Impaired Opsin Targeting and Cone Photoreceptor Migration in the Retina of Mice Lacking the Cyclic Nucleotide-Gated Channel CNGA3

Abstract

purpose. To characterize molecular and cellular changes in the mouse retina caused by the genetic deletion of the cone cyclic nucleotide-gated channel (CNG) subunit CNGA3.

methods. Retinas of wild-type and CNGA3-deficient (CNGA3^−/−) mice from 9 days up to 22 months of age were analyzed by immunohistochemistry, electron microscopy, and molecular biological methods.

results. CNGA3^−/− cones failed to transport opsins into outer segments, downregulated various proteins of the phototransduction cascade, and induced apoptotic death. Loss of CNGA3 did not affect the transcription of cone-specific genes. Cone degeneration was evident from the second postnatal week on, and it proceeded significantly faster in the ventral than in the dorsal part of the retina. Ventral cones were completely missing after the third postnatal month, whereas residual dorsal cones were present, even in 22-month-old knockout mice. CNGA3^−/− cone somata exhibited profoundly delayed migration during postnatal development. At the time of eye opening, most CNGA3^−/− cones had displaced somata localized close to or in the outer plexiform layer. These cones lacked the characteristic synaptic pedicle, but revealed synapselike contacts to second-order neurons at their somata. At later stages, most of the surviving CNGA3^−/− cones had correctly located somata and morphologically normal synapses.

conclusions. The loss of CNGA3 impairs the targeting of cone opsins and the expression of other visual cascade proteins. In addition, CNGA3 appears to be essential for normal postnatal migration of cone somata. After loss of cone outer segment proteins, CNGA3^−/− cones induce apoptotic cell death.

Achromatopsia (ACHM, rod monochromatism) is a rare, autosomal recessively inherited ocular disorder characterized by specific loss of cone function but preserved rod photoreceptor function. 

The functional aspects of ACHM have been well characterized. ¹ By contrast, there is only a little information about the morphologic and structural alterations in the retinas of patients with ACHM. The few pathohistologic studies of human ACHM indicate that this disease is associated with a progressive loss of cone photoreceptors. ^{2 3 4} However, time course, regional distribution, and extent of retinal cell damage have not been systematically determined so far. Moreover, the molecular mechanism underlying photoreceptor degeneration in ACHM is not known. 

Genetic analysis of patients with ACHM revealed that approximately 25% carry mutations in the CNGA3 gene, ⁵ which encodes the A subunit of the cone cyclic nucleotide-gated (CNG) channel. ⁶ Another 40% to 50% of the patients with ACHM were found to have mutations in the gene encoding the B subunit of the same channel. ^{7 8} The CNG channel plays a key role in visual transduction. The protein is expressed in cone outer segments (COSs) where it couples light-induced alterations of the cGMP concentration with changes in membrane potential and cellular Ca²⁺ concentration. ⁹ As a consequence, loss-of-function mutations in this channel completely disrupt cone phototransduction. Moreover, mutations in the CNGA3 subunit that do not completely destroy channel function but affect its biophysical properties have been identified in patients with incomplete achromatopsia. ^{5 10}  

Recently, we developed a mouse line deficient in the CNGA3 gene. CNGA3-deficient (CNGA3^−/−) mice have a complete loss of cone function but normal rod-mediated vision, which indicates that they represent a valid mouse model of ACHM. ¹¹ In this initial analysis, we observed a degeneration of COSs and loss of cone photoreceptors in adult CNGA3^−/− mice. ¹¹ However, time course and spatial distribution of cone cell death, as well as adaptive alterations in the morphology and connectivity of neurons in the CNGA3^−/− retina, have not been systematically determined so far. In the present study, we investigated these important issues. Moreover, we set out to study the molecular basis of cone cell death in the CNGA3^−/− retina and studied alterations in the expression of key proteins of the visual transduction cascade. 

Materials and Methods

Animals

The generation of CNGA3^−/− mice has been described elsewhere. ¹¹ For all experiments, age-matched wild-type (CNGA3^+/+) and CNGA3^−/− mice, backcrossed for 10 or more generations on the 129sv background, were used. Mice were kept under a normal 12-hour light–dark cycle. The animals were treated in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Generation of Anti-CNGB3 Antibody

A polyclonal rabbit antibody directed against the carboxyl terminal peptide of murine CNGB3: NH₂-KVDLGRLLKGKRKTTTQK-COOH was generated and tested by standard techniques. ¹²  

Histology

Immunohistochemistry was performed on cryosections, vibratome sections, or retinal wholemount preparations, as described previously. ¹³ The sources and working dilutions of primary antibodies are listed in Table 1 . For secondary detection, we used FITC donkey anti-rabbit or anti-mouse IgG (Dianova, Hamburg, Germany), Alexa 488 or 594 goat anti-mouse or anti-guinea pig IgG (Molecular Probes, Eugene, OR), and biotin donkey anti-goat IgG (Vector Laboratories, Burlingame, CA), followed by DY-547 streptavidin (Dyomics, Jena, Germany). To visualize cell nuclei, retinal slices were treated with 5 μg/mL Hoechst 33342 (Molecular Probes). 

The lectin peanut agglutinin (PNA), a specific extracellular surface marker of cone photoreceptors, ²⁹ either FITC-labeled PNA (1:100; Sigma-Aldrich, Deisenhofen, Germany) or biotinylated PNA (1:100; Sigma-Aldrich) followed by detection with FITC-avidin (Rockland, Gilbertsville, PA), DY-547 or DY-631 avidin or streptavidin (Dyomics), was used to visualize all cones. 

Specimens were analyzed by fluorescence microscopy (Axioplan 2; Carl Zeiss, Oberkochen, Germany) or confocal microscopy (LSM510 Meta; Carl Zeiss). Electron microscopy was performed as described previously. ¹³  

Terminal deoxynucleotide transferase (TdT)-mediated dUTP nick end labeling (TUNEL) ³⁰ was performed on retinal cryosections with an apoptosis peroxidase in situ detection kit (ApopTag; Intergen, Purchase, NY), used according to the manufacturer’s instructions. 

Quantification

TUNEL-positive cells were quantified in subfields (0.05 mm²) of the outer nuclear layer (ONL) of retinal images taken at 200× magnification. Cone outer segments were quantified on images (400×) from the central part of the dorsal and ventral quadrants of retinal wholemounts stained with PNA and anti-opsin antibodies, respectively. Quantitative analysis was performed using a software quantification tool (Axiovision; Carl Zeiss). 

Semiquantitative RT-PCR

Retinas were dissected, and total RNA was isolated (RNeasy; Qiagen, Hilden, Germany) and subsequently treated with DNAse I (Roche, Mannheim, Germany). First strand-cDNA was synthesized from 6 μg of RNA (Superscript II H⁻-Kit; Invitrogen, Karlsruhe, Germany) using oligo-dT-primers. cDNAs of CNGB3, G_αt₂, middle-wavelength–sensitive (MWS) opsin, and hypoxanthine phosphoribosyl-transferase (HPRT; internal control) were amplified with the primers and conditions listed in Supplementary Table S1. Amplicons were separated on 5% polyacrylamide gels, stained with ethidium bromide, and analyzed densitometrically on a system (Gel Doc 2000; Bio-Rad, Munich Germany), with accompanying computer software (Quantity One; Bio-Rad). The intensity of the amplicons was normalized to the intensity of the HPRT amplicon obtained from the same cDNA. All primer pairs used were intron-spanning to avoid amplification of genomic DNA. 

Western Blot Analysis

Membrane proteins were isolated from murine retinas as previously described. ³¹ Equal amounts of protein were separated by 10% SDS-PAGE followed by Western blot analysis according to standard procedures. The following antibodies were used: anti-MWS or anti-short-wavelength–sensitive (SWS) opsin (Chemicon, Hofheim, Germany; 1:1000) and anti-neuron-specific enolase (NSE; Biomol International LP, Exeter, UK; 1:1000). 

Statistics

Data were analyzed by unpaired t-test or analysis of variance (ANOVA) for repeated measures, followed by the Newman-Keuls post hoc test, and are presented as mean ± SEM. Statistical significance was accepted if P < 0.05. 

Results

Lifespan of Cone Photoreceptors in Different Regions of the CNGA3^−/− Retina

To determine the lifespan of cones in CNGA3^−/− mice, we made use of the cone marker PNA, which labels the extracellular matrix surrounding cone photoreceptors. Figure 1shows a series of PNA-stained retinal wholemounts derived from wild-type and CNGA3^−/− mice of different ages. The focus of the microscope was always adjusted to visualize COSs. Although in wild-type mice the density of PNA-positive COSs did not significantly differ between postnatal week (PW)4 and postnatal month (PM)17 (Figs. 1B 1C) , CNGA3^−/− mice revealed a progressive loss of PNA-stained COSs (Figs. 1D 1E 1F) . In the ventral and nasal portion of the retina, loss of PNA-positive COSs was first detectable after PW3 (not shown) and proceeded rapidly (Figs. 1D 1E 1F 1I) . The number of PNA-labeled COSs in this part of the CNGA3^−/− retina was reduced by 56.6% ± 2.9% (n = 6, P < 0.0001) and 86.8% ± 3.3% (n = 5; P < 0.0001) at PW4 and PM17, respectively. At PW4, approximately 75% of COSs were still present in the temporal and dorsal sectors (loss of 25.4% ± 5.5%; n = 6; P < 0.05), and even at PM17 only 46.4% ± 5.9% (n = 5; t-test: P < 0.01) of the cones were lost in comparison with the wild-type retina (Figs. 1G 1H) . In conclusion, cone loss was much more profound in the ventral and nasal part than in the dorsal part of the CNGA3^−/− retina. Recently, it has been speculated that NSE may serve as a survival factor in cortical neurons and cone photoreceptors. ^{32 33} To test for this hypothesis, we determined the expression levels of NSE in CNGA3^−/− mice. In the mouse retina the γ,γ-isoform of NSE (46 kDa) was the predominant form, as detected by Western blot using a polyclonal anti-NSE antibody (Fig. 2A) . In retinal slices from wild-type mice, NSE was abundantly expressed in inner segments of rods and cones, whereas in COSs it was found only at very low density (Fig. 2B) . Starting from PW3, we observed increased NSE immunoreactivity in a subpopulation of COSs in the CNGA3^−/− retina (Figs. 2C 2D) . The number of COSs with upregulated NSE was profoundly higher in the dorsal than in the ventral part of the CNGA3^−/− retina (Figs. 2E 2F) . We noted that all surviving cones in the dorsal retina of aged knockout mice (PM17) exhibited strong NSE immunoreactivity (Fig. 2G) . By contrast, cones in age-matched wild-type retina did not reveal upregulation of NSE (Fig. 2H) . 

Downregulation and Disturbed Trafficking of Opsins in CNGA3^−/− Cones

Our previous electron microscopy experiments suggest that COS disks are formed but are disorganized in the CNGA3^−/− retina. ¹¹ To achieve an additional view on the structural status of COSs, we performed immunostaining for cone opsins, which represent a major component of COS proteins. A monoclonal antibody (COS1) directed against the MWS (red/green) cone opsin labeled almost all PNA-positive COSs in the dorsal wild-type retina, a region where MWS opsin is the predominant cone opsin (Figs. 3A 3B) . By contrast, in the PW4 knockout retina, very few PNA-positive cones revealed MWS opsin immunoreactivity (Figs. 3C 3D) . Similarly, at PW4, a monoclonal antibody (OS2) directed against the SWS (blue) cone opsin labeled only a small percentage of PNA-positive cells in the ventral retina (Figs. 3E 3F) , a region where SWS opsin is abundantly expressed in the wild-type retina. Thus, in CNGA3^−/− mice, outer segments of both types of cones were essentially devoid of opsins. We addressed this unexpected finding in another set of experiments using polyclonal anti-MWS and anti-SWS opsin antibodies (Figs. 3G 3H 3I 3J 3K 3L) . The specificity of both antibodies was confirmed in Western blot analysis and immunostaining of wild-type retinal sections (see Supplementary Fig. S1). As expected, the MWS-specific antibody intensely stained the COSs of a PW3 wild-type retina (Fig. 3G) , whereas PNA-labeled COSs of age-matched CNGA3^−/− retinas revealed only residual, if any, MWS opsin staining (Figs. 3H 3I) . However, the MWS-specific antibody detected substantial amounts of opsins in the inner segments, somata, and terminals of knockout but not of wild-type cones (Fig. 3I) . The same staining pattern was also observed with the anti-SWS antibody (not shown), as well as with another polyclonal anti-opsin antibody (anti-SWS opsin, JH455, ²⁸ data not shown). Even aged MWS cones revealed an opsin signal in the inner segment, the soma, and the terminal, but never showed opsins in the outer segment (Figs. 3J 3K) . It is noteworthy that mislocalized opsins were not labeled by the monoclonal COS1 and OS2 antibodies, probably because these antibodies are raised against retinal membrane fractions and therefore detect opsin epitopes that are present only if the proteins are correctly folded and expressed in the outer segment. To assess the total opsin levels of wild-type and CNGA3^−/− retinas, we performed Western blot analysis with retinal membrane fractions. Figure 3Lshows a blot from 3-week-old mice probed with the polyclonal anti-MWS antibody. The antibody recognized a specific 42-kDa band with an intensity that was much weaker in CNGA3^−/− than in wild-type mice. The difference in intensity indicates that the amount of mislocalized opsin in a CNGA3^−/− cone photoreceptor is much lower than the amount of opsin normally present in a wild-type COS. The downregulation of opsins could be caused by inhibition of gene transcription or by posttranscriptional effects. To differentiate between these two options, we determined mRNA levels by semiquantitative RT-PCR. In support of a posttranscriptional mechanism, there was no significant difference in the level of MWS opsin mRNA up to PM3 (Fig. 3M , left). Even in 1-year-old mice, in which only a few cones contain opsin, the mRNA levels were approximately 35% of the control level (Fig. 3M , right). 

Downregulation of Proteins of the Cone Visual Cascade

We next asked whether other proteins of the visual transduction cascade were also downregulated in COSs of CNGA3^−/− retinas (Fig. 4) . The B subunit of the cone CNG channel (CNGB3) was completely missing in CNGA3^−/− cones at all time points investigated (Figs. 4A 4B) . Cone transducin (G_αt₂) was profoundly downregulated at eye opening (Figs. 4C 4D) , and expression levels of guanylyl cyclase E (GCE) as well as guanylyl cyclase activating protein 1 (GCAP1) declined after the second postnatal week (Figs. 4E 4F 4G 4H) . As observed with opsins, expression of mRNA for CNGB3 (Fig. 4I)and G_αt₂ (Fig. 4J)was not different from wild-type retina at PM3. 

With regard to the labeling patterns for rhodopsin, the rod CNG channel subunits (CNGA1 and -B1), guanylyl cyclase F (see Supplementary Fig. S2), and GCAP2 (not shown) knockout mice did not differ from wild-type control subjects. 

Gliosis and Apoptotic Cell Death in the CNGA3^−/− Retina

The loss of CNGA3 triggered degeneration of COSs and the eventual loss of cone photoreceptors. Müller glial cells are known to respond immediately to retinal injury by profound upregulation of intermediate fibers, such as glial fibrillary acid protein (GFAP). ^{34 35} To define the onset of degeneration, we investigated the expression of GFAP in the CNGA3^−/− retina. In the wild-type retina GFAP immunoreactivity was restricted to the neurofilament layer (NFL), where it was found in astrocytes surrounding retinal blood vessels and at the end feet of Müller glia (Fig. 5A) . In CNGA3^−/− mice, the first signs of Müller gliosis were evident shortly after eye opening (P12). At PW3 we observed strong upregulation of GFAP across the retina (Fig. 5B) . In aged CNGA3^−/− mice, this phenomenon was even more pronounced (not shown). 

CNGA3^−/− cones frequently revealed blebbing of outer and inner segments (Fig. 5C , arrows) and nuclear fragmentation (Fig. 5D , arrow) indicating ongoing apoptosis. To detect and quantify the number of photoreceptors displaying DNA fragmentation, a hallmark feature of apoptosis, we performed TUNEL histochemistry (Figs. 5E 5F 5G) . In wild-type mice, the number of TUNEL-positive photoreceptor nuclei continuously decreased during retinal maturation (Fig. 5G) . After PW4, nuclear fragmentation was only rarely observed. In contrast, DNA fragmentation in the ONL of the CNGA3^−/− retina increased after P12 and reached a maximum at PW3. Thereafter, the number of TUNEL-positive cells declined but was still significantly higher than in the wild-type retina (Fig. 5G) . In 3- to 4-week-old CNGA3^−/− mice, we frequently detected activation of caspase-3, one of the key executor proteases involved in apoptosis (Figs. 5H 5I) . Furthermore, we observed release of cytochrome c, another apoptosis-specific event (Fig. 5K) . In the age-matched wild-type retina, activation of caspase 3 and release of cytochrome c were not observed (Figs. 5J 5L) . 

Delayed Migration of Cone Somata in CNGA3^−/− Mice

COSs in the rodent retina begin to develop at the end of the first postnatal week and express opsins from then on. ³⁶ We stained vertical retinal sections of premature wild-type and CNGA3^−/− retinas with the polyclonal anti-opsin antibodies to investigate opsin expression during postnatal development. Figures 6A 6B 6Cshow wild-type retinal sections at time points shortly before (postnatal day [P]9) and after eye opening (P12) stained with anti-SWS opsin (Fig. 6A)and anti-MWS opsin (Figs. 6B 6C) . At P9, COSs were already strongly opsin positive, whereas cone somata were labeled more weakly. The identified cone somata were exclusively located in the upper half of the ONL. Other than in the wild-type retina, opsin immunoreactivity in the CNGA3^−/− retina was concentrated in cone somata, synapses, and inner segments (Figs. 6D 6E 6F) . Most of the CNGA3^−/− COSs did not show opsin immunoreactivity (Fig. 6F) , in line with results from older mice (Figs. 3H 3I 3J 3K) . 

Unlike the wild-type somata, most cone somata of CNGA3-deficient mice were found at P9 in the outer plexiform layer (OPL; Figs. 6D 6E ). This phenomenon was observed to the same extent in the dorsal and the ventral part of the CNGA3^−/− retina and for both MWS and SWS cones. At P12 a significant number of mutant cone somata were still present in the OPL (Fig. 6F , arrows). Staining with Hoechst dye confirmed that displaced opsin-positive somata contained a nucleus (Fig. 6G) . At PW3, three classes of cones coexisted in the mutant retina: cones with somata located in the OPL (Fig. 6H , arrows), with somata located in the middle of the ONL (Fig. 6H , asterisk), or with normally positioned somata (Fig. 6H , arrowheads). Rarely, single opsin-positive cones were found in the INL (not shown). Only in aged mice, most of the (few) remaining cone somata (which were exclusively MWS cones) were correctly localized in the outermost part of the ONL (see Fig. 3K ). Costaining with an antibody against the cone-specific glycogen phosphorylase (Ref. ²⁴ and Haverkamp S, unpublished data, 2004) and anti-SWS opsin (Fig. 6I)confirmed that displaced cones featured a long axon reaching to the outer limiting membrane and possessed inner segments (IS; Fig. 6I ). 

As shown in Figure 6 , displaced cone somata were frequently localized close to or within the OPL, and some of these seemed to be merged with structures of this synaptic cell layer. We analyzed the ultrastructure of displaced cone somata and found that the vitreal part of the soma (base) exhibited structural features, including multiple synaptic ribbons and synaptic contacts, that are characteristic of cone pedicles (Figs. 7A 7B 7C) . 

The horizontal cells of CNGA3^−/− mice reacted with the formation of sprouting neuronal extensions penetrating into the ONL (Fig. 7D) . We found calbindin-positive horizontal cell processes extending into the ONL in CNGA3^−/− mice of any age from 12 days to 22 months. By contrast, in wild-type mice older than 12 days, horizontal cell sprouting was never observed (Fig. 7E) . 

Presence of Normal-Shaped Cone Synapses in CNGA3^−/− Mice

The deletion of CNGA3 profoundly slowed down the migration of cone somata but did not abolish it. Thus, most surviving cones in the adult CNGA3^−/− retina had normally localized somata. We analyzed the synaptic pedicles of these cones and found wild-type–like morphology. The electron microscopic image in Figure 7Fshows a synapse in the dorsal part of a PM6 CNGA3^−/− retina that contains typical elements of functional cone pedicles. For example, this cone pedicle bears several synaptic ribbons (Fig. 7F , arrows) and flat synaptic contacts (Fig. 7F , arrowheads). Notably, we found immunoreactivity for the glutamate receptor subunit 5 (GluR5) in the OPL of CNGA3^−/− retinas that was colocalized with PNA staining (Figs. 7G 7H) , a hallmark feature of functional cone synapses contacting OFF-type bipolar cells. ²³  

Discussion

In this study, we investigated the consequences of the genetic deletion of CNGA3 on the structure and postnatal development of the murine retina. CNGA3^−/− mice revealed a progressive loss of cone photoreceptors. In contrast, the number and morphology of rod photoreceptors was unaffected in these mice. This finding is in line with our previous ERG recordings showing that CNGA3^−/− mice display a normal rod but lack any cone photoresponse. ¹¹ Cone loss was most severe in the inferior part of the CNGA3^−/− retina. Cones containing SWS opsin, the predominant cone-type in this part of the retina, were hardly detectable after PM3, whereas several MWS cones survived in the dorsal retina, even in aged (PM22) CNGA3^−/− mice. A comparable asymmetrical pattern of cone loss was also observed in the retina of GCE-deficient mice. ¹⁸  

Unexpectedly, most of the PNA-positive COSs of CNGA3^−/− mice were devoid of opsins. However, opsins were found in inner segments, somata, and terminals. Thus, mutant cones are principally able to synthesize opsins, but they fail to route the proteins into the outer segments. The molecular details of cone opsin trafficking are not known so far. In rod photoreceptors, rhodopsin is synthesized in the inner segments and then actively transported through the connecting cilium into the outer segment. ³⁷ An equivalent mechanism may exist in cones and could be impaired in CNGA3^−/− mice. The lower total amount of opsins in mutant retina, as detected in Western blot analysis, suggests that opsin synthesis is downregulated and/or that misrouted opsin is degraded. The latter mechanism is supported by our finding that MWS opsin mRNA levels were not different between wild-type and mutant mice up to PM3. We also observed a rapid downregulation of other proteins of the cone phototransduction cascade. Moreover, the B subunit of the cone CNG channel (CNGB3) was missing in CNGA3^−/− COSs. This finding suggests that CNGB3 needs CNGA3, to be normally targeted in cone photoreceptors, and that it is rapidly degraded in the absence of CNGA3. Analogous to the opsins, mRNA levels of other COS proteins were not decreased up to PM3 favoring a posttranscriptional mechanism underlying protein downregulation. Only in aged retina (PM12) did we observe a decrease in mRNA levels, which very likely reflects the fact that a major fraction of cones had induced apoptosis up to this time point. 

Despite the fact that some dorsal cones in the CNGA3^−/− retina survived for long periods we found clear evidence of an early onset of retinal degeneration. Induction of GFAP in Müller glial cells, a well-established indicator of stress and degeneration in the central nervous system (CNS) and the retina ^{38 39} was observed shortly after eye opening (P12). CNGA3^−/− cones most likely died via apoptotic pathways. The peak phase of apoptosis in CNGA3^−/− mice correlated well with the onset of cone photoreceptor loss (PW3–4). During this peak phase, we found 20% to 30% more TUNEL-positive photoreceptors in the ventral than in the dorsal part of the CNGA3^−/− retina (data not shown). This is in agreement with the faster progression of loss of cones in the ventral CNGA3^−/− retina. The cellular pathways leading to apoptosis in CNGA3^−/− cones remain to be determined. One possible mechanism explaining the early onset of degeneration is based on the idea that the cone CNG channel, besides its role as the Na⁺/Ca²⁺ influx pathway, probably serves as an important structural protein in cones. Although proteins binding to the cone channel have not yet been identified, it is reasonable to assume that the cone channel, in analogy to the related rod CNG channel, ^{40 41 42} is part of a highly structured protein complex. Thus, the loss of CNGA3 would not only impair the phototransduction cascade but could also lead to a structural alteration of COSs. In support of this idea we have previously shown that CNGA3^−/− COS disks appear partly disorganized. ¹¹ Cone opsin may fail to be targeted into these irregular COSs and, as a consequence, accumulate in inner segments and somata. Given the high expression levels of opsins, the mislocalization and accumulation of these proteins could well induce cellular stress and apoptosis. Misrouting and accumulation of proteins have been found frequently in degenerative processes in neurons. ⁴³  

Most surviving cones in adult CNGA3-deficient mice were located in the dorsal retina. Because ventral cones preferentially express SWS opsin, whereas MWS opsin is mainly found in dorsal cones, ^{44 45} one might speculate that mislocalized SWS opsin is more toxic than MWS opsin. It might also be argued that ventral cones disappear earlier because they are more susceptible to light damage. However, this option appears unlikely, because recent studies in rodents indicate that light damage is most profound in central and dorsal regions of the retina. ⁴⁶ Finally, MWS cones may possess additional survival mechanisms. For example, it has been reported that brain-derived neurotrophic factor and its receptor TrkB are found in COSs of MWS but not SWS opsin cones. ⁴⁷ Recently, it has also been proposed that NSE serves as a survival factor in cone photoreceptors. ³³ We found enhanced levels of NSE in the outer segments of surviving dorsal CNGA3^−/− cones. To date, it is not clear whether NSE upregulation causes the elongated lifespan of these cones or is only an epiphenomenon. 

Another key finding of this study is the identification of CNGA3 as a determinant of postnatal migration of cone somata. In the embryonic and early postnatal murine retina, cone photoreceptors are situated in the outermost part of the neuroblastic layer. ⁴⁸ Cone photoreceptor somata undergo a first migratory phase after P4, moving downward toward the synaptic cell layer, and then start to be repositioned upward. ⁴⁹ Whereas at P12, most wild-type cone somata had reached their final destination in the outermost third of the ONL, most of the CNGA3^−/− cone somata were still present in the lower half of the ONL at the same time point. However, in aged knockout mice, most surviving cones possessed normally localized somata. Thus, the migration of cone somata is delayed but not completely abolished in the CNGA3^−/− retina. At present, the molecular pathways controlling migration of photoreceptor somata are unknown. Our results suggest that the CNG channel plays an important role in this process. It is noteworthy that TUNEL-positive photoreceptors in CNGA3^−/− mice were found primarily in the upper half of the ONL. Thus, displaced cones do not seem to be more susceptible to apoptosis than normally localized cones. 

A large number of displaced cone somata were found in the OPL, containing characteristic elements of cone pedicles at their bases (e.g., multiple triad-associated synapses [composed of two lateral horizontal cell dendrites and one central bipolar cell dendrite opposite a ribbon], as well as flat synaptic contacts). These unusual synapses may represent an immature stage occurring during cone differentiation. In the dorsal part of the retina of older knockout mice, we consistently found, though at low density, normal-shaped cone pedicles invaginated by bipolar and horizontal cell dendrites. This implies that CNGA3^−/− cones are able to form and maintain synaptic contacts to second-order neurons. In support of this observation, we have previously found that mice lacking both functional cones and rods possess structurally normal synaptic contacts up to the time of complete loss of photoreceptors. ¹³ Despite the seemingly normal morphology of synapses, signaling between cones and second-order neurons is probably disturbed. The horizontal cells in the CNGA3^−/− retina extended sprouting neurites into the ONL. This phenomenon may be a general response to degeneration, since it was also observed in other mouse models of retinal degeneration. ^{13 50}  

Despite the morphologic differences between murine and human retina, the molecular and cellular changes observed in CNGA3-deficient mice could well be the same in human achromats. For example, mislocalized cone nuclei have also been observed in the retina of a human achromat. ² In our study, CNGA3-deficient cones displayed several hallmark features of wild-type cones (e.g., expression of cone opsins and formation of synaptic contacts to second-order neurons). In addition, a subpopulation of MWS-opsin–positive cones survived for long periods in the CNGA3-deficient retina, in agreement with pathohistologic data from patients with achromatism who display incomplete cone cell loss. ^{2 3 4}  

In principle, it seems to be feasible to rescue the CNGA3-null phenotype functionally by reintroducing the cDNA of the CNGA3 subunit (e.g., by using a viral vector system). Our results may be useful for approaches that have the goal of restoration of cone photoreceptor function in human achromats. 

 Supported by Deutsche Forschungsgemeinschaft.

 Submitted for publication December 20, 2004; accepted January 2, 2005.

 Disclosure: S. Michalakis, None; H. Geiger, None; S. Haverkamp, None; F. Hofmann, None; A. Gerstner, None; M. Biel, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Antibodies and Dilutions

Antigen	Host, Antibody, Dilution	Source
Active-Caspase 3	Rabbit, anti-cleaved caspase 3, 1:100	Cell Signaling 14
Calbindin	Mouse, anti-calbindin, 1:2000	Swant 15
CNGA1	Mouse, PMc1D1, 1:50	16
CNGA3	Rabbit, anti-CNGA3, 1:100	11
CNGB1	Rabbit, C-AbmCNGB1, 1:500	17
CNGB3	Rabbit, anti-CNGB3, 1:100	Present study
Cone-transducin	Rabbit, anti-Gαt2 (sc-390), 1:100	Santa Cruz Biotechnology 18
Cytochrome c	Mouse, anti-cytochrome c, 1:100	BD Pharmingen 19
GCAP1	Rabbit, anti-GCAP1 (UW14), 1:500–1000	20
GCE	Rabbit, anti-GCE (K 286), 1:200	21
GCF	Rabbit, anti-GCF (L 671), 1:200	21
GFAP	Mouse, anti-GFAP (Cy3-labeled), 1:1000	Sigma-Aldrich 22
GluR5	Goat, anti-GluR5 (sc-7617), 1:100	Santa Cruz Biotechnology 23
Glycogen phosphorylase	Guinea pig, anti-glycogen phoshorylase, 1:1000	24
MWS opsin	Mouse, COS1, 1:100	25
MWS opsin	Rabbit, anti-MWS opsin, 1:500	Chemicon International Inc.
NSE	Rabbit, anti-NSE, 1:200	Biomol International LP 26
Rhodopsin	Mouse, Ab-1 (clone RET-P1), 1:100	Lab Vision 27
SWS opsin	Mouse, OS2, 1:100	25
SWS opsin	Rabbit, anti-SWS opsin, 1:500	Chemicon International Inc.
SWS opsin	Rabbit, JH455 anti-SWS opsin, 1:5000	28

Figure 1.

 

Time- and region-specific loss of cone photoreceptors in the retina of CNGA3-deficient mice. Photoreceptors were visualized in retinal wholemount preparations by labeling with PNA. (A) Representation of the four retinal quadrants: D, dorsal; N, nasal; T, temporal; and V, ventral part of the retina. (B, C) Similar cone density in young (PW4) and aged (PM17) wild-type retina. (D–F) Time course of cone photoreceptor loss in the CNGA3^−/− retina. Representative wholemounts from CNGA3-deficient mice at PW4 (D), PM7 (E), and PM17 (F). (G–I) Higher-magnification images of PNA-labeled PM17 wild-type (G) and CNGA3^−/− (H, I) retinal wholemounts. A significant number of PNA-positive cones was still present in the dorsal quadrant of the CNGA3^−/− retina (H). Almost no cones remained in the ventral quadrant (I). Scale bar: (B–F) 250 μm; 20 μm in (G–I).

Figure 2.

 

Induction of NSE in CNGA3^−/− cones. (A) Western blot analysis of a retinal membrane protein preparation (30 μg protein) from adult wild-type mice using a polyclonal anti-NSE antibody. Bars to the right: 37 (top) and 50 (bottom) kDa. (B–H) Confocal images of retinal slices double labeled for NSE (green) and PNA (red). Yellow: overlapping signals. (B) At PW3, NSE was expressed in all cone and rod inner segments of the wild-type retina. Only very faint staining was observed in the outer segments. (C, D) Upregulation of NSE-labeling in cone outer segments of a PW3 CNGA3^−/− retina. The NSE-positive cone outer segment labeled with an arrow in (C) is shown at higher magnification in (D). (E, F) Most cone outer segments in the dorsal part of a PW4 CNGA3^−/− retina showed increased NSE-labeling (E, arrows, yellow). In contrast, cone outer segments in the ventral part showed PNA staining but no increased NSE-labeling (F). (G) The remaining cones in the dorsal part of a PM17 CNGA3^−/− retina were highly labeled with the anti-NSE antibody (arrow). (H) A PM17 wild-type retina revealed no NSE-labeling in the outer segments. Scale bar, 20 μm.

Figure 3.

 

Expression of cone opsins in the CNGA3^−/− retina. (A–D) Costaining of mid-dorsal wholemount preparations from PW4 wild-type (A, B) and CNGA3-deficient retinas (C, D) with a monoclonal antibody against MWS opsin (COS1) and PNA. (A, C) Opsin signal (red); (B, D) overlay with PNA (green). (E, F) Costaining of a PW4 mid-ventral wholemount preparation with a monoclonal anti-SWS opsin antibody (OS2, red) and PNA (green). (E) Opsin signal. (F) Overlay. (G–K) Confocal images of retinal sections from PW3 wild-type (G) and CNGA3^−/− mice (H) stained with polyclonal anti-MWS antibody. (I) Same cone as in (H) costained with anti-MWS opsin (red) and PNA (green). (J) Costaining of a section from a PM17 CNGA3^−/− mouse with anti-MWS opsin (red) and PNA (green). The differential interference contrast (DIC) image is overlaid to visualize the retinal cell structure. (K) MWS opsin (red)-positive cones in the dorsal retina of a 22-month-old CNGA3^−/− mouse. (L) Top: Western blot with 30 μg retinal membranes of PW3 wild-type and CNGA3^−/− mice probed with polyclonal anti-MWS. Bottom: same blot probed with α-tubulin as a loading control. (M) Semiquantitative RT-PCR analysis of expression levels of MWS opsin in the retina of wild-type (□, n = 3) and CNGA3^−/− mice (▪, n = 3). At PM3 and PM12 the mRNA levels of MWS opsin were 94.8% ± 10.6% and 34.2% ± 6.9% of those in wild-type retina (n = 3, t-test **P < 0.005). OS, outer segment; IS, inner segment; ONL, outer nuclear layer. Scale bars: (A–F) 20 μm; (G–J) 5 μm; (K) 10 μm.

Figure 4.

 

Downregulation of proteins of the cone phototransduction cascade in the CNGA3^−/− retina. (A–H) Confocal scans of retinal sections costained with PNA (red) and a specific antibody (green). Yellow: staining overlay. (A, B) Colabeling with an antibody against CNGB3 and PNA. CNGB3 was specifically expressed in COSs of wild-type mice but was completely missing in the CNGA3-deficient mice. (C, D) Cone transducin (G_αt₂) was expressed in all PNA-labeled COSs of the P12 wild-type retina (C). At the same time point, very few cones expressed G_αt₂ protein in CNGA3^−/− cones (D). (E–H) Projections of confocal z-stack images showing that GCE and GCAP1 were present in rods and cones (PNA-labeled) of wild-type mice (E, G, respectively) but downregulated in PW3 CNGA3^−/− cones (F, H, arrows). (I, J) Semiquantitative RT-PCR analysis of expression levels of CNGB3 (I) and Gαt₂ (J) in the retina of PM3 wild-type (n = 3) and CNGA3^−/− mice (n = 3). mRNA levels of CNGB3 and G_αt₂ were 83.8% ± 8.4% of 91.8% ± 8.0% of those in wild-type retina, respectively (n = 3). Scale bar, 20 μm.

Figure 5.

 

Gliosis and apoptosis in the CNGA3^−/− retina. (A) In the wild-type retina, an anti-GFAP antibody (red) labeled Müller cell end feet and astrocytes of the neurofilament layer (NFL) (B) In the CNGA3-deficient retina GFAP-staining revealed extensive induction of intermediate fibers of retinal Müller cells. (C) Blebbing (arrows) of cone outer and inner segments in a CNGA3^−/− retina visualized with anti-SWS opsin staining. (D) Fragmentation of a cone nucleus (arrow) identified by costaining with anti-SWS cone opsin antibody (green) and Hoechst dye (blue). (E, F) TUNEL-labeling on PW4 retinal slices of CNGA3^−/− (E) and wild-type (F) mice. Arrows: TUNEL-positive photoreceptors. (G) Quantitative analysis of nuclear fragmentation in the outer nuclear layer of wild-type (n = 4 to 5) and CNGA3^−/− (n = 4–7) mice. (H–J) Activation of caspase 3 (green) (H, I) and release of cytochrome c (K) in PW4 CNGA3^−/− retinas, but not in age-matched control subjects (J, L). Shown are confocal scans overlaid with the differential interference contrast (DIC) images (H–L). *P < 0.05, ***P < 0.001 (ANOVA). Scale bars: (A, B) 50 μm; (C) 10 μm; (D, I) 5 μm; (E, F, H, J–L) 20 μm.

Figure 6.

 

Impaired migration of cone somata in the CNGA3^−/− retina. (A–I) Confocal images of retinal slices of wild-type and CNGA3^−/− mice labeled with polyclonal anti-opsin antibodies (green). At P9, COSs of the wild-type retina expressed high levels of SWS (A) and MWS (B) opsin. Almost all cone somata (arrows) were positioned in the upper half of the ONL. (C) At P12, most opsin-staining in the wild-type retina was found in the COS (shown: MWS opsin). In the P9 CNGA3^−/− retina, no opsin (SWS in D, MWS in E) was found in COSs and opsin-positive cone somata were distributed throughout the ONL. (F) At P12, many opsin-positive cone somata (shown: MWS opsin) were still displaced in the CNGA3^−/− retina and were found close to the synaptic layer (arrows). At this stage, some cones were correctly located in the outermost part of the ONL. At that point, more cone inner segments were opsin positive (overlay with PNA; red). (G) Displaced cone somata (arrows) identified by double-labeling with Hoechst nuclear dye (blue) and anti-SWS opsin (green). (H, I) Projections of confocal z-stack images of PW3 CNGA3^−/− retina labeled with anti-MWS or anti-SWS opsin (green) and either PNA (red) or glycogen phosphorylase (GlyPh; red). (H) The three classes of cone somata present in the PW3 CNGA3^−/− retina are labeled by arrows (displaced close to the INL), arrowheads (normal localization), and an asterisk (in the middle of the ONL). (I) A CNGA3^−/− cone photoreceptor containing a displaced cell soma (arrow), a long axon, and an inner segment. INL, inner nuclear layer; IS, inner segments; OS, outer segments; ONL, outer nuclear layer. Scale bars: (A–H) 20 μm; (I) 10 μm.

Figure 7.

 

Synaptic contacts in the CNGA3^−/− retina. (A–C) Electron microscopic images of a displaced cone somata in the OPL of a CNGA3^−/− retina at PW8. The soma in (A) fused with the pedicle and lay at the inner border of the OPL surrounded by several rod spherules (rs). (B) High-magnification image of (A) with focus on the pedicle base. Several postsynaptic elements and a presynaptic ribbon (arrow) are visible. (C) Another example of a displaced soma. Two presynaptic ribbons and several postsynaptic elements are visible. h: invaginating lateral horizontal cell dendrites at the cone pedicle base. (D, E) Calbindin staining of horizontal cells in adult CNGA3^−/− (D) and CNGA3^+/+ (E) retinas. Horizontal cells in the CNGA3^−/− retina show neuronal sprouting into the ONL (arrow). (F) Electron microscopic image of a normal-shaped cone pedicle (cp) found in the dorsal part of a PM6 CNGA3^−/− retina showing synaptic ribbons (arrows) and flat synaptic contacts (arrowheads). (G, H) Confocal images illustrating the colocalization of GluR5 (G, arrows) and PNA (H) signals in the dorsal part of a CNGA3^−/− retinal wholemount at PM7. h, horizontal cell dendrites; m, mitochondria; nu, cone nucleus. Scale bars: (A) 2 μm; (B, C) 0.5 μm; (D–E) 20 μm; (F) 1 μm; (G, H) 10 μm.

Supplementary Materials

Supplementary Table S1 - 23.3 KB 

Supplementary Figure S1 - 224 KB 

Specificity of anti-opsin antibodies. (A-D) Confocal images of wild-type retinal slices double-labeled with polyclonal anti-MWS or anti-SWS antibody (green) and the nuclear dye Hoechst (blue). Anti-MWS labels the majority of cone outer segments in the dorsal (A) but not in the ventral (B) mouse retina. By contrast, anti-SWS labels only a few cones in the dorsal retina (C) but the majority of cone outer segments in the ventral mouse retina (D). (E, F) Co-staining of a wild-type dorsal section with polyclonal anti-MWS opsin antibody (green), monoclonal COS1 antibody (red, specific for MWS cones) and PNA (blue). There is complete overlap between MWS and COS1 staining (yellow overlay in (E)). Triple staining shown in (F) reveals that all cones in this section are MWS-type cones. (G) Whole-mount preparation from the central retina stained with polyclonal anti-MWS antibody and the monoclonal OS2 antibody (specific for SWS-type cones). There is only sparse overlap of signals (yellow) confirming the specificity of the two antibodies. (H) Western blots with 30 �g retinal membranes from adult wild-type mice using polyclonal anti-MWS (left) or anti-SWS (right) detected a 42kDa and 46 kDa band, respectively. Scale bars: (A-D) 100 �m; (E) 20 �m; (G) 10 �m. 

Supplementary Figure S2 - 190 KB 

Preserved rod morphology in CNGA3-deficient mice. (A, B) Confocal images of retinal sections from the ventral part of PM17 wild-type (A) and CNGA3^-/- (B) retina labeled with anti-rhodopsin antibody (red) and PNA (green). (C-E) Normal expression of CNGA1 (C), CNGB1 (D) and GCF (E) in the retina of a 22-month-old CNGA3^-/- mouse. The confocal scans in (C, D) were overlaid with the respective differential interference contrast images to demonstrate the retinal morphology. Scale bar: 20 �m. 

The authors thank David L. Garbers, Bernd Hamprecht, Robert S. Molday, Jeremy Nathans, Krzysztof Palczewski, and Agoston Szel for the gift of antibodies.