Endoplasmic Reticulum Stress-associated Cone Photoreceptor Degeneration in Cyclic Nucleotide-gated Channel Deficiency*

Background: Photoreceptors undergo degeneration when phototransduction is impaired. Results: The endoplasmic reticulum stress markers and processing of the associated caspases are elevated in retinas with cone photoreceptor CNG channel deficiency. Conclusion: The endoplasmic reticulum stress-associated apoptotic pathways play a crucial role in cone degeneration. Significance: Understanding of the mechanism(s) of photoreceptor degeneration is essential for development of therapeutic strategies. Cyclic nucleotide-gated (CNG) channels play a pivotal role in phototransduction. Mutations in the cone CNG channel subunits CNGA3 and CNGB3 account for >70% of all known cases of achromatopsia. Cones degenerate in achromatopsia patients and in CNGA3−/− and CNGB3−/− mice. This work investigates the molecular basis of cone degeneration in CNG channel deficiency. As cones comprise only 2–3% of the total photoreceptor population in the wild-type mouse retina, we generated mouse lines with CNG channel deficiency on a cone-dominant background, i.e. CNGA3−/−/Nrl−/− and CNGB3−/−/Nrl−/− mice. The retinal phenotype and potential cell death pathways were examined by functional, biochemical, and immunohistochemical approaches. CNGA3−/−/Nrl−/− and CNGB3−/−/Nrl−/− mice showed impaired cone function, opsin mislocalization, and cone degeneration similar to that in the single knock-out mice. The endoplasmic reticulum stress marker proteins, including Grp78/Bip, phospho-eIF2α, phospho-IP3R, and CCAAT/enhancer-binding protein homologous protein, were elevated significantly in CNGA3−/−/Nrl−/− and CNGB3−/−/Nrl−/− retinas, compared with the age-matched (postnatal 30 days) Nrl−/− controls. Along with these, up-regulation of the cysteine protease calpains and cleavage of caspase-12 and caspase-7 were found in the channel-deficient retinas, suggesting an endoplasmic reticulum stress-associated apoptosis. In addition, we observed a nuclear translocation of apoptosis-inducing factor (AIF) and endonuclease G in CNGA3−/−/Nrl−/− and CNGB3−/−/Nrl−/− retinas, implying a mitochondrial insult in the endoplasmic reticulum stress-activated cell death process. Taken together, our findings suggest a crucial role of endoplasmic reticulum stress in cone degeneration associated with CNG channel deficiency.

Rod and cone photoreceptor cyclic nucleotide-gated (CNG) 3 channels are localized to the plasma membrane of the outer segment (OS) and play a pivotal role in phototransduction. In darkness, rod CNG channels are activated by binding of cyclic guanosine monophosphate (cGMP), allowing a steady inward cation (Na ϩ and Ca 2ϩ ) current. Light triggers a sequence of enzymatic reactions that leads to the hydrolysis of cGMP resulting in CNG channel closure, reduction of the inward cation (Na ϩ and Ca 2ϩ ) currents, and membrane hyperpolarization (1). A similar transduction scheme exists in cones. Because it provides the only source of Ca 2ϩ influx into the OS, the CNG channel is crucial to the control of intracellular Ca 2ϩ concentration. Structurally, CNG channels belong to the superfamily of voltage-gated ion channels. The rod CNG channel is formed from CNGA1 and CNGB1 subunits, whereas the cone CNG channel is formed from CNGA3 and CNGB3 subunits. In the heterologous expression system, the A subunits by themselves form a functional channel, whereas the B subunits do not form channels in the absence of the A subunits. However, co-expression of the A and B subunits forms heteromeric channels displaying a number of properties of typical native CNG channels (1,2). Biochemical characterization has demonstrated that the native photoreceptor CNG channels are the heterotetrameric complexes with a 3A:1B stoichiometry (3)(4)(5)(6).
Photoreceptor degeneration is represented by two categories: induced degeneration such as that caused by bright light exposure (18) or retinal detachment (19) and inherited degeneration resulting from a variety of genetic disorders. Inherited rod and cone degeneration causes incurable visual diseases such as retinitis pigmentosa and cone dystrophy. Despite a remarkable genetic heterogeneity of photoreceptor degeneration, apoptotic cell death is recognized as a common pathway in many types of human retinal degeneration (17,20). Recently, endoplasmic reticulum (ER) stress-associated apoptosis has been implicated in a wide variety of neuronal degenerative diseases, including retinal degeneration (20). The ER is a large membrane-enclosed cellular organelle in which membrane and secretory proteins are synthesized and folded into stable conformations and in which free Ca 2ϩ ions are stored. Accumulation of incorrectly folded proteins in the ER is known to cause unfolded protein response (UPR) and ER stress, which often triggers cytotoxicity and cell death. ER stress-associated photoreceptor death has been shown in animal models of retinitis pigmentosa, including those caused by mutations in rhodopsin (such as P23H mutation) (21,22) and by deficiency of the rod phosphodiesterase (PDE6, rd mice) (20,23). Nevertheless, our understanding of the mechanism(s) of cone degeneration is limited, and we know little about the molecules and pathways involved in cone death in CNG channel deficiency. As CNG channel is the main source of the Ca 2ϩ inward currents in the OS, deficiency of this channel may interfere with the cellular calcium homeostasis and lead to ER stress and subsequent cell death.
We investigated a potential role of ER stress in cone degeneration in CNG channel deficiency. An obvious challenge to study the cone system is the low population of cones in a roddominant mammalian retina. Cones comprise only 2-3% of the total photoreceptor population in the wild-type mouse retina. To overcome this, we generated mouse lines with CNG channel deficiency on a cone-dominant background, i.e. CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice. The double knock-out mice showed a retinal phenotype similar to that in their respective single knock-out mice, i.e. impaired cone function and cone degeneration. Biochemical characterization of the CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice (at postnatal 30 days) showed significantly elevated ER stress marker proteins, including Grp78/Bip, phospho-eIF2␣, IP 3 R, and CCAAT/enhancer-binding protein homologous protein (CHOP) in the retina. The up-regulation of the cysteine protease calpains and processing of caspase-12 and caspase-7 in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas suggest an ER stress-associated cell death process. Hence, ER stress might play a crucial role in cone degeneration in CNG channel deficiency.

EXPERIMENTAL PROCEDURES
Mice, Antibodies, and Other Materials-The CNGA3 Ϫ/Ϫ and CNGB3 Ϫ/Ϫ mouse lines (on a C57BL/6 background) were generated as described previously (14,16). The Nrl Ϫ/Ϫ mouse line (on a C57BL/6 background) was kindly provided by Dr. Anand Swaroop (National Eye Institute, Bethesda, MD). The CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mouse lines were generated by cross-mating. Wild-type mice (C57BL/6J background) were purchased from Charles River Laboratories (Wilmington, MA). All animal maintenance and experiments were approved by the local Institutional Animal Care and Use Committee (Oklahoma City, OK) and conformed to the guidelines on the care and use of animals adopted by the Society for Neuroscience and the Association for Research in Vision and Ophthalmology (Rockville, MD).
Primary antibodies used in this study are listed in Table 1. Horseradish peroxidase (HRP)-conjugated anti-rabbit and anti-mouse were purchased from Kirkegaard & Perry Laboratories, Inc. (Gaithersburg, MD). Fluorescent goat anti-rabbit and goat anti-mouse conjugated to Alexa Fluor 488, 568, or 647 were purchased from Invitrogen. All other reagents were purchased from Sigma, Bio-Rad, and Invitrogen.
Recordings of Electroretinograms (ERG)-Full-field ERG recordings were carried out as described previously (16). Briefly, after overnight dark adaptation, animals were anesthetized by intraperitoneal injection of 85 mg/kg ketamine and 14 mg/kg xylazine. ERGs were recorded using an LKC Technologies system (Gaithersburg, MD). Potentials were recorded using a platinum wire contacting the corneal surface through a layer of 2.5% methylcellulose. For assessment of scotopic responses, a white light stimulus intensity of 1.89 log cd s m Ϫ2 was presented to dark-adapted, dilated mouse eyes in a Ganzfeld (GS-2000; Nicolet Instruments, Inc. Madison, WI). To evaluate photopic responses, mice were adapted to a 1.46 log cd s m Ϫ2 light for 5 min, and then a light intensity of 1.89 log cd s m Ϫ2 was given (24). Responses were differentially amplified, averaged, and stored using a Nicolet Compact-4 signal averaging system. The ERG testing was performed between 10:00 a.m. and 12:00 p.m.
Retinal Protein Sample Preparation, SDS-PAGE, and Western Blot Analysis-Retinal protein sample preparation, SDS-PAGE, and Western blotting were performed as described previously (25). Briefly, retinas were homogenized in homogenization buffer (20 mM HEPES-NaOH, pH 7.4, 1 mM EDTA, 200 mM sucrose, containing protease mixture (Sigma)). The homo-genate was centrifuged at 1000 ϫ g for 10 min at 4°C to pellet down nuclei and cell debris. The pellet was used as nuclei-enriched fraction after three washes. The supernatant of the homogenate was further centrifuged at 16,000 ϫ g for 30 min at 4°C to separate out cytosolic (supernatant) and membrane fractions (pellet). Protein concentration of the preparations was measured using a Bio-Rad Protein Assay kit (Bio-Rad Laboratories).
The protein samples (8 to 20 g) were solubilized in SDS-PAGE sample buffer and separated on an SDS-PAGE gel (7 or 10% acrylamide) and transferred onto polyvinylidene difluoride (PVDF) membrane (Bio-Rad Laboratories). After 1 h of blocking in 5% milk or 5% BSA containing Tris-buffered saline with 0.1% Tween (v/v) at room temperature, the membranes were incubated with primary antibodies overnight at 4°C (see Table  1 for antibody dilutions). The membranes were then washed with Tris-buffered saline with 0.1% Tween three to four times and incubated with HRP-conjugated secondary anti-rabbit or anti-mouse antibodies for 1 h at room temperature. After washing, the antigen and antibody binding was detected using SuperSignal West Dura Extended Duration chemiluminescent substrate (Pierce). The blots were scanned, and images were captured using a KODAK Image Station 4000R digital imaging system (Carestream Molecular Imaging, New Haven, CT). Densitometry analysis was performed by quantifying the intensities of the bands of interest using KODAK molecular imaging software with ␤-actin or acetyl-histone H3 serving as a loading control. Data for each group were obtained from three to four independent Western blot experiments performed using retinas prepared from four to five mice and analyzed and graphed using GraphPad Prism software (GraphPad Software, San Diego, CA).
Eye Preparation, Immunofluorescence Labeling, and Confocal Microscopy-Mouse eye cross-sections were prepared for immunohistochemical analysis as described previously (16). Briefly, euthanasia of mice was performed by CO 2 asphyxiation, and mouse eyes were enucleated and fixed with 4% formalde-hyde (Polysciences, Inc., Warrington, PA) in 0.1 M sodium phosphate buffer, pH 7.4, for 16 h at 4°C. The superior portion of the cornea was marked with a green dye for orientation prior to enucleation. Fixed eyes were transferred to PBS or 0.1 M sodium phosphate buffer, pH 7.4, containing 0.02% sodium azide, for storage until processing and embedding in paraffin. Paraffin sections (5-m thickness) passing vertically through the retina were prepared using a Leica microtome (Richmond, IL). Immunofluorescence labeling was performed as described previously (16). Briefly, eye sections were blocked with PBS containing 5% BSA and 0.5% Triton X-100 for 1 h at room temperature. When necessary, antigen retrieval was performed by incubating tissues in 10 mM sodium citrate buffer, pH 6.0, for 30 min in a 65°C water bath. Primary antibody incubation (rabbit anti-CHOP, 1:200; goat anti-S-opsin, 1:500; and rabbit anti-Grp78/Bip, 1:200) was performed at room temperature for 2 h. Following Alexa Fluor 488 or 568 or FITC-conjugated secondary antibody incubation and rinses, slides were mounted and coverslipped. Fluorescent signals were imaged using an Olympus AX70 fluorescence microscope (Olympus Corp., Center Valley, PA) with QCapture imaging software (QImaging Corp., Surrey, BC, Canada) or an Olympus IX81-FV500 confocal laser scanning microscope (Olympus, Melville, NY) (using excitation wavelengths of 543 nm for Alexa Fluor 568 and 488 nm for FITC) and FluoView imaging software (Olympus, Melville, NY).
TUNEL Assay-The TUNEL assay was performed to evaluate photoreceptor apoptotic death in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ , CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and Nrl Ϫ/Ϫ mice as described previously (16). An apoptosis detection kit (ApopTag plus peroxidase in situ apoptosis detection; Chemicon, Temecula, CA) and paraffin-embedded sections were used in this analysis. Immunohistochemical labeling was imaged using an Olympus AX70 fluorescence microscope (Olympus Corp.) with QCapture imaging software (QImaging Corp.). Quantification of the positive labeling was performed by using Image Pro 6 software (Media Cybernetics, Inc., Bethesda, MD). Briefly, images were taken using a 40ϫ objective on an Olympus microscope; the image scale was calibrated; and positive labeling was counted in regions with dimensions of 336 ϫ 256 m (8.6 ϫ 10 4 m 2 ) using the software. The averages from four to five regions of each retinal section were obtained, and data from sections were prepared from four to five mice from each group were analyzed and graphed using GraphPad Prism software (GraphPad Software).
Enhanced Expression of ER Stress Marker Proteins in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ Retinas-As CNG channel plays a key role in OS calcium homeostasis; loss of the functional channel is likely to alter the cellular calcium balance. Moreover, cones in CNG channel-deficient mice display opsin mislocalization, accompanied by photoreceptor apoptotic death. Hence, we examined whether there is ER stress in CNG channel-deficient retinas. The expression levels of the ER stress marker proteins, Grp78/Bip, phospho-eIF2␣, and CHOP were examined by immunoblotting. Grp78/Bip is a member of the heat-shock protein-70 family and is involved in the folding and assembly of proteins in the ER. Grp78/Bip interacts transiently with many ER proteins and plays a key role in monitoring protein transport through the cell. When there is ER overload and stress, Grp78/Bip dissociates from its interacting proteins (such as protein kinase RNA-like endoplasmic reticulum kinase), initiating activation or phosphorylation of a variety of proteins, including eIF2 (eukaryotic initiation factor 2, a protein required in the initiation of translation) (27). Here, we show the elevated levels of the ER stress markers in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas. As shown in Fig. 4A, the levels of Grp78/Bip and phospho-eIF2␣ were increased significantly in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas. Densitometric analysis shows that the levels of phospho-eIF2␣ were enhanced by ϳ38 and 20% in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas, respectively, compared with the Nrl Ϫ/Ϫ control (Fig. 4A). CHOP (also known as GADD153 (growth arrest and DNA damage-inducible protein 153)) is a member of the C/EBP (CCAAT enhancer-binding protein) transcription factor family and functions as a dimer to inhibit the DNA-binding activity of C/EBP and liver-enriched activator protein. This protein is not expressed at detectable levels under normal conditions and is induced in ER stress through the Grp78/Bip-phospho-eIF2␣ pathway (28). We therefore examined CHOP expression using retinal nuclear preparations and found that the expression levels of CHOP were increased in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas (Fig. 4A). The elevated CHOP expression was also shown by immunofluorescence labeling. The CHOP signal was barely detected in the ONL of Nrl Ϫ/Ϫ retinas (Fig. 4B, left panel); in contrast, a markedly increased signal was detected in the ONL of CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ retinas (Fig. 4B, right panel).
We have shown opsin mislocalization in CNG channel-deficient retina (13,15). To determine whether opsin is mislocalized in the ER, we performed double labeling on retinal sections of CNGA3 Ϫ/Ϫ , CNGB3 Ϫ/Ϫ , and wild-type mice using anti-Grp78/Bip and anti-S-opsin. Analysis of high magnification images of confocal microscopy shows a co-labeling of S-opsin and Grp78/Bip in the inner segment of CNGA3 Ϫ/Ϫ and CNGB3 Ϫ/Ϫ but not wild-type retina, and both proteins are barely detected in the inner segment of the wild-type mice (Fig. 4C).

Enhanced Levels of Phospho-IP 3 R and Calpains in CNGA3
Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ Retinas-Inositol 1,4,5-trisphosphate (IP 3 ) is a second messenger for many growth factors, hormones, and neurotransmitters. Upon binding to its receptor (IP 3 R) on the ER membrane, IP 3 triggers the phosphorylation of IP 3 R. In the channel-deficient photoreceptors there is an altered calcium balance, which may stimulate the release of Ca 2ϩ from the ER store. Therefore, we examined the levels of phospho-IP 3 R, and we found its elevation in CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas (Fig. 5A). Densitometric analysis showed that the levels of phospho-IP 3 R were enhanced ϳ30 and 27% in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ / Nrl Ϫ/Ϫ mice, respectively, compared with the Nrl Ϫ/Ϫ controls (Fig. 5B). We further examined the expression levels of the cysteine protease calpains (calpain I and calpain II, also known as -calpain and m-calpain) in the channel-deficient retinas. Calpains have been shown to be activated by Ca 2ϩ under ER stress and are involved in the processing of caspase-12 (29). As shown in Fig. 5A, expression levels of these proteases were increased significantly in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas. Densitometric analysis showed that the levels of calpain II were enhanced by ϳ28 and 19% in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas, respectively (Fig. 5B). It has also been shown that the ER membrane permeability to Ca 2ϩ is altered by activation of the Bcl-2 proteins in cells undergoing apoptotic death (30). We therefore examined the expression levels of Bcl-2 and Bcl-x and found a significant elevation of these proteins in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retina (Fig. 5C). This result along with an elevated phospho-IP 3 R supports a poten-tial Ca 2ϩ release from ER storage in CNG channel-deficient cones.
Enhanced Processing of Caspase-12 in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ Retinas-We found evidence of ER stress in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas and further examined whether this ER stress is associated with cone death. It has been shown that the cysteine protease caspase-12 is activated upon ER stress and Ca 2ϩ release from the ER store, and its processed forms can translocate into the nucleus to induce apoptosis (31,32). Therefore, we examined the levels of caspase-12 and its processing in the CNG channeldeficient retina. The processing of the protein was evaluated by Western blotting using an antibody recognizing the full-length protein and its cleaved products. As shown in Fig. 6A, procaspase-12 was detected in the retinal cytosols only, whereas the cleaved forms were detected in both cytosolic and nuclear fractions. The expression levels of procaspase-12 in CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retina were similar to that in Nrl Ϫ/Ϫ retina; however, significantly elevated levels of the cleaved forms were detected in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas, which was particularly obvious in the nuclei-enriched preparations (Fig. 6A). Densitometric analysis showed that the levels of the cleaved forms were enhanced by ϳ33 and 15% in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ / Nrl Ϫ/Ϫ mice, respectively (Fig. 6B). Thus, CNG channel-deficient retinas show an enhanced processing of caspase-12, suggesting an ER stress-associated, caspase-12-mediated cell death process. Furthermore, caspase-7, which resides in the ER neighborhood along with caspase-12, can also be activated under ER stress (33) and can subsequently cleave procaspase-12 to generate active forms of the protease (34). We therefore examined the levels of caspase-7 and its cleaved products. The full-length protein was detected in the retinal cytosolic preparations of both channel-deficient and control mice; however, the cleaved forms were detected only in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ / Nrl Ϫ/Ϫ retinas (Fig. 6C), suggesting an activation of caspase-7. We did not detect any caspase-7 immunoreactivity in the nuclear fractions (data not shown), suggesting that caspase-7 may act only as an initiating (activating) protease.
Expression of Mitochondrial Apoptotic Proteins in CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ Retinas-Although we obtained evidence supporting the ER stress and ER stress-associated apoptotic death process in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retina, it was tempting to test whether mitochondrial-associated cell death is involved. Apoptosis-inducing factor (AIF) is the protein that triggers chromatin condensation and DNA degradation. Normally, it is found behind the outer membrane of the mitochondria and therefore is sheltered from the nucleus. When the mitochondria are damaged, it moves to the cytosol and to the nucleus and initiates a caspase-independent cell death process. We therefore examined the expression levels of AIF in the retinal nuclear and cyto-solic fractions. We found a significant elevation of AIF in the nuclear fractions of CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ / Nrl Ϫ/Ϫ retinas, compared with the Nrl Ϫ/Ϫ controls (Fig. 7A), and only a trace amount of the protein was detected in the cytosol fractions of both channel-deficient and Nrl Ϫ/Ϫ retinas (data not shown). We also examined the nuclear translocation of endonuclease G (Endo G), an endonuclease that is localized in the mitochondrion and has been shown to act as an apoptotic DNase when released from mitochondria (35). As shown in Fig.  7A, expression levels of Endo G in the nuclear fractions of the channel-deficient retina were increased significantly, com-  Both cytosolic and nuclear fractions were used for this examination. Actin and TATA binding protein were used as loading controls for cytosolic and nuclear fractions, respectively. B, densitometric analysis of the relative levels of processed forms of caspase-12 (casp-12) in the nuclear fractions of CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ , CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ , and Nrl Ϫ/Ϫ retina. Data are represented as means Ϯ S.E. of measurements from three to four independent experiments using retinas from four to five mice. Unpaired Student's t test was used for determination of the significance (*, p Ͻ 0.05). C, shown are representative images of the Western blot detection of caspase-7. Retinal cytosolic protein preparations were used for this examination. Actin was used as a loading control. IB, immunoblot. pared with the Nrl Ϫ/Ϫ controls. Densitometric analysis shows that the levels of AIF and Endo G were elevated by ϳ25 and 16%, respectively, in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice and elevated by ϳ19 and 14%, respectively, in CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice, compared with the age-matched Nrl Ϫ/Ϫ controls (Fig. 7B). This observation suggests a potential mitochondrial insult in CNG channel-deficient retina.
Activation of caspase-3 by caspase-9 following the leakage of cytochrome c from mitochondria and the formation of apoptosome is a well characterized cellular event in apoptotic cell death. We therefore examined the levels of cytochrome c, caspase-9, and caspase-3 in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ / Nrl Ϫ/Ϫ retinas and found that the levels of these proteins were not altered, and no cleaved forms of caspase-3 or caspase-9 were detected in the channel-deficient retina (data not shown). This observation does not favor a role of the caspase-3-mediated classical pathway in cone death.

DISCUSSION
This work investigated the mechanism(s) of cone degeneration in CNG channel deficiency using CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice. Nrl is a basic motif leucine zipper (bZIP) transcription factor essential for the normal development of rods and the Nrl Ϫ/Ϫ mouse line is a cone-dominant mouse model. Nrl Ϫ/Ϫ mice have no rods but have increased numbers of S-cones, functionally manifested as a loss of rod function coupled with super-normal cone function (36). Retinas of Nrl Ϫ/Ϫ mice have cone-like nuclear morphology, short and disorganized OS, and a rosette-like structure (36), and Nrl Ϫ/Ϫ cones undergo a slow degeneration that can result in decreased ERG recordings (ϳ30% reduction) by 2-3 months of age (37,38). Nevertheless, as a unique mammalian cone-dominant model, the Nrl Ϫ/Ϫ mouse line has been widely used for the study of cone biology and disease (26,38,39). We have previously shown that the retinas of Nrl Ϫ/Ϫ mice express an abundant amount of cone CNG channel and lack the expression of the rod channel (25). In this study, we show that the double knock-out mice showed a retinal phenotype similar to that in their respective single knock-out mice, i.e. impaired cone function and cone degeneration. Hence, the CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mouse lines can serve as unique models to explore the mechanism(s) of cone degeneration, particu-larly to study the biochemical alterations occurring in these mice at young ages.
ER stress has been implicated in a variety of retinal degeneration including those caused by rhodopsin mutations (22,40) and by deficiency of PDE6 (20,23). The present work shows ER stress and its associated cell death process in CNG channeldeficient retinas. Although the exact mechanism(s) of ER stress is not known, an altered calcium homeostasis may in part be responsible. As a cellular Ca 2ϩ storage organelle, the ER is sensitive to the cellular Ca 2ϩ levels. In rd mice, lack of the functional PDE6 enzyme causes failure of cGMP degradation, constant opening of the channel and subsequent cellular Ca 2ϩ overload, which leads to ER stress (20,23). In contrast, lack of functional CNG channels might affect the calcium homeostasis differently from that caused by PDE6 deficiency. CNG channel is the only source for Ca 2ϩ influx in the OS (1), and lack of the functional CNG channel abolishes the inward Ca 2ϩ currents and might potentially lower the cellular Ca 2ϩ level. The presumption of a lowered cellular Ca 2ϩ level is supported by the observation of cGMP accumulation in CNG channel-deficient retinas (41). Retinal cGMP production is controlled by the enzyme guanylate cyclase-activating protein, which is tightly regulated by the cellular Ca 2ϩ level (42). A lowered cellular Ca 2ϩ level enhances the activity of guanylate cyclase-activating protein, which in turn stimulates retinal guanylate cyclase to produce cGMP. A low cellular Ca 2ϩ level in CNG channeldeficient retina also is supported by the increased levels of phospho-IP 3 R and Bcl-2 (which has been shown to modulate Ca 2ϩ release from ER storage (43)) (see Fig. 5). It is likely that the lowered cellular Ca 2ϩ level evokes Ca 2ϩ release from ER storage (via IP 3 R). Thus, ER stress in photoreceptors can be evoked by both cellular Ca 2ϩ overload (in rd mice) (23) and cellular Ca 2ϩ insufficiency (in CNG channel deficiency). Of note, our results show that the increases in the ER stress marker proteins (Grp78/Bip, phospho-elF2␣, and phospho-IP 3 R) were more prominent in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas compared with that in CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas, which may suggest a more severe ER stress in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retina. This observation is consistent with the consequences caused by deficiency of each subunit. Deficiency of CNGA3 causes a complete loss of cone phototransduction, whereas lack of CNGB3 leads to reduced cone function. Thus, CNGA3 deficiency might cause a much lower cellular Ca 2ϩ level compared with that in CNGB3 deficiency. Indeed, we observed a much higher level of retinal cGMP in CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice compared with that in CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ mice (see supplemental Fig. 1). In a separate experiment, we measured expression of CNGA3 in CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas by Western blotting. Similar to that in CNGB3 Ϫ/Ϫ mice (16), CNGA3 was detected but at a significantly reduced level in CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retina (see supplemental Fig. 2). Hence, the remaining light response (mediated by CNGA3 homomeric channels) likely helps maintain the cellular Ca 2ϩ levels in CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ cones.
The ER stress in CNG channel-deficient retinas might also be related with opsin mislocalization and ER accumulation. Accumulation of proteins or misfolded proteins in the ER is known to cause UPR and ER stress. ER stress has been documented in rhodopsin mutation both in vitro and in vivo (21,22,40,44). CNGA3 Ϫ/Ϫ (13) and CNGB3 Ϫ/Ϫ (15) mice display opsin mislocalization, and we have shown in this study the co-localization of opsin with Grp78/Bip in the inner segment of CNGA3 Ϫ/Ϫ and CNGB3 Ϫ/Ϫ retina, supporting an ER accumulation of cone opsin. How deficiency of CNG channel leads to ER accumulation of opsin is still not clear. The phenotype might not be caused directly by the channel deficiency but is more likely a consequence of the loss of (CNGA3 deficiency) or impaired (CNGB3 deficiency) phototransduction. It is worth mentioning that though lack of cone function in the absence of CNG channel (such as in CNGA3 Ϫ/Ϫ mice) causes opsin mislocalization, subunit mislocalization arising from diseaselinked missense mutations in CNGA3 has been reported extensively in heterologous expression system. Enhanced cytosolic aggregation and activation of ER stress were observed in many CNGA3 mutants, including those with mutations located in the subunit transmembrane S1 and S4 domains (45,46) and at the carboxyl terminals (47)(48)(49). Thus, the UPR and ER stress in CNGA3 missense (and loss of function) mutations could be caused not only by opsin mislocalization (due to lack of normal channel function/phototransduction) but also by mislocalization of mutant CNGA3 subunits, and the ER stress might be more severe than the situation of lack of CNGA3.
Persistent ER stress is known to trigger apoptotic cell death (32). The elevated expression and processing of CHOP, calpains, and caspase-12 suggest an ER stress-associated cell death mechanism in CNG channel deficiency. Based on our findings, we postulate that cone death is likely mediated by the following pathways (Fig. 8).
Grp78/Bip-eIF2a-CHOP Pathway-CHOP is known to be induced by ER stress (28) and by alterations of Ca 2ϩ flux across the ER membrane (50). We observed increased levels of CHOP and elevations of Grp78/Bip and phospho-eIF2␣ in CNG channel-deficient retinas, implying an ER stress-activated, CHOPmediated cell death. Indeed, CHOP activation has been shown in P23H rhodopsin transgenic rats (51), Lrat Ϫ/Ϫ mice (52), and in rat models of experimental retinal detachment (19).
Calpain-caspase-12 and Calpain-caspase-7-caspase-12 Pathways-It has been shown that calpains are activated in ER stress and mediate processing of caspase-12 (31,32), and the processed forms translocate to the nucleus to induce apoptosis (53). We found elevated expression of calpains, enhanced processing of caspase-12, and nuclear localization of its processed forms in the CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas. These observations suggest that caspase-12 might act as an executioner caspase, and the calpain-caspase-12 pathway might be potentially activated in CNG channel-deficient cones. Indeed, activation of calpains and caspase-12 has been linked to a variety of neurodegenerative conditions, including photoreceptor death in rd mice (20,29,54). Calpains have also been shown to cleave caspase-7 (34). This processing generates highly active and distinct fragments of caspase-7, which cleave caspase-12 (34). We examined the expression and processing of caspase-7 and found that the full-length protein was detected universally, whereas the processed forms were detected only in the CNGA3 Ϫ/Ϫ /Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retinas. Hence, activation of caspase-7 might also be involved in cone death. Unlike caspase-12, the processed forms of caspase-7 were not detected in the nuclear fractions, suggesting that caspase-7 acts primarily as an activating caspase. It is likely that caspase-12 is processed by both calpains and caspase-7 in CNG channel-deficient cones. Loss of functional CNG channel leads to impaired phototransduction. Impaired cone function interferes with the calcium homeostasis and opsin OS localization, which elicits ER stress. The substantial and constant ER stress ultimately triggers apoptotic death through activation of CHOP, caspase-12, and AIF/Endo G pathways. CHOP is induced by the Grp78/Bip and phosho-elF2␣ pathway; activation of caspase-12 occurs through caspase-7 and calpains; and activation of mitochondrial AIF and Endo G is mediated by activation of calpains. The altered calcium homeostasis stimulates Ca 2ϩ release from ER storage (via IP 3 R), which activates calpains. The accumulation of cGMP might be a potential factor to induce ER stress and mitochondrial insult through unknown mechanism(s).
Calpain-AIF Pathway-ER stress and activation of calpains have been shown to insult mitochondria, manifested as an increased permeability of the mitochondrial membrane and release of AIF. The released AIF can further translocate into the nucleus to induce DNA fragmentation and chromatin condensation (55). The activation of calpains and AIF have been shown in variety of retinal degeneration animal models, including rd mice (20,54), RCS rat (56), and Uchl3 Ϫ/Ϫ mice (57). We found an enhanced AIF level in the nuclear fractions of CNG channeldeficient retinas. This observation, along with the elevated calpains, suggests an activated calpain-AIF pathway, a process in which both the ER and mitochondria functions are engaged. Although we observed an increase of AIF in the CNGA3 Ϫ/Ϫ / Nrl Ϫ/Ϫ and CNGB3 Ϫ/Ϫ /Nrl Ϫ/Ϫ retina, the mitochondria-mediated, caspase-dependent pathways do not seem to play a major role in cone degeneration. This was supported by the findings that expression levels of cytochrome c, procaspase-9, and procaspase-3 were not altered, and no cleaved forms of caspase-3 or caspase-9 were detected in CNG channel-deficient retina. Indeed caspase-3, the executioner caspase in the classical mitochondrion-mediated pathway, appears to play a different role in varying types of photoreceptor degeneration. Activation of caspase-3 was detected in transgenic rats with rhodopsin mutations (58), and use of a caspase-3 inhibitor reduced the apoptotic cell death in tubby mice (59); however, photoreceptor death in rd1 mice was shown to be caspase-3independent (60), and lack of caspase-3 activation was also reported in animal models of induced retinal degeneration (light damage and retinal detachment) (19). Thus, though apoptotic cell death has been shown as a common cause of many types of photoreceptor degeneration (17), multiple and distinct pathways are likely engaged (61), and the mechanisms might vary between different pathological disorders.
It is important to mention that UPR and ER stress is an evolutionarily conserved cellular program that is characterized by down-regulation of de novo protein synthesis, up-regulation of ER chaperones and folding enzymes, and enhanced ER-associated degradation. Thus, UPR and ER stress are protective under certain circumstances. The consequences of ER stress include initiation of both cell death and cell protection from further death, and the balance of the two forces determines the final outcome (62). Recently, delivery of the Grp78/Bip transgene has been shown to reduce CHOP levels and apoptosis and restore visual function in P23H rhodopsin transgenic rats (51). Bcl2 family proteins have been shown to suppress the ER stressinduced caspase activation and cell death (63). Indeed, Bcl and Bcl-x were found to be up-regulated in CNG channel-deficient retina, and they may act as anti-apoptotic factors over the course of cone degeneration.
In summary, cone photoreceptors with CNG channel deficiency undergo ER stress as characterized by elevated levels of Grp78/Bip, phospho-eIF2␣, and phospho-IP 3 R. Accumulation of CHOP in the nucleus, elevated expression of calpains, and processing of caspase-12 and caspase-7 imply an ER stress-associated apoptosis. In addition, the nuclear translocation of AIF and Endo G suggests a mitochondrial insult in the ER stressassociated cell death process.