Inhibition of immune response reduces pathology in dorsal root ganglia and peripheral nerves in cynomolgus macaques following AAV gene therapy

Abstract

Administration of adeno-associated virus (AAV) gene therapies via blood or cerebrospinal fluid (CSF) in non-human primates (NHPs) can lead to degeneration of dorsal root ganglion (DRG) neurons and nerve fibers in the spinal cord and peripheral nerves. AAV cargo expression is implicated in AAV DRG toxicity, but the underlying mechanism(s) is unknown. Here, we performed a time course study of intra-cisterna magna (ICM) administration of an AAV9 variant encoding human survival of motor neuron 1 (hSMN1) to identify molecular and cellular changes preceding pathology. Increases in inflammatory gene modules, cerebrospinal fluid (CSF) cytokines, and immune cell infiltrates as early as day 5 prior to neuron and nerve fiber degeneration on days 15 and 29 suggested a role for the immune response in AAV-mediated toxicity. Prophylactic treatment with a glucocorticoid steroid dexamethasone and a calcineurin inhibitor tacrolimus diminished pathology in NHPs following administration of three AAV gene therapy vectors. Collectively, these data demonstrate a causal role for the immune response to AAV in AAV-mediated DRG and nerve fiber toxicity. The efficacy of immunosuppression with three different AAV cargos suggests broad utility across AAV vectors and provides a clinically feasible approach to mitigating this potential toxicity in patients.

Graphical abstract

Grubor and colleagues investigated events preceding dorsal root ganglion (DRG) toxicity following intra-CSF administration of adeno-associated virus (AAV) gene therapy in non-human primates. They identified an immune response prior to neuronal degeneration and were able to mitigate toxicity with a novel immunosuppression regimen.

Introduction

Adeno-associated virus (AAV) is the most prevalent vector for in vivo gene therapy delivery with >200 AAV-based gene therapies in the clinic.¹ Yet, AAV gene therapies have encountered hurdles for safe and efficacious in vivo administration, including histopathological findings of dorsal root ganglia (DRG) toxicity in preclinical animal models following intra-cerebrospinal fluid (CSF) and high-dose intravenous (i.v.) administration.² A meta-analysis of AAV-mediated histopathological lesions consisting of DRG neuronal degeneration and spinal cord nerve fiber degeneration in 256 nonhuman primates (NHPs) across 33 preclinical AAV studies demonstrated that this pathology occurred with various capsids and transgenes, suggesting AAV-mediated DRG toxicity is a class-specific effect.³ Originally identified in NHPs and juvenile piglets,² AAV-mediated DRG toxicity has since been reported in mice,⁴ rats,⁵ and rabbits.^6,7 While human clinical data are sparse, autopsy findings of DRG neuron loss⁸ and T lymphocytic infiltrates,⁹ as well as reports of peripheral neuropathy^9,10 following AAV gene therapy are consistent with AAV-mediated DRG toxicity. Based on the weight of evidence from nonclinical and clinical data, we predict that this AAV toxicity is likely to translate to humans. Thus, as more patients are dosed with AAV gene therapies, it is conceivable that there will be an increased incidence of reports consistent with DRG toxicity.

Understanding the mechanisms underlying AAV-mediated DRG toxicity is paramount to devising mitigation strategies to address this potential toxicity in human patients. Currently, the mechanisms for AAV-mediated DRG toxicity are largely understood through key variables, such as dose and promoter strength,² which implicate vector cargo overexpression as a contributing factor in AAV-mediated DRG toxicity. Consistent with these variables, administration of empty AAV particles in the absence of full AAV particles^4,11 or promoter-less vectors¹¹ fail to cause DRG toxicity. Yet, how AAV vector transgene overexpression contributes to DRG toxicity is unclear. Furthermore, the mechanisms underlying the differences in sensitivity of sensory neurons in the DRG versus motor neurons in the spinal cord (SC) is unknown; most studies fail to detect motor neuron toxicity.^2,4,5,12,13 When motor neuron degeneration has been reported in NHPs receiving AAV gene therapies, the incidence was rare and the severity minimal.¹⁴

To inform on the mechanism(s) underlying DRG toxicity, we conducted a detailed lesion time course study in cynomolgus macaques following a single intra-cisterna magna (ICM) administration of an AAVhu68 vector encoding human codon-optimized survival of motor neuron 1 (hSMN1). Previous studies with ICM dosing of this vector have shown the presence of DRG lesions on day 29 following administration on day 1, which allowed us to anchor our time course on day 29. We included three additional time points (days 5, 9, and 15) to capture the onset of DRG lesions, as well as to understand the biological and pathological changes that precede and/or lead to the development of DRG toxicity. We analyzed molecular and cellular pathways at time points preceding (days 5 and 9) and coincident (days 15 and 29) with DRG lesion formation (e.g., DRG neuron degeneration/necrosis and spinal cord nerve fiber degeneration). Based on these results, we identified key pathways and immune responses that differ in the DRGs versus the SC and precede DRG lesion formation. Furthermore, the addition of an immunosuppression regimen consisting of dexamethasone and tacrolimus reduced pathology in three independent NHP studies, demonstrating a causal role for the immune response in AAV-mediated DRG toxicity.

Results

ICM administration of 3.68 × 10¹³ genomic copies of AAVhu68-hSMN1 was well tolerated in NHP with minimal clinical chemistry changes, consistent with AAV administration

Administration of vehicle or a clinically relevant dose (3.68 × 10¹³ GC) of AAVhu68-ss-coSMN1 (AAVhu68-hSMN1) in five groups of NHPs (N = 2–3/group) via ICM route for a lesion time course study (days 5, 9, 15, and 29) was generally well tolerated. All NHPs survived to the day of scheduled necropsy, except for one male monkey in the day 15 group, which was euthanized due to ICM dosing-procedure-incurred injury at 7 h post-dose. As a result, this NHP was omitted from further analysis. AAVhu68-hSMN1 had no effect on body weight, food consumption, body temperature, and clinical pathology parameters for hematology and coagulation (data not shown). No remarkable findings were noted during physical and neurological examinations (data not shown). AAVhu68-hSMN1-related minimal-to-mild elevations (1.5×–2.3× over baseline on day −6) in alanine aminotransferase activity (ALT) and/or aspartate aminotransferase activity (AST) were observed in five NHPs on 9, 11, 15, and/or 29 days following AAVhu68-hSMN1 administration (Table S1). These changes were suggestive of liver injury, although only minimal hepatocellular single-cell necrosis was observed in the liver microscopically in one female on day 5 and one female on day 9. Discordance between clinical chemistry changes and liver histopathology findings may be due to differences in the kinetics of ALT/AST release versus hepatocyte necrosis and/or difficulties in detecting acute necrosis due to necrotic cell replacement by proliferating hepatocytes. Thus, the use of both clinical chemistry measures and histopathology provides a more comprehensive assessment of AAV-associated changes.

Histological lesions were observed by day 15 in DRGs, trigeminal ganglia, and spinal cord

AAVhu68-hSMN1-related histopathology findings in the nervous system were first observed on day 15 in the lumbar DRGs and cervical, thoracic, and lumbar SC in only one of the two animals (14001) (Figure 1A). Histopathology changes included minimal mononuclear cell infiltration in two DRGs (L4 and L5-6), accompanied by minimal neuronal degeneration/necrosis in one of the DRGs (L5-6) and minimal nerve fiber degeneration in dorsal and/or lateral funiculi of the spinal cord. Animal 14001 also exhibited mild motor neuron degeneration in C2. Trigeminal ganglia (TG) were affected in both animals from the day 15 group with mild mononuclear cell infiltration and minimal neuronal cell body degeneration/necrosis (Figure S1).

Figure 1.

AAVhu68-mediated toxicity detectable by day 15 in NHPs

Cynomolgus macaques received vehicle or 3.68 × 10¹³ of AAVhu68-hSMN1 by ICM administration. Tissue and serum were collected on days 5, 9, 15, and 29 for AAVhu68-hSMN1-dosed and day 29 for vehicle-dosed subjects. (A) DRG and spinal cord tissues were assessed for histopathology findings. The incidence and severity of findings in the DRGs and spinal cord (DRG neuronal degeneration/necrosis, DRG mononuclear cell infiltration, and spinal cord nerve fiber degeneration) is reported per subject at the different spinal column or spinal cord levels, respectively [y axis: cervical (C), thoracic (T), lumbar (L)] and for the spinal cord subregions [x axis: cauda equina (C), and dorsal (D), lateral (L), and ventral (V) funiculi]. Each finding is represented by a dot where the diameter of the dot is scaled to indicate the severity of the finding. Histopathology findings from additional tissues are reported in Figure S1B. DRG tissue was stained for cleaved caspase 3. Representative images at 20x are shown. Arrows indicate positive cleaved caspase 3 staining. (C) Quantification of number of cleaved caspase-positive cells/mm² of DRG tissues in subjects per spinal column level. Dot represents detection of cleaved caspase-positive cells with size and color reflecting number of cells/mm². (D) Serum levels of NF-H (pg/mL) in each subject as determined by ELISA at their terminal time point. Each symbol represents the mean of three technical replicates per sample. The horizontal line indicates the median value within the group. FDR adjusted ∗p_adj < 0.05, one-way ANOVA was performed on log-transformed values comparing all AAV-hSMN1-dosed subjects to the vehicle controls.

On day 29, AAVhu68-hSMN1-related histopathology findings had a greater severity and higher incidence and were more widely spread throughout the nervous system as compared to day 15 (Figure 1A). All levels (cervical, thoracic, and lumbar) of the DRGs were affected, with higher severity of findings observed at the cervical and the lumbar compared to the thoracic levels. These findings included minimal-to-moderate mononuclear cell infiltration in all three NHPs, with minimal-to-mild neuronal cell body degeneration/necrosis present in DRGs from two of the three NHPs. In addition, minimal-to-marked nerve fiber degeneration in the dorsal nerve roots was present in two of the three NHPs and was observed at all levels of the spinal cord, with the most severe changes at the lumbar level; mild nerve fiber degeneration of ventral nerve roots was present only at the lumbar level in one NHP (Figure S1). Similar to DRGs and dorsal nerve roots, all SC levels were affected with the highest severity of findings observed at the lumbar SC/cauda equina, with minimal-to-moderate nerve fiber degeneration in the dorsal and/or lateral funiculi and cauda equina in all three NHPs (Figure 1A). Minimal nerve fiber degeneration was noted in the lateral funiculus of the thoracic spinal cord (T6 and T9) in one vehicle-treated monkey on day 29. This spinal cord finding likely represents an incidental background finding. In addition, two of the three NHPs dosed with AAVhu68-hSMN1 had minimal or moderate gliosis in the gray matter of the lumbar spinal cord (L3 or L1, L3, and L5-6, respectively) on day 29 (data not shown). Mild neuronal central chromatolysis (L3) and motor neuron necrosis (L5-6) were only present in the NHP with more severe gliosis on day 29 (data not presented). Trigeminal ganglia were affected in all three monkeys from day 29 group and findings included mild-to-moderate mononuclear cell infiltration with minimal neuronal cell body degeneration/necrosis (Figure S1). Peripheral nerves were also affected in all three animals on day 29, with findings of mild-to-marked nerve fiber degeneration (Figure S1) and mild mononuclear cell infiltration in the sciatic nerve (data not shown) and minimal-to-mild nerve fiber degeneration and minimal mononuclear cell infiltration in the median nerve (data not shown).

Notably, electron microscopy of lumbar DRGs (L5 and L6) revealed persistent AAVhu68-hSMN1-related ultrastructural changes starting on day 5 and at all later time points (Table S2). These ultrastructural changes included endoplasmic reticulum dilation/vesiculation and mitochondrial swelling with peripherally placed to disintegrating cristae in the sensory neurons and endoplasmic reticulum dilation/vesiculation in the satellite glial cells. Although mitochondrial swelling was also observed in DRG-L5 sensory neurons from one control animal, the severity of this finding was higher in AAVhu68-SMN1-treated animals on days 9 and 15. Additional AAVhu68-hSMN1-related ultrastructural changes in the sensory neurons on days 15 and 29 were consistent with histopathology findings of neuronal degeneration/necrosis, i.e., shrunken neuronal cell body, condensed cytoplasmic structures, membrane remnants, occasional dense bodies, and/or fragmented nucleolus. Mononuclear cells were observed as early as day 9 in L5/L6 DRG electron microscopy samples from AAVhu68-hSMN1-dosed animals and were also observed on days 15 and 29 (Table S2), consistent with histopathology findings (Figure 1A). While ultrastructural changes were observed in the lumbar spinal cord samples, these findings were considered unrelated to AAVhu68-hSMN1 treatment as they were similarly observed in vehicle-treated animals (data not shown).

Histological lesions in DRGs, trigeminal ganglia, and spinal cord correlated with serum neurofilament heavy

Given ultrastructural changes in neurons were present as early as day 5 and histopathology findings on day 15, we sought to discern whether apoptosis was an underlying mechanism of cell death. DRG tissues were stained for cleaved caspase 3, a marker of apoptosis. Cleavage of caspase 3 was only detected in DRG neurons from AAVhu68-hSMN1-dosed animals on day 29 and specifically in NHPs with the greatest incidence and/or severity of histopathology findings (i.e., 15001 and 15002) (Figures 1B and 1C). AAV-mediated neuronal damage was also assessed using serum neurofilament heavy (NF-H), a peripherally accessible biomarker. Though some individual serum NF-H levels [average (SD)] on days 9 and 15 were elevated, day 5 [1496 (1047) pg/mL], day 9 [3581 (2362) pg/mL], and day 15 [32936 (32501) pg/mL] did not differ significantly from vehicle-dosed controls [2382 (552.5) pg/mL] (Figure 1D). Day 29 AAVhu68-dosed NHP had significantly increased serum NF-H compared to the vehicle control [42.97 (20.82)-fold increase, p value = 0.036]. On an individual level, serum NF-H elevations corresponded to the NHPs with greatest incidence/severity of AAVhu68-hSMN1-related DRG neuronal degeneration/necrosis and mononuclear cell infiltration and spinal cord nerve fiber degeneration (i.e., animals 14001, 15001, 15002, and 15501) (Figures 1A and 1D).

Transgene expression levels in DRGs corresponded to observed pathology

ICM administration of the vector successfully reached tissues of interest for the time course study (DRG and SC) with approximately 1–1,000 vector genomes (VG)/diploid genomes (DG) (Figure S2). Other CNS regions (motor cortex and brainstem) showed 10- to 100-fold lower exposures compared to the DRGs and SC. Vector DNA biodistribution to the liver was similar to DRGs and SC, as expected. Interestingly, there was a trend toward decreasing VG per diploid genome (DG) in DRGs and liver over time, most notably on day 29, that was not observed in other tissues (motor cortex, brainstem, and spinal cord) (Figure S2).

Transgene RNA expression within the DRGs was present on day 5 [2.1 (1.7) transcripts per million (TPM)] (Figure 2A), peaking by day 15 [4.4 (2.2) TPM] and maintained on day 29 [4.6 (2.7) TPM]. DRG neuronal transduction, as determined by any neuronal RNAscope reactivity for coSMN1, was evident at all time points, was highly variable among animals, and was highly variable throughout spinal levels within animals (Figures 2B, 2C, S3, and S4). Neuronal coSMN1 reactivity density, indicative of coSMN1 transgene expression per neuron, peaked in DRG tissue on day 15 (Figure 2D), coinciding with the first detectable histopathology findings and serum NF-H elevations in animal 14001. Consistent with a role of transgene expression in neuronal damage and degeneration, animal 14001 had greater levels of DRG neuron transduction and transgene density per neuron than the other animal at the day 15 time point (14501), which lacked both DRG histopathology findings and elevated serum NF-H.

Figure 2.

Transgene RNA expression peaks by day 15 in DRGs, and expression is associated with neuronal cell death

Cynomolgus macaques received vehicle or 3.68 × 10¹³ of AAVhu68-hSMN1 by ICM administration. (A) Human SMN1 transgene expression in NHP DRGs reported as transcripts per million (TPM) as determined by bulk RNA-seq at various days post-AAVhu68-hSMN1 or day 29 post-vehicle administration. Each time point consists of cervical, thoracic, and lumbar DRGs from 2 to 3 subjects. A mixed-model ANOVA was performed comparing transgene expression at each time point, accounting for subject effects. (B) Representative 40x images of Hs-SMN1-O1 RNAscope in situ hybridization on cervical, thoracic, and lumbar DRGs on day 29 for vehicle or days 5, 9, 15, and 29 post-AAVhu68-hSMN1 administration. (C) Mean percent transduction of DRG neurons per spinal cord level in each subject as defined as Hs-SMN-O1-positive cells from (B). Size of bubble and color density increases with increasing percent neurons with hSMN1 transgene expression. (D) Mean density of Hs-SMN-O1 signal in transduced DRG neurons per spinal cord level in each subject from (B). Size of bubble and color density increases with increasing density of transgene expression. Refer to Figure S4 for image analysis for transduced DRG neurons. Relationship of (E) neuronal transduction and (F) transgene expression density with cleavage of caspase 3 in animals 5001 and 5002. Spearman rho correlation values are indicated.

A role for transgene expression in neuronal degeneration and necrosis was further supported by the observations from NHPs sacrificed on day 29; animals 15001 and 15002 had greater neuronal transduction and transgene expression than animal 15501, and 15001 and 15002 also had a higher incidence and severity of histopathology findings as well as higher serum NF-H levels. To discern whether the level of transgene expression was also related to apoptotic neuron death on day 29, we correlated transduction and single-cell transgene expression to cleaved caspase 3 staining. The number of cells detected with cleaved caspase 3 had a significant and moderate correlation with neuronal transduction and transgene expression levels/neuron in those regions (Figures 2E and 2F). Collectively, these results support that transgene expression plays a role in AAV-mediated neuronal degeneration and necrosis and more specifically, neuronal apoptosis in DRGs.

Transcriptomic signatures from DRGs after AAV administration reveal potential molecular pathways contributing to neuronal degeneration

To elucidate molecular pathways underlying the susceptibility of DRGs to neuronal degeneration, necrosis, and apoptosis in response to AAV transgene expression, DRGs from this time course study were evaluated for transcriptomic changes following AAV delivery. Due to the observed inter- and intra-subject variability in histology in different spinal cord regions (Figure 1A), a weighted cluster gene network analysis (WGCNA) was utilized for RNA sequencing (RNA-seq) analysis to improve signal detection and functional annotation over differential gene expression (DGE) analyses.

Eleven gene co-expression modules were identified in the DRG transcriptome by WGCNA (Table S3; Figure S5). Four modules, DRG1, 7, 8 and 11, were enriched in genes involved in synaptic, innate immune, mitochondrial, or ribosomal/RNA processing and complement processes, respectively, and displayed bimodal kinetics with altered gene expression at early (day 5) and delayed time points (days 15 and/or 29). The ribosomal/RNA processing/complement module displayed trends for increased expression on days 5 and 29, whereas the expression of genes from the innate immune module exhibited trends for increased expression on days 5, 15, and 29 (Figure 3A; Table S3). In contrast, synaptic and mitochondrial modules’ genes revealed trends for decreased expression on days 5 and 29 (Figure 3A; Table S3). The contemporaneous changes in the four modules’ expression suggest a cause-effect relationship between these pathways or a common upstream driver for these four modules. Furthermore, early pathway changes (increased or decreased module expression) on day 5 do not appear to correlate with transgene expression levels, whereas the samples with the greatest increases or decreases on days 15 and 29 are also the samples with the highest transgene expression (Figure 3A).

Figure 3.

AAVhu68 transduction in DRGs increases expression of transcriptional modules associated with innate immune response and ribosomal/RNA processing and complement activity and decreases transcriptional modules associated with synaptic and mitochondrial function

Cynomolgus macaques received vehicle or 3.68 × 10¹³ of AAVhu68-hSMN1 by ICM administration. DRG and spinal cord tissues were collected on days 5, 9, 15, and 29 for AAVhu68-hSMN1 and day 29 for vehicle-dosed subjects. Total RNA was isolated from homogenized tissues and evaluated by RNA-seq. Fold change of principal component (PC1) from each module (innate immune, ribosomal/RNA processing/complement, synaptic, and mitochondrial) in (A) DRG tissue samples or (B) spinal cord tissue samples from each group is shown. Each symbol represents one sample. The color of the symbol indicates relative hSMN1 transgene expression level within that sample. (C) Heatmaps depicting relative gene expression (relative Z score) in DRGs and spinal cord tissue from vehicle- (day 29) or AAVhu68-hSMN1-dosed NHP (days 5, 9, 15, and 29) for innate immune module genes. Higher expression over mean gene expression is indicated by red scale, and lower expression is indicated by blue scale.

Despite similar levels of transgene expression in the DRGs and SC (Figures 2A–2D and S6A–S6E), the incidence of motor neuron degeneration/necrosis was lower with a focal distribution compared to DRG neurons. Thus, we sought to understand the transcriptomic differences in these two transduced tissues. To do so, WGCNA modules were also identified from spinal cord tissue samples. While four spinal cord modules (SC12, 14, 15, and 16) demonstrated significant changes in response to AAV, none of these changes preceded histopathology findings (i.e., were present only on day 15 and/or 29) (Table S4). To better compare transcriptomic changes between DRGs and spinal cord, we compared spinal cord and DRG specific gene expression before and after AAVhu68-hSMN1 administration in the four previously mentioned DRG modules: synaptic, innate immune, mitochondrial, or ribosomal, RNA processing and complement. DRGs and spinal cord samples exhibited baseline transcriptomic differences (Figures S7 and S8). DRGs from vehicle control animals had higher expression of innate immune genes compared to vehicle-treated spinal cord samples. In contrast, ribosomal, RNA processing and complement, mitochondrial, and synaptic genes were differentially expressed in vehicle-treated spinal cord and DRG tissues, with different gene subsets being more highly expressed in one tissue than the other (Figure S8).

Transcriptomic responses to AAVhu68-hSMN1 differed between the DRGs and spinal cord samples when compared to expression in vehicle-treated tissue. AAV-induced perturbations to synaptic and mitochondrial modules identified from DRG WGCNA were largely absent in the spinal cord at early (day 5) or late (days 15 and 29) time points (Figure 3B). Ribosomal, RNA processing and complement genes were significantly increased in the spinal cord on day 29, but they were unaltered at time points preceding pathology (Figure 3B; Table S4). Additionally, genes involved in the innate immune response appeared modestly elevated in spinal cord samples on day 5 and more robustly on days 15 and 29 (Figure 3B) but genes involved in the immune response (module SC12) were significantly increased in the spinal cord only at the later time points (Table S4). However, given the higher baseline levels of innate immune module genes in DRG as compared to spinal cord tissue (Figure S8), the overall expression of innate immune module genes in the spinal cord of AAVhu68-hSMN1 subjects was still lower than that observed in the DRG (Figure 3C). Overall, these studies reveal distinct baseline and early transcriptomic changes between DRGs and spinal cord tissue that may contribute to their differential susceptibility to AAV toxicity.

Type I interferon signaling and immune cell recruitment occurs in DRGs prior to neuronal degeneration/necrosis in response to AAVhu68-hSMN1 ICM administration

Given our observations that inflammatory processes were elevated in the DRG transcriptome at baseline as well as prior to neuronal degeneration in animals administered AAVhu68-hSMN1, we evaluated CSF samples from the time of injection and necropsy for the presence of cytokines and chemokines. Interferon gamma (IFN-γ), interleukin (IL)-1β, IL-2, IL-8, IL-10, tumor necrosis factor alpha (TNF-α), and IFN-α2a were all below the limit of quantitation (data not shown). CSF CXCL10 and MIP1-α were significantly elevated on days 5 and 29 and days 9 and 29, respectively (Figures 4A and S9A). Overall, there was a trend for increased IL-6, CXCL10, MIP1α, and monocyte chemoattractant protein 1 (MCP1) as early as day 5, prior to time points with detectable neuronal degeneration (day 15). Interestingly, CSF CXCL10 and MIP1α levels correlated positively with serum NF-H levels from subject-matched samples (Figures 4B and 4C). Neuronal CXCL10 transcript levels had a moderate correlation with the apoptotic cell marker, cleaved caspase 3 (Figure 4D), though correlation is likely limited by undetectable levels of cleaved caspase 3 prior to day 29.

Figure 4.

Local innate immune response in DRGs to AAVhu68-hSMN1 correlates with neuronal damage

Cynomolgus macaques received vehicle or 3.68 × 10¹³ of AAVhu68-hSMN1 by ICM administration. (A and B) CSF was sampled prior to ICM dosing (baseline) and at time of necropsy (days 5, 9, 15, and 29) and evaluated for IL-6, CXCL10, MIP1α, and MCP-1 in the CSF by MSD ELISA. (A) Cytokine levels in CSF for each group. Blue symbol reflects mean ± standard deviation at baseline and day 29 in the vehicle-dosed group. Open white symbol represents mean ± standard deviation for baseline levels prior to AAVhu68-hSMN1 administration. Yellow, orange, brown, and black symbols represent mean ± standard deviation of cytokine levels in NHP subjects from days 5, 9, 15, and 29, respectively. One-way ANOVA was performed on linear (MIP1α) or log-transformed data (IL-6, MCP-1, and CXCL10) comparing each time point to its baseline. FDR adjusted ∗p < 0.05. Individual subject cytokine CSF levels can be found in Figure S9A. (B) Relationship of CSF CXCL10 at necropsy time point compared to matched serum NF-H. Spearman rho correlation values are indicated. (C) Relationship of CSF MIP1α at necropsy time point compared to matched serum NF-H. Spearman rho correlation values is indicated. (D) DRG tissue collected from AAV-hSMN1-dosed animals on day 29 was evaluated for neurons that were Cxcl10 RNA (Mfa-Cxcl10-C1) and cleaved caspase 3 in serial sections. Each symbol reflects an individual DRG sample. Relationship of Cxcl10-positive neurons and presence of cleaved caspase 3 on day 29 is shown. Spearman rho correlation value is indicated.

To address limitations surrounding infrequent sampling of the CSF, longitudinal serum and whole blood samples were evaluated for cytokine/chemokine analysis and transcriptomics, respectively. Serum IFN-γ, IL-1β, IL-6, IL-10, IFN-α2a and TNF-α, IL-8, macrophage-derived chemokine (MDC), MCP-1, and MIP1α were either below the threshold for detection or had minimal AAV-responsive changes (data not shown). AAVhu68-hSMN1 treatment increased serum CXCL10 levels compared to vehicle (Figure S9B). However, despite red blood cell contamination (Table S5), levels of CXCL10 in the CSF (maximum of 9,400 pg/mL) were ∼9x higher than in the serum (maximum of 1,100 pg/mL, mean of 720 pg/mL), suggesting a local innate immune response and synthesis of CXCL10 in the CNS. Given that CXCL10 can be induced in response to type I IFN, and type I IFN is known to be involved in apoptosis, we also evaluated the type I IFN transcriptional response in the DRGs and in the blood. The DRGs and whole blood exhibited both early (day 5) and delayed (days 15 and 29) type I IFN responses in response to AAVhu68-hSMN1 treatment, punctuated by reduced expression of these genes on day 9 (Figures S10A and S10B).

Type I IFN signaling in response to AAV may activate AAV-specific adaptive immune responses. Therefore, we evaluated whether there was evidence for an adaptive immune response to AAV. Transcriptomics revealed an adaptive immune response signature within the DRGs primarily on days 15–29, including genes associated with antigen presentation, such as Erap1, Tap1, and Tap2 (Figure S10C). In contrast, the adaptive immune signature was present in the blood at early (12 h–5 days) but not late (days 15–29) time points (Figure S10D). However, there were no detectable capsid- or antigen-specific T cells from peripheral blood mononuclear cells (PBMCs) (positive sample defined as >50 SFU/1 × 10⁶ PBMCs and 3x-fold increase above its respective negative control) (Figure S11).

Further analysis by in situ hybridization (ISH) and immunohistochemistry (IHC) revealed immune infiltrates in DRG tissue as distinct foci as early as day 5 for natural killer (NK) cells, monocytes, B cells, and CD4+ T cells. Other innate and adaptive immune cells, such as plasmacytoid dendritic cells (pDCs) and CD8+ T cells were detected at later time points (days 15 and 29) (Figures 5 and S12). These ISH/IHC findings demonstrate recruitment of different immune cell types into DRG tissue prior to detectable neuronal cell body degeneration/necrosis observed on day 15 by both histopathology and electron microscopy, suggesting that immune cells are not solely entering the DRG tissue in response to neuronal degeneration/necrosis but may contribute toward DRG neuron degeneration.

Figure 5.

Immune cell infiltrates observed on days 5 and 9, preceding neuronal degeneration/necrosis in DRGs

Representative images of RNAscope or IHC assays employed on DRGs from NHP on day 29 post-ICM administration of vehicle or days 5, 9, 15, and 29 post-ICM administration of 3.68 × 10¹³ of AAVhu68-hSMN1. In situ hybridization was performed to evaluate infiltration of natural killer cells (Nkp46; NCR1) and plasmacytoid dendritic cells (Clec4c; CD303) by RNAscope, macrophages (CD68), B cells (CD20), CD4+ T cells, and CD8+ T cells staining by IHC. Quantitation of immune cell infiltrates as determined via RNAscope and IHC assays can be found.

Immunosuppressants reduce incidence and severity of DRG, SC, TG, and peripheral nerve histopathological findings following ICM administration of 3.5 × 10¹³ GC AAVhu68-hSMN1

Although less immunogenic than first-generation gene therapies, e.g., adenovirus, AAV gene therapies are now known to induce immune responses in preclinical models and in the clinic,¹⁵ including the generation of capsid-^16,17,18 and/or transgene-specific¹⁹ T cells, both of which have been linked to other AAV-mediated toxicities, e.g., hepatotoxicity,^20,21,22 cardiotoxicity,¹⁹ etc. Thus, to test whether DRG toxicity following AAV gene therapy administration is immune-mediated, we evaluated the histopathology in NHPs in response to AAVhu68-hSMN1 (3.5 × 10¹³ GC) in the absence or presence of an immunosuppressant regimen (IMS#1). IMS#1 was delivered orally on days 0–28 and consisted of drugs used to prevent T-cell-mediated organ transplant rejection: a CNS-penetrant glucocorticoid steroid dexamethasone (0.5 mpk, SID) to reduce inflammation, a calcineurin inhibitor tacrolimus (1 mpk, SID) to target T cells, and an anti-proliferative agent mycophenolate mofetil (MMF; 50 mpk, BID) to inhibit B cell and T cell proliferation. To further ensure distribution of dexamethasone to the CNS, NHPs receiving IMS#1 received 0.5 mg of dexamethasone via intrathecal (IT) route on day 0 as a single dose. On day 28, we confirmed peripheral exposure of drugs in our IMS#1-dosed NHPs achieved clinically relevant concentrations at 70.1 ± 6.0; 10,216.7 ± 8,610.5; and 12.0 ± 2.9 ng/mL for dexamethasone, mycophenolic acid (the active metabolite of prodrug MMF; MPA), and tacrolimus, respectively (Figure 6A). Since transgene expression is required for AAV-mediated DRG toxicity,^3,4,23 we confirmed that IMS#1 treatment did not reduce hSMN1 transgene expression in the DRGs compared to subjects that received no IMS treatment (7.1 ± 4.8 vs. 11.7 ± 9.1, p = 0.45) (Figure 6B).

Figure 6.

Dexamethasone, MMF, and tacrolimus reduced AAVhu68-hSMN1-mediated DRG toxicity following ICM administration on day 29

Cynomolgus macaques received vehicle or 3.5 × 10¹³ of AAVhu68-hSMN1 by ICM administration in the presence or absence of an immunosuppression (IMS) regimen consisting of dexamethasone, MMF, and tacrolimus. (A) Plasma and whole blood were collected on day 28 at 6 h post-IMS dosing to measure IMS drug concentrations for drug exposure. Each symbol represents the amount of drug in one IMS-dosed subject on day 28. (B) hSMN1 transgene expression in DRGs on day 29 post-AAVhu68-hSMN1 administration was determined by RNAscope. Median % Area hSMN1 was determined in each subject at cervical, thoracic, and lumbar DRG levels. A different symbol (circle, square, triangle) is assigned to each animal, and each symbol represents the median value from cervical, thoracic, or lumbar DRG tissue. No significant differences in expression were found via mixed model ANOVA on raw values and log-transformed values. (C) The incidence and severity of findings in the DRGs and spinal cord [DRG neuronal degeneration/necrosis, DRG mononuclear cell infiltration, spinal cord nerve fiber degeneration in the dorsal funiculus (D) or cauda equina (C) and spinal cord motor neuron degeneration] is reported as the maximum severity observed per subject at the different spinal cord levels [y axis: cervical (C), thoracic (T), lumbar (L)]. Each finding is represented by a dot where the diameter of the dot is scaled to indicate the severity of the finding.

As expected, there were no histopathology findings in the DRG of vehicle-treated animals. Two of three subjects dosed with AAVhu68-hSMN1 without IMS#1 had minimal neuronal cell body degeneration/necrosis, which was accompanied by minimal-to-moderate mononuclear cell infiltrates (Figure 6C) and minimal-to-mild satellite glial cell proliferation (data not shown). There were no DRG histopathology findings in animal 22501. Two of three animals (24002 and 24501) that were co-administered IMS#1 with AAVhu68-hSMN1 presented with no histopathology findings in their DRGs. One subject (14001) presented with minimal neuronal cell body degeneration/necrosis in only one DRG sample (L2), consisting of two degenerate/necrotic neuronal cell bodies surrounded by minimal mononuclear cell infiltration (Figure 6C) and the absence of satellite glial cell proliferation (data not shown). Nerve fiber degeneration in the dorsal funiculi of the spinal cord was present bilaterally in multiple sections and ranged from minimal-to-moderate severity in two of three NHP administered AAVhu68-hSMN1 without IMS#1. Minimal spinal cord nerve fiber degeneration was noted in two of three monkeys that received IMS#1; it was bilateral and present in one (24501) or multiple sections (24001) (Figure 6C). Minimal motor neuron degeneration, consisting of shrunken motor neurons in the ventral horn surrounded by mononuclear and/or glial cells, was observed in two of three animals with AAVhu68-hSMN1 without IMS. No motor neuron degeneration was observed in any animals receiving AAVhu68-hSMN1 with IMS#1. Collectively, IMS#1 reduced the severity and/or incidence of histopathological findings in the DRG and SC in response to ICM delivery of AAVhu68-hSMN1.

Mononuclear cell infiltrates were present in all NHPs that received AAVhu68-hSMN1 with or without IMS#1 ranging from minimal-to-mild severity as well as mononuclear cell infiltrates of minimal severity in the TG of two of the three vehicle-treated NHPs (Figure S13). TG neurons exhibited minimal-to-mild multifocal neuronal cell body degeneration accompanied by minimal-to-mild mononuclear cell infiltration in all three AAVhu68-hSMN1-treated subjects without IMS#1 (Figure S13). In comparison, neuronal degeneration was present in the TG of one of three NHPs with AAVhu68-hSMN1 with IMS#1 (Figure S13); the finding consisted of a single degenerate neuron surrounded by mild mononuclear infiltrates and mononuclear cell nests where neuronal bodies had presumably been located. Minimal-to-moderate sciatic nerve fiber degeneration was present in two of three NHPs receiving AAVhu68-hSMN1 without IMS#1, as well as minimal-to-mild nerve fiber degeneration in all three NHPs dosed with AAVhu68-hSMN1 and IMS#1 (Figure S13). Minimal-to-marked nerve root nerve fiber degeneration was present in the dorsal and ventral nerve roots in two of the three NHPs that received AAVhu68-hSMN1 without IMS. In contrast, there were no observations of nerve root nerve fiber degeneration in NHPs that received AAVhu68-hSMN1 with IMS#1 (Figure S13). Thus, the presence of IMS#1 lessened the severity and incidence of TG neuronal degeneration and nerve root nerve fiber degeneration. However, IMS#1 did not appear efficacious in decreasing TG mononuclear cell infiltrates and sciatic nerve fiber degeneration in NHP that received AAVhu68-hSMN1.

Immunosuppressants reduce incidence and severity of DRG, SC, TG, and peripheral nerve histopathological findings following ICM administration of 3.0 × 10¹³ GC AAV9-hGBA1

Overexpression of hSMN1 in wild-type mice causes late-onset sensorimotor deficits and proprioceptive and motor neuron degeneration, accompanied by neuroinflammatory and innate immune gene expression.²⁴ We reasoned it was possible that the benefit of IMS#1 on reducing AAVhu68-hSMN1-mediated lesions could reduce inflammation and associated neurodegeneration related to hSMN1 overexpression, thereby limiting the benefit of IMS to hSMN1 cargo. To test this hypothesis, we evaluated the benefit of IMS in reducing DRG toxicity following administration of an AAV encoding another transgene protein cargo, human acid beta-glucosidase 1 (hGBA1). NHPs were administered AAV9-hGBA1 (3.0 × 10¹³ GC) via ICM administration in the presence or absence of IMS#2, which consisted of orally administered dexamethasone (0.5 mpk, SID), tacrolimus (1 mpk, SID), and MMF (50 mpk, BID) from day −2 to day 43. In an internal independent study (data not shown), we confirmed that dexamethasone was CNS-penetrant, with a high CSF/unbound plasma ratio (0.71) at 6 h post-dose. Therefore, in subsequent studies we omitted IT injection of dexamethasone to minimize the burden on animals, as the oral delivery was predicted to maintain sufficient concentrations (>EC50) in the CSF for up to 16 h. On day 1, we confirmed peripheral drug exposure reached clinically relevant concentrations: 96.8 ± 34.4 ng/mL dexamethasone, 8,063.3 ± 3,560.4 ng/mL MPA, and 7.3 ± 4.8 ng/mL tacrolimus (Figure 7A). We further confirmed no statistically significant difference in hGBA1 transgene expression in DRG tissue from subjects that received AAV9-hGBA1 with or without IMS#2 (p = 0.997; Figure 7B).

Figure 7.

Dexamethasone, MMF, and tacrolimus reduced AAV9-hGBA1-mediated DRG toxicity following ICM administration on day 43

Cynomolgus macaques received vehicle or 3.0 × 10¹³ of AAV9-hGBA1 by ICM administration in the presence or absence of an immunosuppression (IMS) regimen consisting of dexamethasone, MMF, and tacrolimus. (A) Plasma and whole blood were collected on day 1 at 1 h post-IMS dosing to measure IMS drug concentrations for drug exposure. Each symbol represents the amount of drug in one IMS-dosed subject. (B) hGBA1 transgene expression in DRGs on day 29 post-AAV9-hGBA1 administration was determined by RNAscope. Median % Area hGBA1 was determined in each subject at cervical, thoracic, and lumbar DRG levels. A different symbol (circle, square, triangle) is assigned to each animal, and each symbol represents the median value from cervical, thoracic, or lumbar DRG tissue. No significant differences in expression were found via mixed model ANOVA on raw values and log-transformed values. (C) The incidence and severity of findings in the DRGs and spinal cord [DRG neuronal degeneration/necrosis, DRG mononuclear cell infiltration, spinal cord nerve fiber degeneration in dorsal funiculus (D) or cauda equina (C) and spinal cord motor neuron degeneration] is reported as the maximum severity observed per subject at the different spinal cord levels [y axis: cervical (C), thoracic (T), lumbar (L) and sacral (S)]. Each finding is represented by a dot where the diameter of the dot is scaled to indicate the severity of the finding. Missing samples are indicated by white “x” at the intersection of the line for the sample. C3 is annotated as missing from subject 1001 and 5002 to reflect absence of an unspecified cervical DRG (C2, C3, or C6) from each; as a result, only two of three cervical DRGs were evaluated for subject 1001 and 5002.

DRG and spinal cord tissue from vehicle and AAV9-hGBA1-dosed subjects with and without IMS#2 was evaluated for histopathology. Vehicle-treated animals had few findings that were limited to minimal spinal cord nerve fiber degeneration in two of three animals in the thoracic dorsal funiculus of the spinal cord (Figure 7C). Minimal-to-moderate DRG neuron degeneration/necrosis in cervical, lumbar, and/or sacral DRGs was noted in all three AAV9-hGBA1-dosed animals (10 of 26 DRGs evaluated) in the absence of IMS#2. Neuronal DRG lesions were accompanied by mononuclear cell infiltrates of minimal-to-moderate severity (Figure 7C). Additionally, cervical, thoracic, and lumbar spinal cord nerve fiber degeneration in the dorsal funiculi of minimal-to-moderate severity was observed in 3/3 animals that received AAV9-hGBA1 without IMS#2, as well as moderate-to-marked nerve fiber degeneration of the cauda equina in two of three NHPs with AAV9-hGBA1 (Figure 7C). Furthermore, minimal-to-marked nerve root nerve fiber degeneration (3/3), mild-to-moderate sciatic nerve fiber degeneration (2/3), and minimal mononuclear cell infiltration of the TG (2/3) were observed in the subjects that received AAV9-hGBA1 without IMS#2 (Figure S14). Only 1 out of 29 DRGs evaluated from animals that received AAV9-hGBA1 with IMS#2 had a finding of neuronal degeneration/necrosis (35001; Figure 7C). While two of three subjects dosed with IMS#2 had DRG mononuclear cell infiltrates, it was reduced in incidence and severity compared to their counterparts without IMS#2. Likewise, spinal cord nerve fiber degeneration was observed in only one of three IMS#2 subjects, and it was of minimal severity. Finally, minimal mononuclear cell infiltration in the TG was observed in one of three subjects that were administered IMS#2, and there were no findings of sciatic or nerve root nerve fiber degeneration (Figure S14). These data demonstrate the benefit of our IMS regimen on reducing histopathology in DRGs, SC, and peripheral nerves in NHPs, with a second gene therapy vector encoding a protein transgene. The benefit of the IMS#2 regimen in preventing motor neuron degeneration could not be evaluated, as none of the nine animals reported in this study had detectable motor neuron degeneration (Figure 7C). However, it should be noted that we have observed, albeit rarely, minimal-to-mild motor neuron degeneration with AAV9-hGBA1 in other subjects (data not shown).

Immunosuppressants reduce incidence and severity of DRG, SC, TG, and peripheral nerve histopathological findings following IT administration of 4.0 × 10¹³ GC AAV9-mir-SOD1 encoding a non-protein cargo

Our immunosuppressant regimen consisting of dexamethasone, MMF, and tacrolimus reduced peripheral and CNS AAV-mediated toxicities in NHPs. Others have demonstrated that a steroid alone¹⁴ or MMF in combination with rapamycin^12,13 is unable to reduce AAV-mediated DRG toxicity. Thus, it suggests an important role for the T cell inhibitor tacrolimus in reducing toxicities associated with ICM delivery of AAVs encoding protein transgenes, hSMN1 and hGBA1. Human transgenes can elicit T-cell-mediated toxicities in NHP,²⁵ and thus, adaptive immune responses to human transgenic proteins expressed in NHPs may contribute toward the immune-mediated pathology observed following administration of AAVhu68-hSMN1 and AAV9-hGBA1. To exclude the potential for anti-transgene immune responses in AAV-mediated DRG toxicity, we evaluated the benefit of immunosuppression in NHPs that received AAV9 encoding a non-protein cargo transgene, a microRNA targeting superoxide dismutase 1 (hereafter referred to as AAV9-mir-SOD1), which has previously been reported to elicit DRG toxicity in NHP and mice.²⁶ In contrast to the NHP studies with human SMN1 or human GBA1 transgenes, the only foreign protein in the AAV-mir-SOD1 studies to drive a T-cell-mediated immune response is the capsid protein itself.

Cynomolgus macaques were dosed with vehicle or AAV9-mir-SOD1 (4.0 × 10¹³ GC) via intrathecal route of administration in the presence or absence of an IMS regimen consisting of dexamethasone (0.5 mpk, SID) and tacrolimus (1 mpk, SID), hereafter referred to as IMS#3. IMS#3 was delivered orally from day −3 to day 43. Peripheral concentrations for dexamethasone and tacrolimus at 2 h post-dose on day 42 were 144.7 ± 11.9 and 10.4 ± 7.7 ng/mL, respectively (Figure 8A). IMS#3 had no impact on transgene expression levels in DRGs in AAV9-mir-SOD1-dosed subjects compared to those without IMS#3 (p = 0.33; Figure 8B).

Figure 8.

Dexamethasone and tacrolimus reduced AAV9-mir-SOD1-mediated DRG toxicity following IT administration on day 43

Cynomolgus macaques received vehicle or 4 × 10¹³ of AAV9-mir-SOD1 by ICM administration in the presence or absence of an immunosuppression (IMS) regimen consisting of dexamethasone and tacrolimus. (A) Plasma and whole blood were collected on day 42 at 1 h post-IMS dosing to measure IMS drug concentrations for drug exposure. Each symbol represents the amount of drug in one IMS-dosed subject on day 42. (B) SOD1 miR transgene expression in DRGs on day 43 post-AAV9-mir-SOD1 administration was determined by RNAscope in situ hybridization for WPRE. Median % Area WPRE was determined in each subject at cervical, thoracic, and lumbar DRG levels. A different symbol (circle, square, triangle) is assigned to each animal/group, and each symbol represents the median value from cervical, thoracic, or lumbar DRG tissue. No significant differences were found via mixed model ANOVA. (C) Longitudinal serum levels of NF-H (pg/mL) in each treatment group as determined by ELISA. Data are plotted as mean ± standard deviation for each treatment group. ∗p < 0.05; One-way ANOVA was performed comparing NF-H values on days 15, 28, and 43 for all groups. (D) The incidence and severity of findings in the DRGs and spinal cord [DRG neuronal degeneration/necrosis, DRG mononuclear cell infiltration, spinal cord nerve fiber degeneration in cauda equina (C) and dorsal funiculus (D) and spinal cord motor neuron degeneration] is reported as the maximum severity observed per subject at the different spinal cord levels [y axis: cervical (C), thoracic (T), lumbar (L), and sacral (S)]. Each finding is represented by a dot where the diameter of the dot is scaled to indicate the severity of the finding. Missing samples are indicated by white “x” at the intersection of the line for the sample.

Through serial sampling of NHPs in this study, we observed an increase in serum NF-H over time in NHPs that received AAV9-mir-SOD1 without IMS#3, with average levels of 27,699 (±14,885) pg/mL at the terminal day 43 time point (Figure 8C). In contrast, vehicle- and AAV9-mir-SOD1 +IMS#3-treated NHP had similar levels of serum NF-H, with peak levels of 1,368 (±214) and 1,224 (±601) pg/mL, respectively (Figure 8C). Serum NF-H levels suggested minimal AAV-associated DRG toxicity in the NHPs that received AAV9-mir-SOD1 alongside the IMS#3 regimen. We further confirmed this with histopathology. Minimal-to-mild DRG neuronal degeneration/necrosis was observed in the lumbar and sacral DRGs of AAV9-mir-SOD1-injected subjects (11 of 36 DRGs; 3 of 3 subjects) in the absence of IMS#3, compared to no observations in the vehicle control group (33 DRGs evaluated; 0 of 3 subjects) (Figure 8D). Minimal-to-moderate DRG mononuclear cell infiltration was also present in three of three AAV9-mir-SOD1 subjects, predominantly in lumbar and sacral DRGs, with a couple of findings of minimal severity in thoracic DRGs. Minimal mononuclear cell infiltration was also observed in some vehicle control animals but with reduced incidence and severity compared to the AAV9-mir-SOD1 subjects without IMS#3 (Figure 8D). Control subject 41501 exhibited minimal-to-mild spinal cord nerve fiber degeneration in the thoracic and lumbar regions of the dorsal funiculus and a single finding of minimal nerve fiber degeneration in the cauda equina. This was absent in the other two vehicle-treated subjects. Spinal cord nerve fiber degeneration was exacerbated by AAV9-mir-SOD1 treatment in the absence of IMS#3 with minimal-to-moderate severity of lesions in the dorsal funiculus of the cervical, thoracic, and lumbar spinal cord in all three animals and a few additional lesions in the cauda equina of the lumbar spinal cord (Figure 8D). Administration of dexamethasone and tacrolimus strikingly reduced the incidence of AAV9-mir-SOD1-mediated DRG neuronal degeneration/necrosis (one finding of mild severity of 36 DRGs evaluated), limited the incidence and severity of DRG mononuclear cell infiltrates to five findings of minimal severity, and precluded any findings of spinal cord nerve fiber degeneration (Figure 8D).

Nerve fiber degeneration in lumbar and sacral dorsal nerve roots and bilaterally in sciatic nerves were present in all three NHPs administered AAV9-mir-SOD1 without IMS#3. Severity grades were minimal to mild in the nerve roots and minimal to moderate in the sciatic nerve (Figure S15). In the IMS#3-treated group, there was a single finding of minimal nerve fiber degeneration in the dorsal sacral nerve root (46001) but was otherwise absent at other levels within animal 46001 as well as in the other NHPs in that group (Figure S15). None of the IMS#3-treated NHPs exhibited findings of sciatic nerve fiber degeneration, demonstrating an immune-mediated mechanism that contributes toward AAV-mediated sciatic nerve fiber degeneration (Figure S15). Minimal TG neuron degeneration was observed in a single (1 of 3) NHP that received AAV9-mir-SOD1 without IMS#3, which was accompanied by immune cell infiltration in the TG. No findings of TG neuron degeneration were observed in NHP dosed with AAV9-mir-SOD1 with IMS#3, despite unilateral minimal mononuclear cell infiltration in the TG of two of the three NHPs (Figure S15).

Discussion

AAV-mediated DRG toxicity requires transgene expression and is further influenced by route of administration and dose level.^3,4,23 Yet the molecular mechanisms underlying how these factors lead to neuronal degeneration and necrosis are unknown. The initial work described herein includes a prospective study performed in cynomolgus macaques to dissect the molecular and cellular changes that precede neuronal damage in response to intra-CSF AAV delivery. The use of different necropsy time points allowed for the dissection of the kinetic responses to the AAV gene therapy administration. We identified transcriptional responses to AAV as early as day 5, including increases in ribosomal, RNA processing and complement pathways, and corresponding reductions in synaptic and mitochondrial genes. This was accompanied by innate immune activation at early time points, the production of chemokines in the CSF, and recruitment and transmigration of immune cells. These events preceded peak transgene expression and neuronal damage in DRGs, suggesting that they could contribute to AAV-mediated DRG toxicity. Contemporaneous with neuronal damage, there was an additional influx of immune cells and a second wave of transcriptional changes. The findings from this time course study (summarized in the graphical abstract) implicate a potential role for the immune response in AAV-mediated DRG toxicity. Modulation of the immune response with a glucocorticoid steroid dexamethasone and calcineurin inhibitor tacrolimus with and without MMF significantly reduced the pathology associated with the administration of three distinct AAV constructs, including an AAV vector encoding non-protein cargo. Collectively, these studies support a causal role for the immune response in AAV-mediated DRG toxicity and provide a strategy to prevent AAV-mediated DRG toxicity in gene therapy recipients.

Administration of AAVhu68-hSMN1 resulted in an increase in antiviral immune response genes, in particular interferon-stimulated genes (ISGs) in the DRGs as early as day 5. Although not detected on a transcriptomic level, chemokines known to be responsive to type I IFN [MIP1-α (CCL3), MCP-1 (CCL2) and CXCL10 (IP-10)]²⁵ trended toward (MCP-1) or exhibited (MIP1-α and CXCL10) increased protein in the CSF in AAV-dosed animals. Thus, the presence of CXCL10 and MIP1 may indirectly indicate the presence of low levels of type I IFNs, which is consistent with the increase in ISGs in the DRG transcriptome. Interestingly, CSF CXCL10 and MIP1α transcript levels correlated with serum NF-H, implicating a contribution of the immune response toward AAV-mediated neuronal damage and toxicity. As such, CSF CXCL10 and MIP1-α may be a useful biomarker for AAV-mediated toxicity in conjunction with serum neurofilament levels. Recently, elevations of CSF CXCL10 have been reported in clinical trial participants at three and six months post-intrathalamic and intrathecal administration of AAV vectors encoding HexA and HexB for treatment of Tay-Sachs and Sandhoff disease. These recent clinical reports of elevated CSF CXCL10 levels following AAV administration support the translatability of our NHP findings to humans. Longitudinal CSF chemokine measurements may better inform whether early and/or late elevations are predictive of AAV-mediated DRG toxicity. If successful, CSF CXCL10 and/or MIP1α may be especially useful in patients with neurodegenerative-disease-related increases in serum neurofilament, which obscures the use of serum neurofilament measures for AAV-mediated DRG toxicity in these patients.

While other cytokines may be induced in response to AAV exposure, we were unable to detect significant changes to the other cytokines and chemokines tested. This may be due to many factors such as route of administration, limitations in sampling CSF, and the small number of NHPs in this study along with inter-subject variability. For example, we did observe elevations in IL-6 and MCP-1 in the CSF in some subjects compared to their pre-dose values, but these changes did not reach statistical significance, especially when accounting for multiple comparisons. Levels of IL-8 and IFN-α2a were below the threshold of detection in the CSF. IL-8 can be secreted by PRR engagement on monocytes, macrophages, epithelial cells, endothelial cells, and fibroblasts. This typically occurs early after exposure to a pathogen to facilitate recruitment of neutrophils. Since our first time point post-AAV exposure was on day 5 for CSF collection, we may have missed elevations in early cytokines, such as IL-8. Furthermore, IFN-α2a may be missed due to sensitivity of the assay. Though highly potent, type I IFNs are notoriously difficult to detect due to low fg/mL abundance.

While the initial stimulus for this type I IFN signature is unknown, initial transcriptional changes on day 5 occurred prior to peak transgene expression and did not correlate with hSMN1 transgene levels, arguing against a direct role for AAV transgene expression in this primary transcriptional response. Regardless of the initial stimulus, the early production of chemokines in response to AAV is likely responsible for the primary immune cell recruitment observed in the DRGs by day 5. Thus, future work should utilize empty vectors, non-coding vectors, CpG-depleted vectors, etc. to refine the mechanisms required for the early immune response and to identify the required pathogen recognition receptors (PRRs) as well as whether this initial response is necessary for subsequent degeneration.

In the four studies reported herein in which individual DRG level histopathology findings with AAV were reported from day 15 onward, mononuclear cell infiltrates were identified in 79 of the 249 DRGs by histopathology. In contrast, only 35 DRGs exhibited neuronal degeneration; yet 97% of DRGs with neuronal degeneration (34 of these 35 DRGs) were among those showing mononuclear cell infiltrates. These data support our contention that mononuclear cell infiltrates contribute toward DRG neuronal degeneration since degeneration in NHP is rarely detected in the absence of infiltrates, while mononuclear cells are present without detectable neuronal degeneration. This is further supported by our immunohistochemistry data that identified immune cell infiltrates increasing as early as day 5 post-AAV. Our hypothesis that immune cells contribute toward DRG neuron loss is further supported by the consistent and reproducible reduction in histopathology findings across three immunosuppression studies with dexamethasone and tacrolimus with and without MMF. The addition of immunosuppression regimens consisting of dexamethasone and tacrolimus with and without MMF also reduced the incidence and severity of pathology associated with AAV administration in the sciatic nerve and nerve roots, supporting a role for the immune response in AAV-mediated pathology in peripheral nerves.

Other groups have tested the efficacy of a glucocorticoid steroid prednisolone,¹⁴ MMF, and mTOR inhibitor rapamycin,^12,13 or B-cell-depleting antibody rituximab and mTOR inhibitor everolimus¹⁴ without apparent reductions in AAV-mediated DRG toxicity. To our knowledge, this is the report of the use of dexamethasone and tacrolimus with or without MMF to pharmacologically modulate AAV-mediated DRG toxicity. While we cannot conclude the “optimal” regimen, our data support that dexamethasone and tacrolimus are sufficient to reduce AAV-mediated DRG toxicity, and MMF is not required for this effect. Since prednisolone alone was unable to reduce DRG lesions, we reason that the inhibition of T cells via the addition of Tacrolimus to an anti-inflammatory glucocorticoid steroid is responsible for observed effect. Patel et al. (2023) have reported that a triple immunosuppression regimen consisting of rapamycin, methylprednisolone, and rituximab reduced DRG pathology in NHPs receiving an AAV vector to an incidence of one in three, compared to a dual regimen of rapamycin and methylprednisolone in which the incidence was two in three NHPs.²⁷ Patel et al. (2023) posited that the apparent benefit of their triple immunosuppression regimen was derived from rituximab’s removal of CD20+ T cells. CD20+ T cells are predominantly CD8+ with an effector memory phenotype,^28,29 are present in blood as well as CSF,²⁸ and have an enhanced activation status,^28,29 which concurs with our hypothesis for a prominent role of CD8+ T cells in AAV-mediated DRG toxicity in NHP. It is unclear why rituximab-containing regimens have been unsuccessful in reducing DRG toxicity in other preclinical studies, and so more studies are required to understand the reproducibility and mechanism of rituximab-mediated efficacy.

Consistent with immune-mediated elimination of transduced cells, both DRG samples and liver samples, the only two non-CNS tissues evaluated in our time course study, had a trend toward reduced genomic copies of AAVhu68-hSMN1 on day 29 compared to earlier time points.^17,21,30 While we were unable to detect AAVhu68 capsid- or transgene-specific T cells in PBMCs by ELISpot, others have reported discordance in AAV-specific and transgene-specific T cell responses in PBMCs with tissue-isolated lymphocytes.² Unfortunately, we were unable to successfully isolate infiltrating lymphocytes from DRGs in AAV-treated NHPs to test the antigen specificity of infiltrating T cells. However, the presence of lesions in NHP following administration of AAV9-mir-SOD1 argues against an immune response to AAV protein cargo. Since the only foreign protein expressed in the DRGs following AAV9-mir-SOD1 administration is AAV capsid protein and AAV capsid protein can activate capsid-specific T cells,^{31,32,33,34,35,36,37,38} it is reasonable to postulate that the infiltrating T cells in the DRG are targeting transduced neurons following cross-presentation of capsid protein peptides in major histocompatibility complex (MHC) class I. Assuming antigen processing and presentation of AAV capsid in NHP and humans is the same as mice,³⁴ we predict that the immunosuppression regimen would need to be maintained for at least 6 months to prevent capsid-specific T cell targeting and killing of DRG neurons presenting capsid peptide-MHC.

Interestingly, we noted animal-to-animal and study-to-study variability in DRG pathology, most notably with the AAVhu68-hSMN1 study with and without immunosuppression. This is consistent with the observations from Hordeaux et al. in which they reported 83% of animals that received ICM or IT AAV developed DRG lesions.³ In our study, from day 15 onward 11/12 (91%) NHPs that were dosed with AAV via ICM or IT without immunosuppression developed DRG pathology. Subject 22501 (AAVhu68-hSMN1 no IMS#1) lacked DRG pathology but had similar levels of DRG transgene expression and AAV vector biodistribution as subjects 22001 and 22002, both of which had histopathology findings. Therefore, we do not believe the lack of findings in subject 22501 is due to neutralization from anti-AAV antibodies. However, it is possible that subject 22501 failed to mount a robust immune response to the AAV vector and as such, did not develop DRG histopathology lesions. This is also consistent with the observed variability in clinical trials of the kinetics and magnitude of T cell responses to AAV.²⁰ In contrast, we theorize that subject 24001 that received AAVhu68-hSMN1 with immunosuppression may have had a previous endemic AAV exposure, resulting in pre-existing memory capsid T cells that were insufficiently inhibited by our immunosuppression regimen. Unfortunately, we did not evaluate peripheral T cell responses from this study. Future studies should evaluate peripheral T cell responses to better understand subject-to-subject variability.

AAV-mediated DRG pathology, though adverse, is rare³ and can vary even within an animal in the same general region (e.g., L1 vs. L2 or right vs. left DRGs). Histopathology findings for ICM studies have generally been enriched in the cervical and lumbar DRGs (Figures 1, 6, and 7). In contrast, lumbar puncture IT injections have resulted in histopathology, which is mainly confined to the lumbar (and sacral) regions (Figure 8). These data suggest that route of administration, biodistribution of AAV within the CSF, and thus transduction contribute toward the regions impacted by histopathology. Yet, high transgene expression levels alone are not necessarily sufficient to cause histopathological findings in NHPs, as evidenced by relatively high transgene expression on day 9 in the DRGs in the absence of microscopic findings at this time point, as well as high levels of transgene expressing neurons present at the terminal time points, suggesting other mechanisms may be required. Yet, previous reports,^3,23 as well as data presented herein, strongly support a role for transgene expression in AAV-mediated DRG toxicity in NHPs. Consistent with a role for transgene expression, our transcriptomic analysis demonstrates that the late transcriptional changes in DRGs on days 15 and 29 correlate with transgene expression levels, which differs from cargo-expression-independent changes on day 5, suggesting two distinct mechanisms. In fact, mononuclear cells tend to surround neurons with the highest transgene expression levels. Thus, we propose that high transgene levels drive an increase in genes involved in antigen presentation such as Erap1, Tap1, and Tap2 and facilitate specific cell targeting of highly transduced neurons by cytotoxic CD8+ T cells. Unfortunately, we did not evaluate specific differences in transcriptomics between healthy and degenerating neurons. Future studies may utilize in situ transcriptomics to characterize differences between healthy versus degenerating transduced neurons. We predict that these latter transcriptomic changes would be absent in non-degenerating DRG neurons as well as in NHPs that receive only empty capsids or non-expressing vectors, and thus, they would fail to develop lesions, consistent with a previous report utilizing a non-expressing vector.⁴

While mouse DRGs induce a similar transcriptional profile as NHPs following intrathecal administration of an AAV9 vector, neuronal degeneration in mice can occur in the absence of immune infiltrates, and thus, immune infiltration into the DRG in mice is considered a secondary response to neuronal degeneration.²⁶ However, the incidence of pathology in mice is lower than in NHPs, necessitating higher numbers of mice for these studies and may account for differences in the underlying mechanism(s) of pathology. This is particularly interesting given that mice elicit poor or dysfunctional T cell responses to AAV vectors, so much so that a particular immunogenic capsid variant AAVrh32.33^38,39,40 or chimeric capsids (e.g., capsids expressing the highly immunogenic SIINFEKL peptide from ovalbumin³⁴) have been utilized by the field to study T-cell-mediated immune responses to AAV in mice. It would be of interest to test whether the use of immunogenic capsids in mice would further exacerbate this pathology and/or increase the incidence in mice.

Given the differential susceptibility to AAV-mediated neurodegeneration in the spinal cord and DRGs, we compared the baseline transcriptomic profiles of the top modules in DRG or spinal cord samples from vehicle-dosed animals. Spinal cord tissue revealed lower baseline levels of innate immune pathway genes and differential patterns of synaptic, mitochondrial, and ribosomal, RNA processing and complement gene expression when compared to the DRG transcriptomic profiles. These differences in baseline transcriptomics in the spinal cord may impact motor neurons’ ability to withstand the stress of AAV transduction (day 5) and AAV transgene expression (days 15–29) or vice versa and may contribute to DRG neurons’ susceptibility. Alternatively, differences in the magnitude of transcriptional responses at early time points following AAV administration may contribute to the differential outcomes of neuronal toxicity observed in DRGs versus the spinal cord. For example, there are no changes in mitochondrial or synaptic transcripts within the spinal cord tissues in response to AAV treatment. Similarly, there is no induction of ribosomal, RNA processing and complement transcripts within the spinal cord at any time point. The early innate immune response on day 5 that occurs prior to DRG neuron degeneration is lower in the spinal cord compared to the DRG, potentially due to differences in immune cell composition in these tissues. In contrast, the innate immune response is elevated in the spinal cord at the days 15 and 29 time points, but as these time points coincide with pathological lesions, it is unclear whether these changes contribute to the pathology or are a response to the pathology. Furthermore, overexpression of SMN1 protein in wild-type mice can lead to delayed onset motor neuron degeneration.²⁴ While our studies are on a much shorter timescale, it is debatable whether the motor neuron degeneration learnings from overexpression of SMN1 protein in healthy NHP can be broadly applied across AAVs with different transgenes.

Additionally, motor neuron degeneration was only observed in NHPs that received AAVhu68-hSMN1 without immunosuppression (4 of 14 NHPs) but not in our AAV9-hGBA1 or AAV9-mir-SOD1 groups. No motor neuron degeneration was observed in NHP receiving AAVhu68-hSMN1 in the presence of immunosuppression (0 of 3). However, since limited in number, it is hard to conclude on the benefit of immunosuppression in reducing AAV-hSMN1-dependent motor neuron degeneration. It is unclear whether AAV-hSMN1 gene therapy would cause toxicity in SMA patients who lack normal human SMN1 expression. However, elevations in neurofilament following onasemnogene abeparvovec in SMA patients⁴¹ is reminiscent of AAV-induced elevations in neurofilament in healthy NHPs, which is correlated with AAV-mediated DRG toxicity in NHPs.¹¹ Collectively, these data suggest that SMA patients would also be susceptible to the AAV-mediated DRG toxicity with the current technology, though more studies are required.

The findings described herein further our understanding of the molecular pathways altered in NHP DRGs following ICM or IT administration of an AAV gene therapy. We observed molecular (e.g., transcriptomic alterations) and cellular (i.e., immune cell infiltrates, structural) changes in the DRGs that preceded neuronal degeneration/necrosis and nerve fiber degeneration. These changes were enriched in the DRGs compared to the spinal cord, suggesting a causal role in DRG sensory neuron degeneration. Based on these results, we tested whether pharmacologic modulation of the immune response to AAV was able to blunt the pathology associated with AAV gene therapy and found reduced pathology in DRGs and peripheral nerves of NHPs that received dexamethasone and tacrolimus immunosuppression. Finally, this dataset also provides compelling rationale for the incorporation of CXCL10 or MIP1-α as an additional biomarker for monitoring AAV-mediated toxicity in the clinic, which will broaden our understanding of the frequency of asymptomatic AAV-mediated toxicity in humans.

Materials and methods

AAV vectors

AAVhu68-ss-HSMN1 is an AAVhu68 serotype vector harboring a single-stranded (ss) cargo encoding the CMV enhancer, chicken b-actin promoter, and codon-optimized (co) human survival of motor neuron 1 (SMN1) transgene, as previously described.² AAV9-ss-GBA1 is an AAV9 serotype vector containing ss cargo encoding a codon-optimized version of human acid beta-glucosidase 1 (GBA1) transgene under the control of the CAGGS promoter, followed by a woodchuck hepatitis virus post-regulatory element and a strong polyadenylation signal. AAV9-mir-SOD1 is an AAV9 serotype vector containing ss cargo encoding an artificial microRNA (amiR) targeting superoxide dismutase1 (SOD1), as previously described.²⁶ Vectors were produced by triple transfection in HEK293 cells, as previously described.²¹ Vectors were purified using POROS Capture Select AAV9 resin (Thermo Fisher Scientific, Waltham, MA) followed by affinity chromatography. Vectors were formulated in artificial CSF (0.2 mM NaH₂PO₄ꞏ2H₂O, 0.8–1 mM Na₂HPO₄, 150 mM NaCl, 3 mM KCl, 1.4 mM CaCl₂, 0.8–1 mM MgCl₂ ꞏ6H₂O) with 0.001% Pluronic F-68 at a pH 7.2. Genome copy titer was determined by droplet digital PCR. AAVhu68-hSMN1 Lot PPD-19-1759 and Lot AD-19-0424 were quantified as 3.68 × 10¹³ and 3.5 × 10¹³ genomic copies (GC)/mL, respectively. AAV9-hGBA1 and AAV9-mir-SOD1 were quantified as 3.0 × 10¹³ and GC/mL and 4.0 × 10¹³ GC/mL, respectively. Each vector had ≤0.13 endotoxin units/mL.

Study designs

Time course study with AAVhu68-hSMN1

Asian cynomolgus macaques (2–4 years of age, N = 2–3 per group), eight male and seven female (N = 15 total), were randomized into five groups and dosed with vehicle (group 1) or 3.68 × 10¹³ GC AAVhu68-hSMN1/subject (groups 2–5) by a single slow bolus (target rate 1 mL/min), percutaneous intra-cisterna magna (ICM) injection in a total dose volume of 1 mL with a 22-gauge syringe on day 1.

AAVhu68-hSMN1 immunosuppression study #1

Asian cynomolgus macaques (2–4 years of age, N = 3/group), six male and three female (N = 9 total) were sex-matched and randomized into three groups and dosed with vehicle or 3.5 × 10¹³ GC of AAVhu68-hSMN1/subject by a single slow bolus, percutaneous ICM injection in a total dose volume of 1 mL with a 22-gauge syringe on day 1. Immunosuppression (IMS) dosing started on day 2 and continued throughout the study until euthanasia (day 29). Dexamethasone (in ultrapure water), tacrolimus (in ORA-Plus), and MMF (in 2%HPMC E5/1% Tween 80) were administered SID at 0.5 mpk, SID at 1 mpk, and BID at 50 mpk/dose, respectively. Additionally, dexamethasone in artificial CSF was dosed as a single 0.5 mg injection via lumbar IT route 10 min prior to AAV injection. Peripheral blood was collected in K₂EDTA on day 28 at 6 h post-IMS administration and analyzed as whole blood or plasma for IMS concentrations. Terminal collections for tissues were performed on day 29.

AAV9-hGBA1 immunosuppression study #2

Asian cynomolgus macaques (2–3 years of age, N = 3/group), six male and three female (N = 9 total) were sex-matched and randomized into three groups and dosed with vehicle or 3.0 × 10¹³ GC of AAV9-hGBA1/subject by a single slow bolus, percutaneous ICM injection in a total dose volume of 1 mL with a 22-gauge syringe on day 1. IMS dosing started on day −2 and continued throughout the study until day 43. Dexamethasone (in ultrapure water), tacrolimus (in ORA-Plus), and MMF (in 2%HPMC E5/1% Tween 80) were administered SID at 0.5 mpk, SID at 1 mpk, and BID at 50 mpk/dose, respectively. Peripheral blood was collected in K₂EDTA 1 h post-IMS administration on day 1 and analyzed as whole blood or plasma for IMS concentrations. Tissue collection at necropsy was performed on day 43.

AAV9-mir-SOD1 immunosuppression study #3

Asian cynomolgus macaques (2–3 years of age, N = 3/group), five male and four female (N = 9 total) were randomized into three groups and dosed with vehicle or 4.0 × 10¹³ GC of AAV9-mir-SOD1/subject by a single slow bolus, percutaneous lumbar IT injection in a total dose volume of 1 mL with a 25-gauge syringe on day 1. Immunosuppression dosing started on day −3 and continued until study termination (day 43). Dexamethasone (in ultrapure water) and tacrolimus (in ORA-Plus) were administered SID at 0.5 mpk and 1 mpk SID, respectively. Peripheral blood was collected in K₂EDTA 1 h post-IMS administration on day 42 and analyzed as whole blood or plasma for IMS concentrations. Tissue collection at necropsy was performed on day 43.

In vivo study execution

ICM injections were performed by freehand-guided injections with a 22-gauge, 1.5-inch spinal needle (BD, 405161) and confirmed via visual verification of CSF flow during pre-dose CSF collections (AAVhu68-hSMN1) or by fluoroscopy without contrast (AAV9-hGBA1). Additionally, a 0.25 mL flush of the vehicle (artificial CSF [aCSF]) was administered following ICM injection of AAVhu68-hSMN1. IT injections for AAV9-miR-SOD1 were performed by freehand-guided injections with an atraumatic needle [B Braun (P2510), 25-gauge, 1 inch]. Animals were fasted a minimum of 6 h prior to dosing and anesthetized with cocktail of ketamine, dexmedetomidine, and glycopyrrolate and then maintained with isoflurane. Subjects were dosed in a lateral recumbency position and remained in a prone position for 15 min after dosing was complete before administration of atipamezole. Animals were monitored at 4 h (±30 min) following dosing. In-life parameters and endpoints evaluated in this study included morbidity and mortality, body weight and food intake, clinical signs, body temperature, physical examinations, neurological evaluations, clinical pathology (hematology, coagulation, and clinical chemistry), and CSF analyses (cell counts and clinical chemistry). Hematology and coagulation were performed on blood and plasma samples, respectively. Clinical chemistry was performed on serum. CSF cell counts and clinical chemistry were analyzed on day 1 prior to ICM injection and at necropsy time points.

Additional ex vivo analyses (gross findings, organ weights, microscopic evaluation, vector DNA biodistribution, transgene RNA expression, RNA-seq, ELISpot, serum and CSF analysis for cytokines, in situ hybridization (ISH), immunohistochemistry (IHC), and/or electron microscopy) were performed using whole blood, serum, and/or tissue samples obtained during the study. Whole blood for PBMC isolation was collected in sodium heparin tubes. Additionally, 2.5 mL whole blood was collected in PAXgene tubes for RNA isolation and sequencing. Blood was processed to serum or plasma as indicated for relevant analyses. CSF was collected prior to dosing and at necropsy.

The care and use of the Asian cynomolgus macaques on the study were conducted in accordance with the guidelines of the Guide for the Care and Use of Laboratory Animals and the Canadian Council on Animal Care. The study was approved by the Institutional Animal Care and Use Committee (IACUC) at the testing facility where the study was conducted. All cynomolgus macaques were socially housed, when possible, fed two times daily with Lab Certified Primate Diet 5048, provided with toys and food enrichment, and water ad libitum, except during designated procedures.

PAXgene whole-blood collection and processing

Whole blood (2.5 mL) was collected in PAXgene tubes (Qiagen, Germantown, MD) with 7 mL of additive, inverting gently 8–10 times to mix. Tubes were incubated at room temperature for at least 2 hours before transferring to a −20^oC freezer for at least 24 h. PAXgene tubes were then stored at −80^oC until use.

CSF collection and processing

CSF samples were collected pre-dose on day 1 and at necropsy while animals were under anesthesia using a 22-gauge spinal needle; 0.4 mL CSF sample was collected in K₂EDTA (BD, Franklin Lakes, NJ) for CSF cell counts or 0.3 mL was collected without anticoagulant for clinical chemistry. Remnant CSF was placed on dry ice and stored at −80^oC until use.

Serum collection and processing

Blood samples were collected in serum separator tubes and centrifuged for 10 min at 1,400 × g at 4^oC. Serum was collected, aliquoted, and stored at −80^oC until use.

Plasma collection and processing

Blood samples were collected in K₂EDTA tubes and placed on wet ice until processing. Samples were centrifuged for 10 min at 1,400 × g at 4^oC. Plasma was collected, aliquoted, and stored at −80^oC until use.

Tissue collections and histopathology

Cervical, thoracic, and lumbar spinal cord with dorsal/ventral nerve roots and DRGs, trigeminal ganglia (TG) with trigeminal nerve, brain, median nerve, sciatic nerve, liver, and kidney were collected and preserved in 10% neutral buffered formalin for 1–2 days, paraffin embedded, sectioned, and stained with hematoxylin and eosin (H&E). Tissues were evaluated microscopically and graded for histopathology features (e.g., neuronal degeneration/necrosis, nerve fiber degeneration, and mononuclear cell infiltration) by a board-certified veterinary pathologist. Microscopic findings were graded, as previously described.¹² Briefly, findings were graded on a 1–5 scale, with grade 1 being minimal in severity and affecting less than 10% of tissue; grade 2 being mild and affecting approximately 10%–25% of tissue; grade 3 being moderate and affecting approximately 25%–50% of tissue; grade 4 being marked and affecting 50%–90% of tissue; and grade 5 being severe and affecting greater than approximately 95% of tissue.

Electron microscopy

DRG L5-L6 and spinal cord L4 were collected, trimmed, and preserved in McDowell-Trumps fixative (4.0% formaldehyde–1.0% glutaraldehyde in phosphate buffer) and stored at 4°C until routine processing to epoxy resin blocks. The blocks were thick-sectioned at approximately 1 micron and stained with toluidine blue for review by light microscopy to identify the region of interest for ultrastructural evaluation. The region of interest was thin-sectioned in the range of 70–90 nm and post-stained with methanolic uranyl acetate and lead citrate for transmission electron microscopy. A minimum of 10 digital images were collected per tissue sample per animal (DRG L5, L6 and spinal cord L4; total of 528 digital images) at direct magnifications ranging from 500× to 15,000× (print magnifications of 2,850×–86,300×; AMT XR16MP digital camera system) and provided to a board-certified veterinary pathologist for evaluation.

In situ hybridization and immunohistochemistry

Tissues collected for ISH and IHC were fixed in 10% neutral buffered formalin for 1 to 2 days and then processed into paraffin blocks. Blocks were processed into 5 μm sections. For time course tissues, RNAscope was performed at Invicro (Needham, MA) for transgene expression (Hs-SMN1-O1); at Advanced Cell Diagnostics (ACD, Newark, CA) for pDC marker CD303 (Mfa-Clec4c), NK cell marker NkP46 (Mfa-Ncr1), positive control probe (Mfa-PPIB), and negative control probe (DapB); and at Biogen for type-I-IFN-stimulated gene Cxcl10 (Mfa-CXCL10-C1), according to manufacturer’s instructions. RNAscope analyses for transgene expression in immunosuppression studies for human codon-optimized SMN1 (Hs-SMN1-O1), human codon-optimized GBA1 (Hs-GBA1-O1), and WPRE for miR-SOD1, as well as the positive control probe (Mfa-PPIB) and negative control probe (DapB), were performed at Biogen, according to manufacturer’s instructions. Signal was detected using the fully automated RNAscope assay on a BOND Rx platform (Leica Biosystems, Deer Park, IL) or Ventana Discovery Ultra (Roche, Switzerland). IHC was performed on a Ventana Discovery Ultra platform for CD8 (clone SP57, Roche, Switzerland), CD20 (clone L26, Roche, Switzerland), CD68 (clone KP-1, Roche, Switzerland), CD4 (clone EPR6855, Abcam, Waltham, MA), and cleaved caspase 3 (clone 5A1E, Cell Signaling Technology, Danvers, MA). Slides were scanned with Pannoramic-250 (3DHISTECH, Hungary), and quantitative whole slide analysis was performed with Visiopharm software (Denmark). Using an AI-based algorithm, DRG neurons were segmented for analysis of transduction efficiency and/or RNA expression analysis within neurons and non-neuronal cells. Percent area was used as the metric for quantification of hSMN1, hGBA1, and WPRE transgenes, CLEC4C and NCR1 ISH, as well as CD4, CD8, CD68, and CD20 IHC using Visiopharm software.

Vector DNA biodistribution

Briefly, 25 mg of frozen tissue was treated with Proteinase K and RNase and lysed. Genomic DNA was extracted using QIAamp DNA Mini Kit (Qiagen, Germantown, MD) per the vendor’s instructions. Quantitative PCR was performed (TaqMan Universal PCR Master Mix, No AmpErase UNG, Applied Biosystems, Waltham, MA) to evaluate vector DNA biodistribution (vector transgene DNA) normalized to the endogenous cynomolgus macaque housekeeping gene RPP30, ribonuclease P protein subunit p30, in single-plex reactions on CFX96 System (Bio-Rad, Hercules, CA). RPP30 from IDT (Coralville, IA) (Forward: 5′-TGAAATGAAGCTGT TGATTTCAACACACAAATTCTG-3’; Reverse: 5′-CCTCAGGCATGCTGGGAGC-3’; Probe 5’/5HEX/CCCTAGTCA/ZEN/ACTGGAGGTAGAGACGGAC/3IABkFQ-3′), and the information of transgene primers and probes is proprietary to Biogen. The cycling conditions were 10 min at 95°C, 40 cycles of 15 s at 95°C, and 1 min at 60°C.

RNA sequencing

PAXgene tubes (Qiagen, Germantown, MD) were thawed at 4°C overnight. Total RNA was extracted as recommended by the vendor protocol.

DRG and spinal cord tissues were homogenized in Qiazol lysis reagent (Qiagen, Germantown, MD) in 2 mL deep-well plates (Costar) with 2.3 mm chrome beads (BioSpec Products). Homogenate was mixed 1:5 with chloroform, incubated for 3 min at room temperature, and centrifuged at 6 000 rpm at 4^oC for 15 min. The aqueous layer was collected and mixed with equal volume 70% ethanol. The samples were then applied to RNeasy 96-well plate, and RNA was extracted according to manufacturer’s instructions. RNA was quantified by Nanodrop 8000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA), and RNA integrity was evaluated by Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA).

RNA-seq libraries were prepared from whole-blood RNA extracted from PAXgene tubes using KAPA RNA HyperPrep with RiboErase (HMR) + Globin (Roche, Switzerland) from 400 ng of input RNA according to manufacturer’s instructions. RNA-seq libraries from tissue RNA were prepared using KAPA RNA HyperPrep with RiboErase (HMR) (Roche, Switzerland) from 100 ng RNA from DRGs and 200 ng RNA from spinal cord. Libraries were sequenced on a NovaSeq 6000 (Illumina, San Diego, CA), paired end 2 × 50 bp targeting a minimum of 40 million fragments for each sample.

The FASTA sequence for the AAVhu68-hSMN1 vector was appended to a custom Macaca fascicularis genome reference. Raw sequencing data were aligned using an in-house pipeline; reads were aligned to the custom reference genome using STAR⁴² and transcripts quantified with RSEM.⁴³ Expression data and comparisons were uploaded to Quickomics for downstream visualization.⁴⁴ TPM count data were used as input into Weighted Gene Correlation Network Analysis.⁴⁵ TPM values were log-transformed, and the WGCNA package function pickSoftThreshold was run on the resulting expression matrix to select a soft thresholding power based on scale-free topology fit index and mean connectivity. To reduce the effect of sample outliers on observed co-expression patterns, robust WGCNA was performed.⁴⁶ The expression matrix was randomly subsampled for 75% of samples 100 times, and for each of these subsamples a signed adjacency matrix was constructed. A consensus adjacency matrix was generated by keeping the 20^th percentile value across all subsamples for each gene-gene adjacency. This consensus matrix was used as input to generate a signed topological overlap (TO) matrix. 1 – TO was used as a dissimilarity measure to construct a gene dendrogram using average linkage hierarchical clustering, and the cutreeDynamic function was used to assign genes to modules using deepSplit of four and a minimum module size of 50. Modules were subsequently cleaned by removing genes with kME less than 0.4, and modules no longer reaching the minimum module size, or with fewer than 17 core genes with kME greater than 0.9, were disbanded. Finally, modules with similar co-expression patterns were merged using the mergeCloseModules function, iteratively merging modules with Pearson correlation of eigengenes greater than 0.9. This final module set was annotated for gene ontology enrichment with the topGO package.⁴⁷

Cytokine and chemokine ELISAs

Blood from time-course study subjects was collected into serum separator tubes at week −1; day 1 at 12, 24, 48, and 72 h post-dose; and on days 5 (all groups), 7 and 9 (groups 1, 3, 4, and 5), 13 and 15 (groups 1, 4, and 5), and 19, 22, and 29 (groups 1 and 5). Samples were centrifuged at 1,400 × g at 4^oC for 10 min and stored at −80^oC until use. CSF was collected at 0 h and at necropsy and stored at −80^oC until use. IFN-γ, IL-1β, IL-2, IL-6, IL-8, and IL-10 were evaluated by Meso Scale Discovery (MSD) V-plex ELISA (MSD, Rockville, MD). MIP-1β, eotaxin-3, TARC, CXCL10, MIP-1α, IL-8, MCP-1, MDC, and MCP-4 were evaluated by V-plex ELISA (MSD, Rockville, MD). IFN-α2a and TNF-α were evaluated by U-plex ELISA (MSD, Rockville, MD). Assays were qualified for performance in CSF and serum matrices for dilution linearity and spike recovery before testing experimental samples. All assays were performed using manufacturer’s instructions, and samples were diluted 2-fold in Diluent 43 (MSD, Rockville, MD). Due to limited CSF volume, samples from day 9 were only evaluated on the chemokine panel (CXCL10, MIP1α, MCP1, and MDC).

Serum neurofilament heavy

Serum samples were evaluated by ELISA for NF-H (Bio-Techne, Minneapolis, MN). Samples were diluted 6-fold in the provided sample diluent, and 50 μL were loaded per well in a 72-sample cartridge.

ELISpot assay

PBMCs were isolated from heparinized whole blood samples (sample volume 5 mL in-life, 10 mL at necropsy) using a Lympholyte-Mammal (Cedar Lane Labs, Burlington, NC) density gradient for pre-dose and day 5 or Ficoll-Paque (Cytiva Life Sciences, Marlborough, MA) to improve PBMC recovery for days 9, 15, and 29. Isolated PBMCs were divided into three aliquots of 1 × 10⁷ cells/aliquot or more if available and cryopreserved until analysis of the T cell response by an ex vivo ELISpot assay (R&D Systems, Minneapolis, MN) for measurement of IFN-γ release upon restimulation. Briefly, 200,000 cells were loaded per well of an ELISpot plate coated with anti-human IFN-γ antibody. Cells were stimulated in triplicate overnight, as PBMC numbers allowed, using the following conditions: (1) negative control (medium alone), (2) positive control (50 ng/mL Phorbol 12-myristate 13-acetate [PMA] with/500 ng/mL ionomycin; Sigma), or (3) 2 μg/mL of antigen [2 μg/mL of each antigen-specific peptide (15 amino acid peptide pools from the (3a) AAVhu68 capsid or (3b) hSMN1 transgene) from JPT (Germany)].

Immunosuppression bioanalytical sample processing and analysis

Blood was processed to plasma for bioanalysis of dexamethasone and MMF, as outlined above. Whole blood was collected for bioanalysis of tacrolimus. Blood and plasma samples were analyzed for the concentration of dexamethasone, mycophenolic acid (the active metabolite of prodrug MMF), and tacrolimus by LC-MS/MS.

Statistics

Correlations were evaluated using Spearman rho value. Other quantitative datasets were analyzed by one-way ANOVA and mixed model ANOVA, adjusting for multiple comparisons testing. Specific tests are noted in figure legends. Significance was determined using a 0.05 p value cutoff after multiple comparisons adjustments.

Data availability

Data are available upon request from the corresponding author.

Acknowledgments

The authors thank the Wilson Lab and the former Gene Therapy Program of University of Pennsylvania for the design of the AAVhu68-hSMN1 vector. The authors also thank Chelsea Worley, Michael Nguy, Merilla Michael, PJ De Giovanni, Shawn Borque, and Terrance Dobrowsky for vector production and characterization. This work was funded by Biogen.

Author contributions

Conceptualization, A.P., J.B., B.G., E.T., D.M., and L.B. Data curation, S.J.C., K.L.H., M.S., B.G., A.S., V.C., S.W.-H., S.H., S.X., P.C., E.D.P., F.Z., A.W., D.B., P.C., and K.N. Analysis and/or supervision, S.J.C., M.S., B.G., A.S., V.C., S.W.-H., N.P.v.d.M., S.-C.L., S.H., S.X., A.J.G., M.K., J.D., G.B., K.G., D.G., J.G., J.C., E.L., E.D.P., A.G., T.R., E.T., K.Z., D.H., J.S., J.F., T.M.C., P.T., P.N., S.B., D.M., and L.B. Writing, K.L.H., B.G., M.S., V.C., and E.T. Review and editing: all authors.

Declaration of interests

All authors, except V.C. and S.W.-H., were employed by Biogen at the time of their contribution to this work. Biogen employees own Biogen stock and other equities.