Advertisement
JBC

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Originally published In Press as doi:10.1074/jbc.M405085200 on June 24, 2004

J. Biol. Chem., Vol. 279, Issue 40, 41903-41910, October 1, 2004
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Data
Right arrow All Versions of this Article:
279/40/41903    most recent
M405085200v1
Right arrow Submit a Letter to Editor
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowRequest Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Rohrer, B.
Right arrow Articles by Almeida, J. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rohrer, B.
Right arrow Articles by Almeida, J. S.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?

Multidestructive Pathways Triggered in Photoreceptor Cell Death of the RD Mouse as Determined through Gene Expression Profiling*{boxs}

Baerbel Rohrer{ddagger}§, Francisco R. Pinto||**, Kathryn E. Hulse§, Heather R. Lohr§, Li Zhang{ddagger}{ddagger}, and Jonas S. Almeida||

From the Departments of {ddagger}Ophthalmology, §Physiology and Neuroscience, and ||Biostatistics, Bioinformatics, and Epidemiology, Medical University of South Carolina, Charleston, South Carolina 29425, the **Biomathematics Group, Instituto de Tecnologia Química e Biológica, University Nova Lisboa, 2780 Oeiras, Portugal, and the {ddagger}{ddagger}M. D. Anderson Cancer Center, University of Texas, Houston, Texas 77030

Received for publication, May 7, 2004 , and in revised form, June 24, 2004.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
In the rd/rd mouse, photoreceptor degeneration is due to a mutation of the rod-specific enzyme cGMP phosphodiesterase, resulting in permanently opened cGMP-gated cation channels in the rod outer segment membrane that allow Na+ and Ca2+ ions to enter the cell, resulting in possibly toxic levels of Ca2+. To identify pathways involved in cell death of the rd/rd rods, we evaluated gene expression in the rd/rd and wild type (wt) mouse retina (U74A oligonucleotide arrays (Affymetrix)) over the known time course of photoreceptor degeneration. 181 genes passed the selection criteria (low standard deviation and high correlation between replicates), falling into six clusters. For any given pair of genes, an expression profile correlation distance and a semantic distance (one for each class of gene ontology terms) were established using newly designed software. Gene expression in rd/rd started to deviate from wt by postnatal day 10. The reduction in photoreceptor-specific genes followed the known time course of photoreceptor degeneration. Likewise the increase in transcription factors and apoptosis- and neuroinflammation-specific genes followed the kinetics of the rise in intracellular cGMP in the rod photoreceptors. In addition, genes coding for calcium-binding proteins and those implicated in tissue and vessel remodeling were increased. These results suggest that photoreceptor degeneration in the rd/rd mouse is a process starting with Ca2+ toxicity followed by secondary insults involving multidestructive pathways such as apoptosis and neuroinflammation, presumably boosting morphological changes. All of these components need to be addressed if rods are to be successfully protected.


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Photoreceptors are very stable but at the same time extremely fragile cells (1). Photoreceptor degeneration can be caused by any changes that alter the composition of the signal transduction cascade, influence the energy metabolism or the oxygen tension in the outer retina, or disturb the phagocytic process by the retinal pigment epithelium. In the rd/rd mouse, photoreceptor degeneration is due to an autosomal recessive mutation of the enzyme cGMP phosphodiesterase, which is an important component in the phototransduction cascade (2). The genetic defect is due to the integration of a provirus (Xmv-28) into intron I of the {beta} subunit of the enzyme and subsequent incorrect splicing (3). Degeneration starts at about P8,1 which is the time point at which naturally occurring cell death starts in the wild type mouse (4), and commences very quickly.

The resulting lack in cGMP phosphodiesterase activity causes a dramatic increase in cytoplasmic cGMP concentration (5). This in turn results in permanent opening of the cGMP-gated cation channels in the photoreceptor membrane, allowing the excessive entry of extracellular ions, particularly calcium. It has been suggested that this increase in intracellular calcium causes a metabolic overload of the cells, eventually leading to cell death by apoptosis (6). Likewise it has been suggested that either disruption of the phototransduction cascade or its continued activation can lead to photoreceptor degeneration in humans (e.g. Refs. 7 and 8).

With the advent of genomic technology, transgenic and double knock-out mice have been generated to determine which genes are involved in causing or preventing photoreceptor cell death in the rd/rd mouse (e.g. Refs. 9–11). In addition, drug studies have been designed to identify factors that could delay or interfere with photoreceptor degeneration (e.g. Refs. 12–15), immunohistochemical analyses have been performed to describe changes in protein expression in the rd/rd retina (16, 17), and calcium-induced cell death in a photoreceptor culture model has been studied to unravel the apoptotic pathway (64). However, a clear picture has yet to emerge that describes the apoptotic pathways involved in photoreceptor degeneration in the rd/rd mouse.

To obtain a more complete view of potential pathways involved in cell death of the rd/rd mouse photoreceptors we evaluated gene expression over the known time course of photoreceptor degeneration in the rd/rd and wild type mouse retina. Our results indicate that the rd gene-triggered cell death alters the expression of genes involved in diverse cellular pathways including calcium homeostasis, catabolism, neuroinflammation, and tissue remodeling. Temporal gene expression profiles are discussed in the context of the cGMP and calcium accumulation and photoreceptor cell death kinetics. To conclude, a hypothetical transcriptional network is proposed that will require further validation in the future.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Animals
These studies adhered to the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research as well as the Medical University of South Carolina regulations for the use of animals in medical research. C57BL/6 mice (wild type (wt)) were purchased from Harlan (Indianapolis, IN); the rd/rd mouse strain (on a C57BL/6 background) was a generous gift from Dr. Debora Farber (University of California, Los Angeles, CA). Animals were bred on homozygous backgrounds and raised in a 12-h light/12-h dark cycle with unlimited access to food and water.

RNA Preparation
All chemicals used in this study were at least molecular biology grade material and purchased from Fisher Scientific unless otherwise noted. Age-matched wt and rd/rd animals (P6, P10, P14, P17, and P21) were sacrificed by decapitation, and retinas were quickly isolated and stored in RNAlater (Ambion) at –20 °C. Retinas from four animals per geno-type per time point were pooled, and each data point was obtained in duplicate. Total RNA was isolated using TRIzol (Invitrogen) followed by a clean-up using RNeasy minicolumns (Qiagen). The quality of the RNA was examined by gel electrophoresis (to visualize the quantity and integrity of the 18 and 28 S ribosomal RNA bands) and spectrophotometry (accepting a 260/280 ratio of 1.95–2.1).

Microarray Procedures
Sample preparation for microarray hybridization was carried out as described in the Affymetrix Expression Analysis Technical Manual with minor modifications. Five micrograms of total RNA were used to generate double-stranded cDNA (Invitrogen), which was purified using phase lock gel columns (Eppendorf) followed by ethanol precipitation. The purified cDNA served as a template for the generation of biotinylated cRNA using the BioArrayTM HighYieldTM RNA transcript labeling kit (ENZO Diagnostics). The labeled probes were purified using the RNeasy minikit (Qiagen), fragmented (8 mol/liter Na+-citrate buffer), and stored at –80 °C. The length of the cRNA and fragmentation was confirmed by agarose gel electrophoresis. U74A oligonucleotide arrays (Affymetrix) containing ~6,000 genes (all sequences contained in the Mouse UniGene data base, Build 74) and ~6,000 expressed sequence tag clusters were used in this study. Hybridization with equal amounts of labeled cRNA (15 µg/array) and readout were performed by the DNA Microarray Core Facility at the Medical University of South Carolina using the Affymetrix Fluidics Station.

Expression Estimation and Normalization
Gene chips were scanned using the Affymetrix scanner (Microarray Suite 5.0 software) to obtain probe level data. Outputs were scaled to a target intensity of 250. The raw Affymetrix data (absolute expression level and perfect match values) were normalized using a quantile normalization method, and gene expression levels were obtained using the positional dependent-nearest neighbor model, which integrates probe sequence information for estimating probe binding affinities to the target sequences on microarrays (18). Parameters for filtering the differentially expressed genes are presented under "Results."

Functional Analysis
All algorithms and computations referred to in this section were implemented (or have already been implemented) in Matlab (Version 6.5), including its Statistics Toolbox. A complete description of the implementation of tools and the link to the program can be obtained from the author (65).

Hierarchical Clustering—The mouse gene set was clustered by gene expression profiles. The distance between gene profiles was defined by the correlation distance. Hierarchical clustering was performed using unweighted pair-group average linkage (19). The resulting dendrogram was divided into separate clusters. The cutting points were searched by identifying peaks in the distances between adjacent genes in the dendrogram order. Five multipeaks were identified and smoothed with a 3-point moving average.

Gene Ontology (GO) Annotations—The annotations for the mouse genome and the relation scheme between the GO terms used were obtained at the GO Consortium site (Gene Ontology Consortium 2000, www.geneontology.org). Annotations for molecular functions, biological processes, and cellular components were used separately.

Computation of Semantic Distances—The three annotation categories of function, process, and cell location were used. The following description of the procedure is valid for the three classes, and the procedure was repeated for each one of them independently. The fact that the terms are interrelated enables the direct calculation of a distance measure between each two terms. A measure based in information theory concepts, called semantic distance, was used. This measure was previously defined and discussed (20) and is briefly described. For each two terms it is possible to find parent terms in common. From these, the most specific one is chosen. The semantic distance value between two terms is as high as the difference between the specificity of the common parent term and of the two child terms. The specificity of a term is measured by the information content of that term and defined as follows,

(Eq. 1)
where pi is the frequency of genes annotated with the term i, or with child terms of the term i, in the genome. The semantic dissimilarity is then calculated as follows,

(Eq. 2)
where k varies between 1 and the number of ancestral common nodes of the two terms i and j. Thus, the distance of interest is between two genes, not between two terms, and this distance can only be calculated when each of the two genes has an annotation with at least one term. In the case of a gene having an annotation with more than one term, all the possible distances between the terms of each gene are calculated with the final ontological distance being defined by the minimum value obtained (i.e. having the shortest path in the ontology graph).

Attribution of Colors to GO Terms—The list of all the unique GO terms used to annotate the selected set of genes for which the semantic term pairwise distance had been determined was sorted by hierarchical clustering using unweighted pair-group average. This way similar terms should be close to each other in the dendrogram order (with the exception being the ones at the edge of each cluster), which consequently will have similar colors. These colors are then used to graphically represent the distribution of GO terms across different expression profiles.

Computation of GO Clustering Coefficients—With the aim of detecting in an objective way groups of genes that were similarly expressed and at the same time were similarly annotated, GO clustering coefficients were computed (i.e. local correlation coefficients). Statistical significance of the local correlation coefficients for each gene was evaluated by bootstrapping.

Over-represented GO Terms—GO terms that appeared to be over-represented in this set of 181 genes when compared with their frequency in the entire mouse genome were further examined by {chi}2 and binomial testing for significant deviation (21). The selected genes had to have a p value lower than 0.05 for at least one of the tests, and if they only achieved that value for one of the tests, the other p value should not be higher than 0.075. This double test strategy was a way to overcome the possible flaws of the {chi}2 test for low expected frequencies and other situations where the binomial test assumptions were not completely valid. It should be noted that the binomial test uses the "term frequency in the genome" as a constant characteristic of the entire population, but some term frequencies could be biased because some areas of biology are more intensively studied than others. Additionally the binomial test is ideally applied when one takes finite samples from an infinite population; this is not the case here, but samples can be comparable if the sample size (181 in this case) is small when compared with the finite population size (~12,500 probe sets on the array). From the set of terms that were considered significantly over-represented, only the ones that appeared in more than five genes (of the 181 set) were subsequently analyzed.

Analysis of Kinetics
The kinetics of photoreceptor degeneration, cGMP accumulation, and changes in gene expression for select clusters was analyzed. Outer nuclear layer thickness in the rd/rd mouse retina was fitted with an exponential decline (22), whereas cGMP accumulation, which rises quickly between P6 and P14 (5), could be fitted with a single exponential. Gene clusters were analyzed with the same two equations.

Real Time PCR
The same RNA samples that were used to generate the biotinylated cRNA were also used for real time PCR. Equal amounts of RNA (1 µg) were used in reverse transcription reactions (Invitrogen). PCR amplifications were conducted using the QuantiTect Syber Green PCR kit (Qiagen) with 0.2 µmol/liter forward and reverse primers (see Table I for primer sequences, gene accession numbers, and expected lengths; each primer was designed to span an intron) and equal amounts of complimentary DNA (1 µl of 1:10 dilutions). Reactions were treated with 0.01 unit/µl AmpErase® uracil-N-glycosylase enzyme (Applied Biosystems) to prevent carryover contamination. Real time PCR was performed in triplicate in a GeneAmp® 5700 sequence detection system (Applied Biosystems) with the following cycling conditions: 50 °C for 2 min, 94 °C for 15 min, 40 cycles of 94 °C for 15 s, and 58 °C for 1 min. Quantitative values were obtained by the cycle number (Ct value) (i.e. by using the point at which fluorescence starts to increase above background and therefore at a fixed threshold level (set at 0.15 relative fluorescence units)), which is inversely proportional to the amount of a specific mRNA species in the tissue sample from which the cDNA was derived. Relative gene expression levels were calculated using the equation y = (1 + AE){Delta}{Delta}Ct where AE is the amplification efficiency of the target gene (set at 1.0 for all calculations), and {Delta}{Delta}Ct is the difference between the mean experimental and control {Delta}Ct values. The {Delta}Ct value is the difference between the Ct value for a retina-associated gene and the {beta}-actin internal reference control gene (23).


View this table:
[in this window]
[in a new window]
  		TABLE I
Primer sequences used in quantitative real time PCR

 

   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Our main goal was to molecularly characterize the pathways of photoreceptor degeneration in the rd/rd mouse. We therefore chose to assay gene expression in the retina of rd/rd mice from postnatal day 6 (2 days prior to the onset of retinal degeneration) through postnatal day 21 (the time point when the majority of the photoreceptors have been eliminated, and relatively little reorganization in the inner retina has occurred) (24) in steps of 3–4 days.

Normalization of Data—To extract gene expression levels from microarray data we used a recently developed model, called the positional dependent-nearest neighbor model. This model takes into account that hybridization on the microarrays is affected by the stacking energy of the nearest neighboring base pairs and the microarray surface. It has been demonstrated that this approach gives more precise and accurate expression estimates than the conventional alternatives (18).

The identification of differentially expressed genes is based on the variance structure of the expression data and the reproducibility of gene expression profiles. Fig. 1A shows the S.D. of the log-transformed expression levels (LELs) plotted against the LEL average (Avg(LEL)) for each gene on the arrays. Fig. 1B shows the correlation of LEL profiles between replicates against the Avg(LELs). There seems to be a trend that better correlations are observed with higher LELs. At the lower end (Avg(LEL) < 5.5), the median of the correlation coefficients is below zero, indicating poor reproducibility due to gene expression values below detection threshold. Very good correlations were observed for genes with S.D.(LEL) < 0.45 as shown in Fig. 1C. Therefore, we adopted the following criteria to identify differentially expressed genes: Avg(LEL) > 5.5; S.D.(LEL) < 0.45; and correlation between replicates, >0.80; from these criteria a list of 181 genes was produced.



View larger version (33K):
[in this window]
[in a new window]
  		FIG. 1.
Data filtering. A, plot of standard deviation versus average of LEL. LEL is the log-transformed expression level according to the positional dependent-nearest neighbor model, revealing higher S.D. of data at low expression levels (<5.5) than at higher expression levels. Correlation between replicates was used to identify variable genes. In B, all genes are included, while in C, only those genes with Avg(LEL) > 5.5 were selected. Please note that if Avg(LEL) or S.D.(LEL) are low, the correlation between replicates is also low. expr, expression.

 
Hierarchical Clustering—Hierarchical cluster analysis was performed with the expression profiles of the 181 genes (Fig. 2D; see also supplemental material, data and cluster assignment, which lists genes and their correlation distances). The clustering tree shows that the samples fall into three clusters: one consisting of young animals (irrespective of genotype), the second one of wt mice older than P10, and the final one of rd/rd animals older than P10. The tight correlation between P6 wt and rd/rd mice indicates that there is no large scale expression difference between the two genotypes prior to the onset of {beta} phosphodiesterase expression and the start of photoreceptor cell death (see also Fig. 3). The somewhat closer correlation of older rd/rd mice with young animals (P6) reflects the dominance of photoreceptor genes in the retina. As rod photoreceptors outnumber all other retinal cells combined by 3.5:1 in the mouse retina (25), it is not surprising that samples lacking the majority of rod-specific genes (i.e. retina at P6, which is prior to final photoreceptor differentiation) and those that lose the majority of their rods (older rd/rd mice) would cluster together. Although samples from the same age and phenotype are tightly clustered, one or two samples show less consistency than the others when replicate samples are compared (e.g. P10C57–1 and P14RD-2).



View larger version (44K):
[in this window]
[in a new window]
  		FIG. 2.
Cluster analysis. Cluster analysis of filtered data is shown. Please note that for clarity, the figure panels are described out of order. D, hierarchical clustering of gene expression profiles. Selection criteria were as follows: Avg(LEL) > 5.5; S.D.(LEL) < 0.45; correlation between replicates, >0.80. Note that the clustering tree of genes shows a striking modular behavior. The data are expressed as the log2(expression ratio) from ≤2 (red) to ≥2 (green). B, to identify the modules, the dendrogram was split into clusters with coherent expression profiles based on the distance between adjacent genes in the dendrogram (3-point moving average). C, to assign function to the genes, data were represented based on three categories: molecular function (C1), biological process (C2), and cellular component (C3) derived from GO. Within each category, data were visualized based on GO terms (right-hand part of each panel, see A for color assignments) or GO term clustering (left-hand part of each panel). GO term clustering is based on the semantic distance or the degree of dissimilarity between the terms describing each gene, and GO clustering p values were assigned (see GO clustering p value for color assignment). Time points: 1, P6; 2, P10; 3, P14; 4, P17; 5, P21.

 



View larger version (30K):
[in this window]
[in a new window]
  		FIG. 3.
Functional analysis. The six clusters identified in Fig. 2B were further characterized in terms of expression levels over time (diagrams, average ± S.D.) and function (biological process terms, pie charts). Please note that the color assignment is not identical to that in Fig. 2C2 as the assignment between categories would not be sufficiently different.

 
Cluster Assignment—The hierarchical cluster analysis (Fig. 2D) demonstrated that the genes show a striking modular behavior. To determine the number and size of the modules, the distance between adjacent genes in the dendrogram was plotted using a 3-point moving average. The resulting peaks defined the separation between clusters (Fig. 2B). However, the peaks separating the clusters may include genes and not just separate the clusters (e.g. peak between clusters 3 and 4). Broad peaks contain genes with large distances to all of their neighbors that make their automated classification highly sensitive to the clustering threshold considered. Four genes were found under these conditions (genes 95–98) and consequently were manually associated with cluster 3. Using this procedure, six clusters were identified: cluster 1, genes 1–69 (in the dendrogram order); cluster 2, genes 70–79; cluster 3, genes 80–98; cluster 4, genes 99–115; cluster 5, genes 116–155; and cluster 6, genes 156–181 (see also Fig. 3 and supplemental material, data and cluster assignment, for further analysis of the clusters).

Gene Ontology—The GO terms available for all the genes on the U74A arrays were used to classify the genes in the six clusters. Color schemes were built to identify the GO terms based on molecular function, biological process, and cellular component (Fig. 2, C1, C2, and C3, right-hand color bars, respectively; see supplemental material, legend lists, for a list of GO terms used). The color coding allows for evaluation of clusters that are dominated by particular colors (i.e. dark blue for phototransduction in biological processes (Fig. 2, C2 and see A2 for legend)). However, due to the large number of annotation terms required for GO, the resulting figures are time-consuming to interpret. Thus, we set out to determine whether genes with similar annotation cluster together according to their expression profiles. For any given pair of genes, an expression profile correlation distance and a semantic distance (one for each class of GO terms) was established. The semantic distance expresses the degree of dissimilarity between the terms with dissimilarity being based on the distance between two terms within a tree of terms (directed acyclic graph). Therefore, the genes of interest are the ones for which small expression profile distances are associated with small semantic distances. The results of this analysis demonstrated that the clustering varied significantly based both on which class of GO terms was used for the analysis and which cluster was analyzed (Fig. 2, C1–C3, left-hand color bars, see legend for GO clustering p values at the bottom; see also supplemental material, data and cluster assignment). For example, in the molecular function category of cluster 1 (Fig. 2C1), the GO term clustering method detected a significant number of calcium-binding proteins (light green; n = 9, p < 0.001); however, they participate in different biological processes (Fig. 2C2) and are localized to different cellular compartments (Fig. 2C3). In cluster 5, 17 vision-related genes were identified by analyzing the biological processes category (dark blue; p < 0.001) (Fig. 2C2). As expected, these genes did not cluster based on molecular function or cellular compartment. In other words, the GO clustering p values for the vision-related terms in cluster 5 were correlated (red) in the biological processes category but dispersed (green) in both the molecular function and cellular component categories. Cluster 6 was found to be the most consistent. GO clustering p values in all three categories revealed clustered genes. This cluster contains mostly genes involved in cell cycle and development, such as transcription factors and cell cycle proteins, which are co-localized in the nucleus and serve similar molecular functions (see supplemental material, data and cluster assignment). Yet these genes were part of different categories when analyzed in terms of the classical gene ontology method (comparing color codes of the right-hand columns among those genes with low p values (red genes)). Taken together, our results suggest that a combined analysis for the three terms is both useful and necessary to identify co-regulated genes.

Temporal Analysis—The clusters isolated in our analysis identified significantly different time-dependent expression patterns (Fig. 3). As shown by hierarchical clustering (Fig. 2D), at P6 there is no difference between the genotypes in any of the clusters, whereas by P10 (or possibly earlier if time points with smaller intervals had been chosen) clusters revealed time-dependent changes: up-regulated at all age points beyond P6 in the rd/rd retina (clusters 1 and 3), up-regulated in the rd/rd retina at P10 but not later in development (cluster 2), down-regulated in the rd/rd retina (clusters 4 and 5), and not different between the rd/rd and the wt retina (cluster 6). Based on this temporal expression pattern, one would predict that the genes most detrimental for the rd/rd photoreceptors might be represented in cluster 3. Gene expression of rd/rd samples in this cluster peeled off the quickest and with the greatest difference.

Over-represented GO Terms—GO terms that were over-represented in this subset of genes when compared with their frequency in the entire mouse array were identified by {chi}2 and binomial testing for significant deviation (21). 18 function terms, 25 process terms, and 10 cellular component terms passed the criteria of low odd ratio and significant p values (see supplemental material, over-represented GO terms). Of particular interest with respect to the apoptotic pathway analyses were the two functional terms (complement component (p{chi}2 = 0.003 and pbin = 0.004) and caspase-3 (p{chi}2 < 0.0001 and pbin < 0.001)) and the two process terms (catabolism (p{chi}2 = 0.001 and pbin = 0.001) and energy pathways (p{chi}2 < 0.0001 and pbin < 0.0001)).

To analyze the clusters in more detail, the genes in each cluster were represented based on a set of 11 process terms that were considered significantly over-represented (having a frequency in the set of >0.03; see supplemental material, over-represented GO terms) and interpreted based on their GO (Fig. 3, pie charts). The most significant genes represented by these GO terms are discussed below under "Discussion." Cluster 6 contains mainly cell cycle and development genes (i.e. GO terms: nucleic acid metabolism and cell proliferation) that change in parallel in rd/rd and wt mice over time, suggesting that development of the retina progresses normally in the rd/rd mice. As expected, one module contains mainly photoreceptor genes (41%) (cluster 5; i.e. GO terms: signal transduction, response to abiotic stimuli, and perception of external stimuli). The expression level of these genes drops progressively in the rd/rd mice after the onset of photoreceptor degeneration (P14–P21). This subset contains the gene for {beta} phosphodiesterase. And finally, clusters 1 and 3 contain the genes with the most drastic changes early in retinal degeneration and might therefore represent genes involved in the early processes of apoptosis. Cluster 1 is characterized by the molecular function term "calcium binding" and contains genes for various calcium sensors that appear to be up-regulated by the massive calcium influx due to the {beta} phosphodiesterase defect (no common over-represented GO terms available to characterize those genes). In addition, genes involved in angiogenesis and tissue remodeling were significantly elevated. Cluster 3 on the other hand is characterized by GO terms such as nucleic acid metabolism (i.e. transcription factors), immune response, catabolism, and response to biotic stimuli. A more detailed promoter study is currently being conducted to establish the potential transcriptional network that links the expression of these genes.

Real Time PCR Analysis—Real time PCR analysis was used to confirm the expression levels of a small subset of interesting genes (two genes from cluster 1 (aquaporin 4 and clusterin), three genes from cluster 3 (cathepsin S, ceruloplasmin, and the complement factor 1q-{beta} (C1q{beta})), and one gene from cluster 5 (tubby-like protein 1)). Genes from clusters 1 and 3 were chosen because of their known involvement in apoptosis and neuroinflammation, whereas the one photoreceptor-specific gene was examined for control purposes. As shown in Fig. 4, the expression profiles identified by gene expression arrays were matched by those determined by real time PCR, although the absolute levels were higher in the latter (Fig. 4).



View larger version (29K):
[in this window]
[in a new window]
  		FIG. 4.
Validation by quantitative reverse transcription PCR. A subset of genes was confirmed by quantitative reverse transcription-PCR (see Table I for primers used). The -fold changes obtained for these six genes were compared between the two methods (gene expression analysis (GEA) versus quantitative reverse transcription-PCR (QPCR)).

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
In the rd/rd mouse, an accepted model for photoreceptor degeneration in retinitis pigmentosa, the lack of functional cGMP phosphodiesterase causes an increased number of cGMP-gated channels to be in the open state due to the significant increase in intracellular cGMP (5), presumably allowing intracellular Ca2+ to rise to potentially neurotoxic levels and resulting in rapid photoreceptor degeneration (6). The present experiments were designed to examine rd-induced changes in gene expression in the retinas of these mice over time. Hierarchical clustering based on gene expression levels and gene ontology clustering based on molecular function revealed four important findings. First, postnatal rd/rd mouse retinas appear to develop normally prior to the onset of cGMP-gated channel function and the intracellular rise in cGMP. Second, early changes include increases in expression of genes encoding transcription factors and calcium-binding proteins as well as genes involved in cellular immune defense and catabolism. Third, later changes are characterized by loss of photoreceptor-specific gene expression and an increase in genes involved in blood-retina barrier breakdown. Fourth, analysis of the kinetics of the rise in cGMP and the degeneration of photoreceptor revealed kinetics similar to those of some of the dominant clusters in the gene expression analysis. Taken together, our results suggest that in photoreceptor degeneration in the rd/rd mouse, like other neurodegenerative diseases such as Alzheimer's disease, a primary insult is followed by secondary insults involving multidestructive pathways. In the rd/rd mouse these two insults appear to involve Ca2+ toxicity followed by apoptosis, blood-retina barrier breakdown, and neuroinflammation.

Data Filtering and Processing—Microarray analysis is a powerful tool to examine the expression level of thousands of genes within tissue samples. However, to obtain significant results, stringent filtering criteria and appropriate processing tools are required. Here we pooled tissue from four animals of the same genotype for each sample to obtain good quality RNA, to reduce sample variability, and to reduce the number of arrays needed to generate reliable data (26). Two samples were generated per genotype and age (five age points in total) spanning the period of photoreceptor differentiation (P6–P10) and degeneration (P14–P21) in the rd/rd mouse (27). As shown in Fig. 1, samples with low expression values (Avg(LEL)) tended to have large standard deviation (Fig. 1A) and low correlation between replicates (Fig. 1B). Thus, stringent criteria (Avg(LEL) > 5.5; S.D.(LEL) < 0.45; and correlation between replicates, >0.8) were used for the hierarchical cluster analysis, resulting in the selection of 181 genes. Quantitative real time PCR confirmed that this level of filtering is appropriate as all genes tested (six reported here and six additional not shown) had similar levels using either method. Our results differ from those in other reports in which only some of the identified genes could be verified by real time PCR (28, 29). However, on the flip side, the rigorous filtering probably limited the results to mainly highly expressed genes and suggests that we probably missed genes with low expression levels of possibly important functions. We are currently working on additional tools to increase the reliability of less stringent filtering techniques to prevent high false discovery rates when using a small number of replicates focusing on temporal trend analysis.

To identify the number of gene clusters, an algorithm was developed based on the correlation distance between adjacent genes in the tree (Fig. 2B), which identified six clusters in the data set. GO clustering, a tool that is available in other microarray analysis programs (e.g. Dchip (30)), was implemented and extended for further analysis. The GO terms were represented by color schemes for easy identification (Fig. 2, C, righthand columns, and A1–A3), and semantic distances were established (Fig. 2C, left-hand columns) and analyzed for overexpressed terms. Data for each cluster could then be expressed as a time series or as a summary pie chart focusing on genes separated by small semantic distances to aid in the identification of relevant genes (Fig. 3). Although the kinetics of gene expression coupled with correlation distance analysis was used to extract information about co-regulation, a more detailed promoter study will be necessary to establish the potential transcriptional network regulated by calcium.

Gene Expression Network—The hierarchical cluster analysis revealed that the 181 genes were clustered in six distinct patterns with different temporal patterns (Fig. 3). Clusters 4 and 5 and to some degree cluster 2 contain genes that were down-regulated in the rd/rd mouse retina. The levels of expression of these genes, which are predominantly photoreceptor-related, did not deviate significantly from wt levels until >P10, the onset of degeneration. The temporal pattern of gene expression in clusters 4 and 5 followed the known time course of photoreceptor cell loss in the rd/rd mouse (31) (Fig. 5B), although, as expected, the loss of photoreceptor-specific genes (cluster 5) preceded the loss of cells. In contrast, clusters 1 and 3, which contain transcription factors, genes for calcium-binding proteins, and genes whose expression is known to be activated by calcium and/or stress, were up-regulated as early as P10, peaked at P14, and remained elevated until P21 (the latest time point studied). The temporal pattern of gene expression in particular in cluster 3 followed the known time course of intracellular rise in cGMP in the rd/rd mouse outer nuclear layer that peaks at ~P14 (5) with a somewhat slower time constant (Fig. 5A). This temporal pattern suggests that genes in cluster 3 might be induced by the rise in intracellular cGMP followed by genes in cluster 1 activated by Ca2+. Together they may be involved in the initiation and/or execution of cell death and subsequently trigger events leading to phagocytosis of these dead cells.



View larger version (26K):
[in this window]
[in a new window]
  		FIG. 5.
Kinetics of cell death, cGMP accumulation, and gene expression in the rd/rd mouse retina. A, cGMP levels increase exponentially over time in the rd/rd mouse (5) (solid squares), and the kinetics of gene expression in cluster 3 (open squares, dotted line) were found to be tightly linked to this rise. Please note that for cGMP levels and cluster 3 only the rising phases were examined. B, photoreceptor degeneration (black symbols) in the rd/rd mouse retina follows an exponential decline (22). The kinetics of gene expression in clusters 4 (closed symbols, dashed line) and 5 (open symbols, dotted line) followed this decline of photoreceptors in the retina. ONL, outer nuclear layer.

 
Early Genes—As soon as the components of the rod signal the transduction cascade, and in particular guanylate cyclase and the cGMP-gated cation channel are expressed in the rd/rd photoreceptors, cGMP, and consequently calcium levels, begin to rise (>P6; Refs. 5, 6). Thus, early events were expected to include transcription factors as well as calcium-related transcription. Three transcription factors, c-Fos, Egr-1, and DKK-3, were identified in cluster 3. c-Fos is expressed in the outer nuclear layer during apoptosis in the rd/rd mouse retina. However, photoreceptor degeneration could not be prevented in a c-fos rd/rd double mutant mouse presumably due to the presence of multiple apoptotic pathways (Ref. 9 and this report). Egr-1 is a transcription factor involved in programmed cell death (32), and it has been shown to be up-regulated in the retina and lens cells in response to stress (33).2 Finally DKK-3 is a regulator of the Wnt pathway, a signaling network that has been shown to regulate apoptosis. In addition, it has been suggested that Wnt signaling may participate in degenerative processes leading to cell death in the aging brain (34).

Calcium is an important intracellular signaling molecule requiring tight regulation of intracellular concentrations for optimal triggering of signaling cascades and cell survival (for a review, see Ref. 35). In the mouse photoreceptor, normal calcium levels range between ~250 nmol/liter (in complete darkness) and ~60 nmol/liter (in the light) (36). However, in the rd/rd photoreceptors, calcium levels are increased up to ~190% over wt levels (37). Consequently messages for calcium-binding proteins and in particular calcium sensors such as calbindin, visinin-like 1, and sparc were up-regulated in the rd/rd retina. Genes activated by calcium through cAMP-response element-binding protein sites in their promoters include genes involved in the immune response such as {beta}2-microglobin and major histocompatibility complex class I receptors (cluster 3 (38)). The focus has been on calcium, while the question of cytotoxicity of high levels of cGMP remains unanswered at this time.

Late Genes—The central nervous system, like other tissues, has an efficient innate immune system that not only fights invading pathogens but also clears away toxic debris (e.g. debris caused by cell death (39)). Apoptotic cells bind so-called "eat me" signals to promote their removal by microglia and astrocytes. Normal photoreceptor eat me signals presented to the retinal pigmented epithelium and required for normal disk shedding include the presence of detached rod outer segments (40), Gas6 (cluster1 (41)), and its receptor MerTK (42). Pathological eat me signals include members of the classical complement pathway. C1q, which is up-regulated in the rd/rd retina (cluster 3), is produced in the brain by glial cells and neurons, and its levels are increased by tissue injury. C1q appears to bind preferentially to apoptotic cells, whereas non-dying cells protect themselves by expressing inhibitors to C1q such as clusterin (for a review, see Ref. 43). Co-regulation of C1q{beta} and clusterin has been reported in association with focal cerebral ischemia in the mouse (44), 6-hydroxydopamine lesions in the rat (45), and neurodegeneration due to sporadic amyotrophic lateral sclerosis in humans (46) and in this report. Clusterin (apoJ) has been reported to be localized in the retinal pigment epithelium cells and all three nuclear layers of the retina (47) and has been implicated in a number of cellular processes including lipid transport, membrane integrity, apoptosis, and neurodegeneration. Interestingly clusterin was found to be increased in patients with drusen (age-related macular degeneration (48)), in the inner nuclear layer of the rat retina after light-induced photoreceptor degeneration (47), or in streptozotocin-induced diabetic retinopathy (49).

Efficient clearance requires catabolic enzymes. Two enzymes were identified, cathepsin S, a cysteine protease, and peroxiredoxin 6 (cluster 3), which have been shown to be activated by oxidative stress (50, 51). In the retinal pigment epithelium cathepsin S affects the proteolytic processing by cathepsin D of diurnally shed photoreceptor outer segments (52). However, cathepsins have been shown to be involved in antigen processing and presentation by microglia, which are the initial steps of the immune response (53, 54). Peroxiredoxin 6 appears to be a bifunctional enzyme serving as a lysosomal phospholipase A and a protein with antioxidative function (55).

Photoreceptor degeneration is followed by remodeling of both neuronal and glial cells as well as the vasculature (for a review, see Ref. 56). Genetic evidence for tissue and vessel remodeling in the rd/rd retina is based on the overexpression of aquaporin 4, matrix metalloproteinase 9, tissue inhibitor of metalloproteinase 3, and fibroblast growth factor-1 (cluster 1). Matrix metalloproteinases and their tissue inhibitors have been shown to be involved in extracellular matrix remodeling in a variety of brain pathologies (57) including optic nerve inflammation (58). Fibroblast growth factor-1 is a potent angiogenic factor and is overexpressed in the retina after injury (59). Finally aquaporin 4, which is one of the components necessary for the development and establishment of the blood-brain barrier (60), when overexpressed in the injured brain leads to blood-brain barrier breakdown (Ref. 61; for a review, see Ref. 62). Aquaporin 4 is expressed in Mueller cells in the mouse retina (63), and its high expression in the internal limiting membrane suggests that aquaporin 4 might be involved in maintaining a fluid barrier. Taken together, these results suggest that tissue and vessel remodeling and potentially fluid barrier breakdown, possibly triggered by neuroinflammation and the resulting immune response, may contribute to the phenotype observed in these mice.

Conclusion—Photoreceptor degeneration, neuroinflammation, and vasculature remodeling are part of a major reorganization that occurs in all mammalian retinas during degeneration. Retinal degeneration can be divided into three phases: rod degeneration, cone degeneration, and a protracted remodeling phase that includes neuronal, glial, and vasculature remodeling (56). Based on our genetic evidence, it appears that in a retina that is undergoing very fast degeneration (i.e. the rd/rd retina), these phases may not occur sequentially but rather may happen simultaneously. Our observations in the context of data from neurodegeneration in other systems suggest that photoreceptor degeneration in the rd/rd mouse consists of two insults, the primary and unavoidable insult posed by the rise in cGMP and the concomitant Ca2+ influx and a secondary, possibly preventable insult triggered by the presence of dying cells, leading to oxidative stress, neuroinflammation, and tissue and vessel remodeling (i.e. leading to the activation of multidestructive pathways). Gene expression analysis has proven to be a powerful method to assign genes to the anatomical changes reported in the literature; however, further proof about coregulation using promoter studies and single cell analysis will be necessary to further corroborate our conclusions.


   FOOTNOTES
 
* This work was supported in part by National Institutes of Health Grants EY-13520 (to B. R.) and EY-14793 (Core Grant for Vision Research at Medical University of South Carolina (MUSC)), the NHLBI, National Institutes of Health Proteomics Initiative through Contract N01-HV-28181 (to J. S. A.), Portuguese Foundation for Science and Technology Grant SFRH/BD/6488/2001 (to F. R. P.), Fundação para a Ciência e a Tecnologia of the Portuguese Ministério da Ciência e do Ensino Superior Grant SAPIENS/34794/99 (to J. S. A.), the Kirchgessner Foundation (to B. R.), and an unrestricted grant to MUSC from Research to Prevent Blindness, Inc., New York. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Back

{boxs} The on-line version of this article (available at http://www.jbc.org) contains supplemental Worksheets 1 (data and cluster assignment), 2 (legend list), and 3 (over-represented GO terms). Back

To whom correspondence should be addressed: Dept. of Ophthalmology, Medical University of South Carolina, 167 Ashley Ave., Charleston, SC 29425. Tel.: 843-792-5086; Fax: 843-792-1723; E-mail: rohrer{at}musc.edu.

1 The abbreviations used are: P, postnatal day; wt, wild type; LEL, log-transformed expression level; Avg, average; C1q, complement factor 1q; GO, gene ontology. Back

2 Tari, S. R., Lee, S. E., Tseng, J. J., Onta, D., Pachydaki, S. I., Hörig, H., Moroziewicz, D. N., Yan, S. F., Schmidt, A. M., and Barile, G. R. (2004) Investig. Ophthalmol. Vis. Sci. ARVO E-Abstract 1187. Back


   ACKNOWLEDGMENTS
 
We thank Michael Schowe for helpful discussions on a previous version of the manuscript, Michael Mitas for the use of the GeneAmp® 5700 sequence detection system, and Luanna Bartholomew for critical review.



   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Stone, J., Maslim, J., Valter-Kocsi, K., Mervin, K., Bowers, F., Chu, Y., Barnett, N., Provis, J., Lewis, G., Fisher, S. K., Bisti, S., Gargini, C., Cervetto, L., Merin, S., and Peer, J. (1999) Prog. Retinal Eye Res. 18, 689–735[CrossRef][Medline] [Order article via Infotrieve]
  2. Farber, D. B. (1995) Investig. Ophthalmol. Vis. Sci. 36, 263–275[Free Full Text]
  3. Bowes, C., Li, T., Frankel, W. N., Danciger, M., Coffin, J. M., Applebury, M. L., and Farber, D. B. (1993) Proc. Natl. Acad. Sci. U. S. A. 90, 2955–2959[Abstract/Free Full Text]
  4. Lolley, R. N. (1994) Investig. Ophthalmol. Vis. Sci. 35, 4182–4191[Free Full Text]
  5. Farber, D. B., and Lolley, R. N. (1974) Science 186, 449–451[Abstract/Free Full Text]
  6. Chang, G. Q., Hao, Y., and Wong, F. (1993) Neuron 11, 595–605[CrossRef][Medline] [Order article via Infotrieve]
  7. Rosenfeld, P. J., Cowley, G. S., McGee, T. L., Sandberg, M. A., Berson, E. L., and Dryja, T. P. (1992) Nat. Genet. 1, 209–213[CrossRef][Medline] [Order article via Infotrieve]
  8. Robinson, P. R., Cohen, G. B., Zhulovsky, E. A., and Oprian, D. D. (1992) Neuron 9, 719–725[CrossRef][Medline] [Order article via Infotrieve]
  9. Hafezi, F., Abegg, M., Grimm, C., Wenzel, A., Munz, K., Sturmer, J., Farber, D. B., and Reme, C. E. (1998) Investig. Ophthalmol. Vis. Sci. 39, 2239–2244[Abstract/Free Full Text]
  10. Ogilvie, J. M., Speck, J. D., and Lett, J. M. (2000) Exp. Neurol. 161, 676–685[CrossRef][Medline] [Order article via Infotrieve]
  11. Wu, J., Trogadis, J., and Bremner, R. (2001) Mol. Vis. 7, 101–106[Medline] [Order article via Infotrieve]
  12. LaVail, M. M., Yasumura, D., Matthes, M. T., Lau-Villacorta, C., Unoki, K., Sung, C. H., and Steinberg, R. H. (1998) Investig. Ophthalmol. Vis. Sci. 39, 592–602[Abstract/Free Full Text]
  13. Ahuja, P., Caffe, A. R., Holmqvist, I., Soderpalm, A. K., Singh, D. P., Shinohara, T., and van Veen, T. (2001) Neuroreport 12, 2951–2955[CrossRef][Medline] [Order article via Infotrieve]
  14. Caffe, A. R., Soderpalm, A. K., Holmqvist, I., and van Veen, T. (2001) Investig. Ophthalmol. Vis. Sci. 42, 275–282[Abstract/Free Full Text]
  15. Frasson, M., Sahel, J. A., Fabre, M., Simonutti, M., Dreyfus, H., and Picaud, S. (1999) Nat. Med. 5, 1183–1187[CrossRef][Medline] [Order article via Infotrieve]
  16. Doonan, F., Donovan, M., and Cotter, T. G. (2003) J. Neurosci. 23, 5723–5731[Abstract/Free Full Text]
  17. Kim, D. H., Kim, J. A., Choi, J. S., and Joo, C. K. (2002) Ophthalmic Res. 34, 150–157[CrossRef][Medline] [Order article via Infotrieve]
  18. Zhang, L., Miles, M. F., and Aldape, K. D. (2003) Nat. Biotechnol. 21, 818–821[CrossRef][Medline] [Order article via Infotrieve]
  19. Everitt, B. S., and Dunn, G. (2001) Applied Multivariate Data Analysis, 2nd Ed., pp. 125–158, Arnold Publishing, London
  20. Lord, P. W., Stevens, R. D., Brass, A., and Goble, C. A. (2003) Pac. Symp. Biocomput. 601–612
  21. Daniel, W. W. (1999) Biostatistics: a Foundation for Analysis in the Health Sciences, pp. 600–606, John Wiley & Sons, New York
  22. Clarke, G., Collins, R. A., Leavitt, B. R., Andrews, D. F., Hayden, M. R., Lumsden, C. J., and McInnes, R. R. (2000) Nature 406, 195–199[CrossRef][Medline] [Order article via Infotrieve]
  23. Mitas, M., Mikhitarian, K., Walters, C., Baron, P. L., Elliott, B. M., Brothers, T. E., Robison, J. G., Metcalf, J. S., Palesch, Y. Y., Zhang, Z., Gillanders, W. E., and Cole, D. J. (2001) Int. J. Cancer 93, 162–171[CrossRef][Medline] [Order article via Infotrieve]
  24. Strettoi, E., Porciatti, V., Falsini, B., Pignatelli, V., and Rossi, C. (2002) J. Neurosci. 22, 5492–5504[Abstract/Free Full Text]
  25. Jeon, C. J., Strettoi, E., and Masland, R. H. (1998) J. Neurosci. 18, 8936–8946[Abstract/Free Full Text]
  26. Kendziorski, C. M., Zhang, Y., Lan, H., and Attie, A. D. (2003) Biostatistics 4, 465–477[Abstract]
  27. Carter-Dawson, L. D., LaVail, M. M., and Sidman, R. L. (1978) Investig. Ophthalmol. Vis. Sci. 17, 489–498[Abstract/Free Full Text]
  28. Calestani, C., Rast, J. P., and Davidson, E. H. (2003) Development 130, 4587–4596[Abstract/Free Full Text]
  29. Gold, D., Coombes, K., Medhane, D., Ramaswamy, A., Ju, Z., Strong, L., Koo, J. S., and Kapoor, M. (2004) BMC Genomics 5, 2[CrossRef][Medline] [Order article via Infotrieve]
  30. Li, C., and Wong, W. H. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, 31–36[Abstract/Free Full Text]
  31. Sanyal, S., and Hawkins, R. K. (1981) Exp. Eye Res. 33, 213–222[CrossRef][Medline] [Order article via Infotrieve]
  32. Thiel, G., and Cibelli, G. (2002) J. Cell. Physiol. 193, 287–292[CrossRef][Medline] [Order article via Infotrieve]
  33. Paron, I., D'Elia, A., D'Ambrosio, C., Scaloni, A., D'Aurizio, F., Prescott, A., Damante, G., and Tell, G. (2004) Biochem. J. 378, 929–937[CrossRef][Medline] [Order article via Infotrieve]
  34. Patapoutian, A., and Reichardt, L. F. (2000) Curr. Opin. Neurobiol. 10, 392–399[CrossRef][Medline] [Order article via Infotrieve]
  35. Nicotera, P., and Orrenius, S. (1998) Cell Calcium 23, 173–180[CrossRef][Medline] [Order article via Infotrieve]
  36. Woodruff, M. L., Wang, Z., Chung, H. Y., Redmond, T. M., Fain, G. L., and Lem, J. (2003) Nat. Genet. 35, 158–164[CrossRef][Medline] [Order article via Infotrieve]
  37. Fox, D. A., Poblenz, A. T., and He, L. (1999) Ann. N. Y. Acad. Sci. 893, 282–285[CrossRef][Medline] [Order article via Infotrieve]
  38. Gobin, S. J., van Zutphen, M., Westerheide, S. D., Boss, J. M., and van den Elsen, P. J. (2001) J. Immunol. 167, 5175–5184[Abstract/Free Full Text]
  39. Elward, K., and Gasque, P. (2003) Mol. Immunol. 40, 85–94[CrossRef][Medline] [Order article via Infotrieve]
  40. Colley, N. J., and Hall, M. O. (1986) Exp. Eye Res. 42, 323–329[Medline] [Order article via Infotrieve]
  41. Hall, M. O., Obin, M. S., Prieto, A. L., Burgess, B. L., and Abrams, T. A. (2002) Exp. Eye Res. 75, 391–400[CrossRef][Medline] [Order article via Infotrieve]
  42. Feng, W., Yasumura, D., Matthes, M. T., LaVail, M. M., and Vollrath, D. (2002) J. Biol. Chem. 277, 17016–17022[Abstract/Free Full Text]
  43. Morgan, B. P., and Meri, S. (1994) Springer Semin. Immunopathol. 15, 369–396[CrossRef][Medline] [Order article via Infotrieve]
  44. Van Beek, J., Chan, P., Bernaudin, M., Petit, E., MacKenzie, E. T., and Fontaine, M. (2000) Glia 31, 39–50[CrossRef][Medline] [Order article via Infotrieve]
  45. Pasinetti, G. M., Hassler, M., Stone, D., and Finch, C. E. (1999) Synapse 31, 278–284[Medline] [Order article via Infotrieve]
  46. Grewal, R. P., Morgan, T. E., and Finch, C. E. (1999) Neurosci. Lett. 271, 65–67[CrossRef][Medline] [Order article via Infotrieve]
  47. Jomary, C., Darrow, R. M., Wong, P., Organisciak, D. T., Neal, M. J., and Jones, S. E. (1999) J. Neurochem. 72, 1923–1929[Medline] [Order article via Infotrieve]
  48. Crabb, J. W., Miyagi, M., Gu, X., Shadrach, K., West, K. A., Sakaguchi, H., Kamei, M., Hasan, A., Yan, L., Rayborn, M. E., Salomon, R. G., and Hollyfield, J. G. (2002) Proc. Natl. Acad. Sci. U. S. A. 99, 14682–14687[Abstract/Free Full Text]
  49. Kim, Y. S., Kim, Y. H., Cheon, E. W., Park, J. M., Yoo, J. M., Kang, S. S., Cho, G. J., and Choi, W. S. (2003) Brain Res. 976, 53–59[Medline] [Order article via Infotrieve]
  50. Luhe, A., Hildebrand, H., Bach, U., Dingermann, T., and Ahr, H. J. (2003) Toxicol. Sci. 73, 315–328[Abstract/Free Full Text]
  51. Wang, X., Phelan, S. A., Forsman-Semb, K., Taylor, E. F., Petros, C., Brown, A., Lerner, C. P., and Paigen, B. (2003) J. Biol. Chem. 278, 25179–25190[Abstract/Free Full Text]
  52. Wasselius, J., Hakansson, K., Johansson, K., Abrahamson, M., and Ehinger, B. (2001) Investig. Ophthalmol. Vis. Sci. 42, 1901–1906[Abstract/Free Full Text]
  53. Katunuma, N., Matsui, A., Le, Q. T., Utsumi, K., Salvesen, G., and Ohashi, A. (2001) Adv. Enzyme Regul. 41, 237–250[CrossRef][Medline] [Order article via Infotrieve]
  54. Nakanishi, H. (2003) Ageing Res. Rev. 2, 367–381[CrossRef][Medline] [Order article via Infotrieve]
  55. Fujii, J., and Ikeda, Y. (2002) Redox Rep. 7, 123–130[CrossRef][Medline] [Order article via Infotrieve]
  56. Marc, R. E., Jones, B. W., Watt, C. B., and Strettoi, E. (2003) Prog. Retinal Eye Res. 22, 607–655[CrossRef][Medline] [Order article via Infotrieve]
  57. Kaczmarek, L., Lapinska-Dzwonek, J., and Szymczak, S. (2002) EMBO J. 21, 6643–6648[CrossRef][Medline] [Order article via Infotrieve]
  58. Agapova, O. A., Kaufman, P. L., Lucarelli, M. J., Gabelt, B. T., and Hernandez, M. R. (2003) Brain Res. 967, 132–143[CrossRef][Medline] [Order article via Infotrieve]
  59. Cao, W., Li, F., Steinberg, R. H., and LaVail, M. M. (2001) Exp. Eye Res. 72, 591–604[CrossRef][Medline] [Order article via Infotrieve]
  60. Nico, B., Frigeri, A., Nicchia, G. P., Quondamatteo, F., Herken, R., Errede, M., Ribatti, D., Svelto, M., and Roncali, L. (2001) J. Cell Sci. 114, 1297–1307[Abstract]
  61. Vizuete, M. L., Venero, J. L., Vargas, C., Ilundain, A. A., Echevarria, M., Machado, A., and Cano, J. (1999) Neurobiol. Dis. 6, 245–258[CrossRef][Medline] [Order article via Infotrieve]
  62. Nguyen, M. D., Julien, J. P., and Rivest, S. (2002) Nat. Rev. Neurosci. 3, 216–227[CrossRef][Medline] [Order article via Infotrieve]
  63. Li, J., Patil, R. V., and Verkman, A. S. (2002) Investig. Ophthalmol. Vis. Sci. 43, 573–579[Abstract/Free Full Text]
  64. Sharma, A. K., and Rohrer, B. (2004) J. Biol. Chem. 279, 35564–35572[Abstract/Free Full Text]
  65. Pinto, F. R., Rohrer, B., Cowart, A., and Almeida, J. (2004) Bioinformatics, in press

Add to CiteULike CiteULike   Add to Complore Complore   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us   Add to Digg Digg   Add to Reddit Reddit   Add to Technorati Technorati    What's this?


This article has been cited by other articles:


Home page
 IOVSHome page
J. Sancho-Pelluz, K. A. Wunderlich, U. Rauch, F. J. Romero, T. van Veen, G. A. Limb, P. R. Crocker, and M.-T. Perez
Sialoadhesin Expression in Intact Degenerating Retinas and Following Transplantation
Invest. Ophthalmol. Vis. Sci., December 1, 2008; 49(12): 5602 - 5610.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
R. J. Davis, J. Tosi, K. M. Janisch, J. M. Kasanuki, N.-K. Wang, J. Kong, I. Tsui, M. Cilluffo, M. L. Woodruff, G. L. Fain, et al.
Functional Rescue of Degenerating Photoreceptors in Mice Homozygous for a Hypomorphic cGMP Phosphodiesterase 6 b Allele (Pde6bH620Q)
Invest. Ophthalmol. Vis. Sci., November 1, 2008; 49(11): 5067 - 5076.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
S. Ahuja, P. Ahuja-Jensen, L. E. Johnson, A. R. Caffe, M. Abrahamson, P. A. R. Ekstrom, and T. van Veen
rd1 Mouse Retina Shows an Imbalance in the Activity of Cysteine Protease Cathepsins and Their Endogenous Inhibitor Cystatin C
Invest. Ophthalmol. Vis. Sci., March 1, 2008; 49(3): 1089 - 1096.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
B. Rohrer, Y. Guo, K. Kunchithapautham, and G. S. Gilkeson
Eliminating Complement Factor D Reduces Photoreceptor Susceptibility to Light-Induced Damage
Invest. Ophthalmol. Vis. Sci., November 1, 2007; 48(11): 5282 - 5289.
[Abstract] [Full Text] [PDF]

Home page
 Nucleic Acids ResHome page
S. Calza, W. Raffelsberger, A. Ploner, J. Sahel, T. Leveillard, and Y. Pawitan
Filtering genes to improve sensitivity in oligonucleotide microarray data analysis
Nucleic Acids Res., August 28, 2007; (2007) gkm537v2.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
R. E. Swiderski, D. Y. Nishimura, R. F. Mullins, M. A. Olvera, J. L. Ross, J. Huang, E. M. Stone, and V. C. Sheffield
Gene Expression Analysis of Photoreceptor Cell Loss in Bbs4-Knockout Mice Reveals an Early Stress Gene Response and Photoreceptor Cell Damage
Invest. Ophthalmol. Vis. Sci., July 1, 2007; 48(7): 3329 - 3340.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
R. B. Wilson, K. Kunchithapautham, and B. Rohrer
Paradoxical Role of BDNF: BDNF+/- Retinas Are Protected against Light Damage-Mediated Stress
Invest. Ophthalmol. Vis. Sci., June 1, 2007; 48(6): 2877 - 2886.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
C. Punzo and C. Cepko
Cellular Responses to Photoreceptor Death in the rd1 Mouse Model of Retinal Degeneration
Invest. Ophthalmol. Vis. Sci., February 1, 2007; 48(2): 849 - 857.
[Abstract] [Full Text] [PDF]

Home page
 Mol. Cell. ProteomicsHome page
S. M. Hauck, P. A. R. Ekstrom, P. Ahuja-Jensen, S. Suppmann, F. Paquet-Durand, T. van Veen, and M. Ueffing
Differential Modification of Phosducin Protein in Degenerating rd1 Retina Is Associated with Constitutively Active Ca2+/Calmodulin Kinase II in Rod Outer Segments
Mol. Cell. Proteomics, February 1, 2006; 5(2): 324 - 336.
[Abstract] [Full Text] [PDF]

Home page
 IOVSHome page
F. Doonan, M. Donovan, and T. G. Cotter
Activation of Multiple Pathways during Photoreceptor Apoptosis in the rd Mouse
Invest. Ophthalmol. Vis. Sci., October 1, 2005; 46(10): 3530 - 3538.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Data
Right arrow All Versions of this Article:
279/40/41903    most recent
M405085200v1
Right arrow Submit a Letter to Editor
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowRequest Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Rohrer, B.
Right arrow Articles by Almeida, J. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rohrer, B.
Right arrow Articles by Almeida, J. S.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
 All ASBMB Journals   Molecular and Cellular Proteomics 
 Journal of Lipid Research   ASBMB Today 
Copyright © 2004 by the American Society for Biochemistry and Molecular Biology.
Advertisement
spacer
Advertisement
Advertisement