Perineural delivery of AAV2/9 in non-human primates is a safe and efficient route for gene therapy in Charcot-Marie-Tooth diseases

Abstract

AAV-based gene therapy represents an attractive treatment for hereditary peripheral neuropathies Charcot-Marie-Tooth diseases. We recently showed that AAV2/9 vector-expressing GFP, locally injected into mouse and rat sciatic nerves, transduced a large amount of myelinating Schwann cells (mSCs). The local delivery of a similar vector expressing a small hairpin RNA (shRNA) targeting PMP22 mRNA prevented CMT1A disease in a rat disease model. Here, we investigated the regional anesthesia standard perineural injection route in non-human primates using AAV2/9 expressing GFP and AAV2/9 expressing shRNA targeting human and monkey PMP22 mRNA. Injecting at multiple sites to cover the largest part of arm and leg nerves, we found that AAV2/9 easily crossed perineurium sheaths to transduce up to 90% of mSCs. Injections on 10 nerves in each animal did not generate any adverse effects. Injected nerves functioned properly, and fine motor dexterity remained unaffected. Enzymatic, metabolite, and blood cytometry were marginally affected, and no systemic inflammation was detected. Locally, we detected a slight to moderate perineural infiltration of monocytes only with vector-expressing GFP. Vector-expressing PMP22 shRNA decreased PMP22 expression in a dose-dependent manner and levels remained physiological. Perineural injections appear to be safe, well-tolerated, and efficient to deliver AAV2/9-based gene therapy vector to treat Charcot-Marie-Tooth diseases.

Graphical abstract

Charcot-Marie-Tooth diseases are rare genetic disorders affecting peripheral nerves. Here, we provide a proof of principle in non-human primates that perineural delivery—the standard technique in regional anesthesia involving injection around nerves—may enable safe and efficient localized AAV9-based gene therapy targeting these conditions.

Introduction

Charcot-Marie-Tooth (CMT) diseases are a group of hereditary peripheral neuropathies that affect up to one in 2,500 persons.¹ They result from the length-dependent dysfunction of the fibers that cross peripheral nerves. While all fibers are defective, as defects occur all along the fiber, long fibers are more heavily impacted than shorter ones. These diseases are usually not lethal but are largely disabling in everyday life. The main symptoms of CMT are distal muscle weakness and waste, loss of motor dexterity, severe fatigue, walking problems occasionally leading to necessary wheelchair use, hands and feet deformities, and pain. Seventy percent of patients display a reduced nerve conduction velocity (NCV), indicating a demyelinating disease, while others show reduced amplitude of the compound of action potential (CMAP), indicating an axonal defect or a mix of both phenotypes.^2,3,4

Demyelinating CMT1A disease is the most common subtype of these diseases, affecting up to 30% of CMT patients. The genetic defect is a duplication of the PMP22 gene that leads to protein overexpression in myelinating Schwann cells (mSCs) that form myelin in peripheral nerves. This overexpression induces defects in myelination (dysmyelination) and eventually myelin sheath degeneration (demyelination).⁵

Several therapies and treatments are in development for this disease, including disease-modifying technologies that aim to decrease PMP22 expression in mSCs. A CRISPR-Cas9-based approach that targets the promoter region of PMP22 gene in a local administration has been described.⁶ Antisense oligonucleotides (ASOs)⁷ and small inhibitory RNAs (siRNAs)^8,9,10 specific for human PMP22 have been developed for systemic administration. A more sustainable expression of siRNA in target mSCs has been described using an AAV9 vector expressing a microRNA delivered intrathecally in rodents.¹¹ All these approaches have been successful in reducing PMP22 expression in mSCs in rodent nerves and bringing significant benefit to rodent models of the disease. This suggests that reducing PMP22 expression in these cells is both amenable and efficient to treat CMT1A in humans.

Recently, we investigated a gene therapy based on an AAV2/9 virus expressing a small hairpin RNA (shRNA) downregulating PMP22 expression in peripheral nerves.¹² Vector load was directly injected in both sciatic nerves of a rat model of CMT1A¹³ at young age, leading to a large transduction of mSCs.¹² Treating animals with this therapy decreased PMP22 protein and increased myelination to physiological levels. Treated animals displayed an NCV close to non-diseased animals and sensory-motor abilities were improved.¹² When injected into nerve tissue, the vector biodistribution was mostly limited to injected nerves and immune response remained low.¹² Therefore, nerve delivery of AAV2/9 vectors appears to be an attractive way to treat demyelinating CMT diseases.

Despite these results in rodents, translation to human therapy is not straightforward as anatomy and size significantly differ between rodents and human nerves. First, while an adult rat sciatic nerve is 2.5 cm long, a human sciatic nerve length goes from 40 to 55 cm according to the stature. Second, peripheral nerves collect axons of sensory and motor neurons in bundles called fascicles. While in relatively small organisms, such as rodents, nerves are made of a single fascicle, in larger organisms like humans, nerves are made of several fascicles that are pooled together in a matrix called epineurium. Consequently, intraneural injections with fine glass needles that were described in rodents¹² occur within the single fascicle and are called intrafascicular. This intraneural intrafascicular delivery is not translatable to humans as needles used for local anesthesia are too large and there are far too many fascicles (up to 50 fascicles in sciatic nerves¹⁴). While several reports showed that the intraneural non-intrafascicular delivery route is more potent to block nerve conduction and does not lead to more significant adverse events,¹⁵ this technique remains discouraged by professional societies. Indeed, standard care is a perineural delivery around the epineurium to deliver local anesthetics. To avoid intraneural delivery, injections are done using ultrasound-guided blunt needles,¹⁶ high-frequency ultrasonography,¹⁷ and electrical nerve stimulation with the metal needle to evaluate the distance between the needle and the nerve fibers through conductance.¹⁵ In addition, it is now possible to measure in real time the injection pressure at the needle tip as it is thought to avoid overpressure and a potential collapse of capillaries leading to local ischemia and axonal damage.¹⁸

In the present study, we used this controlled perineural injection route to set up procedure and parameters for a safe and efficient delivery of AAV2/9 vector into several nerves of arms and legs of non-human primate cynomolgus fascicularis (NHP). In these conditions, the vector transduced up to 90% of mSCs without generating serious adverse effects. This serves as a model to design safe and efficient gene therapy for demyelinating CMT diseases.

Results

Delivery route

NHP remains a key model for designing and optimizing AAV-based gene therapies. Indeed, toxicology and immune response following AAV treatments have been extensively documented in this model, and the NHP physiology remains generally close to human physiology. However, peripheral nerve structure diverges depending on the animal species. As an example, mice, rats, rabbits, dogs, and cynomolgus monkeys show an oligofasciculated structure with all fibers pooled in a single fascicle.¹⁹ On the other hand, pigs, sheep, and humans show a plurifasciculated structure with several bundles of pooled fibers.¹⁹ Either oligo- or pluri-fasciculated, fascicles (Figures 1A and 1B, labels 1) are surrounded by a tight sheath made of successive layers of perineurial cells and called perineurium (Figures 1A and 1B, labels 2). Fascicles are held together in the nerve by a delicate connective tissue containing collagen fibers, fibroblasts, and adipocytes, a matrix called the epineurium (Figures 1A and 1B, labels 3), which contains capillaries (Figures 1A and 1B, asterisks). This epineurium is not a barrier but a flexible holding structure that histologically identifies the nerve from the surroundings. The nerve is then structurally surrounded by several other loose layers of collagen fibers, fibroblasts, and adipocytes connected with the epineurium that form a surrounding fibrous matrix called circumneurium or paraneurium (Figures 1A and 1B, labels 4), which contains large blood vessels (Figures 1A and 1B, asterisks). Regarding this structure, unlike other nerves such as the radial nerve (Figure 1A), the sciatic nerve (Figure 1B) is the pool of several nerves, among which tibial (Figure 1B, label a) and fibular (also termed peroneal) (Figure 1B, label b) nerves are the largest. This pooling occurs through the fusion of their epineurium and a shared circumneurium.

Figure 1.

Cynomolgus monkey nerves, injection, and sampling sites

Cynomolgus monkey nerves have an oligofasciculated structure. Radial (A) and sciatic (B) nerve cross-sections hematoxylin and eosin stained. 1 = Fascicle, 2 = Perineurium, 3 = Epineurium, 4 = Circumneurium, a = Tibial nerve, b = Fibular/peroneal nerve, c = Lateral sural nerve, d = Cutaneous nerve, ∗ = Capillaries and blood vessels. 20X, Scale bars, 500 μm. Sciatic nerves are formed by tibial and common fibular nerves wrapped by common sheath (C). The main muscles innervated by sciatic nerves fibers are labeled accordingly.

In anesthesiology, regional anesthesia is obtained through the local delivery of anesthetics within or around the circumneurium using perineural (also termed paraneural to avoid confusion with perineurium) injections. However, a significant number of these injections occurs close to or in the perineurium or in the endoneurium.²⁰ In humans, these injections are not dangerous.²¹ Nevertheless, as local anesthetics are toxic at high concentration for nerve fibers and because needles, when they are sharp, can damage fibers, injections in the perineurium or in the endoneurium are not recommended. A combination of techniques has been developed to help physicians control the safety of regional anesthesia. First, ultrasound-reflective and “blunt” commercial needles are used to reach the nerve circumneurium without nerve puncture. Next, an electrical nerve stimulator is used to depolarize myelinated motor fibers through the metal needle and to slightly move muscles. Muscle contractions following electrical stimulation indicate the needle’s proximity to nerve fibers. Finally, a pressure-controlled injector controls the bolus delivery to avoid high injection pressure and potential nerve injury. We combined these techniques to perform safe perineural injections in cynomolgus monkey nerves: none of our injections were done at less than 0.1mA and more than 180 mm Hg (3.5 PSI), two parameters considered safe in human practice.

Dye diffusion along nerves

Two clinical anesthesiologists performed several Indian ink injections on the main nerves of upper and lower limbs of two anesthetized young adult cynomolgus monkeys. The diffusion along nerves was then analyzed immediately after animal sacrifice. Dye diffused on both sides of the injection site and relatively linearly along the nerves (Figure S1). In sciatic nerves, 700-μL and 300-μL injections diffused over 6.2 and 3 cm, respectively (Figure S1). Diffusion remained more variable in other targeted nerves, but it roughly followed the same diffusion pattern (Table S1). The nerve surface coverage by the dye ranged from 39% to 84% (Table S1).

NHP experimental design

We first treated three young adult male cynomolgus monkeys (3 to 4 years old) with AAV2/9-CAG-GFP-expressing green fluorescent protein (GFP) under the control of a strong and broad cytomegalovirus (CMV) enhancer combined with the chicken beta-actin promoter (CAG). Sciatic, fibular, and tibial nerves in legs and radial and median nerves in arms were injected with the vector solution or vehicle based on results obtained in the dye study: each injection was a 300-μL bolus and delivery speed was 100 μL/min; contiguous injections in the same nerve were spaced by 3 cm; total dose per nerve is the addition of each bolus dose injected on the nerve (Figure 1C).

According to their body weight, NHP 200014 received 3 × 10¹² vg/kg in six injections, NHP 200015 8 × 10¹² vg/kg in 11 injections, and NHP 200021 2.1 × 10¹³ vg/kg in 10 injections (Table 1). Including vehicle, the total number of injections was 20 (Table 1).

Table 1.

Animal specifications, injection sites, and doses

Nerve injected	AAV2/9-CAG-GFP				AAV2/9-U6-shRNA hPMP22
	Nb of injection	NHP 200014 male 3.35 kg	NHP 200015 male 3.45 kg	NHP 200021 male 2.8 kg	Nb of injection	NHP 171721 female 3.8 kg	NHP 191863 female 4.1 kg	NHP 191879 female 4.15 kg
		Dose (× 1012 vg/nerve)				Dose (× 1012 vg/nerve)
Sciatic R	3	5	10	25	3	2	5	5
Sciatic L	3	5	10	26	3	2	5	5
Fibular R	1	Veh	Veh	2.5	2	1	1	2.5
Fibuar L	1	Veh	2.5	2.5	2	1	1	2.5
Tibial R	2	Veh	2.5	Veh	2	1	1	2.5
Tibial L	2	Veh	Veh	Veh	2	1	1	2.5
Radial R	2	Veh	Veh	2.5	3	1	1	2.5
Radial L	2	Veh	Veh	Veh	3	1	1	2.5
Upper Median R	2	Veh	2.5	Veh	2	1	1	2.5
Upper Median L	0	/	/	/	2	1	1	2.5
Lower Median R	2	Veh	Veh	Veh	2	1	1	2.5
Lower Median L	0	/	/	/	2	1	1	2.5
Sural R	/	/	/	/	1	2.5	2.5	2.5
Sural L	/	/	/	/	1	2.5	2.5	2.5
Total	20	10	27.5	58.5	30	19	25	40
Dose/kg	/	3	8	21	/	5	6	9.6

L, left; Nb, number; NHP, non-human primate; R, right; Veh, vehicle solution; vg, vector genome.

Next, we treated three young adult female cynomolgus monkeys with AAV2/9-U6-shRNA hPMP22 expressing a shRNA targeting both human and cynomolgus PMP22 protein expression under the control of an ubiquitary human POLIII promoter U6. This promoter was previously shown to express shRNA in mouse and rat models.^12,22 The injection parameters (Table 1) were identical to the AAV2/9 CAG-GFP injection except for fibular nerves that received two injections, radial nerves that received three injections, and median nerves that received four injections (two in upper median and two in lower median). Sural nerves were also targeted.

Relative to their body weight, NHP 171721 received 5 × 10¹² vg/kg, NHP 191863 6 × 10¹² vg/kg, and NHP 191879 9.6 × 10¹² vg/kg all of them in 30 injections (Table 1).

Safety of the delivery route

To assess the impact of several nerve injections on monkeys' health, body, and organ weight (Figure S2), general behavior and sensory-motor reflexes were analyzed before and after injections. No adverse events were reported. In addition, nerve conduction velocity and compound of action potential amplitude were recorded in sciatic and median nerves before and after injections (Figures 2A and 2B). No significant difference was found.

Figure 2.

Perineural injections do not change electromyogram recordings or fine motor dexterity in upper limbs

Nerve conduction velocity (NCV) (A) and compound of muscle action potential (CMAP) (B) recordings before (baseline) and 28 days (day 28) after injections in sciatic (orange circles) and median (black squares) nerves. Each connected symbol represents data collected in one nerve. N = 6 animals and n = 9 nerves (some recordings were done unilaterally). (C) Graphic representation of the time to retrieve 10 pellets pre- and post-AAV2/9 administration during multiple sessions for all tested NHPs. The line shows the injection date. Each dot represents one successful trial. N = 3 (days 7, 15, 22) to 5 animals (days −3, −1, 2, 28). Three trials were performed at each time point.

We next analyzed the fine dexterity of the upper limb fingers using Brinkman’s tables.²³ These tables display multiple small horizontal and vertical rectangular holes filled with dried raisins (pellets). We measured how fast monkeys picked up all pellets. No significant alteration of the fine dexterity of the fingers was detected before and after nerve injections (Figure 2C).

Vector toxicity

Plasma was collected before treatment with AAV2/9-CAG-GFP or AAV2/9-U6-shRNA hPMP22 and then 6 and 30 days after treatment. We found no test item-related changes in plasma chemistry parameters 6 days after treatment with either vector (Figures S3 and S4). Thirty days post injection, animals dosed with AAV2/9-CAG-GFP displayed higher plasma concentrations of alanine aminotransferase, aspartate aminotransferase, and creatine kinase compared with pre-dose values (Figure 3, black circles). Values remained close to or within the normal range regarding reference values of clinical pathology parameters in male and female cynomolgus monkeys,²⁴ except at day 30 for AAV2/9-CAG-GFP treated NHPs. There were no test item-related changes in plasma chemistry parameters after dosing with AAV2/9-U6-shRNA hPMP22 (Figures S3 and S4).

Figure 3.

AAV2/9-CAG-GFP, but not AAV2/9-U6-shRNA hPMP22, alters the levels of alanine and aspartate aminotransferases and creatine kinase

Plasma values for alanine aminotransferase (A), aspartate aminotransferase (B), and creatine kinase (C) measured before (pre-dose), 6 days (D6) and 30 days (D30) after administration of AAV2/9-CAG-GFP (black circles) and AAV2/9-U6-shRNA hPMP22 (red squares) in nerves. Dotted lines show normal range values.²⁴ Each dot shows value for one animal.

The blood cellular fraction was also analyzed at the same time. No test item-related changes were observed in hematology, coagulation parameters, or C-reactive protein blood concentration on any sampling day (Figures S5–S7).

Histopathological analysis

Tissue samples of liver, heart, lung, kidneys, spleen, adrenals, spinal cord (three levels), skeletal muscle, brain, dorsal root ganglia, lymph nodes, and nerves (sciatic, tibial, fibular, median, and radial) were collected in all animals at when euthanized 30 days post injection. The histopathological analysis of these samples did not show any test item-related abnormalities. In particular, the elevation of some plasmatic liver enzymes did not correlate with abnormalities in the liver. Samples of the treated or vehicle-injected nerves were also collected and analyzed. Some of the samples collected in AAV2/9-CAG-GFP dosed animals displayed mild mononuclear cell infiltration in the tissues adjacent to nerves (Table S2). No correlation was found with treated nerves as some nerves injected with vehicle also showed this infiltration. No microscopic changes were found in nerves treated with AAV2/9-U6-shRNA hPMP22.

Immune response

Blood samples collected pre-dose and at day 30 post treatment in animals dosed with AAV2/9-U6-shRNA hPMP22 were analyzed for anti-AAV9 neutralizing factors, anti-AAV9 immunoglobulin (Ig)G levels, and cellular activation against AAV9 in the peripheral blood mononuclear cells (PBMCs) fraction to detect an immune response.

Before dosing, all animals had a low level of AAV9 neutralizing factors, and one had low-level anti-AAV9 IgG (Table 2). Treatment with AAV2/9-U6-shRNA hPMP22 generated the production of anti-AAV9 IgG in naive animals and increased their production in the non-naive animal. However, this increase of anti-AAV9 IgG only moderately increased AAV9 neutralizing factors. No T cell reactivity against AAV9 was detected before dosing and AAV2/9-U6-shRNA hPMP22 treatment did not increase this reactivity (Table 2). No correlation was found between doses and the immune response.

Table 2.

AAV2/9-U6-shRNA hPMP22 induces neutralizing factors and IgG production, but no T cell activation

Animal ID	Vector dose vg	AAV neutralizing factora		IgG anti-AAV9b		Anti-AAV9 cellular responsec
		Pre-dose	Day 30	Pre-dose	Day 30	Pre-dose	Day 30
172121	19 × 1012	+	++	–	+++	–	+
191863	25 × 1012	+	++	–	+++	–	–
191879	40 × 1012	+	++	+	+++	–	–

Day 30, 30 days after dosing; IgG, immunoglobulin G; vg, vector genome.

^a+: 1/10 ≤ titer ≤1/10², ++: 1/10² < titer ≤1/10⁴, +++: titer >1/10⁴, −: titer <1/10.

^b+: 1/10 ≤ titer ≤1/40, ++: 1/40 < titer ≤1/640, +++: titer >1/640, −: titer <1/10.

^cIFN-gamma ELISPOT; −: all three AAV9 peptides pool values ≤3 × negative control sample value; +: one out of three AAV9 peptides pool value >3 × negative control sample value; ++: two out of three AAV9 peptides pool value >3 × negative control sample value; +++: three out of three AAV9 peptides pool value >3 × negative control sample value (see Table S3 Source Data File for more details).

Biodistribution

On the day animals were euthanized, 30 days after treatment, three samples of all treated nerves as well as samples of several organs were collected in all animals dosed with AAV2/9-U6-shRNA hPMP22. The transduction efficiency was determined as the ratio of vector genome (vg)/diploid genome (dg) (Figure 4A). While the average transduction rate was 0.9 vg/dg in treated nerves, it reached 10.9 and 3.6 vg/dg in liver and spleen, respectively. On the other hand, low levels (≤0.05 vg/dg) were found in the heart, lung, and kidney. Levels were under the lower limit of quantification (LLOQ, <0.004 vg/dg) in spinal cord and dorsal root ganglia.

Figure 4.

Biodistribution and vector shedding

Absolute quantification of vector genome copy numbers per diploid genome (vg/dg) levels in nerves and organs of the three AAV2/9-U6-shRNA hPMP22-treated NHPs (A). Absolute vector genome quantification in urine (vg/mL) of the three AAV2/9-U6-shRNA hPMP22-treated NHPs (B). LLOQ, lower limit of quantification. Error bars show SEM in A and SD in B.

Vector shedding

Urine samples were also collected at different time points to measure vector shedding. As expected, no vector genome was detected in urine before vector injection. The amount of vector genome was highest 2 days after injection and sharply dropped by 14 days to be undetectable 28 days after vector injection (Figure 4B).

Vector efficacy

Nerves injected with AAV2/9-CAG-GFP were collected at euthanization to perform immunohistochemical analysis. Cryo cross-sections of the sciatic nerve showed many GFP-expressing cells in the two oligofasciculated nerves that constitute the sciatic nerve (Figure 5A), showing that the vector was able to cross the perineurium and reach fibers.

Figure 5.

AAV2/9-CAG-GFP efficiently transduces myelinating Schwann cells

NHP sciatic nerve cross-section unstained and imaged for GFP after AAV2/9-CAG-GFP injection (A). Injection localization shown by the white star. White dotted lines surround the monkey sciatic nerve consisting of two fascicles. Scale bar, 200 μm. High magnification showing GFP (green) and immunostaining for myelin FluoroMyelin (red) and axonal TuJ1 (blue) and merge image (B). Scale bar, 10 μm. Left sciatic nerve, animal 200015. GFP, green fluorescent protein; Tuj1, neuron-specific class III beta-tubulin.

To analyze the nature of these transduced cells, nerve cross-sections were immunostained for axonal (Tuj1) and myelin (FluoroMyelin) markers (Figure 5B). GFP was expressed in the cytoplasm of mSCs that surrounded FluoroMyelin staining, which itself surrounded axons (Figure 5B). These data were consistent with previously published data showing the high specificity of AAV2/9 for mSCs when injected locally in mouse, rat, and monkey nerves.¹²

Expression efficiency was calculated by counting cells positive for both GFP and FluoroMyelin over the total number of FluoroMyelin positive cells (Figures S8A and S8B). This expression efficiency was measured at three sampling sites: one was central regarding injection sites and two others were 4.5 cm distal and proximal regarding injection sites (Figure 1C). So, expression analysis covered up to 15 cm of sciatic nerves (Figure 1C).

In sciatic nerves, low, medium, and high doses allowed transducing an average of 55.1%, 58.4%, and 73.7% of mSCs along the 15-cm nerve (Figure S8C). The correlation dose/expression efficiency was not linear but plateaued (Figure S8B). The highest expression efficiency occurred in the injections area; however, expression remained high both in distal and proximal samples showing that the virus diffused beyond the 9-cm expected dye diffusion. Taken together, this suggests that expression efficiency is homogeneous along the 15 cm of collected cynomolgus sciatic nerves.

In other nerves, the expression efficiency was heterogeneous except in the tibial nerve where it reached 65.5% in middle but also distal and proximal samples (15 cm of nerve) (Figure S8D). This heterogeneity may reflect the difficulty of reaching the perineurium of small monofascicular cynomolgus nerves with a needle designed for large plurifascicular human nerves.

Three samples of each nerve injected with AAV2/9-U6-shRNA hPMP22 were collected as described previously (Figure 1C) and these samples were analyzed using qPCR to detect the vector genome per diploid genome. AAV2/9-U6-shRNA hPMP22 was detected in all injected nerves, but with a high variability between nerves (Figure S9). The most homogeneous transduction efficiency was found in sciatic nerves. There, the mean transduction efficiency was 0.73 and 1.36 vg/dg for 2 × 10¹² and 5 × 10¹² vg injected, respectively (Figure 6A). A similar range of transduction efficiency was obtained in tibial nerves (0.5–1.2 ± 0.3 vg/dg). In other nerves, the lowest chosen dose (1 × 10¹² vg/nerve) resulted in a mean transduction efficiency close to 0.8 vg/dg. We found no correlation between anti-AAV9 IgG positivity prior to treatment and lower transduction efficiency.

Figure 6.

AAV2/9-U6-shRNA hPMP22 transduction efficiency and dose/response in sciatic nerves

Transduction efficiency in vg/dg in the three sampling sites of left (squares) and right (dot) sciatic nerves of three injected cynomolgus monkeys, NHP 171721, 191863, and 191879 (A). Error bars show SEM. PMP22 protein level in sciatic nerve of NHPs 1 month after AAV2/9-U6-shRNA hPMP22 administration (B). PMP22 levels are normalized to β-tubulin level as loading control. Each symbol shows one sample value. Error bars show SEM. Statistical tests are one-way ANOVA followed by Dunnett’s multiple comparisons test vs. non-injected control values.

Besides ultrasonography imaging, we used a nerve stimulator to prevent intraneural delivery. Reporting the transduction efficiency to the nerve stimulation intensity (mA) showed that a low intensity (≤0.2 mA) resulted in a higher heterogeneity of the transduction efficiency than at high intensity (≥0.3 mA) (Pearson’s correlation R-values −0.06 vs. 0.29, respectively; Figure S9). Injections at higher stimulating currents, more distant to the fibers, were as successful. This confirmed that a delivery distant to the fibers, such as the standard perineural or paraneural route, is an efficient option for AAV delivery in NHP nerves.

Efficacy of AAV2/9-U6-shRNA hPMP22 in decreasing PMP22 expression

As AAV2/9 efficiently transduced mSCs following a local injection and AAV2/9-U6-shRNA hPMP22 targets both human and monkey PMP22, we tested the efficiency of AAV2/9-U6-shRNA hPMP22 to decrease PMP22 expression in cynomolgus monkey nerves. One month after injection, samples of sciatic, tibial, and fibular nerves treated in increasing doses of vector were analyzed for PMP22 and myelin basic protein (MBP) expression using automated western blot analysis (Simple WesternTechnology; Biotechne, USA). PMP22 expression decreased significantly in a dose-dependent manner in sciatic (Figure S10A), tibial (Figure S10C), and fibular (Figure S10E) nerves, while MBP expression remained unchanged or slightly increased (Figures S10B, S10D, and S10F).

As we could not process control non-treated nerves in this first analysis for logistical reasons, we next used classical western blotting to compare PMP22 protein expression in treated vs. non-treated sciatic nerves (Figure 6B). AAV2/9-U6-shRNA hPMP22 significantly decreased cynomolgus PMP22 expression by 25% and 40% with 2 × 10¹² and 5 × 10¹² vg/nerve, respectively, when compared with non-treated nerves (Figure 6B). The decrease was relatively homogeneous among the three samples (proximal, central, distal) collected 15 cm apart in each nerve (Figure 1C). This suggests that, while the dye diffusion after 1 h occurs over 9 cm in sciatic nerves (Figure S1), the vector can reliably diffuse over at least 15 cm, and this diffusion allows for relatively homogeneous decreases of PMP22 expression over this distance.

Discussion

As mSCs are highly differentiated cells with a very low turnover,^25,26 the use of long-term AAV-based gene therapy to treat CMT1A is relevant. However, the systemic administration of AAV vectors has also shown some associated risks.²⁷ In addition, the intrathecal delivery of AAV9 is both relatively inefficient to transduce mSCs²⁸ and toxic for dorsal root ganglia neurons in NHPs.^29,30,31 The data we present here are discussed in this context.

Perineural delivery route: Safety, efficacy and limitations

In our study, none of the injections were associated with any adverse events concerning animal behavior, sensorimotor reflexes, electrophysiology, fine motor dexterity, or histopathology. Monocytic infiltration in some nerves injected with vehicle or AAV2/9-CAG-GFP may be linked to the delivery route. Indeed, like injection site reactions (ISRs) in subcutaneous injections,³² local tissue dilatation during injection may lead to monocytic infiltration around nerves. Nevertheless, no infiltration was observed after similar injections of AAV2/9-U6-shRNA hPMP22, so sufficient evidence is lacking to validate this hypothesis.

The efficiency of the delivery route to address the vector to target mSCs along nerves is demonstrated through several datasets. First, Indian ink injections showed that dye diffusion preferentially occurs along the nerve, suggesting that small molecules tend to follow the tubes formed by the several layers that surround the fascicles. Second, AAV2/9-CAG-GFP injections showed that AAV2/9 vector can cross these concentric layers and the perineurium to reach many fibers in fascicles. With this method, vector transduction efficiency reached more than 70% of target cells over 15 cm of sciatic nerve. We did not analyze the vector specificity for mSCs, but these cells represented a large majority of transduced cells. This observation was consistent with our previous data on rodent and NHP nerves showing that more than 90% of transduced cells are mSCs in sciatic nerves of adult animals.¹² An even higher transduction was obtained with AAV2/9-U6-shRNA hPMP22 vector in NHP sciatic nerves. Assuming mSCs account for 50% of cells in peripheral nerves²⁶ and that more than 95% of AAV2/9-transduced cells are mSCs,¹² our data suggest that each mSC received an average of 1.5 (dose 2 × 10¹² vg) to 3 (dose 5 × 10¹² vg) vector genomes in sciatic nerves. Despite the high variability of the transduction rate in non-sciatic nerves and if the same parameters apply to all nerves, each mSC received one vector genome in non-sciatic nerves. While the efficiency of the two tested vectors appears different, we suspect that the use of GFP fluorescence to detect transduction is less sensitive than the quantitative PCR to detect vector genome. Indeed, some cells expressing a lot of GFP tend to mask cells expressing less. We conclude that qPCR detecting the vector genome is probably the most reliable method to measure the transduction rates along nerves.

It is worth noting that while dye diffusion was limited to 9 cm in sciatic nerves, we found a relatively homogeneous high transduction efficiency over at least 15 cm. This suggests that the vector diffused over a larger distance than dyes along nerves after a local delivery. An interesting follow-up would be to define the optimal volume and delivery speed to reach maximal diffusion. However, as sciatic nerve length hardly reaches 15 cm, the NHP is probably not the best animal model for that study.

The main characteristic of GFP expression efficiency in non-sciatic nerves was heterogeneity. As these nerves are approximately twice thinner and thus more difficult to inject than the sciatic nerve, we hypothesized that variability in the place of injection underlined this heterogeneous transduction. Correlating the transduction efficiency with the localization of the injection indicated that delivery close to the fibers resulted in a significant efficacy of transduction despite some heterogeneity. A more distant delivery at higher current intensity resulted in similar efficacy. While intraneural delivery, very close to the fibers, may be more efficient to transduce target cells, our data indicate that delivery in the circumneurium is also effective. In addition, intraneural injection is not recommended by anesthesiology associations in Europe and in the United States as it is believed to be associated with a higher probability of adverse events. We therefore conclude that the perineural delivery route is appropriate for the local delivery of gene therapy vectors in NHP.

Taken together and according to the practice of perineural injections in anesthesiology, our data suggest that perineural delivery of AAV2/9 vectors in the main nerves of arms and legs of humans is both safe and efficient to target mSCs.

Low toxicity and high efficiency of AAV2/9-U6-shRNA hPMP22 following local delivery

AAV2/9-U6-shRNA hPMP22 was designed to target both human and cynomolgus PMP22 expression. No change in body weight, organ weight, or blood biochemical and cellular parameters of monkeys treated with this vector 6 days and 30 days after injections with any of delivered doses. In comparison, AAV2/9-CAG-GFP delivery resulted in a moderate elevation of alanine aminotransferase, aspartate aminotransferase, and creatine kinase 30 days after injections at all doses. All AAV2/9-U6-shRNA PMP22 doses were like the low and medium doses of AAV2/9-CAG-GFP, indicating that for a similar dose, AAV2/9-U6-shRNA PMP22 is less toxic. This may be explained by the immunogenic nature of GFP in several animal models.³³ It is worth noting that, even with the highest dose and immunogenicity of GFP, the local delivery of AAV2/9-CAG-GFP resulted in no remarkable test-item-related histopathological defects and in a moderate elevation of the liver enzymes compared with the systemic delivery of a similar dose of AAV2/9 vector in monkeys.^30,31 In addition, no histopathological defect was seen in dorsal ganglia neurons with this vector, suggesting that the local delivery of AAV2/9-CAG-GFP makes it less toxic than after a systemic or intrathecal administration.^30,31

Biodistribution of AAV2/9-U6-shRNA hPMP22 showed that the vector can cross the blood vessels around nerves and reach the liver and spleen. A very low amount of AAV2/9-U6-shRNA hPMP22 was also observed in heart, kidney, and lung, but none in spinal cord and dorsal root ganglia. The level of vector genome observed in non-treated tissues was around 100 times lower than when AAV2/9 is delivered in a systemic administration,³⁰ indicating that local delivery is limiting the distribution of the vector in the organism.

A clear immune response was observed after delivering AAV2/9-U6-shRNA hPMP22 in nerves with the generation of antibodies and neutralizing factors against the vector. However, no T cell response was detected, suggesting that the level of exposure to the antigen is limited after local delivery. All animals were positive for AAV9 neutralizing factors, and one already had antibodies against AAV9. Yet, this was not correlated with a decreased transduction efficiency or any detected toxicity, suggesting that a pre-existing immunity is not preventing the local transduction of mSCs.

The efficiency of AAV2/9-U6-shRNA hPMP22 to decrease monkey PMP22 expression was first measured in several treated nerves comparing low and high doses. Then, we investigated it specifically in treated sciatic nerves compared with non-treated animals. In these conditions, we found that the vector decreased PMP22 expression in a dose-dependent manner, reaching up to 40% downregulation with the highest dose.

Variability among distal, central, and proximal samples was low in two animals, suggesting that the amount of vector that reaches the extremities of the sciatic nerve is sufficient to decrease PMP22 relatively homogeneously all along it. This relative homogeneity of the molecular effect is an important factor of the success of the therapy in patients. Indeed, molecular and cellular defects are homogeneously distributed along nerve fibers in CMT1A animal models and in patients.³⁴ This allows characterizing the disease in nerve biopsies. This suggests that the slow-down of nerve conduction in patients results from the cumulation of slow-down or conduction failure all along fibers rather than on a particular area of the nerve. In these conditions, bringing a benefit all along the nerve fiber is critical for improving patients' health. One of the next steps of the development of the local gene therapy will be to confirm the homogeneity of the decrease of PMP22 expression along treated nerves.

Forty percent decrease in PMP22 protein expression in treated sciatic nerves may seem a relatively low number regarding the high transduction rate we reported. We calculated that an average transduction of 1 vg per mSC (0.5 vg/dg) led to a decrease of PMP22 protein expression by 20%. As the cesium chloride-gradient purification that was used to produce our vector batches results in more than 90% full particles,³⁵ the limited efficiency of the vector to decrease PMP22 expression is unlikely to be due to emptyvector particles. It may also reflect the high level of expression of this myelin protein in mSCs and the very large size of the cell (it can reach up to 2 mm long in large animals). Indeed, the expression of the siRNA in transduced cells must catch up with the high expression rate of the protein in target cells. A third parameter is that PMP22 inserted in the myelin is highly stable, constituting a pool of protein hardly affected by the siRNA expression. In any case, this means that a meaningful decrease of PMP22 expression will probably require multiple copies of the vector DNA in the same cell. However, our data indicate that a meaningful downregulation of PMP22 (up to 40%) can be reached in just 1 month in a large animal such as a monkey with a dose of 5 × 10¹² vg/sciatic nerve and no detected toxicity.

Significance regarding a treatment for CMT1A disease

In a preclinical and proof-of-concept situation, our data show that the perineural delivery route developed by anesthesiologists for regional anesthesia is both safe and efficient to decrease PMP22 expression in a large amount of mSCs of an adult monkey.

Our data suggest also that the amount of siRNA to be delivered in mSCs to reach a meaningful decrease of PMP22 expression is relatively high. A highly targeted delivery of the siRNA to mSCs thanks to the AAV2/9 vector administrated locally may therefore represent an advantage over a systemic administration of the siRNA. Indeed, the high dose of siRNA required to meaningfully decrease PMP22 may constitute a risk in other tissues and organs after a systemic delivery.

After a local delivery of AAV2/9-U6-shRNA hPMP22, the extent of this downregulation is limited to the main nerves of arms and legs that were targeted. As CMT1A is a genetic disease that affects all nerves, a benefit limited to a set of nerves may appear unsatisfying. However, regarding the pathophysiology of CMT1A disease that affects mostly the longest myelinated fibers of the body and the main symptoms of many patients, a correction of PMP22 expression in mSCs of these nerves should bring enough benefit to completely change the life of the patients. This is supported by the significantly high benefit observed in a CMT1A rat mimicking the disease that was treated in sciatic nerves only.¹² Taken together, our data suggests that the local AAV9- and shRNA-based gene therapy constitutes a safe and efficient approach to treat CMT1A.

Materials and methods

Study design

To investigate dye diffusion, two male NHPs were injected with black China Ink (Staedtler, Black, Reference #745 M2-9) (1:1 water volume). Three males received either vehicle (sterile phosphate-buffered saline) or AAV2/9-CAG-GFP and three females received AAV2/9-U6-shRNA hPMP22, shRNA directed against human and monkey PMP22 mRNA. Animals received between 1 and 26 × 10¹² vg/nerve in 100–300 μL. During the procedure, animals were placed on warming blankets and physiological parameters were monitored. After full recovery from anesthesia, they were replaced in their cages, and clinical observations were daily performed. Biological matrix collection (urine and blood) was performed at pre-dose and after injection on day 2, day 14, and day 28. Urine collection was performed in metabolism cages and immediately frozen and stored at ≤ −18°C. Blood samples were drawn in EDTA tubes (hematology), citrated tubes (coagulation parameters), and lithium heparin tubes (blood chemistry, C-reactive protein analysis, immune response) from the cephalic vein.

Animals

Eight male and female cynomolgus macaques (Macaca fascicularis) (CynoConsulting, Toulouse, France) were included in the study. Animals were imported from licensed primate breeding centers in Mauritius (Le Tamarinier) and Vietnam (Nafovanny). All procedures were performed in an E1802301-authorized facility under APAFIS number #25875-2020052714098186v2. Non-human primates (NHPs) remained under veterinary care during the full study.

Animals were housed collectively by group and by gender in an indoor enclosure with daily observations. Single housing was limited to the periods necessary for the correct performance of animals' care and experimental procedures. Animals were placed in an air-conditioned (20°C–24°C) animal house kept at relative humidity between 40% and 70% with non-recycled filtered air changed 10–15 times per hour. The artificial day/night cycle involved 12 h light and 12 h darkness with light on at 7:30 a.m. SDS DIETEX OWM (E) Short Banane SQC was available ad libitum except during the fasting experimental period. Food was withheld during the night before scheduled blood sampling for hematology/coagulation parameters and blood clinical chemistry, during urine collection for urinalysis, and during the night before scheduled necropsy. The diet was supplemented with fresh vegetables, fruits, and other supplements after the approval of the attending veterinarian. Drinking water was available ad libitum.

Vector production

Cloning of the enhanced green fluorescent protein (GFP) and shRNA in pAAV and AAV vector productions was provided by INSERM UMR1089-TaRGeT (Nantes, France).

Cloning, cell amplification, transfection, harvest, and supernatant PEG-precipitation

The vector plasmids were generated by cloning the shRNA17.PMP22 sequence in the pAAV.U6 plasmid kanamycin-resistant backbone or the CAG.GFP.SV40 polyA in pAAV.MCS ampicillin-resistant backbone and amplified in StBl E. coli bacteria. Cloning of vector plasmids was validated by restriction enzyme profile and sequencing.

The HEK293 cells seeding in Cell Stack Cells 5 chambers with vent caps (CS5) were cultured with DMEM supplemented with 10% FBS and 1% Pen/Strep. At approximately 80% confluency, cells were co-transfected with vector and helper plasmids (containing helper genes from adenovirus and the rep cap genes) using the CaPO₄ precipitate technique. The culture medium was removed from the CS5 and exchanged with the transfection medium; the cells were then incubated 6 to 15 h at 37°C and 5% CO₂. The transfection medium was then removed from the CS5 and replaced by a fresh exchange medium (DMEM, 1% Pen/Strep) prior to a 3-day incubation at 37°C and 5% CO₂. The cells of the CS5 transfected were then harvested. The supernatant was precipitated at 5°C overnight with polyethylene glycol (PEG). The precipitated supernatant was then centrifuged. The supernatant was discarded, and the PEG-pellet resuspended in TBS before Benzonase digestion.

Vector purification by double cesium chloride-gradient ultracentrifugation

The vector suspension was centrifuged, and the vector-containing supernatant was loaded on a step density CsCl gradient in UltraClear tube for SW28 rotor. The gradient was centrifuged at 28,000 rpm for 24 h at 15°C. The full-particle band was collected and transferred to a new UltraClear tube for SW41 rotor. The second gradient was centrifuged at 38,000 rpm for 48 h. The enriched full-particle band was then collected (usually >90% of full particles).³⁵ The vector suspension was then subjected to four successive rounds of dialysis in a Slide-a Lyzer cassette against DPBS 1X. The purified vector was finally collected, sampled for vg titer and purity assay, and stored at −70°C in polypropylene low-binding cryovials.

Vector titration

The recombinant AAV vector genome was titered by quantitative polymerase chain reaction (qPCR); the target amplicons correspond to ITR-2.³⁶

In vivo delivery

Animals were sedated through an intramuscular treatment in deltoid muscle of ketamine hydrochloride (Imalgene, 10 mg/kg) and xylazine (2% Rompun, 0.5 mg/kg). Anesthesia was then maintained with propofol (Propovet, 10 mg/kg/h, intravenous infusion in the external saphenous vein). Under anesthesia, animals were injected into the main nerves by two anesthesiologists.

Electromyogram CMAP/NCV

The measurement of compound muscle action potentials (CMAPs) and nerve conduction velocity (NCV) was carried out on the right median and sciatic nerves using Dantec system Keypoint Focus (Natus Neurology Incorporated, USA). Electromyographic evaluation was performed at pre-treatment and day 29. The measurements were performed under anesthesia by isoflurane inhalation after chemical immobilization with ketamine (10 mg/kg, intramuscularly) and anesthesia induction with propofol at 4 to 5 mg/kg.

Brinkman table test

The manual dexterity was assessed using a modified Brinkman table test.²³ The modified Brinkman table (14 × 22 cm) is a Perspex board pierced with 25 horizontally and 25 vertically oriented wells 15-mm long, 6-mm deep, and 8-mm wide. Animals were trained to grasp food rewards (dried raisins) within a limited time slot. The training sessions were conducted over at least 3 weeks to allow the establishment of a baseline level for each animal. After the start of the study (the day of administration of the vectors), the test was carried out weekly. For each session, the unrestrained animals were placed in an individual box, then a Brinkman board filled with food rewards was presented to each animal. The animals were free to grasp rewards within a limited time slot of 1 minute. This procedure was repeated two additional times within a weekly session. Each attempt was separated by a fixed time slot of 5 minutes. The manual dexterity was assessed according to the time required to pick up the 10 pellets during a session.

Immune response

Peripheral blood mononuclear cell isolation

PBMCs from blood samples in Li-Heparin tubes were isolated using a Ficoll gradient-based strategy, followed, when necessary, by a red blood cell lysis. All cells were cryopreserved at −65°C in an FBS-10% DMSO solution using a Mr Frosty box system for a maximum of 7 days before a transfer in liquid nitrogen for long-term storage. PBMC isolation was performed from blood samples of animals before injection (BD) and when animals were euthanized (D33).

Anti-AAV9 cellular response

An ELISpot assay was performed on isolated PMBCs for the detection of interferon (IFN)γ-secreting cells to AAV9 VP1 capsid protein. It is based on an enzyme-linked immunospot (ELISpot) assay and performed according to the manufacturer’s recommendations (MABTECH). Briefly, PBMCs were restimulated in vitro with overlapping peptides spanning the sequence of interest (i.e., AAV9 VP1). For each tested sample, a negative control consisting of unstimulated cells (RPMI complete medium) without any antigen, and a positive control for cytokine secretion consisting of Concanavalin A (Sigma; C0412-5MG) were used. After an incubation of 24 h, spot-forming colonies (SFCs) were determined using an ELISpot reader ELR07IFL (AID iSpot) and analyzed with AID ELISpot Reader Software V7.0. Independently of the peptide pool used, responses were considered positive when the number of SFCs per 1 × 10E6 cells were >50 and at least 3-fold higher than the control condition (unstimulated cells with medium only).

Anti-AAV9 neutralizing factors

The detection of anti-AAV9 neutralizing factors (NFs) was performed using a neutralization assay on a 96-well plate. Briefly, the method consists in a cell transduction inhibition assay using a recombinant AAV9 expressing the reporter gene system LacZ. Gene expression is measured using a luminometric method. Each serum was tested using a range of six dilutions (1/10, 1/10², 1/10³, 1/10⁴, 1/10⁵, and 1/10⁶). Each assay is validated on acceptance criteria such as cell number range, AAV9 transduction level in the absence of serum, limit of detection (LOD), or a neutralizing serum positive control. NF titer corresponds to the last dilution leading to the inhibition of more than 50% of the transduction (the 100% corresponding to the transduction control with AAV9 alone). Anti-AAV9 neutralizing factor analysis was performed for animals before injection (BD) and on day 28 post dosing.

Anti-AAV9 humoral response

The method is based on a qualitative enzyme-linked immunosorbent assay (ELISA). Briefly, Nunc MaxiSorp 96-well plates were coated with recombinant AAV9 particles. After saturation of the wells, six dilutions of the samples were added into the wells (4-fold dilutions from 1/10 to 1/10,240); each dilution was tested in triplicate. After incubation, anti-NHP IgG horseradish peroxidase (HRP)-conjugated antibody was revealed by measuring absorbance at 450 nm and 570 nm after addition of tetramethylbenzidine (TMB) substrate (transformed in a measurable colored compound in presence of HRP) and of stop solution consisting of phosphoric acid. The optical density signal 450 nm–570 nm was considered for the analysis (570 nm being measured and subtracted to eliminate unspecific absorption). For each specimen, results are expressed as “negative” or “positive.” For a positive sample, antibody titer was determined as the last serum dilution above the positivity threshold.

Biodistribution and transduction rate

Genomic DNA extraction and quantification

Genomic DNA (gDNA) was extracted from tissue samples using the Gentra Puregene kit and the Tissue Lyser II (both from Qiagen), according to the manufacturer’s instructions. Final gDNA pellets were resuspended in 100 μL of DNA Hydration Solution. Concentration and purity of the gDNA samples were determined using a Multiskan Go reader (Thermo Scientific), by measuring optical densities at 260 nm (A260) and 280 nm (A280).

qPCR analysis

Simplex qPCR analyses were conducted in duplicate using 50 ng of gDNA (5 μL of gDNA diluted at 10 ng/μL). Transgene-specific qPCR was conducted on a StepOne PlusTM Real-Time PCR System (Applied Biosystems, ThermoFisher Scientific). Vector genomes (vg) were determined using the following primer/probe combination, designed to amplify a specific region of the BGH polyA sequence, present in the rAAV9-U6-shRNA17 vector:

Forward: 5′-TCTAGTTGCCAGCCATCTGTTGT-3′

Reverse: 5′-TGGGAGTGGCACCTTCCA-3′

Probe: 5′-FAM-TCCCCCGTGCCTTCCTTGACC-TAMRA- 3′

In parallel, endogenous gDNA copy number was determined using primer/probe combination designed to amplify the NHP Epsilon globin gene³⁷:

Primer forward: 5′-TGGCAAGGAGTTCACCCCT-3′

Primer reverse: 5′-AATGGCGACAGCAGACACC-3′

Probe: 5′-TGCAGGCTGCCTGGCAGAAGC-3′

For each sample, cycle quantification (Cq) values were compared with those obtained with different dilutions of a linearized standard plasmid (containing either the BGH polyA sequence or the NHP Epsilon globin gene). The absence of matrix effect on BGH polyA qPCR performances was determined by analyzing 100 ng of different NHP gDNA (liver, heart, lung, kidney, spleen, spinal cord, and peripheral nerve) obtained from an untreated control NHP tissues, spiked with different dilutions (10⁷ to 25 copies) of the linearized plasmid carrying the BGH polyA sequence. No matrix effect was observed on BGH polyA qPCR performance parameters (i.e., efficacy, linearity, and specificity). The absence of matrix effect on NHP Epsilon globin qPCR performances was determined by analyzing different dilutions (10⁷ to 100 copies) of the linearized plasmid carrying the NHP Epsilon globin sequence, as well as different quantities (1,000, 500, 250, 100, 50, 25 and 10 ng) of different NHP gDNA (liver, heart, lung, kidney, spleen, spinal cord, and peripheral nerve) obtained from an untreated control NHP tissues. No matrix effect was observed on NHP Epsilon globin qPCR performance parameters, i.e., efficacy, linearity, and specificity. Results were expressed as vector genome copy number per diploid genome (vg/dg). The lower limit of quantification (LLOQ) of the assay was 0.003 vg/dg.

Vector shedding

DNA extraction

rAAV DNA was extracted from 140 μL of urine using the QiAamp viral RNA minikit (Qiagen) and according to the manufacturer’s instructions. Final elution was done in 80 μL of elution buffer.

qPCR analysis

Simplex qPCR analyses were performed in duplicate, using 5 μL of urine DNA. Transgene-specific qPCR was conducted on a StepOne PlusTM Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). Vector genome copy number was determined using the same BGH polyA qPCR assay as the one used for tissue sample analysis as described above. The absence of matrix effect on BGH polyA qPCR performances was determined by analyzing 2.5 μL of urine DNA obtained from an untreated control NHP urine sample, spiked with different dilutions (1E+7 to 25 copies) of the linearized plasmid carrying the BGH polyA sequence. No matrix effect was observed on BGH polyA qPCR performance parameters (i.e., efficacy, linearity, and specificity). The final recovery of vector genomes during extraction of NHP urine was evaluated by spiking known quantities of the AAV9-U6-shRNA17-BGHpA (Batch ID 7482, i.e., the same one as those injected into the NHPs of this study) in an NHP urine sample obtained from an untreated control animal. DNA was extracted from the resulting samples and analyzed using BGH polyA qPCR. The final recovery of vector genomes in NHP urine was 30%. Therefore, a correction factor of 30% was applied to the results.

CRP analysis

Aviva Systems Biology Crp ELISA Kit (Monkey) (OKWB00291) based on standard sandwich enzyme-linked immunosorbent assay technology was used to evaluate C-reactive protein plasma concentration at pre-dosing, day 14, and day 28 after AAV injections. Tests were performed according to the manufacturing protocol.

Sample preparation

On the day animals were euthanized, 30 days after treatment, the animals were premedicated with Ketamine HCl and killed by subtotal exsanguination following sodium pentobarbital anesthesia by the intravenous route. Killing and necropsy was performed under the responsibility of a veterinarian. The organ samples were weighed, paired organs together, before freezing on dry ice (biodistribution) or fixed in 10% buffered formalin (histopathology). The nerves were dissected and then samples were frozen on dry ice (biodistribution) or fixed in 10% buffered formalin (histopathology).

Immunohistochemistry

Following fixation in 10% buffered formalin, samples were incubated 24–48 h in two successive baths of 6% and 30% sucrose and then embedded in optimal cutting temperature (OCT, NEG-50, MM France) and stored at −80°C. Sections (10 μm of thickness) were cut using cryostat apparatus (LEICA CM3050). Cryosections were blocked with 5% normal goat serum (NGS) and 0.1% Triton X-100 in PBS, incubated overnight at 4°C with primary antibody mouse anti-Tuj1 (Invitrogen, Ref. MA1-118, dilution 1/400), diluted in NGS/Triton/PBS, washed with PBS, and then incubated 1 h at room temperature with secondary antibody donkey anti-mouse Alexa Fluor 647 (Invitrogen, Ref. A31571, dilution 1/1,000) and Red FluoroMyelin (Invitrogen, Ref. F34652, dilution 1/300) diluted in NGS/Triton/PBS. After several PBS washes, cryosections were mounted in Dako fluorescent mounting medium (S3023). LMS700 confocal microscope (Zeiss, France), AxioScan slide scanner (Zeiss, France), NanoZoomer 2 slide scanner (Hamamatsu, Japan), and an Apotome fluorescence microscope (Zeiss, France) were used to obtain images. The percentage of transduced mSCs (GFP-positive cells) over all mSCs (FluoroMyelin positive) was calculated in three areas of a nerve section using ImageJ software. The percentage was calculated at the injection site (central), proximally (toward the spinal cord) and distally (toward the hand) regarding the injection sites (Figure 1C).

Histopathology

Histological slides were stained with hematoxylin and eosin, and examined (Dr. Longeart, DMV, ACVP, Pathology consultant) under light microscopy using a Leica DMLB stereomicroscope. Diagnoses were recorded with a purpose designed validated computer software, the RoeLee v.3.1 system.

Automated western blotting and analysis

Nerve samples were lysed with a buffer consisting of CelLyticMT reagent with 1% of protease and phosphatase inhibitor cocktail (60 μL per well). Lysates were stored at −80°C and processed at +4°C. For each condition, the quantity of proteins was determined using the micro kit BCA (Pierce). Jess Automated western blot system (Simple Western Technology, Bio-Techne, USA) was used to detect PMP22 and MBP proteins in nerve samples. The run was performed according to the manufacturer’s recommendations. Capillaries, samples, antibodies, and matrices were loaded inside the instrument. Each protein was evaluated independently. Capillaries were incubated for 2 h with primary antibodies (anti-PMP22, Sigma-Aldrich, SAB4502217 and anti-MBP, Novus, MBP2-46632) at room temperature (23°C, ±3°C). Capillaries were washed and then incubated with HRP-conjugated secondary antibodies for 1 h at room temperature (23°C, ±3°C). After removal of unbound secondary antibody, the capillaries were incubated at room temperature (23°C, ±3°C), with the luminol-S/peroxide substrate and chemiluminescent signal collected using the charge-coupled device (CCD) camera with six different exposure times (30, 60, 120, 240, 480, and 960 s). Areas under the curve were determined using the high dynamic range 4.0 method, that integrates the intensity obtained at the different exposure times. Data analysis was performed using the Compass Software (Bio-Techne, USA).

Western blotting

Frozen sciatic nerve samples were solubilized in 4°C lysis buffer (0.5 mL Tris HCl 1 M pH 8, 0.375 mL NaCl 4 M, 40 μL EDTA 0.5 M, 100 μL Triton and 8.9 mL H₂O) completed with protease inhibitors (Fisher Scientific, France). Each nerve was cut into small parts (around 0.5 or 1 mm each part) and sonicated three times for 10 s on ice (Microson ultrasonic cell disruptorXL, Microsonic), vortexed five times for 2 min and in rotation overnight at 4°C. Then, 10 μg of protein samples were denatured using Laemmli/β-mercaptoethanol (Merck, S3401) incubation at 95°C for 10 min, separated on a denaturing 10% SDS-polyacrylamide gel and transferred onto nitrocellulose membranes. Membranes were blocked for 60 min with 5 mL of LI-COR Blocking Buffer. The following primary antibodies were incubated overnight at 4°C in the same blocking buffer: rabbit anti-PMP22 (1:100, abcam, ref. 211052) and mouse anti-b-tubulin (1:500, Sigma-Aldrich, T8578-100μL). The following day the membranes were washed three times for 10 min in TBS Tween 20 (0.1% v/v) and then incubated for 1 h at room temperature with the secondary fluorescence antibodies: goat anti-mouse IRDye 800 (1:10,000, LI-COR Biosciences, ref. 926–32210), and donkey anti-rabbit IRDye 680 (1:10,000, LI-COR Biosciences, ref. 926–68073). After secondary antibody incubation, the membranes were washed three times for 10 min with TBS Tween 20 (0.1% v/v). Visualization of the bands was performed using the Odyssey CLX LI-COR Imaging System and band quantification was performed using ImageJ software (version 4.0).

Statistical analysis

Data were analyzed with GraphPad Prism version 10 (GraphPad Software) and expressed as the mean ± standard error of the mean (SEM) or ± standard deviation (SD) as indicated in the figure legends. Statistical differences between mean values were tested using one-way ANOVA followed by Dunnett’s multiple comparison test. Differences between values were considered significant with ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. All exact p values are available in the Source Data File.

Data availability

Raw data are available in the “Espallergues et al. Source Data” Excel file.

Acknowledgments

The present work was supported by a European Research Council grant (FP7-IDEAS-ERC 311610) to N.T., SATT AxLR and an ATIGE grant from Genopole to N.T. All authors thank Sergio Gonzalez for his contribution to the study. F. Swisser and O.C. thank the Department of Anesthesiology and Critical Care Medicine, Lapeyronie University Hospital, Montpellier, France.

Author contributions

N.T. and V.M. managed the study. J.C., F. Souab, F. Swisser, and O.C. performed perineural injections. P.B. provided expertise and analyzed images. J.E., V.M., M.-L.S., and N.T. analyzed the experimental data. J.E. and N.T. were major contributors in writing the manuscript. N.T. set up, directed, and supervised the study. All authors read and approved the final manuscript.

Declaration of interests

N.T. is co-founder and CEO/CSO of Nervosave Therapeutics. J.E. and G.v.H. are employees of Nervosave Therapeutics. M.-L.S. is an employee of ERBC. V.M. is an employee of Biotrial. All other authors declare no competing interests. N.T. and INSERM hold a patent US10801040B2 and a patent application EP22758538.7/US18/294,325 related to this work.