|Molecular Vision 2007;
Received 27 March 2007 | Accepted 14 August 2007 | Published 10 September 2007
Comparison of human RPE gene expression in macula and periphery highlights potential topographic differences in Bruch's membrane
Simone S. van Soest,1 Gerard. M.J. de Wit,1 Anke
H.W. Essing,1 Jacoline B. ten Brink,1 Willem
Kamphuis,1 Paulus T.V.M. de Jong,1,2,3
Arthur A.B. Bergen1,4
1Department of Clinical and Molecular Ophthalmogenetics, The Netherlands Institute for Neuroscience, an institute of the Royal Netherlands Academy of Arts and Sciences (KNAW), Meibergdreef, Amsterdam, 2Department of Ophthalmology, AMC Amsterdam, 3Department of Epidemiology and Biostatistics, Erasmus Medical Center Rotterdam, and 4Department of Clinical Genetics, AMC Amsterdam, The Netherlands
Correspondence to: Dr. A. A. B. Bergen, The Netherlands Institute for Neuroscience, an institute of the Royal Netherlands Academy of Arts and Sciences (KNAW), Meibergdreef 47, 1105 BA Amsterdam, The Netherlands; Phone: (+31) 20 566 6101; FAX: (+31) 20 566 6121; email: A.Bergen@nin.knaw.nl
Purpose: To describe gene expression differences between healthy, young human retinal pigment epithelium (RPE) cells from the macular area and RPE cells from two locations in the retinal periphery.
Methods: RPE cells from six human donor eyes, ages 17-36, without histopathological abnormalities, were dissected by laser and isolated from cryosections. Total RNA was isolated, amplified, and hybridized to a custom made oligonucleotide array containing 22,000 genes. Bioinformatic analysis was carried out using the computer programs Rosetta Resolver and the webtools EASE/David and GoStat. Confirmatory real time PCR (RT-PCR) and immunohistochemistry were performed according to standard protocols.
Results: Microarray and statistical analysis yielded 438 genes that were differentially expressed between macular RPE, and at least one out of two peripheral RPE locations. Out of these genes, 33 that showed fold changes of four, or higher, were selected for RT-PCR confirmation. For 17 genes (51%), a significant differential expression was found, while 11 additional genes (33%) showed a similar trend. Immuno-staining of one target (WFDC1) confirmed its differential expression on the protein level. Functional annotation and overrepresentation analysis independently defined extracellular matrix (ECM) genes as a statistically overrepresented class of genes in our RPE dataset. In total, 33 ECM genes were differentially expressed between macular and peripheral RPE regions. A subset of proteins corresponding to these genes is known to be present in Bruch's membrane.
Conclusions: Our data showed that consistent topographical gene expression differences in the human RPE constitute around 1-5% of the RPE transcriptome. These changes may underlie topographical differences in RPE physiology, and pathology and may reflect local differences in the molecular composition and turnover of Bruch's membrane.
The retinal pigment epithelium (RPE) is a pigmented neuro-epithelium, that forms a functional unit with photoreceptor cells, interphotoreceptor matrix and Bruch's membrane (BM). The RPE is a cellular monolayer that acts as a selective barrier between the choriocapillaris and the neuro-retina, and is essential for maintaining normal vision .
The RPE has a wide range of functions including transport of nutrients, phagocytosis of photoreceptor discs, recycling of photoreceptor material, isomerization of rod and cone visual pigments, and protection against oxidative stress [1-4]. On the apical side, the pseudopodial processes of the RPE tightly surround the photoreceptors. The RPE and photoreceptors are separated by the interphotoreceptor matrix. The tight interaction between RPE and photoreceptors is essential for the survival and excitability of photoreceptors. On the basal side, the RPE borders on BM, which separates the RPE from the fenestrated endothelium of the choriocapillaris. The RPE is part of the outer blood-retina barrier.
The high number of ophthalmic diseases that involve the RPE illustrates the crucial role of this tissue in visual function. These disorders include retinitis pigmentosa, several juvenile macular degenerations, and age-related macular degeneration (AMD). Several of these disorders show a distinct topographical pattern of pathology in the retina. For example, neovascular membranes in wet AMD tend to grow centripetally toward the center of the macula, the fovea . While there is currently no satisfying explanation for these phenomena, it is conceivable that intrinsic topographical differences in the RPE and adjacent tissues correlate with local disease predisposition. Indeed, Chong and colleagues recently found that BM has a three to six times thinner elastic layer in the macula compared with BM in the retinal periphery, which apparently correlates with the location of AMD lesions . Topographical RPE differences may be present at an early age, but could possibly disappear or become more explicit during aging. After all, in the aging macula, multiple changes occur. Among others, aging foveal RPE cells loose their normal hexagonal shape, an increased number of RPE and photoreceptor cells die via apoptosis, and multiple structural changes occur in or near BM . To determine if, and which genes showed consistent topographical gene expression differences in the human RPE, we performed a large scale microarray study using laser dissected healthy RPE cells from the fovea, the macular border and the retinal periphery.
The studies on human eye tissue were carried out in accordance with the Declaration of Helsinki on the use of human material for research. The Corneabank Amsterdam provided human eyes. Available medical history of the donors revealed no pre-existing disorders, prolonged medicine use or treatment, or other prolonged agonal states that could possibly influence RPE gene expression or mRNA quality , and thus the nature of this study.
Tissue processing and sectioning
Eyes were enucleated and transported to the Amsterdam eye bank in a moist chamber on ice. Upon receipt and removal of the corneal button for transplantation, the fundus was, within 24 h post-mortem, visually inspected using a Zeiss light preparation microscope (Carl Zeiss, Sliedrecht, The Netherlands) and photographed. The intact eighteen bulbi used in this study were snap-frozen in isopentane (-80 °C) and stored at -80 °C prior to further processing. In total, six eyes from donors (<40 years old), were used for microarray and RT-PCR studies. Ten eyes were used to create a common reference hybridization sample for current and future microarray studies (common reference design), and two eyes were used for immunohistochemistry (Table 1).
From the six frozen bulbi selected for microarray and RT-PCR studies, a macular tissue fragment (M) of 16 mm2, centered on the fovea, was excised (Figure 1). In addition, peripheral fragments of similar size from both the paramacular area (P1; 6-8 mm distant from the center of the macula) and the postequatorial periphery (P2; outside the vascular arcade) were isolated. Throughout these tissue fragments, alternating series of eight 6 mm and ten 20 mm cryosections were cut and mounted, respectively, on poly-L-lysine coated slides and on slides covered with 1.35 mm PEN-membrane (both PALM, Microlaser Technologies AG, Bernried, Germany). At least 10 sections, no more than 220 mm apart through the tissue fragments, were stained with periodic acid Schiff and evaluated by at least two independent observers for the presence of gross abnormalities or drusen.
Ten additional donor eyes (ages 42-80 years old, average age 60 years) were processed for direct isolation of RPE/choroid, to be used as the common reference sample for our microarray studies (not shown in Table 1). After removal of the anterior segment, vitreous and neural retina from these eyes, the RPE/choroid was gently detached from the eyecup, collected in TRIZOL and processed for RNA isolation according to manufacturers protocol (Invitrogen, Carlsbad, CA).
Finally, retinal cryosections from two eyes (age <40 years) were used for immunohistochemistry using standard technology.
Laser dissection microscopy
Unstained 20 mm sections were quickly dehydrated and briefly air-dried. Up to 20,000 RPE cells were isolated using a PALM laser dissection microscope (P.A.L.M. Microlaser Technologies AG). Total RNA isolation was carried out with RNeasy Mini (Qiagen Benelux, Venlo, The Netherlands).
RNA isolation and amplification
Total RNA was amplified once using the MessageAmp aRNA Kit in the presence of 5-(3-aminoallyl)-UTP (Ambion, Austin, TX). The aRNA yield and quality were assessed using a 2100 Agilent BioAnalyzer (Agilent Technologies, Santa Clara, CA). The aRNA was coupled to Cy3 and Cy5 monoreactive dyes (Amersham, Buckinghamshire, UK), in 0.05 M Na2CO3 (pH 9.0) during 1 h. Free dye was quenched by the addition of 0.44 vol. 4 M hydroxylamine and removed by filtration over a Chromaspin-30 column (Clontech, Mountain view, CA) using 10 mg glycogen as a carrier. Incorporation efficiency and yield were determined on a microscale spectrophotometer (Nanodrop; Agilent Technologies).
We designed a 22K custom oligonucleotide microarray enriched for sequences expressed in RPE and neural retina, based on information derived from public databases and on experimental data generated through hybridizations with total RNA from RPE/choroid, neural retina, and brain (results not shown). Design of the oligonucleotides and microarray manufacture was performed by Agilent Technologies. A common reference type of design was employed, and dye swaps were performed for each sample. Hybridizations and washings were carried out according to the protocol supplied (Agilent Technologies). Scanning was performed with the GeneArray scanner (Agilent Technologies).
Analysis of microarray results
Images were extracted using Feature Extraction software (Agilent Technologies). Raw data were normalized using linear and Lowess normalization (Agilent Technologies). Resulting data were submitted and analyzed using Rosetta Resolver (RR) gene expression data analysis system (Rosetta Inpharmatics, Seattle, WA). Genes showing significant differences in expression were identified by error-weighted ANOVA t-test analysis on three sets of hybridizations (M: n=4; P1: n=3; P2: n=3), p<0.01. Technical replicates were precombined before the statistical analyses were carried out on the resulting groups. The prior error taken into account by RR included propagated error and scatter error derived from the feature extraction data. When the number of replicates was small, this input provided additional information to yield more reliable variance estimates, and minimized both false-positive and false-negative rates. Overrepresentation analyses were carried out with EASE/DAVID2006  and GOstat  using Benjamini-Hochberg FDR correction.
Direct comparisons between M, P1, and P2 were analyzed by error-weighted ANOVA t-test with Benjamini-Hochberg correction for multiple testing.
Quantitative real-time polymerase chain reaction analysis
Template cDNA for the real-time PCR was generated by reverse transcription of 200 ng aRNA with Superscript III (Invitrogen). Real-time qPCR reactions were carried out in a 20 ml volume using qPCR Core Kit Sybr Green I (Eurogentec, Seraing, Belgium). Amplified product levels were detected by real time monitoring of SYBR Green I dye fluorescence in the ABI Prism 7300 (Applied Biosystems). Data analysis of the real-time PCR was carried out according to methods described earlier [11,12]. All Ct (threshold cycle) values were converted to absolute amounts of cDNA present and the geomean was calculated from the reference genes. The normalized expression level was calculated by dividing the absolute amount of the samples by a geomean. The non-parametric Mann-Whitney test was used to calculate the statistical significance in expression levels between the macular and peripheral RPE samples. A p-value <0.05 was considered significant. RT-PCR reference primers are presented in Table 2; experimental primers are available on request.
WFDC1/ps20 was localized using ps20ab-1 (a kind gift of Dr D.R. Rowley), an affinity purified and previously characterized rabbit polyclonal antibody . Air-dried cryosections were briefly fixed (PFA 4%). Sections to be stained with ps20ab-1 were pre-treated with 4 M guanidine HCL, 50 mM DTT, 20 mM Tris (pH 8.0) for 15 min followed by a wash in 20 mM Tris (pH 8.0) and treatment with 100 mM iodoacetamide in the dark for 15 min to enhance antigenicity. Staining was performed using an indirect immunoperoxidase procedure. Optimal dilution (1:25) was determined by titration starting from the 1:4 dilution used in the original paper, to dilutions over 1:100 . Sections were incubated with ps20ab-1 (1:25) for at least 16 h at 4 °C. Sections were washed, then incubated with goat anti-rabbit IgG conjugated to horseradish peroxidase (PowerVision Poly-HRP-Anti Rb/Ms, Immuno Vision Technologies, Brisbane, CA) for 1 h. Sections were washed and staining was performed using 10 mg/ml 3,3-amino-9-ethyl carbazole (AEC) in N,N-dimethylformamine (Sigma) + 0.01% H2O2 . Counterstaining was performed with hematoxylin. Primary antibody was omitted in control sections. For (semi-quantitative) measurements of differences in staining intensity between different retinal regions, sections from these regions were combined on slides in order to ensure equal staining conditions. Intensity of the staining was then determined by measuring optical density (Image-Pro Plus software, Media Cybernetics, Bethesda, MD).
Donors, tissue processing, and RNA quality
Relevant data of the six eyes selected for this study are presented in Table 1. Inspection of the fundus revealed no overt abnormalities or signs of posterior segment disease. Upon light microscopy and histological analysis, no obvious abnormalities, such as thickening of BM, presence of pigmentary changes, presence of basal laminar deposits or drusen, were observed. The M, P1, and P2 tissue fragments (Figure 1) were cut from six donor eyes and used for cryo-sectioning. From each region, up to 20,000 RPE cells per eye were laser dissected. Total RNA was isolated and linearly amplified. aRNA showed size distributions from 300 to 2500 bp with a peak at 800-1000 bp, with similar results in all samples.
After hybridization of the 22K oligonucleotide arrays with RPE RNA samples and feature extraction, ANOVA analysis was used to compare the expression profiles of the M, P1 and P2 regions. For 1,155 genes, a significant (p<0.01) difference among these groups was found. (For a complete list, see Appendix 1). Pairwise comparisons between M, P1, and P2, including Benjamini-Hochberg correction for multiple testing, resulted in 319 genes differentially expressed genes between M and P1, 167 between M and P2, and 20 between P1 and P2. Eliminating the redundancy between these sets yielded 438 genes up- or down regulated between macula and periphery (see Appendix 2).
Confirmatory real-time polymerase chain reaction analysis and immunohistochemistry
RT-PCR was carried out for at least 33 genes showing fold changes of four or higher between M and either P1 or P2 in the microarray analysis (Table 3). Statistically significant differences in expression were found for 17 out of 33 genes (51%), while another 11 entries (33%) showed a similar trend. In summary, a correlation of 84% was observed between the fold changes in the microarray and those in the real-time PCR.
Next, we studied the protein expression pattern of the serine-protease inhibitor WFDC1, which may be involved in extracellular matrix (ECM) turnover. We chose this entry for confirmatory immuno-staining, for several reasons: (a) It was one of the top entries of our differential expression list in terms of fold changes; (b) the antibody ps20ab-1, against WFDC1, was fully characterized previously [13,14]; (c) RPE specificity, and inversed regional RPE expression differences were previously shown during embryonic development ; and (d) the antibody was readily available.
Our mRNA expression study (Table 3), predicted a stronger immunostaining of ps20ab-1 in the macula. While interpretation of (semi-)quantative studies by immunohistochemistry has its limitations, we observed a intense staining of the RPE underlying the macula. The staining was clearly less in the P2 area (Figure 2). To formalize this observation, we measured the optical density of the RPE in the different regions. First, we measured the negative control sections, which showed that the optical density of the RPE (melanin) slightly increased toward the periphery. Despite of this increasing absorption due to melanin, the measurements of the sections stained with the antibody showed a two-fold reduction in optical density of the RPE between M and P2. The latter indicated, at least semi-quantitatively, that the intensity of the immune staining was higher in the macula than in P2. Blood vessels and capillaries in the neural retina and inner segments of the photoreceptors showed no immunoreactivity. The choriocapillaris was not stained.
Functional overrepresentation analysis
We used the Web tools EASE/DAVID and GOstat for functional overrepresentation analysis of the 1155 differentially expressed gene entries from the ANOVA analyses on M, P1, and P2 (Appendix 1). The significant overrepresented ontologies were ECM (EASE score 0.002), and 'ECM structural constituent' (EASE score 0.003), and 'skeletal development' (EASE score 0.003). GOstat analysis with FDR correction identified the 'ECM' (GO:0005578, p=5.4*10-5) as the only statistically significant overrepresented group of genes. Thus, both EASE and GOstat analyses independently defined ECM genes as a statistically overrepresented class of genes in our RPE dataset. In total, 33 ECM genes were differentially expressed between macular and peripheral RPE cells. All but two genes (COMP, THSB4) were more abundantly expressed in the periphery than in the macula (Table 4).
In this study, we describe the analysis of 22K gene expression profiles of macular and peripheral RPE cells. There have been a number of other published studies that had a similar research question but employed different methodologies [8,17-21]. However, all of these studies suffered from one or more relative weaknesses in terms of sample selection, RPE cell handling, target selection, as well as microarray- or bioinformatics methods. Our aim was to avoid these limitations as much as possible (Table 5 and Table 6).
While the number of samples we used was limited, our entire strategy was focused on reduction of gene expression differences between individual eyes. In this study, we detected only consistent differences, more or less present in the RPE of each individual eye. We neither detected transient gene expression differences nor potential differences which co-occurred with high interindividual variation. The complexity of our material was greatly reduced by stringent selection of donor eyes (age-range, cause of death, light microscopic examination and histological evaluation), the use of laser dissected RPE cells, stringent RNA quality control, and a reduced genetic complexity since the macular and peripheral samples were derived from the same eyes. The strength of the laser dissection approach was that minimal manipulation prior to freezing was needed, which could otherwise have possibly influenced gene expression. In addition, compared with other methods, RPE cells or RPE RNA was less contaminated with adjacent photoreceptor- or endothelial cells or RNA [8,17,18,20]. Finally, our oligo microarray and bioinformatic tools were more up to date as compared to previous studies.
Differentially expressed retinal pigment epithelium genes from the macula and periphery
A large set of significant differentially expressed genes (438) resulted from the ANOVA analysis of the macular RPE versus the two peripheral regions (P1 and P2; Appendix 2). If we divided this number by 10.000-30.000 genes expressed in the RPE, we approximate that around 1-5% of the RPE transcriptome was different between macula and periphery.
The top differential RPE expression data-set of 65 genes (presented in Table 3), included expression of a few cone photoreceptor specific genes (PDE6H, OPN1-SW, -MW, and -LW). As Ishibashi et al. already observed in their study that employed retinal laser dissection microscopy (LDM) , there was apparently photoreceptor mRNA present at the photoreceptor-RPE interface. The presence of cone opsin mRNA (most abundant mRNA in cones) in itself was a limitation of this study, but validated our approach as an internal control as well. We observed that "RPE" gene expression of cone opsin was higher in the macula. If we assume that equal photoreceptor material contamination were present in each LDM RPE sample used, that finding must be due to the higher number of cones present in the macula. No obvious contamination of endothelial specific genes/material was noted. At least 16 of our RPE entries (24%) were previously already known, or suspected, to be expressed by the RPE (ALDH3, IGF2, OCA2, PDGFA, ANG, PGF, GAS1, WFDC1, COL1, COL18A1, FLRT2, TIMP2, SOD3, and the laminin genes).
We compared our list of differentially expressed RPE genens with those of Ishibashi et al.  and Bowes Rickman et al. , which are the two closest related studies, and we found that only a few gene entries overlap. From our current data (Appendix 1) and those of Bowes-Rickman  (Table 4 of the supplementary results of Bowes-Rickman "4MacRPE" over "4PeriRPE") we can extract only four overlapping genes (Hs.30570, Hs.532853, Hs.21162, and Hs.532768) which show 1 to 2.5 higher expression in the macular RPE compared to the periphery in both studies. When we compared our data with those of Ishibashi et al. (RT-PCR confirmed targets in the manuscript of Ishibashi) , we found that both (c)KIT and ALDH1A3 (annotated AA455235 and NM_000693) were more highly expressed in the peripheral RPE.
The small number of overlapping gene entries between these studies may be explained as follows: First, a comparison between our current data and those derived from the literature is difficult. The various published studies on this subject have different strengths and weaknesses (Table 5 and Table 6). Furthermore, our data indicated that gene expression differences between M and P1 were much less pronounced than those between M and P2 (Table 3). The latter suggested that gene expression (differences) may depend on the specific peripheral location of the RPE sample used. Unfortunately, it is impossible to compare the exact sampling location in between previously published studies. Finally, in contrast with our study, Ishibashi et al.  used RPE from old donors, in which gene expression patterns may be markedly different. At the time of their study, only filter arrays with a limited set of genes (4K), and non-linear exposure methods were available . If we compared our study specifically with that of Bowes Rickman et al. , we can see that patient selection, sample handling, methodology, and bioinformatic tools used in both studies are quite different. Most notably, their RPE sample appears to be heavily contaminated by photoreceptor and endothelial cell material (Table 6).
Retinal pigment epithelium expression points at topographic differences in Bruch's membrane composition
Data driven overrepresentation analyses of all our differentially expressed RPE genes yielded a cluster of 33 genes related to the gene ontology term ECM. Further functional annotation of the genes from this cluster resulted in four categories: (a) elastin/collagen related genes; (b) genes involved in cell adhesion; (c) matrix metallo-proteinase genes; and (d) a miscellaneous group. (Table 4). Obviously, given it's location between the RPE and choroid, BM is an extracellular matrix largely formed by secreted molecules from the RPE and choroidal (endothelial) cells. From the first group (a), at least collagens types I, VI, and XVIII as well as the laminins are known constituents of BM . Furthermore, data from the first three GO-term groups (a-c) revealed that all putative RPE ECM genes, except COMP, THBS4, and perhaps SPARC, showed a more abundant expression in the retinal periphery. This possibly reflects an increasingly active production or turnover of BM from the macula toward the retinal periphery. The latter may correlate well with the observed increase in thickness and integrity of BM and it's elastic lamina from the macula toward the periphery . Finally, it is interesting to note, that RPE gene expression from macula or periphery did not significantly change for a number of other known molecular constituents of BM, such as most fibronectin types and collagen IV chains (data not shown). Taken together, our data suggest that the molecular composition of BM may alter depending on it's central or peripheral location in the retina.
Potential implications for retinal disease
Topographic differences in BM composition and turnover rate, which, in part, is reflected by local RPE gene expression patterns, may be connected with local differences in physiology and disease susceptibility. The latter hypothesis prompted us to survey the literature for possible links between the selected genes and pathophysiology. Interesting data was found for Fibrillin 1, EMILIN1, Collagen XVIII, and perhaps WFDC1.
Fibrillin 1 is mutated in Marfan syndrome, in which the phenotype includes angioid streaks. The latter are breaks in BM that may be due to loss of elasticity of this layer. The EMILIN1 gene interacts with fibulin 5, which was previously associated with age- related macular degeneration [22,23]. Bhutto et al.  demonstrated a possible role for collagen XVIII, and more specifically, its cleavage product endostatin (an angiogenesis inhibitor) in AMD pathology. Mice lacking collagen XVIII show AMD and basal laminar deposit formation between the RPE and BM, indicating a direct interaction between this collagen and the RPE cells . Finally, the WFDC1 gene encodes the ps20 protein, a serine protease inhibitor that is most likely involved in turnover of ECM, in this case BM. In murine eye development, the expression of WFDC1 is largely restricted to the periphery of the developing RPE . This is in contrast to our finding of preferential macular expression in adult human eyes. Ps20 stimulates wound healing and endothelial cell migration in vitro. Thus, given it's expression pattern and putative function, WFDC1 is an interesting target for further macular disease studies.
Conclusion and summary
Our results present an overview of topographical changes in gene expression in healthy RPE cells in unaffected and relatively young human eyes. Our data illustrate the tight functional interaction between RPE and BM and may give an entry toward the understanding of topographical differences in RPE and BM biology and pathophysiology.
We thank Dr L. Pels and co-workers of the Corneabank, Amsterdam, for providing donor eyes. We also thank Drs. T. Gorgels and J. Booij for critically reading the manuscript and C. Ottenheim and E. Groot for excellent technical assistance. This study was supported by grants from the Foundation Fighting Blindness (T-GE-0101-0172), The Netherlands Organization for Scientific Research (NWO; project 948-00-013), The Royal Netherlands Academy of Arts and Sciences (KNAW) and de Algemene Nederlandse Vereniging ter Voorkoming van Blindheid (ANVVB).
1. Strauss O. The retinal pigment epithelium in visual function. Physiol Rev 2005; 85:845-81.
2. Bok D. The retinal pigment epithelium: a versatile partner in vision. J Cell Sci Suppl 1993; 17:189-95.
3. Marmor MF. Structure and function of the retinal pigment epithelium. Int Ophthalmol Clin 1975; 15:115-30.
4. Marmor MF. The retinal pigment epithelium and ocular disease. Int Ophthalmol Clin 1975; 15:131-46.
5. de Jong PT. Age-related macular degeneration. N Engl J Med 2006; 355:1474-85.
6. Chong NH, Keonin J, Luthert PJ, Frennesson CI, Weingeist DM, Wolf RL, Mullins RF, Hageman GS. Decreased thickness and integrity of the macular elastic layer of Bruch's membrane correspond to the distribution of lesions associated with age-related macular degeneration. Am J Pathol 2005; 166:241-51.
7. Del Priore LV, Tezel TH, Kaplan HJ. Maculoplasty for age-related macular degeneration: reengineering Bruch's membrane and the human macula. Prog Retin Eye Res 2006; 25:539-62.
8. Ishibashi K, Tian J, Handa JT. Similarity of mRNA phenotypes of morphologically normal macular and peripheral retinal pigment epithelial cells in older human eyes. Invest Ophthalmol Vis Sci 2004; 45:3291-301.
9. Hosack DA, Dennis G Jr, Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol 2003; 4:R70.
10. Beissbarth T, Speed TP. GOstat: find statistically overrepresented Gene Ontologies within a group of genes. Bioinformatics 2004; 20:1464-5.
11. Kamphuis W, Schneemann A, van Beek LM, Smit AB, Hoyng PF, Koya E. Prostanoid receptor gene expression profile in human trabecular meshwork: a quantitative real-time PCR approach. Invest Ophthalmol Vis Sci 2001; 42:3209-15.
12. Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol 2002; 3:RESEARCH0034.
13. McAlhany SJ, Ayala GE, Frolov A, Ressler SJ, Wheeler TM, Watson JE, Collins C, Rowley DR. Decreased stromal expression and increased epithelial expression of WFDC1/ps20 in prostate cancer is associated with reduced recurrence-free survival. Prostate 2004; 61:182-91.
14. Larsen M, Ressler SJ, Lu B, Gerdes MJ, McBride L, Dang TD, Rowley DR. Molecular cloning and expression of ps20 growth inhibitor. A novel WAP-type "four-disulfide core" domain protein expressed in smooth muscle. J Biol Chem 1998; 273:4574-84.
15. Graham RC Jr, Karnovsky MJ. The histochemical demonstration of monoamine oxidase activity by coupled peroxidatic oxidation. J Histochem Cytochem 1965; 13:604-5.
16. Rowan S, Chen CM, Young TL, Fisher DE, Cepko CL. Transdifferentiation of the retina into pigmented cells in ocular retardation mice defines a new function of the homeodomain gene Chx10. Development 2004; 131:5139-52.
17. Sharon D, Blackshaw S, Cepko CL, Dryja TP. Profile of the genes expressed in the human peripheral retina, macula, and retinal pigment epithelium determined through serial analysis of gene expression (SAGE). Proc Natl Acad Sci U S A 2002; 99:315-20.
18. Bowes Rickman C, Ebright JN, Zavodni ZJ, Yu L, Wang T, Daiger SP, Wistow G, Boon K, Hauser MA. Defining the human macula transcriptome and candidate retinal disease genes using EyeSAGE. Invest Ophthalmol Vis Sci 2006; 47:2305-16.
19. Wistow G, Bernstein SL, Wyatt MK, Fariss RN, Behal A, Touchman JW, Bouffard G, Smith D, Peterson K. Expressed sequence tag analysis of human RPE/choroid for the NEIBank Project: over 6000 non-redundant transcripts, novel genes and splice variants. Mol Vis 2002; 8:205-20 <http://www.molvis.org/molvis/v8/a27/>.
20. Kociok N, Joussen AM. Varied expression of functionally important genes of RPE and choroid in the macula and in the periphery of normal human eyes. Graefes Arch Clin Exp Ophthalmol 2007; 245:101-13.
21. Chowers I, Esumi N, Campochiaro P, Zack DJ. Studies on retinal and retinal pigment epithelial gene expression. Novartis Found Symp 2004; 255:131-45;discussion145-6,177-8.
22. Stone EM, Braun TA, Russell SR, Kuehn MH, Lotery AJ, Moore PA, Eastman CG, Casavant TL, Sheffield VC. Missense variations in the fibulin 5 gene and age-related macular degeneration. N Engl J Med 2004; 351:346-53.
23. Lotery AJ, Baas D, Ridley C, Jones RP, Klaver CC, Stone E, Nakamura T, Luff A, Griffiths H, Wang T, Bergen AA, Trump D. Reduced secretion of fibulin 5 in age-related macular degeneration and cutis laxa. Hum Mutat 2006; 27:568-74.
24. Bhutto IA, Kim SY, McLeod DS, Merges C, Fukai N, Olsen BR, Lutty GA. Localization of collagen XVIII and the endostatin portion of collagen XVIII in aged human control eyes and eyes with age-related macular degeneration. Invest Ophthalmol Vis Sci 2004; 45:1544-52.
25. Marneros AG, Keene DR, Hansen U, Fukai N, Moulton K, Goletz PL, Moiseyev G, Pawlyk BS, Halfter W, Dong S, Shibata M, Li T, Crouch RK, Bruckner P, Olsen BR. Collagen XVIII/endostatin is essential for vision and retinal pigment epithelial function. EMBO J 2004; 23:89-99.