Title: Targeting channelrhodopsin-2 to ON-bipolar cells with vitreally administered AAV restores ON and OFF visual responses in blind mice

Authors: Emilie Macé1,2,3*, Romain Caplette1,2,3*, Olivier Marre1,2,3*, Abhishek Sengupta1,2,3, Antoine Chaffiol1,2,3, Peggy Barbe1,2,3, Mélissa Desrosiers1,2,3, Ernst Bamberg6, Jose-Alain Sahel1,2,3,4,5, Serge Picaud1,2,3, Jens Duebel1,2,3# and Deniz Dalkara1,2,3#

1

INSERM, U968, Paris, F-75012, France.

2

Sorbonne Universités, UPMC Univ Paris 06, UMR_S 968, Institut de la Vision, Paris, F-75012, France.

3

CNRS, UMR_7210, Paris, F-75012, France.

4

Centre Hospitalier National d’Ophtalmologie des Quinze-Vingts, INSERM-DHOS CIC 503, Paris, F-

75012, France. 5

Fondation Ophtalmologique Adolphe de Rothschild, Paris, France.

6

Max Planck Institute of Biophysics, Department of Biophysical Chemistry, Frankfurt am Main, Germany.

*Equal Contribution #

Correspondence should be adressed to:

Deniz Dalkara ([email protected], +33153462532) Vision Institute, 17 rue Moreau, Paris, F-75012, France Jens Duebel ([email protected], +33153462595) Vision Institute, 17 rue Moreau, Paris, F-75012, France

1

Abstract Most inherited retinal dystrophies display progressive photoreceptor cell degeneration leading to severe visual impairment. Optogenetic reactivation of retinal neurons mediated by adeno-associated virus (AAV) gene therapy has the potential to restore vision regardless of patient specific mutations. The challenge for clinical translatability is to restore a vision as close to natural vision as possible, while using a surgically safe delivery route for the fragile degenerated retina. To preserve the visual processing of the inner retina, we targeted ON bipolar cells, which are still present at late stages of disease. For safe gene delivery, we used a recently engineered AAV variant that can transduce the bipolar cells after injection into the eye’s easily accessible vitreous humor. We show that AAV encoding channelrhodopsin under the ON bipolar cell specific promoter mediates long-term gene delivery restricted to ON-bipolar cells after intravitreal administration. Channelrhodopsin expression in ON bipolar cells leads to restoration of ON and OFF responses at the retinal and cortical levels. Moreover, light induced locomotory behavior is restored in treated blind mice. Our results support the clinical relevance of a minimally-invasive AAV-mediated optogenetic therapy for visual restoration.

Introduction The remarkable success in clinical trials for the childhood-onset blindness, Leber’s Congenital Amaurosis (LCA) established the proof-of-concept for AAV-mediated retinal gene therapy1–4. Gene replacement approach is effective for treating diseases resulting from recessive null mutations but remains difficult to apply to dominantly inherited retinal dystrophies, affecting the majority of visually impaired patients. Furthermore, inherited retinal degenerations (RD) display wide variation in their mode of inheritance,

underlying

genetic

defects,

age

of

onset,

and

phenotypic

severity

(https://sph.uth.edu/retnet/disease.htm). The genetic origin of the disease remains unknown in half of the patients. These present enormous obstacles for the development of broadly applicable gene therapy strategies for RD. As one such example, over 80 gene loci are involved in retinal diseases that result in photoreceptor cell death, with the most common subtype being retinitis pigmentosa (RP). Given this constraint, mutation independent gene therapeutic approaches have been widely developed over the past twenty years5–7. One such approach is optogenetics. Optogenetics aims at restoring vision in blind

2

patients by expressing microbial opsins8–10, endogenous opsins11 or engineered photosensitive ion channels12 to re-activate residual retinal neurons in late stage photoreceptor diseases. The non-selective expression of optogenetic light switches in inner retinal cells, does not restore the diversity of retinal output responses as they activate ON and OFF cells indistinctly11–13. On the other hand, it has been shown that expressing halorhodopsin in non-functional but surviving ―dormant‖ cones preserves the processing of visual inputs by all layers of the retina10. Clinical data shows that dormant cones appear in a restricted area of the macula, but it remains unclear what percentage of the patient population displays this phenotype. Histopathologic studies of post-mortem retinas from patients with retinitis pigmentosa (RP) show that 78% to 88% of the inner nuclear layer cells are preserved in patients with severe and moderate RP, respectively14. An attractive cell target for optogenetic therapies is thus the ON-bipolar cell. A pioneering study used electroporation to insert channelrhodopsin cDNA under the control of the ON bipolar cell promoter into bipolar cells of the rd1 mouse retina. This lead to the recovery of visually evoked potentials and visually guided behaviors after the intervention9. Electroporation however, is not a viable delivery method for clinical application as it leads to transient and low expression levels (~7% of the targeted ON bipolar cells). To use a microbial opsin in vision restoration, the transgene expression must be stable and robust in specific cellular targets, and this can be best achieved using AAVs. Natural AAVs have been shown to transduce effectively retinal ganglion cells following intravitreal injection15,16, and photoreceptors using subretinal injection in normal retinas17. However, bipolar cells were more difficult to target and they require engineered vectors. After degeneration of photoreceptors in the rd1 retina, a tyrosine capsid-mutated serotype, AAV8-Y733F, was effective at transducing bipolar cells via subretinal administration18. In this study, hChR2–GFP was expressed in the ON-bipolar cells and it was shown that continuous channelrhodopsin expression after AAV delivery is safe from immunological standpoint. However, recently published results from clinical trials show that subretinal injections are associated with procedural risks in the foveal region19. The progression of the disease likely affects retinal structure making it prone to damage by surgical detachment. The risk of compromising residual central vision in late stage RP patients may represent a roadblock for this therapeutic option. Furthermore, subretinal injections only treat a fraction of the retina. To overcome these hurdles, new AAV variants with

3

the ability to deliver genes deep into the retina via intravitreal route have been engineered20–22. We, for instance, recently developed a new AAV vector, 7m8, a genetic variant of AAV2, which has a peptide on its heparin-binding site21. This variant was isolated from a large library of AAV mutants injected intraocularly into transgenic mice expressing a rhodopsin-GFP fusion protein in rod photoreceptors23. Rods were subsequently isolated using flow cytometry, and AAV variants were recovered from these cells. This process enabled us to iteratively enrich for AAV variants capable of reaching the outer retina from the vitreous. All of the variants isolated through this screen had 7mer insertions in their heparin-binding site pointing to the relevance of heparin binding. The increased retinal transduction efficiency of 7m8 compared to its parental serotype AAV2 may arise in part from this variant’s reduced heparin affinity24, which may both decrease capsid sequestration in the ILM and enable enhanced penetration through retinal layers. In this study, we combined this potent AAV capsid variant designed for gene delivery into deep layers of the retina21 with the previously described 200–base pair enhancer sequence of the mouse Grm6 gene, fused to the SV40 eukaryotic promoter9,18. Large numbers of ON-bipolar cells were targeted across the retina using this capsid-promoter combination. Unlike preceding studies, treated rd1 retinas showed both ON and OFF responses indicating that these pathways are reactivated jointly by our approach. Furthermore, we show that these ON and OFF responses are transmitted to the visual cortex and lead to light induced locomotory behavior. Restoring vision at the bipolar cell level with a surgically less complex intravitreal injection of engineered AAV-channelrhodopsin is thus a promising strategy for RP patients with no remaining cones and fragile retinal architecture.

Results Gene delivery to ON bipolar cells For all experiments we used a humanized version of channelrhodopsin-2 with H134R mutation (ChR2/H134R). The H134R mutation provides a reduction in desensitization and increased light sensitivity25. It was fused to GFP to facilitate cellular localization. Gene expression however was restricted to the ON bipolar cells by the use of a 200–base pair enhancer sequence of the mouse Grm6 gene, which encodes the ON bipolar cell–specific metabotropic glutamate receptor, mGluR626. We used 7m8, a

4

genetic variant of AAV2, to deliver ChR2 across the retinal layers. ChR2 was delivered to the retinas of C57BL/6J wild-type and C3HeN rd1 mice. Rd1 mice were used for this study as their photoreceptors rapidly degenerate and their electroretinogram is undetectable by 8 weeks of age27,28. Four to 8 weeks old rd1 mice were injected intraocularly. Four to six weeks after intravitreal delivery of AAV2-7m8 vector encoding hChR2/H134R-GFP lead to strong pan-retinal expression as seen by in vivo fundus imaging (Figure 1a). Retinas were then harvested and directly visualized using two-photon microscopy to characterize the cell types expressing hChR2/H134R-GFP (Figure 1b). Observation of retinal slices revealed strongly labeled axon terminals in the inner half of the inner plexiform layer (ON-sublamina), consistent with localization of axons of ON bipolar cells29. These cells were imaged in a region of the retina where bipolar cells were sparsely labeled with GFP fluorescence to ensure that their axon terminals are not masked by processes of neighboring bipolar cells. Based on their branching pattern, dendritic morphology and stratification of their axon terminals, we were able to identify both rod and cone bipolar cell types30, as shown in Figure 1 b: The narrowly branching bipolar cell with compact dendritic field and short axon terminals was identified as Type 7

30

. Another ON Cone bipolar cell type with sparse, long

meandering dendrites was identified as Type 9. The rod bipolar cell type was identified by its typical bulbous shaped axon-terminal. Some eyes from the same group were enucleated, fixed and embedded in OCT for histology and confocal microscopy 120 to 140 days post injection. 15 µm thick retinal cryosections were incubated with anti-PKC alpha antibodies (in red) and GFP signal was amplified with GFP antibodies (Figure 1c). There was strong labeling of the ON strata of the inner plexiform layer across the retina (Supplemental Figure 1). GFP expression was localized to the plasma membrane, where it strongly labeled cell somas, dendritic arbors, axons and axon terminals. Another series of animals were sacrificed 204 to 254 days post injection. These eyes were used for cross sections shown in Supplemental figure 1. In some areas, we observed sparse labeling of amacrine cells. No off-target expression was observed in OFF bipolar cells (Supplemental Figure 1). To examine the distribution of GFP across treated rd1 retinas, low magnification images were acquired after anti-GFP staining (Figure 1d). Retinal flat-mounts co-stained with ON bipolar cell antibodies (PKC-alpha and Go-alpha) the percentage of ON bipolar cell transduction was studied using confocal microscopy. To evaluate the percentage of ON bipolar cells labeled across the retina, low

5

magnification confocal images were acquired at 8 different locations, at two different eccentricities from the optic nerve head. Confocal image stacks were acquired across the bipolar cell layer and cell bodies that stain positive for the ON bipolar cell markers and GFP were counted (Supplemental Table 1). We found that six to ten thousand ON bipolar cells were labeled per square millimeter across the retinas, which represents 52 to 74% of total ON bipolar cells counted (Figure 1e). Expression was stable for at least 12 months post-injection in both wild-type and rd1 retinas (Supplemental Figure 1). Immunoreactivity against the rod bipolar cell marker PKC alpha was observed in about half of the GFP-expressing cells in the AAV injected retinas (Figure 1 d). Our data show that both rod and cone ON bipolar cells expressed the hChR2/H134R-GFP fusion protein.

ChR2-evoked spike activity in retinal ganglion cells To demonstrate that selective ON-bipolar cell targeting of hChR2/H134R-GFP can restore visual function in blind rd1 retinas (age > 18 weeks), we recorded spiking activity from retinal ganglion cells using a multi-electrode array (MEA). Mice were injected 4-8 weeks after birth and MEA recordings were performed at 18 to 36 weeks of age. Light evoked spiking activity was observed when stimulating treated rd1 retinas with full field flashes (Figure 2b), whereas control rd1 retinas did not show any increase in spiking in response to light (data not shown). We measured the light spectrum of the responses at 1017 photons.cm-2.s-1 using different wavelengths. The spectrum recorded matches the excitation spectrum of channelrhodopsin231,32. Firing frequency was highest in response to 480 nm light, which corresponds to the excitation peak of ChR233 (Figure 2a). Firing rate frequency was intensity dependent, with a threshold of >1014 photons.cm-2.s-1 (Figure 2d). The ratio of responding cells increased linearly with increasing light intensities (Figure 2c). Importantly, treated rd1 retinas showed both ON and OFF responses to light stimuli. To investigate if these responses were produced by residual cone photoreceptors, we blocked the cone to ON bipolar synapse with a metabotropic glutamate receptor agonist. ON responses remained after application of metabotropic glutamate receptor agonist L-(+)-2-Amino-4-phosphonobutyric acid (LAP4) at 50µM in ChR2 treated rd1 retinas (Figure 2c-d). Conversely, LAP-4 blocked all ON responses in wild-type retinas by blocking the transmission between photoreceptors and ON bipolar cells (data not shown). These results demonstrate that all responses in treated rd1 retinas originated from ChR2/H134R

6

expressing bipolar cells, bypassing the endogenous phototransduction step. Furthermore, OFF responses (92%) were blocked by the glycine receptor antagonist, strychnine, at 10µM (Figure 2b and d). This observation is consistent with an indirect activation of the OFF pathway, likely through rod-bipolar cells to AII amacrine cells, in the treated rd1 retina.

Optogenetic photo-responses in the visual cortex To assess whether the ChR2-mediated retinal photo-responses were transmitted to higher visual centers of the brain, visually evoked responses were recorded from the visual cortex of AAV2-7m8ChR2/H134R treated rd1 mice and compared with responses obtained from untreated rd1 and wild-type mice. The treated eye (contralateral to the recording hemisphere) was stimulated with 200 ms pulses of blue light (light intensity 1.7 1017 photons cm–2 s–1) repeated 200 times at 1 Hz. Both local field potentials (LFP) and multiunit spiking activity (MUA) were recorded. As the amplitude and shape of a visually evoked LFP response, also called visually-evoked potential (VEP), depends on cortical depth, we used linear multisite silicon microprobes (sixteen electrodes at 50 μm intervals) to perform VEP recordings. For each acquisition, after averaging over the 200 trials, the electrode showing the VEP with maximal peak amplitude was selected for quantification. No VEPs were visible (flat traces) on recordings from untreated rd1 mice. For comparison, we measured the maximal amplitude of the LFP trace due to the noise level in these recordings from untreated rd1 mice. The maximal VEP amplitude was significantly larger in treated animals (54 +- 6 µV, N=5) than maximal LFP amplitudes recorded in untreated rd1 mice (17 +- 2 µV, N=5) (P = 2.3 10-4, P < 0.001, unpaired one-tailed t-test) (Figure 3a). The amplitude of the VEP in treated rd1 mice was consistently lower than wild-type VEPs (464 +- 37 µV, N=7) (P= 1.8 10-6, P < 0.001, unpaired one-tailed t-test). Interestingly, the shape of the VEPs were similar between wild type and treated mice, with clearly visible ON and OFF responses (Figure 3). The age of the 5 treated rd1 animals analyzed were 245, 323, 304, 337 and 346 days at the time or recordings and all animals in this group were injected 8 weeks after birth. VEP responses are dominated by synaptic currents because the synaptic currents are slow events that can overlap locally to build a macroscopic current response, and a large fluctuation in the field potential34. The presence of VEPs does not guarantee that visual stimulation induces spiking responses in

7

neurons in the vicinity of the electrode. Therefore VEPs and spikes in the primary visual cortex convey independent information35. We thus studied, for the first time, the multiunit activity (MUA) recorded by the silicon probe, extracted from the same data set. Light stimuli produced strong ON and OFF spike discharges in treated rd1 mice similar to those observed in sighted animals (Figure 3 b). No modulation of spike discharges was observed in untreated rd1 mice. We calculated the latency of both ON and OFF responses in the two groups after building the peri-stimulus time histogram (Figure 3 c). The latency of the ON response was consistently shorter in treated rd1 mice (20.3 +- 0.6 ms) than in wild-type animals (50.9 +- 3.3 ms) (Figure 3d). On the contrary, OFF responses exhibit similar latencies in both treated (55.2 +1.5 ms) and wild-type mice (53.0 +- 4.5 ms). Altogether our data demonstrate that cortical neurons generate light-induced spike discharges at ON and OFF light onsets in blind mice whose retinas have been re-activated in a circuit-specific manner.

Light Induced Locomotory Behavior We tested whether a light induced locomotory behavior was visible in rd1 blind mice lacking photoreceptors treated with bilateral injection of AAV2-7m8-ChR2/H134R. The mouse was placed into a cylindrical transparent tube with a bright blue (470 nm) LED light source at one end (Figure 4a). The LED light source was controlled by a function generator generating light flashes at 2 Hz. Initially, the light source was kept off and the mice were allowed to acclimate in the chamber for 2 minutes. Mice were filmed by an infrared camera during the whole experiment. The LED light source was switched on (at an intensity of 3000 lux) when the mice approached the light source, facing the LED. We observed that wild type mice quickly turned their whole body away from the light (n= 8, mean latency to first turn = 4.62± 1.25 s) once the light was switched on. Untreated rd1 mice took significantly longer than the wild-type mice to perform a first body turn after the light was switched on (n= 15, mean latency to first turn = 15.8 ± 3.04 s, significance wt vs rd1: p