Research ArticleAmyotrophic Lateral Sclerosis

Safe and effective superoxide dismutase 1 silencing using artificial microRNA in macaques

See allHide authors and affiliations

Science Translational Medicine  31 Oct 2018:
Vol. 10, Issue 465, eaau6414
DOI: 10.1126/scitranslmed.aau6414

A silent attack on ALS

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder for which there are currently no available therapies. Gain-of-function mutations in the gene encoding superoxide dismutase 1 (SOD1) are responsible for 20% of familial ALS cases. Silencing SOD1 using artificial microRNA has been shown to have therapeutic effects in mouse models of ALS. Here, Borel et al. tested the efficacy and safety of intrathecal delivery of SOD1-targeting artificial microRNAs in nonhuman primates. The therapy efficiently reduced SOD1 protein expression without side effects in macaques. The results suggest that this approach is safe and effective, paving the way for further development of this potential therapy.

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal neurological disease caused by degeneration of motor neurons leading to rapidly progressive paralysis. About 10% of cases are caused by gain-of-function mutations that are transmitted as dominant traits. A potential therapy for these cases is to suppress the expression of the mutant gene. Here, we investigated silencing of SOD1, a gene commonly mutated in familial ALS, using an adeno-associated virus (AAV) encoding an artificial microRNA (miRNA) that targeted SOD1. In a superoxide dismutase 1 (SOD1)–mediated mouse model of ALS, we have previously demonstrated that SOD1 silencing delayed disease onset, increased survival time, and reduced muscle loss and motor and respiratory impairments. Here, we describe the preclinical characterization of this approach in cynomolgus macaques (Macaca fascicularis) using an AAV serotype for delivery that has been shown to be safe in clinical trials. We optimized AAV delivery to the spinal cord by preimplantation of a catheter and placement of the subject with head down at 30° during intrathecal infusion. We compared different promoters for the expression of artificial miRNAs directed against mutant SOD1. Results demonstrated efficient delivery and effective silencing of the SOD1 gene in motor neurons. These results support the notion that gene therapy with an artificial miRNA targeting SOD1 is safe and merits further development for the treatment of mutant SOD1-linked ALS.

INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is a devastating, invariably fatal neurological disease caused by degeneration of motor neurons leading to rapidly progressive paralysis. It has an incidence of about 1.5 to 2.5 cases in 100,000 persons in the United States (1, 2) and in Europe (3), or up to about 30,000 new cases of ALS per year in those areas. Survival in ALS is typically 3 to 5 years. There is no cure for ALS, and the U.S. Food and Drug Administration (FDA)–approved treatments extend survival only modestly (4).

Familial ALS, which represents about 10% of all ALS cases, is inherited as a dominant trait. About 20% of these cases arise from mutations in the gene encoding cytosolic Cu/Zn superoxide dismutase 1 (SOD1) (5). An estimated 12 to 23% of patients with familial ALS and 1 to 3% of patients with sporadic ALS carry a mutation in this gene; 185 mutations in SOD1 have been identified (http://alsod.iop.kcl.ac.uk). Multiple mechanisms have been proposed to explain why mutant SOD1 proteins are neurotoxic, including the observation that mutant SOD1 acquires toxicity via conformational instability, misfolding, and some degree of aggregation (6). In turn, this activates multiple adverse events (AEs) that include the unfolded protein response (7), endoplasmic reticulum (ER) stress (8), mitochondrial damage (9), heightened cellular excitability (10), impaired axonal transport (11), and some elements of apoptotic (12) and necrotic (13) cell death. Some data suggest that misfolded mutant SOD1 protein can spread from cell to cell in a prion-like fashion (14). Additionally, it is proposed that mutant SOD1 can cause toxic misfolding of wild-type SOD1 (12, 15).

Because the pathology initiated by mutant SOD1 thus reflects acquired, gain-of-function properties, a potential strategy to treat SOD1-associated ALS is to suppress expression of the SOD1 gene. In recent years, we and others have investigated this strategy in depth using various modalities [reviewed in (16)]. In the present study, we elected to silence SOD1 in cynomolgus macaques using an artificial microRNA (miRNA) targeting multiple SOD1 mutations (17). We selected a recombinant adeno-associated viral vector serotype rh.10 (rAAVrh.10) because of its excellent central nervous system (CNS) transduction (18) and safety profile in nonhuman primates. In previous studies in adult SOD1G93A mice, a model of ALS, we demonstrated that SOD1 silencing profoundly delays disease onset and death and significantly preserved muscle strength and motor and respiratory functions (17). We further documented that intrathecal delivery of the candidate artificial miRNA in marmoset monkeys (Callithrix jacchus) safely silenced SOD1 in lower motor neurons (17).

Here, we have administered an artificial miRNA to the CNS in cynomolgus macaques (Macaca fascicularis) and tested the efficacy in silencing SOD1 expression and the safety profile.

RESULTS

Optimized intrathecal rAAV delivery at the lumbar spinal cord section is well tolerated and leads to widespread vector biodistribution and green fluorescent protein expression in primate spinal cord and brain

To silence SOD1, we used a previously developed artificial miRNA targeting multiple SOD1 mutations driven by polymerase II (pol II) or polymerase III (pol III) promoters, and delivered with a recombinant AAV serotype rh.10 (17). Here, we chose to work with large nonhuman primates, the cynomolgus macaques (M. fascicularis). To minimize intersubject variability and therefore allow for robust testing of the preclinical candidate, macaques for study were of the same gender and close in age and body weight (table S1). Ten days before infusing our therapy, a polyethylene-lined polyurethane catheter [outer diameter (OD), 1.0 mm; inner diameter (ID), 0.38 mm] equipped with a MIN-LOVOL subcutaneous titanium access port for infusions was placed via radiographic guidance; the catheter entered the intrathecal space and extended to the low thoracic region (Fig. 1A). After 10 days, the entry site was well healed, permitting slow infusion without backflow. Previous work showed that placing the subject in the head-down, Trendelenburg position improved CNS transduction (19). Accordingly, our animals were restrained in a prone position with the restraint table tilted about 30° head-down (Fig. 1B) for the duration of the dosing procedure. The animals were not anesthetized during this procedure; to allow better monitoring during the infusion, they were trained to be restrained in this position.

Fig. 1 Optimized intrathecal rAAV delivery at the lumbar spinal cord section is well tolerated and leads to widespread vector biodistribution and GFP expression in primate spinal cord and brain.

(A) Myelogram showing the magnified thoracolumbar region with the catheter tip, implanted at least 10 days before the vector delivery. (B) The subject is restrained and placed in the Trendelenburg position, at a 30° angle, for vector infusion. (C) Vector biodistribution throughout the CNS and peripheral organs determined by droplet digital polymerase chain reaction (ddPCR). CB-miR-SOD1 (n = 4); H1-miR-SOD1-GFP (n = 4); U6-miR-SOD1-GFP (n = 4). Error is represented as SD. (D and F) Detection of GFP protein expression in the cervical spinal cord (D) and brain (F). (E and G) GFP expression in lower motor neurons (E) and pyramidal neurons in the Brodmann area 4 in brain (G) of primates injected with the preclinical H1 vector. CSC, cervical spinal cord; LSC, lumbar spinal cord; TSC, thoracic spinal cord; CB, clinical CB vector; H1, preclinical H1 vector; U6, preclinical U6 vector.

We designed the study to also compare the use of both pol II (CB) and pol III (H1 and U6) promoters driving the expression of the artificial miRNA against SOD1 (miR-SOD1). The vector was administered as two 2.5-ml infusions through the intrathecal lumbar (IT-L) port/catheter system, the second infusion being separated from the first one by at least 6 hours to allow for cerebrospinal fluid (CSF) turnover. The animals were split into four groups, and each group received either no vector, a CB-miR-SOD1 [clinical CB, without green fluorescent protein (GFP)] vector, an H1-miR-SOD1 (preclinical H1, included GFP) vector, or a U6-miR-SOD1 (preclinical U6, included GFP) vector, as described in table S1.

The procedure was well tolerated during both the infusion and the postprocedure period, during which experienced animal caretakers frequently monitored the animals. No gross adverse side effects were observed during the course of the postprocedure monitoring; the treated animals demonstrated normal behavior and food intake. This injection protocol achieves widespread transduction in the spinal cord, brain, and peripheral organs, as shown by analysis of the vector biodistribution (Fig. 1C); the numbers of vector genome copies per diploid genome (vg/dg) ranged from 10 to 100 in the spinal cord.

The vector biodistribution was then confirmed at the protein level with analysis of the GFP expression in the spinal cord and in the brain. Our results show transduction in the spinal cord of both the anterior and posterior horns (Fig. 1, D and E, and fig. S1) and widespread transduction of the brain. Coronal brain sections show GFP staining patterns consistent with corticospinal tracts emanating from the cortex and traversing the cerebral peduncle, as well as the pontocerebellar fibers in the pons and oculomotor nucleus in the midbrain. We also observed GFP-positive staining of cortical layer V pyramidal cells in the primary motor cortex (Brodmann area 4) (Fig. 1, F and G, and figs. S2 and S3).

Delivery of preclinical H1 and U6 vectors leads to elevation of liver transaminases

Although the animals showed no behavioral evidence of toxicity from this infusion, we observed an elevation in liver transaminases outside of the normal range on day 22 in animals treated with both H1-miR-SOD1 and U6-miR-SOD1 vectors carrying the GFP sequence. Conversely, animals treated with CB-miR-SOD1 (without GFP) did not show liver transaminase increase. This was not present at day 3 after injection (necropsy samples; Fig. 2, A and B, and table S2). In previous clinical studies with AAV targeting the liver, an increase in serum transaminases has been associated with the reactivation of a response of cytotoxic T lymphocytes (CTLs) to AAV capsid (20, 21). This seemed unlikely in our study, because liver function elevation was observed in only two of the three groups, yet all three cohorts had received the same AAV capsids at equivalent doses. We therefore formulated the alternative hypotheses that the liver toxicity reflected either (i) the presence of the GFP protein or (ii) the high expression of miR-SOD1 precursors in the two cohorts with pol III promoters (H1 and U6). Conceivably, the miRNA precursors might saturate the endogenous miRNA pathway. To discriminate between these two hypotheses, we analyzed the cellular immune response to GFP protein with a standard interferon γ (IFN-γ)–ELISpot assay. Spleens were harvested at necropsy, and splenocytes were isolated after enzymatic dissociation (22 days after dosing). These cells were expanded for 6 days before running the ELISpot assay. For the IFN-γ–ELISpot assay, splenocytes were stimulated for 48 hours with overlapping peptides spanning the entire GFP protein sequence and divided into three pools. Two of eight animals developed a cellular immune response to the GFP protein detectable by ELISpot (fig. S4).

Fig. 2 Delivery of preclinical H1 and U6 vectors leads to elevation of liver transaminases.

(A and B) Aspartate aminotransferase (AST) (A) and alanine aminotransferase (ALT) (B) quantification in serum after vector injection. (C) SOD1 expression in the liver determined by ddPCR and normalized to HPRT. Unpaired two-tailed t test; CB versus H1, P ≤ 0.0045 and CB versus U6, P ≤ 0.0447. (D) Mature miR-SOD1 expression in the liver quantified by ddPCR and normalized to U47. (E) Hepatocellular pathology score [two-tailed analysis of variance (ANOVA) CB versus H1 and U6, P ≤ 0.0001] and (F) Ki67 score (two-tailed ANOVA; CB versus H1 and U6, P ≤ 0.0178). Error is represented as SEM. Control, uninjected animals (n = 3); CB (n = 4); H1 (n = 4); U6 (n = 4). *P ≤ 0.05; **P ≤ 0.01; ****P ≤ 0.0001.

An in-depth analysis of the liver phenotype was then performed, both to identify the pathological basis for the observed liver toxicity and to assess efficacy of SOD1 suppression. Silencing of SOD1 in the liver was determined by ddPCR (Fig. 2C). Silencing was more complete with the pol II (CB) promoter than with the pol III (H1 and U6) promoters; as compared to untreated animals, the mean relative SOD1 expression was 0.08 in the CB group, 0.54 in the H1 group, and 0.33 in the U6 group (H1 versus CB, P ≤ 0.0045; U6 versus CB, P ≤ 0.0447). Next, mature miR-SOD1 molecules were quantified by ddPCR (Fig. 2D). Mature miR-SOD1 was detected in the liver of the animals treated with the pol II (CB) promoter and with the pol III (H1 and U6) promoters; compared to controls, the mean relative miR-SOD1 expression was 22.44 in the CB group, 0.19 in the H1 group, and 0.36 in the U6 group (statistical analysis is provided in table S3). Liver samples were subsequently subjected to a blind analysis by an experienced pathologist. Scoring of the liver histopathology (severity scores for portal mononuclear infiltrate, hepatocyte cell death, and hepatocyte regeneration) revealed a significant increase in the scoring (reflecting increased pathology) in the two pol III–treated groups compared to the untreated control animals (Fig. 2E; P ≤ 0.0001). Additionally, a Ki67 staining was performed and Ki67-positive cells were counted (Fig. 2F). The Ki67 immunostaining score, which reflects the degree of cellular proliferation, was significantly increased in the H1 and U6 groups compared with untreated controls (P ≤ 0.0178). Hepatocellular damage and Ki67 scores were positively correlated (P ≤ 0.0001; fig. S5). Although the pathology results confirmed the liver injury, they did not elucidate which of the two hypothetical causes for the liver toxicity was driving this process (GFP toxicity or pol III–mediated toxicity) (22, 23). However, quantification of the expression of miR-122, the most abundant endogenous miRNA in the liver, revealed stable miRNA expression among treatment groups (fig. S6), suggesting that the toxicity is not caused by saturation of the endogenous miRNA machinery due to high expression of miR-SOD1 from the pol III promoters.

SOD1 is profoundly and reproducibly silenced in motor neurons

The mature miRNA, miR-SOD1, was quantified in the spinal cord at three sections: lumbar (LSC), thoracic (TSC), and cervical (CSC) (Fig. 3A). The SD within each group was limited, suggesting that there was good technical reproducibility. The results show that the quantity of miR-SOD1 increased with the relative strength of the promoter (CB < H1 < U6) (Fig. 3A). Ninefold higher miR-SOD1 expression levels were detected in the H1 group compared with the CB group (P ≤ 0.0003), and a 15-fold increase was found for the U6 group compared with the CB (P ≤ 0.0002). Last, to precisely assess silencing in the motor neurons, which are the primary cellular target for this therapy, an average of 400 cells were microdissected using laser capture. The resulting expression profile in motor neurons demonstrated significant silencing (Fig. 3B and table S3). Overall, the extent of the silencing positively correlated with the strength of the promoter (CB < H1-U6).

Fig. 3 SOD1 is profoundly and reproducibly silenced in motor neurons.

(A) Relative mature miR-SOD1 expression in the spinal cord determined by ddPCR and normalized to U47. Unpaired two-tailed t test; control TSC versus CB TSC, P ≤ 0.0188; control TSC versus H1 TSC, P ≤ 0.0036; control LSC versus H1 LSC, P ≤ 0.0034; control CSC versus U6 CSC, P ≤ 0.0105; control TSC versus U6 TSC, P ≤ 0.0162. (B) Relative SOD1 expression in laser-capture microdissected motor neurons quantified by ddPCR and normalized by HPRT. Unpaired two-tailed t test; control LSC versus CB LSC, P ≤ 0.0014; control CSC versus H1 CSC, P ≤ 0.0034; control TSC versus H1 TSC, P ≤ 0.0009; control LSC versus H1 LSC, P ≤ 0.0001; control CSC versus U6 CSC, P ≤ 0.0097; control TSC versus U6 TSC, P ≤ 0.0007; control LSC versus U6 LSC, P ≤ 0.0002. Error is represented as SEM. (C and D) GFP and SOD1 immunostaining in the lumbar spinal cord of control and vector-injected animals. DAPI, 4′,6-diamidino-2-phenylindole. Control, uninjected animals (n = 3); CB (n = 4); H1 (n = 4); U6 (n = 4); Scale bars, 50 μm. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Our results show a 1.7-fold (P ≤ 0.0053) and 1.6-fold (P ≤ 0.0074) decrease in the H1 and U6 groups compared with the CB group, respectively, in the CSC, a 2.1-fold (P ≤ 0.0152) and 2.5-fold (P ≤ 0.0052) decrease in the TSC, and a 3.6-fold (P ≤ 0.0033) decrease in H1 versus CB in the LSC. There was no difference between the CB and the U6 group in the LSC. Overall, differences between control versus H1 and control versus U6 throughout the spinal cord were statistically significant, and differences between control versus CB were statistically significant in the LSC (table S3).

SOD1 silencing was generally lower in the cervical section (CSC), intermediate in the thoracic section (TSC), and higher in the lumbar section (LSC), with up to 93% silencing in the LSC of the H1 group (table S3). These results were then confirmed at the protein level. LSC sections were stained with anti-GFP and anti-SOD1 antibodies. As expected, an uninjected control primate shows no GFP expression but strong SOD1 expression in motor neurons (Fig. 3C). Animals injected with the preclinical H1 vector (that included GFP) show GFP expression in most but not all motor neurons, confirming that rAAVrh.10 targets motor neurons. The GFP and SOD1 co-staining shows that the motor neurons expressing the GFP protein are not expressing SOD1, whereas GFP-negative motor neurons in the same tissue section do have SOD1 protein (Fig. 3D).

GFP-devoid clinical constructs are not linked to elevation of liver transaminases

A second experiment was subsequently carried out using the same delivery protocol. The goals of this experiment were to test whether the GFP was indeed the cause of the aforementioned mild liver toxicity and to demonstrate safety of the GFP-devoid clinical vectors. This experiment also afforded an opportunity to assess the impact of preexisting neutralizing antibodies (NAbs) to AAV in the periphery on intrathecal AAV injection. The animals were split into four groups; each group received either no vector, an H1-miR-SOD1 (clinical H1) vector (studied at day 43), an H1-miR-SOD1 (clinical H1) vector (studied at day 92), or a CB-miR-SOD1 (clinical CB) vector (also evaluated at day 92), as described in table S4. To determine the impact of preexisting NAb on vector biodistribution, we noted that three of seven animals selected in the H1-miR-SOD1 (short-term) group were seropositive for AAVrh.10 before dosing (table S5). Analysis of the CNS and peripheral organs revealed a widespread biodistribution of the virus in the spinal cord, brain, and peripheral organs (Fig. 4A). These results show that there was neither a difference in the biodistribution of the AAV in the CNS in the group with preexisting NAbs (Table 1) nor a reduction of vector genomes per cell in the spinal cord (table S6). It should be noted that the first study ended after only 22 days, whereas the second study had day 43 and day 92 end points. To verify that there is no loss of vector genome due to a clearance of extracellular AAV genomes between day 22 and days 43 and 90 in transduced motor neurons, which are the main target in this study, we performed a biodistribution on 200 laser-captured motor neurons from animals injected with preclinical or clinical H1 vectors and euthanized at days 22 and 92, respectively. The results showed no difference in viral genome copy number per diploid genome in these cells at the two time points (table S7).

Fig. 4 GFP-devoid clinical constructs are not linked to elevation of liver transaminases.

(A) Vector biodistribution throughout the CNS and peripheral organs determined by ddPCR. Error is represented as SD. (B and C) AST (B) and ALT (C) expression in serum after vector injection. (D) SOD1 silencing in the liver determined by ddPCR and normalized to HPRT. Error is represented as SEM. Control, uninjected animals (n = 3); CB (n = 3); H1 [n = 7 (43 days) and n = 4 (92 days)]. The numbers in parentheses below the vector name denote the number of days at end point for each group.

Table 1 Preexisting NAbs have no impact on spinal cord transduction and silencing.

MN, motor neurons; no NAb, low NAb titer before dosing; NAb+, animal positive for NAb before dosing.

View this table:

In this second primate study, blood was sampled frequently (at days 8, 15, 22, 29, and 43) to allow for close monitoring of clinical chemistry parameters. Liver transaminases remained within the normal range and were stable throughout the course of the study (Fig. 4, B and C). These results suggest that the previously observed liver toxicity was not linked to the pol III promoters but was likely caused by the GFP (CB versus H1; first study, day 22: ALT, P ≤ 0.0001; AST, P ≤ 0.0001; second study, days 22, 29, and 43: ALT, not significant; AST, not significant). Analysis of liver SOD1 expression is shown in Fig. 4D.

Clinical constructs led to reproducible SOD1 silencing

To assess silencing of SOD1 in motor neurons, cells were laser captured and used to isolate RNA and subsequently quantify SOD1 expression (Fig. 5, A to C). At 92 days after injection, silencing exceeded 30% in the CB group and 60% in the H1 group in the lumbar section (Fig. 5A). An intermediate degree of silencing was observed in the thoracic section (Fig. 5B), and in the cervical section, the difference between groups was reduced, with about 35% silencing in the CB group and about 45% silencing in the H1 group (Fig. 5C). The amount of mature miR-SOD1 was detected only in the treated groups (Fig. 5, D to F) and was fairly stable along the spinal cord in the lumbar (Fig. 5D), thoracic (Fig. 5E), and cervical (Fig. 5F) sections. The difference in miR-SOD1 expression between pol II and pol III groups at 92 days is fivefold in the lumbar section (P ≤ 0.0155, table S8), and there was no difference further away from the injection site (thoracic and cervical). SOD1 silencing was confirmed by branched fluorescence in situ hybridization (FISH) (RNAscope) assay where transduced motor neurons show an important decrease in SOD1 signal (Fig. 5G).

Fig. 5 Clinical constructs led to reproducible SOD1 silencing.

SOD1 silencing measured by ddPCR in laser-capture microdissected motor neurons at the lumbar [(A) unpaired two-tailed t test; control versus H1 (43), P ≤ 0.0001; control versus H1 (92), P ≤ 0.0008; CB (92) versus H1 (92), P ≤ 0.0284], thoracic [(B) unpaired two-tailed t test; control versus H1 (43), P ≤ 0.0045; control versus H1 (92), P ≤ 0.0087], and cervical [(C) unpaired two-tailed t test; control versus CB P ≤ 0.0457, control versus H1 (43) P ≤ 0.0067, control versus H1 (92) P ≤ 0.0099] section of the spinal cord. Mature miR-SOD1 detection in primates by ddPCR at the lumbar [(D) unpaired two-tailed t test; control versus CB, P ≤ 0.0131; control versus H1 (92), P ≤ 0.0038; CB (92) versus H1 (92), P ≤ 0.0155], thoracic [(E) unpaired two-tailed t test; control versus CB, P ≤ 0.0094], and cervical (F) section of the spinal cord. Control, uninjected animals (n = 3); CB (n = 3); H1 [n = 7 (43 days) and n = 4 (92 days)]. The numbers in parentheses below the vector name denote the number of days at end point for each group. (G) Motor neurons stained by branched FISH (RNAscope) assay, ChAT (choline acetyltransferase; red), vector genomes (vg, green), and SOD1 (yellow). Scale bars, 10 μm. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001.

Clinical constructs are safe

Last, small RNA sequencing (RNA-seq) was performed to more precisely examine the small RNA species generated by the vector-derived artificial miRNA at 92 days after treatment. Depending on the flanking sequences used (here that of hsa-miR-155), miRNA processing can vary. Important parameters for safety include the abundance of the predicted mature miRNA over alternate species (which should be high, hence avoiding potential off-targeting), as well as the guide/passenger strand ratio (which should be high as well, again limiting the risk of off-targeting). Although preliminary work using these flanking sequences to silence another target demonstrated precise processing (24), we sought to confirm these findings in the context of miR-SOD1 in an in vivo setting by small RNA-seq. Alignment of all reads mapping to the miR-SOD1 precursor showed that the most abundant species for both the guide and the passenger strand are the predicted sequences (Fig. 6A). In the case of the guide strand, the predicted species was 100-fold more abundant than alternate species (Fig. 6A). Moreover, the predicted/most abundant mature guide was about 100-fold more abundant than the predicted/most abundant mature passenger (Fig. 6A). Quantification of mature miR-SOD1 (guide strand) showed that, in vivo, the H1 promoter expressed about fivefold more miR-SOD1 than the CB promoter (Fig. 6B, P ≤ 0.0030).

Fig. 6 Clinical constructs are safe.

(A) Small RNA-seq catalog of all the artificial miRNA species generated in vivo by alternative processing of the vector-encoded construct. CB, n = 3; H1, n = 4. (B) Comparison of the expression of mature miR-SOD1 generated in vivo by the CB or the H1 vector. CB, n = 3; H1, n = 4. Unpaired two-tailed t test; CB versus H1, P ≤ 0.003. Analysis of potential off-target effects by ddPCR on control animals (n = 3 to 4) and animals injected with H1 clinical vector (n = 4, 43 days) at the cervical (C) and lumbar (D) sections of the spinal cord and in the liver (E). Each gene was normalized to HPRT. Error is represented as SEM. Unpaired t test; control versus H1.

We then analyzed potential off-target effects in the group injected with the clinical H1 vector and euthanized at day 43 after injection. We identified four potential off-target genes on the basis of high sequence identity (52 to 71%) between the 21-nucleotide-long mature miR-SOD1 and the messenger RNA of genes cataloged in the M. fascicularis RefSeq_RNA database (table S9). We analyzed the expression levels of these genes (Fig. 6, C to E). This analysis was performed on the CSC (Fig. 6C) and LSC (Fig. 6D), as well as on the liver (Fig. 6E) (because the liver demonstrated the highest vector copy number per cell in periphery). Our results show no off-target silencing in terms of mRNA relative expression for each target comparing uninjected animals to animals injected with the clinical H1 vector.

Intrathecal AAV delivery results in peripheral immune responses

Immune responses to AAV capsid are a major concern in clinical trials (20, 21). Initiation or reactivation of a T cell response to AAV capsid as well as high titer of NAbs can prevent AAV transduction and vector readministration. In this study, we monitored both NAb response and cellular immune response to AAV capsid in all animals (Table 2). NAbs were detected in the serum and in the CSF after gene transfer in all animals seronegative before dosing, and we observed an increase in titers for the animals seropositive before gene transfer (from 1:80−1:320 to 1:640−1:5120) (table S5). We also observed that the presence of preexisting NAb in the periphery did not affect the CNS vector biodistribution or the silencing in our model; the vector copy number and the silencing were comparable between the two groups (clinical H1, no NAb versus clinical H1, NAb+) in the three sections of the spinal cord (lumbar, thoracic, and cervical; Table 1).

Table 2 NAb responses and cellular immune responses to AAVrh.10 after intrathecal injection.

PBMC, peripheral blood mononuclear cell.

View this table:

The T cell response to AAVrh.10 was also monitored by a standard IFN-γ–ELISpot assay to determine the number of antigen-specific IFN-γ–secreting cells in response to stimulation with AAV.rh10 VP1 peptide pools. PBMCs were collected before dosing and at necropsy, and splenocytes were collected at necropsy. None of the animals showed a positive response to AAVrh.10 capsid proteins after PBMC restimulation with AAV.rh10 VP1 peptide pools for 48 hours. Four of 28 animals showed a cellular immune response detectable by ELISpot when splenocytes were expanded for 6 days and restimulated for 48 hours with AAV.rh10 VP1 peptide pools. The response was low, between 55 and 200 spot-forming units (SFU) per million cells (Table 2). Three of them were positive at day 22, one at day 43, and none at day 92, suggesting that the cellular immune response was not systematic, could be transient, and did not persist few months after gene transfer. These data confirmed that a NAb response to AAVrh.10 occurred after intrathecal delivery and did not correlate with the cellular immune response. The results also suggest that these responses did not affect gene transfer efficacy.

DISCUSSION

The therapeutic silencing of SOD1 has been pursued by many groups, using various modalities: antisense oligonucleotides (ASOs) (25), RNA interference (RNAi) (26), viral vector-delivered RNAi (17, 18, 2733), and CRISPR-Cas9 (34). From a clinical perspective, one of the major disadvantages of ASOs and small interfering RNAs is the repeated dosing of the patients, whereas rAAV-mediated gene therapy (including gene transfer and RNAi-based gene silencing) relies on a one-time dosing paradigm. Technological improvements allow the ASO doses to be less frequent than in the past as illustrated, for example, by nusinersen (Spinraza), a recently approved ASO developed as a treatment for spinal muscular atrophy (SMA) by Biogen and Ionis Pharmaceuticals. With this drug, a typical patient would receive three intrathecal doses yearly upon completion of the loading doses [estimated half-life of 135 to 177 days in CSF and 63 to 87 in plasma (nusinersen US Label); for updates, see FDA index page for NDA 209531 (www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview.process&varApplNo=209531)], and this would be continued for life to maintain therapeutic benefit. In contrast, as an example, AVXS-101, a gene therapy treatment developed by AveXis as a treatment for SMA type 1, has a therapeutic effect for up to 24 months after a single intravenous injection of a rAAV9 vector (35). Although AVXS-101 has not been approved by the FDA, all the five AEs reported with this therapy consisted of clinically asymptomatic liver enzyme elevations. These types of AE have been observed with other gene therapy trials (21, 36). An additional difference between these two modalities in terms of safety remains, as in the case of an AE, an ASO may be discontinued whereas gene therapy may not (one-time treatment). However, the field has significantly advanced over the past 20 years, particularly with the development of AAVs, which have an excellent safety profile in both animal and human studies (37). The work presented here builds upon previous studies demonstrating that suppression of SOD1 expression ameliorates the disease course in a mouse model of ALS, preserving motor and respiratory functions (17). Additionally, it extends our observation that in nonhuman primates, intrathecal delivery of rAAVrh.10 harboring a SOD1-targeting artificial miRNA reduces SOD1 expression throughout the spinal cord (17). In the present study, we have optimized delivery of the gene therapy vector in cynomolgus macaques to accommodate a larger volume of viral vector and ensure reproducible and widespread biodistribution of the viral vector and the miRNA. We showed here that intrathecal infusion of a large volume of rAAV vector (5 ml) was safe and well tolerated by the cynomolgus macaques up to 92 days after administration.

A distinctive and clinically important finding in this study is the high amount of vector genomes in the periphery after intrathecal vector delivery. Given the relevance of transduction efficacy on therapeutic properties of AAV-based therapies, these results highlight the clinical relevance of the strategy developed in this study.

The presence of GFP in our vectors caused mild liver toxicity, as previously described (38), and a cellular immune response in two of eight animals. The fact that the immune response is not detected in all the injected animals can be explained by the early sacrifice point (22 days). A previous study in nonhuman primates injected intravenously with an AAV vector reported transaminase elevation between weeks 3 and 4 after injection and a delayed cellular immune response to GFP between weeks 4 and 5 after dosing (38). This hepatotoxicity, evident both serologically and pathologically, was not detected when GFP was eliminated from the AAV construct. The interventions that excluded GFP were safe without evidence of RNAi-mediated toxicity, regardless of whether the artificial miRNA miR-SOD1 is expressed from a pol II (CB) or pol III (H1) promoter. The absence of toxicity is particularly noteworthy given the relatively high vector load per cell. We speculate that it reflects the fact that the RNAi-mediating molecule of choice is an artificial miRNA and not a short hairpin RNA; the latter has been associated with cellular miRNA-related toxicity in previous reports (22, 23). Also important in the safety profile in our studies is the processing of mature miR-SOD1, which affects the possibility of off-target toxicity. Inaccurate processing of the premature-to-mature miRNA (where the mature sequence could “slip” toward the 5′ or 3′ end) would lead to alternate mature sequences, as would higher abundance of the passenger strand versus the guide strand (where the complementary sequence would be in higher abundance than the guide strand that targets SOD1). Both of these events could lead to sequence identity of the mature miR-SOD1 with transcripts other than the one intended, also known as off-target effects. The accurate processing and the favorable guide/passenger strand ratio we observed in our studies contribute to the relative derisking of the off-target toxicity.

Our SOD1 silencing data demonstrate a broad correlation between the amount of the miRNA and the resulting SOD1 gene silencing. Thus, silencing measured in motor neurons was enhanced by use of the pol III promoters (H1 and U6), which drive higher expression of the artificial miRNA than the CB promoter. Using the H1 promoter, we achieved up to 93% silencing of SOD1 at the lumbar section. More broadly, silencing by the pol III promoters was in the range of 50%, whereas the pol II (CB) promoter achieved a more limited range of silencing, up to 68%. We note as well that an apparent spatial difference in silencing (lumbar greater than cervical cord) is less evident after longer periods of follow-up in these animals. A difference in silencing with the CB vector between the lumbar section and the thoracic-cervical sections evident after 22 days (P ≤ 0.0110) was not seen in the 92-day study (P ≤ 0.6300). At 92 days, the extent of SOD1 silencing is comparable throughout the entire spinal cord. This observation may be a consequence of the kinetics of onset of promoter activity, whereby a later end point may allow higher expression of the effector miRNA miR-SOD1. If confirmed, this point would be important in the design of preclinical and clinical studies.

An expanded access clinical trial of this intervention has been authorized by the FDA and is now under way (IND# 17179). The pilot human trial will ascertain whether the safety profile in humans parallels that described above in nonhuman primates. An important limitation of the present nonhuman primate trial is that it may not accurately predict results in the human study. For example, the cellular immune response characterized by IFN-γ–secreting T cells in humans cannot be faithfully modeled in animals because the preclinical models do not mimic initiation or reactivation of CTLs to AAV capsid as observed in humans (39). Our present study shows that low titers of NAb (between 1:80 and 1:320) before dosing do not have an obvious impact on spinal cord transduction, suggesting that low NAb titers might be acceptable for future clinical studies and that it may not be necessary to exclude seropositive patients from intrathecal AAV therapies. This view is consistent with studies of the impact of systemic NAb to AAV on intrathecal delivery of rAAV in nonhuman primates (40, 41). These issues notwithstanding, our present findings support cautious optimism that intrathecally delivered, rAAV-mediated silencing of SOD1 may offer meaningful therapy for mutant SOD1-associated familial ALS. If, as some data suggest, misfolding of wild-type SOD1 participates in the pathogenesis of sporadic ALS, the same therapy may be more widely applicable to that population as well.

MATERIALS AND METHODS

Study design

The purpose of this study was to explore the safety of an AAV gene therapy candidate encoding for an artificial miRNA targeting SOD1 in a preclinical model. We also compared the use of pol II (CB) and pol III (H1 and U6) promoters driving the expression of our miRNA. A total of 32 male cynomolgus monkeys close in age and body weight were included in both studies [first study, n = 12, n = 4 per group (table S1); second study, n = 16, n = 3 to 4 per group (table S3)]. Monkeys were randomly assigned to different groups except for the group seropositive for NAb to AAV.rh10. Animals received either clinical CB, preclinical H1 (that included GFP), or preclinical U6 (that included GFP) in the first study and were followed for 22 days after dosing to prevent any toxicity related to GFP protein expression. In the second study, animals were injected with either clinical CB or clinical H1 (without GFP) and were followed for 43 or 92 days. The vector was administered as two 2.5-ml infusions through the IT-L port/catheter system, the second infusion being separated from the first one by at least 6 hours. All animal experiments were approved by the Institutional Animal Care and Use Committees and were conducted in agreement with institutional guidelines. First, we investigate the vector biodistribution in the spinal cord, brain, and peripheral organs by determining viral vector copy number per cell in each tissue and analyzing GFP expression by immunostaining. Second, we analyzed silencing of SOD1 in the different sections of the spinal cord (cervical, thoracic, and lumbar) as well as in laser-captured motor neurons at the mRNA level using ddPCR and branched FISH (RNAscope) assay. Last, we evaluated the safety of the clinical vector by analyzing the abundance of mature miR-SOD1 (guide strand) by small RNA-seq, off-target effects, and the humoral and cellular immune responses to AAV capsid using NAb assay and ELISpot, respectively. Experiments were carried out in an unblinded fashion using technical replicates and at least two independent repeats.

Gene therapy vector

The miR-SOD1 sequence was previously generated and validated (17). Briefly, the sequence of the guide strand matches nucleotides 275 to 295 of RefSeq NM_000454.4 (located in exon 2). Research grade plasmid (endotoxin < 100 EU/mg) was produced by Aldevron. Recombinant AAV serotype rh.10 was produced, and titers of vector batches were determined by the University of Massachusetts Vector Core.

Animal experiments

All animal experiments were approved by the Institutional Animal Care and Use Committees and were conducted in agreement with institutional guidelines. The experiments were conducted at and by Northern Biomedical Research (NBR). Fifteen animals were included in the first study (table S1) and 17 were included in the second study (table S3), after being prescreened for NAbs to the AAV vector by the University of Massachusetts Vector Core. Animals were preimplanted with an IT-L catheter. For catheter implantation, the animals were pretreated with atropine sulfate (0.04 mg/kg) and sedated with ketamine (8 mg/kg). The animals were then intubated and maintained on oxygen (about 1 liter/min) and about 2.0% isoflurane. Prednisolone sodium succinate (30 mg/kg) and flunixin meglumine (2 mg/kg) were administered before surgery. For the IT-L catheter, a hemilaminectomy was made at approximately the L5 vertebra, and the catheter (polyethylene-lined polyurethane; OD, 1.0 mm and ID, 0.38 mm) was inserted and advanced intrathecally with the tip located near the thoracolumbar junction. The catheter was secured with a suture. The proximal end of the catheters terminated in a subcutaneous titanium access port (MIN-LOVOL; Solomon Scientific). The actual locations of the catheter and catheter tip were verified by a myelogram with Isovue 300. Upon recovery from anesthesia, the animals were provided butorphanol tartrate (0.05 mg/kg) for analgesia and placed on postsurgical antibiotic ceftiofur sodium (5.0 mg/kg, one injection during or before surgery followed by three injections). A recovery period of at least 10 days was provided before dosing. For the duration of the dosing procedure, the animals were restrained in a prone position with the restraint table tilted about 30° head-down. Animals were trained to the restraint device before dose administration. The vector was administered as two 2.5-ml infusions through the IT-L port/catheter system. The infusion rate was about 7.5 ml/hour over about 20 min. A calibrated syringe infusion pump and primed extension set were used. The test article was followed by about 0.3 ml of vector excipient to flush the dose from the catheter system. For the necropsy, the animals were sedated with ketamine (8.0 mg/kg), maintained on an isoflurane/oxygen mixture, and provided with an intravenous bolus of heparin sodium (200 IU/kg). The animals were perfused via the left cardiac ventricle with 0.001% sodium nitrite in chilled saline. Organs were harvested and either fixed in 10% neutral buffered formalin (NBF) for possible histopathological analysis or frozen in liquid nitrogen for bioanalytical analysis and stored at −60°C or below. The spinal cord with dorsal root ganglia was removed intact and sectioned into cervical, thoracic, and lumbar segments. A 0.5-cm sample was collected from the rostral end of each residual section and saved in 10% NBF for possible histopathological analysis. The remaining tissue was frozen for bioanalytical analysis in 0.5-cm sections and stored at −60°C or below. The brain was cut in a brain matrix at 3-mm coronal slice thickness. The first slice and every other slice thereafter will be saved in 10% NBF for possible histopathological analysis. The second slice and every other slice thereafter was frozen for bioanalytical analysis and stored at −60°C or below. NBR was also responsible for the clinical chemistry.

Isolation of nucleic acids

Genomic DNA was isolated using Gentra Puregene kit (Qiagen) from flash-frozen tissues, according to the manufacturer’s protocol. Frozen tissue was homogenized in TRIzol (Thermo Fisher Scientific), and total RNA was isolated according to the manufacturer’s instructions. The protocol used for RNA isolation from laser-capture motor neurons is described in the dedicated section.

Laser-capture microdissection

Spinal cords were stored at −60°C or below until cryosectioning. Tissue was sectioned at 20 μm using a cryostat and mounted on silane-treated slides (catalog no. 16004-406, VWR). Slides were stored at −60°C or below until used for staining. Individual slides were thawed and stained following the manufacturer’s instructions (catalog no. KIT0401, Thermo Fisher Scientific). Slides were then stored in xylene until used for capture. About 200 stained motor neuron cell bodies from the ventral horn were laser microdissected (ArcturusXT, Thermo Fisher Scientific), collected onto a single cap (catalog no. LCM0211, Thermo Fisher Scientific, and immediately lysed in 50 μl of extraction buffer (catalog no. KIT0204, Thermo Fisher Scientific) for 30 min at 42°. Samples were stored at −60°C or below until RNA isolation. RNA was isolated following the manufacturer’s instructions (catalog no. KIT0204, Thermo Fisher Scientific) and eluted in 11 μl of nuclease-free water (catalog no. AM9937, Thermo Fisher Scientific). The quality of the RNA was determined on an Agilent 2100 Bioanalyzer or on a Fragment Analyzer (Advanced Analytical Technologies Inc.). RNA samples with an RNA Integrity Number (RIN) or RNA Quality Number inferior to 8 were strictly excluded (table S10). Two RNA samples of satisfactory quality (designated #1 and #2 in table S10) were pooled, and this pooled RNA corresponding to about 400 motor neurons constituted the sample used for downstream expression analysis.

Biodistribution

ddPCR was performed according to the manufacturer’s recommendations using 50 ng of DNA as input.

Expression analysis . RNA was appropriately retrotranscribed (messenger RNA: mix of random hexamer and oligo dT; small RNA: stem-loop primers), and ddPCR reactions were prepared and run as described in the manufacturer’s instructions, using gene-specific primers/probes (table S11). For gene expression, custom assays were designed to detect SOD1 (FAM-labeled, target) and hypoxanthine-guanine phosphoribosyltransferase (HPRT) (HEX-labeled, normalizer) and ordered from Bio-Rad. For small RNAs, a custom assay was designed to detect miR-SOD1 (target), and a commercial assay was ordered to detect U47 (normalizer); both were FAM labeled and ordered from Thermo Fisher Scientific. To be considered valid, each run had to have at least 10,000 droplets, of which at least 100 are negative.

Small RNA-seq

Ventral horn tissue was sampled by 2-mm biopsy punch and homogenized in TRIzol (Thermo Fisher Scientific), and RNA was isolated according to the manufacturer’s protocol. RNA with RIN > 8 was used to generate small RNA libraries that were subsequently sequenced (BGI). For data analysis, reads that mapped to the miR-SOD1 precursor were aligned, attributed to the guide or passenger strand, and quantified.

Immunostaining

Frozen-fixed spinal cord and brain tissues were cryosectioned at a thickness of 40 μm and stained as floating sections. Sections were fixed in 0.5% H2O2 and 10% methanol for 15 min, heated for 30 min at 80°C in 10 mM sodium citrate buffer (pH 8.5), and blocked with avidin and biotin block for 10 min each. Blocking consisted in a 1-hour incubation in 10% normal goat serum (NGS) (Jackson ImmunoResearch Laboratories Inc.). Sections were then incubated for 5 days at 4°C in primary rabbit anti-GFP (1:2500, Invitrogen) antibody diluted in 10% NGS. Secondary antibody consisted in incubation for 2 hours at room temperature (RT) in alkaline-phosphatase goat anti-rabbit immunoglobulin G (IgG) H&L (1:200, Vector Labs). Substrate was added for 20 to 30 min at RT. Sections were mounted with phosphate-buffered saline and dried in a desiccator for 2 days before dehydration and coverslipping. Images were then captured using a Leica DM5500 B microscope and Leica Application Suite AF 3.1.0 build 8587 imaging software.

GFP and SOD10 immunofluorescence staining

Frozen-fixed spinal cords were cryosectioned at a thickness of 40 μm and stained as floating sections. Briefly, sections were fixed in 0.5% H2O2 and 10% methanol for 15 min, heated for 30 min at 80°C in 10 mM sodium citrate buffer (pH 8.5), and blocked with avidin and biotin block for 10 min each. Blocking consisted in a 1-hour incubation in 10% NGS (Jackson ImmunoResearch Laboratories Inc.) and a 40-min incubation in 2.5% RTU horse serum (Vector Labs). Sections were then incubated for 4 days at 4°C in primary rabbit anti-GFP (1:2500, Invitrogen) and goat anti-SOD1 (1:2500, R&D Systems) antibodies diluted in 10% NGS. Secondary antibody consisted in incubation for 2 hours at RT in goat anti-rabbit IgG H&L (1:1000, Abcam) and donkey anti-goat IgG H&L (1:1000, Abcam). DAPI was applied, and sections were mounted with fluorescent mounting medium (PermaFluor, Thermo Scientific). Images were captured using a Leica DM5500 B microscope and Leica Application Suite AF 3.1.0 build 8587 imaging software.

RNA multiplexed branched FISH

Frozen spinal cord tissue was cryosectioned at a thickness of 20 μm. Multiplexed branched FISH was performed according to the manufacturer’s recommendations (ACDBio). Briefly, the sections were fixed for 15 min at 4°C in 10% NBF and pretreated with a protease-based solution for 30 min at RT to increase availability of the target mRNA. DNA probes were hybridized with the target mRNAs for 1 hour at 40°C, and bound probe signal was subsequently amplified through branching before imaging on a Leica DM5500 B microscope using Leica Application Suite.

Pathology analysis

Fixed tissue was sectioned and stained at the University of Florida Molecular Pathology Core. The pathology analysis was carried out blindly by an experienced pathologist at Pathogenesis, LLC.

Off-target analysis

Targets were identified by BLASTn against M. fascicularis RefSeq_RNA database, using the mature miR-SOD1 as input. The search was performed on 27 March 2018 and gave 2135 results. Results in predicted genes were excluded, and the validated genes with a percentage of identity higher than 50% were selected. Additionally, genes for which the identity included at least part of the seed region of miR-SOD1 were considered of particular relevance. RNA from spinal cord (cervical and lumbar sections) and liver was extracted using TRIzol according to the manufacturer’s protocol. RNA was then deoxyribonuclease treated using a TURBO DNA-free kit (Ambion), and reverse transcription was performed on 100 ng of RNA using a High-Capacity RNA-to-cDNA kit (Applied Biosystems). ddPCR was then run as described by the manufacturer using QX200 ddPCR EvaGreen Supermix (Bio-Rad). FAS and POLR2H were custom designed, and FSTL1 (Unique Assay ID: dHsaEG5020261) and RNF146 (catalog no. 10031258, Unique Assay ID: dHsaEG5008535) were ordered from Bio-Rad. All these targets were normalized using HPRT1 (Unique Assay ID: dHsaEG5189658).

NAb assay

Sera sampled before dosing and at necropsy and CSF sampled at necropsy were heat inactivated for 30 min at 56°C. AAVrh.10-CMV-LacZ (109 gc per well) was incubated with twofold serial dilutions of samples for 1 hour at 37°C and 5% CO2. The mixture was added to Huh7 cells (1 × 105 cells) previously infected with an adenovirus (100 particles per well) and incubated for 18 to 22 hours at 37°C and 5% CO2. Cells were washed and developed with Galactor-Star kit (Applied Biosystems). Luminescence was measured with a luminometer. The NAb titers are expressed as the highest dilution that inhibited β-galactosidase expression by at least 50% compared to a negative mouse serum control.

IFN-γ–ELISpot assays

Immune cellular immune responses to AAVrh.10 capsid were analyzed by ELISpot assay by measuring IFN-γ secretion after PBMC or splenocyte restimulation in vitro. PBMCs were isolated before dosing and at different time points after injection. Splenocytes were isolated at necropsy. Cells were stimulated in vitro with overlapping peptides spanning the AAVrh.10 capsid VP1 sequence and divided into three pools (15-mers overlapping by 10 amino acids). The negative control consisted of unstimulated CD3/CD28 cells whereas the positive control consisted of stimulated CD3/CD28 cells for cytokine secretion. Responses were considered positive when the number of SFU per 1 × 106 cells were >50 and at least threefold higher than the control condition.

Statistics

Differences between control and treatment groups were compared using one-way ANOVA. Differences between control and treatment groups and spinal cord sections were assessed using two-way ANOVA. Differences between specific groups were assessed using unpaired two-tailed t test. Differences between groups were considered to be significant at a P value of <0.05. Statistical analyses were performed with GraphPad Prism version 7.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/465/eaau6414/DC1

Fig. S1. GFP protein is detected in the spinal cord at the lumbar section.

Fig. S2. GFP protein is detected in the brain.

Fig. S3. GFP protein is detected in the pons and midbrain.

Fig. S4. Cellular immune response to GFP protein is quantified by the IFN-γ–ELISpot assay.

Fig. S5. Hepatic histopathology scores correlate with Ki67 immunostaining.

Fig. S6. Stable expression of miR-122 in the liver.

Table S1. Experimental groups, first M. fascicularis study.

Table S2. Liver transaminases, first M. fascicularis study.

Table S3. Statistical analysis for Fig. 3.

Table S4. Experimental groups, second M. fascicularis study.

Table S5. NAb titers to AAVrh.10 in serum and CSF after intrathecal delivery.

Table S6. Detection of viral genome in laser-capture microdissected motor neurons.

Table S7. Statistical analysis for Fig. 5.

Table S8. List of potential off-target genes based on sequence identity.

Table S9. RIN values.

Table S10. List of primers/probes.

REFERENCES AND NOTES

Acknowledgments: We acknowledge G. Gao and Q. Su at University of Massachusetts Medical School who supervised the vector production. We thank C. Simon, E. Adams, and R. Reed at NBR for their support. We are grateful to M. Zieger for assistance with histology, microscopy, and figure assembly. We thank M. Difiglia for help with nonhuman primate brain histology. Funding: This work was funded by ALS Alliance Therapy (R.H.B.), the Rizzuto fund (R.H.B.), and NINDS NS088689 (C.M. and R.H.B.). R.H.B. also received support from the NINDS (NS079836), the ALS Association, the Angel Fund, the Pierre L. de Bourgknecht ALS Research Foundation, Project ALS, ALS Finding A Cure, ALSOne, and P2ALS. F.B. received support from the Alpha-1 Foundation. Author contributions: F.B., G.G., R.H.B., and C.M. designed experiments; F.B., G.G., H.S., R.S., and M.B. performed experiments; and F.B., G.G., R.H.B., and C.M. wrote and edited the manuscript. F.B. performed the SOD1 silencing experiments, miR-SOD1 ddPCR quantification, in silico off-target predictions, and RNAscope assay; G.G. performed the immunology experiments, vector biodistribution, and off-target ddPCR analysis; H.S. and F.B. analyzed the small RNA-seq data; R.S. performed laser-capture microdissection and immunostaining; and M.B. helped with the immunology experiments, immunostaining, and RNAscope assay. Competing interests: R.H.B. and C.M. are inventors on the patent for the technology described within the paper (rAAV-Based Compositions and Methods for Treating Amyotrophic Lateral Sclerosis, WO2015143078A1) and may be entitled to royalty payments in the future. The patent has been licensed to Apic Bio for which C.M. and R.H.B. are founders with equity. C.M. and R.H.B. were formerly paid consultants for Voyager Therapeutics. Voyager Therapeutics was not involved in funding or any other aspect of this study. The other authors (F.B., G.G., H.S., R.S., and M.B.) declare no competing financial interests. Data and materials availability: All the data are present in the paper or in the Supplementary Materials.
View Abstract

Navigate This Article