Knockout of GARPs and the β-subunit of the rod cGMP-gated channel disrupts disk morphogenesis and rod outer segment structural integrity

Phototransduction in mammalian rod photoreceptors is initiated by light impinging on the receptor, rhodopsin, which produces conformational changes that result in an active state. Light-activated rhodopsin then initiates a signaling cascade by activating the G-protein transducin, which in turn disinhibits a rod-specific cGMP phosphodiesterase (PDE6). The subsequent reduction of cytoplasmic cGMP causes closure of a cGMP-gated cation channel (CNGC1) in the plasma membrane of rod outer segment (ROS), resulting in photoreceptor hyperpolarization, which is the primary signal for the photoreceptor communicated to the retina and ultimately to the visual cortex.

The rod CNGC1 is a member of a family of cyclic-nucleotide-gated channels found in several tissues (Kaupp and Seifert, 2002). The channel is composed of three α-subunits and one β-subunit (Weitz et al., 2002; Zheng et al., 2002; Zhong et al., 2002). In heterologous expression systems, the α-subunit is targeted to the membrane and is functional in the absence of the β-subunit (Chen et al., 1994; Kaupp et al., 1989). The β-subunit, however, is required to mimic several physiological and pharmacological properties of the native channel (Korschen et al., 1995). Like the rod α-subunit, the β-subunit contains six putative transmembrane segments, a pore loop region, a voltage sensor-like motif and a cyclic-nucleotide-binding domain but unlike the α-subunit it cannot form functional channels when heterologously expressed (Chen et al., 1994; Colville and Molday, 1996; Korschen et al., 1995).

In addition, the rod β-subunit contains a unique proline- and glutamic-acid-rich N-terminal extension called glutamic-acid-rich protein (GARP). This region exists in an intrinsically unfolded state that may be critical for interaction with target proteins (Batra-Safferling et al., 2006). Two soluble GARP proteins of unknown function are also encoded by the Cngb1 locus and are designated GARP1 and GARP2 (Ardell et al., 2000; Colville and Molday, 1996; Korschen et al., 1995). GARP2 is a 32 kDa protein that is abundant in ROS, and GARP1 is about 65 kDa and is thought to be of very low abundance based on evidence from immunoblots. GARP2 binds tightly to rod PDE6 (Sugimoto et al., 1991) and may inhibit its activation (Korschen et al., 1999). Recently it was suggested that GARP2 only binds to PDE6 in the dark-adapted inhibited state, thereby acting to attenuate dark level noise (Pentia et al., 2006). Previously, a Cngb1 3′-knockout of exon 26 (Cngb1-X26) encoding the pore and six transmembrane-like segments of the β-subunit was reported (Hüttl et al., 2005). This mouse lacks expression of the β-subunit in all tissues but has normal GARP expression and normal photoreceptor structure prior to the onset of degeneration but lacks active phototransduction.

In vitro studies have shown that GARP2 and the N-terminal GARP extension on the β-subunit can both bind peripherin-2 (also known as Rds) (Poetsch et al., 2001), a membrane protein required for disk morphogenesis and structural integrity of the ROS. To date, the functional consequence of this interaction in vivo has not been determined. There have been no direct studies to demonstrate a structural role for the β-subunit or GARP proteins in the photoreceptor. Here, we provide evidence demonstrating that the absence of the β-subunit and GARP proteins disrupts disk morphogenesis and compromises structural integrity, and we show that the knockout reduces circulating current but does not abolish the rod photoresponse.

To facilitate design of a targeting vector, we determined the structure of the murine Cngb1 gene locus using the available mouse genome sequence and RT-PCR to fill identified gaps (not shown). Like the human gene (Ardell et al., 2000), the murine Cngb1 locus consists of at least 36 exons and undergoes multiple modes of alternative splicing, generating transcript and encoded protein diversity in the retina and other tissues (Fig. 1A). Based on coding potential, the gene can be subdivided into exons encoding the GARP portion, which includes an acidic proline repeat domain, and exons encoding a channel-like domain including a Ca²⁺-CaM binding domain, six transmembrane segments, a pore like region, and the cyclic-nucleotide-binding domain.

To abolish expression of all three GARP-containing proteins in the photoreceptor we targeted an upstream region of the gene that contains exons 1 and 2 and a predicted core promoter region common to all three transcripts and replaced this region with a neomycin resistance gene (Neo) driven by a phosphoglycerate kinase (PGK) promoter (Fig. 1B). Homozygous Cngb1-X1 knockout (KO) mice were differentiated from wild type (WT) by multiplex PCR analysis with three primers (primers a-c) that amplify 1.6 and 1.8 kb fragments in KO and WT samples (Fig. 1C). The absence of the first protein coding exon from the genome of the KO mouse was verified using exon-2-specific primers that amplify a 254 bp product in WT and heterozygous cDNA samples but not in cDNA generated from KO mice (Fig. 1D). PCR-independent confirmation of the absence of Cngb1 transcripts in the KO mice was obtained by Northern blot analysis using a probe covering exons 4-9 of the cDNA (Fig. 1E). In WT retina, prominent transcripts of 1.6 and 6.2 kb were apparent (lane +/+) that are not observed in KO mice (lane –/–). A diffuse band just below the 6.2 kb transcript may represent GARP1, which was also not apparent in the KO sample.

Morphology of the KO retina was comparable to WT prior to outer segment maturation at postnatal days 8-10 (PN 8-10) (not shown) but became structurally compromised as outer segments reached maturation. Normal retinal stratification of all layers in the KO retina (Fig. 2B,C) was apparent at 1 month; however, rod outer segments were shorter [WT 26.6±1.3 μm (n=10), KO 20.6±1.4 μm (n=11)] and nonuniform in appearance compared with those in the WT retina (Fig. 2A). Measurements were taken from regions where the cell appeared to be fully elongated and at least part of the tip was visible in the image. At later stages there was a progressive degeneration that progressed to complete loss of rod photoreceptors by 3-4 months. We cannot rule out, however, that the total length may have been slightly longer due to the irregular ROS orientation. Clearly, however, at the light microscopic level, the primary effect of the knockout is confined to ROS. Retinas of heterozygous mice (+/–) appeared structurally normal even up to 2 years of age (see supplementary material Fig. S1). By contrast, homozygous KO mice retina (–/–) exhibited complete loss of the photoreceptor layer within 1 year.

To demonstrate that the KO mouse is a true Cngb1 null in the photoreceptor, western analysis was performed on purified ROS from PN30 mice with an antibody recognizing an N-terminal epitope shared by murine GARPs and the β-subunit. As shown in Fig. 3A, GARPs and the β-subunit (Cngb1) were apparent in WT and heterozygous (+/–) mouse retinal extracts but not in KO (–/–) mice samples consistent with the KO mouse carrying a true photoreceptor null allele of Cngb1.

To assess the levels of the CNG channel α-subunit, quantitative western analysis was performed on enriched ROS from WT and KO mice. A greater than 90% reduction in the α-subunit was observed in ROS from KO mice compared with WT mice (Fig. 3B; Table 1). To differentiate a global effect on ROS protein expression and distribution from a specific effect on channel subunits, we performed quantitative western analysis of ROS protein homogenate with antibodies that recognize disk membrane proteins (peripherin-2, Rom-1, GC1, Rho and ABCA4) and a membrane-associated protein (PDE6). Surprisingly, we found that whereas most proteins were found at comparable levels in WT and KO retina, ABCA4 and GC1 were reduced in abundance by about fivefold (Fig. 3C; Table 1). To determine whether changes in protein levels were due to changes in RNA levels, quantitative PCR was performed and demonstrated that all transcripts examined, including CNGA1, ABCA4 and GC1, were at comparable levels in KO and WT mice (see supplementary material Fig. S2). Only the retinitis pigmentosa guanine nucleotide regulator interacting protein 1 (RPGRIP1) transcript showed a moderate reduction in the KO, which has not been examined at the protein level. Thus, the protein reductions observed are the result of post-transcriptional changes in the KO ROS.

We next examined the spatial distribution of protein expression in 1-month-old retinas to evaluate possible ectopic localization of outer segment proteins. In agreement with the western analysis (Fig. 3A), the antibody against GARP and β-subunit (Cngb1) detected proteins only in the ROS of WT and of heterozygous mice; no GARP region-containing proteins were detected anywhere in the KO retina (–/–) (Fig. 4, upper row). Labeling with a channel α-subunit antibody (Cnga1) also showed a strong signal confined to the outer segment layer of WT and heterozygous mice retinas (middle row). In the KO mouse retina, Cnga1 labeling was greatly reduced but specifically localized to ROS. By contrast, an antibody against peripherin-2 showed comparable labeling in all three genotypes (lower row). The distribution of rhodopsin, ABCA4, Rom-1 and GC1 was also similar for all three genotypes (data not shown). These results indicate a specific effect of Cngb1-X1 KO on channel assembly or ROS targeting.

To determine the basis for the observed alterations of morphology at the fine microscopic level, PN 20 mice retinas were examined by transmission electron microscopy. At this age, outer segments are near maturity. As shown in Fig. 6A, WT mice showed well-developed, nicely aligned, tightly packed outer segments surrounded by a distinct plasma membrane. In the KO mice, tightly packed but misaligned outer segments were apparent (Fig. 5A, Fig. 6B). Arrowheads in Fig. 6B-D point to areas of abnormal elongation of disks. To follow the progression of the ectopic disk morphogenesis, we examined PN 60 mice retina (Fig. 5B, Fig. 7). WT ROS appeared as uniform stacks of disks surrounded by a thin plasma membrane (Fig. 7A). In Cngb-X1 KO mice, in addition to loss of cells, rod outer segments were shorter and had a unique abnormal morphology defined by elongated membranous material (Fig. 5B, Fig. 7B-C, arrowheads) that based on the appearance of PN 20 KO retina (Fig. 6B-D) originates from disks. As shown in Fig. 7D, whorls of abnormally elongated membranes were sometimes observed. To examine the appearance of the outer surface of the intact rod cell, SEM was performed on WT and KO retinas of PN 35 mice (Fig. 8). ROS in WT mice appeared as thinner uniform cylinders throughout the entire length (Fig. 8A,C). KO ROS (Fig. 8B,D) appeared swollen and nonuniform in shape with expanded and constricted regions consistent with the elongated disks observed by TEM. There was no indication that abnormal membranes extended beyond the plasma membrane; rather it appeared that all of the extra membranous material observed by TEM at PN 60 was confined within the cell and most likely represents overgrown disks.

To directly demonstrate rod function, we measured the light-induced current responses from individual photoreceptors (Fig. 9). Initial observation of the tissue with infrared illumination to visualize the cells revealed that the ROS in the KO were shorter (23% shorter histologically; variably shorter in single-cell recordings ranging from 1/3 to 2/3 of WT length); they were also often bent and less rigid than the ROS of WT rods. The electrical responses demonstrated that rod cells were nevertheless functional in the KO mouse but they had a smaller circulating current and were significantly less sensitive to light stimulation compared with WT rods. Fig. 9 shows representative single-cell responses from WT (Fig. 9A) and KO (Fig. 9B) mice. The average amplitude of the light response as a function of the stimulus strength was plotted for both WT and KO rods in Fig. 9C, and the data have been fitted with a Michaelis-Menton function. Our results showed approximately a ninefold decrease in maximum amplitude [total dark current: WT, 12.6±0.9 pA (n=15) and KO, 1.35±0.25 pA (n=9)] and about a 0.8 log-unit shift to higher intensities in the value of I_1/2 for the KO rods. The average value of I_1/2 for 13 KO rods was 4.8-fold higher than for WT rods. The flash sensitivity of the KO rods (pA per unit light intensity) was 34 times less sensitive in the KO mouse (S_f=0.006 pA/photon/μm²) compared with the WT rods (S_f=0.2 pA/photon/μm²). This difference can partly be attributed to the difference in ROS length and the smaller circulating current of the KO rods. The time-to-peak and integration times of the KO rod responses were 1.5-2.5 times slower than those of WT. In spite of these differences in the kinetic properties of the response, the dominant time constant (T_d), measured by the time required after bright flashes for the rod to recover 20% of its circulating current (Pepperberg et al., 1992), was found to be very similar: 219±20 msecond (n=9) in KO rods compared with 205±17 msecond (n=15) in WT rods. This dominant time constant was distinctly different to that measured for WT mouse cones (69±5 msecond) (Nikonov et al., 2008).

Despite more than a decade of study no clear role for the abundant GARP protein has been found. To analyze GARP function we created a murine Cngb1-null allele that eliminates GARP and rod CNG β-subunit expression. These mice exhibit a distinct retinal phenotype without evidence of gross systemic abnormality. The fact that rods in the Cngb1-X1 KO are functional suggests that retinal degeneration in our animal model is due to structural failure that does not abolish signal transduction. This is consistent with a reported model for a structural role for GARP2 and the β-subunit in maintaining the structural integrity of ROS (Poetsch et al., 2001). Physical interaction of the GARP extension of the β-subunit and peripherin-2 alone or in a complex with Rom-1–peripherin-2 (ROM-P2) is predicted to create a physical link between the disks and the plasma membrane. The model also could be extended to include a physical link between disks in the stack mediated by the interaction of GARP2 and ROM-P2 complexes, and perhaps additional unknown proteins. Moreover, both plasma-disk and disk-disk membrane physical links have been observed by freeze-fracture analysis (Kajimura et al., 2000; Roof and Heuser, 1982), and more recently by cryo-electron microscopy (Nickell et al., 2007).

It is hypothesized that the normal signal to indicate the completion of disk synthesis to the appropriate length has been lost or perturbed in the KO mice. This would be consistent with a critical role for the β-subunit (and/or GARP) in regulating disk morphogenesis. It is clear, however, that only a fraction of disks are abnormal in each ROS so other components of the cell are likely to partially compensate for the absence of Cngb1 gene products. Alternatively, Cngb1 gene products may only be required at a critical as yet undefined point in disk morphogenesis.

Comparison of the Cngb1-X1 KO (5′ KO) of exons 1 and 2 and the upstream flanking region described here and the Cngb1-X26 KO (3′ KO) of exon 26 (Hüttl et al., 2005) reveals important phenotypic differences that are indicative of a requirement for GARP in the photoreceptor. In the 3′ KO, the authors observed that disk stacks were intact, which is in contrast to disk stacks reported here that showed signs of disruption of disk morphogenesis and compromised ROS structure. This is consistent with a critical role for GARP proteins in maintaining the structural integrity of the ROS. This idea is supported by studies demonstrating a physical interaction in vitro of GARP2 and the GARP portion of the β-subunit with peripherin-2 (Poetsch et al., 2001).

Hüttl et al. (Hüttl et al., 2005) also observed a lower weight during early postnatal development and high postnatal mortality presumably due to loss of olfaction, which was not observed in our Cngb1-X1 mice (see supplementary material Table S1). The alternatively spliced β-subunit variant used in the olfactory channel (Bonigk et al., 1999; Sautter et al., 1998) is likely to utilize a different promoter just upstream of the novel olfactory 5′-exon predicted from the cDNA sequence (Ardell et al., 2000), and would therefore be predicted to maintain expression in our Cngb1-X1 mice. A photoreceptor-specific promoter would also be predicted to exist in the upstream deleted region in our Cngb1-X1, which has been suggested from analysis of the upstream sequence of the gene (Ardell et al., 2000).

Curiously, the Cngb1-X26 shows a slower retinal degeneration based on morphology but only 3 of 32 rods tested yielded a weak current response (0.9 pA) and rod sensitivity could not be determined. In our 5′-KO mice most of the rods recorded gave a measurable response from which a sensitivity curve was readily generated. There are two major differences in the 5′- and 3′-KO mice that may account for the functional properties observed: whereas channel α-subunit levels were greatly reduced in both KO mouse models, they were significantly lower in the Cngb1-X26 mouse, which retains GARP expression. It is possible that GARP, acting in a dominant negative fashion in the absence of the β-subunit directly inhibits α-subunit transport to the outer segment leading to the even greater reduction of α-subunit in ROS and nonfunctional rods. This also correlates with a critical role for the β-subunit in transport of the channel complex to the plasma membrane. Indeed, it has been demonstrated that the olfactory β-subunit (an alternatively spliced product of the Cngb1 locus) is required for ciliary transport of the olfactory cAMP-gated cation channel (Jenkins et al., 2006; Michalakis et al., 2006). Our observed reduction of GC1 and ABCA4 may similarly reflect a broader role for the photoreceptor β-subunit in transport of select proteins to ROS.

Despite the significant reduction in the amount of the α-subunit, rods in the Cngb1-X1 mice are functional as indicated by electroretinography (data not shown) and as demonstrated by the single-cell responses. The photocurrents observed are presumably carried by homomeric α-subunit channels that manage to be formed and functionally placed in the plasma membrane albeit at significantly lower density than the WT channel in normal rods. Although the cells were tremendously insensitive, the recordings were certainly from rod cells, not cones. The dominant time constant (PDE^* inactivation) for response inactivations was close to 200 mseconds, similar to rods and much slower than the inactivation observed in cones. The abnormal photoresponse observed in the single-cell recordings of others indicates impaired electrophysiological function of photoreceptors after Cngb1 gene knockout. In our Cngb1 KO mice, however, a significant rod signal still remains. The decreased amplitude of the responses observed in the suction electrode records reflects in part the decreased length of the ROS, as well as the decreased expression level of the channel α-subunit.

Defects within the channel-like coding region of the human CNGB1 locus were shown to cause distinct forms of autosomal recessive retinitis pigmentosa (arRP) (Bareil et al., 2001; Kondo et al., 2004). In a large screen of patients with arRP we found a few patients with potential defects in the GARP region of the gene (S.J.P., unpublished results); however, the families analyzed were too small to establish a causal relationship. Thus, our animal model may be an excellent model for human disease.

In summary, this study provides insight into the role of GARP proteins in ROS. Previously, little was known about the importance of GARP proteins in the visual response and in outer segment structure. Indeed, loss of GARP expression could have resulted in a normal phenotype, a complete absence of outer segments akin to a peripherin-2 KO, or something in between. Our studies show in vivo that the absence of GARP proteins including the β-subunit reduces α-subunit channel expression in ROS but that the visual response is not completely abolished. Furthermore, deletion of all GARP proteins has a distinct effect on disk morphogenesis and alters the structure of the outer segments, suggesting an important structural role. Additionally, the surprising finding of significant reduction of ABCA4 and GC1 may indicate a broader transport function for the β-subunit. The precise molecular mechanisms by which the GARP proteins affect outer segment structure and channel expression is beyond the scope of this study. Future studies will be directed towards the selective knockout of GARP proteins to further address their function in the rod photoreceptor.

Homozygous Cngb1^–/– and Cngb1^+/+ mice were propagated by brother-sister matings. All mice used for analysis were at least fifth generation. Homozygotes were genotyped periodically to verify the KO status. All mice used in this study were treated in accordance with the ARVO statement on the use of laboratory animals and the UAB IACUC committee.

A 1.4 kb fragment beginning upstream of the predicted promoter region and a 7 kb region spanning exons 3-10 were cloned from a 129/sv mouse genomic DNA library and transferred into the vector pCngb1-KO flanking a 1.8 kb Neo cassette. All manipulations of embryonic stem (ES) cells and the establishment of mouse chimeras were done at InGenious Targeting Laboratory, Inc. (Stony Brook, New York, NY). Briefly, 10 μg of linearized (NotI) targeting DNA was electroporated into 129SvEv ES cells. Following selection in G418, resistant colonies were expanded and DNA was analyzed by PCR with primer a (Fig. 1B), 5′-GAAGAGCCTACCTTGGAAGCAGAG-3′; and primer b, 5′-TGCGAGGCCAGAGGCCACTTGTGTAGC-3′, which is specific for homologous recombination at the Cngb1 locus (see Fig. 1). Primer c, 5′-TGTCAGCGGCTAGCCAGGAA-3′ was used for amplification with primer a to discern the WT allele. Correctly targeted ES cells were injected into C57BL/6 blastocysts to generate chimeras that were crossed to WT mice to generate stable heterozygotes. All results shown are from a mixed C57BL/6 × 129SvEv background.

Mouse tails were cut and genomic DNA was extracted with a genomic tail DNA purification kit (Promega, Madison, Wisconsin, USA). PCR was performed under standard conditions with Taq DNA polymerase for the following cycling parameters: 94°C denaturation, 5 minutes, for one cycle; 33 cycles of 94°C, 45 seconds; 61°C, 45 seconds; 72°C, 105 seconds; and 1 cycle of 72°C, 7 minutes. Products were visualized with ethidium bromide following separation in agarose gels.

For RT-PCR, total RNA extraction from single mouse eyes of each genotype, and subsequent reverse transcription were done following the manufacturer's protocol (Ambion, Austin, TX). Approximately 1.5 μg total RNA was used in a 20 μl RT reaction, and 1 μl of product was used in each PCR reaction with standard PCR conditions. For QPCR, total RNA was isolated from retinas of 18-day-old wild-type and Cngb1 KO mice using a Versagene RNA tissue kit (Gentra Systems) followed by DNaseI digestion. The quantity and quality of RNA were determined using an ND-1000 spectrophotometer (Nanodrop). Primer sequences were obtained from Primer Bank (http://pga.mgh.harvard.edu/primerbank/) or designed in house with Vector NTI v.10 (Invitrogen). Amplicons ranged in size from 150 bp to 300 bp. Equal amounts of total RNA were used for the cDNA reaction. Briefly, 1 μg of total RNA was reverse transcribed with Improm-II Reverse Transcriptase (Promega) using random decamers and oligo dT primers. The first strand cDNA was diluted fivefold and 1-2 μl of the diluted reaction was used to performing quantitative real-time RT-PCR using SYBR Green/fluorescein (Super Array) on the iQ5 multicolor real-time PCR detection system (Bio-Rad). Each experiment was performed at least in triplicate in 20 μl reaction volumes from three different biological samples, each of which was extracted from a pool of 6-16 retinas. Results were normalized to Gapdh and TBP. For each gene, the threshold value (Ct) was calculated for the average of the triplicates. The fold-change of the gene of interest was then calculated with the ΔCT method.

A PCR product generated with primers ME12F, 5′-CCTGGGATGTGTGACGTACA-3′ and ME15R, 5′-CTAGCTCTGACGGTGTCCTT-3′ was excised from an agarose gel and purified using a PCR product purification kit (Qiagen, Valencia, CA), and sequenced in the Heflin Center for Human Genetics at UAB. Sequence analysis was performed on both DNA strands.

Total RNA was isolated from 6-8 pooled retinas of the same genotype with an RNeasy kit (Qiagen). Then, 5 μg of total RNA of each genotype was size-fractionated through a 0.8% denaturing agarose gel, transferred to a 0.2 μm nylon membrane (Schleicher and Schuell, Keene, NH) overnight and crosslinked with 120 J/cm² UV light. A 323 bp [α-³²P] dCTP-labeled fragment spanning exons 4-9 was used to probe the blot. The probe was incubated overnight and the membrane was washed and exposed to X-ray film at –80°C for 7 days. Methylene blue staining and reprobing with a G3PDH radiolabeled probe was used to demonstrate comparable loading of RNA in each lane.

The entire murine Cngb1 locus was found within two clones available in the NCBI mouse genome sequence database designated BC016201 (covers exons 1-12a) and XM-286113 (covers exons 15-33), except for a small gap from exon 12a to 15. The gap was filled by PCR amplification with primers within exons 12 and 15, to generate a 540 bp product that was sequenced. Freely available genome sequence manipulation programs were used to determine the gene structure: http://www.genome.ucsc.edu/cgi-bin/hgBlat; http://www.ncbi.nih.gov/; http://www.genebee.msu.su/services/malign_reduced.html.

Three independent experiments were carried out for comparative analysis of the relative level of protein expression by western blottings. In each experiment, eight retinas dissected each from C57BL/6 wild-type (+/+), heterozygous (+/–) and homozygous (–/–) GARP KO mice were each immersed in 400 μl of Tris buffered saline (TBS; 20 mM Tris, pH 7.3, 0.15 mM NaCl, 2 mM MgCl₂) containing Complete protease inhibitor (Roche) at 4°C. The retinas were homogenized by passing the solution through a 22 gauge needle 12 times. The homogenate was applied on top of a 50% (w/v) sucrose solution in TBS and spun in Beckman Optima TLS55 rotor at 26,000 rpm for 30 minutes at 4°C. The ROS membranes were collected from the top of the 50% sucrose layer and washed once with 3.0 ml TBS. The pellet was resuspended in 140 μl TBS containing Complete protease inhibitor and added to SDS denaturing cocktail. The proteins were fractionated on a 9% SDS polyacrylamide gel (20 μl/well) and transferred to Immobilon-FL membranes (Millipore) using a Bio-Rad semidry transfer apparatus. The blots were blocked with 1.0% milk in PBS and subsequently labeled for 1 hour with hybridoma culture fluid containing monoclonal antibodies to the N-terminus of CNGB1 and GARP proteins (GARP 4B1) (Poetsch et al., 2001); CNGA1 (PMc 2G11); ABCA4 (Rim 3F4); rhodopsin (Rho 1D4), peripherin-2 (Per 5H2), Rom-1 (Rom 1C6), and guanylate cyclase-1 (GC 2H6), or a polyclonal antibody against PDE6 (Affinity BioReagents). The blots were subsequently labeled with goat anti-mouse Ig or goat anti-rabbit Ig conjugated with Alexa 680 (Molecular Probes, Eugene, OR) at a dilution of 1:20,000 in PBS. Finally, the blots were rinsed three times in PBS containing 0.1% Tween 20 and the bands were visualized and quantified by infrared scanning using the LI-COR Odyssey system (Lincoln, NE). The values from the three experiments were averaged and the standard deviation determined.

For immunocytochemistry, the retinal cryosections were blocked with 0.1 M phosphate buffer (PB), pH 7.3, containing 0.2% Triton X-100 and 10% goat serum for 20 minutes and labeled overnight with the primary antibody. The sections were washed in PB and labeled for 1 hour with Alexa Fluor 488-conjugated secondary antibody (green). Images were merged with DAPI nuclear staining (blue). The sections were viewed on a Zeiss LSM510 confocal microscope.

Morphologic analysis was done with minor modification of a procedure we have previously used to analyze the effect of drugs on the morphology of rat retina (Pittler et al., 1995). Briefly, for light microscopy following sacrifice, the eyes were oriented with marker dyes and, fixed with a mixture of 1.2% paraformaldehyde and 0.8% glutaraldehyde in 0.1 M PBS, pH 7.4. The anterior segment including the lens was removed and post-fixed in 2% osmium tetroxide. The eye cup was embedded in Epon-Araldite, and thin and semithin sections were cut with an ultramicrotome beginning at the nasal cutting toward the temporal quadrant. Sections for examination were always from a 50 μm region around the optic nerve. Semithin (0.5-1 μm) sections were stained with 2% toluidine blue. Light micrographic images were obtained on a Zeiss Axioplan 2 imaging and Axiophot 2 universal microscope equipped with a 4 megapixel digital camera. Ultrathin sections for transmission electron microscopy (TEM) were prepared as previously described (Pittler et al., 1995) and visualized with an FEI Technai Spirit electron microscope (Phillips) with AMT camera imaging software. For scanning electron microscopy (SEM), fixed samples were osmicated and dehydrated through a series of ethanol washes to 95%. Samples were transported to the University of Alabama at Tuscaloosa Optical Analysis Facility where they were dehydrated with 100% ethanol, critical point dried, and sputter coated for visualization in a Hitachi S2500 SEM microscope.

At least 1-3 hours prior to enucleation, the animal was dark adapted. Details of the recording procedure have been reported (Baylor et al., 1984; Schnapf et al., 1988; Woodruff et al., 2008). Briefly, a small portion of retina was chopped into fine pieces and placed in a recording chamber perfused with bicarbonate-buffered Lockes solution equilibrated with 95% O₂, 5% CO₂ and kept at 37°C. The recording solution contained: NaC1 (140 mM), KC1 (3.6 mM), CaC1₂ (1.2 mM), MgC1₂ (2.4 mM), H₂CO₃ (20 mM), HEPES buffer (3 mM, pH 7.4), D-glucose (10 mM), EDTA (0.02 mM), supplemented with amino acid and vitamin mixtures (Gibco, Grand Island, NY). Recordings also were made with bicarbonate-buffered DMEM (D2902 Sigma, St Louis, MI), supplemented with the Na⁺ salts of gluconate, succinate and glutamate (Woodruff et al., 2002). Photocurrent recordings were made by drawing a single ROS, still attached to a small retinal fragment, into a suction electrode. Signals were recorded with a patch-clamp amplifier (Axon Instruments, Burlingame, CA or Warner Instruments Co., Hamden, CT), and digitized as described previously (Kraft et al., 2005; Woodruff et al., 2008).