Defective photoreceptor phagocytosis in a mouse model of enhanced S-cone syndrome causes progressive retinal degeneration

Abstract

Enhanced S-cone syndrome (ESCS), featuring an excess number of S cones, manifests as a progressive retinal degeneration that leads to blindness. Here, through optical imaging, we identified an abnormal interface between photoreceptors and the retinal pigment epithelium (RPE) in 9 patients with ESCS. The neural retina leucine zipper transcription factor-knockout (Nrl^−/−) mouse model demonstrates many phenotypic features of human ESCS, including unstable S-cone-positive photoreceptors. Using massively parallel RNA sequencing, we identified 6203 differentially expressed transcripts between wild-type (Wt) and Nrl^−/− mouse retinas, with 6 highly significant differentially expressed genes of the Pax, Notch, and Wnt canonical pathways. Changes were also obvious in expression of 30 genes involved in the visual cycle and 3 key genes in photoreceptor phagocytosis. Novel high-resolution (100 nm) imaging and reconstruction of Nrl^−/− retinas revealed an abnormal packing of photoreceptors that contributed to buildup of photoreceptor deposits. Furthermore, lack of phagosomes in the RPE layer of Nrl^−/− retina revealed impairment in phagocytosis. Cultured RPE cells from Wt and Nrl^−/− mice illustrated that the phagocytotic defect was attributable to the aberrant interface between ESCS photoreceptors and the RPE. Overcoming the retinal phagocytosis defect could arrest the progressive degenerative component of this disease.—Mustafi, D., Kevany, B. M., Genoud, C., Okano, K., Cideciyan, A. V., Sumaroka, A., Roman, A. J., Jacobson, S. G. Engel, A., Adams, M. D., Palczewski, K. Defective photoreceptor phagocytosis in a mouse model of enhanced S-cone syndrome causes progressive retinal degeneration.

The retina of all vertebrates contains rod photoreceptors for dim light environments and cone photoreceptors for brighter environments; cones are further divided into short-wavelength-sensitive (S) and long- and middle-wavelength-sensitive (L/M) subtypes. These photoreceptors, which are organized into mosaic structures with characteristic rod/cone ratio in retinal position and species specific manner, are organized to provide useful visual sensation for the animals during their entire life span. Thus, a normal retina is controlled by a multitude of interacting factors that determine the precise developmental organization and lifetime maintenance of interconnected neurons for optimal visual function (1, 2). Disruption of these complex interactions during development or in the mature retina can give rise to cellular pathology, mainly manifesting as loss of vision. In addition to the development and maintenance processes of the photoreceptor itself, homeostatic processes in neighboring cells contribute to photoreceptor health. A key example of this support function is continuous phagocytosis of shed photoreceptor discs by the neighboring retinal pigment epithelium (RPE; refs. 3–5).

Enhanced S-cone syndrome (ESCS) is a human visual disorder first recognized for its unique feature of showing increased S-cone vision. With noninvasive studies, ESCS was demonstrated to result in supernormal sensitivity to blue colors and an excess number of S cones, normally the minority photoreceptor in the human retinal mosaic consisting mainly of rods and L/M cones. In contrast, rod and L/M-cone vision are reduced in ESCS. A hypothesis emerging from these results was that abnormal retinal development causes ESCS involving a disturbance in photoreceptor cell specification (6–8). A search for the causative genes ensued, and mutations in most patients mapped to the gene encoding the human photoreceptor-specific nuclear receptor, NR2E3, while a few mapped to the neural retina leucine zipper, NRL, gene (9–11). NR2E3 and NRL are now known to play key roles in the regulatory transcriptional networks controlling photoreceptor cell fate (1). Identification of the causative genes, however, did not account for the degenerative component of this disease.

Knockout of the Nrl transcription factor in mice produces a retina overpopulated with S-cone-like photoreceptors along with absence of rod photoreceptors. Precise identification of changes in transcriptional networks in the Nrl^−/− mouse retina and resulting aberrant composition of expressed proteins would likely provide information concerning critical factors that dictate cone-like photoreceptor maintenance/survival as well as proper retinal lamination. Previous studies have also suggested abnormal association between photoreceptors and the RPE in Nrl^−/− mice (12, 13), and differences in RPE appearance such as discontinuity and depigmentation compared with normal RPE have been noted in human postmortem donor ESCS retinas (14, 15). Our own electron microscopy (EM) revealed aberrant bulbous outer segment (OS) structures containing abnormal internal structures (16).

Early stages of photoreceptor development and maintenance involve Notch signaling through basic helix loop helix (bHLH) transcription factors (17, 18), as well as through Hedgehog, which also converges on downstream Notch targets (19). The interplay of these factors, among others, dictates the proper transcriptional environment for photoreceptor maintenance, but the precise relationship between them is not yet fully elucidated. Previous studies of the retinal tissue transcriptome in mouse models employed microarray-based serial analysis of gene expression (SAGE; refs. 20–23), expressed sequence tag (EST; refs. 24, 25), and hybridization microarrays (26–28), but a complete transcriptome analysis could not be achieved by these approaches. Recent developments have revolutionized transcriptome analysis. RNA-sequencing (RNA-Seq) technology now combines the advantages of previous large-scale RNA analytical methods with a larger dynamic range of detection (29, 30), and already has provided new genetic insights in different model systems (31–33).

We now report RNA-Seq transcriptome analyses and quantification of transcript levels to identify genes differentially expressed in mature wild-type (Wt) and Nrl^−/− mouse eyes and retinas. We identified changes in expression of the set of genes involved in the formation/maintenance of cone-like photoreceptors of the Nrl^−/− retina and noted changes in key homeostatic genes involved in OS disc phagocytosis and toxic metabolite removal that suggested a potential molecular mechanism for the degenerative component of ESCS. The genetic findings were complemented by a set of new imaging technologies (34, 35) used to assess the disrupted retinal layering in Nrl^−/− mice, the abnormal photoreceptor-RPE interface, and the process of phagocytosis (3). Imaging and biochemical experiments revealed fewer detectable phagosomes in Nrl^−/− as compared to Wt retina. Compromised phagocytosis was attributed to a defect in the aberrant Nrl^−/− photoreceptors rather than phagocytes because RPE cell cultures from Nrl^−/− retina exhibited phagocytosis comparable to RPE cell cultures from Wt animals. These results are consistent with the hypothesis that impairment in phagocytosis stemming from aberrant ESCS photoreceptors leads to retinal degeneration in both patients with ESCS and the Nrl^−/− mouse model.

MATERIALS AND METHODS

Materials

All chemicals, unless otherwise stated, were purchased from Sigma-Aldrich (St. Louis, MO, USA). Reagents for cDNA library preparation for Illumina sequencing, unless otherwise indicated, were bought from Illumina (San Diego, CA, USA). Reagents for cDNA synthesis and quantitative real-time PCR (RT-PCR) were obtained from Applied Biosystems (Foster City, CA, USA). Primary antibodies anti-red/green pigment opsin and anti-blue opsin were acquired from Chemicon International (Billerica, MA, USA), anti-phosphatidylserine (PS) and anti-annexin V were purchased from Abcam (Cambridge, MA, USA), peanut aggulutinin (PNA) was obtained from Invitrogen (Carlsbad, CA, USA), and anti-rhodopsin was generated in the K.P. laboratory from hybridoma cells (36). Cy3 and Alexa488 conjugated secondary antibodies were acquired from Jackson Immuno-Research (West Grove, PA, USA) or Invitrogen. Nuclear staining was achieved with Hoechst, 4′,6-diamidino-2-phenylindole (DAPI), or quinolinium, 4-[3-(3-methyl-2(3H)-benzothiazolylidene)-1-propenyl]-1-[3-(trimethylammonio)propyl]-diiodide (ToPro3; Invitrogen).

Human studies

All patients with ESCS studied had mutations in the NR2E3 gene (11). Informed consent was obtained, and procedures followed the Declaration of Helsinki guidelines and were approved by the institutional review board at the University of Pennsylvania. Patients had complete ocular examinations including kinetic perimetry quantified by published methods (37). Psychophysical thresholds were measured with a modified automated perimeter (Humphrey Field Analyzer; Humphrey Instruments, San Leandro, CA, USA) to determine S-cone function (440-nm stimulus on a yellow background, 170 cd/m²), L/M cone function (650-nm stimuli, dark-adapted) and rod function (500-nm stimuli, dark-adapted). Details of visual function techniques and analyses have been described previously (6, 7, 15). Spectral-domain (SD) optical coherence tomography (OCT) was used (RTVue-100; Optovue Inc., Fremont, CA, USA) with published recording and analysis techniques to perform retinal cross-sectional imaging (38–41). RPE lipofuscin imaging was performed as described previously (38, 42).

Animals

Mice were housed in the animal facility at the School of Medicine, Case Western Reserve University (CWRU; Cleveland, OH, USA), where they were maintained on a standard chow diet in a 12-h light-dark cycle (light ∼10 lux). Wt mice on C57BL/6 background were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). Nrl-deficient mice in the C57BL/6 background were from Dr. Anand Swaroop (University of Michigan, Ann Arbor, MI, USA; ref. 12). Genotyping of mice was done by PCR with primers NRL-A (5′-gtgttccttggctggaaaga-3′) and NRL-B (5′-ctgttcactgtgggctttca-3′) for Wt and NRL-KO1 (5′-tgaatacagggacgacacca-3′) and NRL-KO2 (5′-gttctaattccatcagaagctgac-3′) for targeted deletion of the Nrl gene. All animal procedures and experiments were performed in accordance with U.S. animal protection laws and were approved by the CWRU Animal Care Committees and conformed to both the recommendations of the American Veterinary Medical Association Panel on Euthanasia and the Association of Research for Vision and Ophthalmology.

Ultra-high-resolution SD-OCT

Nine Wt and 9 Nrl-deficient mice aged 4 wk were each anesthetized by intraperitoneal injection of a mixture (20 μl/g body weight) containing ketamine (6 mg/ml) and xylazine (0.44 mg/ml) in 10 mM sodium phosphate (pH 7.2) and 100 mM NaCl. Pupils were dilated with 1% tropicamide. Mice were placed in a specialized holder to permit ultra-high-resolution SD-OCT (Bioptigen, Research Triangle Park, NC, USA) for in vivo imaging of mouse retinas at λ = 870 nm with a superluminescent diode. Each 2-dimensional (2-D) B scan was acquired at a speed of 1000 scans/s, and each final SD-OCT image was an average of 3 individual B-scans. Three-dimensional (3-D) scans were taken around the optic nerve with a scanning radius of 1.6 mm. Images were postprocessed by using commercial Bioptigen software and ImageJ (U.S. National Institutes of Health, Bethesda, MD, USA) (43).

Library preparation for sequencing

Mice were euthanized by cervical dislocation. Eyes were enucleated and immediately placed in RNA later stabilization reagent (Qiagen, Valencia, CA, USA) to preserve RNA content and integrity (44) for whole-eye runs. Alternatively, the retina was rapidly dissected out and similarly preserved. One mouse eye or 2 retinas were homogenized at once and passed through a QIAShredder column (Qiagen) as per manufacturer's directions to further homogenize the eye tissues. Total RNA was then purified by using the RNeasy Mini Kit (Qiagen) with on column DNase treatment (Qiagen) as per manufacturer's directions. Poly(A) RNA was isolated with the Oligotex kit (Qiagen) as per the manufacturer's instructions. Pooled total RNA samples of 5 Wt and 5 Nrl^−/− female mice at 4 wk of age were used for the whole-eye library preparation, and pooled total RNA samples from 5 Wt and 5 Nrl^−/− female mice at 4 wk of age were used for the retina library preparation.

For first-strand cDNA synthesis, instructions from the SuperScript III kit protocol (Invitrogen) were followed. About 400–450 ng of isolated poly(A) RNA was mixed with 50 ng of random primers and 1 mM deoxyribonucleotide triphosphate (dNTP), incubated at 65°C for 5 min, and then placed on ice for 5 min. A reaction mixture comprising 5 mM MgCl₂, 10 mM DTT, 40 U RNaseOUT, and 200 U SuperScript III reverse transcriptase was added to the initial mix to achieve a total volume of 20 μl. The mixture was incubated at 25°C for 10 min, followed by 50 min at 50°C. The reaction was terminated at 85°C for 5 min and then chilled on ice for 10 min. At this point, 2 U RNase H was added, and the mix was incubated at 37°C for 20 min. The first-strand cDNA synthesis reaction was immediately used for second-strand synthesis. To the first-strand product, 300 μM dNTP, Escherichia coli DNA polymerase I buffer, and water were added to obtain a total volume of 95 μl and allowed to incubate on ice for 10 min. Then, 0.05 U E. coli DNA polymerase I (New England Biolabs, Beverly, MA, USA) was added, and the mixture was incubated at 16°C for 2.5 h. The resulting double-stranded cDNA was purified with the Wizard SV Gel and PCR Clean-up System (Promega, Madison, WI, USA), eluted in 100 μl nuclease-free water, and then fragmented by the Covaris S2 instrument (Covaris, Woburn, MA, USA) to generate ∼200-bp fragments as follows: 10% duty cycle, intensity of 5, 100 cycles/burst, with a bath temperature of 7.7°C and an acoustic power of 24 W.

The Illumina library was prepared according to the manufacturer's instructions and purified using the Wizard SV Gel and PCR Clean-up system. Overhangs were converted into blunt ends with T4 DNA polymerase and Klenow DNA polymerase by incubating the mixed sample at 20°C for 30 min. cDNA was purified and eluted in 32 μl of nuclease-free water with the Wizard Plus Minipreps DNA purification system (Promega). The purified sample was then mixed with Klenow fragment (3′ to 5′ exo minus) and incubated at 37°C for 30 min to add an A base to the 3′ end of the blunt phosphorylated DNA fragments. The cDNA was then purified and eluted in 23 μl of nuclease-free water with the Wizard Plus Minipreps system. Eluted DNA was mixed with Illumina Adapter Oligo mix and T4 DNA ligase and incubated at room temperature for 15 min to ligate adapters to the ends of the DNA fragments to prepare them for hybridization to the flow cell. cDNA then was purified and eluted in 10 μl of nuclease-free water with the Wizard Plus Minipreps system (Promega). cDNA templates were purified by running samples on a 1% agarose gel at 120 V for 60 min and excising the region of the gel in the 200-bp range. The 200-bp cDNA enriched fragments were purified and eluted in 30 μl of nuclease-free water with the Wizard Plus Minipreps system. cDNA in the library was then amplified by a 15 cycle PCR with two primers that annealed to the ends of the adapters. The amplified cDNA was purified and eluted in 30 μl of nuclease-free water with the Wizard Plus Minipreps system. The size, purity and concentration of the final library was checked with the Bio-Rad Experion DNA specific chip prior to sequencing by using the Illumina Genome Analyzer. The concentration of the sample was also measured using 1 μl of purified sample with the Qubit Quantitation Platform (Invitrogen) to estimate loading conditions for the Illumina Cluster Station.

Genome analyzer RNA-Seq runs, read mapping, and quantification

Each library was run on 3 lanes of the Genome Analyzer II in the Genomics Core Facility at CWRU by using 36- or 49-bp single-end sequencing. The numbers of mapped single reads from different experiments were 38,166,142 from the whole-eye technical replicates; 45,431,330 from the Wt whole eye; 66,643,381 from the Nrl^−/− eye; 85,159,191 from the Wt retina; and 104,081,398 from the Nrl^−/− retina. Technical replicates of the whole eye entailed running the sample library preparation on independent lanes on different day runs and analyzing them separately. Primary data transformation included image analysis, intensity scoring, base calling, and alignment, all carried out with Illumina pipeline software running on Linux. Image analysis identified distinct clusters and created digital intensity files describing the signal intensity of each cluster per cycle. Signal intensity profiles for each cluster were used to call bases, and quality scores for each base call were calculated for alignment. Efficient Large-Scale Alignment of Nucleotide Databases (ELAND; Illumina) was then used for read mapping to the University of California–Santa Cruz (UCSC; Santa Cruz, CA, USA) mouse genome assembly and transcript annotation (mm9) (45). For each read, ELAND determined the position in the genome to which the read substrings matched with a maximum of 2 errors. Base quality scores and the positions of the mismatches in a candidate alignment were used to calculate a probability score for each candidate, with the highest probability score indicating the best candidate. Eligible reads were defined by having a unique alignment to the genome or a single most probable alignment to the genome. Other reads with failed quality control measures were not used in subsequent processing. The ELAND alignment was loaded onto Consensus Assessment of Sequence and Variation (CASAVA; Illumina) software for calculation of fragments per kilobase of exon model per million mapped reads (FPKM) statistics by gene, transcript, and exon. CASAVA counted the number of bases that belonged to exons and genes, and the numbers of bases that fell into the exonic regions of each gene were summed to obtain gene level counts. Normalized values were then calculated as FPKM. The output for CASAVA was visualized with the GenomeStudio RNA Sequencing Module (Illumina), which allowed comparison between the samples based on the CASAVA output files. The raw files (fastq) and processed FPKM value data can be found online at the National Center for Biotechnology Information gene expression omnibus site with the series accession number GSE29752 (http://www.ncbi.nlm.nih.gov/projects/geo/query/acc.cgi?acc=GSE29752).

RT-PCR

Isolated total retinal RNA (2 μg) from 3 pooled Wt and Nrl^−/−samples was converted to cDNA with the High Capacity RNA-to-cDNA kit (Applied Biosystems). RT-PCR was done with TaqMan chemistry and Assays on Demand probes (Applied Biosystems) for mouse Abca4 (Mm00492035_m1), Atp8a2 (Mm00443740_m1), Atoh7 (Mm00844064_s1), Bmp15 (Mm00437797_m1), Crx (Mm00483995_m1), Egr1 (Mm00656724_m1), Eya1 (Mm00438796_m1), Gdf11 (Mm01159973_m1), Neurod1 (Mm01946604_s1), Notch1 (Mm00435249_m1), Prdm1 (Mm01187284_m1), Opn1sw (Mm00432058_m1), Otx2 (Mm00446859_m1), Six6 (Mm00488257_m1), Six6os1 (Mm01290652_m1), Thrb (Mm00437044_m1), Rxrg (Mm00436411_m1), and Wnt9b (Mm00457102_m1). The 18S rRNA (4319413E) probe set (Applied Biosystems) was used as the endogenous control. All real-time experiments were done in triplicate with the ABI Step-One Plus qRT-PCR machine (Applied Biosystems). Fold changes were calculated based on differences in threshold cycles (C_t) between the Nrl^−/− and Wt samples after normalization to 18s rRNA.

Analysis of data

Genes were categorized using AmiGO 1.8 software (http://www.geneontology.org/). Fold differences in RNA-Seq experiments were compared by examining the ratio of FPKM between Wt and Nrl^−/− sample runs. A 1.5-fold or greater change in threshold was used to identify differential expression, thereby allowing comparisons with previous experiments. Statistical significance of fold expression changes in RT-PCR were analyzed with Microsoft Excel software (Microsoft, Redmond, WA, USA). P values were calculated from a Student's 2-tailed t test to confirm that fold changes were statistically significant (P<0.05). Power analysis was calculated to detect the sample size required to detect significant changes with RNA-Seq using a 1.5-fold difference cutoff. The parameters were detecting a 0.33=FPKM difference (a 1.5-decreased fold of 1 FPKM, representing an expressed transcript, is 0.67, yielding a difference of 0.33 FPKM), a standard deviation of 10% in the FPKM value (estimated from technical replicates), an α value of 0.05, and a β value of 0.10, with the ratio of Wt to Nrl^−/− samples as 1.

Cryosectioning

Twenty Wt and 20 Nrl-deficient mice aged 4 wk were sacrificed 1.5 h after lights went on in the morning, a time when phagocytosis of photoreceptor OS in Wt is maximal. Eye cups were dissected out under a surgical microscope and incubated in 4% paraformaldehyde overnight at 4°C. Eye cups then were dehydrated in successive solutions of 5, 10, 15, and 20% sucrose in phosphate buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 4.3 mM Na₃HPO₄, and 1.4 mM KH₂PO₄, pH 7.3) for 30 min each on a shaker. Subsequently, eyes were placed in a 1:1 solution of 20% sucrose in PBS:Optical Cutting Temperature Compound (Tissue-Tek-Sakura, Torrence, CA, USA) for 30 min on a shaker, when the solution was replaced and the eye cups were kept at 4°C overnight. Eye cups were frozen the next day by placing them in cryomolds and submerging them into 2-methyl-butane in a tank of liquid nitrogen. Cryoblocks were then cut with a Leica cryosectioner (Leica Microsystems, Wetzlar, Germany) and 10-μm sections around the optic nerve were collected on glass slides for immunohistochemical staining.

Immunohistochemistry

All procedures used were reported previously (46, 47). Cross-sections of mouse eyecups were incubated with primary antibodies, namely anti-rabbit red/green pigment opsin, anti-rabbit opsin blue anti-mouse rhodopsin, anti-mouse PS, anti-rabbit annexin V, and PNA. Signals were detected with either Cy3-conjugated secondary antibody or Alexa488-conjugated secondary antibody. Nuclear staining was achieved with DAPI. Sections were analyzed with a Leica 6000B microscope.

Whole-mount retinal confocal microscopy

Ten Wt and 10 Nrl-deficient 4-wk-old mice were sacrificed, and eye whole mounts were prepared and incubated overnight with primary antibodies; i.e., anti-rabbit red/green pigment opsin, anti-rabbit opsin blue anti-mouse rhodopsin, anti-mouse PS, and PNA. Signals were detected with either Cy3-conjugated secondary antibody or Alexa488-conjugated secondary antibody. Nuclear staining was achieved with ToPro3. Thick Z stacks (30–40 μm) were collected at ×40 view with 1 μm between each slice and visualized with a Leica SP5 confocal microscope. Obtained images were postprocessed with ImageJ to adjust contrast and brightness.

Scanning EM (SEM)

Seven Wt and 9 Nrl-deficient mice 4 wk of age were sacrificed, and their retinas and the RPE were separated and fixed in 2.5% glutaraldehyde, 0.1 M cacodylate buffer, and 2% sucrose (pH 7.4) for 24 h. Samples were washed in 0.1 M cacodylate buffer and 2% sucrose, fixed with 1% OsO₄ in washing buffer, dehydrated with ethanol, dried by a critical point drying method (48), and sputter-coated with a 5–10 nm gold layer. Samples were imaged with a JSF-6300F scanning electron microscope (Jeol, Akishima, Japan) at the University of Washington Department of Pathology (Seattle, WA, USA). The emission current was set to enable acquisition of backscattered electron scanning images at ×2000 to ×10,000.

Transmission EM (TEM)

Five Wt and 5 Nrl-deficient mice aged 4 wk and 5 Wt and 5 Nrl-deficient mice aged 8 wk were sacrificed at 1.5 h after lights went on in the morning. Eyes were removed, and whole eye cups were dissected out under a surgical microscope and placed in 4% paraformaldehyde at 37°C for 4 h. Eye cups then were rinsed in PBS and incubated in a 1:1 solution of 2% OsO₄:3% potassium ferrocyanide for 1 h. This was followed by incubation in a new mixture of 2% OsO₄:3% potassium ferrocyanide for 1 h, after which eye cups were washed in filtered water and placed in 0.25% uranyl acetate overnight at 4°C. Eye cups were dehydrated the next day for 10 min each in sequential solutions of 30, 50, 75, 85, 95, and 100% ethanol in water, then for 15 min each in sequential solutions of 50, 75, and 100% propylene oxide in ethanol, followed by 2 h in 30% epon in propylene oxide, and finally kept in 50% epon in propylene oxide overnight. Next day, eye cups were placed in 75% epon in propylene oxide for 4 h, then in 100% epon for 2 h under vacuum, and finally in a mold with epon kept at 73°C for 4 d to enable cross-linking. Then blocks were cut with a microtome, and ultrathin sections (0.07 μm) were stained with uranyl acetate and adsorbed onto carbon grids. A Tecnai T12 electron microscope (FEI, Eindhoven, Netherlands) operating at 120 kV with a tungsten filament was used for final imaging.

Serial block face SEM (SBF-SEM), SEM data analyses, and 3-D reconstruction

The same blocks prepared for TEM were used for SBF-SEM (34). To prepare a sample, we used an ultramicrotome (Leica UCT) and a diamond knife (Diatome, Hatfield, PA, USA), and trimmed the block so that only resin-embedded tissue of the region of interest remained. The final tissue block was adhered by conductive carbon cement to an aluminum SEM stub to preserve conductivity. The prepared sample was fixed on the microtome (3View; Gatan, Pleasanton, CA, USA) attached on the door of the scanning electron microscope (Quanta 200 FEG ESEM; FEI). Cutting was initiated in the evacuated specimen chamber. To perform serial cutting of the block face, a 100-nm slice was cut from the face with a diamond knife, and the freshly cut surface of the block was imaged from the backscattered electron signal. This process was repeated sequentially in an automatic computer-controlled fashion to collect 500 successive images over ∼12 h. Imaging was performed at an accelerating voltage of 3 kV in a low-vacuum mode (0.23 Torr) at 4096- × 4096-pixel resolution at a rate of 3 μs/pixel.

After serial sectioning, images were opened with Fiji-win32 (a version of ImageJ; http://imagej.nih.gov/ij/index.html) and merged to form a stack. The stack was registered and aligned to account for any drift that may have occurred over the time course of sectioning. The registered stack then was opened using the Reconstruct program (49), and structural elements were mapped to provide 3-D reconstructions.

Phagocytosis assays of RPE cell cultures

RPE was isolated from 10- to 12-d-old Wt and Nrl^−/− mice as described previously (50). Briefly, eyes were removed from animals and washed twice in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) supplemented with nonessential amino acids. Eyes were incubated in 2% dispase (Invitrogen) solution for 45 min in a 37°C water bath with occasional tube inversion. Eyes were washed twice in cold DMEM plus streptomycin/penicillin (Invitrogen), 10% fetal bovine serum (Invitrogen), and 20 mM HEPES (pH 7.2). Eyes were enucleated, and the cornea, lens, and iris were removed. Eye cups were incubated in DMEM plus streptomycin/penicillin, 10% fetal bovine serum, and 20 mM HEPES (pH 7.2) in a 37°C incubator for 15 min to facilitate removal of the neural retina. After removal of neural retina, sheets of continuous RPE were peeled from choroid and pipetted into a tube containing DMEM plus streptomycin/penicillin and 10% fetal bovine serum. RPE sheets were subsequently filtered over a 40-μm cell strainer (Fisher Scientific, Pittsburgh, PA, USA) to remove contaminating cell types. Sheets were spun at 200 g for 3 min and then resuspended in DMEM plus streptomycin/penicillin and 10% fetal bovine serum before gentle disruption by pipetting. Disrupted cells were seeded onto 24-well 0.4-μm transwell permeable supports (Corning Inc., Corning, NY, USA) with the RPE from ∼2 eyes/well to allow polarization of cells. Cells were grown for 5–6 d at 37°C, 5% CO₂ before use in phagocytosis assays.

For phagocytosis challenge assays, photoreceptor OS membranes were isolated from Wt and Nrl^−/−mice. Photoreceptor OS membranes from Wt mice were isolated as described previously (51), whereas OS membranes from Nrl^−/− mice were obtained by a similar protocol with a 10–100% continuous gradient of OptiPrep (Nycomed, Norway) to improve the yield. Photoreceptor OS membranes isolated from Wt and Nrl^−/− mice were covalently labeled with fluorescein isothiocyanate (FITC; Invitrogen) by using established protocols (52). FITC-labeled photoreceptor OS membranes were resuspended in DMEM plus streptomycin/penicillin, 10% fetal bovine serum, and 2.5% sucrose; 50 μl of this mixture was added to the top of the transwell membrane, while 700 μl of DMEM alone plus streptomycin/penicillin and 10% fetal bovine serum was added to the well of the plate. Assay mixtures were incubated in the dark at 37°C for 1 h. Cells were washed 3 times with PBS plus 1 mM MgCl₂ and 0.2 mM CaCl₂ (PBS-MC). FITC fluorescence of externally bound photoreceptor OS was quenched by incubation with 0.2% trypan blue (Invitrogen) for 10 min, after which cells were washed 3 times with PBS-MC. Cells were fixed with ice-cold methanol for 5 min at 4°C followed by 3% paraformaldehyde at room temperature for 10 min. Cells were washed 2 times with PBS-MC and permeablized with 0.2% Triton X-100 in PBS for 30 min at room temperature. Nuclear staining was performed by incubation with Hoechst stain (10 μm final) for 30 min at room temperature. Cells were washed in PBS-MC an additional 3 times. Transwell membranes were removed from supports and mounted onto microscope slides with ProLong Gold Antifade Agent (Invitrogen).

RESULTS

Key features of human ESCS: relationship to the Nrl^−/− mouse model

The diagnosis of ESCS is based on a quantitative comparison of S-cone and L/M-cone visual and/or retinal parameters (6, 7, 53, 54). Normally, L/M-cone vision is far more sensitive than S-cone vision, but in ESCS, surprisingly, it is just the opposite. ESCS manifests heightened sensitivity of S-cone vision relative to L/M-cone vision in the presence of little or no rod function. Comparison between a 13-yr-old boy with ESCS [patient 1 (P1)] and a healthy control subject exemplifies the increased S-cone function and reduced L/M-cone vision compared with results in a healthy subject (Fig. 1A). The sensitivity difference is positive (Fig. 1A, right panels, denoted by + symbols at test loci) in contrast to negative numbers when normal L/M-cone sensitivities are subtracted from S cones (7). P2, another patient at 2 different ages (40 and 48 yr), shows the same diagnostic difference at loci with persistent function despite reduced S- and L/M-cone vision because of progressive retinal degeneration (Fig. 1A). Progressive degenerative retinopathy of ESCS is further illustrated by plotted kinetic visual field data from 9 patients followed longitudinally for at least a decade (Fig. 1B). Relatively full visual fields tend to become reduced with age, leaving only central and peripheral islands separated by blind spots (Fig. 1B, insets).

Figure 1.

Key features of human ESCS and the Nrl^−/− mouse model. A) Topographic maps of visual sensitivity for S cones (left panels) and L/M cones (middle panels) in a healthy subject and patients with ESCS at different ages. Normally, S-cone sensitivities are less than those of L/M cones. In the patients with ESCS, S-cone sensitivities are greater than L/M-cone sensitivities at the same loci. P1, a 13-yr-old patient, has supernormal S-cone sensitivities; in P2, at 2 ages (40 and 48 yr), S-cone sensitivities are normal or subnormal but still greater than those of colocalized L/M cones; in the 8-yr interval, a progressive loss of vision was found. Loci showing this positive (enhanced) difference are marked (+, right panels). B) Kinetic visual field extent for a large bright target (V-4e) as a function of age in 9 patients with ESCS with longitudinal measurements spanning at least a decade. A decline with age seen in a proportion of these patients is attributable to progressive retinal degeneration. Inset: kinetic fields at 2 ages in ESCS P3 illustrates loss of field extent over a 19-yr interval. C) Cross-sectional OCT scans of retinal architecture along >10 mm of the horizontal meridian through the fovea (F) of ESCS P1 (top panel) compared with a healthy subject (top panel). Outer nuclear layer (ONL) and inner segment (IS) thicknesses are labeled at right. Rectangles show temporal retinal region quantified in next panel. D) Longitudinal reflectivity profiles (LRPs) of outer retinal lamina in 2 representative healthy subjects (left panel) and 3 patients with ESCS (middle panel). Identifiable layers are labeled and colored. Among notable LRP features are thicker ONL and IS layers in these patients and abnormal structures between the IS and RPE. Quantification of IS thickness in 6 patients with patients with ESCS vs. control subjects (right panel) showed a significant difference. E) En face autofluorescence images of the central fundus of 2 patients with ESCS illustrating hyperautofluorescent features (white dots). Insets: cross-sectional images in colocalized regions show dysmorphology with intraretinal hyperreflective lesions. F) OCT of Wt and Nrl^−/− mouse retinas illustrates phenotypic changes in the Nrl^−/− retina resembling those in human ESCS disease; e.g., a hyperreflective RPE-photoreceptor interface and nuclear layer rosette formation (asterisk). ONL thicknesses are labeled at right. G) Three-dimensional spectral domain OCT of Wt and Nrl^−/− mice indicates retinal disorganization caused by rosette formation as well as an abnormal photoreceptor-RPE interaction in the Nrl^−/− retina.

In vivo histopathology in early stages of ESCS shows a hyperthick photoreceptor outer nuclear layer (ONL) in the more central retina but a variably reduced ONL with increasing retinal eccentricity (Fig. 1C). In the extracentral retina of patients with ESCS, there can be noticeable dysmorphology of the ONL, with intraretinal hyperreflective lesions extending to the inner retina (for example, in P1). Longitudinal reflectivity profiles of the outer retinal laminar architecture in 2 healthy subjects at 2.5 mm from the fovea show layers of ONL, photoreceptor inner segments (ISs), rod OSs, cone OSs and RPE (Fig. 1D). Three patients with ESCS (ages 17, 13, 31; Fig. 1D, middle panel, left to right) definitely have a thickened ONL (55) and appear to have a thickened IS layer as well. When the normal IS layer thickness (n=6, ages 8–29; mean±2sd, 27±2.8 μm) is compared with IS thickness in 6 patients with ESCS (ages 13–31; mean±2sd, 35±6.7 μm), the IS layer in ESCS is significantly thicker (t test, P<0.001). This finding may relate to the longer IS in human S cones seen in morphological studies (56). The interface between photoreceptor OS and RPE is also abnormal and ill-defined; i.e., the normal stereotypical multipeaked profile is not evident in patients with ESCS (Fig. 1D). The reason for the abnormal interface between photoreceptors and RPE found in these imaging studies is not known. En face imaging further illustrated abnormalities in patients with ESCS. In normal subjects, autofluorescence (AF) emissions on short-wavelength excitation are dominated by spatially homogeneous lipofuscin granules accumulated in the RPE (42, 57), but patients with ESCS exhibit hyperautofluorescent loci in the macular and midperipheral retinal regions. Cross-sectional imaging of colocalized regions shows dysmorphology of the ONL extending to the inner retina (Fig. 1E, insets; dysmorphology is also seen in the temporal retina of P1, Fig. 1C). Abnormal deposition of retinal or RPE fluorophores, unmasking of natural fluorophores by localized loss of RPE melanin, or lipofuscin-laden macrophages, alone or in concert, could contribute to these hyperautofluorescent features (58–60).

The Nrl^−/− mice exhibit many phenotypic features of human ESCS disease and thus provide a model to study ESCS pathophysiology. Retinal degeneration is not yet evident at 4 wk of age in the Nrl^−/− murine eye. However, cross-sectional optical imaging by SD-OCT revealed abnormal retinal lamination in the Nrl^−/− retina compared with Wt retina (Fig. 1F). Three-dimensional reconstructions allowed visualization of abnormal intraretinal hyperreflective lesions, presumed to be rosettes, and how this distortion affected the retinal laminar architecture (Fig. 1G). Plastic block and cryosectioning of retinas further highlighted the dynamic changes resulting from the excessive S-cone photoreceptor population at higher resolution. Compared with normal Wt retina (Supplemental Fig. S1A–D), the Nrl^−/− retina displayed aberrant photoreceptor packing and abnormal association with the RPE (Supplemental Fig. S1E–H).

Although these data demonstrate comparable retinal degeneration in human and mouse models of ESCS, the fundamental cellular aberrations that cause this pathophysiology are unknown. We initially approached the molecular basis of this degeneration with a comprehensive global sequencing approach.

Transcriptome analysis by RNA-Seq of retinas from Wt and Nrl^−/− mice

By sequencing mature Wt and Nrl^−/− ocular tissues, we determined global changes resulting from knockout of the Nrl transcription factor. The reproducibility of RNA-Seq for murine eye was verified by carrying out technical replicates of Wt eye samples (R²=0.98; Supplemental Fig. S2A), indicating a minimal variability from run to run. This reproducibility indicated that any differences between tissue samples were not inherent to instrument read errors. Wt retinal tissue (Fig. 2A) generated 11,677 unique transcripts at a level of 1 FPKM or higher (Fig. 2B) whereas Nrl^−/− retinal tissue (Fig. 2C) generated 11,778 unique transcripts at a level of 1 FPKM or higher (Fig. 2D). As expected, in Wt and Nrl^−/− retinal tissues, a large proportion of transcripts had no annotated function. A complete categorization of Gene Ontology (GO) terms from these tissues is shown in Table 1. RNA-Seq expression analysis on whole-eye tissues of Wt and Nrl^−/−mice is summarized in Supplemental Fig. S2C–F. The RNA-Seq findings revealed that, across experiments, the number of transcripts detected between runs was not drastically different, nor was their categorization, indicating there is not a gross difference between Wt and Nrl^−/− retinas and eyes, but rather more subtle differences that require RNA-Seq single-gene resolution to tease out their precise differential expression patterns.

Figure 2.

RNA-Seq of Wt and Nrl^−/− retinas reveals new differentially expressed genes arising from transcriptional misregulation. A) Wt retina RNA-Seq run, represented by robust Nrl transcript detection, detected 11,677 transcripts at 1 FPKM or higher. B) Pie chart shows breakdown of GO term categories to which the transcripts are assigned, with the number of transcripts in each category indicated. C) Single-base resolution of the RNA-Seq run of Nrl^−/− retina reveals ablation of Nrl transcript detection in the regions of exon 2 and 3 where there is a neomycin cassette. D) Run detected 11,778 transcripts at 1 FPKM or higher; pie chart shows breakdown of GO term categories to which the transcripts are assigned, with the number of transcripts in each category indicated. E) RNA-Seq runs of Wt and Nrl^−/− retina are plotted to show their differential expression pattern. Plots of Log FPKM of the retinal runs of Wt and Nrl^−/− illustrate that, whereas the majority of reads fall along the line representing equal expression, a range of transcripts falls either above or below the line that represent differentially expressed transcripts. Cnga1, Esrrb, Gnat1, Kcnj14, Nr2e3, Nrl, Rho, Slc24a1, and Susd3 (arrows) are among the highest expressed transcripts in the Wt retina, whereas Clca3, Cngb3, Fabp7, Gnat2, Gnb3, Gngt2, Opn1sw, Pde6c, Pde6h, and Six6os1 (arrows) are among the highest expressed transcripts in the Nrl^−/− retina RNA-Seq run. F) RT-PCR validated differential expression patterns detected by RNA-Seq. To validate differences from RNA-Seq experiments, retinal tissue from Wt and Nrl^−/− mice was used for RT-PCR using probes against well characterized targets from previous studies as well as newly identified targets from the current RNA-Seq study. RT-PCR results validated the RNA-Seq differential expression pattern that ranged from those genes that were highly differentially expressed (Egr1, Opn1sw) to those with more subtle differential expression (Gdf11, Otx2, Thrb) and even those without a significant fold change (Crx). Blue bars indicate RT-PCR of retina; red bars, RNA-Seq of whole eye; green bars, RNA-Seq of retina. Current RNA-Seq experiment, when compared to 2 previous microarray studies looking at differential expression between Wt and Nrl^−/− retina, reveals more comprehensive and quantitative data. G) RNA-Seq data reveal 3659 unique transcripts up-regulated in the Nrl^−/−retina compared to previous data sets, indicating a considerable amount of newly differentially expressed transcripts compared to previous findings. Moreover, the bottom panel shows that whereas microarray studies can assess gross changes well (5-fold or greater), more subtle changes in differential expression are more robustly characterized using RNA-Seq. H) RNA-Seq reveals 2230 unique transcripts down-regulated in the Nrl^−/−retina compared to previous data sets. The bottom panel again highlights the greater coverage of differential expression at lower thresholds using RNA-Seq.

Table 1.

GO term breakdown of transcript reads across different RNA-Seq experiments with Wt and Nrl^−/− tissues

GO term	Retina		Eye
	Wt	Nrl−/−	Wt	Nrl−/−
Autophagy	28 (0.24)	27 (0.23)	28 (0.21)	28 (0.21)
Binding activity	979 (8.38)	995 (8.45)	1186 (8.85)	1190 (8.90)
Biogenesis	92 (0.79)	94 (0.80)	102 (0.76)	103 (0.77)
Catalytic activity	167 (1.43)	172 (1.46)	189 (1.41)	192 (1.44)
Cell adhesion	265 (2.27)	257 (2.18)	330 (2.46)	329 (2.46)
Cell cycle	292 (2.50)	296 (2.51)	353 (2.63)	349 (2.61)
Cell death	165 (1.41)	173 (1.47)	203 (1.51)	202 (1.51)
Cytoskeleton organization	158 (1.35)	161 (1.37)	183 (1.37)	181 (1.35)
Developmental process	159 (1.36)	163 (1.38)	203 (1.51)	201 (1.50)
DNA repair	138 (1.18)	140 (1.19)	159 (1.19)	157 (1.17)
Homeostatic process	112 (0.96)	113 (0.96)	123 (0.92)	122 (0.91)
Metabolic process	1704 (14.59)	1729 (14.68)	1983 (14.79)	1979 (14.78)
No term	2314 (19.82)	2341 (19.88)	2570 (19.17)	2599 (19.44)
Protein folding	98 (0.84)	98 (0.83)	101 (0.75)	102 (0.76)
Protein modification	421 (3.61)	418 (3.55)	508 (3.79)	500 (3.74)
Regulatory process	228 (1.95)	231 (1.96)	271 (2.02)	271 (2.03)
RNA processing	362 (3.10)	360 (3.06)	392 (2.92)	391 (2.92)
Signal transduction	1126 (9.64)	1131 (9.60)	1354 (10.10)	1337 (10.00)
Structural molecule activity	79 (0.68)	83 (0.70)	107 (0.80)	107 (0.80)
System process	55 (0.47)	61 (0.52)	86 (0.64)	91 (0.68)
Transcription	1164 (9.97)	1166 (9.90)	1255 (9.36)	1250 (9.35)
Translation	284 (2.43)	283 (2.40)	286 (2.13)	286 (2.14)
Transport	1287 (11.02)	1286 (10.92)	1434 (10.70)	1404 (10.50)
Total transcripts (≥1 FPKM)	11677	11778	13406	13368

All transcripts detected at a level of 1 FPKM in Wt and Nrl^−/− tissue samples were categorized by GO term categorization using AmiGO 1.8. Values represent number of transcripts, with percentage of total transcripts in parentheses. We had fewer total transcripts in the retina sample runs compared to the eye, since we examined a more specialized tissue of the whole. The number of transcripts in each category did not greatly vary between Wt and Nrl^−/−samples. This finding indicated that the knockout of Nrl causes relatively subtle differences in the overall gene network to mimic the human ESCS phenotype.

Verification of sequencing data by RT-PCR

Differential expression of transcripts was analyzed by comparing FPKM values between Wt and Nrl^−/−retinal (Fig. 2E) and whole-eye (Supplemental Fig. S2B) tissues. Previous microarray studies have used an empirical cutoff of a 1.5-fold average change as the minimum to identify a difference between the Wt and Nrl^−/− genotypes (26). Using this same cutoff, we identified 7316 and 6203 differentially expressed transcripts in whole-eye and retinal samples, respectively. Previous studies have used in situ hybridization and RT-PCR to verify results of microarray analyses (26, 28), results of which overlapped with our current RNA-Seq data (Fig. 2).Therefore, we chose to use RT-PCR to verify differentially expressed transcripts identified by RNA-Seq. Compared with the 66 transcripts that were differentially expressed and overlapped with findings of all 3 studies, RT-PCR had been done previously with 31 of these targets (26), and our RNA-Seq data indicated an excellent correlation of fold differences between Wt and Nrl^−/− retinal tissues for these 31 targets (R²=0.91). We also performed RT-PCR on targets representing a new signaling pathway that may dictate the formation and maintenance of the cone-line environment in the Nrl^−/− retina. We chose transcripts with large differential expression (Egr1, Opn1sw), more subtle changes (Gdf11, Otx2, Thrb) and no significant change (Crx). The results of our RT-PCR strongly correlated with the differential expression detected by RNA-Seq analyses of the retina and whole eye (Fig. 2F). Given the precise quantification of the transcript levels, we calculated that the signal needed for detecting a fold change of 1.5 with statistical significance required only 2 biological samples. In addition, because this trend in differential expression was preserved in RNA-Seq runs of the whole eye and retina (Table 2), this also indicates that the detected unique transcripts truly represent significant changes between the Wt and Nrl^−/− genotypes. Therefore, a thorough analysis of the murine retina transcriptome was well warranted.

Table 2.

Fold changes of selected transcripts in Nrl^−/− relative to Wt tissue across different experiments

Gene	RT-PCR	Retina RNA-Seq	Eye RNA-Seq
Abca4	−2.94	−3.05	−3.98
Atp8a2	−2.22	−3.02	−3.90
Atoh7	11.44	6.07	4.83
Bmp15	3.48	8.75	4.96
Crx	−1.22	−1.02	−1.02
Egr1	18.26	7.46	11.20
Eya1	37.48	24.35	5.26
Gdf11	−1.85	−3.45	−2.50
Neurod1	2.57	2.20	2.80
Notch1	1.08	1.21	0.62
Opn1sw	39.32	25.89	22.93
Otx2	1.73	1.66	1.39
Prdm1	4.74	5.33	2.93
Rxrg	2.81	5.96	3.52
Six6	2.75	5.26	4.87
Six6os1	7.89	25.83	37.91
Thrb	1.24	1.27	1.37
Wnt9b	−4.55	−5.88	−3.92

Fold change of selected transcripts was well preserved across different experiments. Targets chosen for validation included genes whose expression changed greatly (Opn1sw, Eya1, Egr1), genes with more subtle fold changes (Gdf11, Otx2, Thrb) and Crx, which had no significant change. Fold changes determined by RNA-Seq correlated well with changes in tissue expression of transcripts in both eye and retina.

Characterizing differentially expressed transcripts

Direct comparison of previous retinal microarray studies (26, 28) with the RNA-Seq of the Wt and Nrl^−/− retina revealed that of the 6203 differentially expressed transcripts, 5889 were unique to this study, with 3659 of the transcripts up-regulated (Fig. 2G) and 2230 transcripts down-regulated (Fig. 2H) in the Nrl^−/−retina. The greater number of up-regulated as compared to down-regulated transcripts using the same thresholds is similar to what was reported in previous microarray experiments (28). Breakdown of transcripts by their fold change revealed that RNA-Seq identified more subtle changes in transcript levels than microarrays (Fig. 2G, H).

Examination of transcripts with large differences between Wt and Nrl^−/− expression revealed 248 transcripts to differ by 5-fold or more, of which 134 were unique to this study. Pathway analysis was done for all 248 transcripts. In particular, analysis of the 134 unique transcripts identified several that were critical in pathways involved in photoreceptor differentiation and maintenance, such as atonal homologue 7 (Atoh7/Math5), a bHLH factor involved in Notch signaling (61); Six6, a sine oculis-related homeobox gene in the pathway of the master regulatory genes of eye development, Pax6 and Eya1 (62–68); and desert hedgehog (Dhh), a sonic hedgehog signaling molecule (69, 70). Furthermore, several key developmental maintenance pathways, such as the Wnt (71, 72) and Bmp signaling pathways (73, 74) had misexpressed transcripts in the Nrl^−/− retina, such as Wnt9b and Bmp15. Pathway analyses of these transcripts, coupled with other transcripts found in this and previous studies to be misregulated, provide a more comprehensive transcriptional landscape to examine S-cone commitment and maintenance in the Nrl^−/− mouse retina. Notably, 29 of 134 transcripts have no annotated function and thus represent new targets for study of cone-like photoreceptor maintenance and function.

Examination of homeostatic processes involved in photoreceptor function in the visual cycle revealed cone signature genes (Pde6c and Cngb3, among others). It also revealed down-regulated genes important for retinoid metabolism and clearance of potentially toxic photooxidized compounds (Abca4, Rdh12, and Rdh5, among others; Table 3). In contrast, expression of genes encoding putative proteins involved in RPE-mediated phagocytosis responsible for toxic metabolite removal and recycling were unchanged. However, key photoreceptor ligands necessary for phagocytosis, Tub and Tulp1 (75), were down-regulated in the Nrl^−/− retina (Table 4). Down-regulation of key retinoid metabolic genes, coupled with down-regulation of Tub and Tulp1, suggested a potential mechanism involving defective phagocytosis underlies the photoreceptor degeneration seen in ESCS. Therefore, to verify a potential aberrant phagocytotic process, we sought to verify the differential expression changes identified by RNA-Seq by complementary methods, including high-resolution imaging studies.

Table 3.

Transcript levels of visual cycle proteins in Nrl^−/− relative to Wt tissue across different RNA-Seq runs

Gene	Retina RNA-Seq			Eye RNA-Seq
	Wt	Nrl−/−	Fold differencea	Wt	Nrl−/−	Fold differencea
Abca4	175.71	57.62	0.33	90.25	22.51	0.25
Arr3	88.55	578.24	6.53	62.25	302	4.85
Cnga1	198.7	0.04	—	148.55	0	—
Cnga3	4.63	44.45	9.60	2.18	21.64	9.93
Cngb3	2.66	84.83	31.89	1.34	42.98	32.07
Gnat1	7281.64	1.53	—	4311.21	1.02	—
Gnat2	100.61	1924.72	19.13	67.4	951.07	14.11
Gnb1	1472.44	67.23	0.05	1400.74	121.52	0.09
Gnb2	76.18	88.58	1.16	124.9	92.93	0.74
Gnb3	214.35	1481.55	6.91	144.04	815.01	5.66
Gngt1	304.41	574.91	1.89	209.38	591.24	2.82
Gngt2	60.66	1379.01	22.73	91.18	1465.48	16.07
Guca1a	720.15	1150.14	1.60	749.33	553.63	0.74
Guca1b	827.36	64.27	0.08	625.54	25.57	0.04
Gucy1a3	33.47	39.42	1.18	19.16	25.01	1.31
Gucy1b3	32.67	39.5	1.21	38.38	56.48	1.47
Gucy2e	114.93	22.17	0.19	48.5	15.56	0.32
Gucy2f	20.73	0.25	0.01	10.82	0.37	0.03
Lrat	2.36	3.5	1.48	14.64	15.01	1.03
Opn1mw	137.65	246.42	1.79	49.12	136.56	2.78
Opn1sw	191.93	4968.88	25.89	120.49	2762.3	22.93
Pde6a	492.09	30.48	0.06	243.91	12.2	0.05
Pde6b	736.22	26.53	0.04	527.87	16.87	0.03
Pde6c	30.13	564.61	18.74	31.17	512.77	16.45
Pde6d	68.19	118.06	1.73	54.08	85.73	1.59
Pde6g	2250.16	1151.56	0.51	1561.58	639.69	0.41
Pde6h	166.22	1700.2	10.23	105.58	959.4	9.09
Ppp2r4	55.13	42.87	0.78	52.08	35.72	0.69
Rbp3	1093.27	1017.07	0.93	440.05	430.68	0.98
Rcvrn	1068.31	446.34	0.42	787.61	295.54	0.38
Rdh12	215.6	70.46	0.33	132.14	44.55	0.34
Rdh5	21.19	23.91	1.13	130.08	69.71	0.54
Rdh8	78.11	33.36	0.43	45.53	13.02	0.29
Rgs9	105.31	53.94	0.51	119.26	81.97	0.69
Rgs9bp	151.7	86.18	0.57	91.56	38.6	0.42
Rho	11745.08	0.83	—	6575.59	0.67	—
Rpe65	9.21	20.42	2.22	52.64	65.66	1.25
Sag	2021.36	1659.71	0.82	1181.97	768.89	0.65

Differential expression pattern in Nrl^−/− mice compared to Wt in transcript levels (FPKM) elucidates those transcripts that are essential for rod and cone function and maintenance. Fold differences, preserved across eye and retina RNA-Seq runs, illustrate noticeable enrichment of Cnga3, Cngb3, Gnb3, Gngt2, Pde6c, and Pde6h in Nrl^−/− mice, highlighting their role in cone-like photoreceptor function and maintenance. Conversely, the noticeable reduction of Cnga1, Gnat1, Gnb1, Guca1b, Gucy2f, Pde6a, and Pde6b in Nrl^−/− mice points out the importance of these transcripts in rod photoreceptor function and maintenance. In addition, many visual cycle proteins involved in retinoid metabolism, such as Abca4, Rdh12, and Rdh5, have attenuated transcript expression in Nrl^−/− mice, suggesting a misregulation of this process compared to Wt mice.

^aFold differences cannot be accurately determined for transcripts that are almost completely absent in Nrl^−/− mice.

Table 4.

Transcript levels of putative phagocytosis proteins in Nrl^−/− relative to Wt tissue across different RNA-Seq runs

Gene	Retina RNA-Seq			Eye RNA-Seq
	Wt	Nrl−/−	Fold difference	Wt	Nrl−/−	Fold difference
Anxa2	12.02	17.64	1.47	491	285.3	0.58
Axl	4.25	4.85	1.14	16.08	13.42	0.83
Cd36	0.45	1.13	2.51	11.18	30.48	2.73
Cd81	131.05	138.36	1.06	121.55	166.18	1.37
Gas6	125.26	78.05	0.62	175.97	95.2	0.54
Itgav	5.99	11.85	1.98	9.53	14.55	1.53
Itgb5	21.75	20.38	0.94	91.7	56.22	0.61
Mertk	1.82	2.44	1.34	3.38	2.97	0.88
Mfge8	190	155.59	0.82	253.06	173.92	0.69
Ptk2 (Fak)	9.37	12.39	1.32	15.04	14.8	0.98
Src	16.61	9.39	0.57	40.92	18.87	0.46
Tub	75.26	30.46	0.40	50.16	17.99	0.36
Tulp1	478.55	325.02	0.68	295.93	180.83	0.61
Tyro3	8.77	1.93	0.22	13.4	4.35	0.32

Differential expression pattern in Nrl^−/− mice compared to Wt in transcript levels (FPKM) elucidates those transcripts that are essential for rod and cone function and maintenance. Fold differences, preserved across eye and retina RNA-Seq runs, illustrate noticeable enrichment of Cnga3, Cngb3, Gnb3, Gngt2, Pde6c, and Pde6h in Nrl^−/− mice, highlighting their role in cone-like photoreceptor function and maintenance. Conversely, the noticeable reduction of Cnga1, Gnat1, Gnb1, Guca1b, Gucy2f, Pde6a, and Pde6b in Nrl^−/− mice points out the importance of these transcripts in rod photoreceptor function and maintenance. In addition, many visual cycle proteins involved in retinoid metabolism, such as Abca4, Rdh12, and Rdh5 have attenuated transcript expression in Nrl^−/− mice, suggesting a misregulation of this process compared to Wt mice.

Disrupted ESCS retinal architecture and patchy loss of photoreceptors

Overproduction of short-wavelength-sensitive photoreceptors in patients with ESCS causes retinal disorganization and has been analyzed only by conventional optical methods. The Nrl^−/− mouse retina offers the possibility to investigate the structural phenotype in greater detail. Confocal microscopy imaging of Nrl^−/− mouse retinal whole mounts dramatically illustrated disrupted architecture in 3-D space (Fig. 3). When compared to cone-like photoreceptor packing in Wt mouse retina (Fig. 3A, B), Nrl^−/− retina exhibited aberrant clustering of photoreceptors with empty patches where nuclear rosettes had formed (Fig. 3G, H). Wt retinal whole mounts stained for both rods and cones displayed the entire area populated by photoreceptors (Fig. 3C, D), whereas Nrl^−/− retinal mounts featured irregular photoreceptor packing (Fig. 3G, H). Notably, Nrl^−/− photoreceptor density was clearly reduced, as indicated by empty patches of retina that lacked photoreceptor staining (Fig. 3I, J).

Figure 3.

ESCS photoreceptors of Nrl^−/− mice display aberrant packing and OS morphology caused by buildup of material in OS heads and aberrant photoreceptor phagocytosis. Retinal whole-mount confocal microscopy displays the tight packing of Wt mouse retinal photoreceptors in 3-D space. A, B) Cone OS (S-cone opsin antibody; A) and cone sheath (peanut agglutinin; B) signals overlap and illustrate the packing of cones in normal retina. C, D) Staining of cone OS (C) and rod OS (rhodopsin C-terminal 1D4 antibody; D) reveals that the retina is fully occupied by photoreceptors, especially densely packed rods. E) SEM imaging of critical point dried Wt retina further emphasizes that photoreceptors pack tightly in the retina. F) Closer examination by SEM shows that rod photoreceptors display their characteristic cylindrical shape. In contrast, Nrl^−/− retinas exhibited disrupted photoreceptor packing with clusters of densely populated cones separated by empty patches. G, H) Cone-like OS (G) and extracellular sheath (H) signals overlap but also are absent from some retinal patches. I, J) Staining of cone OS (I) and rod OS (J) reveals only ESCS photoreceptors. K) SEM imaging of critical point dried Nrl^−/− retina shows disrupted packing of photoreceptors in the retina, with a less dense population of photoreceptors than Wt. L) Closer examination by SEM highlights abnormal OS morphology with enlarged head structures. M–R) Thin-sectioned retinas from Wt and Nrl^−/− mice were prepared for TEM imaging. M) Thin sectioning of Wt retina reveals the internal structure of photoreceptors. N) Discrete stacked discs are seen in rod photoreceptors. O) Because these samples were prepared at the peak of photoreceptor turnover, TEM imaging captures the disc shedding process and RPE mediated phagocytosis (asterisk). P) Thin sectioning of Nrl^−/− retina shows a distinctive OS disc arrangement that differs from Wt rods. Q) Discs retain some interconnections (arrow) as well as connections to the plasma membrane. R) Closer examination of ESCS photoreceptors reveals that most photoreceptors have enlarged head structures owing to buildup of material at the photoreceptor-RPE interface (asterisk), which would not occur with normal phagocytosis. Scale bars = 5 μm (A–D, G–J); 1 μm (E, F, K, L, O, R); 250 nm (M, N, P, Q).

To probe the photoreceptor morphology, we prepared critical point dried retinas separated from the RPE of both Wt and Nrl^−/− 4-wk-old mice. As noted from previous work (51), SEM imaging revealed that Wt retina contained tightly packed, cylindrically shaped rod photoreceptors (Fig. 3E, F). In contrast, Nrl^−/− photoreceptors exhibited a different structure and packing arrangement. In agreement with whole-mount confocal microscopy imaging, SEM imaging showed patches of ESCS photoreceptors clustered together and separated by patches devoid of photoreceptors (Fig. 3K). When these photoreceptors were probed at higher magnification, their OS appeared bulbous (Fig. 3L). This abnormal phenotype could explain why the packing density of photoreceptors is greatly reduced in ESCS (n=5, 0.38±0.04 photoreceptors/μm²) relative to Wt retina (n=5, 0.86±0.06 photoreceptors/μm²) when analyzing the SEM data. To better understand the structural defect contributing to this bulbous OS phenotype in Nrl^−/− photoreceptors, we studied thin sections of photoreceptors by TEM to examine their internal architecture.

ESCS photoreceptors exhibit abnormal accumulations of material

Thin sections of retina were prepared and examined by TEM. Wt rods displayed neatly stacked discrete discs (Fig. 3M, N). Because these blocks were prepared shortly after the onset of light, coinciding with the peak of OS disc shedding (3), imaging of the photoreceptor-RPE interface also revealed phagosomes that dispose of shed OS disc membranes (Fig. 3O). TEM imaging of photoreceptors in Nrl^−/− retina showed an OS disc arrangement distinctly different from rods (Fig. 3P), with some discs exhibiting interconnections to each other and the surrounding plasma membrane (Fig. 3P, Q). An abnormal buildup of material in ESCS photoreceptors was found at the photoreceptor-RPE interface (Fig. 3R). OS tips of these photoreceptors were enlarged because of an internal buildup of vacuole-like structures. By 8 wk of age, when degeneration is present, thin sections of Nrl^−/− retina revealed photoreceptor OS that exclusively contained vacuole-like structures with only a few disc elements present (Supplemental Fig. S3A, B).

Evidence for aberrant phagocytosis in ESCS disease

Phagocytosis is a dynamic process that occurs across the entire retina. Therefore, a hybrid SEM technique, SBF imaging, was used to cover a large area of the retina and capture serial sections of the photoreceptor-RPE interface to identify any abnormalities that might relate to this process. The block face was imaged by backscattered electrons after a 100-nm section was removed by a microtome inside the scanning electron microscope. This process was repeated to allow sectioning and imaging of the entire RPE to visualize the phagocytic process that occurs during OS disc shedding. SBF-SEM imaging revealed that the photoreceptor-RPE interface in the Wt sample has an orderly architecture with tightly stacked rods apposed to the interface where shed disc membranes are clearly engulfed (Fig. 4A, B and Supplemental Movie S1). Moreover, 3-D reconstructions revealed that these phagosomes were present throughout the RPE (Fig. 4C), demonstrating that this technique can capture a dynamic process occurring across multiple planes of the retina. When Nrl^−/− mouse retina was subjected to SBF-SEM imaging, the photoreceptors were not tightly packed at the RPE interface, and they also exhibited the abnormal associations with the RPE previously noted. Furthermore, the aberrant diseased OS head phenotype was visible in some sections (Fig. 4D–E and Supplemental Movie S1), but no phagosomes were seen throughout the retina, despite sectioning through multiple RPE cells. Given that a proportion of cone photoreceptors is renewed daily (4, 5), it is noteworthy that not a single phagosome was identified, indicating severely reduced RPE phagocytosis of OS in this ESCS model. Three-dimensional reconstructions further illustrated the lack of phagosomes in the RPE and confirmed the presence of abnormal photoreceptor OS heads at the RPE interface (Fig. 4F).

Figure 4.

SBF-SEM allows visualization of impaired phagocytosis present in ESCS retinal degeneration. Because photoreceptor disc phagocytosis is a dynamic process that occurs throughout the retina, SBF-SEM imaging was used to collect precise serial sections and investigate phagocytosis of shed discs. A) In Wt retina, the photoreceptor-RPE interface is clearly visible, with tight packing of rods opposed to the RPE (see Supplemental Movie S1). B) Moreover, phagosomes (asterisk) ingested by the RPE are clearly visible. C) Three-dimensional reconstructions of collected data with Reconstruct allow visualization of multiple phagosomes (red), including the one indicated by asterisk in panels A, B, throughout a RPE cell (gray, with nucleus in blue) and also reveal the tight packing of rods (green and blue) in a plane. D) In Nrl^−/− retina, the photoreceptor-RPE interface is visible, but photoreceptors (asterisks) are not as densely packed against the RPE (see Supplemental Movie S1). E) Enlarged OS head structures of ESCS photoreceptors (asterisk) are seen with less electron-dense material at the tips, indicating loss of OS material in that area. F) Resulting 3-D reconstruction illustrates these enlarged headed photoreceptors (asterisks) and their interactions with the RPE (gray, with nucleus in blue). Of note is the absence of any visible phagosomes within the modeled Nrl^−/− RPE. Scale bars = 1 μm.

The compromised phagocytic process identified in Nrl^−/− mice was subsequently confirmed by staining for phagosomes at the RPE-photoreceptor interface. Cryo-sectioned retinas were stained with an antibody against PS, the most abundant anionic phospholipid (76) and a key component of phagosomes (77, 78), asymmetrically situated in the inner leaflet of their plasma membranes (79). Because PS is present in most cell membranes (80), low detergent concentrations (0.3% Triton) were used to selectively detect PS staining of shed disc membranes. In Wt mouse retina, PS staining identified phagosomes at the photoreceptor-RPE interface (Fig. 5). PS staining colocalized with shed discs labeled for cone opsin, confirming that staining at the photoreceptor-RPE interface truly captured shed disc phagosomes (Fig. 5A, D). Light and confocal microscopy of cryosectioned retinas and stained eye whole mounts also revealed that this phagosome staining was located at the RPE-photoreceptor plane interface (Fig. 5E, F). When the same experiments were performed with Nrl^−/− mice, no such staining was found at the photoreceptor-RPE interface, either in cryosectioned retinas (Fig. 5G) or eye whole mounts (Fig. 5H). Similar experiments were done with an anti-annexin V antibody, which also recognizes PS, albeit less specifically (81), with results consistent with those obtained with the anti-PS antibody. Because phagocytosis depends on both proper signaling from shed disc packets and recognition of this signal by the neighboring RPE, we finally sought to understand whether the Nrl^−/− phagocytotic defect was due to an inherent abnormality of the photoreceptors or the RPE.

Figure 5.

Absent phagosome staining at the photoreceptor-RPE interface confirms impaired phagocytosis in Nrl^−/− mice. Absence of proper phagocytosis was confirmed by biochemical staining for phosphatidylserine (PS) found on phagosomes and its localization with shed cone disc packets. A) In Wt retina, shed cone opsin discs (red; cone opsin antibody) colocalize with phagosomes stained for PS (green; indicated by arrow) at the photoreceptor-RPE interface. (Note that not all the PS staining indicates phagosomes as PS dye also stains the RPE cell nucleus). B–D) Zoom views show a cone opsin disc (red; indicated by arrow; B), PS phagsome staining (green; indicated by arrow; C), and colocalization of the two stains (indicated by arrow; D). E) In samples of Wt mouse retina examined at the peak of phagocytosis, staining of phagosomes (red) was found at the photoreceptor-RPE interface (arrows). F) Through optical sectioning of the collected data, confocal imaging of the retina-RPE interface reveals that the PS signal is present at the photoreceptor-RPE interface. Still image of the tangential plane of these collected data shows 3 corresponding slices of data at right, indicating layers containing nuclear (blue), cone sheath (green), and phagosome (red) signals. G) In contrast, comparable Nrl^−/− retinal samples fail to exhibit staining for PS at the photoreceptor-RPE interface. H) Confocal imaging of the Nrl^−/− retina-RPE interface shows that no detectable PS signal is evident at the photoreceptor-RPE interface. Note that it appears that there is some PS staining in the Nrl^−/− retina, but it is not located at the interface. Still image of the tangential plane of the collected data shows 3 slices of data at right, indicating layers containing nuclear (blue), ESCS photoreceptor (green), and phagosome (red) signals. Staining: cone opsin for cone OS; PNA for cone sheaths; DAPI or Topro 3 nuclear stain; phosphatidylserine for phagosomes. Scale bars = 5 μm.

ESCS phenotype attributed to photoreceptor abnormalities rather than an RPE defect

The RPE plays a critical role in retinal maintenance. Thus, the defect in Nrl^−/− mice and humans with ESCS may be attributable not only to aberrant photoreceptors that result in this disease but also to defective RPE cells that interface with these photoreceptors. To investigate these possibilities, we cultured isolated RPE cells from Wt and Nrl^−/− eyes (50). The yield from both types of mice was comparable. Purified OS membrane vesicles from Wt and Nrl^−/− animals then were used to test the phagocytotic activity of cultured RPE cells (52). The OS vesicles were not always intact when isolated, especially those from Nrl^−/− animals, due to interconnections of the cone-like discs, and thus would produce membranes with PS exposed to RPE cells that would promote normal phagocytosis. Signals were noted to emanate selectively from OS membrane vesicles that had been ingested by the RPE rather than any other source (Fig. 6A–C). Wt RPE cells, when challenged with OS from Wt mice, phagocytosed these membranes as expected (Fig. 6D–F). Ingestion of OS membranes by the RPE was also confirmed by TEM imaging of thin plastic blocks (Supplemental Fig. S3C, D). When the Wt RPE cells were challenged with OS from Nrl^−/− mice, the RPE phagocytosed these membranes as well (Fig. 6G, I). Moreover, RPE cells cultured from Nrl^−/− mice also phagocytosed OS from Wt mice (Fig. 6J, L) as well as from Nrl^−/− animals (Fig. 6M, O). These results indicate that the Nrl^−/−defect was not in the RPE, but rather it is related to photoreceptors and their interface with the RPE.

Figure 6.

Wt and Nrl^−/− mouse RPE phagocytose both Wt and Nrl^−/− photoreceptor OS membranes. First, the phagocytosis assay was run with no OSs as a negative control. A–C) Wt mouse RPE cells in culture not subjected to photoreceptor challenge (A) were just washed with FITC-labeled dye (B) where only faint background fluorescence was detected; images were overlaid (C). D–F) Wt RPE cells (D) were challenged with isolated Wt photoreceptor OS membranes (E); phagocytosis of the OS membranes was seen in the overlapping images (F). G–I) Wt RPE cells (G) challenged with isolated Nrl^−/− photoreceptor OS membranes (H) phagocytosed the OS membranes, as evident in the overlapping images (I). J–L) Nrl^−/− RPE cells (J) challenged with isolated Wt photoreceptor OS membranes (K) phagocytosed the OS membranes, as evident in the overlapping image (L). M–O) Similarly, Nrl^−/− RPE (M) challenged with isolated Nrl^−/− photoreceptor OS membranes (N) phagocytosed the OS membranes, as evident in the overlapping images (O).

DISCUSSION

The details of abnormal photoreceptor development resulting from mutations in NR2E3 and NRL genes have captured the interest of developmental biologists for nearly 2 decades. However, an equally important but less explored feature of ESCS is the progressive retinal degeneration that leads to blindness in these patients (6, 7, 15, 82). Retinal degeneration in ESCS has been postulated to be secondary to a defective photoreceptor maintenance function of NR2E3 or a combination of cell proliferation and death (14, 15, 55), but specifics have been lacking. To understand this unique human condition, an appropriate animal model is required. Knockout of the Nrl transcription factor in mice produces a retina that is overpopulated with S-cone-like photoreceptors along with a complete absence of rod photoreceptors. Morphological assessment of this Nrl^−/− mouse model revealed that the postnatal perturbation of retinal organization (12, 83, 84) was similar to the disorganized retinal layering noted in postmortem donor retinas of patients with ESCS (14, 15). In this study, by comparing and following patients with ESCS, we show by in vivo and ex vivo imaging that the Nrl^−/− mouse model approximates the phenotypic features of human ESCS. Thus, the Nrl^−/− mouse model allowed us to probe the molecular mechanisms of ESCS-induced retinal degeneration.

Comprehensive analysis of the murine eye and retina transcriptomes of Wt and Nrl^−/− mice by RNA-Seq allowed greater understanding of the global transcriptional misregulation that results in aberrant, unstable photoreceptors of the Nrl^−/− mouse retina. Interestingly, our RNA-Seq analysis reveals that many signaling networks, such as Notch and Hedgehog, needed for normal photoreceptor maintenance and retinal lamination are misregulated in Nrl^−/−mice. Proper retinal cell type specification is heavily reliant on both Notch (17, 85) and Hedgehog signaling (69, 70), because depletion or pharmacological inhibition of Notch1 or Dhh in the retina causes progenitor cells to differentiate prematurely. Notably, Notch1 inhibition causes commitment of retinal progenitor cells to a cone photoreceptor fate, preferentially specifying S cones (17, 18, 86). Furthermore, both Notch1 and Dhh are critical for proper retinal morphology, as their depletion in the retina causes rosette formation in the ONL (58, 87, 88). Precocious S-cone formation and ONL rosette formation resulting from aberrations in Notch and Hedgehog in the retina are consistent with what is observed in the mature Nrl^−/− retina.

The misregulation of Hedgehog and Notch pathways is evident by almost complete absence of Dhh in Nrl^−/− retina compared to Wt and up-regulation of bHLH factors in Nrl^−/− retina. The latter are normally suppressed by proper Notch signaling. Examples include neurogenic differentiation 1 (NeuroD1) and Atoh7, genes responsible for committing cells to the earliest cell fates of S-cone photoreceptors and ganglion cells, respectively. NeuroD1 alone cannot commit cells to the S-cone fate, but can do so in cooperation with Six6, another transcript up-regulated in the Nrl^−/− retina. Six6 can be activated by NeuroD1 (89). Further, Eya1, in the pathway of canonical Pax signaling, is another transcript highly elevated in the Nrl^−/−retina. Although elevated levels of NeuroD1 and Six6 can commit cells to a S-cone fate, a prolonged increase of Six6 expression leads to disruption of photoreceptor maturation, indicating a regulatory factor is needed to down-regulate Six6 activity after its induction (89). We hypothesize that the regulatory factor that could control terminal photoreceptor differentiation in this pathway is sine oculis-related homeobox 6 homologue opposite strand transcript 1 (Six6os1), which is highly up-regulated in the Nrl^−/− eye and retina (Fig. 2F and Fig. 7A). Opposite strand transcripts are natural antisense transcripts that can be involved in gene regulation, and recently, it was predicted that both Six6os1 and Crxos1 might encode putative protein products and thereby play a major role in photoreceptor development (90). Altogether, RNA-Seq based elucidation of the eye and retinal transcriptomes revealed transcriptional mis-regulation in the mature Nrl^−/− mice. These factors along with others identified in this study can potentially play a crucial role in cone photoreceptor maintenance (Fig. 7B). Indeed, proteins encoded by those unique genes found from this study to have no annotated function could prove especially attractive candidates for this role.

Figure 7.

Transcriptional misregulation causes precocious development of cone-like cells in the Nrl^−/− retina, which are then maintained by transcriptional networks that alter key homeostatic processes. A) Increased levels of Eya1 in the Nrl^−/− retina can activate the expression of Six6, which causes retinal progenitor cell (RPC) proliferation. Commitment of these retinal cells to an early cell fate requires a premature cell cycle exit. This is mediated by altered levels of Notch and Hedgehog transcriptional networks that produce increased levels of NeuroD1 and decreased levels of Dhh in the Nrl^−/− retina. Six6 and NeuroD1 together with Six6os1, exercising a possible regulatory role on Six6, synergize to promote the S-cone fate rather than alternative early cell fates. B) Maintenance of the cone-like cells in the mature retina can be attributed to a series of genes involved in transcriptional control of retinoid metabolism, transport, cell cycling, and signal transduction. Unique transcripts identified by RNA-Seq to be ≥5-fold differentially expressed in Nrl^−/− vs. Wt mouse retina provide a resource for identifying maintenance factors required for cone cell survival and the alterations that accompany disease. Up and down arrows indicate transcripts that are up- and down-regulated, respectively, in Nrl^−/− vs. Wt retina, as determined by RNA-Seq. C) Transcriptional misregulation causes changes in the expression of key homeostatic genes involved in phagocytosis, leading to the pathological degeneration in ESCS. The most critical receptor tyrosine kinase, Mertk, involved in RPE phagocytosis is unchanged in the Nrl^−/− retina compared to Wt. However, key homeostatic genes involved in photoreceptor OS phagocytosis and toxic metabolic movement, such as Abca4, Atp8a2, Tub, Tulp1, are down-regulated in the Nrl^−/− retina compared to Wt, thus contributing to the defect in photoreceptor phagocytosis. Down arrows indicate transcripts that are down-regulated in the Nrl^−/− retina compared to Wt; the up/down arrow indicates transcripts that were unchanged in expression, as determined by RNA-Seq.

Although the RNA-Seq study identified transcriptional mis-regulation that could affect normal cone photoreceptor maintenance, it also provided a potential molecular mechanism for ESCS induced retinal degeneration due to defective photoreceptor phagocytosis. RNA-Seq revealed down-regulation of genes involved in photoreceptor phagocytosis, such as tubby (Tub) and tubby-like protein 1 (Tulp1) (77) and possibly a recently described PS flippase in photoreceptor disc membranes, Atp8a2 (91). This prompted a more detailed study of the photoreceptor-RPE interface with high-resolution imaging methods. Compared to Wt rods correctly apposed at the RPE interface, ESCS photoreceptors examined by TEM demonstrated abnormal interactions with the RPE. Phagocytotic material shed from OS could not be detected across the retinas of multiple Nrl^−/− animals. Instead, the OS layer displayed a buildup of vacuole-like material in the heads of the photoreceptor (Fig. 3P, R), likely accounting for the bulbous OS head structures identified by SEM imaging (Fig. 3K, L). This phenotype changed with increasing age such that the OS became devoid of discs. Thin sections also supported this view. However, such data do not illustrate the dynamic process of phagocytosis across the retina. Phagosomes could be lacking in any given section, because only ∼10% of photoreceptors may be shedding at any one time. Therefore, we used SBF-SEM to section through an entire RPE cell in contact with hundreds of photoreceptor cells. This strategy allowed all phagocytotic events to be identified in Wt mouse retina at the morning peak of phagocytosis. In contrast, when the Nrl^−/− retina was studied using the same approach, no phagosomes were detected. This apparent defect in phagocytosis was then validated biochemically. Retinas from Wt mice stained with phagosomal markers, such as PS and annexin V, exhibited phagosomes at the photoreceptor-RPE interface. There was a complete absence of such staining in Nrl^−/− mouse retina. Thus, both the defect in phagocytosis and the degenerative component of ESCS seem attributable to aberrant photoreceptors in the retina rather than a combination of photoreceptor and RPE cell dysfunction. This was consistent with the RNA-Seq study where the critical receptor tyrosine kinase involved in phagocytosis, Mertk, was unchanged at the transcriptional level between Wt and Nrl^−/−mice. The aberrant photoreceptor hypothesis was further supported by challenging cultured Wt and Nrl^−/−RPE cells with OS membranes. In the challenge assay, the fed OSs are not intact and thus broken pieces presenting PS will readily be phagocytosed by the RPE cells. Both Wt and Nrl^−/−RPE cells revealed comparable phagocytotic activity, indicating that the defect in phagocytosis was independent of an impairment in RPE function. Instead, the aberrant packing and spacing of the photoreceptors in ESCS disrupts the normal phagocytosis machinery of shed photoreceptor discs.

Based on this phagocytotic defect found in Nrl^−/− mice, we hypothesize that it is the precipitating cause of the retinal degeneration that occurs in human and murine ESCS (Fig. 7C). Previous investigators have speculated that retinal degeneration in ESCS is secondary to a postnatal photoreceptor maintenance function of NR2E3, for example (15, 92). However, the dysplasia in human ESCS and murine models, evidenced by rosettes (or whorls), is not unique to ESCS but is a feature of many retinal pathological processes (93). Patchy loss of laminar integrity and abnormal photoreceptor-RPE interactions are dramatic features accompanying the rosettes in many diseases, but no direct evidence indicates that this causes progressive retinal degeneration. Even the hypothesis that overcrowding due to retinal folding or rosette formations may be relieved by photoreceptor degeneration has been disputed (93). Phagocytosis of shed disc packets from the OS of photoreceptors is essential for normal function of these cells (3). Without this process, the buildup of material becomes toxic to the cell over time as demonstrated in the Royal College of Surgeons rat, which is defective for Mertk (94, 95). Similarly, impairment of phagocytosis in the Nrl^−/− retina could progressively cause a buildup of toxic materials that leads to degeneration. This buildup of toxic compounds would produce a fluorescent signal in the retina, much like the one we identified in patients with ESCS (Fig. 1E).

In summary, we have shown that photoreceptors in the Nrl^−/− retina have robust expression of S-cone opsin and they display an aberrant packing and morphology leading to progressive degeneration attributable to a defect in normal photoreceptor phagocytosis. Changes in the transcriptional landscape of the Nrl^−/− eye result in the expression of a unique subset of photoreceptor genes at levels that differ from those of native rods or cones. The developmental defect that affects photoreceptor cell fate also appears to have a detrimental effect on the normal retinal microenvironment. Thus, the inherent defect in phagocytosis in the Nrl^−/− retina observed in this work is likely caused by changes in the normal transcriptional landscape that causes an overpopulation of ESCS photoreceptors in the retina. In animals, such as the tree shrew, that possess retinas that are almost completely populated with cones, phagocytosis occurs normally (5). The produced mutant photoreceptor cells in ESCS have a lower density than photoreceptor cells in Wt rod or cone-dominated retina as well as disruption in expression of key homeostatic genes, including genes involved in proper photoreceptor phagocytosis and maintenance. This could account for their instability. These changes make ESCS photoreceptors unstable, producing retinal degeneration at an early age.