Molecular Vision 2016; 22:1318-1331
Received 11 November 2015 | Accepted 08 November 2016 | Published 10 November 2016
1Department of Ophthalmology, Hamilton Eye Institute, The University of Tennessee Health Science Center, Memphis, TN; 2Department of Anatomy and Neurobiology, The University of Tennessee Health Science Center, Memphis, TN; 3Department of Ophthalmology, the Second Xiangya Hospital, Central South University, Changsha, Hunan, China; 4Department of Genetics, Genomics and Informatics, The University of Tennessee Health Science Center, Memphis, TN; 5Department of Pharmaceutical Sciences, The University of Tennessee Health Science Center, Memphis, TN
Correspondence to: Monica Jablosnki, The University of Tennessee Health Science Center, Hamilton Eye Institute, 930 Madison Avenue, Suite 758, Memphis, TN 38163, Phone: (901) 448 7572; FAX: (901) 448 5028; email: firstname.lastname@example.org
Purpose: Photoreceptor degenerative diseases are among the leading causes of vision loss. Although the causative genetic mutations are often known, mechanisms leading to photoreceptor degeneration remain poorly defined. We have previously demonstrated that the photoreceptor membrane-associated protein XAP-1 antigen is a product of the HSPA5 gene. In this study, we used systems genetic methods, statistical modeling, and immunostaining to identify and analyze candidate genes that modulate Hspa5 expression in the retina.
Methods: Quantitative trait locus (QTL) mapping was used to map the genomic region that regulates Hspa5 in the cross between C57BL/6J X DBA/2J mice (BXD) genetic reference panel. The stepwise refinement of candidate genes was based on expression QTL mapping, gene expression correlation analyses (direct and partial), and analysis of regional sequence variants. The subcellular localization of candidate proteins and HSPA5 in mouse and human retinas was evaluated by immunohistochemistry. Differences in the localization of extracellular HSPA5 were assessed between healthy human donor and atrophic age-related macular degeneration (AMD) donor eyes.
Results: In the eyes of healthy mice, extracellular HSPA5 was confined to the area around the cone photoreceptor outer segments. Mapping variation in Hspa5 mRNA expression levels in the retina revealed a statistically significant trans-acting expression QTL (eQTL) on Chromosome 2 (Chr 2) and a suggestive locus on Chr 15. Sulf2 on Chr 2 was the strongest candidate gene based on partial correlation analysis, Pearson correlation with Hspa5, expression levels in the retina, a missense variant in exon 14, and its reported function in the extracellular matrix and interphotoreceptor matrix. SULF2 is localized to the rod and cone photoreceptors in both human and mouse retinas. In human retinas with no pathology, extracellular HSPA5 was localized around many cones within the macular area. In contrast, fewer HSPA5-immunopositive cones were observed in the retinas from AMD donors.
Conclusions: We identified Sulf2 as a candidate gene modulating the Hspa5 expression in the retina. The preferential loss of HSPA5 in the interphotoreceptor matrix around cone photoreceptors in atrophic AMD retinas opens up new avenues for exploring the changes in interphotoreceptor matrix (IPM) that are associated with macular disease.
Heat shock proteins protect and maintain cell viability primarily by acting as molecular chaperones under stressful conditions . HSPA5 (GRP78, also known as binding immunoglobulin protein [BiP]) is a well-established endoplasmic reticulum (ER) chaperone. It is involved in many cellular processes, including the translocation of newly synthesized polypeptides across the ER membrane, facilitating the folding and assembly of proteins, targeting misfolded proteins for ER-associated protein degradation, regulating calcium homeostasis, and serving as an ER stress sensor [2-5]. An upregulation of HSPA5 is an indication of ER stress in various mouse models of retinal degeneration, such as light-induced retinal degeneration , the rd1 mouse , and retinal detachment .
In addition to being an ER resident protein, HSPA5 has also been characterized as a soluble protein localized to the interphotoreceptor
matrix (IPM) surrounding the cones in pig and mouse retinas and to the IPM surrounding both rods and cones in Xenopus laevis [9-11]. Previously, our laboratory identified the
Systems genetics integrates and interprets high-throughput quantitative molecular data to model cellular responses and disease outcomes using a genetic reference population [13-16], such as the BXD family of advanced crosses, which is the largest and best-characterized murine genetic reference population available. These strains were generated by crossing two widely used inbred strains of mice: C57BL/6J (B6) and DBA2/J (D2) [17,18]. Most molecular and genetic data for the BXD family has been archived in the GeneNetwork, including many datasets that are relevant to the genetics and genomics of the eye and central visual system [19-23].
The purpose of this investigation was to identify the genes that modulate the expression of Hspa5 in the mouse retina. A secondary goal was to determine the localization pattern of HSPA5 in healthy and diseased human retinas. In this study, we used various systems genetics methods and ex vivo studies to identify Sulf2 as the best candidate gene to modulate the expression of Hspa5 in the retina. HSPA5 is confined to the IPM surrounding the cone photoreceptors in human donor eyes that have no ocular pathology, yet it is nearly absent in atrophic age-related macular degeneration (AMD) eyes. We hypothesized that Hspa5 may play an important role in maintaining cone photoreceptor structural integrity, and that its expression was modulated by Sulf2.
This study was approved by The Animal Care and Use Committee at The University of Tennessee Health Science Center (UTHSC). Mice were handled in a manner consistent with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, Public Health Service Policy on Humane Care and Use of Laboratory Animals). A total of 80 strains (n = 326) of BXD mice from both sexes were used for the systems genetics study. Three BXD100 mice (2 months old, including both sexes) were used in the immunohistochemistry study. No abnormal phenotypes were detected in the eyes of these mice. More importantly, they lacked the pigmentary dispersion phenotype that is present in DBA/2J mice. The animals were maintained on a 12 h: 12 h light-dark cycle, at a temperature from 20 °C to 24 °C and were allowed food and water ad libitum.
Our research and protocol for the collection of human retinas was conducted in accordance with the tenets of the Declaration of Helsinki and approved by University of Tennessee Health Science Center Institutional Review Board (IRB). The control donor eyes had no history of eye disease; the age and sex of the donors were 47 years (M), 59 years (F), 69 years (F), 77 years (F), 87 years (M), and 93 years (M). The age-matched human donor eyes had a clinical diagnosis of atrophic AMD; the age and sex of the donors were70 years (M), 79 years (M), 87 years (M), and 94 years (F). All eyes were obtained from the Mid South Eye Bank and the Lions Eye Bank of Oregon. All eyes were procured and processed in compliance with the Medical Standards of the Eye Bank Association of America (EBAA) and Government regulations. Eyes were fixed within 5 h of death in formalin solution. Anterior segment structures were removed and the fundus was photographed. The clinical diagnosis of each eye was confirmed by an ophthalmologist. Full thickness biopsies of the macula region centered on the fovea, and the far peripheral temporal retinas of each eye were removed with a 6-mm diameter punch (Miltex, York, PA).
To obtain the light level immunolocalization of the antigens of interest, the mouse and human retinas were embedded in low molecular weight agarose and sectioned at a thickness of 50 µm. Immunofluorescence staining for both intracellular and extracellular HSPA5 was performed according to our previously published protocol. Briefly, tissue sections were blocked with 10% goat serum and permeabilized with 0.1% Triton X-100 or left un-permeabilized to differentiate intracellular and extracellular localization patterns, respectively based on our previously published protocol . Tissue sections were then either labeled with anti-XAP-1 (HSPA5; Clone 3D2, monoclonal antibody, Developmental Studies Hybridoma Bank, University of Iowa, IA, 1:1 dilution) or anti-sulfatase 2 antibody (SULF2; rabbit polyclonal IgG, Santa Cruz Biotechnology, Dallas TX, 1:200 dilution). This was followed by Alexa Fluor 488 or 568-conjugated secondary antibody (Invitrogen, Carlsbad, CA; 1:200 dilution) to detect the antigen of interest. Peanut agglutinin (PNA)-tagged with Alexa Fluor 488 or 568 (Molecular Probes, Eugene, OR; 1:25 dilution) and wheat germ agglutinin (WGA)-tagged with Alexa Fluor 633 (Molecular Probes; 1:50 dilution) were used to differentiate the cone and rod photoreceptors, respectively. TO-PRO3 iodide (1:4000 dilution; Invitrogen) was used to label the nuclei. Sections were viewed and images were obtained using a Nikon C1 confocal microscope within the Imaging Core Facility in the Hamilton Eye Institute. All microscope settings, including laser levels and gain, were held constant to allow for relative comparisons of signal intensity within and between experiments.
Electron microscopic (EM) immunohistochemical analyses were performed using our previously published protocol . Briefly, formalin fixed human donor eyes were embedded in LR White (Electron Microscopy Sciences, Hatfield, PA). Ultra-thin sections were collected on mesh nickel grids. The sections were blocked with 10% normal goat serum (Vector Laboratories, Burlingame, CA) and incubated in anti-XAP1 antibody (1:1 dilution) at 4 °C overnight, followed by incubation with a gold-conjugated secondary antibody (1:10 dilution, 10 nm gold particles; Sigma-Aldrich, St. Louis, MO). Sections were washed and images were captured with an electron microscope (JEM 200EX II; JEOL, Tokyo, Japan).
The Hamilton Eye Institute (HEI) Retina Database contains the data analysis of 346 Illumina Sentrix® Mouse Whole Genome-6 version 2.0 arrays (Illumina, San Diego, CA). This database contains the retinal transcriptome profiles of 80 strains of BXD mice aged between 2 and 4 months. We used the Normal HEI Retina dataset available on the GeneNetwork to perform QTL analysis and define genetic networks. These data are globally normalized to produce arrays that have a mean of 8 and a standard deviation of 2 [23,24].
On the Illumina array, Hspa5 is represented by a single probe set (ILMN_1255561) that hybridizes to the proximal 3′ untranslated region (UTR) of Hspa5 on Chromosome 2 (Chr 2) at 34.631669 Mb on the plus strand. The eQTL mapping was performed using our previously published methods [19,21].
QTL mapping was carried out using the WebQTL module available in GeneNetwork. Simple interval mapping was performed at regular intervals to identify eQTLs that regulate Hspa5 expression. The level of significance was estimated by permutations analyses. Mapping analyses produced a likelihood ratio statistic (LRS) score, providing us with a measure of the strength of linkage between Hspa5 expression level variation and genotype markers.
Using the tools available on the GeneNetwork, partial correlation analyses were performed using our previously published methods [25-28]. In the current study, the partial correlation feature on the GeneNetwork was used to identify a candidate gene that modulates Hspa5 expression in the retina. A partial correlation is defined as the relationship between a primary variable and a target variable after the influence of one or more related variables are removed . In this case, the expression of Hspa5 was the primary variable and the expression levels of the candidate regulatory genes within the retina database were the target variables. The trans-effect of the eQTL on Chr 2 was mathematically controlled using the markers rs3143843 (Chr 2: 169.778380) and rs3673248 (Chr 2: 172.849974), which straddle the trans-eQTL to strengthen the relationship between the Hspa5 expression and the potential candidate genes. Using the partial correlation function, we generated the top 100 gene correlates of Hspa5. To evaluate the relationship between Hspa5 and the QTL region, this list of 100 correlates was analyzed by constructing a heat map in which the more intense colors indicated the chromosomal regions with comparatively high linkage statistics. The top candidate genes with a biologic function or a localization pattern associated with the extracellular or interphotoreceptor matrix were selected for further analyses.
The Pearson correlation coefficients with Hspa5 were calculated for the candidate genes using the tools available in the GeneNetwork. Candidate genes with significant Pearson correlation (threshold≥0.5) values with Hspa5 were examined for the presence of non-synonymous single nucleotide polymorphisms (SNPs) or insertions and deletions (InDels). A detailed list of polymorphisms between the D2 and B6 parental strains is freely available on the GeneNetwork and the original sequence data are available at 327618. A second heat map that included only the candidate modulatory genes was also generated to evaluate their relationship with Hspa5. All correlation values for candidate gene prioritization were adjusted for multiple comparisons using the Bonferroni correction.
Previously, we determined that the XAP-1 antibody labels extracellular HSPA5 in Xenopus retinas [10,12]. To determine which cell types in the mouse retina were labeled by the HSPA5 antibody, we performed immunohistochemical analysis. In mouse retinas, HSPA5 preferentially co-localized with PNA around the cone photoreceptors. A minor amount of HSPA5 was co-localized with WGA around the rods (Figure 1A-D). To differentiate between extracellular and intercellular HSPA5, mouse retina sections were un-permeabilized or permeabilized, respectively, before immunohistochemistry was performed. In the non-permeabilized tissue, intense HSPA5 immunolabeling co-localized with PNA, which indicated that extracellular HSPA5 expression was predominantly present in the IPM surrounding the cone photoreceptors (Figure 2A-D). After permeabilization, a similar pattern of extracellular HSPA5 localization was observed in the IPM surrounding the cone photoreceptors (Figure 2E-H). No intracellular labeling was detected using the XAP-1 antibody.
Using the HEI retina database available on the GeneNetwork, we determined that there was a significant variation in the level of Hspa5 transcript in the retina across the BXD strains (Figure 3A). The average expression level of Hspa5 was 7.6±0.0 and the range was 6.9±0.2 (BXD80) to 8.4±0.3 (BXD102). Simple interval mapping for Hspa5 revealed a significant trans-eQTL with the likelihood ratio statistics (LRS) of 18.4 (genome wide significance at p≤0.05) on Chr 2 with a peak between 170 and 173 Mb. A suggestive peak on Chr 15 between 85 and 90 Mb with an LRS of 14.3 was also observed (Figure 3B).
To identify the candidate genes that could modulate the expression of Hspa5 in the retina, we performed a partial correlation analysis within the GeneNetwork. By mathematically controlling for the QTL on the Chr 2 peak, we generated a list of the top 2,000 correlates of Hspa5. To select the top candidate genes within this list that modulated the variation in the expression of Hspa5 in the retina, we used the following criteria:
• The gene was expressed in the retina;
• The gene was located within 5 Mb of the QTL peaks on Chr 2 and Chr 15 that were identified using simple interval mapping;
• The expression of the gene exhibited a high (≥0.5) correlation with Hspa5; and
• The gene had functional relevance to the extracellular matrix, including the IPM.
Of the 71 genes residing within the trans-eQTLs on Chr 2 and Chr 15, the following genes fulfilled the selection criteria: Dpm1 and Sulf2 (located in the trans-eQTL interval on Chr 2), and Fbln1, Glt8d3, and Pmm1 (located in the eQTL interval on Chr 15). Table 1 provides the details of each gene.
By performing a direct Pearson correlation calculation between the variation in the levels of Hspa5 and that of the six candidate genes, we determined that the variation in the levels of Sulf2, Pmm1, and Fbln1 had very high correlation coefficients of 0.775 (Adjusted [adj]. p = 6.0×10−16), −0.725 (adj. p = 1.7×10−12), and 0.611 (adj. p = 4.4×10−08), respectively, with variation in the expression of Hspa5. Variation of B4galt5, Dpm1, and Glt8d3 had moderate correlation coefficients of 0.486 (adj. p = 6.9×10−05), −0.486 (adj. p = 6.6×10-0.5), and 0.493 (adj. p = 4.8×10−05), respectively, with variation in the level of Hspa5.
A heat map that was generated using the top 100 transcripts with the highest Pearson correlation coefficients with Hspa5 revealed tight bands on Chr 2 and Chr 15 at the location of the QTL peaks found when mapping the Hspa5 expression (Figure 4A). A second heat map generated with only Hspa5 and the six candidate genes had a banding pattern similar to the heat map for the top 100 transcripts (Figure 4B). These data indicated that both groups of genes shared common signature QTLs, which suggested that they were all part of the same genetic signature network.
To further evaluate the strength of these six candidate genes, we determined if there were any sequence variants between the parental strains. The data in the UCSC Genome Browser on Mouse Assembly (UCSC mm9, July 2007) showed no missense/nonsense SNPs in B4galt5, Dpm1, Fbln1, Glt8d3, or Pmm1. One non-synonymous SNP was present in Sulf2 in exon 14 (Table 1). The SNP in Sulf2 resulted in a change in the amino acid sequence in which an arginine was replaced by a lysine, which can lead to the synthesis of an altered protein product (Table 1). Only Sulf2 met all of the criteria for the selection of a candidate gene.
To strengthen Sulf2 as a candidate gene modifier of Hspa5, we analyzed the localization pattern of SULF2 by immunofluorescence in cross sections of mouse retinas. In the tissue lacking permeabilization, SULF2 was predominantly present in the extracellular space around the cone outer segments (OS; Figure 5A-D), which demonstrated that a small amount of secreted SULF2 was present in the IPM around the cone photoreceptors. After tissue permeabilization, an intense immunoreactivity for SULF2 was observed in the outer nuclear layer (ONL), as well as around the inner segments (IS) and OS of the cone photoreceptors (Figure 5E-H). A similar pattern was detected in the normal human donor retina sections (Figure 5I-P, representative image).
In our previously published study, we determined that HSPA5 was associated with the IPM surrounding the rod and cone photoreceptors in Xenopus laevis retinas . However, in mouse retinas, extracellular HSPA5 was preferentially localized to the cone photoreceptors  (Figure 1 and Figure 2). Building on this observation, we sought to determine if the localization pattern of extracellular HSPA5 was conserved in mammals, and if macular disease affected the presentation of HSPA5. In human donor retinas from eyes with no pathology, intense HSPA5 expression was co-localized with PNA in the macula (Figure 6A-D). This indicated that in the human retina, HSPA5 immunolabeling was more abundant in the IPM surrounding the cone photoreceptors than in the IPM surrounding the rod photoreceptors. The cross sectional view of the macular retina showed the HSPA5 expression around the cone photoreceptors along the entire IPM (Figure 6E-H). The results were representative of all the donor eyes.
To better characterize the localization of HSPA5 in human retinas, we performed an EM immunolocalization analysis. At the EM level, HSPA5 was preferentially associated with flocculent material on the surface of the plasma membrane of the cone photoreceptors. In contrast, there was minimal immunolabeling around the rod photoreceptors (Figure 6I,J).
In ocular diseases, such as age-related macular degeneration, the primary cellular target leading to vision loss is the cone photoreceptors. To compare the localization pattern of HSPA5 in photoreceptors from human donor eyes with no pathology and age-matched human donor eyes with a diagnosis of atrophic AMD, immunohistochemistry was performed on retinal punches obtained from the macular and peripheral regions. The en face images represented the staining pattern from all the donor eyes used in the study. In the macula of the normal human donor retinas, HSPA5 was present in abundance surrounding the cones, but was detected at low levels around the rod photoreceptors (Figure 7A-C). In contrast, HSPA5-positive cones were less frequent throughout the peripheral retina (Figure 7D-F) in eyes with no ocular pathology. Atrophic AMD donor retinas exhibited an aberrant immunolabeling pattern in which there were very few HSPA5 immunopositive cells in both the macular (Figure 7G-I) and peripheral areas (Figure 7J-L) compared with that of retinas from the age-matched control eyes.
In this investigation, we confirmed that extracellular HSPA5 was a cone photoreceptor-specific marker in mouse retinas. It was also a cone marker in the macula of human donor eyes that lacked AMD-like pathology. In contrast, HSPA5 immunolabeling was drastically reduced in the donor eyes from atrophic AMD patients. We also integrated a novel application of expression genetics with biochemical analysis to dissect the molecular pathways, and identified Sulf2 as a candidate gene that modulated Hspa5 expression in the retina.
A major cause of adult blindness in developed countries is the progressive dysfunction and death of retinal photoreceptors. Photoreceptors are structurally unique photosensitive cells of the neural retina . The extracellular space between the photoreceptors of the retina and the apical surface of the retinal pigment epithelium is occupied by the IPM . Research has demonstrated that some of the soluble and insoluble components of the IPM may contribute to visual pigment chromophore exchange, retinal adhesion, metabolite trafficking, and growth factor presentation [31-33]. Because this matrix resides in a key location and is putatively crucial in supporting photoreceptor function, documenting its structure and its components, and understanding the gene products found in the IPM can lead us closer to identifying the causes of photoreceptor degeneration.
HSPA5 is a central regulator of ER homeostasis due to its multiple functional roles in protein folding, ER calcium binding, and controlling the activation of transmembrane ER stress sensors. ER stress induction of HSPA5 represents a major pro-survival arm of the unfolded protein response (UPR). Most neurodegenerative disorders, including Parkinson’s disease, Alzheimer’s diseases, and progressive retinal degeneration, are characterized by the activation of the UPR and the modified expression of HSPA5 [6,7,34-36]. However, the function played by the extracellular form of HSPA5 in photoreceptor degeneration is unknown.
In the current study, we used immunohistochemistry and the XAP-1 antibody to demonstrate that extracellular HSPA5 binds to the IPM surrounding the cone photoreceptors both with and without tissue permeabilization. We confirmed that the antibody was specific for extracellular HSPA5 in mouse and human retinas. The presence of HSPA5 in the IPM around the cones in healthy human retinas and its near absence in atrophic AMD retinas suggested that HSPA5 might play a role in the cone structural integrity in the macula. However, additional studies to target the disruption of Hspa5 synthesis are required to confirm this hypothesis.
To determine which candidate genes modulate Hspa5 transcript variation, we performed simple interval mapping. Our mapping studies in BXD mice indicated a single highly genome wide significant trans-eQTL on Chr 2 and a suggestive genome wide trans-eQTL on Chr 15, both of which suggested that Hspa5 expression was modulated, either directly or indirectly, by a different gene in the retina. We pursued a parallel strategy of heat map analyses and partial correlation analyses to corroborate the findings of our QTL analysis for Chr 2 and 15. Our results suggested that one or more transcripts that predominantly regulate Hspa5 in a trans manner resided within these strong eQTLs. We focused on six positional and biologically relevant genes, B4galt5, Dpm1, Sulf2, Fbln1, Glt8d3, and Pmm1, located on Chr 2 and Chr 15 under the QTL peaks. All of these genes have been shown to play key roles in proteoglycan maintenance and cellular signaling in extracellular matrices [37-42]. Of the six genes, only Sulf2 had a non-synonymous SNP in the mouse genome. Interestingly, a recent genome wide association study by the AMD Gene Consortium revealed 19 AMD loci, one of which showed enrichment for genes involved in extracellular matrix remodeling . This provided further evidence to support our assertion that genes modulating IPM structure or function may contribute to the pathogenesis of photoreceptor degeneration in AMD.
SULF2 is an extracellular neutral-pH sulfatase. It provides a post-synthetic mechanism for the regulation of heparan sulfate proteoglycan (HSPG) function by removing 6-O-sulfate groups from the HSPG. SULF2 can thereby modulate the signaling pathways in the retina associated with Wnt-dependent signaling , as well as the activities of the glial cell line-derived neurotrophic factor (GDNF) , sonic hedgehog (SHH) , vascular endothelial growth factor (VEGF) , fibroblast growth factor 2 (FGF2) , and Noggin .
Our immunohistochemistry data demonstrated that secreted SULF2 was present in the extracellular space surrounding the cone inner and outer segments. This expression pattern raised the possibility that the secreted form of SULF2 binds to the IPM surrounding the cone inner and outer segments. An interesting implication of this observation is that the SULF2 may regulate heat shock protein signaling in the cone photoreceptors, although the mechanism is not known.
The loss of HSPA5 around the cone photoreceptors in atrophic AMD retinas provides new avenues for exploring the changes in IPM that accompany cone photoreceptor degeneration. Because HSPA5 binds to the IPM surrounding the cone photoreceptors in human retinas, it may be important in organizing the IPM, which may support photoreceptor function in higher vertebrates. In addition, our data suggested that SULF2 might play a role in modulating HSPA5 expression. This finding can narrow our focus in our attempts to understand the mechanisms of cell-intrinsic programs that alter the IPM. These future outcomes can aid our understanding of why the macula, which is a region rich in cone photoreceptors, is especially vulnerable to photoreceptor degeneration. Our findings provide the first evidence that Hspa5 may be involved in cone photoreceptor health.
The authors acknowledge the financial support from NEI Grant EY021200, NEI Core Grant P30EY013080, and an unrestricted grant from Research to Prevent Blindness, Inc. We also thank Dr. Eldon Geisert, Dr. Lu Lu, and Mr. Bill Orr for their assistance in generating the BXD microarray datasets that were used in these analyses. The XAP-1 monoclonal antibody (Clone 3D2 developed by D.S. Sakaguchi and W.A. Harris) was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242.