Animals were housed under standard white cyclic light (200 lux, 12 h dark-light periods), had free access to food and water, and were used irrespective of gender. All procedures involving animals were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The study and the protocol were approved by the competent local authority, Regierungspräsidium Tübingen, basing on the assessment by the appointed regional ethics board.

Cloning and mutagenesis were performed by standard techniques. All sequence manipulations were confirmed by sequencing. Single-strand genome AAV2/5 Y719F (Michalakis et al., 2010) or AAV2/8 Y733F (Koch et al., 2012) pseudotyped AAV2 vectors expressing mouse Cnga3 cDNA (Michalakis et al., 2010) under control of a 0.5 kb mouse blue (S) opsin (Michalakis et al., 2010) were produced as described recently (Becirovic et al., 2016). Physical titers (in vector genomes/μl) were determined by quantitative PCR on a LightCycler 480 (Roche Applied Science, Mannheim, Germany) using KAPPA SYBR FAST kit (Peqlab, Erlangen, Germany) and the following primer set: WPREF: 5′-AGTTGTGGCCCGTTGTCAGG-3′ and WPRER: 5′-AGTTCCGCCGTGGCAATAGG-3′.

A detailed protocol for optimized subretinal injection technique was described previously (Mühlfriedel et al., 2013). In brief, mice were anesthetized by subcutaneous injection of ketamine (66.7 mg/kg) and xylazine (11.7 mg/kg), and their pupils were dilated with tropicamide eye drops (Mydriaticum Stulln; Pharma Stulln, Germany). Young mice (postnatal 14 days) received a single subretinal injection of 1 μl of AAV particles (~5 × 10⁹ total vector genomes) at the ventral part, whereas in aged animals (from postnatal month one), due to the ventral-dorsal gradient of cone degeneration, the viral particles were applied at the dorsal part. The injection was performed free hand under a surgical microscope (Carl Zeiss, Germany) using a NanoFil injection kit (WPI, Germany) equipped with a 34 gauge beveled needle. In the experimental design, one eye was treated whereas the untreated eye served as the untouched control. Special care was taken to avoid damage of the lens. After injection, in vivo imaging was performed (described in next section). Animals were placed on a warming blanket and eye ointment (Vidisic®, Dr. Mann Pharma, Berlin, Germany) was applied to the corneas to prevent drying while the animals recovered from anesthesia. In addition, an antibiotic eye ointment was applied two times daily for 48 h (Dexamytrex®, Bausch & Lomb, Berlin, Germany).

The quality of the injection and retinal morphology were examined immediately after subretinal injection using confocal scanning laser ophthalmoscopy (cSLO) and spectral domain optical coherence tomography (SD-OCT). The cSLO procedure was reported previously (Seeliger et al., 2005). For consecutive examinations at later time points, mice were anesthetized and pupils were dilated as described in section “subretinal AAV injections.” cSLO images were obtained with a Heidelberg Retina Angiograph (HRA I) at 10 and 20° field of view. Two different imaging modalities were used to visualize the retina: (1) 514 nm laser excitation, red free imaging mode, (2) and 830 nm laser excitation, infrared imaging mode. SD-OCTs were performed using a Spectralis TM HRA + OCT device from Heidelberg Engineering (Fischer et al., 2009; Huber et al., 2009). Imaging data were analyzed using the software package Eye Explorer version 3.2.1.0 from Heidelberg Engineering. Resulting data were exported as 24 bit color image files and processed in Adobe Photoshop CS3 (Adobe Systems, San Jose, CA, USA) (Garcia Garrido et al., 2015). If the procedure was not successful (severe damage like a full retinal detachment, lens opacity due to injury), the mice were excluded from further analysis.

Electroretinograms (ERGs) were recorded according to the previously described procedures (Tanimoto et al., 2009). The ERG system consisted of a xenon light source, a Ganzfeld bowl, a signal amplification system (band-pass 0.3–300 Hz), a PC-based control and recording unit, and a monitor (Multiliner Vision; VIASYS Healthcare GmbH, Hoechberg, Germany). Animals were dark adapted overnight, anesthetized and pupils were dilated as described in section “subretinal AAV injections.” Single flash intensity series data were obtained under photopic (light-adapted 10 min at 30 cd/m²) conditions. Flash stimuli ranged from −2.0 to 1.5 log cd^*s/m² divided into ten steps. Responses to trains of brief flashes (flicker) of fixed luminance (1 log cd^*s/m²) but varying frequency (0.5, 1, 2, 3, 5, 7, and 10 Hz) were recorded at the cornea under light-adapted conditions with a background illumination of 30 cd/m². Flicker responses were averaged either 20 times (for 0.5–3 Hz) or 30 times (for 5–10 Hz). Single flash intensity and flicker frequency series data were obtained from treated animals (one treated eye, one untreated eye) at indicated time points after treatment.

Vertical cryosections (10 μm) of the mouse retina were prepared for immunohistochemical staining as described previously (Michalakis et al., 2005). For an appropriate localization of the treated and untreated retinal part, eyes were marked temporally before harvesting. Antibodies used were rabbit anti-mouse CNGA3 polyclonal antibody (Michalakis et al., 2005, applied at 1:3,000 dilution) and rabbit anti-mouse S opsin polyclonal antibody (Millipore, applied at 1:300 dilution). For secondary detection donkey Cy3 anti-rabbit IgG was used (Jackson ImmunoResearch). Cone photoreceptors were stained with fluorescein isothiocyanate conjugated peanut agglutinin (FITC-PNA; 1:100, Sigma-Aldrich). Confocal images were collected on a LSM 510 (Carl Zeiss, Oberkochen, Germany) microscope.

Animals were sacrificed, eyes were enucleated and immediately fixed in buffered 4% paraformaldehyde for 10 min. For an appropriate localization of the treated and untreated retinal part, eyes were marked temporally before harvesting. For immunostaining, retinal whole mounts were washed with 0.1M PB (3 × 10 min) and co-stained with FITC-PNA and rabbit anti-mouse CNGA3 polyclonal antibody over night. For secondary detection, a donkey Cy3 anti-rabbit IgG was used. After washing with 0.1M PB the retinas were flat mounted on objective slides, orientated and embedded in mowiol mounting medium. Pictures were taken using a fluorescence microscope connected to a video camera (Zeiss ApoTome, AxiocamMRm, Germany).

In this study, two different types of Student's t-tests were used for statistical analysis to determine if two sets of data are significantly different from each other. The independent samples t-test (unpaired independent samples) was used when two separate sets of independent and identically distributed samples were obtained, one from each of the two populations being compared were non-overlapping (e.g., when comparing the effects of the two capsid variants). The paired t-test (paired dependent samples) was used to when one group was tested on two time consecutive points (e.g., repeated measurements at 1 and 12 months PI).

Each animal received a single subretinal injection of AAV vector in one eye, while the other eye remained untouched and served as control. To minimize any possible damage due to the surgical procedure a total volume of 1 μl AAV vector suspension was injected. The technical outcome was monitored via in vivo imaging of the mouse eye using confocal scanning-laser ophthalmoscopy (cSLO) and spectral domain optical coherence tomography (SD-OCT) within 10 min after subretinal delivery of the AAV vector (Figure 1). Following the injection of 1 μl vector suspension to the dorsal retina, a defined local subretinal bleb became evident in cSLO fundus imaging (Figures 1A,B). A higher magnification (10° field of view) revealed the trajectory of retinal blood vessels at the border of the transitory subretinal bleb (Figure 1C). Subsequent assessment of horizontal and vertical SD-OCT scans confirmed the detachment of the retina from the RPE at the subretinal bleb area (Figure 1D). This subretinal detachment resolved spontaneously within a few days, and after 4 weeks only subtle morphological abnormalities around the injection site and within the former detached area remained detectable (Figures 1E–H).

To assess the extent of transduction obtained within the area of the subretinal injection, we harvested the retina from mice treated with a dose of ~5 × 10⁹ total vector genomes of AAV2/8 (Y733F)-S opsin-Cnga3, expressing murine Cnga3 under control of the mouse S opsin promoter. After labeling with a specific anti-CNGA3 antibody and the cone photoreceptor marker peanut agglutinin (PNA), the retina was flat-mounted and imaged under an epifluorescence microscope (Figures 2A–D). As described earlier, the natural course of cone photoreceptor degeneration in this model follows a ventral to dorsal gradient (Michalakis et al., 2005). Consequently, at 4 months post-injection (PI), only a few remaining cones were labeled by the cone marker PNA in the ventral half (V) of the flat-mounted Cnga3 KO retina (Figure 2A). Injections in this set of experiments were placed in the ventral part, however, transduction of the temporal or nasal parts of the retina due to vector diffusion could not be excluded. Immunolabeling confirmed the presence of mouse Cnga3 protein in the treated part of the retina surrounding the injection site (Figure 2B). The area of cone-specific expression of CNGA3 (red) following the single subretinal injection spanned about 1/3 of the entire retina. A closer inspection of the overlay of the Cnga3 expression pattern with the PNA staining revealed a substantial number of PNA-positive cones in the ventral half of the treated Cnga3 KO retina (Figures 2C,D). As this area normally lacks cones at this stage of degeneration (Michalakis et al., 2005), one can conclude that the extended cone survival is related to the treatment.

To evaluate potential differences in the therapeutic capacity of two different modified pseudotyped AAV2 vectors expressing mouse Cnga3 cDNA (AAV2/5 Y719F and AAV2/8 Y733F), we compared the outcome in two groups of respectively treated Cnga3 KO mice for a period of up to 3 months PI. Retinal function was assessed via Ganzfeld ERG to investigate the contribution of the rod and the cone system (Figure 3). Figure 3A shows a comparison of wild-type (wt) mouse data to the ERGs of the AAV2/8 (Y733F)-treated and untreated eyes of a Cnga3 KO animal. Under scotopic (dark-adapted) conditions, the specific rod system response (Figure 3A, top) revealed no difference between the treated eye (TE, red traces), untreated eye (UE, black traces), and the wild-type (wt) eye (gray traces). This finding indicates that the injection procedure had no measurable deleterious effects on retinal functionality, and corroborates the published data for AAV2/5 (Y719F) (Michalakis et al., 2010). Use of a more intense flash under scotopic conditions [ISCEV standard flash, Marmor et al., 2004], which potentially evokes mixed rod/cone system responses, revealed a higher b-wave amplitude in the TE relative to the UE, indicating a contribution of the cone system (Figure 3A, lower part of top graph). Cone function may usually best be observed under photopic, light adapted conditions, as in the bottom graph of Figure 3A. The respective photopic single flash recording clearly shows the treatment effect on cones in the TE, while there is merely a zero baseline in the UE. Again, these results in AAV2/8 (Y733F) corroborate the published data for AAV2/5 (Y719F) (Michalakis et al., 2010), including the fact that the approximate size of the treated area of 1/3 of the retina corresponds to an ERG amplitude of 1/3 to 1/4 of wt.

A particularly valuable biomarker for cone function is the photopic flicker ERG. For the direct comparison between AAV2/5 (Y719F) and AAV2/8 (Y733F), we chose a flash frequency of 7 Hz (Figure 3B). The presence of three consecutive flash responses in a trace enhances the visibility of differences and the detection of potential rescue effects. In general, Figure 3B again clearly shows the benefit of the intervention on cone functionality for both serotypes, and for both observation intervals at 1M and 3M post-injection (PI). Due to the complete lack of both cone (Cnga3 KO) and rod responses (which were desensitized with background light), the UEs are unable to respond, while there are distinct responses in the TEs (Figure 3B). We found no visible differences in cone system functionality between the AAV2/5 (Y719F) and the AAV2/8 (Y733F) groups, neither in the waveforms (representative traces in Figure 3B) nor in the group amplitude data (box plots) summarizing the results at 1M and 3M PI for each serotype (Figure 3C). The traces of the UEs did not contain any discernible waveforms, so that no amplitudes were measurable, as marked by “n.d.” for “non-detectable” in Figure 3C. A statistical work-up of the data summarized in Figure 3C revealed

As (b) and (c) involve follow-up data from the same individual animals, a paired t-test was applicable, while an unpaired t-test was used in (a) as the data originate from different mice. These findings suggest that a robust rescue of the Cnga3 KO phenotype was achieved over a 3 months period with both AAV vector serotypes, AAV2/5 (Y719F) and AAV2/8 (Y733F).

A key outcome of any gene supplementation therapy is a sustained, optimally life-long rescue. Published data (Michalakis et al., 2010) and the results shown above document a beneficial effect of the treatment up to 3 months PI. Here, we investigated the performance of the treatment over a period of 12 months PI, as the average lifespan of mice does not permit longer periods without unduly increasing the risk of systematic errors. In these long-term experiments, the AAV2/5 (Y719F)-S opsin-Cnga3 vector was used, and 1 μl subretinally injected into the ventral retina of 2-week old Cnga3 KO mice. The functional assessment of cone photoreceptor function was done at 1 M and 12 M PI (Figure 4A). Figures 4B,C summarizes the Ganzfeld ERG recordings with a focus on the biomarker used in previous sections (7 Hz photopic flicker ERG). For comparison to Figure 3A, the results of the photopic single flash ERG are included in Figure 4C. As illustrated in Figures 4B,D, cone-mediated function persisted at a similar level over the entire observation period of 12 months. Consequently, a statistical analysis as described above based on the data summarized in Figure 4D indicated no significant differences in treatment effects (p = 0.5). This result is particularly encouraging as the treated area covered parts of the ventral retina where, without therapy, cones are completely lost after 4 months (Michalakis et al., 2005), indicating that these targeted cones were saved.

Upon completion of the ERG measurements, the retinas were harvested for immunohistological analyses. We used the cone marker PNA and a CNGA3-specific antibody on vertical cryo-sections through the retina of treated Cnga3 KO mice. A representative confocal overview image from a retinal section including a treated (right) and untreated (left) area is shown in Figure 5A. AAV vector-encoded CNGA3 protein is still clearly detectable at 12 M PI in cones within the treated, but not in the untreated part of the Cnga3 KO retina (Figure 5A). These CNGA3-positive cones were also positive for the cone marker PNA (Figure 5B) whereas only a few surviving CNGA3-negative cones were found (see left, untreated part in Figure 5B). Higher magnification images show that most of the CNGA3 protein localized to cone outer segments, but some protein was also observed in cone cell bodies and synapses (Figures 5C,D). S opsin-positive cones are not found in untreated Cnga3 KO retina older than 4 months of age (Michalakis et al., 2005). However, S opsin-positive cones still survived at 12 M PI in the treated ventral part of Cnga3 KO retina (Figure 5E), suggesting that the treatment may halt degeneration of the fast degenerating S opsin-positive cones.

Finally, we treated 3-month-old (PM3) Cnga3 KO mice in the same way as before the 2-week-old animals, using the AAV2/8 (Y733F)-S opsin-Cnga3 vector (Figure 6A). Since the peak of cone cell death is at about postnatal day 35 (Arango-Gonzalez et al., 2014) and cone degeneration is particular present in the ventral Cnga3 KO retina, we aimed at delivering the vector entirely to the dorsal retina to target the slower degenerating dorsal cones. We found that although presumably a substantial number of cones has already been lost at time of treatment, it was still possible to endow the remaining cones with the ability to generate specific light responses (Figure 6B). The treatment effect persisted up to 12 M PI (Figure 6B). The functional benefit was confirmed by single flash ERG recordings (Figure 6C). In summary, the preclinical results in Cnga3 KO mice show that even at advanced stages of degeneration up to at least PM3, surviving cones are still treatable.

The handling Editor and author SM declare that they are co-hosting the Research Topic “Retinal Gene Therapy: From Basic Research to Translational Studies” but have no other collaborations. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.