Impact of pre-existing immunity on safety and biodistribution of a single AAV9 vectorintrathecal injection in cynomolgus monkeys

Abstract

Adeno-associated virus (AAV) vectors are widely used for gene therapy as they deliver therapeutic genes with minimal toxicity. However, pre-existing anti-AAV immunity in a large percentage of humans poses a challenge to their efficacy and safety; only patients with low or no AAV-specific antibodies are currently eligible for systemically administered gene therapies. The impact of pre-existing anti-AAV total antibodies (TAbs) following local delivery is less characterized. This study explored the impact of pre-existing anti-AAV9 TAbs on biodistribution and safety of an AAV9 tool vector administered intrathecally to cynomolgus monkeys. Although high-serum AAV9 titers did not affect central nervous system vector biodistribution, it reduced distribution to peripheral tissues, except the spleen in which vector genome copies increased. Animals with high pre-existing antibodies had rapid and transient systemic interferon response, faster therapy-emergent antibody increases in the cerebrospinal fluid (CSF), and higher cellular responses in the spleen but no major safety concerns. Our findings suggest that using serum AAV titers as an eligibility criterion for intrathecal AAV-based gene therapy could exclude patients who might benefit from treatment. Stronger anti-vector immune responses in animals with high anti-AAV9 TAb titers suggest the need to closely monitor treatment-emergent immune responses and potential consequences for highly seropositive patients.

Graphical abstract

Huang and colleagues evaluated the effect of high pre-existing serum anti-AAV9 TAb titers (up to 1:327,680) on the biodistribution and safety of a scAAV9-CBA-mCherry vector delivered to cynomolgus monkeys via a single intrathecal injection at a nominal dose of 1.2 × 10¹³ vg/animal.

Introduction

The number of approved or late-stage investigational viral vector gene therapies has been rapidly growing, and adeno-associated virus (AAV) vectors^1,2 remain the leading platform for gene delivery. AAV vectors have become the vector of choice because of their broad tissue tropism and demonstrated safety profile compared with other viral vectors.^3,4 As the number of AAV-based gene therapies continues to grow, there is increasing need to better understand the impact that host immune responses may play in the safety, efficacy, and long-term durability of these treatments.^5,6,7

The presence of pre-existing anti-AAV immunity in a large percentage of the human population, through natural exposure to AAV infections,^8,9,10,11,12 remains a major challenge for AAV-based gene therapies, especially following systemic delivery. Pre-existing immunity can reduce their efficacy and safety through neutralization of the vector with transduction inhibition, clearance of the transduced cells, and inflammatory responses.¹³

Consequently, only patients with low or no vector-specific pre-existing antibodies are currently eligible for systemic gene transfer, although the titer threshold for excluding patients varies substantially between trials and programs: neutralizing antibodies (NAbs; range, 1:1–1:1,200) and total antibodies (TAbs; range, 1:5–1:1,600).¹⁴ In some studies, eligibility thresholds appeared to be less restrictive given the availability of additional preclinical¹⁵ and clinical data.¹⁶ Preclinical nonhuman primate (NHP) studies¹⁷ informed the eligibility criteria for subsequent clinical studies. For example, in a phase 1/2 clinical trial of a systemically administered gene therapy, delandistrogene moxeparvovec for Duchenne muscular dystrophy (DMD; NCT03375164), a pre-existing binding antibody titer of 1:400, was used as the cutoff.¹⁸ In contrast, patients with anti-AAVRh74var NAb titers ≥1:1 are not eligible for fidanacogene elaparvovec, a recently approved treatment for adults with moderate to severe hemophilia B.¹⁹ Importantly, assay differences could lead to variability in titers, which are not always comparable with each other.²⁰ Furthermore, functional data linking selected pre-existing titer thresholds with observed transduction inhibition or safety events are often sparse or absent.

The potential impact of pre-existing antibodies on the effectiveness of gene therapies is influenced by their route of administration. Local administration (e.g., intrathecal, intravitreal, and subretinal) to an immune-privileged site may pose a lower risk of antibody-mediated capsid neutralization or immune complex formation compared with systemic administration.²¹ Clinical and nonclinical studies have consistently demonstrated that there are lower levels of antibodies, including lower immunoglobulin G (IgG), within the intra-cerebrospinal fluid (CSF) compared with blood,²² as well as AAV9-specific antibody titers in both adults and children.²³ This finding has since been verified with other AAV serotypes, including AAV5.²⁴ In addition, the rate of IgG clearance from the CSF is much more rapid than the rate of IgG diffusion from the blood into the CSF,²⁵ resulting in lower anti-AAV9 antibody titers in the CSF. Therefore, reliance on serum anti-AAV antibody titers as an eligibility criterion for treatment with locally administered gene therapies could lead to the unnecessary exclusion of some patients.

CSF administration of AAV-based gene therapies to the central nervous system (CNS) has been recently used in numerous nonclinical and clinical studies with several potential advantages, such as lower dosages, direct exposure to the spinal cord and brain parenchyma, and potential evasion of circulating pre-existing anti-AAV antibodies.^26,27,28,29 Previous studies in NHPs examined AAV intra-CSF administration in the presence of low-to-medium levels of serum pre-existing antibodies, albeit using different intra-CSF administration routes and vector doses.^27,29 These studies reported no noticeable difference in AAV9-mediated gene transfer in the CNS between seronegative (defined as absence of neutralizing activity in an NAb assay) and seropositive animals (maximum TAb titer 1:400; maximum NAb titer 1:3,125),²⁹ although differences were observed in liver enzyme activities, histopathology, and protein biomarkers of neural injury, inflammation, and organ damage.

There is also evidence of faster therapy-emergent humoral responses in animals with pre-existing immunity; however, data on innate and adaptive cellular immune responses are limited. These data are important for maximizing efficacy and minimizing potential safety concerns.

Here, we investigated the effect of high pre-existing serum anti-AAV9 TAb titers (up to 1:327,680) on the biodistribution and safety of a scAAV9-CBA-mCherry vector delivered to cynomolgus monkeys via a single intrathecal injection at a nominal dose of 1.2 × 10¹³ vector genomes (vg)/animal. A comprehensive assessment of AAV9-vector-mediated innate and adaptive immune responses, including cellular responses, was performed in addition to clinical and pathology safety evaluations. Local CNS delivery can bypass pre-existing serum antibodies but cannot prevent recalling AAV capsid-specific memory immune responses. The data presented here demonstrated that high-serum AAV9 capsid-specific antibody titers do not impact the vg CNS biodistribution. We observed stronger AAV9-driven immune responses in the presence of high levels of pre-existing antibodies, but clinical evaluations and pathology assessments excluded major safety concerns for all animals over a 4-week period. Altogether, our findings suggest that using serum anti-AAV TAb titers as a criterion for patient eligibility for intrathecal AAV-based gene therapy could unnecessarily exclude patients who may benefit from these treatments. The stronger anti-vector immune responses observed in the presence of high levels of pre-existing vector-specific antibodies suggest a potential need to closely monitor treatment-emergent immune responses and potential consequences in seropositive patients with high pre-existing anti-AAV TAb titers.

Results

Development of treatment-emergent serum and CSF anti-AAV9 TAb titers

The study design and animal assignment to dose groups are summarized in Figure 1A. Animals in group 1 received the vehicle control article and did not have a specific threshold for pre-dose serum anti-AAV9 TAb titers. Animals in groups 2 through 4 received a single intrathecal injection of scAAV9-CBA-mCherry vector at a dose of 1.2 × 10¹³ vg/animal but had increasing levels of pre-existing anti-AAV9 TAb titers: low (group 2, ≤1:20), medium (group 3, 1:160–1:640), and high (group 4, ≥1:20,480) and were assigned based on the day −9 or day –14 anti-AAV9 TAb titers. AAV9-treated animals had a similar body weight at day 1, except for one animal in group 3 (5.4 kg) that received a slightly lower vector dose (Figure 1A). Results for this animal were in the same range as all the other animals across the different read-outs. Importantly, study animals did not receive immunosuppressants that could have dampened treatment-emergent immune responses.

Figure 1.

Anti-AAV9 TAb response pre-dose and post-scAAV9-CBA-mCherry intrathecal administration

Study design (A). Graphs demonstrate endpoint anti-AAV9 TAb titer pre-dose (day −14 or day −9) and following scAAV9-CBA-mCherry intrathecal administration in the serum (B) or CSF (C). Dashed line denotes lowest reportable titer (1:20). Data are presented as median values with range. A nominal value of 10 is assigned to all titer endpoints <1:20. For serum titers, all groups are significantly different from the others across time points with the exception of the low- and medium-titer groups (mixed model analysis; not shown on graphs). AAV9, adeno-associated virus serotype 9; CSF, cerebrospinal fluid; TAb, total antibody.

Prior to dosing (day −14/−9 to day 1), changes in anti-AAV9 TAb serum titers occurred for a few animals in the low- (n = 2/4 group 2) and medium-titer (n = 1/4 group 3) groups, with two animals in the low-titer group reaching TAb titers similar to those in the medium-titer group prior to AAV9 dosing, which might be related to seroconversion or normal titer fluctuation over time. Therefore, pre-dosing titer ranges at day 1 for the low- and medium-titer groups fell outside the targeted titer ranges and overlapped. The AAV9 titer ranges of the high-titer group (group 4) remained well separated from the low- and medium-titer groups. This had no impact on the conclusion of the study.

In contrast, pre-existing CSF anti-AAV9 TAbs were absent for animals in the control, low-, and medium-titer groups. Animals in the high-AAV9-titer group (group 4) had measurable but low anti-AAV9 TAb titers in the CSF compared with serum titers (64- to 256-fold; Table S1).

Following scAAV9-CBA-mCherry vector intrathecal dosing, all animals developed serum treatment-emergent anti-AAV9 TAbs, reaching high levels resulting from humoral memory by day 8. Animals in group 4 had higher serum anti-AAV9 TAb titers than those in groups 2 and 3 at all tested time points. Serum titers remained elevated at the end of study (day 30) (Table S2).

In the CSF, treatment-emergent antibodies were also observed in all animals administered scAAV9-CBA-mCherry but at lower levels compared with serum. Animals in group 4 had higher levels of treatment-emergent antibodies in the CSF by day 15 postinjection compared with groups 2 and 3 (Figures 1B and 1C; Table S3).

High-serum anti-AAV9 TAb titers did not impact CNS vg biodistribution

We next analyzed mCherry vg biodistribution from 16 different tissues collected at necropsy (4 weeks post-dosing). Because of the limited number of animals exhibiting large inter-animal variability in mCherry vector DNA tissue concentrations and overlapping titer ranges for groups 2 and 3, analysis of the impact of pre-existing anti-AAV9 TAbs by descriptive statistics was challenging. Therefore, statistical analysis (linear trend test) was used for between-group comparisons; we also tested whether the high-titer group was different from the low- and medium-titer groups combined (Figure 2).

Figure 2.

Vg biodistribution at 4 weeks

Biodistribution of scAAV9-CBA-mCherry vg normalized to reference gene, CFTR, across different tissues (A, B). Bar graphs show median values, with each dot being an individual value. Mixed-effects statistical analysis between group 4 (high anti-AAV9 TAb titer) vs. groups 2 (low anti-AAV9 TAb titer) and 3 (medium anti-AAV9 TAb titer) combined. Numbers at the bottom of each bar denote the number of animals for each assessment. ∗p < 0.05, ∗∗p < 0.01. Note: for values below the LOQ, such as with the vehicle group (group 1), 1/10 the value of the lowest data point above LOQ was plotted so that the data are represented on graphs. AAV9, adeno-associated virus serotype 9; DRG, dorsal root ganglia; LOQ, limit of quantification; TAb, total antibody; vg, vector genome.

As presented in Figure 2A, high serum levels of AAV9-specific TAbs (up to 1:327,680) did not impact vg biodistribution to the CNS, including the spinal cord. However, vg biodistribution to some peripheral tissues, such as to the dorsal root ganglia (DRG), liver, and heart, was reduced in these animals. As previously reported by Wang and colleagues,³⁰ high levels of pre-existing anti-AAV9 TAbs directed the vector to the spleen. No statistical differences between the tested groups were observed in muscle tissues, lung, kidney, and ovary.

Higher AAV9-specific T cell responses were observed in the spleen of treated animals with high levels of pre-existing anti-AAV9 TAbs

In the spleen, AAV9 capsid-specific T cell responses, which play a critical role in immune responses against viral-vector-based therapies, were assessed in all study animals using a multiplexed activation-induced marker (AIM) assay, which measures the upregulation of T cell activation surface markers following T cell receptor stimulation with the cognate antigen (AAV9-capsid-protein-derived peptides).

Inclusion of T cell lineage markers (CD3, CD4, and CD8) and memory markers (CD197 and CD45RA) in the flow cytometry staining panel allowed us to monitor AAV9-specific responses in different CD4 and CD8 subsets as demonstrated in a representative gating strategy (Figure S1A). For both CD4 and CD8 populations, frequencies of naive cells (CD197+CD45RA+), memory subsets (CD197+CD45RA− T central memory and CD197−CD45RA− T effector memory), and terminally differentiated (CD197−CD45RA+) cells were similar between study groups (Figure S1B).

Stimulation with AAV9 peptides recalled AAV9-specific CD4+ T cell responses in most AAV9-treated animals. Stronger AAV9-specific CD4+ T cell responses were observed in the spleens of animals with high pre-existing anti-AAV9 TAbs (group 4; Figure 3). Data from the six different AIM pairs led to the same conclusion, although the high anti-AAV9 TAb titer group (group 4) was significantly different from groups 2 (low) and 3 (medium) combined for only three AIM pairs (CD137+CD25+, CD69+CD25+, and CD134+CD25+) for both total CD4 (Figure 3C) and CD4 memory (Figure 3D) responses. AAV9-responding T cells were predominantly CD4+ memory T cells, including central (Tcm) and effector (Tem) memory cells (Figures 3B and 3D). Because numerous cells were transitioning from Tcm to Tem cells, we analyzed AAV9-specific responses in the whole CD4+ T cell memory compartment. All AAV9-treated animals had similar frequencies of Tcm and Tem, and AAV9-specific activated memory T cells were falling in both Tcm and Tem gates with no qualitative differences (Figures 3E and 3F).

Figure 3.

Stronger AAV9-specific CD4 T cell responses observed in the spleen of immunized animals with high levels of pre-existing anti-AAV9 Tabs

Representative flow cytometry plots depicting selected AIM pairs demonstrating AAV9-specific total CD4 (A) and CD4 memory responses (B). Boxplot graphs depict data expressed as stimulation index (ratio of the signal from cells stimulated with AAV9 peptides over the signal from cells stimulated with DMSO only) for all AIM pairs for both total CD4 (C) and CD4 memory responses (D); dots indicate individual animals. Mixed-effects statistical analysis between group 4 (high anti-AAV9 TAb titer) versus groups 2 (low anti-AAV9 TAb titer) and 3 (medium anti-AAV9 TAb titer) combined. ∗p < 0.05, ∗∗p < 0.01. t-SNE plots include concatenated data from all AAV9-treated animals together and show the abundance of CD4 central memory (Tcm) and CD4 effector memory (Tem) cells within the memory compartment (E) and the distribution of AAV9-specific activated memory T cells within the whole CD4 memory compartment (F). AAV9, adeno-associated virus serotype 9; AIM, activation-induced marker; TAb, total antibody; t-SNE, t-distributed stochastic neighbor embedding.

Frequencies of AAV9-specific activated CD8+ T cells were lower compared with CD4+ T cell responses for all tested AIM pairs. A trend toward higher responses was observed in treated animals with high pre-existing anti-AAV9 TAbs for most of the AIM pairs. However, a substantial difference was only observed for the CD137+CD134+ combination (Figure 4). The very low frequencies of AAV9-specific activated CD8+ T cells prevented examination of specific responses from CD8+ memory T cell subsets.

Figure 4.

AAV9-specific CD8 T cell responses in the spleen

Representative flow cytometry plots depicting selected AIM pairs to identify AAV9-specific total CD8 responses (A). Box plot graph depicts data expressed as stimulation index (ratio of the signal from cells stimulated with AAV9 peptides over the signal from cells stimulated with DMSO only) for AIM pairs for total CD8 cells; dots show individual animals (B). Mixed-effects statistical analysis between group 4 (high anti-AAV9 TAb titer) versus groups 2 (low anti-AAV9 TAb titer) and 3 (medium anti-AAV9 TAb titer) combined. ∗p < 0.05. AAV9, adeno-associated virus serotype 9; AIM, activation-induced marker; TAb, total antibody.

High levels of pre-existing anti-AAV9 TAbs contributed to a systemic early transient interferon response without enhancing complement activation

To explore whether different levels of pre-existing anti-AAV9 TAbs may lead to a different circulating cytokine profile and complement activation following intrathecal AAV9 administration, we examined the concentration of 48 different cytokines and complement component C3a in the plasma of study animals longitudinally at pre-dose and on days 2, 8, 15, 22, and 30 post-vector administration. Of 48 measured analytes, only 17 cytokines were greater than the limit of quantification, including monocyte chemoattractant protein 1 (MCP-1), interferon (IFN)-induced protein of 10 kDa (IP-10), IFN-inducible T cell alpha chemoattractant (I-TAC), B lymphocyte chemoattractant 1 (BCA-1), interleukin-1 receptor A (IL-1RA), IL-2, granzyme A, IL-21, IL-18, vascular endothelial growth factor A (VEGF-A), transforming growth factor α (TGF-α), macrophage inflammatory protein 1β (MIP-1β), perforin, IL-16, fractalkine, IL-8, and RANTES.

An early, transient plasma IFN response was indicated by increased concentrations of IP-10 and I-TAC at day 2 post administration only in animals with high levels of pre-existing antibodies (group 4) (Figures 5A and 5B). These chemokines are known to be potent chemoattractants for T cells and may play a role in cellular responses, such as the recruitment of activated T cells. Animals from group 4 also had increased levels of IL-1RA and MCP-1 at day 2 compared with pretest values, although these increases were not statistically significant (Figures 5C and 5D). Changes in the other detected markers were observed in a few individual animals only, independently of the level of pre-existing anti-AAV9 TAbs (data not shown).

Figure 5.

Longitudinal assessment of plasma cytokines and complement factor C3a

Plasma concentrations of IP-10 (A), I-TAC (B), MCP-1 (C), IL-1RA (D), and C3a (E). Lines indicate individual data. Within each study group, comparisons over time were performed by using mixed-effects analysis. ∗p < 0.05, ∗∗p < 0.01. IFN, interferon; IL-1RA, interleukin-1 receptor antagonist; IP-10, IFN-induced protein of 10 kDa; I-TAC, interferon-inducible T cell alpha chemoattractant; LLOQ, lower limit of quantification; MCP-1, monocyte chemoattractant protein 1.

In addition, a trend toward transient complement activation as demonstrated by increased plasma C3a levels was observed in some AAV9-treated animals between day 8 and day 22 post-dosing, which was independent of the initial level of pre-existing anti-AAV9 TAbs (Figure 5E).

Clinical evaluations and pathology assessments over 4 weeks excluded major safety concerns

Clinical observations, body weights, qualitative food consumption, and clinical laboratory evaluations were performed for all groups. No test-article–related clinical observations, impact on body weight, or alterations in qualitative food consumption were noted. Other clinical observations included liquid/non-formed feces, skin sores and scabs, and vomitus/emesis (noted on day 1 shortly after dosing). These appeared infrequently, were transient, associated with an injury or dosing/anesthesia, or comparable with incidences as controls; therefore, they were considered not test article related.

Test-article–related clinical chemistry findings were limited to minimally to mildly increased alanine aminotransferase (ALT) and glutamate dehydrogenase (GLDH) activities for one to two animals in each treated group (groups 2, 3, and 4) without impact of pre-existing immunity (Figure S2). In group 2 (low anti-AAV9 TAb titer), increased ALT and GLDH activities (1.8- to 4-fold increase over baseline) were observed on day 30 in two of four animals. In groups 3 and 4, increased ALT (up to 1.5- to 2-fold increase over baseline) was observed on days 8 and/or 15 in two of eight animals. These findings were of small magnitude and likely reflect minor hepatocellular perturbation associated with hepatic transduction (Figure S2). No other changes in liver parameters or microscopic correlates were observed.

The only test-article–related hematology finding included transiently decreased absolute lymphocyte count (0.6-fold lower than baseline) in group 4 on day 2 (data not shown), which is likely reflective of lymphocyte trafficking and lacked microscopic correlates. No test-article–related findings were observed in coagulation, urinalysis, and urine chemistry parameters (data not shown).

Test-article–related microscopic findings were consistent with those commonly known for AAV9-derived vectors in cynomolgus monkeys. These histopathology findings were mostly minimal to slight in severity with high individual variabilities and included mononuclear cell infiltrate, neuronal degeneration, axonal degeneration, and gliosis in the brain, intrathecal injection site, spinal cord, cauda equina/spinal nerve roots, and/or DRG (Figure 6A). In the brain, there were trends toward increased severity of mononuclear cell infiltration in the meninges and/or perivascular region in the neuropil in animals with high pre-dose anti-AAV9 TAbs (group 4) compared with control animals or animals with low anti-AAV9 TAbs (Figures 6B and 6C). These changes were asymptomatic and not accompanied with any detrimental effects in the vascular or perivascular cells, including inflammatory, degenerative, or necrotic changes, and were considered non-adverse.

Figure 6.

scAAV9-CBA-mCherry vector-related pathology findings and CSF IL-12p40 concentrations

Representative photomicrograph of scAAV9-CBA-mCherry-related findings observed in animals across scAAV9-CBA-mCherry-administered groups without impact of pre-existing AAV9-Ab titers (A). Top: DRG (sacral): neural and axonal degeneration that accompanied mononuclear cell infiltration in the ganglia, nerve root, and perineurium/fascia. Middle: nerve root close to the level of the intrathecal injection site: axonal degeneration and mononuclear cell infiltration. Bottom: spinal cord (cervical): axonal degeneration in the lateral funiculus. Representative photomicrographs of mononuclear cell infiltration in the brain of study animals across groups (B). Top: minimal severity in a control animal. Middle: slight severity from low (group 2; left) and medium (group 3; right) titer levels. Bottom: moderate severity in a high-titer animal (group 4). Summary of the incidence and severity of scAAV9-CBA-mCherry-related microscopic findings in the brain (C). Longitudinal assessment of CSF IL-12p40 concentrations (pg/mL) (D). Graph depicts mean values ± SEM; ∗p < 0.05 mixed-effects analysis. AAV9-Ab, adeno-associated virus serotype 9 antibody;; CSF, cerebrospinal fluid; DRG, dorsal root ganglia; IL, interleukin; SEM, standard error of the mean.

In line with the increased incidence of mononuclear cell infiltration in the brain of AAV9-treated animals, increased IL-12p40 levels were observed in the CSF of AAV9-treated animals compared with controls at day 30 post-dose, confirming test-article–induced local changes in inflammatory status. A trend toward higher IL-12p40 levels was observed in a subset of animals with medium and high anti-AAV9 TAb titers, but the results were not conclusive because of high inter-animal variability (Figure 6D).

Discussion

Pre-existing anti-AAV immunity resulting from natural exposure to wild-type AAV in a large percentage of the human population^11,31 poses a challenge to gene therapy development because it can impact both efficacy and safety. Although strategies aimed at overcoming pre-existing AAV antibodies are being explored (e.g., use of antibodies-cleaving endopeptidase,³² immunoadsorption;³³ or combination strategies), they are currently not approved for use. Immunodepletion using therapeutic plasma exchange and rituximab reduced anti-AAV9 TAb levels to within the eligibility range in young twins with SMA.³⁴ However, both individuals experienced acute hypersensitivity reactions of unknown origin during systemic gene therapy administration, suggesting that such combinatory regimens may not be enough to completely blunt all aspects of pre-existing immunity and highlighting the need for further investigation into these approaches.³⁴ Therefore, seropositive patients are often excluded from gene therapy treatment and clinical trials at present. For programs using intravenous dosing, even low levels of pre-existing neutralizing antibody titers diminished vector transduction and negatively impacted efficacy in preclinical models.^15,35

However, the influence of pre-existing serum antibodies on the efficacy and safety of gene therapies delivered locally is less characterized. Our results demonstrated that even high serum levels of AAV9 capsid-specific antibodies did not impact vg biodistribution to the CNS (e.g., spinal cord) following intrathecal delivery, although it did influence vector biodistribution to peripheral tissues, including reduced vg copies in the DRG, liver, and heart and increased vg copies in the spleen. Thus, local administration of gene therapy has the advantage of providing transduction of the CNS tissue while avoiding the inhibitory effect on biodistribution by pre-existing serum anti-AAV TAbs. These data also suggest that using serum anti-AAV TAb titers as a criterion for patient eligibility for treatments delivered intrathecally could potentially lead to the unnecessary exclusion of patients who may benefit from these treatments.

The present study is consistent with previous preclinical studies evaluating the impact of pre-existing AAV-specific antibodies. In rhesus macaques, naturally occurring pre-existing AAV8 NAb titers of ≤1:160 had no effect on expression of a secreted transgene after intramuscular delivery of the vector,³⁶ suggesting that circulating antibodies may not be a limitation when local delivery is planned. In addition, previous studies reported no noticeable difference in AAV9-mediated gene transfer in the CNS tissues of seronegative and seropositive animals (maximum TAb titer 1:400).²⁹

In addition to having no impact on CNS transduction, our data suggest that high titers of circulating pre-existing antibodies may even be beneficial in de-targeting the liver and other peripheral organs in line with data reported by Wang and colleagues.³⁵ However, pre-existing antibodies direct the vector to lymphoid organs, such as the spleen, as previously demonstrated with other AAV serotypes (e.g., AAV8).³⁵ The increase in splenic vector load in primates with high pre-existing antibody titers could result from its unique microvasculature and fenestrated endothelium with high blood flow³⁷ in which antigen-antibody complexes are trapped by follicular dendritic cells within germinal centers (GCs), leading to B cell activation.^38,39,40 Studies with different AAVs revealed the presence of AAV capsid in the GCs of the spleen up to 28 days post-administration.⁴¹ Increased vg copies in the spleen could also be related to clearance of AAV immune complexes through the spleen.

In seropositive animals with high anti-AAV9 TAb titers, memory immune responses contributed to higher AAV9 capsid-specific antibody responses. A rapid and substantial increase in newly generated anti-AAV9 TAbs in both serum and CSF was observed as quickly as days 8 and 15, respectively. AAV9-specific antibody titers remained high at day 30, the last available time point. Higher AAV9-specific T cell responses were also observed in the spleens of group 4 animals and were composed primarily of memory CD4 T cells, which was in line with the contribution of memory responses and the higher biodistribution of vector in the spleen.

Altogether, our findings suggest that local CSF delivery can bypass pre-existing serum antibodies for therapeutic efficacy but cannot prevent vector-specific memory immune responses.

In vitro human data reported by Smith et al.⁴² and West et al.⁴³ suggested that pre-existing humoral immunity (NAb titers ≥1:100) to AAV vectors may also contribute to increased pro-inflammatory cytokine/chemokine secretion and complement activation. Increased plasma concentrations of the IFN-related cytokines, IP-10 and I-TAC, as well as MCP-1 and IL-RA, were detected at day 2 post vector administration only in animals with high levels of pre-existing antibodies (group 4). Those changes were transient and resolved by day 8 but represent in vivo evidence of different AAV-driven innate responses in the presence of high levels of pre-existing antibodies. However, high levels of pre-existing antibodies did not enhance complement activation in our study (i.e., C3a levels).

Despite higher vector-specific immune responses in animals from group 4, no substantial safety concerns were observed at the vector dose used. A mild increase in ALT and GLDH activities, transiently decreased lymphocyte counts, and vector-related microscopic findings (brain, spinal cord, cauda equina, DRG, and heart) were observed for some treated animals, indicating minor liver and hematology perturbations that were independent of the pre-existing AAV9 antibody titers. In addition, a trend toward increased mononuclear cell infiltration in the meningeal and perivascular regions of the brain was more pronounced for animals with high pre-existing AAV9 titers, with a subset of animals in groups 3 and 4 having higher levels of IL-12p40 in the CSF. Immune cell infiltrates are most likely responsible for the increased IL-12p40 levels in the CSF, which was previously observed in both NHP²⁹ and mouse⁴⁴ studies. However, these changes were considered non-adverse.

In a clinical trial for giant axonal neuropathy (GAN), patients with or without positive AAV9 NAb titers at baseline were included and treated with scAAV9/JeT-GAN therapy via intrathecal administration. Seropositive patients had higher serum anti-AAV9 NAb titers at 3 weeks compared with those who were seronegative, indicating that seropositive individuals may be prone to stronger humoral responses after CSF-AAV administration,⁴⁵ which is consistent with our findings in NHPs. In the GAN trial, vector clearance from blood occurred faster in patients who were originally seropositive,⁴⁵ highlighting the need for studies of longer duration.

The present data may be specific to intrathecal administration, AAV9 serotype, the selected dose, and the NHP model, but they are in line with previous findings, including those using animals of a different sex, different AAV serotypes (AAV2.5²⁷ and AAV8³⁵), vector doses, and administration routes. This consistency supports the notion that excluding patients with pre-existing antibodies may not be necessary when targeted intrathecal delivery is used, especially for targeting spinal motor neurons; however, supportive clinical data are needed. The observed stronger immune responses in the presence of high levels of pre-existing AAV9 capsid-specific antibodies suggest a need to closely monitor for treatment-emergent immune responses, including to the transgene, and potential consequences in AAV seropositive patients as patient intrinsic factors may affect outcomes.

Materials and methods

Animal specifications

This non-GLP study included 15 female cynomolgus macaques (Mauritius). In total, 12 NHPs were dosed with scAAV9-CBA-mCherry vector, and three animals were dosed with vehicle material (tangential flow filtration 3 [TFF3] buffer) and used as concurrent controls. At the initiation of dosing, animals were approximately 41 months of age and weighed 2.8 to 5.4 kg. All procedures were performed in compliance with the Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, and the Office of Laboratory Animal Welfare and were approved by the Global Risk Review Team at Novartis Biomedical Research and by the animal welfare group at Labcorp (Madison, WI, USA). The Mauritius-origin primates were sourced from Bioculture Mauritius Ltd. (Immokalee, FL, USA). The in-life portion of this study was performed at Labcorp. Animals were group-housed in pens compliant with European guideline ETS 123 (up to three animals/pen), enriched with bedding material when possible. Animals were given various cage enrichment devices, including fruits, vegetables, or dietary enrichments. Prior to study assignment, animals were preselected based on their total anti-AAV9 TAb titers.

Evaluation of anti-AAV9 TAbs

In serum and CSF samples, anti-AAV9 TAb titers were detected using a bridging electrochemiluminescence immunoassay (ECLIA). All serum samples were tested in titration assays, in which samples were serially diluted using 2-fold dilutions with qualified blanks. After incubation with a “master mix” to allow anti-AAV9 TAbs to bind to biotinylated-AAV9 capsid and sulfo-tagged AAV9 capsid to form an antibody-bridge complex, samples were added to an MSD streptavidin-coated plate in which the biotinylated-AAV9 capsid binds to the streptavidin. Following removal of any unbound material by washing after an incubation step, an MSD read buffer containing tripropylamine was added, and the sulfo-tagged associated with the AAV9 capsid produced a chemiluminescent signal when electrical voltage was applied. The signal produced was proportional to the amount of anti-AAV9 TAbs present in the sample.

Serum titers were determined as the reciprocal of the value, which was calculated by multiplying the final minimal required dilution (MRD) of the assay (1:20), and the highest serial dilution yielding a response greater than or equal to the plate-specific cut point. Negative titers (<20) were defined as undiluted sample falling below the plate-specific cut point at final MRD of 1:20.

CSF samples were tested using a two-tiered approach composed of an initial confirmation assay and a titration assay. Only CSF samples that tested positive in the confirmation assay were further analyzed in the titration assay. The confirmation assay followed the same format as the titration assay except that samples were incubated with unlabeled, empty AAV9 capsid in a competitive binding format to demonstrate the specificity of the binding interactions in the antibody/labeled AAV9 complexes. The presence of anti-AAV9 TAbs was determined by comparing the signal in the sample with a statistically derived threshold, the assay cut point. CSF samples for which no immunodepletion was observed were reported as negative, which can be interpreted as the absence of anti-AAV9 TAbs. Samples confirmed as positive but below the titer threshold were reported as <20 negative titer.

AAV vector and vehicle control

A scAAV9-CBA-mCherry vector was used as a pre-formulated solution at a nominal level of 1 × 10¹³ vg/mL (actual level of 9.8 × 10¹² vg/mL) and administered as a single dose of 1.2 × 10¹³ vg/animal to three groups (n = 4 animals/group) of female cynomolgus monkeys via intrathecal injection. The dose volume was 1.2 mL/animal. An additional group of cynomolgus monkeys (n = 3 females) was administered TFF3 formulation buffer at an equivalent dose volume and served as controls. Dose formulations were prepared by diluting the vector in TFF3 buffer to target levels and testing per quantification by droplet digital polymerase chain reaction (ddPCR). The TFF3 formulation buffer consisted of tromethamine (20 mM), magnesium chloride (1 mM), sodium chloride (200 mM), and poloxamer 188 (0.005%) that was adjusted to pH 8.1 (±0.1) with hydrochloric acid (6 M), quantum satis to volume with sterile water for injection. The genetic payload of the vector consisted of a self-complementary AAV9 genome. The reporter transgene was the fluorescent reporter, mCherry, which was expressed under the cytomegalovirus (CMV) early enhancer and a hybrid CMV enhancer/chicken β actin (CBA) promoter. The vector used in this study was produced in a mammalian system using preclinical research processes comparable with vector lots produced under good manufacturing practice conditions. The vector met acceptance and release criteria for in vivo research use for the percentage of empty capsid (<10%), plasmid DNA, production cells, endotoxin concentrations (≤0.05 EU/mL), bioburden (<1 colony-forming units [CFU]/mL), and purity (>95%). Vector titer was determined by ddPCR after manufacturing and confirmed after reformulation (when deemed necessary) before dosing.

Dosing procedure and intrathecal administration

On day 1, animals were anesthetized with 10.0 mg/kg of ketamine 10 to 15 min prior to dosing, followed by 0.02 mg/kg dexmedetomidine. Animals were also administered buprenorphine after ketamine and dexmedetomidine before the dosing procedure.

Doses were infused via intrathecal injection into the intervertebral space of L5 to L6 (lumbar puncture [LP], slow bolus over at least 1 min). Once the needle was inserted, up to 1 mL of CSF was collected prior to dosing. Following dosing, the needle was flushed with approximately 0.25 mL of CSF to clear the needle. Animals were maintained in dorsal recumbence with the hindlimbs elevated (Trendelenburg-like position) for at least 10 min following the completion of dosing. Atipamezole (0.2 mg/kg) was administered intramuscularly as a reversal agent for the anesthesia shortly after the dosing procedure. Animals did not receive immunosuppression.

Clinical assessments (in-life procedure)

Clinical observations included, but were not limited to, twice-daily general observations (morbidity/mortality health checks), once-daily cage-side observations for behavioral changes or apparent abnormalities (posture, activity level, coordination, and any discharge), weekly detailed observations for signs of toxicity by removing the animals from their cages, and weekly body weights and qualitative daily food consumption determinations for all groups prior to dosing and during the observation period.

Blood collection and handling (in-life procedure)

Blood collections were performed during the pre-dosing phase and/or on day 1, prior to administration of AAV9 vector or vehicle control, at specific intervals during the in-life observation phase and on the day of scheduled euthanasia. Blood samples for standard clinical laboratory procedures (hematology, coagulation, and clinical chemistry) and anti-AAV9 TAbs were collected from the femoral vein in appropriate collection tubes. Samples were centrifuged within 60 min of collection for serum and/or plasma. Following harvesting, samples were stored at −60°C to −80°C until used.

Clinical laboratory procedures

Blood for hematology, coagulation, and clinical chemistry tests was collected twice during the pre-dose phase and on days 2 (hematology only), 8 (all except coagulation), 15, and 30 of the post-dosing phase. The following clinical pathology parameters were analyzed using standard laboratory procedures and instrumentation: (1) hematology: red blood cell count, hemoglobin, hematocrit, mean corpuscular volume, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, platelet count, white blood cell count, reticulocyte count, and blood smear/cell morphology; (2) coagulation: prothrombin time, fibrinogen, and activated partial thromboplastin time; and (3) clinical chemistry: glucose, urea nitrogen, creatinine, total protein, albumin, globulin, albumin:globulin ratio, cholesterol, total bilirubin, ALT, GLDH, alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), aspartate aminotransferase (AST), calcium, inorganic phosphorus, sodium, potassium, chloride, triglycerides, magnesium, creatine kinase, and C-reactive protein. Urine for urinalysis and urine chemistry (urine creatinine, urine total protein, and urine microalbumin concentrations and urine protein:urine creatinine, and urine microalbumin:urine creatinine ratios) tests were collected twice during the pre-dose phase and on day 29 of the post-dosing phase.

Terminal procedures

At necropsy, animals were anesthetized with sodium pentobarbital and exsanguinated via whole body perfusion with chilled nuclease-free phosphate buffered saline. Tissue samples were collected and either snap-frozen in liquid nitrogen for molecular biodistribution analyses or processed to paraffin-embedded blocks for pathology assessments (histopathology). In addition, spleen biopsies were collected for preparation of splenocytes on the day of the necropsy.

Light microscopy evaluation (histopathology)

Paraffin-embedded tissues (brain, spinal cord [cervical, thoracic, and lumbar], intrathecal injection site, cauda equina/nerve root, DRG [cervical, thoracic, lumbar, and sacral], liver, and heart) were routinely processed, sectioned, and stained with hematoxylin and eosin (H&E). Macroscopic and microscopic examinations were conducted by American College of Veterinary Pathologists (ACVP)-board-certified veterinary pathologists experienced in toxicologic pathology.

Vg biodistribution

Tissue samples were collected using instruments cleaned with sterile saline and/or sterile instruments to avoid contamination. Samples for ddPCR were collected, flash frozen in liquid nitrogen, and stored between −60°C and −80°C until they were packed on dry ice and shipped for analysis. DNA from tissue samples of each study animal was extracted using Qiagen DNeasy Blood and Tissue Kit (Cat. 69504, Qiagen; Germantown, MD, USA). The extracted DNA was analyzed for vg content using duplex ddPCR quantitation with a C1000 Thermal Cycler and a QX200 droplet reader (Bio-Rad; Hercules, CA, USA). mCherry-specific primers and probes were used for ddPCR analysis of the scAAV9-CBA-mCherry vector. In parallel, a two-copy reference gene or diploid gene (dg), CFTR, was also quantified for normalizing purposes. CFTR primers and probe were added to the master mix along with specific vg primers and probe, and both were quantified in the same reaction using multiplex ddPCR. The resulting values are presented as vg/dg, thus normalizing vg to the CFTR reference gene. The lower limit of quantitation (LLOQ) determined for mCherry was 21 copies/reaction. The following primers were used for ddPCR: CFTR forward 5′- GAA TTC ATG CGG TTC AAG GTG -3', reverse 5′- TCT GTG TTC CCT CGT AAG GT -3', probe 5′- AAC GGC CAC GAG TTC GAG ATT GAA -3' and mCherry forward 5′- AGA AAA GAG ACT GTC CCT AG -3', reverse 5′- GTG AAC GTC ATC AGA TCC AAA -3', and probe 5′- CCC ACA TTT CCA GGC AGA AG -3'.

AIM T cell assay

A multiplexed AIM T cell assay was established for detecting AAV9-specific T cell responses. Briefly, splenocytes were thawed and rested for 3 h in 96-well round-bottom plates in X-Vivo medium supplemented with 1% Glutamax and 10 μg/mL DNAse. Cells were then stimulated with a peptide pool from AAV9 VP1 capsid protein (pool of 68 peptides of 15-mer length, purchased by peptides & elephants, Hennigsdorf, Germany) dissolved in DMSO. A DMSO-treated condition and a Staphylococcus enterotoxin B (SEB)-treated (10 ng/mL) condition served as negative and positive controls, respectively. Samples were tested in duplicates, given enough cells. After 20-h incubation (37°C, 5% CO₂), cells were washed and stained with a viability dye, live/dead-eFluor780 (BioLegend; San Diego, CA, USA), and then a cocktail of antibodies (30 min, 4°C) as follows: CD137- Super Bright 436 (Thermo Fisher; Waltham, MA, USA); CD8-BV510, CD69-BV605, CD197-PE-Cy7, CD20-APC Fire750, and CD16-APC Fire750 (all from BioLegend); and CD3-BUV395, CD25-BUV737, CD4-BV711, CD45RA-BB700, and CD134-RY586 (all from BD Biosciences; Franklin Lakes, NJ, USA). Our assay integrates and co-analyzes four AIMs: CD69, CD137 (4-1BB), CD134 (OX40), and CD25. In the absence of a consensus combination of activation markers across tissues and species, we opted for the inclusion of several AIM pairs, enhancing the likelihood for detecting AAV9-specific T cells.⁴⁶ Samples were acquired on a flow cytometer FACSFortessa Cell Analyzer (BD Biosciences). Analyses were performed using OMIQ software, Microsoft Excel, and Prism v10 (GraphPad Software).

Plasma cytokine and C3a analysis

Plasma samples were received and stored at –60°C to −80°C until measurement. Plasma samples were analyzed for the presence of cytokines and chemokines using a 48-plex MILLIPLEX Non-Human Primate cytokine/chemokine/growth factor premixed kit (MilliporeSigma; Burlington, MA, USA; Cat. # PRCYTA-40K) following the manufacturer’s instructions. Analytes included BCA1, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), granzyme A, granzyme B, IFN-γ, IFN-α2, IL-2, IL-4, IL-5, IL-6, IL-7, IL-1α, IL-1β, IL-1RA, IL-10, IL-12p70, IL-15, IL-16, IL-17A, IL-17E, IL-18, IL-21, IL-22, IL-23, IL-28A, IL-31, IL-33, I-TAC, IP-10, MCP-1, monokine induced by IFN-γ (MIG), MIP-1α, MIP-1β, MIP-3a, perforin, sCD137, sCD40L, soluble Fas ligand (sFASL), tumor necrosis factor alpha (TNF-α), TNF-β, TGF-α, VEGF-A, eotaxin, fibroblast growth factor 2 (FGF-2), fractalkine (all tested at a 4-fold dilution), and IL-8 and RANTES (tested at a 200-fold dilution).

Results were fit to a curve and extrapolated in MyAssays Desktop Pro software v.9.2 (MyAssays, Brighton, United Kingdom). C3a measurements in plasma samples (tested at a 200-fold dilution) were performed using an NHP complement C3a ELISA kit (Quidel Ortho, San Diego, CA, USA; Cat. #A031) following the manufacturer’s instructions. Results were fit to a log-log line and extrapolated in Softmax Pro software v.5.4.6 (Molecular Devices, San Jose, CA, USA).

IL-12p40 protein concentrations in the CSF

CSF samples were collected by lumbar puncture and analyzed for the presence of IL-12p40 using the IL-12 ELISA Kit (Meso Scale Discovery LLC; Rockville, MD, USA) following the manufacturer’s instructions.

Statistical analysis

Data are expressed as median with range or mean ± standard error of measurement (SEM) and analyzed using either Prism 10.0 or SAS 9.4. The low number of animals used for this study did not lend itself to high-power statistical interpretation. Nonetheless, statistical analyses were run to examine differences between the different study groups. A mixed-model analysis was conducted using a log2 transformation of the serum titer data to examine the influence of group, time, and the interaction between the two variables. For biodistribution and T cell data, a mixed-model analysis was conducted to investigate differences between a combination of groups 2 (low anti-AAV9 TAb titer) and 3 (medium anti-AAV9 TAb titer) compared with group 4 (high anti-AAV9 TAb titer). For cytokine and complement data, a mixed-model analysis was conducted to examine the influence of group and time using Prism 10. For all analyses, statistical significance was set as p value < 0.05.

Data and code availability

All data supporting the findings of this study are available within the paper and its supplemental information.

Acknowledgments

The authors acknowledge Cameron McElroy, Elena Hallaren, Laura Beasley-Topliffe, David Kagan, Terrence Oday, and Christie Watters of Novartis Biomedical Research for their support with the execution of this study, data generation, and analysis. Editorial and medical writing support was provided by Marjet Heitzer, PhD, of Kay Square Scientific, Butler, PA, USA. This support was funded by Novartis Pharma AG, Basel, Switzerland.

Author contributions

M.V., F.F.T., and G.H. wrote the manuscript. M.V. and N.R.L. contributed to T cell data generation and analysis. T.d.R., E.M., and F.O. generated the biodistribution data. P.J. performed statistical analysis. M.P.S. analyzed and interpreted clinical pathology data. R.K. generated and interpreted histopathology findings. K.M. generated and interpreted data. D.B., E.H., T.R.-S., F.F.T., F.McB., and G.H. designed the study or were fundamental for study execution. All authors contributed to the drafting of the manuscript and reviewed and approved the final version for submission.

Declaration of interests

All authors are current (M.V., T.d.R., N.R.L., P.J., C.W., R.K., K.M., F.McB., M.P.S., D.B., E.H., T.R.-S., F.F.T., F.O., and G.H.) or former E.M. employees of Novartis Pharma AG. All authors disclose salaries and compensations received for the execution of this study funded by Novartis Biomedical Research.