Superior Retinal Gene Transfer and Biodistribution Profile of Subretinal Versus Intravitreal Delivery of AAV8 in Nonhuman Primates

Abstract

Purpose: To investigate shedding and biodistribution characteristics of recombinant adeno-associated virus serotype 8 (rAAV8) after single-dose subretinal or intravitreal injection in nonhuman primates (NHP, Macaca fascicularis) as a surrogate for environmental hazard and patient safety.

Methods: In a study for regulatory submission, 22 NHP were divided into four cohorts receiving either single subretinal injections of vehicle or clinical grade rAAV8 (1 × 10¹¹ or 1 × 10¹² vector genomes [vg]) versus single intravitreal application of 1 × 10¹² vg. Viral shedding and biodistribution were monitored in biofluids for up to 91 days, followed by necropsy and tissue harvesting of all major organs, the visual pathway, and lymphatic tissue. Quantification of vector genomes was done by quantitative (q)PCR.

Results: Shedding occurred in a dose-dependent manner in all biofluids and persisted for a maximum of 7 days. Intravitreal delivery led to increased and persistent (up to 13 weeks) distribution of vector genomes in blood and draining lymphatic tissue, increased off-target deposition, and inefficient gene transfer to the retina. No vector targeting of the germ line was observed in any cohort.

Conclusions: These data illustrate that subretinal application of rAAV8 leads to a more favorable biodistribution profile compared to intravitreal injections. Extraocular biodistribution is limited after subretinal delivery, while intravitreal injection leads to both greater and more persistent systemic exposure, evident in blood and lymphatic tissues. With the knowledge on the dynamics of shedding in a setting mimicking clinical application, guidelines can be developed to refine clinical trial protocols to reduce the risk for trial subjects and their environment.

This study was performed to meet the requirements for a regulatory submission of a “first in man” clinical trial, testing an investigational new drug (IND) for the treatment of achromatopsia (ACHM). ACHM is a hereditary retinal disorder with an estimated prevalence of 1:30,000¹ for which currently no treatment exists. The prevalence is greater in certain genetically isolated areas like the Micronesian atoll Pingelap, where up to 6% of the population are affected by the condition.² ACHM is a congenital disease that exclusively impairs cone function. Affected individuals suffer pronounced photopic defects such as total color blindness, reduced visual acuity, hemeralopia (day blindness), nystagmus, and photophobia.³ To date, six genes have been identified that link to ACHM. Five of them constitute essential parts of the cone phototransduction cascade as they encode either for cyclic nucleotide–gated channels (CNGA3⁴/CNGB3⁵), phosphodiesterases (PDE6C⁶/PDE6H⁷), or a related G protein (GNAT2⁸). The sixth gene is AFT6. It encodes a protein that is part of the unfolded protein response in the endoplasmic reticulum.⁹ At the University Eye Hospital Tübingen, we performed the first clinical gene therapy for ACHM caused by mutations in CNGA3 using subretinal administration of rAAV8.hCNGA3 (NCT02610582) in 2015. Here, we describe findings from the preclinical biodistribution study, which evaluated safety aspects including the extent of biodistribution within the visual system, the extraocular tissues (e.g., germ line), and shedding of any genetically modified organism (GMO) into the environment after subretinal versus intravitreal injections. Cynomolgus monkeys were chosen because their ocular anatomy is rather humanlike (i.e., presence of a macula) and suited to mimic surgical procedures used in patients. These data are expected to be relevant to all ocular gene therapies utilizing the adeno-associated virus serotype 8 (AAV8) capsid and possibly other serotypes with similar tropism, as the affinity of capsid surface epitopes toward local receptors helps to define their distribution pattern. Recent evidence of innate and adaptive immunity toward AAV epitopes increases the significance of these findings.^10,11 

Materials and Methods

Cynomolgus monkeys (nonhuman primates, NHP) were treated and cared for at the Covance Preclinical Services GmbH test facility in Muenster, Germany. The study was conducted with great care to ensure the well-being of the animals and was approved by the local authorities (Regierungspraesidium of North-Rhine Westphalia). All animal procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and in full compliance with the guidelines of the European Community (EUVD 86/609/EEC) for the care and use of laboratory animals, as well as in accordance with Good Laboratory Practice (GLP) standards as defined by German GLP monitoring authorities and in compliance with U.S. Food and Drug Administration and Good Laboratory Practice regulations. 

Animals

A total of 22 NHP were assigned to four study cohorts. Six animals (three of each sex) were assigned to cohorts 1 to 3 and treated with subretinal injections (SR). Cohort 1 received vehicle only: balanced salt solution, BSS (Alcon Laboratories [UK] Ltd., Camberley, UK) with 0.001% PF-68 surfactant (BASF, Ludwigshafen, Germany). Cohorts 2 and 3 received single SR of either 1 × 10¹¹ vector genomes (vg) (low dose) or 1 × 10¹² vg (high dose), respectively. Cohort 4 consisted of four animals (two of each sex) and received single intravitreal injections (IVT) of 1 × 10¹² vg (high dose) to mimic leakage from the subretinal space (e.g., through retinotomy or macular hole) or via falsa injection. 

Production of Recombinant AAV8

Recombinant AAV8 carrying a transgene cassette based on an AAV2 genome (pseudotype AAV2/8) was manufactured according to good manufacturing practice (GMP) guidelines. It contained a cone-arrestin 3 promoter driving CNGA3 expression that has been shown to have therapeutic effect in the Cnga3^−/− mouse model.¹² GMP grade cis and trans plasmid DNA was provided by Aldevron (Fargo, ND, USA) using a high-quality characterized Escherichia coli master cell bank (MCB), and Atlantic BioGMP (Nantes, France) produced the GMP grade viral particles (rAAV.hCNGA3) utilizing a transient double-transfection protocol of an HEK293 MCB fully characterized according to the European Pharmacopeia. Harvested cells were lysed and supernatant PEG-precipitated, treated with benzonase, and purified by two rounds of caesium chloride gradient ultracentrifugation followed by a tangential flow filtration step for diafiltration and concentration. After formulation, the resulting drug substance was stored at ≤−70°C until application. 

Surgery and Postsurgical Care

Animals received general volatile anesthesia with isoflurane; (peri-)orbital regions were treated with 10% povidine iodine solution and sterile surgical drapes applied as in the clinical setting. A temporal canthotomy was performed for improved access, and three transconjunctival sclerotomies were made after transillumination confirmed the location of pars plana. Vitrectomy was performed as completely as possible without affecting the lens. A localized retinal detachment was induced through SR of 50 μL BSS (Alcon Laboratories) using a 41-gauge cannula (DORC 1270.EXT; Dutch Ophthalmic Research Center [International], B.V., Zuidland, The Netherlands). Virus solution (200 μL) was injected into the preformed bleb using a foot pedal–controlled injection system (PentaSys II; Ruck GmbH, Eschweiler, North Rhine-Westphalia, Germany). Before recovery, subconjunctival cefuroxime (125 mg; Ratiopharm GmbH, Ulm, Baden-Württemberg, Germany) and dexamethasone (2 mg; Ratiopharm Gmbh) were administered to the operated eye. Postoperative prophylactic treatment consisted of antibiotic (0.5% moxifloxacine; Ratiopharm GmbH), and anti-phlogistic (1% prednisolone; Ratiopharm GmbH) eye drops given 3×/day each in the treated eye for 2 weeks and prednisone (Ratiopharm GmbH) 1 mg/kg intramuscularly from day −2 until day 5. 

Biofluid and Tissue Sampling

Biofluid samples were harvested from all animals before dosing and on days 2, 3, 5, 7, and 31 for quantitative (q)PCR analysis. Blood was collected before dosing and 24, 48, 72, and 168 hours and at weeks 4 and 13 post dosing. Predose samples and samples collected on days 2 and 3 were analyzed in the first instance. The remaining samples were analyzed until there were two consecutive time points that were negative for vector genomes. Sample volumes used for analyses were 220 μL for blood, tear film, and nasal secretions, 100 μL for urine, and 50 μL for aqueous humor. For large organs, tissue samples were harvested at necropsy (91 days post dosing) with sterile, DNase-free single-use instruments and stored at <−70°C until DNA extraction. Where reported, whole organ weights, as well as individual sample weights, were documented. This enabled extrapolation of copy numbers per mg sample to copy numbers per (whole) organ. 

DNA Extraction and qPCR

DNA was extracted from all tissue and blood samples prior to qPCR using QIAsymphony DSP DNA Mini kit with Qiagen Reagent DX, and the QIAsymphony DSP DNA Mini kit, respectively (Qiagen, Hilden, North Rhine-Westphalia, Germany). DNA concentrations were measured using spectrophotometry. Eight cryosections (20 μm thick) from whole eye cups were processed for each animal and pooled into one sample for extraction. Also, for heart, lung, and liver, three samples were extracted separately and the eluates pooled to result in one sample analyzed by qPCR. For all tissue samples and blood, 1 μg DNA was used per qPCR reaction. QIAsymphony DSP Virus/Pathogen kit for biofluids was used for the remaining samples. DNA concentration of these samples was not determined, due to the inclusion of carrier RNA in the extraction procedure. Here, samples of 5 μL were analyzed neat by qPCR. For presentation, results are normalized to 1 mL source material to improve interpretability. To validate the qPCR assay, a positive control (transgene plasmid) was serially diluted from 5 × 10⁷ to 50 copies/reaction. The data generated from this dilution series were used to construct a standard curve for quantitative data analysis. Samples that tested positive after 40 cycles of amplification, but below 50 copies, were supposed to contain a nonquantifiable number of genomes per reaction (between 1 and 50), and thus deemed “<LLOQ.” Samples without amplification after 40 cycles were deemed “negative.” This validated qPCR assay was used to analyze each sample in triplicate, with one of the replicates spiked with a known quantity of the qPCR positive control to assess for the presence of any potential inhibitors in the sample. 

Results

Animal Dosing

All randomly assigned animals successfully underwent dosing per protocol within 3 consecutive days. Complete subretinal placement was evident in direct visualization through the operating microscope and involved the targeted macular area in all animals. All biofluid and tissue sampling could be performed according to standard operating procedures and in strict accordance with GLP guidelines. 

Animals Injected With Vehicle Only

All tissue and biofluid samples taken from animals in the control cohort injected with vehicle tested negative for presence of transgenic DNA, with exception of one tear sample and two nasal swabs taken from one female animal. In these samples, transgenic DNA quantities detected were all below the lower level of quantification (LLOQ = 50 copies per reaction). All consecutive tear samples and nasal swabs from this animal tested negative, as did all other biofluid and/or tissues samples at all time points. 

Animals Injected With Vector Solution

Shedding occurred to some degree in all animals injected with vector solution and in all types of biofluids tested. An overview of maximal amount of detected vector genomes/mL biofluid and duration of shedding by route of administration and dose is given in Table 1. Of the biofluids that shed into the environment (tears, nasal secretions, and urine) no sample tested above the LLOQ later than day 7. Maximal shedding in most animals was observed 2 days after injection. After SR, shedding was most pronounced in tear fluid, followed by nasal secretions and urine. After IVT, the largest amount of vector was also shed via tears, but was followed by urine and lastly nasal secretions. Table 1 presents time points when shedding of all samples of a cohort were below LLOQ (residual shedding), or below level of detection (no shedding). Figure 1 summarizes the qualitative changes of shedding status (positive, residual, or negative) per cohort and biofluid over the observed time frame. To characterize time dependency it introduces T_1/2 as a semiquantitative “shedding status half-life”: the point in time (in days) after administration on which ≥50% of a cohort's samples test below LLOQ. Individual shedding results of all animals are shown in Figures 2 to 4, plotted in log scale against time of harvest, capturing both the variability between animals of the same cohort and differences between cohorts. 

Table 1

 

Surrogates Relevant to Clinical Practice

 

Figure 1

 

Shedding prevalence per cohort and biofluid during the observed time frame. Switching between columns compares results of the same sample type (i.e., shedding via tear) over different cohorts (i.e., subretinal low versus high dose), while switching between rows compares different sample types in the same cohort. The most notable difference is visible in the bottom row of the graph, displaying vector presence in blood across the three cohorts. y-axes: sample status in % of animals in the given cohort, color-coded. x-axes: days after injection. T_1/2: ≥50% of the cohort's samples test below LLOQ or negative. Individual sample status is color-coded.

Figure 2

 

Vector genomes in tear film detected by qPCR plotted against time of harvest. Cohort means per time point indicated by colored lines. Of note, the high-dose subretinal cohort mean on day 3 was skewed by a single animal. It tested with 4.6 × 10 E+7 vg/mL on day 3, 53 times more than it tested on day 2, and 33 times more than it tested on day 4. The measurement was not omitted, although such variation was not seen in any other instance across the study. LLOQ = lower limit of quantification (50 copies/reaction).

Figure 3

 

Vector genomes in nasal secretions detected by qPCR plotted against time of harvest. Dose escalation increased copy numbers and shedding duration. Intravitreal administration led to fewer copy numbers, but similar prevalence and duration compared to subretinal administration of the same dose. Cohort means per time point indicated by colored lines. LLOQ = lower limit of quantification (50 copies/reaction).

Figure 4

 

Vector genomes detected in urine by qPCR plotted against time of harvest. In the low-dose subretinal cohort, no relevant shedding was detected. Both high-dose cohorts displayed similar overall shedding via urine. Cohort means per time point indicated by colored lines. LLOQ = lower limit of quantification (50 copies/reaction).

Individual Lacrimal Shedding Results

Shedding via tear film was subject to a pronounced dose effect (Figs. 1, 2; Table 1). In animals from the low-dose group, fewer copies were found in positive samples, and tears were negative after a shorter time, compared to the high-dose group. Subretinal application of the high dose led to a 33-fold increase in maximal shedding (cohort mean) compared to the low dose, and shedding persisted at quantifiable levels for at least 1 week after surgery. Route of administration also had distinct effects on shedding via tears. While fewer copies were detected following intravitreal compared to SR (with the same amount of vector applied), both groups featured quantifiable samples after 1 week. Where only four out of six animals from the low-dose group shed vector via tears at any given time point, all (6/6) animals from the high-dose subretinal and (4/4) intravitreal cohorts shed vector via tears over the course of the study. Individual results are displayed in Figure 2 and copy numbers/mL are given in Supplementary Material S1. 

Individual Shedding Results via Nasal Secretions

Shedding via nasal secretions was less dependent on dosage than shedding through tears and urine (Figs. 1, 3; Table 1). Shedding in the low-dose subretinal cohort receded more quickly (T_1/2 after 4 vs. 7 days for high dose) and was slightly less prevalent (5/6 animals), compared to the high-dose cohort (6/6). In line with the slightly increased prevalence after high-dose injection, dose escalation by one log unit also entailed a modest 1.4-fold increase in maximal copy numbers (cohort mean). Like dose escalation, route of delivery had only a minor effect on shedding via nasal secretions. Between subretinal and intravitreal cohorts of the same dose, there was an equal prevalence (all animals in both groups) and duration of shedding, with 50% of animals in both cohorts featuring quantifiable qPCR results 7 days after surgery. A slight difference between the high-dose groups was found in a 1.7-fold elevation of copy numbers after SR, relative to intravitreal administration. Individual results are displayed in Figure 3 and copy numbers/mL are given in Supplementary Material S1. 

Individual Shedding Results via Urine

Shedding via urine was strongly dose dependent (Figs. 1, 4; Table 1). Foremost, it was barely detectable and nonpersistent in the low-dose cohort. Due to the low-dose cohort's low baseline, dose escalation by one log unit led to an 18.5-fold increase in maximal shedding (cohort mean) compared high-dose SR. 

In contrast to the clear dose dependency, route of administration had subtler effects. Shedding prevalence was similar between IVT (4/4) and SR (5/6). Between these high-dose groups, copy numbers were moderately increased (2.7-fold) in subretinally injected animals, while shedding above LLOQ lasted 2 days longer after IVT. Individual results are displayed in Figure 4 and copy numbers/mL are given in Supplementary Material S1. 

Transduction of the Visual System

At necropsy (91 days post injection), tissues of the visual pathway (Fig. 5; Table 2) were collected for formal biodistribution analyses. The segments of the visual system were sampled individually along the pathway taken by light (anterior segment, retina, optic nerve and chiasm, lateral geniculate nucleus [LGN], and visual cortex). Presence of vector in the aqueous humor was exclusively dependent on route of administration, while upstream distribution along the visual pathway was strongly associated with both dose and route of administration. 

Figure 5

 

Individual visual pathway sample results (day 91) of all cohorts plotted anterior to posterior. Both aqueous humor deposition after IVT and retinal deposition after SR administration were highly uniform across their respective cohorts. Individual samples further up the optic pathway tested above LLOQ, with a single sample being positive at the level of the lateral geniculate nucleus. Cohort mean per segment indicated by colored lines. AH, aqueous humor; ON, optic nerve; OC, optic chiasm; LGN, lateral geniculate nucleus; V₁, visual cortex; LLOQ = lower limit of quantification (50 copies). *LLOQ for AH = 2.2 × 10 E+4 copies/mL sample.

Table 2

 

Vector Deposition Along the Visual Pathway in vg/μg Extracted DNA After 91 Days. Vector Deposition, Visual Pathway, and Vector Genomes per mL for Aqueous Humor

 

Ninety-one days after surgery, only one aqueous humor sample across both subretinal cohorts exhibited vector presence, while all samples in the intravitreal cohort were clearly above LLOQ. This was reversed within the eye, where compared to IVT, subretinal application of the same dose resulted in 53 times higher copy numbers in whole retinal sections. A 10-fold dose escalation in SR caused a 37-fold increase of copy numbers detected in retinal tissue. The respective presence of vector in the retina carried over to following stations along the visual pathway. In optic nerve samples, dose escalation led to a 20-fold increase in copy numbers, while compared to IVT, SR resulted in 2.9-fold increased copy numbers. In line with the difference in copy numbers, five of six animals demonstrated vector presence in optic nerve samples after subretinal high-dose injections, as opposed to two of six animals in the low-dose cohort. A similar pattern, although with markedly lower copy numbers, also carried over to the optic chiasm, LGN, and visual cortex. Only one sample in the study (high-dose SR cohort) tested above the LLOQ at the level of the LGN, while not a single sample from the visual cortex tested positive for vector DNA. Table 2 shows copy numbers detected in each segment. Figure 5 illustrates the distribution of vector along the visual pathway in each cohort. 

Biodistribution Outside the Visual System

The study also analyzed samples from blood (Figs. 1, 6; Table 1), major organs, and draining lymphatic tissue. Biodistribution in the systemic circulation and lymphatic tissues was strikingly dependent on the route of administration, but barely dose dependent. For example, while dosage had a moderate effect on biodistribution via blood, injecting intravitreally resulted in a 464-fold increased blood vector load compared to SR of the same dose, and vector presence in blood persisted in all animals of the IVT cohort until 91 days. Individual results are displayed in Figure 6 and copy numbers/mL are given in Supplementary Material S1. 

Figure 6

 

Vector genomes in blood detected by qPCR plotted against time of harvest. Blood samples did not show any vector genomes after low-dose subretinal injection, and only limited vector load after high-dose subretinal injection. Intravitreal administration led to immediate, strong, and persistent presence of vector in the blood. Cohort means per time point indicated by colored lines. 1 μg DNA = 40.1 ± 17.8 μL blood (mean ± SD). LLOQ = lower limit of quantification (50 copies).

Large Organs and Lymphatic Tissue

In line with elevated vector load in blood after IVT, draining lymph node and spleen contained significant amounts of vector genomes in animals of the IVT cohort, but virtually no copies after subretinal delivery (Figs. 7, 8). Relative amounts of vector genomes found in the lymph nodes (Table 3) after IVT were increased around 430-fold for deep cervical, 1000-fold for retropharyngeal, 510-fold for mandibular, and 490-fold for mesenteric lymph nodes, compared to SR. IVT led to an approximately 7400-fold increase in copy numbers in spleen compared to SR. Liver samples were also positive in the intravitreal cohort, while heart, tongue, spinal cord, adrenal glands, and, importantly, the gonads were free of vector genomes in all animals of all cohorts (Table 4). 

Figure 7

 

Individual internal large organ sample results of all cohorts. Off-target transduction of spleen and liver occurred to a similar degree in all animals from the intravitreal cohort. Corresponding tissue amounts are given in Supplementary Material S2. LLOQ = lower limit of quantification (50 copies).

Figure 8

 

Individual lymph node sample results of all cohorts. Fourteen of 16 lymph node samples from the intravitreal group feature a noticeable degree of vector deposition. DC, deep cervical; RP, retropharyngeal; MA, mandibular; ME, mesenteric. Corresponding tissue amounts are given in Supplementary Material S2. LLOQ = lower limit of quantification (50 copies).

Table 3

 

Vector Deposition in Lymph Nodes Given as vg per Organ After 91 Days

 

Table 4

 

Vector Deposition in Large Organs Given as vg per Organ After 91 Days

 

Summary

The study was designed to explore the shedding and biodistribution characteristics after subretinal delivery of AAV8 when mimicking the clinical scenario in patients. The IVT control group was added to test effects of inadvertent primary injection into the vitreous (via falsa) and/or a secondary delivery into the vitreous cavity through a retinal tear (e.g., macular hole formation) in the context of subretinal delivery. Both dose and route of delivery change the distribution profile, and Figure 9 shows a proposed model based on the major finding that intravitreal placement of AAV8 results in much higher viremia and contact with immune-competent cells. It highlights the potential of immune-competent effector cells to influence local reactions to the viral vector in the eye. 

Figure 9

 

Summary and proposed model of biodistribution following intravitreal administration of AAV8. Intravitreal injection or retinal tear after subretinal injection leads to (1) persistent presence of vector in the aqueous humor and anterior chamber with (2) access to the venous system through Schlemm's canal. This study demonstrated long-term persistence of (3) vector in the blood and (4) deposition in lymphatic organs. The resulting prolonged and close contact with the immune system (5) might help to explain (6) acute or delayed inflammation observed after gene therapy.

Discussion

To the best of our knowledge, this study reports the most comprehensive data on viral shedding and biodistribution of recombinant adeno-associated virus serotype 8 (rAAV8) after subretinal versus intravitreal injection (IVT) in nonhuman primates. This study was conducted in strict adherence to GLP guidelines and designed to generate data on biodistribution/shedding after subretinal and IVT in a protocol mimicking clinical application of rAAV8 gene therapy as closely as possible. In contrast to the protocol of the clinical trial, sclerotomies were not sutured in primates with the aim to reduce postsurgical irritation. However, this did not prevent digital manipulation of the treated eyes by the animals once transferred back to the cages. This may have affected biodistribution/shedding outcomes, and any such manipulation can easily be avoided in the clinical setting. 

One important component of patient safety is avoidance of germ line transduction.^13,14 There was no vector presence in the germ line, regardless of sex, vector dosage, or route of administration. Shedding of GMO into the environment is another important aspect, and our data show that shedding to the environment receded below LLOQ by 1 week after surgery, with 21% of samples without detectable vector genomes and 79% of all samples testing below LLOQ. In line with other reports, vector biodistribution and consequent deposition in internal organs were minimal in the subretinal cohorts.¹⁵ 

Both dose (1E+11 vs. 1E+12) and delivery (subretinal versus intravitreal) had an impact on biodistribution and shedding. Reducing the dose by one log unit resulted in 30% to 97% fewer copies off-target and shorter duration of shedding. Retinal gene transfer to the retina after IVT was at least one order of magnitude less compared to SR of the same vector dose. Vector genomes detected on day 1 in the intravitreal cohort's bloodstream accounted for 0.26% of the total dose applied, in contrast to 0.0006% after subretinal delivery of the same dose. This is in line with IVT leading to increased shedding into the systemic circulation via the anterior chamber. The observation that mesenteric lymph nodes, which do not drain from the injection site, did show a similar degree of vector deposition as deep cervical, retropharyngeal, and mandibular lymph nodes, also supports the notion of circulation-mediated biodistribution after IVT. Taken together, this evidence suggests a biodistribution profile not unlike what is described for intramuscular and intravenous injections.^16,17 Such a systemic visibility after IVT may promote vector recognition by the immune system and might help to explain any form of immune response after gene therapy.^11,18,19 Recent findings complement this hypothesis by showing a qualitative difference in immune response after IVT.¹⁰ 

It is important to note, however, that the absolute risk of any clinically relevant systemic toxicity in a patient after a retinal tear during subretinal delivery or IVT delivery can still be managed effectively. This is evident from clinical trials, where much higher dosages of rAAV8 were delivered primarily into the systemic circulation without causing serious adverse events.²⁰ However, these trials have shown that systemic exposure to significant numbers of any rAAV serotype can induce humoral and cellular immune response followed by directed removal of successfully transduced target cells by cytotoxic T lymphocytes.²¹ 

Previous studies in large animal models have demonstrated that rAAV8 gene therapy has a favorable safety profle²² and efficiently transduces photoreceptors,^23,24 and that vector genome can be found along the visual pathway after subretinal delivery of rAAV.^25,26 Our data are in line with these reports and show that dose and route of application are important determinants of the extent of this distribution pattern.²⁷ While our data do not explain the mechanism of transduction along the visual pathway, they clearly show that both routes of application can lead to anterograde vector genome distribution along the axons of the ganglion cells up to the LGN. Intriguingly, this would indicate that subretinal rAAV8 does transduce ganglion cells. Alternative explanations would involve trans-synaptic mobility and/or vector solution traveling in the subretinal space to reach the optic nerve sheath and access to the cortex, for example, via cerebrospinal fluid. Both seem rather unlikely scenarios and are not supported by the fact that the visual cortex is free from vector genomes. 

When taking all biodistribution and shedding data into consideration, we argue that SR offers the more favorable set of results. With constant improvements of subretinal surgery,^28,29 and the predominant intention of treating the retinal pigment epithelium and photoreceptors, we conclude that SR is therefore the preferable procedure in most current ocular gene therapy scenarios. This may of course change with the advent of new viral vectors^30–35 and improvements in intravitreal surgery,^36,37 which promise to efficiently transduce cells of the outer retina after intravitreal delivery. 

While these results were generated using rAAV8, there is reason to believe that the mechanics underlying shedding and biodistribution are also applicable to other AAV serotypes with similar tropism. As most transgene cassettes in classic gene augmentation strategies feature target cell–specific promoters, off-target transduction with nonintegrating vectors such as AAV can still be regarded as fairly safe—especially as the germ line is shown not to be affected. However, in approaches using either ubiquitous promoters or immunogenic bacterial enzymes (e.g., CRISPR-Cas9) to edit genomic DNA with less than 100% specificity, off-target transduction may be more of a concern.³⁸ 

Acknowledgments

The authors thank Apostolos Bezirgiannidis (University Eye Hospital Tübingen), Sven Korte, and Jörg Luft (both Covance Laboratories GmbH) for their help in the animal study and Daniel Pauleikhoff (Augenzentrum St. Franziskus, Münster) for the supply of 41-gauge needles on very short notice. 

Supported by Tistou und Charlotte Kerstan Stiftung. 

Disclosure: I.P. Seitz, None; S. Michalakis, P; B. Wilhelm, None; F.F. Reichel, None; G.A. Ochakovski, None; E. Zrenner, None; M. Ueffing, None; M. Biel, P; B. Wissinger, None; K.U. Bartz-Schmidt, None; T. Peters, None; M.D. Fischer, Nightstar Ltd (C), Eyeserve GmbH (C), Regenxbio, Inc. (C), Bayer (R), Novartis (R) 

Appendix

RD-CURE Consortium: Bernd Wissinger; Martin Biel; Eberhart Zrenner; Karl Ulrich Bartz-Schmidt; M. Dominik Fischer; Susanne Kohl; Stylianos Michalakis; Francois Paquet-Durand; Tobias Peters; Mathias Seeliger; Marius Ueffing; Nicole Weisschuh; Barbara Wilhelm; Ditta Zobor; Stephen Tsang; Laura Kühlewein; Christian Johannes Gloeckner; Nadine A. Kahle.