Gene therapy induces synaptic ribbon maturation, synaptogenesis and vision recovery in an adult dog model of retinal degeneration

Abstract

The extent to which regeneration is possible in adult mammalian synapses remains an intractable question in neuroscience. Calcium interactions at the first retinal synapse are necessary for normal retinal development and vision. Calcium binding protein 4 (CaBP4) modulates these interactions, and mutations in CaBP4 or the voltage-gated calcium channel lead to similar forms of visual impairment. We identified a spontaneous mutation in CaBP4 in dogs that results in synaptic loss of function, immaturity of synaptic ribbons, and disorganization and thinning of the outer plexiform layer (OPL). Adeno-associated virus (AAV)-mediated gene augmentation therapy restored synaptic function and vision and led to synaptic ribbon maturation and elongation. Therapy also resulted in retinal layer preservation and re-organization, including expansion of the previously thin adult OPL. We show that the restoration of calcium regulation is, therefore, requisite for retinal plasticity and remodeling. This first naturally occurring large-animal model of mutant CaBP4 recapitulates components of the human disease and illustrates the potency of gene therapy in reversing blindness caused by the loss of CaBP4, paving the way for a cure. Structural and molecular changes following gene therapy demonstrate the phenomenal plasticity of the OPL and its synaptic machinery, highlighting the potential of neuroplasticity in the mammalian central nervous system.

Graphical abstract

Beckwith-Cohen and colleagues report on successful gene augmentation therapy in a naturally occurring large-animal dog model of cone-rod synaptic disorder due to a mutation in CaBP4. Affected dogs exhibit synaptic immaturity and disorganization, which are rescued even in later stages of the disease, supporting significant retinal plasticity.

Introduction

Inherited retinal degenerations (IRDs) are an important cause of blindness,¹ and over 338–400 genes underlying IRDs have been identified in people (RetNet, the Retinal Information Network, https://RetNet.org/last searched 10/2025, Genomics England PanelApp v.8.28, Retinal Disorders).² Therapies to prevent or reverse vision loss from IRDs are of major biomedical and socioeconomic importance.^3,4 Major barriers to treatment are related to the inability of mature differentiated neurons to regenerate and the abnormal synaptic connections that they form in the face of disease. While studies into cell transplantations are ongoing, a knowledge gap exists: to what extent can novel synaptic connections be formed between neurons that had not previously communicated?

Gene augmentation therapy is used to treat IRDs, and while there are several ongoing clinical trials, so far only one has resulted in FDA approval.^4,5,6,7,8 Despite the promise of gene augmentation therapy, our understanding of retinal remodeling and plasticity following therapy is lagging. The possible outcomes of gene augmentation therapy are variable. Therapy may halt or slow disease progression or even reverse it by replacing a missing protein.^6,9 Other IRDs are associated with abnormalities that are proving to be more complicated to treat.^4,10 Understanding which aspects of degenerative retinal remodeling are reversible or augmentable is pertinent to most IRDs, as well as to other neurodegenerative diseases. Some recent studies have explored this avenue with respect to photoreceptor plasticity and retinal ganglion cell plasticity, but few have focused on the outer plexiform layer (OPL).^11,12,13 Lessons learned from research in the OPL can serve as a foundation for research in other sensorineural tissues where synaptic ribbons and voltage-gated calcium channels play pivotal roles in cell signaling.

Transmission of light-induced photoreceptor responses is dependent on calcium signaling and subsequent gradation of glutamate release at the first retinal synapse. Calcium binding proteins (CaBPs) are a subfamily of calmodulin-like proteins, which modulate calcium interactions, playing crucial roles in sensorineural function.¹⁴ In people, mutations in CaBP4 (calcium binding protein 4, OMIM: 608965) result in visual deficits from childhood; the clinical phenotype is described as an incomplete congenital stationary night blindness or as Leber congenital amaurosis, but perhaps the most appropriate descriptor is congenital cone-rod synaptic disorder.^{15,16,17,18,19,20} Similar phenotypes were described for other genetic mutations affecting the same synapse, including those in CACNA1F and CACNA2D4.²¹

Dogs present unique opportunities for the investigation of hereditary conditions and particularly for IRDs.^22,23 In 2017, we reported on a retinal atrophy that progressed to blindness in whippet dogs.²⁴ Here, we identify the cause of the atrophy in these dogs to be a spontaneous mutation in CaBP4 that resembles the human disease. Dogs with the mutant gene show synaptic ribbon immaturity, synaptic disorganization, and profound changes in the OPL and visual dysfunction. Notably, we show that gene augmentation therapy in adult CaBP4-mutant dog results in maturation and elongation of synaptic ribbons and expansion of the OPL through the “newly” established intricate synaptic connections. These findings demonstrate exceptional plasticity in adult mammalian retina.

This work lays the foundation for gene therapy clinical studies to treat the analogous human condition and advances our understanding of synaptic maturity and plasticity following restoration of calcium-driven synaptic signaling. The degree of plasticity and synaptic recovery we see here showcases not only the potential of retinal maturation and reorganization following gene therapy but is fundamental to treatments involving other novel retinal therapeutics such as prime or gene editing and cell-based techniques such as those using retinal organoids.

Results

CaBP4-mutant dogs show failure in maturation in the first retinal synapse, diminished synaptic machinery, and altered synaptic structure

Whole-exome sequencing was performed on DNA from whippet breed dogs in which recessive progressive retinal degeneration was segregating. A variant affecting the start codon of CaBP4 was found to segregate with disease status (c.3G>A, p.Met1Ile). Young affected dogs had an abnormally thin OPL (Figures 1A and 1B., 2.87 ± 0.20 μm in CaBP4^−/−, compared to 8.27 ± 0.15 μm in age-matched unaffected control dogs, p = 1.5 x10⁻⁶), while the other retinal layers were of comparable thickness (Figure 1B). Thinning of the OPL was previously reported in mice with a knockout (KO) mutation of cabp4, as well as in those with Ca_v1.4 KO, with both genes reported to participate in calcium regulation in the first retinal synapse in the OPL.^25,26 Nonetheless, this feature is not unique to mutations involved in synaptic calcium regulation, and while not discussed, can be seen in images of mouse retina with a KO mutation in NudC, a mitotic protein involved in photoreceptor function and survival.²⁷ The synaptic ribbons of the photoreceptor to second order neuron of mutant dogs had fewer identifiable ribbons than controls, and when identified, ribbons appeared to be present in immature forms, including short-elongated forms, club-shaped ribbons, and sphere-shaped ribbons (Figure 1C). Ribbons were significantly shorter in CaBP4^−/− (227 ± 14 nm; n = 100) than in wild-type (WT) (642 ± 27 nm; n = 120) or CaBP4^+/− (692 ± 27 nm ; n = 101) (p = 1.5 × 10⁻²⁸, CaBP4^−/− vs. WT; p = 4.85 × 10⁻³², CaBP4^−/− vs. CaBP4^+/−; p = 0.19, CaBP4^+/− vs. WT), supporting immaturity (Figure 1D). Identifiable synaptic ribbons in CaBP4^−/− dogs were seldom associated with a traditional triad (Figure S1). While the synaptic ribbon CtBP2 (i.e., RIBEYE) protein still showed abundant expression in CaBP4^−/− dogs, it appeared ectopically and was displaced in clusters away from the inner nuclear layer, rather than showing the expected dual distribution in association with rod and cone pedicles (Figure 1E). Furthermore, synaptic ribbons seen by immunohistochemistry (IHC) appeared immature and truncated, rather than elongated, consistent with transmission electron microscopy (TEM) findings (Figure 1E). Occasionally ectopic ribbons were seen in the ONL, as previously described for Ca_v1.4 KO mice.²⁶

Figure 1.

Outer plexiform layer thinning and immature synaptic ribbon formation in the CaBP4-mutant canine retina

(A) Whole-retina sections from 3-month-old WT and CaBP4^−/− dogs illustrating that the outer plexiform layer (OPL) (arrowhead) is thinner in CaBP4^−/−dogs. Scale bars, 10 μm. (B) Quantification of layer thickness from thick section triplicates of 4 control dogs (2 each WT and CaBP4^+/−) and 3 CaBP4^−/− dogs, as illustrated in (A). (C) Transmission electron microscopy sections from 3-month-old WT, CaBP4^+/−, and CaBP4^−/− dogs, illustrating numerous and mature synaptic ribbons (arrowhead) in WT and CaBP4^+/− dogs. CaBP4^−/− dogs had immature synaptic ribbons that were at times challenging to identify. Scale bars are 2 μm and 500 nm (insets). (D) Synaptic ribbons did not differ in length between WT and CaBP4^+/−dogs but were considerably shorter in CaBP4^−/− dogs. (E) WT canine retina exhibiting the normal appearance of CtBP2 (immunolabeled in green) in the OPL. Protein clusters in multiple short-elongated ribbons near cone pedicles or form a larger horseshoe shape when associated with a rod spherule. Synaptic ribbons in CaBP4^−/− dogs were present in abundance but appeared in poorly defined forms and clusters. Inset illustrates that even the most exemplary cone and rod ribbon synapses in the section had immature truncated forms. Scale bars, 10 μm.

Abbreviations: OS, outer segment; IS, inner segment; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. ∗∗∗∗, p < 0.0001. ns, non significant. Error bars represent Mean±SEM. Whiskers represent the full range of the dataset.

Aside from a thinned OPL, the retinas of 3-month-old CaBP4^−/− dogs exhibit normal retinal layer thickness. IHC of the retinal sections revealed a profound reduction in the postsynaptic bipolar cell (BC) G-protein coupled receptor 179 (GPR179) (Figures 2A and 2B) in the OPL of affected dogs and a moderate reduction in the binding of peanut agglutinin (PNA), which labels presynaptic cone photoreceptor pedicles. At this age, reduction in GPR179 was most notable in the outer region of the OPL associated with rod pedicle synapses. The postsynaptic density protein 95 (PSD95), which localizes to photoreceptor terminals,²⁸ was present similarly in specimens from CaBP4^−/− and WT dogs. LRIT3 protein, which localizes presynaptically to rod photoreceptor pedicles,²⁹ was present; however, it labeled punctae that are less discrete than in WT (Figure 2C). LRIT3 is instrumental for the assembly of several trans and postsynaptic proteins in the first retinal synapse, including TRPM1.²⁹ It is required for nyctalopin localization to the depolarizing (ON-) BC dendritic tips, as well as for mGluR6 and GPR179 localization to cone depolarizing BCs but not their localization to rod BCs. While LRIT3 presentation appears slightly disrupted in early age, it is largely in place and cannot explain the absence of GPR179 from the synaptic machinery nor the absence of a b-wave early in life.³⁰ Co-labeling of cones with human cone arrestin (hCAR) and rod BCs with protein kinase C-α (PKC-α) demonstrated that the affected dog retina had elongation of BC dendrites beyond the photoreceptor pedicles (Figure 2D). BC and cone synapses not only extended past each other but exhibited a thickening, rather than the typical fine finger-like projections. The intimate association between the cone pedicles and BC dendrites was lost in the affected retina.

Figure 2.

Immunohistochemistry shows that synaptic architecture of the outer plexiform layer is severely disrupted in CaBP4-mutant dogs

(A) GPR179 is present in abundance in WT dogs and co-labels with PNA in the cone photoreceptor pedicles. CaBP4^−/− dogs show significantly diminished GPR179, with GPR179 being nearly absent from rod spherules. (B) The presynaptic protein PSD95 is present in abundance in both WT and CaBP4^−/− dogs in the OPL. (C) In WT dogs, the synaptic ribbon CtBP2 clusters in rod spherules and cone pedicles and is tightly associated with LRIT3 punctae. CaBP4^−/− dogs show abundance of CtBP2 that is dispersed in the OPL, while LRIT3 is still present and associated with CtBP2, and its punctae are poorly defined. (D) When immunolabeling cones and bipolar cells with hCAR and PKC-α, the synaptic pedicles and dendrites show intricate relationship in the WT animal. In CaBP4^−/− dogs, both photoreceptor pedicles and bipolar cell dendrites are stretched and overreach past each other. Z projections of IHC sections.

Scale bars are 50 μm (A–D), 20 μm (inset A), 10 μm (insets B, C, and D).

Staining for calbindin showed normal horizontal cell (HC) somatic morphology in young CaBP4^−/− dogs that was also evident on TEM, although labeling for SNAP25 in this group was more pronounced than in controls, with some ectopic extension of HC dendrites into the outer nuclear layer (ONL) (Figures S2A and S2B). The proximity of HC dendrites to the ONL can be attributed to the previously discussed collapsed OPL. Synaptophysin and glial fibrillary acidic protein (GFAP) showed subtle increased expression and formed clusters in CaBP4^−/− dogs (Figures S2C and S2D). Synaptic pedicles appeared swollen and coarse in the OPL, as can be seen in Figure 2D. TRPM1 labeling of synaptic dendrites appeared to have normal abundance in young CaBP4^−/− dogs (Figure S2E). Photoreceptor soma and inner/outer segment morphology appeared normal in young CaBP4^−/− dogs, with similar cone opsin expression, as well as PNA labeling of the cone outer matrix sheath around the outer segments and anti-rhodopsin antibody (RET-P1) labeling rods (Figures S2F–S2H).

CaBP4-mutant dogs have reduced vision and cone-rod synaptic dysfunction followed by progressive photoreceptor degeneration

Ophthalmoscopic indications of retinal thinning became apparent in CaBP4^−/− dogs between 9 and 12 months of age, namely a hyperreflective appearance to the tapetal fundus (an ophthalmoscopic indicator of thinning of the retina) and attenuation of superficial retinal vasculature (Figure 3A). Except for changes seen in the OPL, retinal thickness measured by spectral domain optical coherence tomography (SD-OCT) was normal in the first 3 months of life (Figures 3B and 3C). On SD-OCT examination, the total retinal thickness of CaBP4^−/− dogs was found to decrease progressively compared with its initial thickness (F(4,25) = 238.5, p < 1 × 10⁻⁵ at 18 months) (Figures 3B and 3C). The retinal thinning was accounted for by progressive thinning of the outer retina, as measured by the receptor plus layer (REC+), which spans the OPL, ONL, photoreceptor inner and outer segments, and retinal pigment epithelium (F(4,25) = 164.6, p < 1 × 10⁻⁵), whereas the thickness of the inner retina remained largely unchanged during the first 18 months of life. Retinal thickness of the visual streak (VS), area centralis (AC), and dorsal retina was compared in 6 eyes of CaBP4^−/− dogs aged 2–4 years old. The results showed significant regional variation supporting a preservation of retinal thickness in the VS and AC compared to elsewhere in the dorsal retina (mean retinal thickness dorsal, 104 ± 4 μm; AC, 151 ± 5 μm, VS, 148 ± 4 μm, F(5,10) = 3.68, p < 0.03) (Figure S3).

Figure 3.

CaBP4 mutation causes a progressive retinal degeneration and extinguished b-wave

(A) Fundus images of WT and CaBP4^−/− dogs showing that CaBP4^−/− dogs have progressive retinal thinning and vascular attenuation. (B) OCT images from WT and CaBP4^−/− dogs showing progressive retinal thinning of the total retina (TR) and receptor plus (REC+) complex. (C) Longitudinal measurements of retinal layer thickness illustrating that the loss in retinal thickness during the first 18 months of life is largely due to a loss in the REC+ complex. The dashed magenta line reflects the average retinal thickness of a mature WT retina. (D) Dark-adapted electroretinograms (ERGs) at varying luminance of stimuli showing a lack of b-wave but the presence of a-wave in CaBP4^−/− dogs. Light-adapted ERGs also lack a b-wave and illustrate a diminished cone-flicker response. (E) When ERGs were repeated over time, the a-wave was lost both in dark- and light-adapted conditions. ∗∗∗∗, p < 0.0001. ns, non significant. Error bars represent standard deviation.

The electroretinogram (ERG) of affected dogs was characterized by an absent rod and cone b-wave (PII response), resulting in a characteristic negative waveform indicating impaired BC function with resulting lack of ERG PII response (Figure 3D). Initially, CaBP4^−/− dogs had an increased amplitude of dark-adapted a-wave due to the lack of a b-wave that normally impinges on the underlying negative photoreceptor-driven PIII response (Figure 3E). With disease progression and outer retinal thinning in CaBP4^−/− dogs, the a-wave amplitude progressively decreased (Figure 3E). Additional luminance response curves and ERG traces are shown in Figure S4.

Though the ERG had an absent b-wave and, within months, became entirely flat, affected dogs had retained some degree of vision in the long term. Affected dogs had reduced visual function in scotopic conditions, which was detected both in a 4-choice exit device and an obstacle course test. Vision impairment paralleled the progression of retinal degeneration. While visual performance in photopic conditions trended to be worse in affected dogs, this did not reach significance in the first year of life (Figure S5).

CaBP4 augmentation therapy restores the ERG b-wave and improves visual performance

All CaBP4^−/− dogs treated with either canine (dCaBP4) or human (hCaBP4) cDNA adeno-associated virus (AAV)-mediated gene augmentation therapy (AAV8(733) GRK1-dCaBP4/hCaBP4) showed improved visual performance and dramatically improved ERG function, and no difference in rescue was detected when using either vector (Figure S6). With vision testing using a 4-choice exit device in scotopic conditions (Figure 4A), the two outcome measures—correct choice of open tunnel and time to exit the device—were improved in the treatment group vs. paired untreated controls (% success p = 3 × 10⁻⁵, exit time, p = 0.001) and were not statistically different from unaffected controls. Using an obstacle course as second method for vision testing, the treated dogs also showed improved scotopic performance (Figure 4B, Video S1). The time to complete the obstacle course in the treatment group vs. paired untreated controls was significantly reduced (p = 0.0004), and the number of collisions was reduced in the treatment group (p = 5 × 10⁻⁵). Performance was not significantly improved in photopic conditions, likely due to the young age of treated dogs, which still had residual photopic vision (Figure S5). No difference was detected in visual performance when using the hCAPBP4 compared with the dCaBP4 vector (Figure S6).

Figure 4.

Gene therapy improves visual performance and recovers the electroretinogram

(A) Visual performance assessed with a 4-choice test showed that scotopic performance was comparable between the unaffected and treatment groups both in terms of the time to exit the apparatus and percent success. The untreated group exhibited significantly longer times and reduced success rate when completing the task. One way ANOVA, F(2,20) = 13.02, p = 2 × 10⁻⁴, n = 5 controls, and n = 9 treated/untreated pairs. (B) Visual performance assessed with an obstacle course showed that the scotopic performance was improved in the treatment group when compared with the untreated group both in terms of time to exit the apparatus and the number of collisions. One way ANOVA, F(2,20) = 23.06, p = 6 × 10⁻⁶, n = 5 controls, and n = 9 treated/untreated pairs. For (A) and (B), the dogs were treated at an average of 12.75 ± 3.11 months of age, and the data presented include the last time point of collection for each condition (Table S2). (C) A representative ERG from a dog treated at 4.5 months of age and recorded 6 months following treatment is compared with the ERG at 3 months of age, showing the recovered b-waves in both scotopic and photopic conditions and increased cone-flicker response amplitudes. (D) Results from 4 dogs treated at an average of 6.35 ± 1.18 months of age and recorded at 9.73 ± 1.01 months of age and compared with 6-month-old WT dogs. The light- and dark-adapted a-wave amplitudes resemble those of young unaffected dogs, and the light- and dark-adapted b-wave amplitudes were increased. Amplitudes are reduced compared with normal as only a part of the retina was treated with a subretinal injection. ∗∗, p < 0.01; ∗∗∗∗, p < 0.0001. For (A) and (B) whiskers represent the full range of the dataset, for D error bars represent standard deviation.

ERGs following treatment showed recovery of the scotopic and photopic a and b-wave amplitudes of the treated dogs (Figure 4C). Before treatment, the scotopic and photopic ERG b-waves were not detectable but, following therapy, were restored and were 91 ± 18 (n = 4) at 0.01 cd⋅s/m² luminance, respectively, for dark-adapted conditions and 35 ± 3 (n = 4) at 3 cd⋅s/m² luminance for light-adapted conditions. Two dogs were treated at later disease stage (28 months of age) when photoreceptor degeneration was well established. Just prior to the treatment, the ONL thickness in the location of the treatment area was 33% and 34% of what it was at 3 months of age. These two dogs also showed recovery of the ERG b-wave, with rescue being maintained to the last recording time point 18-month post-treatment (Figure S7). Three dogs were treated with gene therapy and were followed for over 2 years while maintaining rescue as documented with ERG, vision testing, and SD-OCT. One of these dogs was treated bilaterally at a young age (4 and 6 months) and was followed until 44 months of age and showed robust and sustained ERG recovery (Figure S8). The ERG recovery seen in the treated dogs is not expected to match amplitudes in WT dogs because improved electrophysiologic function is expected only in bleb/treated regions, which do not span the entire retina. Furthermore, if treated when photoreceptor loss was already well established, there would be fewer photoreceptors to drive the response.

CaBP4 augmentation therapy preserves retinal thickness, expands the OPL, and promotes maturation of synaptic ribbons

Fundus examination showed that retinal vasculature and retinal appearance were well preserved within the gene augmentation-treated retinal regions (Figure 5A). SD-OCT scans showed preservation of the total retinal thickness and layer definition within the treated retinal regions compared with untreated retinal regions (188 ± 3.5 μm vs. 140 ± 9.5 μm, p = 1.4 × 10⁻⁴, n = 12) (Figure 5B). Marked thinning of the retina due to progression of the degeneration was present in the adjacent untreated regions (Figures 5A and 5B). Preservation of retinal thickness did not differ between dogs treated with the dCaBP4 or hCaBP4 vectors (p = 0.29 in untreated regions, and p = 0.44 in bleb/treated areas) (Figure S6B).

Figure 5.

Gene augmentation therapy recovers retinal morphology, outer plexiform layer thickness, and synaptic ribbons

(A) Fundus images showing advanced retinal degeneration in the 30-month-old CaBP4^−/− dog eye. In the eye treated at 4.5 months of age, normal morphology was noted in the treatment areas (blebs) 38 months following injection. (B) OCT heatmap of a dog treated at 6.3 months of age and imaged at 24.2 months of age shows retained retinal thickness in the bleb area compared with the non-bleb area. A fundus photograph below shows the region of the heatmap with the ventral bleb margin approximated in magenta. The ventral bleb margin is easily seen in a scanning laser ophthalmoscopy image and highlighted with an arrowhead. A corresponding line scan illustrates normal retinal morphology in the treatment area with loss of retinal thickness, most notably in the outer retina in the untreated area. A boxplot compares the retinal thickness in unaffected controls to the bleb and non-bleb regions of 12 dog eyes treated at 9.33 ± 0.96 months of age and imaged at 21.24 ± 2.65 months of age, showing no difference between the treated eyes and controls in the overall retinal thickness. Whiskers represent the full range of the dataset. (C) Fundus photographs from a dog treated at 9.7 months of age and photographed at 21.5 months of age, along with thick sections obtained from bleb and non-bleb areas of the same dog collected at 22.9 months of age. Two black foci in the treatment eye represent the retinotomy/injection sites. The magenta line highlights the ventral region of the merging blebs, and the two white circles approximate the locations from which the shown plastic sections were obtained. Summary of retinal layer thickness measured from 3 dogs treated at an average of 8.18 ± 0.99 months of age and collected at 29.10 ± 5.86 months of age shows increased retinal thickness in the bleb regions, most notably in the OPL. Scale bars, 10 μm. Error bars represent Mean ± SEM. (D) TEM sections from the bleb and non-bleb regions of the same dogs as in (C) show elongated synaptic ribbon morphology in the treated regions compared with short, immature ribbons in the untreated regions. Scale bars, 2 μm and 500 nm (inset). The boxplot illustrates that the newly matured synaptic ribbons are no different in length from those in control animals. ∗∗∗, p < 0.001; ∗∗∗∗, p < 0.00001. Whiskers represent the full range of the dataset.

Retinal layer thickness was also measured on plastic embedded sections. The gene therapy-treated regions had a thicker OPL than the untreated regions (6.8 ± 0.5 μm in treated vs. 1.7 ± 0.8 μm in untreated regions p = 4 × 10⁻⁶, n = 3), being only slightly reduced compared to that in WT control eyes (6.8 ± 0.2 μm in treated vs. 8.2 ± 0.15 μm in unaffected controls, p = 0.02) (Figure 5C). TEM showed the elongation and maturation of synaptic ribbons in the treated regions compared with untreated regions, such that the ribbon length in the treated regions was not different from that in control WT (604 ± 15 nm, n = 245 treated vs. 642 ± 27 nm, n = 120; p = 0.19). Ribbons in the treated regions were significantly longer than those in the untreated retinal regions of CaBP4^−/− dogs (604 ± 15 nm, n = 245 treated vs. 245 ± 10 nm, n = 171 untreated; p = 2.5 × 10⁻⁶²). Furthermore, ribbons in the treated regions were associated with the formation of the expected synaptic triads involving BC and HC dendrites (Figure 5D).

CaBP4 augmentation therapy restores synaptic machinery and augments synaptic structure

IHC of the retinal sections exhibited a complete recovery of the postsynaptic BC GPR179 labeling in the treated regions (Figures 6A, S6C, and S8C), comparable to that in normal controls (Figure 3A). The average number of punctae in a 70-μm-long section of the OPL was 78.0 ± 11.3 in WT controls and 78.5 ± 14.3 in the treated CaBP4^−/− dogs (n = 4 per group, p = 0.979). There was also preservation of PSD-95 in the photoreceptor pedicles, which diminished with advanced retinal degeneration, compared with the young CaBP4^−/− dogs (Figures 2B and S6). The effect of the treatment was sharply reduced beyond the treated regions (area of the bleb), and this is seen both as a reduction of protein expression and retinal layer thickness. Co-labeling of cones and rod BCs with PKC-α and hCAR exhibited a fine intricate network of synaptic connections that was restored between photoreceptor pedicles and BC dendrites in the region of treatment. This contrasts with the thickening and elongation of BC dendrites beyond the photoreceptor pedicles and the photoreceptor synaptic pedicle thickening seen in the untreated regions (Figure 6B). As the disease progressed in the untreated retina, there was profound loss of retinal thickness and disrupted photoreceptor morphology (Figure 6B), which were also evident as an absence of distinctive outer retinal zones on SD-OCT (Figure 5B). Synaptic ribbons remained in abundance in the treated regions and had improved demarcation (Figure 6C), as also shown in Figure 5D. Untreated regions showed a drop off in both CtBP2 and LRIT3 as degeneration advanced (Figure 6C). Animals treated in late stages of the disease and followed for a long term showed robust expression of synaptic proteins in the treated regions, though those extended into the inner ONL (Figure S7). By contrast, the ONL or residual OPL of the untreated regions was nearly devoid of synaptic proteins. Co-labeling of GPR179 and PKC-α illustrated that in the treated areas, GPR179 was not only abundant as in the WT but was appropriately localized to the tips of the BC dendrites (Figure 6D); by contrast, untreated regions in the same retinal section resembled those of young CaBP4^−/− dogs, with rare GPR179 and ectopic elongation of the BC dendrites into the ONL (Figure 6D). The animals followed for a long term showed more profound degeneration in the untreated regions as well as retraction of BC dendrites and ectopic extension of BC soma into the ONL, which caused obliteration of the OPL and consolidation of the ONL and INL (Figure S8).

Figure 6.

Gene augmentation therapy leads to recovery of synaptic machinery and morphology

(A) Immunohistochemical section of the junction between treated/bleb region (to left) and untreated region on right of a dog treated at 7 months of age and collected at 15 months of age. Insets illustrate normal morphology of the OPL and abundant GPR179 and PSD95 in the treated region. (B) Immunolabeling for PKC-α and hCAR of a dog treated at 5 months of age and collected at 15 months of age shows normal outer retinal morphology and thickness in the treated area, with marked retinal thinning of the untreated area. The BC dendrites and photoreceptor pedicles are fine and intricately connected in the treated area, while the untreated area shows thickening and blunting of the synaptic connections with occasional extension of BC dendrites into the outer retina. (C) Treated area of the same eye as in (A) showing abundant CtBP2 and LRIT3 proteins, with clear LRIT3 punctae associated with CtBP2. The untreated regions illustrate diminished CtBP2 and LRIT3 in a thinner OPL. (D) GPR179 is abundant and co-localizes to bipolar cell dendrites in WT dogs and treated sections of the same eye as in (B). The untreated region of the treated eye resembles that of the CaBP4^−/−control. GPR179 labeling is too scant to detect without a Z projection. Bipolar cell dendrites are stretched into the outer retina in the untreated retinal region. (E) Retinal sections of an animal treated in the late stages of disease at 28 months of age and collected at 44 months of age show abundance of GPR179 and PNA. GPR179 labeling is extending into the outer nuclear layer. The untreated retina is thinner, particularly in the outer retina, and the PNA and GPR179 are difficult to identify.

Z projections of IHC slices (A–C) and a single slice (D and E); all paired images are obtained from a single histological section. Scale bars are 50 μm (A, B, and E), 10 μm (B, C, and E insets), and 20 μm (A and D insets).

Discussion

Our results show that gene therapy is not only capable of restoring retinal function but is also able to establish a nearly normal anatomical arrangement in the OPL of the retina. This recovery involves expansion of a layer that did not form normally during retinal development along with maturation of synaptic features such as ribbon elongation, supporting the lasting blueprint of the retinal neural network through adulthood.

The regenerative potential of the OPL

Using gene augmentation therapy in a large-animal model of CaBP4 cone-rod synaptic disorder, we show remarkable plasticity in the OPL and first retinal synapse. Affected animals failed to develop normal photoreceptor to BC synapses, which led to visual impairment. Gene augmentation therapy induced the development of normal synaptic features including maturation and normalization of synaptic ribbons (Figure 5). Further, normal expression of synaptic proteins such as GPR179 was restored at the correct location (Figure 6). Existing cone pedicles and BC dendritic terminals, which previously extended beyond each other, could form appropriate and intricate connections leading to expansion of the OPL and recovery of the ERG (Figure 6). A surprising and important finding was that the OPL could be expanded beyond the thickness seen in the young CaBP4^−/− dogs prior to therapy, illustrating that photoreceptor pedicles and BC dendrites can extend or retract to establish a normal retinal layer, expanding a synaptic space that had not previously developed. While preservation of retinal thickness has been reported following gene therapy,^31,32 such expansion of the OPL has not been previously demonstrated. Expansion of the OPL by photoreceptor to BC synapses normally occurs in early retinal development, as shown in human retina and retinal organoids.³³ Structural plasticity was also demonstrated by the elongation of photoreceptor inner and outer segments and increased distinction of outer retinal zones seen on SD-OCT. This is most likely due to the ability of surviving retinal photoreceptors to sprout/elongate their inner and outer segments and reestablish the ellipsoid and interdigitation zones. Structural normalization is followed by electrophysiological recovery, and vision is improved and maintained long term. These regenerative properties are possible and similarly illustrated in animals treated in later stages of the diseases (Figures 6E and S7). Recovery of structure is maintained up to 3 years following treatment (Figure S8). These results show promise for the treatment of IRDs in mature mammals and for retinal therapies that require retinal rewiring or reafferentations, such as those employing stem cells and organoids.

The role of calcium in synaptic plasticity

Widespread synaptic immaturity is present physiologically during natural retinal development and in certain pathologic conditions.²⁶ Cacna1 and Cabp4 KO mice exhibit disrupted synaptic organization, signaling, and transmission in the retina.^25,26,34 Our results indicate that disruption to calcium modulation in the dog results in anatomical aberrations similar to those previously reported in the Cacna1 and Cabp4 KO mice.^25,35 The results support that calcium and its channel are required for normal synaptic assembly, and any interference via its binding or voltage-gated channel tends to result in abnormal neuronal sprouting, synaptic ribbon immaturity, diminished-to-absent ON-pathway synaptic transmission, and vision deficits.³⁶ Photoreceptor ribbon synapses are capable of dramatic structural remodeling in cell culture³⁷ and exhibit seasonal plasticity in the retina of hibernating ground squirrels.³⁸ In the torpid 13 line ground squirrel, synaptic ribbon structures undergo degeneration, and the synaptic vesicles migrate away from the synapse, in a reversible process.³⁹ Some synaptic plasticity was also demonstrated in an inducible Cav1.4 KO mouse model.⁴⁰ Our results illustrate that maturation of synaptic ribbons in both cones and rods is possible in an adult mammalian model of retinal degeneration following gene therapy. Furthermore, recovery of synaptic interactions between cones and ON-BCs, as seen by GPR179 and PNA IHC labeling, shows the effects that calcium channel regulation and dysregulation exert on upstream and downstream retinal organization. The blueprint for both ribbons and GPR179 assembly remains in the absence of a functional calcium regulation and can be recovered by gene augmentation, even when treating patients with established photoreceptor degeneration. These findings are consistent with the crucial roles that the voltage-gated calcium channel plays in neural function including gene regulation, neurotransmitter release, and neuronal excitability.⁴¹

Current limitations and potentials of gene augmentation therapy

Recent methods for vision restoration including gene-based therapies such as gene augmentation and prime editing, as well as treatment with retinal organoids require a degree of retinal plasticity for a successful outcome. This is particularly relevant in cases of structural abnormalities where normal synaptic connections are lacking. Our results support the possibility of reafferentation of BC dendrites, enabling connections at the first retinal synapse via a traditional triad of BCs, HCs, and photoreceptors.

In summary, our findings exhibit restoration of visual function in dogs with severe electrophysiologic and synaptic dysfunction. We establish that the OPL of the retina has profound plasticity and that the synaptic ribbons can mature and elongate following gene augmentation therapy well into adulthood.

Material and methods

Animals and ethics statement

Dogs used in this study were from a breeding colony maintained at Michigan State University. The dogs were bred from a whippet sire and dam that originated from Brazil and were previously reported in the literature.^24,42 Animals included homozygous CaBP4-mutant affected breed dogs (CaBP4^−/−), heterozygous dogs that are carriers for the mutation (CaBP4^+/−), and WT dogs. Unaffected controls were either WT or CaBP4^+/− dogs after the latter were confirmed to be phenotypically not significantly different from WT. Table S1 includes the details of the animals and interventions performed on each animal. All dogs were either whippet breed or obtained from a cross of whippet and beagle that originated from the same colony. All procedures were conducted in strict accordance with the guidelines of the Association of Research in Vision and Ophthalmology (ARVO) statement for Use of Animals in Ophthalmic and Vision Research. The study was approved by the Institutional Animal Care and Use Committee at Michigan State University (IACUC approved protocol nos. 202000013, 201900182, and 202300089).

Ocular examination, optical coherence tomography, and fundus imaging

For examination and SD-OCT, pupils were dilated with tropicamide 1% (Bausch + Lomb, Bridgewater Township, NJ, USA). Ophthalmic examinations included slit-lamp biomicroscopy (Kowa SL-17, Kowa American Corporation, NY, USA) and indirect ophthalmoscopy (Keeler Vantage Plus wireless LED, Keeler Instruments, Inc, Broomall, PA, USA). Wide-angle fundus color images were collected with a RetCam II (Clarity Medical Systems, Inc, Pleasanton, CA, USA) or RetCam III Wide-Field Digital Imaging System (Natus Medical Incorporated, Middleton, WI). SD-OCT was performed with dogs under general anesthesia. Briefly, the dogs were pre-medicated with subcutaneous or intramuscular acepromazine (0.02–0.1 mg/kg; Aceprojet, Henry Schein Animal Health or Acepran, 0.2%, Vetnil) and in some cases with buprenorphine (0.01–0.03 mg/kg; Reckitt Benkiser Company). Anesthesia was induced with intravenous propofol (4–6 mg/kg; PropoFlo, Zoetis, or Propovan 1%, Cristália, Itapira, São Paulo, Brazil) and maintained on 1.5%–3% isoflurane (Isothesia Inc, Henry Schein Animal Health, Portland, ME, USA). Eyes were positioned in primary gaze, using conjunctival stay sutures (4–0 or 6–0 silk, Ethicon, LLC, Johnson & Johnson Company). Imaging was performed using combined confocal scanning laser ophthalmoscopy (cSLO) and SD-OCT (Spectralis HRA/OCT, Heidelberg Engineering Inc., Carlsbad, CA), as previously described.^43,44 Briefly, infrared (IR) and autofluorescence (AF) cSLO images of the fundus were acquired using a 55° lens. High-resolution horizontal and vertical SD- OCT single-line scans as well as raster volume scans were obtained using a 30° lens. Scans were also performed at the interface between the treated and untreated regions (the boundary of the subretinal injection bleb). The total retinal thickness (from the internal limiting membrane to the retinal pigment epithelium) as well as the outer retinal thickness (REC+; photoreceptor layers plus the OPL) were measured while avoiding measurement where major superficial retinal blood vessels are present. Retinal thickness was measured using the integrated Heidelberg Eye Explorer (HEYEX) software, either manually or with manual correction.

Assessment of visual behavior

Vision testing was performed using either a four-choice vision exit device or an obstacle avoidance course. The four-choice test was performed as previously described.^45,46 Briefly, the device consists of a central box, which has four exit tunnels. The dog is placed in the box, and their performance is assessed during a series of “runs.” For each “run,” a single randomly selected exit tunnel is open. Performance was determined by two measures; the first was success or failure, which was dependent on whether the first exit tunnel entered by the dog was the open tunnel, and the second measure was the time taken to exit the device. Evaluation was performed either with both eyes open or for each eye individually by placement of an opaque contact lens over the contralateral eye following the application of topical ophthalmic anesthetic (proparacaine hydrochloride 0.5% Bausch + Lomb, Bridgewater Township, NJ, USA). Each eye was tested by 14 repeated trials at 7 different light intensities set at the exit of the tunnels (luminance range, 0.057–750 lux) as measured by a photometer (IL1700 with an SED033/Y/R Illuminance detector, International Light, Peabody, MA, USA), and the results for each trial (individual eye or for both eyes open) at each light intensity were averaged. The obstacle avoidance course device used was previously described in detail.⁴⁷ Briefly, a 3.6-m-long and 1.2-m-wide obstacle course with 6 adjustable panels was used to assess vision. The dog would enter the device at one end, and their performance was assessed until they exited the device at the other end in a series of “runs.” The adjustable panels were alternated in locations between each run to prevent the dog from learning the location of the obstacles (i.e., panels). For the current study, visual performance was evaluated in both scotopic and photopic conditions (0.057 and 750 lux, respectively) and recorded with an infrared video recorder. For each “run,” performance was determined by two measures; the first was the number of collisions, counted as a nose/head collision with a panel within the obstacle course, and the second measure was the time taken to exit the device. Evaluation was performed either with both eyes open or for each eye individually by placement of an opaque contact lens over the contralateral eye, as described for the four-choice device. Each trial included 5 repeated runs, and the results for each trial (individual eye or for both eyes open) at each light intensity were averaged for any tested time point.

Assessment of retinal function via ERG

ERG was performed under general anesthesia (inhaled isoflurane), as previously described and in compliance with the International Society for Clinical Electrophysiology of Vision (ISCEV) standards.^48,49,50,51 Briefly, the pupils were dilated using tropicamide 1% (Bausch + Lomb, Bridgewater Township, NJ, USA) and 10% phenylephrine hydrochloride (Bausch + Lomb, Bridgewater Township, NJ, USA). The globes were positioned in primary gaze by using an eyelid speculum and conjunctival stay sutures (4-0 silk, Ethicon, Inc., Piscataway, NJ, USA). Full-Field flash ERGs were recorded using ERGJet lenses (Fabrinal SA, La Chaux-de-Fonds, Switzerland) and platinum subdermal reference (positioned 1–2 cm from lateral canthus) and ground electrodes (positioned under the skin of the neck) (Grass Instruments, Natus Medical Inc., Pleasanton, CA), using an Espion E3 electroretinography system with a ColorDome Ganzfeld stimulator (Diagnosys, Inc., Lowell, MA, USA). Electrode impedance was maintained at <5 kΩ. Dark-adapted ERG responses were recorded following at least 1 h of dark adaptation, from a series of increasing flash luminance stimuli ranging from below threshold to a strong stimulus (−3.26 log cd⋅s/m² to +1.4 log cd⋅s/m²). Light-adapted (photopic) single-flash and flicker ERGs were recorded following exposure to a standard background light of 30 cd/m² for 10 min.

Histology and TEM

Retinal morphology was evaluated on plastic-embedded retinal sections that were collected and processed as previously described.^32,48 All untreated WT and affected samples were collected between 11 and 13 weeks of age. To obtain semithin sections, globes were fixed in 3% glutaraldehyde and 2% paraformaldehyde. After fixation, a 4-mm punch biopsy was used to remove and collect a circular piece of retina from matched regions, including the AC temporally, VS nasally, and superior (dorsal) and inferior (ventral) retina. In animals that underwent gene therapy, paired biopsy samples were collected in proximity of regions within and outside the boundary of the subretinal injection bleb. The bleb or treatment region was defined using several methods including fundus photography obtained during surgery and follow-up examinations and optical coherence tomography. Due to the degenerative nature of the disease and an easily visible “high-water mark” delineating the bleb margin, the treatment region was most often easily visible on fundus images. The samples were postfixed in 2% osmium tetroxide, dehydrated in acetone series, and infiltrated and embedded in low-viscosity epoxy Spurr resin for electron microscopy.⁵² Serial sections at 2-μm intervals were stained with toluidine blue. The samples were chosen based on quality and retinal orientation and were further sectioned for imaging (JEOL JEM-1400Flash, Jeol, USA, Peabody, MA, USA). To measure retinal layer thickness, ImageJ software (National institute of health, Bethesda, MD, USA) was used to measure each layer at three different locations from three different samples obtained from the dorsal retina, which were collected by a masked microscopist. The sum of all layers was then compared with the measured total retinal thickness at the section to ensure that they were not statistically different (Student’s t test).

IHC

Following humane euthanasia by barbiturate overdose (Fatal Plus, Vortech Pharmaceuticals, Dearborn, MI, USA), eyes were removed and processed for IHC, as previously described in vertical sagittal frozen sections.⁵³ The location of the AC was estimated using SD-OCT and by calculating the horizontal distance of the AC from the optic nerve head (expressed as a ratio of the horizontal diameter measured at the optical rim). The width of the optic nerve head was measured during sectioning and used to calculate the approximate horizontal distance to the AC. IHC was performed as previously described using antibodies for cone arrestin (hCAR; labels all cones; kind gift from Cheryl Craft, University of Southern California),^54,55 L/M- and S-cone opsin were counterstained with PNA (cone subtypes),⁵⁴ rhodopsin (rods),⁵⁴ PKC-α (rod bipolar cells),⁵⁶ and GFAP (activated glial cells). Table S2 details the primary and secondary antibodies used, their origins, and dilutions. 4′,6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich Corp., St Louis, MO, USA) nuclear counterstain was used.

Confocal images from the immunolabeled cryosections were collected using a Nikon A1R confocal scanning laser microscope (Nikon Instruments Inc., Melville, NY, USA) at 20×, 40×, and 60×. Digital images were captured using the NIS Elements software (Nikon Instruments Inc., Melville, NY, USA) and processed using ImageJ software (National institute of health, Bethesda, MD, USA).

Genomic sequencing

Genomic DNA was collected from blood samples of WT or CaBP4^−/− whippet breed dogs. Each DNA sample was quantified using Qubit dsDNA HS kit (Thermo Fisher Scientific, Santa Clara, CA, USA), and integrity was assessed using gel electrophoresis. Only high-quality genomic DNA was utilized for sequencing. Sequencing library was prepared using SEQCAP canine exome enrichment kit (Roche Nimblegen Inc., Pleasanton, CA, USA) and KAPA library preparation kit (KAPA HyperPrep Kit, Roche Nimblegen Inc., Pleasanton, CA, USA). Sequencing was performed using Illumina HiSeq 4000 in a 2 × 150-bp paired-end format, using HiSeq 4000 SBS reagents. Obtained raw fastq files were trimmed for low-quality base pairs and adaptor artifacts with Trim Galore. Trimmed reads were aligned to the CanFam3.1 genome with Bowtie2. Aligned reads were prepared, and duplicated reads were removed using the Picard Tools package (http://broadinstitute.github.io/picard). Freebayes was utilized for variant calling, and subsequent variant annotation was processed with SnpEff. Deleterious mutations were computationally predicted with the Sift4G package (http://sift.jcvi.org). Variants that were shared by the affected samples were identified and filtered to remove any variants present in the unaffected dogs using BCFtools.

AAV vectors and subretinal injections

Vectors were produced at the Powell Gene Therapy Center, University of Florida. Standard cloning techniques were used for vector assembly. All sequence manipulations were verified by sequencing. An AAV construct (serotype 8 with a position 733 capsid mutation) packaged with human or canine CaBP4 cDNA (hCaBP4 or dCaBP4, respectively) with a rhodopsin kinase promoter was used to produce AAV8(733) GRK1-hCaBP4 and AAV8(733) GRK1-dCaBP4, respectively. Standard purification of vectors (iodixanol column) was performed to ensure lack of endotoxin and host cell proteins and their suitability for use in the canine retina.

Subretinal injections in CaBP4^−/− dogs were performed as previously described.^57,58 Briefly, under general anesthesia and direct observation through an operating microscope using a contact vitrectomy lens (Volk1 Single-Use Magnifying, Volk, Mentor, OH, USA) to visualize the posterior segment, a RetinaJect injector (SurModics, Inc., Irvine, CA, USA) was introduced into the vitreous via a sclerotomy performed approximately 5 mm posterior to the limbus aiming to pass through pars plana. To limit post-injection intraocular pressure elevation, aqueocentesis was performed, and approximately 100–150 μL aqueous humor were removed prior to the subretinal injection. Following aqueocentesis, the RetinaJect was advanced across the vitreous to target the central retina superior to the optic nerve head. The vector was delivered by 1–2 subretinal injections until 1–2 well-formed blebs were produced. 40–360 μL doses with vector titers of 5 × 10¹¹ to 2 × 10¹² viral genomes (vg)/mL were administered, resulting in a final dose of 2 × 10¹⁰ to 8.5 × 10¹¹ vg per eye. All surgical procedures were video recorded. Following completion of the procedure, fundus images were captured using a RetCam II video fundus camera (Clarity Medical Systems, Pleasanton, CA, USA) to record images showing the position/extent of the subretinal bleb.

Statistical analysis

Outcomes were analyzed using one-way ANOVA and post hoc multiple comparisons, where appropriate, to evaluate the differences between CaBP4^−/− dogs and controls (unaffected dogs; CaBP4^+/− or WT dogs). Paired t tests were used when comparing pairs of treated and untreated regions in the same eye or between eyes of the same dog.

Data were analyzed using Microsoft Excel 365 for Windows, GraphPad Prism version 10.0.0 for Windows, (GraphPad Software, San Diego, California USA, RRID:SCR_002798), and Sigma Plot for Windows version 14.5 (Systat Software, Inc., San Jose, California, RRID: SCR_003210). Data were considered significant at p < 0.05.

Data and code availability

Data and materials supporting the findings of this manuscript are available from the corresponding author upon reasonable request.

Acknowledgments

This research was funded by Myers Dunlap Endowment for Canine Health (S.M.P.-J), NIH NEI EY027285 (S.M.P.-J.), CVM Endowed Research Funds (S.M.P.-J. and B.B.-C.), and the ACVO Vision for Animal Foundation Resident Grant VAF2022-1 (B.B.-C.). The authors would like to thank the staff of the Michigan State University Center for Advanced Microscopy, the staff of the Michigan State Research and Teaching Technical Support, and the students that supported this work. We would like to thank William W. Hauswirth of the University of Florida, Gainesville, for vector design and production.

Author contributions

Conceptualization, B.B.-C., L.M.O., and S.M.P.-J.; methodology, S.M.P.-J., B.B.-C., L.M.O., P.A.W., and K.S.; genetic studies, M.P., V.Y.-G., P.Z.S., and S.M.P.-J.; formal analysis, B.B.-C. and S.M.P.-J.; investigation, B.B.-C., L.M.O., P.A.W., K.S., L.F.L.P.M., and M.P.; resources, S.M.P.-J., V.Y.-G., F.M.-F., and A.T.S.; data curation, S.M.P.-J., B.B.-C., and L.M.O.; writing – original draft, B.B.-C. and S.M.P.-J.; writing – review & editing, B.B.-C. and S.M.P.-J.; visualization, B.B.-C. and S.M.P.-J.; supervision, S.M.P.-J.; project administration, S.M.P.-J.; funding acquisition, B.B.-C. and S.M.P.-J. All authors have read and agreed to the published version of the manuscript aside from V.Y.-G., as she passed away in the process of preparation of the publication.

Declaration of interests

The authors declare no conflict of interest. The funders had no role in the design of the study; collection, analyses, or interpretation of data; writing of the manuscript; or in the decision to publish the results.

Supplemental information

Video S1. Visual behavior following therapy in a CaBP4-mutant dog

Depiction of an affected dog treated unilaterally with gene augmentation therapy navigating an obstacle course in photopic and scotopic conditions. For assessment of the untreated eye, the treated eye has been covered with an occlusive contact lens. For assessment of the treated eye, the untreated eye has been covered with an occlusive contact lens.