Retinitis pigmentosa (RP) is a major cause of heritable vision loss affecting ~1 in 4,000 people (Hartong et al., 2006). It is a genetically heterogenous group of conditions; at the time of writing 67 mapped genes and loci for non-syndromic RP were listed on the RetNet website (the Retinal Information Network https://sph.uth.edu/retnet/home.htm). In many forms of RP the mutated genes are specific to rod photoreceptors leading to rod degeneration and depending on the function of the gene product, a period of abnormal, or absent, rod function may precede rod death. Although cone photoreceptors in many forms of RP do not express the mutant gene, cone degeneration will occur as a consequence of rod degeneration. Initial symptoms of RP result from reduced, or absence of rod function resulting in night blindness. With progression an impairment of cone mediated vision develops leading to visual field constriction and in many cases, complete blindness. Mutations of the alpha or beta subunits of the rod phosphodiesterase (PDE6) result in autosomal recessive RP (arRP). PDE6-alpha (PDE6A; OMIM 180071) and PDE6-beta (PDE6B; OMIM 180072) mutations each account for ~4% of cases of arRP (McLaughlin et al., 1995; Dryja et al., 1999).

The PDE6 complex in rod photoreceptors is composed of two active subunits, alpha and beta, and a pair of gamma inhibitory subunits. Removal of the gamma inhibitory subunits during phototransduction by the activity of transducin, allows the alpha and beta dimer to hydrolyze many cyclic GMP (cGMP) molecules, causing gated channels to close and plasma membrane hyperpolarization to occur (Farber, 1995). Animals lacking functional alpha or beta PDE6 subunit show a reduction in the phototransduction-induced hydrolysis of cGMP resulting in elevated cGMP levels, unregulated influx of calcium, primary rod cell death, and subsequent secondary cone loss (Wensel et al., 2016).

A variety of mouse models for PDE6B arRP have been widely used to study disease mechanisms and treatment, including Pde6b^rd1/rd1 which has a null mutation, and peak rod loss at day 14, (Keeler, 1966; Pittler et al., 1993; Hackam et al., 2004) and the milder Pde6b^rd10/rd10 missense mutation in which rod loss peaks at day 25 (Gargini et al., 2007; Barhoum et al., 2008). The Pde6b^rd10/rd10 mouse retina initially contains low levels of Pde6b and retains a rod-mediated electroretinogram response prior to degeneration (Chang et al., 2007). Additional mutagenesis models have been described for Pde6b in the mouse, exhibiting a variety of severities of retinal disease (Hart et al., 2005; Chang et al., 2007). Irish setter dogs with a null mutation in Pde6b provide a large animal model for PDE6B arRP (Farber et al., 1992; Suber et al., 1993). Affected dogs lack Pde6b activity and have elevated retinal cGMP levels arresting normal photoreceptor development from day 13, resulting in retinal degeneration beginning at day 25; rod loss is followed by a progressive loss of cones leading to blindness (Aguirre et al., 1978; Farber et al., 1992). Successful gene therapy studies were initially described for the milder phenotypes of the Pde6b^rd10/rd10 mouse (Pang et al., 2008, 2011) and Pde6b^H620Q hypomorphic mouse model (Davis et al., 2008; Tosi et al., 2011). The early onset, rapid degeneration of the Pde6b^rd1/rd1 mouse proved to be more challenging to treat with gene therapy, with only partial rescue of the phenotype described (Bennett et al., 1996; Jomary et al., 1997; Kumar-Singh and Farber, 1998; Takahashi et al., 1999; Guo et al., 2015). A recent description of an additional mutation in Gpr179 in the Pde6b^rd1/rd1 mouse line may explain some of the difficulty in rescuing the Pde6b^rd1/rd1 phenotype (Nishiguchi et al., 2015). Following breeding to remove the co-inherited Gpr179 mutation, gene therapy rescued the Pde6b^rd1/rd1 phenotype. Successful and sustained rescue of the Pde6b mutant dog phenotype with adeno-associated viral vectors encoding canine Pde6b when administered by subretinal injection at 20 days of age has been described (Petit et al., 2012; Pichard et al., 2016).

More recently, several mouse models for PDE6A arRP have been developed. The severity and rapidity of retinal degeneration of each model is dependent on the gene mutation, varying from slow to rapid (Sakamoto et al., 2009; Sothilingam et al., 2015). We previously described a recessively inherited Pde6a null mutation (one base-pair deletion at codon 616 resulting in a frameshift and premature stop codon) in the Cardigan Welsh Corgi breed of dog causing an early-onset and rapid rod and subsequent cone degeneration (Petersen-Jones et al., 1999; Tuntivanich et al., 2009). Outer nuclear layer cell death was present from post-natal day 21 and peaked at post-natal day 27. Cell loss is not related to caspase 3 activity in mouse (Sothilingam et al., 2015) and dog (Tuntivanich et al., 2009) models, indicating a non-caspase dependent mechanism of cell death. Pde6a mouse models have reduced amounts of retinal Pde6b in parallel with the decrease in Pde6a (Sakamoto et al., 2009) and the retina of Pde6a mutant dogs also lacks beta and gamma subunits of Pde6 indicating a requirement for Pde6a for the normal production or maintenance of the other subunits in the rod outer segment (Petersen-Jones et al., 1999; Tuntivanich et al., 2009). This is in contrast to animal models with Pde6b null mutations in which expression of the other Pde6 subunits is retained (Pittler and Baehr, 1991; Suber et al., 1993). Therapeutic trials using the Pde6a^nmf363 mouse model have been reported (Wert et al., 2013, 2014). This particular mouse model has one of the less severe phenotypes. Therapy in this model was successful when performed both prior to photoreceptor loss (Wert et al., 2013) but also when ~50% of photoreceptors had degenerated (Wert et al., 2014). Our initial attempts to rescue the Pde6a mutant dog with AAV serotype 5 gene augmentation therapy were unsuccessful (Petersen-Jones, unpublished findings). We hypothesized that because of the severity of the model, transgene expression would be required at a very early age, requiring delivery of treatment to the retina at a stage at which it is still developing and that we would need highly efficient, rapid-onset, viral vectors. A high efficiency adeno-associated virus serotype-8 with capsid mutation was successful in rescuing the rapid degeneration seen in the Aipl1 mutant mouse (Ku et al., 2015) and the Pde6a^nmf363 mouse model, (Wert et al., 2013) even when delivered at more advanced stages of disease (Wert et al., 2014). We have shown that tyrosine capsid-mutant AAV8 viral vectors have an onset of gene expression within 2–3 days in the canine retina, (Mowat et al., 2014) and therefore show promise for the rescue of the rapid degeneration seen in the Pde6a mutant dog.

Currently there are no definitive treatments for RP associated with retinal PDE6 mutations. Recent phase I/II clinical trials have demonstrated the safety and efficacy of gene augmentation using adeno-associated viral vectors in patients with Leber Congenital Amaurosis due to RPE65 gene mutations (Bainbridge et al., 2008; Cideciyan et al., 2008, 2009a,b; Hauswirth et al., 2008; Maguire et al., 2008, 2009; Simonelli et al., 2010). Experimental efficacy in Pde6b models will allow translation into human clinical trials in the near future, however only limited preclinical trials have been performed in PDE6A models.

In this publication, we present the results of tyrosine capsid-mutant AAV8 Pde6a gene therapy delivered by subretinal injection to young Pde6a mutant dogs. We show that treatment results in partial return of visual function as determined by electroretinography and objective vision assessment. The treatment effect is sustained for at least 4 months, and preserves photoreceptors in the treated area. Treatment reverses retinal cGMP accumulation, reduces mislocalization of rhodopsin, limits glial activation, and maintains dendritic arbors of rod bipolar cells. However, within the treated areas adverse treatment related effects were seen. These include persistent regions of retinal separation and rosette formation.

AAV-Pde6a gene augmentation therapy in a small number of young Pde6a mutant dogs resulted in restoration of Pde6 protein in the outer segments of rods in all treated retinal areas. Restoration of Pde6 in the outer segments was associated with rod function on ERG in 5 of 9 eyes tested as evidenced by development of a scotopic threshold response and alterations in the timing of the b-wave. These same five treated eyes had improved dim-light vision compared to the control eyes. The scotopic threshold response is a rod mediated response (despite the measured electrical signal originating in the inner retina) and appears at a stimulus luminance close to visual threshold (Sieving et al., 1986; Naarendorp and Sieving, 1991). The slow negative STR waveform initially increases in amplitude with increasing stimulus luminance but becomes obscured by the incursion of the b-wave. Previous studies in models have suggested that the STR threshold tracks remaining functional vision better than the ERG b-wave (Bush et al., 1995) The scotopic b-wave in the eyes that showed rescue had a longer implicit time, more in keeping with a rod-response, although the flash luminance needed to induce an ERG b-wave was not reduced by treatment. Previous ERG studies in dogs have shown that the rod-mediated b-wave is a slower waveform than the contribution of dark-adapted cones to the scotopic responses to stronger stimuli that in normal dogs result in a mixed rod/cone response (Narfström et al., 1995) Pde6a mutant dogs that have not been treated, have no evidence of rod ERG responses and exhibit cone-only function (Tuntivanich et al., 2009). The small delayed (compared to control eyes), dark-adapted b-wave in treated eyes reflects a low level of functional rod rescue even allowing for the fact that only a proportion of the retina was treated (blebs typically extended over one-fifth to one-third of the retina). The dark-adapted b-wave amplitudes recorded might also be low because (1) only a proportion of rods in the treated area are functional, or (2) the transduced rods have short outer segments reducing the individual response from each rod, or (3) rod bipolar cell dendritic connections to rods may have been reduced (as reported in other PDE6 models; Strettoi and Pignatelli, 2000), or a combination of these possibilities. The increase in dark-adapted b-wave implicit time combined with the presence of an STR, improved dim light vision, Pde6 localization to outer segments on IHC and photoreceptor structural preservation, provides convincing evidence of AAV-Pde6a mediated rod rescue.

Improvement in visual performance is reported to be a more sensitive measure of visual function than ERG. For example, in RPE65 gene augmentation therapy for Leber Congenital Amaurosis improved visual function can be detected in the absence of an ERG improvement both in human patients in phase I/II clinical trials, but also in the dog model when receiving a low dose of therapeutic vector (Bainbridge et al., 2008, 2015; Hauswirth et al., 2008; Maguire et al., 2008). This difference between the sensitivity of visual function testing and ERG likely results from a difference in the number of functional photoreceptors that are required for improvement in vision vs. the number required to create a large enough electrical signal to be detectable by a corneal electrode. Transgene mediated expression of Pde6 was sufficient to slow the typically very rapid loss of photoreceptors in the Pde6a mutant dog. Pde6 that formed in outer segments in the treated region was able to hydrolyze cGMP as evidenced by the large reduction of cGMP detectable by IHC restricted to the retinal regions that had Pde6 expression. Improved trafficking of rhodopsin to the outer segments in treated retinal regions largely reversed the mislocalization that occurred in untreated and control treated retinal regions. Rhodopsin mislocalization is a common finding in diseased and stressed rod photoreceptors (see Hollingsworth and Gross, 2012 for a review). Structural preservation of the outer nuclear layer was detectable both on SD-OCT, where it corresponded with the topography of the treated region, and in retinal sections where it corresponded with the region of increased Pde6 expression. There also appeared to be a positive effect on rod to bipolar cell synapses. In untreated retinal regions rod bipolar cell dendritic retraction was apparent. Treatment appeared to reverse, or prevent this change.

These results show that some rescue of rod function and rod preservation is possible in this severe retinal degeneration model. Previously we had not been able to achieve rescue using AAV5 vectors and hypothesized that the faster transduction rate achieved by AAV8 vector with a capsid mutation might allow us to achieve rescue, i.e., AAV8(Y733F). The greater success with an AAV8(Y733F) construct over an AAV5 serotype delivering the same construct (Petersen-Jones, unpublished findings) is in contrast to the findings in the Pde6b dog, which also has a rapid photoreceptor degeneration, where results with an AAV5 vector were similar to that achieved with AAV8 (Petit et al., 2012). Conversely in rodent studies the stronger expression achieved by an AAV8 vector and an AAV8(Y733F) over AAV5 has been shown to allow for rescue of some of the more severe photoreceptor degeneration phenotypes such as the Aipl1^−/− mouse model, that has a very rapid retinal degeneration (Sun et al., 2010) and also the Pde6b^rd10 mouse model which has a slower photoreceptor degeneration (Pang et al., 2011). We had also previously shown in a reporter gene study that an AAV8(Y733F) vector had a stronger and faster onset of reporter gene expression compared to AAV5 vectors when delivered by subretinal injection in the dog (Mowat et al., 2014).

Pde6a mutant dogs, in addition to lacking detectable Pde6a on Western blot analysis of pre-degenerate retinal tissue, also lack the other Pde6 subunits (Pde6b and Pde6g) (Tuntivanich et al., 2009). This is in contrast to the findings in Pde6b mutant dogs where the other Pde6 subunits were detectable on Western blotting of retinal tissue (Suber et al., 1993). It is not clear whether the lack of the other Pde6 subunits in Pde6a mutant dogs explains the greater difficulty we have faced in achieving rescue compared with the results reported for the Pde6b mutant dog (Petit et al., 2012; Pichard et al., 2016).

Based on these promising preliminary gene augmentation therapy results using AAV8(Y733F) to deliver Pde6a under control of a ubiquitous promoter further studies are warranted to optimize the degree of rescue and minimize adverse effects. This could include optimization of construct with respect to promoter to specifically target rods and possibly other modifications such as addition of an enhancer element or mini-intron. The Pde6a cDNA is too large to allow packaging in a self-complementary vector which could otherwise hasten the onset of expression (Petersen-Jones et al., 2009). A dose escalation study is also needed to establish the optimal therapeutic dose. It is conceivable that the variability of the degree of rescue achieved could mean that the current dose is on the edge of the most effective range, either giving too little expression, meaning that the threshold for rescue was not achieved in each eye, or too high a level of expression resulting in a range of adverse effects. With a condition with such a rapid degeneration the timing of the injection is also likely to be important and that to improve results younger animals with a larger number of surviving rods should be targeted. However, in our study more consistent results were achieved by injecting slightly older animals. Visualization in younger animals during the injection can be challenging because the ocular media is not always completely clear and remnants of development vasculature (pupillary membrane, tunica vasculosa lentis, and hyaloid vasculature) are often present. Thus, obtaining a good subretinal bleb in younger animals presents a technical challenge thus limiting the lower end of the age range that can be treated efficiently. In mouse Pde6 models, introduction of expression of the missing Pde6 subunit achieved rescue even with more advanced loss of photoreceptors (Wert et al., 2013; Koch et al., 2015). Although unlike the Pde6a mutant dog neither model had a complete absence of Pde6 in the pre-degenerate retina.

Limitations of this study include the necessarily small sample size due to utilization of a large animal model species. Statistical evaluation was therefore limited in many comparisons, particularly as half of the animals underwent no treatment in the contralateral control eye, the other half of the animals were injected with a control vector, more suitable to compare to the treated AAV-Pde6a treated eyes, as the effects of injection are controlled. We chose both untreated and control vector treated controls to allow for possible adverse effects of subretinal injection and expression of the GFP reporter gene and conversely possible trophic factor release due to the same interference. The titer of AAV-GFP used was one that would not be expected to lead to loss of photoreceptors due to high expression levels of GFP. Due to limitations of equipment availability in the laboratory at the time of the experiments, certain studies could not be performed on both cohorts of animals limiting the full evaluation of outcome of therapy. Clearly, further work is necessary to expand on these promising initial results aiming to optimize treatment efficacy, and future experiments would consider these factors.

The gene augmentation therapy in this study was not without complications. Adverse retinal changes occurred with both the AAV-Pde6a and the AAV-GFP vectors. In some eyes this resulted in focal lesions of severe retinal thinning within the injected region. These degenerate lesions occurred directly adjacent to other retinal areas where transgene expression was present and where in AAV-Pde6a injected eyes but not the AAV-GFP injected eyes there was retinal preservation. Retinal folds and rosettes were also apparent in some eyes. Retinal lesions and inflammation following subretinal AAV vectors in dogs has previously been described. Bainbridge et al., reported that 3 out of 8 dogs that were treated by subretinal injection of an AAV2 construct expressing GFP under control of a CMV promoter developed delayed ocular inflammation (Bainbridge et al., 2003). Beltran et al found that some dogs developed patches of retinal degeneration and more severe inflammatory response following subretinal administration of an AAV5 expressing GFP under different promoters (Beltran et al., 2010). In our study, no signs of retinal inflammation were detected in any eye at any timepoint, either in vivo or histologically. In the current study dogs were treated prior to retinal maturation, which in the dog is considered to be complete at about 7–8 weeks of age (Gum et al., 1984). It is not clear if this might be a factor in rosette formation or whether that change might be associated with the retinal detachment that is induced by the subretinal injection. Similar rosettes and areas of retinal thinning were reported in a safety study of an AAV-RPE65 vector in adult dogs. The effect was dose related, with development of lesions occurring in eyes injected with a higher vector dose (Jacobson et al., 2006). A similar gene therapy study in which Pde6b mutant dogs were injected at 20 days of age with AAV vectors expressing Pde6b under control of a tissue specific promoter (Rhodopsin Kinase) to target photoreceptors did not report similar rosette formation (Petit et al., 2012). In contrast, our study used a ubiquitous promoter so off target effects are a possibility, although we did not identify Pde6 protein presence outside of the rod outer segments. It is important to note that within the treated retinal regions the retina adjacent to the degenerate lesions showed Pde6 localization to the outer segments and preservation of the outer nuclear layer. The process of performing a subretinal injection causes a temporary retinal detachment which can result in deleterious effects (Cook et al., 1995; Fisher et al., 2001; Fisher and Lewis, 2003). These could potentially be even more damaging in a diseased retina. Complications potentially associated with subretinal injections have also been reported in RPE65 clinical trials, including foveal thinning and development of a macular hole (Maguire et al., 2008; Jacobson et al., 2012).

Here we report that AAV-mediated gene augmentation in a large animal model of rapid onset arRP caused by a Pde6a mutation results in Pde6 expression in rod photoreceptor outer segments restores a degree of rod photoreceptor function, delays photoreceptor degeneration, and reduces rhodopsin mislocalization, GFAP activation and bipolar cell dendrite retraction. Future studies will be designed to improve the degree and duration of rescue and to avoid adverse treatment effects that were associated with the therapy. These additional studies will be important prior to translation of this early proof-of-concept functional success with the Pde6a mutant dog model into clinical trials for human arRP.

FM: Evaluation of retinal samples and immunohistochemistry, data interpretation and wrote the paper. LO: SD-OCT analysis and some IHC, review of paper. JB: Helped in design, assisted in subretinal injections, critical review of manuscript. KG: Performed outcome measures and data analysis, reviewed manuscript. AB: Measurement of retinal sections and cells, reviewed manuscript. JQ: Assisted in all animal experiments and performed ERG and vision testing. AD: Design and production of viral vectors. SB: Design of vector approach and production of viral vectors, critical review of manuscript. WH: Design of vectors, production of viral vectors, critical review of manuscript. SP: Designed and oversaw experiments, subretinal injections, co-wrote the manuscript.