-fade medium containing DAPI (Vectashield-DAPI; Vector Laboratories) to counterstain the nuclei. Specificity of the immunostaining was verified by substituting the primary antibodies with an equivalent dilution/concentration of the appropriate nonimmune IgG (not shown). 

Müller Cells RNA Isolation and Microchip Array Hybridization

Total RNA was isolated from Müller cells (RNeasy Mini Kit; Qiagen, Valencia, CA) according to recommended Affymetrix (Santa Clara, CA) protocols. In the case of diabetic animals, RNA isolated from Müller-cells obtained from two density gradients, corresponding to the retinas of 12 diabetic rats, was pooled for the gene expression analysis and assayed in duplicate. In the case of control animals, three independent RNA pools (each corresponding to the retinas of 12 control rats) were obtained and independently assayed. Gene expression profiling was performed with the GeneChip Rat Genome RG-U34A array (Affymetrix). Five micrograms of RNA were used for each of the five array hybridizations (two diabetic and three nondiabetic control rats). Preparation of biotinylated cRNA, GeneChip hybridization, staining, and scanning of the arrays were performed according to recommended Affymetrix protocols ¹⁹ by the Harvard Medical School, Partners Healthcare Center for Genetics and Genomics Core Laboratory (Cambridge, MA; http://www.hpcgg.org/Affy/index.html). After hybridization (Model 320 oven; Affymetrix), washing and staining of arrays were performed with the GeneChip Fluidics Station 400 (Affymetrix). Arrays were then scanned (GeneArray Scanner 2500; Affymetrix) using the Microarray Suite 5.0 software (Affymetrix). 

Array Data Analysis

The microarray data were imported into the Rosetta Resolver system for gene expression data analysis (Rosetta Biosoftware, Kirkland, WA). After data preprocessing (background correction and intrachip normalization) the Rosetta Resolver system Affymetrix GeneChip error model was used to create an intensity profile for each microarray. Intensity profiles were first analyzed using the Rosetta Resolver system one-sided error-weighted ANOVA, which uses an input error estimate in conjunction with the observed input value to produce variance estimates within and between the two groups. Next, the Rosetta Resolver system Ratio Builder was used to calculate the changes (x-fold) and ratio probabilities for the differential expression between diabetic and control samples. (Detailed information on the Rosetta Resolver system Affymetrix GeneChip error model, error-weighted ANOVA, and Ratio Building error model can be found at http://www.rosettabio.com/tech.) Only those genes with an ANOVA P ≤ 0.05, a ratio P ≤ 0.05, and a change in expression of twofold or more were considered to be significantly differentially expressed genes. The genes whose expression was affected by diabetes were assigned to functional categories based on Affymetrix and National Center for Biotechnology Information (NCBI; Bethesda, MD) annotations and data from the literature. 

The complete microarray data set has been deposited in the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE1979. 

Northern Blot Analysis

Northern blot analysis was performed with total RNA isolated by guanidine-cesium chloride centrifugation ¹⁸ from the whole retina, liver, and cerebral cortex of individual rats. The membranes were sequentially hybridized to cDNA probes for rat α₂-macroglobulin (α₂M; 3.5-kb BamHI fragment, plasmid pRLα₂M/29J; American Type Culture Collection, Manassas, VA), rat ceruloplasmin (2-kb EcoRI fragment, plasmid pJFM26, gift of Julian F. Mercer, Deakin University, Australia), and chicken β-actin (Oncor, Gaithersburg, MD). Probes were labeled with [³²P] dCTP by using a random labeling system (Rediprime II; Amersham Biosciences, Little Chalfont, UK). Relative band density was determined on scanned autoradiographs with NIH Image software (available by ftp at zippy.nimh.nih.gov/ or at http://rsb.info.nih.gov/nih-image; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD). The intensity of the β-actin signal, which was not different in diabetic and control rats, was used as the endogenous control for loading. Data are expressed as transcript-to-β-actin ratios. 

Western Blot Analysis

Retinas were homogenized in ice-cold RIPA buffer (50 mM Tris [pH 7.4], 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 0.5% Na deoxycholate, and 0.1% SDS) containing protease and phosphatase inhibitors. ²⁰ Protein concentration of the retina lysates and plasma samples was determined with the bicinchoninic acid method using BSA as the standard (Micro BCA Protein Assay Kit; Pierce Biotechnology, Rockford, IL). Proteins were resolved by SDS-PAGE and immunoblotted as described. ¹⁴ For the study of ceruloplasmin, membranes were blocked overnight in 5% nonfat milk in Tris-buffered saline containing 0.05% Tween-20. For the study of α₂M, the overnight block was followed by an additional 1-hour blocking in Superblock (Pierce Biotechnology). The primary antibodies were: rabbit anti-human α₂M (1:200; Dako) and goat anti-human ceruloplasmin (1:1000; Sigma-Aldrich). Immunoreactive bands were visualized by enhanced chemiluminescence (SuperSignal; Pierce Biotechnology), and the relative band density was determined on the scanned autoradiographs with NIH Image software. Data are expressed as densitometric units per microgram of protein. 

Real-Time RT-PCR

cDNA was synthesized from 1 μg of total retinal RNA treated with DNase as described. ²⁰ Real-time PCR was then performed on a sequence-detection system (Prism 7900HT; Applied Biosystems, Inc. [ABI], Foster City, CA) using the separate-tubes method ²¹ and the TaqMan PCR Core Reagent kit (ABI). Primers and probe sets for rat IL-6, tumor necrosis factor (TNF)-α, and IL-1β were purchased from ABI. Primers and probe for rat β-actin (forward primer: 5′-CCT CTG AAC CCT AAG GCC AA-3′, reverse primer: 5′-AGC CTG GAT GGC TAC GTA CA-3′, probe: 5′VIC-TGA CCC AGA TCA TGT TTG AGA CCT TCA AC-TAMRA3′) were designed from the rat β-actin sequence (NM_031144). β-Actin was used as the endogenous control to normalize the amount of cDNA added to each reaction (ΔC_T), and the mean ΔC_T of control samples was used as the calibrator to calculate the ΔΔC_T. Quantitation of each transcript was by the comparative C_T method. In this method, the relative quantity of target mRNA, normalized to the endogenous control and relative to the calibrator, is equal to 2^−ΔΔCT. ²¹  

Statistical Analysis

Northern, Western, and real-time-PCR data are summarized with the mean ± 1 SD. Statistical analysis was performed with the unpaired t-test and linear regression analysis (StatView; SAS, Cary, NY). 

Results

Characterization of Müller Cell Preparations

We produced Müller-enriched cell preparations by means of density gradient centrifugation of freshly dissociated retinas. Immunocytochemistry of the preparations showed that most of the isolated cells were Müller cells, as indicated by positive immunostaining for the Müller cell markers vimentin and GS (Fig. 1) . Preparations from control and diabetic animals had similar enrichments, indicating that the diabetic milieu did not affect the isolation and purification of Müller cells. Contamination of the Müller cell preparations by other retinal cell types was minimal. There were virtually no astrocytes, as indicated by the absence of GFAP⁺/vimentin ^– cells (Figs. 2A 2B) . In preparations from control animals, some Müller cells (vimentin⁺) were also GFAP⁺, although the intensity of the GFAP immunostaining was very low (Fig. 2A) , in agreement with the low levels of GFAP expression in healthy Müller cells. On the contrary, all Müller cells isolated from diabetic animals showed intense GFAP immunostaining (Fig. 2B) , consistent with the described upregulation of GFAP in Müller cells in the diabetic retina. Microglial contamination was <1% in preparations from both control and diabetic retinas (Figs. 2C 2D 2E 2F) . Although neurons were present in the Müller-enriched preparations from both control and diabetic retinas, they accounted for <20% of the cells recovered by density gradient centrifugation (Figs. 2G 2H) . No endothelial cells or pericytes/smooth muscle cells were present in either of the Müller cell-enriched fractions (data not shown). 

Diabetes-Induced Changes in Müller Cells

To determine the impact of diabetes on Müller cells, we compared the expression profile of Müller-cell–enriched fractions derived from the pooled retinas of 12 diabetic rats with that of similar preparations from age-matched nondiabetic control rats. The average body weight of the animals at the time of killing was 326 ± 41 g in diabetic versus 613 ± 95 g in nondiabetic rats. Glycohemoglobin was 17.9% ± 2.4% in diabetic and 4.5% ± 0.5% in control rats. Array analysis identified 97 genes (78 known genes and 19 expressed sequence tags [ESTs]) having significantly different expression levels in diabetic than in nondiabetic animals (Table 1) . Fifty-nine of the 78 differentially expressed known genes were upregulated. Of these, 20 had been reported to be upregulated in the brain glia in response to various injuries, ^{12 13 22 23 24 25} and three (GFAP, ceruloplasmin, and galectin-3) had been shown to be upregulated in reactive Müller cells in several retinal diseases ^{26 27 28} (Table 1) . Two additional genes known to be induced in reactive Müller cells—clusterin and CD81 ^{29 30} —showed a modest, though significant, upregulation (1.45-fold, P < 0.0001 and 1.62-fold, P < 0.0001, respectively). Of note, several changes often found associated with reactive Müller cells in other retinal diseases, were not observed in the Müller cells of diabetic retina. Specifically, the expression of vimentin, carbonic anhydrase 2, GS, Bcl-2, apolipoprotein E, and CD44 ^{26 31 32 33} did not change in diabetic compared with control rats. The finding of GFAP, but not Bcl-2 and GS, upregulation is in agreement with our previous observation of increased levels of GFAP but not of Bcl-2 and GS protein in Müller cells from human diabetic retinas. ²  

Upregulation of Acute-Phase Response Proteins

One third of the known genes differentially expressed in diabetic Müller cells (26/78, 33%) coded for proteins associated with inflammation. Fourteen of those genes coded for acute-phase proteins, making this the single largest functional cluster of genes affected by diabetes (18% of all genes significantly changed in diabetes). This group of genes included α₂M; angiotensinogen; ceruloplasmin; the complement components C1, C3, factor B (C2), and C1 inhibitor; lipocalin 2; metallothionein; serine protease inhibitor 2 (Spin2c)/antichymotrypsin; transferrin; tissue inhibitor of metalloprotease 1 (TIMP-1); and transthyretin, as well as the transcription factor C/EBPδ (CCAAT/enhancer-binding protein δ). ^{34 35 36} The induction of C/EBPδ by proinflammatory cytokines is an essential step in the acute-phase response. ³⁵ Thus, upregulation of C/EBPδ in diabetes identifies a possible pathway to the acute-phase response of Müller cells. 

Other upregulated genes coding for proteins associated with an inflammatory response included major histocompatibility complex (MHC) proteins, intercellular adhesion molecule (ICAM)-1, the p105 subunit of NF-κB, osteopontin, scavenger receptor B1, galectin-3, and annexin 1 (Table 1) . 

To verify the upregulation of acute-phase proteins in the diabetic retina, we focused on α₂M and ceruloplasmin because, although their x-fold increase in diabetes was not as high as that of other acute-phase proteins, they showed the highest expression levels in Müller cells (data not shown). The expression of α₂M and ceruloplasmin was studied in retinas obtained from an independent group of diabetic and age-matched control rats. Duration of diabetes (6 months), average body weight (diabetic rats 309 ± 54 g and nondiabetic rats 607 ± 84 g) and glycohemoglobin levels (diabetic rats 15.7% ± 2.3%, nondiabetic rats 4.4% ± 0.4%) of these animals at time of killing were not different from those of the rats used for the gene expression profile experiment. For both α₂M and ceruloplasmin, Northern blot analysis detected a single mRNA transcript of the expected size (5.0- and 3.8-kb, respectively; Fig. 3A ). In agreement with the array data, the levels of α₂M and ceruloplasmin mRNA were significantly increased in the diabetic compared with the control retinas (α₂M/β-actin ratio: 1.27 ± 0.52 vs. 0.43 ± 0.18, P = 0.0006; ceruloplasmin/β-actin ratio: 1.17 ± 0.52 vs. 0.23 ± 0.08, P = 0.0003; Fig. 3B ). The levels of the two transcripts correlated highly in the diabetic retinas (R = 0.85; P = 0.004), suggesting a common pathway mediating the upregulation of these two transcripts in response to diabetes (Fig. 3C) . The increased mRNA levels were paralleled by a similar increase in the corresponding proteins, as detected by Western blot analysis (α₂M: 1499 ± 810 in diabetic animals versus 388 ± 232 densitometric units/μg protein in control rats, P = 0.01; ceruloplasmin: 2049 ± 633 vs. 1224 ± 298, P = 0.016; Fig. 4 ). 

Retina-Specific Upregulation of α₂M and Ceruloplasmin in Diabetes

To determine whether the upregulation of acute-phase proteins in Müller cells is a specific response of this cell type to diabetes or the result of a systemic effect of the disease, we investigated the expression of α₂M and ceruloplasmin in the liver and cerebral cortex of diabetic and nondiabetic rats. Both α₂M and ceruloplasmin were easily detectable by Northern blot analysis in the liver (Fig. 5A) . The mRNA levels of α₂M were lower in diabetic than in control rats (α₂M/β-actin ratio: 0.24 ± 0.11 vs. 0.97 ± 0.32, P = 0.0002), whereas the expression of ceruloplasmin was similar in the two groups of animals (Fig. 5B) . Consistent with these findings, the levels of circulating α₂M were decreased in diabetic animals (922 ± 424 vs. 2672 ± 431 densitometric units/μg protein, P = 0.02), whereas those of ceruloplasmin were not different in the two groups (Figs. 5C 5D) . 

Neither α₂M nor ceruloplasmin mRNA was detectable by Northern blot analysis in the cerebral cortex of control or diabetic rats (not shown), indicating minimal, if any, baseline expression and no significant induction of these genes in this tissue in response to diabetes. 

Increase in the Proinflammatory Cytokine IL-1β in the Retina of Diabetic Rats

Because IL-6, IL-1β, and TNF-α are the major inducers of acute-phase proteins, ³⁷ we next investigated whether diabetes induces the synthesis of these cytokines in the rat retina. All three cytokines were expressed in the retina of control rats, albeit at low levels (average C_T: 32 ± 0.8 for IL-6, 29 ± 0.8 for TNF-α, and 36 ± 1.5 for IL-1β). Diabetes was associated with a dramatic upregulation of IL-1β expression in the retina (relative mRNA quantity: 17.6 ± 11.8 in diabetic rats versus 1.7 ± 1.9 in nondiabetic control animals, P = 0.0011). By contrast, no significant changes were observed in the expression of IL-6 or TNF-α (Fig. 6) . 

Discussion

Our study demonstrates that retinal Müller glial cells acquire a complex reactive phenotype in response to diabetes, well beyond the known changes in GFAP expression. A major feature of Müller cell reactivity in diabetes is the induction of a cluster of acute-phase response proteins and other inflammation-related genes. The concomitant upregulation of IL-1β in the retina of diabetic rats points to this cytokine as a possible mediator of the acute-phase response mounted by Müller cells in diabetes. 

Müller cells acquire a reactive phenotype in many retinal diseases, such as retinal detachment, retinal degeneration, and glaucoma. As part of this phenotype, a wide array of proteins undergoes dramatic expression changes ^{26 27 28 29 30 31 32 33} that are generally believed to be a response of Müller cells to the loss of retinal neurons. Our data indicate that only some of these changes occur in Müller cells exposed to diabetes. GFAP, ceruloplasmin, and galectin-3 were induced, but clusterin and CD81 were only moderately upregulated, and vimentin, GS, Bcl2, apolipoprotein E, CD44, and carbonic anhydrase 2 did not show significant expression changes. Such dissociation suggests that the reactive phenotype acquired by Müller cells in diabetes is not a stereotypical response of the glia to neuronal loss, although neuronal apoptosis does occur in the diabetic retina. ^{3 11} Rather, the Müller cells response appears to be specific to the diabetic milieu. 

In addition to the known markers of Müller cells reactivity described herein, we have identified a number of genes whose expression in normal or diseased retina has not been described before. Several of these genes are known to be induced in reactive brain astrocytes. ^{12 13 22 23 24 25} However, because our Müller cell preparations were free of astrocytes, this cell type can be excluded as the source of these expression changes. The same applies to microglial, endothelial, and pericyte/smooth muscle cells, which were minimally, if at all, present in those density fractions. The minor degree of contamination in the Müller cell preparations was due to neurons, and therefore we cannot exclude a contribution of neural cells to the changes detected by array analysis of the Müller-enriched preparations. However, most of the differentially expressed genes are glia specific, and thus, are likely to be due to Müller cells. 

A remarkable characteristic of these diabetes-induced genes is their relation to inflammation. Especially intriguing is the presence among the upregulated genes of a large cluster of acute-phase proteins. These proteins are constitutively expressed by the liver, and their synthesis and circulating levels are dramatically induced during systemic inflammation. ³⁴ Induction of acute-phase proteins is also a feature of neuroinflammation. ²² The suggestion of an inflammatory component of diabetic retinopathy dates back to 1964 when Powell and Field ³⁸ reported that chronic consumption of high doses of aspirin for the treatment of rheumatoid arthritis had a protective effect on the development and severity of diabetic retinopathy. In recent years, the concept of diabetic retinopathy as a subtle inflammatory disorder has regained attention based on novel molecular and cellular abnormalities detected in the diabetic retina, such as leukocytes adhesion to the endothelium, ³⁹ complement deposition, ¹⁴ and microglial activation. ¹⁰  

Several observations indicate that the upregulation of these proteins in Müller cells is a coordinated response to a local proinflammatory environment induced by diabetes in the retina. First, we observed the movement of a large cluster of acute-phase proteins. Second, their induction was associated with upregulation of the transcription factor C/EBPδ, which is a major inducer of acute-phase gene transcription and is known to be upregulated in response to proinflammatory cytokines, such as Il-1β and TNF-α. ³⁷ Third, the upregulation/induction of acute-phase proteins in Müller cells was associated with a dramatic upregulation of IL-1β. Many of the genes induced in diabetic Müller cells are known targets of this cytokine: C/EBPδ, ⁴⁰ α₂M, ⁴¹ ceruloplasmin, complement components, metallothionein, and TIMP-1, ⁴² Spin2c/antichymotrypsin, ⁴³ lipocalin 2, ⁴⁴ the FGF receptor, and the adhesion molecule ICAM-1. ⁴² Thus, our data point to IL-1β as the possible inducer of the acute-phase response and other reactive changes observed in Müller cells exposed to diabetes. 

Our finding of increased expression of IL-1β in the diabetic retina is in agreement with the finding by Carmo et al. ⁴⁵ that protein levels of this cytokine increase in the rat retina shortly after the induction of diabetes. The lack of retinal changes in TNF-α is instead at variance with the report by Joussen et al., ⁴⁶ which describes increased level of this cytokine in the retina of diabetic rats. This discrepancy could be due to the different duration of diabetes in the two studies (6 months versus 1 week). An alternative explanation is that the increased protein levels of TNF-α in that report were due to increased plasma levels or increased synthesis of this cytokine by circulating white blood cells, rather than increased production by retinal cells. The retinal cell types responsible for IL-1β synthesis in diabetes are unknown at this time. IL-1β was not among the upregulated genes identified in diabetic Müller cells by array analysis, suggesting that Müller cells are not the source of IL-1β in the diabetic retina. 

Müller cells contain aldose reductase in all species studied to date ¹¹ and may therefore be a site of activation of the polyol pathway when exposed to diabetic hyperglycemia. Indeed, we have recently demonstrated that the aldose reductase inhibitor sorbinil prevents the upregulation of GFAP in Müller cells in the retina of diabetic rats. ¹¹ Whether activation of this pathway of glucose metabolism is responsible for the whole set of changes induced by diabetes specifically in Müller cells, including the upregulation of acute-phase proteins, is currently under investigation. Of note, inhibition of aldose reductase has been shown to prevent the enhancing effect of high glucose on IL-1β–induced synthesis of prostaglandin in smooth muscle cells, as well as on IL-1β–induced production of nitric oxide in rat aortic rings, ^{47 48} suggesting a possible role of the polyol pathway in the regulation of IL-1β activity. 

The consequences of the upregulation of acute-phase proteins and IL-1β in the diabetic retina remain speculative at this time. The induction of acute-phase proteins in response to inflammatory stimuli is generally considered an adaptive response that restores homeostasis. ³⁴ For instance, ceruloplasmin and metallothionein have antioxidant activity ^{27 24} and can therefore exert a protective function. However, excessive or persistent overexpression of acute-phase proteins can lead to tissue and organ damage. Ceruloplasmin has been found to induce endothelial dysfunction, ⁴⁹ transferrin to be proangiogenic, ⁵⁰ and α₂M to modulate many processes by binding several cytokines and growth factors. ^{51 52} IL-1β could play a pathogenetic role in the development of diabetic retinopathy, since several of the abnormalities occurring in the diabetic retina correspond to known biological effects of this cytokines: increased levels of ICAM-1 and endothelin, ⁴² leukocyte adhesion and breakdown of the blood-tissue barrier, ^{42 53} neuronal death, ⁵⁴ and activation of micro- and macroglial cells. ⁵⁵ Because the activity of this powerful cytokine is regulated at multiple levels, ⁴² further studies must specifically target the action of IL-1β to assess its contribution to diabetic retinopathy. 

 Supported by Juvenile Diabetes Research Foundation Center for Diabetic Retinopathy at the Schepens Eye Research Institute (JDRF Grant 4-2000-650). CG is the recipient of a JDRF Career Development Award (JDRF Grant 2-2000-357).

 Submitted for publication July 22, 2004; revised September 1 and October 5, 2004; accepted October 7, 2004.

 Disclosure: C. Gerhardinger, None; M.B. Costa, None; M.C. Coulombe, None; l. Toth, None; T. Hoehn, None; P. Grosu, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Characterization of Müller cells purified by density gradient centrifugation from the retina of control and diabetic rats. Most of the cells in the Müller-enriched preparations were Müller cells, as indicated by positivity for vimentin (A, B) and GS (D, E), known markers of Müller cells. Preparations from control (A, D) and diabetic (B, E) animals had similar enrichments, indicating that the diabetic milieu did not affect the isolation and purification of Müller cells. Higher magnification of vimentin⁺ (C) and GS⁺ (F) cells. Although after density gradient centrifugation Müller cells became shorter than in the intact retina, they still maintained their characteristic elongated bipolar shape. Slides were counterstained with DAPI (blue) to detect nuclei. Scale bars: (A, B, D, E) 100 μm; (C, F) 50 μm.

Figure 2.

 

Müller cells preparations were free of astrocytes and microglia contamination. (A, B) Müller cell preparations were free of astrocytes, as indicated by the absence of GFAP⁺/vimentin ^– cells. In preparation of both control (A) and diabetic (B) retinas, all the GFAP⁺ cells (arrowhead) were Müller cells, as indicated by positive immunostaining for vimentin. In preparations of control retinas (A), the intensity of GFAP immunostaining was weak, and not all Müller cells were positive (arrows). In contrast, all the diabetic Müller cells were strongly positive for GFAP (B), consistent with the known upregulation of GFAP in Müller cells in diabetic retinopathy. (C–F) Microglial cells (arrows), identified by immunostaining for phosphotyrosine (PY; C, D), Mac1 (E), and isolectin B4 (F) accounted for <1% of the total cells recovered by density gradient centrifugation from both control (C, E) and diabetic (D, F) retinas. (G, H) The few contaminating cells present in the Müller-enriched preparations were NSE⁺ (red, arrowheads) consistent with neural origin. The neuronal contamination was modest—<20% of the cells were NSE⁺, compared with >80% of vimentin⁺ Müller cells (green)—and not different in preparation from control (G) and diabetic (H) retinas. Slides are counterstained with DAPI (blue) to detect nuclei. Bars, 50 μm.

Table 1.

 

Expression Profile of Diabetes-Reactive Retinal Müller Cells

Gene Title/Sequence Description	Reference Sequence Accession No.	Relative Change in Diabetes		Upregulated in Reactive Glia, ‡
Antigen presentation
CD74 (MHC II antigen-associated)	NM_013069	3.3	Inflammatory response	BA (12)
Major histocompatibility locus	M64795	5.1	Inflammatory response	BA (12)
MHC class I (RT1.EC2)	AF074608	2.1	Inflammatory response	BA (12)
MHC class Ib antigen (RT1.CI)	AF025308	2.3	Inflammatory response	BA (12)
RT1 class Ib (Aw2)	NM_012645, †	2.1	Inflammatory response	BA (12)
Cell adhesion
ICAM-1	NM_012967	2.8	Inflammatory response	BA (12)
Osteopontin*/secreted phosphoprotein 1	NM_012881	9.8	Inflammatory response
Scavenger receptor class B, member 1	NM_031541	2.3	Inflammatory response
Cell proliferation
Cyclin D2	D16308, †	2.2
Cyclin G1	NM_012923	−2.9
B-cell translocation gene 2 (PC3)	NM_017259	−4.1
RhoB	NM_022542	−2.0
Complement components
Complement component 1s	NM_138900	10.1	Acute-phase protein	BA (22)
Factor B (properdin, C2)	NM_172222	48.6	Acute-phase protein	BA (12)
Complement component 3	NM_016994, †	4.6	Acute-phase protein	BA (12)
C1 inhibitor (clade G member 1)	AA800318	8.9	Acute-phase protein
Cytokines, growth factors, and receptors
FGF receptor 1	NM_024146	13.6		BA (13)
Galectin-3	NM_031832	10.2	Inflammatory response	RMC (28)
IGF binding protein 2	NM_013122	2.6		BA (23)
VEGF	NM_031836	−2.3
Early response
Activity neurotransmitter-induced early gene 4	AF030089	2.0
Heat shock 27kDa protein 1	NM_031970, †	2.5		BA (13)
Metabolism
3-Hydroxy-3-methylglutaryl CoA lyase	NM_024386	2.0
3-Hydroxy-3-methylglutaryl-Co A synthase 2	NM_173094	3.8
CDP-diacylglycerol synthase	NM_031242, †	2.1
Hexokinase II	D26393	2.0
Palmitoyl-protein thioesterase 2	NM_019367	38.9
Proprotein convertase subtilisin/kexin type 2	NM_012746	2.9
Metal ion binding
Ceruloplasmin	NM_012532	3.1	Acute-phase protein	RMC (27)
Lipocalin 2	NM_130741	8.8	Acute-phase protein
Metallothionein 1	NM_138826, †	2.4	Acute-phase protein	BA (24)
Transferrin	NM_017055, †	5.7	Acute-phase protein	BA (12)
Nclone10 (transferrin precursor)	U31866	9.4
Proteases and phosphatases inhibitors
Alpha-2-macroglobulin	NM_012488, †	2.8	Acute-phase protein	BA (22)
Angiotensinogen	M12112, †	−3.1	Acute-phase protein
Serine protease inhibitor Spin2c (α1-ACT)	NM_031531	15.3	Acute-phase protein	BA (22)
TIMP-1	NM_053819, †	15.2	Acute-phase protein	BA (12)
Signal transduction
Annexin 1 (lipocortin 1)	NM_012904	2.5	Inflammatory response	BA (13)
Endothelin converting enzyme-like 1	NM_021776	2.4
Galanin	NM_033237	−4.7
Phosphodiesterase 1B	NM_022710	−2.0
Protein phosphatase 2, regulatory subunit B	NM_022209	2.4
Protein tyrosine phosphatase, non-receptor 1	NM_012637	2.2
Protein tyrosine phosphatase, non-receptor 16	NM_053769, †	−2.8
Regulator of G-protein signaling 1	AA892750	2.7
Reversion induced LIM gene	NM_017062	9.5
Somatostatin receptor-like protein	NM_031758	5.1
Transcription and translation regulation
C/EBP delta	NM_013154, †	7.9	Acute-phase protein
Cold shock domain protein A	NM_031979	2.1
Eukaryotic translation elongation factor 2	NM_017245, †	−6.2
HES-1	NM_024360, †	−3.1
HES-5	NM_024383	−4.0
Inhibitor of DNA binding 1	NM_012797	−2.0
Kruppel-like factor 4	NM_053713, †	−2.1
NF-κB p105 subunit	L26267	2.5	Inflammatory response
v-jun sarcoma virus 17 oncogene homolog	NM_021835, †	−3.4
Transporters and channels
Annexin V	D42137	2.1
Aquaporin 1	NM_012778	2.4		BA (25)
ATPase, Na+K+ transporting, alpha 2	NM_012505	2.3
ATP-binding, sub-family D, member 3	NM_012804	−2.4
Cyclic nucleotide-gated channel beta 1	NM_031809, †	7.9
Cyclic nucleotide-gated potassium channel 1	NM_053375	3.3
Sodium channel, voltage-gated, type 6, alpha	NM_031686	41.4
Sodium channel, voltage-gated, type 1, beta	NM_017288	3.6
Others
Best5 protein interferon induced	NM_138881	2.3
ER transmembrane protein Dri 42	NM_138905	−6.0
FBR-murine osteosarcoma provirus genome.	X03347	−2.2
GFAP	NM_017009, †	4.8		RMC, BA (26, 12)
Guanylate binding protein 2, interferon-inducible	NM_133624	2.5
Interferon induced mRNA	X61381	6.0
Mox2	NM_031518	2.1	Inflammatory Response
N-ethylmaleimide sensitive factor	NM_021748	−11.0
NG,NG dimethylarginine dimethylaminohydrolase	D86041	4.1
Opsin 1	NM_031015	2.1
Proteasome 26S subunit, ATPase 2	NM_033236	−2.6
Rexo70 (Exo70)	NM_022691	8.1
S-100 related protein, clone 42C	NM_031114	2.5
Transthyretin	NM_012681	2.0	Acute phase protein

*  Cell adhesion and a cytokines.

†  Genes represented in the array with two or more sequence/probe sets showing the same direction of change.

‡  From published reports (references are given in parenthesis); BA, brain astrocytes; RMC, Retinal Müller cell.

Figure 3.

 

α₂M and ceruloplasmin mRNA levels were increased in the retina of diabetic rats. Northern blot of total retina RNA from diabetic (D) and age-matched control (C) rats. Ten micrograms of total RNA isolated from the retina of individual rats was loaded into each lane. (A) Representative autoradiographs. (B) Summary of the densitometric quantitation, mean ± SD; *P = 0.0006, **P = 0.0003 versus control. (C) Regression plot of the expression levels of the two transcripts in control (○) and diabetic (•) retinas. n, number of rats studied; Cp, ceruloplasmin.

Figure 4.

 

α₂M and ceruloplasmin protein levels were increased in the retina of diabetic rats. Western blot analysis of retinal lysates from diabetic (D) and age-matched control (C) rats. Twenty micrograms of total protein obtained from the retina of individual rats was loaded into each lane. (A) Representative immunoblots. (B) Summary of the densitometric quantitation, mean ± SD; *P = 0.01, **P = 0.016 versus control. n, number of rats studied; Cp, ceruloplasmin.

Figure 5.

 

The expression of α₂M was decreased and that of ceruloplasmin was unchanged in the liver of diabetic rats. (A, B) Northern blot analysis of liver RNA from diabetic (D) and age-matched control (C) rats. Ten micrograms of total RNA isolated from the liver of individual rats was loaded into each lane. (A) Representative autoradiographs. (B) Summary of the densitometric quantitation, mean ± SD; *P = 0.0002 versus control. (C, D) Western blot analysis of plasma from diabetic and age-matched control rats. Fifteen micrograms of total plasma proteins was loaded into each lane. (C) Representative immunoblots; (D) Summary of the densitometric quantitation; *P = 0.02 versus control rats. n, number of rats studied; Cp, ceruloplasmin.

Figure 6.

 

Increased expression of IL-1β in the retina of diabetic rats. Summary of the real-time RT-PCR quantitation, mean ± SD; *P = 0.0011 vs controls. n, number of rats studied.

The authors thank Mara Lorenzi and Alessandro Doria for comments and suggestions and Reddy Gali, Michael Ethier, Ping Ma, Eitan Rubin (Harvard Bauer Center for Genomics Research) and John Barton, Douglas Bassett, Lee Weng, and Bernadette Boutte (Rosetta Biosoftware) for technical assistance.