Received 4 June 2010 | Accepted 7 October 2010 | Published 15 October 2010
1Department of Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA; 2Department of Ophthalmology, 2nd Xiangya Hospital, Central South University, Changsha, PR China; 3Department of Veterinary Clinical Sciences, College of Veterinary Medicine, Iowa State University, Ames, IA; 4Department of Ophthalmology, Oregon Health and Science University, Portland, OR; 5Iowa City Veterans Administration Center for Prevention and Treatment of Visual Loss, Iowa City, IA
Correspondence to: Markus H. Kuehn, Ph.D., Department of Ophthalmology and Visual Sciences, University of Iowa Hospitals and Clinics, 4120 MERF, Iowa City, IA, 52242; Phone: (319) 335 9565; FAX: (319) 335 6641; email: Markus-Kuehn@uiowa.edu
Purpose: The pathophysiological events that occur in advanced glaucoma are not well characterized. The principal purpose of this study is to characterize the gene expression changes that occur in advanced glaucoma.
Methods: Retinal RNA was obtained from canine eyes with advanced glaucoma as well as from healthy eyes. Global gene expression patterns were determined using oligonucleotide microarrays and confirmed by real-time PCR. The presence of tumor necrosis factor (TNF) and its receptors was evaluated by immunolabeling. Finally, we evaluated the presence of serum autoantibodies directed against retinal epitopes using western blot analyses.
Results: We identified over 500 genes with statistically significant changes in expression level in the glaucomatous retina. Decreased expression levels were detected for large number of functional groups, including synapse and synaptic transmission, cell adhesion, and calcium metabolism. Many of the molecules with decreased expression levels have been previously shown to be components of retinal ganglion cells. Genes with elevated expression in glaucoma are largely associated with inflammation, such as antigen presentation, protein degradation, and innate immunity. In contrast, expression of many other pro-inflammatory genes, such as interferons or interleukins, was not detected at abnormal levels.
Conclusions: This study characterizes the molecular events that occur in the canine retina with advanced glaucoma. Our data suggest that in the dog this stage of the disease is accompanied by pronounced retinal neuroinflammation.
Glaucoma is among the leading causes of human blindness world wide and continues to pose a clinical challenge yet the sequence of the pathophysiological events that accompany and lead to retinal ganglion cell (RGC) death, the ultimate cause of vision loss in glaucoma, remains incompletely understood. Dogs frequently develop glaucoma spontaneously with advanced age and represent an attractive model for glaucoma research due to the size of their eye, the chronic nature of the disease, and the pathophysiological similarities to glaucoma in humans. In this species ocular exams such as gonioscopy, fundus photography, intraocular pressure (IOP) measurements, slitlamp exams, and indirect ophthalmoscopy are routinely performed and even advanced diagnostic methodologies such as optical coherence tomography, ultrasound, or pattern electroretinogram (pERG) recordings can be conducted [1,2].
An important step toward a better understanding of the pathophysiology of glaucoma is to determine the retinal gene expression profile during the progression of the disease. Several excellent studies describing changes in the global gene expression pattern in the retina and optic nerve of rodent models of glaucoma have been published previously [3-7].
Here, we examine the gene expression pattern and immune response changes of the retina in healthy eyes and in eyes of dogs with spontaneous glaucoma. Glaucomatous damage in these eyes was typically advanced, allowing insight into the cellular events that occur during late stage glaucoma.
All studies were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Visual Research and are approved by the Iowa State University Committee on Animal Care. Before inclusion in the study all animals were evaluated by a veterinary ophthalmologist (SDG) to rule out the presence of non-related ocular disease. Examinations included slit lamp biomicroscopy, intraocular pressure measurements, indirect ophthalmoscopy, and gonioscopy.
Glaucoma eyes (n=9) were derived from the patient population of the Iowa State University College of Veterinary Medicine Clinics and enucleations were performed with the animal owner’s consent to ease pain and suffering. Retinal samples from total of five glaucomatous eyes were used for microarray analysis, while retinal samples of all nine animals were used for PCR analysis. All glaucoma donors were diagnosed with primary glaucoma based on abnormal gonioscopy examination, elevated IOP and absence of other ocular disease. IOP of affected eyes ranged from 30 to 48 mmHg. None of the glaucoma animals used in this study received surgical treatment, but all of them were treated with IOP lowering topical medications.
In addition, eyes from five control dogs without ophthalmic findings were used. These animals were euthanized for reasons unrelated to this study (see Table 1).
Eyes were dissected and preserved in RNAlater (Ambion, Austin, TX) immediately after enucleation. Samples were then stored at −80 °C until RNA extraction. The neural retina was isolated and total RNA was extracted from the tissue using Qiagen RNeasy minipreps. Samples were treated with RNase free DNase and the integrity of the RNA was evaluated through analysis with a Bioanalyzer (Agilent Technologies, Foster City, CA). RNA was amplified using a T7 RNA polymerase based approach and hybridized to Affymetrix Canine genome 2.0 gene chips following standard protocols.
Raw data obtained were normalized using the RMA algorithm. Normalized data were log2-transformed and filtered to remove non-expressed genes from the data set. For the purpose of this study, expressed genes are defined as those with corresponding probesets displaying log-expression values above 7.0 in at least 2 samples (either controls or affected). The remaining probesets were analyzed to identify significant expression changes using the Wilcoxon unpaired rank sum test and the significance analysis for microarray (SAM; Version 3.0; Microsoft Excel Add-In; Stanford University, Palo Alto, CA). Data were analyzed four times using 200 permutations and different seeds values for the random number generator. The delta value was set at 0.53 and a minimum twofold expression change was required. Only genes identified as differentially expressed in all four analyses are presented in this manuscript.
The Database for Annotation, Visualization and Integrated Discovery (DAVID) was used to obtain current Entrez Gene IDs (November, 2007) and the corresponding gene names are used throughout this manuscript. Data have been deposited to the NCBI Gene Expression Omnibus (GEO) and are available under the accession number GSE21879.
Total retinal RNA was extracted from retinas of nine glaucomatous and five control eyes and treated with DNase. These included all eyes used for the microarray studies as well as four additional glaucoma samples. Only one eye from each animal was used. From each sample 500 ng RNA was reverse transcribed in a random primed reaction and 5 ng was used as template in each PCR reaction. DNA amplification was monitored using the dye SYBR Green (Perkin Elmer, Waltham, MA). Data from each sample was obtained in triplicate; amplification controls included wells containing genomic DNA only and those containing no target (water controls). Transcript levels were determined based upon standard curves for each primer pair (Table 2). Melt curve analyses were performed following each amplification reaction to ascertain the absence of nonspecific amplification products. Expression values obtained were normalized to transcript levels for ubiquitin C (UBC).
Tissue samples for immunohistochemistry were fixed in 4% paraformaldehyde, embedded in paraffin and sectioned to a thickness of 2 microns. Sections were deparaffinized with heat and xylene and rehydrated by serial rinses in decreasing concentrations of ethanol. Endogenous peroxidase activity was quenched by incubation with 3% H2O2 for 10 min. Following rinses in potassium phosphate-buffered saline (KPBS), cells were incubated in blocking solution containing 5% normal donkey serum (NDS, 017–000–121; Jackson ImmunoResearch, West Grove, PA), 0.1% BSA (BSA, A9647; Sigma, St. Louis, MO), and 0.04% Triton X-100 for 2 h to eliminate non-specific antibody labeling. Tissue was then incubated in primary polyclonal antibodies overnight at room temperature including: anti-glial fibrillary acid protein (1:2,000; Dako, Carpinteria, CA); anti-CD3 (1:75; Dako); anti-TNF-alpha (1:50; Abcam, Cambridge, MA), anti-TNF-R1 (1:500; Abcam) and anti-TNF R2 (1:25; Abcam). Antibody binding was visualized through incubation with appropriate biotinylated secondary antibodies followed by incubation with avidin-peroxidase conjugate and 3,3′-diaminobenzidine (DAB) with nickel sulfate. Care was taken to maintain identical development times in those cases where labeling intensity was measured. Sections were dehydrated through a graded ethanol series, cleared with xylene, and coverslipped. Negative controls were performed in parallel and included the omission of the primary or secondary antibody. Images of anti-GFAP and anti-CD3 were taken withan Axioplan 2 microscope (Carl Zeiss MicroImaging, Inc., Thornwood, NY), equipped with a color camera (AxioCam MRc; Carl Zeiss Meditec, Inc.).
Four images of the central and peripheral retina were taken for each section using a Nikon Microphot Microscope (Nikon Inc. Garden City, NY) and a 40× oil immersion objective. Central retinal images were obtained within two microscope fields of the optic nerve. Peripheral retinal images were obtained 7–8 microscope fields away from the optic nerve. The microscope settings for tissue stained with a particular antibody were left consistent to eliminate variation from one sample to the next. A blank image that did not contain any tissue was obtained to correct for any slight variations in the slides. Metamorph image analysis software (Ver. 7; Molecular Devices, Sunnyvale CA) was used to quantify the percentage of the retina that was immunoreactive for each antibody. Blank images from the data set were used to correct each slide to account for differences in light illumination. A threshold two standard deviations below the median staining intensity for each group stained with an antibody was determined, the immunoreactivity was pseudocolored and the fraction of the retina labeled was calculated using Metamorph. The immunoreactivity of all retinal layers combined was quantified. Additionally, the combined ganglion cell layer and the inner plexiform layers were analyzed independently. Morphometric data were statistically analyzed using Students t-test and Graphpad Prism (ver. 4.0 for Macintosh; Graphpad Software, La Jolla, CA).
Initial screening of dog sera was performed using dog retinal proteins that were extracted from a dog retina with 2% octyl glucoside in phosphate/saline buffer (PBS) with proteolytic inhibitors, pH 7.2. The proteins were separated by SDS-gel electrophoresis on a 10% gel and transferred to an Immobilon membrane (Millipore, Bedford, MA). Individual strips containing 10 μg retinal proteins were blocked with 10% normal goat serum, 1% BSA in PBS for 1 h, and then incubated with 1:100 diluted dog serum (1 h) followed by a 1 h incubation with 1:1,000 diluted anti-dog IgG (H and L chain) conjugated to alkaline phosphatase (Sigma, St. Louis, MO). Color reaction was developed by adding the phosphatase substrate until dark bands, appeared in comparison to the positive controls (anti-recoverin antibody diluted 1:50,000, anti-enolase antibody diluted 1:2,000, anti-crystalline-μ antibodies diluted 1:1,000). Western Blots were run and examined in a masked fashion. As a negative control, serum was omitted and only a secondary antibody was applied.
All animals (Table 1) received an ocular exam (slitlamp biomicroscopy, intraocular pressure measurements, indirect ophthalmoscopy, and gonioscopy) to rule out the presence of the non-related ocular disease before inclusion in the study. In the majority of canine breeds glaucoma develops through a gradual narrowing and eventual closure of the angle resulting in elevation of IOP and the development of functional deficits. All glaucoma animals used in this study presented with closed irido-corneal angles during gonioscopy examination.
Histological evaluation of the retina and optic nerve head (ONH) of glaucomatous canine eyes demonstrated that the morphological findings in this species closely resemble those observed in human eyes. As in human glaucoma, the appearance of the ONH in the glaucomatous eye is often characterized by extensive cupping, reorganization of the extracellular matrix and gliotic changes . In the retina of dogs with advanced glaucoma a general thinning of the peripheral retina, associated with extensive loss of cells in the ganglion cell and inner nuclear layers, is evident (Figure 1). Although some loss of nuclei in the outer nuclear layer is discernable, the integrity and organization of the photoreceptor cells is largely maintained. Immunohistochemical analyses demonstrated enhanced expression of glial fibrillary acidic protein (GFAP) in the glaucomatous retina (Figure 1B). GFAP can also be detected in the nerve fiber layer of normal eyes but at much reduced levels (Figure 1A).
To identify retinal gene expression changes associated with advanced glaucoma, we analyzed the global gene expression patterns in five eyes derived from dogs with glaucoma and from five control animals. As observed in similar analyses of experimental models of glaucoma [3-7] our analyses indicate significant changes in the transcript levels of a large number of genes. Under the stringent parameters used here (predicted false discovery rate=0.3%) 1,111 probe sets were identified as being differentially expressed between the two groups. Of these, signals from 586 probe sets are significantly reduced in the glaucomatous retina. Further exclusion of probe sets corresponding to unannotated genes results in 362 probe sets representing 275 unique genes (Table 3). In addition, 525 probe sets are detected at increased transcript levels in the glaucomatous retina. Of these, 486 probe sets remain after removal of unannotated genes, representing 303 unique genes (Table 4).
Interestingly a plot of the average detected gene expression changes versus the p-value associated with each measurement reveals a marked asymmetry between transcripts with reduced expression and those with elevated expression (Figure 2). While transcripts detected at lower levels in the glaucomatous retina show a good correlation between the degree of expression change and associated p-value, genes with higher average expression in glaucoma are frequently associated with non-significant p-values. This is true even for genes with comparatively large changes in average expression levels. These higher p-values are typically the result of a large standard deviation between the values in the glaucoma group. These data indicate that the factors causing lower gene expression are largely similar between the individuals evaluated, whereas glaucomatous events resulting in elevated transcript levels are much less uniform.
To test the accuracy and reproducibility of the Gene chip data we also sought to confirm the observed gene expression differences between glaucomatous and normal retinas using quantitative PCR. Clearly, verification of all gene changes is impractical and consequently only a subset of genes was selected. In the selection of these genes we avoided those with already well described glaucoma related expression changes, such as GFAP and complement components [8-12], and focused instead on less well characterized genes (Figure 3). Our RT–PCR data indicate that expression levels vary considerably among the nine affected animals evaluated for this part of the study. However, statistically significantly elevated (p<0.05 by t-test) levels of colony stimulating factor 1 receptor (CSF1R), glia maturation factor gamma (GMFG), neuronal cell adhesion molecule (NRCAM), neurotrophin 3 (NTF3), Calgizzarin (S100A1), TNF receptor 1 (TNFRSF1A) and TNF receptor 14 (TNFRSF14) were confirmed.
Our molecular findings, in accord with previously published studies, suggest that modulation of TNF alpha and its receptors is correlated to the development of glaucoma. Morphometric analyses performed on retinas from of healthy and glaucomatous dogs using anti-TNF antibodies demonstrated increased immunereactivity in the glaucomatous retina (Figure 4A,B). Quantitation of the observed signal demonstrated a significant increase in TNF labeling both in the central and the peripheral retina (Figure 5A).
Similar changes were observed for TNF receptor 2 (TNFR2). Immunoreactivity for this molecule is significantly increased in the peripheral and central retina of glaucomatous eyes when compared to normal eyes (Figure 4E,F and Figure 5C). In contrast, our data suggest that expression levels for TNF receptor 1 (TNFR1) remain unchanged in the glaucomatous retina (Figure 4C,D and Figure 5B). The finding that overall TNFR1 immunoreactivity remains relatively unchanged contrasts with our data suggesting elevated mRNA levels of the TNFR1 gene (TNFRSF1A) in glaucoma.
The pronounced appearance of MHC class I and inflammation related gene transcripts could be interpreted as indicative of leukocyte infiltration into the canine glaucomatous retina. Immunohistochemical analyses using sagittal sections of several canine glaucomatous retinas and antibodies directed against the T- and B- cell antigens CD3 and CD79 did not reveal immunopositive cells in the evaluated samples (Figure 6A). CD3 positive cells could readily be detected in a tissue sample of canine optic neuritis, used here as a positive control (Figure 6B), and were occasionally observed in the perivascular space in glaucoma retinas. These findings suggest that leukocyte infiltration into the retina of glaucomatous dogs it is, at most, a rare event.
Despite the apparent paucity of CD3/CD79 positive cells, it was conceivable that canine glaucoma may result in the formation of autoantibodies directed against retinal antigens. To determine if this does indeed occur, we incubated western blots of retinal protein extracts with serum obtained from seven dogs with moderate to advanced glaucoma and ten healthy control animals (Figure 7). While minor immune reactivity can be observed in all samples, the majority of the serum obtained from dogs with glaucoma reacts with retinal proteins more vigorously than that obtained from healthy control dogs. It is noteworthy that the samples with the most pronounced immunoreactivity were derived from dogs with advanced glaucoma. A general pattern of labeled bands is not apparent, rather it appears that each individual displays immunereactivity to a specific subset of molecules. In the majority of cases the labeled molecules do not react with purified Recoverin, α-enolase, or crystallins, suggesting that these proteins are not major autoantigens in dogs with glaucoma.
In this study we characterize the molecular events that are associated with advanced glaucomatous degeneration of the retina in dogs. These data are the first to describe changes in the gene expression pattern and immunological consequences in a species with large eyes and spontaneously occurring chronic elevation of IOP. One shortcoming of this study is that the control and glaucoma groups were are not closely matched in age and are derived from a variety of breeds. These factors will likely influence the expression levels of some genes, but the overall similarity in canine retinal morphology and function suggest that these differences between the two groups will be relatively minor.
In accordance with previous studies, our data demonstrate that advanced glaucomatous retinopathy is associated with pronounced changes in the retinal gene expression pattern. Among genes with elevated expression levels in glaucoma, a preponderance of genes mediating various aspects of a neuroinflammatory response was striking. Prominent functional categories of genes with elevated expression in glaucoma include antigen presentation, complement activation, lysosomal and proteasome activity, and acute phase proteins. In addition, numerous genes with a function in apoptosis and inflammation signaling are more abundant in retinas with glaucoma than in those without. It must be noted that many of the identified genes with altered expression levels are associated with several biologic function, thus inclusion in one functional category does not exclude a gene’s involvement in additional molecular pathways.
Our analyses also indicate that glaucoma leads to lower expression levels for a large number of genes, including various neurofilaments, synucleins, and neuregulin 1. Functionally, no particular group of molecules appears to be disproportionally affected. Rather, we detected decreased expression levels of a large number of functional categories, including neuronal development and maintenance, cell adhesion, calcium transport and binding, transcriptional regulation, synaptic transmission, and many others. Reduced mRNA levels of individual genes could result from transcriptional control mechanisms. Alternatively, reduced levels of a specific mRNA in a tissue comprised from several cell types may be related to loss of the cell type that synthesizes the majority of this mRNA. Genes such as STMN2, NEFH, NRG1, ELAV2, and SNCG appear to be primarily expressed by RGC [13-15] and the decrease of their transcript levels is conceivably due to the loss of RGC and, perhaps, amacrine cells [16,17] in the glaucomatous retina and may not represent transcriptional regulation.
Decreased expression levels were also detected for several photoreceptor cell specific genes in the glaucomatous retina. Whether photoreceptor cell loss or functional decline is a feature of advanced glaucoma has been extensively debated [18-21]. In this study significantly decreased expression levels are only observed for a small number of photoreceptor cell specific genes, while most of photoreceptor cell associated transcripts are present at similar or only mildly reduced levels (i.e., less than the twofold cut off value employed) suggesting only a minor effect of glaucoma on the transcriptional activity of photoreceptor cells in the dog.
The systematic comparison of findings in this study to those presented previously by other investigators is not straightforward due to the different formats of gene arrays used, selection of genes represented on each array, incomplete identification of orthologs between animal species, and the stage of disease investigated. Previous studies had observed a striking decrease in the expression levels for several crystallin genes in rodent models of glaucoma [3-5,22]. Although probes for these molecules are present on the gene chips used in this study, our data do not mirror these findings. Reduced synthesis of retinal crystallins could represent a rodent specific response to elevated IOP. However, several studies have demonstrated that expression levels for various crystallin genes decrease quickly after induction of ocular hypertension, but return to normal levels 2 to 3 weeks later [3,22]. Thus, decreased expression of crystallins may reflect an early, transient event in glaucoma pathology.
Our findings largely agree with previously published reports that clearly demonstrated the involvement of the complement system in the pathophysiology of glaucoma and noted the expression of other inflammation-related molecules by retinal cells [3-5,8,23,24]. While some studies have suggested that a breakdown of the blood brain barrier may occur in the eyes of dogs with severe glaucoma [25,26] our own immunohistochemical data suggest that CD3 or CD79 positive cells in canine glaucoma occur, at most, infrequently. In contrast, most hybridization signals of inflammation associated transcripts are quite strong, indicating that these gene products are abundant in the retina of dogs with glaucoma and thus likely originate from a cell type that is relatively common in this tissue. Micro- and macroglia constitute a sizable share of the retinal cell population and several studies have demonstrated that glia are capable of transcribing several of the inflammation-associated genes detected at elevated levels in this study, e.g., AIF1, CXCL12, and MHC class 2 molecules [27-30]. Retinal glia may also contribute to the observed synthesis of inflammatory molecules, e.g., both CXCL10 and CXCL16 are produced by reactive astrocytes [31,32].
An important function of glia is to phagocytose foreign particles or cell debris, a process activated by the LGALS3 gene . Macrophages that have ingested apoptotic cells appear to inhibit the production of proinflammatory molecules  and is possible that retinal microglia are capable of mediating a similar repression of some proinflammatory stimuli after phagocytosis of RGC debris. In that regard, the increased expression of the TREM2 receptor in the glaucomatous retina is noteworthy, as its interaction with its ligands appears to mediate inhibition of inflammation and stimulation of antigen presentation .
We propose that the massive retinal inflammation and antigen presentation responses described here represent late events in the pathophysiology of glaucoma. Increased expression of individual inflammation-related genes has been reported previously in studies that evaluated earlier stages of glaucomatous damage. While this neuroinflammatory response was not observed to the degree seen here data presented by Ahmed et al. suggest progressively increasing expression levels of several immune response molecules during the development of the disease .
Our data suggest that the anti-inflammatory mechanisms that protect the retina in early glaucoma can eventually fail, leading to the development of autoantibodies in many glaucomatous dogs. Although the number of examined samples is relatively small, the fact that all analyzed serum samples from dogs with glaucoma display immunereactivity against distinct antigens, suggests that glaucoma is not initiated by an immune response to specific retinal antigens. Rather it appears likely that the variety of immunoreactive molecules is due to exposure of numerous epitopes during the rapid neuronal cell death of the retina. Consequently the development of anti-retinal antibodies may be secondary to the degeneration of RGC. Never-the-less, exposure to serum antibodies directed against retinal antigens has been shown to induce RGC loss that resembles glaucoma [36,37]. Thus it is conceivable that once an immune response has occurred, it will further accelerate vision loss in an IOP independent mechanism.
The relevance of our findings to human disease remains to be determined. Glaucoma in dogs often presents with comparatively high intraocular pressure, introducing the possibility that ischemic events to the inner retina contribute to the pathophysiology of glaucoma in this species more than they do in humans. It is well documented that retinal ischemia leads to increased vascular permeability (reviewed in ) and given the chronic nature of IOP elevation in glaucoma it is conceivable that even a rare presence of T-cells could over time result in an antigenic response in this species. The intraocular pressure in human patients receiving ophthalmic care is typically much lower than that in glaucomatous dogs and leukocyte infiltration of the retina does not appear to be a feature of human glaucoma. Yet several reports have suggested the presence of autoantibodies against retinal epitopes in the serum of some glaucoma patients [39-41]. Whether those glaucoma patients with detectable autoimmune titers to the retina share a genetic predisposition toward a vigorous immune response or exhibit other unifying features has, to our knowledge, not been thoroughly investigated.
Regardless of the potential role of an adaptive immune response in glaucoma, data from numerous studies indicate that neuroinflammatory events do occur as a consequence of neuronal degeneration in glaucoma. Consequently it is conceivable that therapeutic modulation of the neuroinflammatory response may be a beneficial augmentation to IOP lowering therapy in glaucoma.
This work was supported by EY019485 (M.H.K.), EY1305 (G.A.), Rehabilitation Research and Development Service Grant C3919R (R.H.K., S.D.G., M.H.K.) from the Department of Veterans Affairs, Iowa VA Center of Excellence for the Prevention and Treatment of Vision Loss, an unrestricted grant from Research to Prevent Blindness to the Department of Ophthalmology and Visual Sciences, and a resident training grant from the American College of Veterinary Ophthalmologists.