Modified AAV5 capsid for improved brain biodistribution following direct injection in preclinical models

Abstract

Gene therapies based on adeno-associated virus vectors hold strong potential for the treatment of central nervous system disorders. However, systemic delivery is limited by the blood-brain barrier, off-target effects, immune responses, and vector loss. Direct intraparenchymal injections can bypass these barriers by targeting specific brain regions, but broader vector distribution is essential to reduce the need for multiple injections and to achieve widespread transgene expression. The brain biodistribution of adeno-associated virus serotype 5 is partially mediated by its interaction with sialic acid. Here, we describe a modified serotype 5 variant, AAV5neo, carrying a single amino acid substitution that alters its sialic acid-binding properties. In cultured cells, AAV5neo exhibits transduction that is independent of sialic acid and is not inhibited by N-acetylneuraminic acid, a form of sialic acid highly abundant in the brain. Following direct striatal injection in mice, minipigs, and non-human primates, AAV5neo consistently demonstrated enhanced transduction efficiency, broader distribution to cortical and deep brain regions, and achieved comparable transgene expression at approximately 10-fold lower doses relative to the parental serotype. These findings highlight AAV5neo as a potent and efficient vector candidate for localized gene therapy applications targeting the central nervous system.

Graphical abstract

Direct intraparenchymal AAV delivery can be enhanced by engineering the capsid’s glycan-binding profile. A single amino acid substitution of AAV5 reduces sialic acid dependency, rendering the modified capsid more potent and showing improved biodistribution in small and large animal models.

Introduction

Central nervous system (CNS) disorders are among the leading causes of global disability and contribute to millions of deaths annually.¹ Genetic abnormalities contribute significantly to the development of many CNS disorders.^2,3 The complexity of CNS disorders, along with physiological barriers such as the blood-brain barrier, limits the effectiveness of conventional drug therapies, leaving approximately 95% of patients without a cure.⁴ In contrast, novel gene therapy approaches offer the potential for transformative treatment.⁵

Recombinant adeno-associated virus (AAV) vectors are widely used for therapeutic gene delivery because they are non-pathogenic, demonstrate limited immunogenicity,⁵ and support long-term transgene expression following one-time administration.^4,6,7,8 For CNS disease indications, intraparenchymal AAV administration allows precise delivery at low doses, reducing the likelihood of immune responses and off-target effects compared to systemic administration.^9,10 Although the CNS is considered relatively immune privileged,¹¹ it has the capacity to elicit both adaptive and innate immune responses,¹² which may limit the efficacy of AAV-based therapies. These limitations highlight the continued need to engineer next-generation AAV capsids with improved properties, even for direct intraparenchymal delivery. Recent advances employing directed evolution and peptide insertion strategies have produced engineered AAV1 and AAV2 variants that exhibit enhanced potency and broader distribution within the brain following injection into the globus pallidus.¹³ These innovations underscore the potential of next-generation engineered capsids to outperform their parental natural serotypes, offering improved outcomes for CNS-targeted gene therapies.

AAV5 has previously been used in clinical and preclinical trials for CNS diseases.^4,5,14 Importantly, antibodies neutralizing AAV5 (1) have a low seroprevalence,¹⁵ (2) have low avidity,¹⁶ and (3) because AAV5 is the most phylogenetically distinct AAV, exhibit limited cross-reactive immunogenicity.^17,18,19 Furthermore, preclinical studies have demonstrated that, unlike some AAV serotypes, the AAV5 vector is an excellent candidate for intrastriatal injection,²⁰ exhibiting widespread dissemination via both retrograde and anterograde axonal transport.^21,22,23,24 Taken together, these features position AAV5 as a highly promising vector for CNS-directed gene therapy applications.^25,26

Given its favorable characteristics, a compelling question is whether the performance of AAV5 can be further improved through rational capsid engineering. The distinct tissue tropism and AAV-host interaction of several serotypes are, in part, determined by capsid interactions with cell surface glycans.^27,28 Manipulating AAV-glycan interactions can improve transduction efficiency and vector biodistribution in the CNS, which has been achieved using strategies including AAV2 receptor blockade in the neuronal membrane by heparin co-injection²⁹ and mutagenesis of AAV2 glycan-interacting residues.^30,31

AAV5 requires a specific glycan, N-linked (α2-3) sialic acid (SIA),^32,33,34,35 a 9-carbon sugar acid moiety for infectivity/transduction.³⁶ Post attachment, AAV5 engages AAVR, a glycosylated proteinaceous AAV receptor, for further cellular uptake/internalization.^32,37,38 Notably, two SIA-binding sites have been identified in the AAV5 capsid that alter viral transduction efficiency. Different mutations of the “A site”—located in a depression at the 3-fold symmetry axis and responsible for regulating SIA-dependent transduction—altered the AAV5 in vitro transduction efficiency.^32,34 These findings suggest that rational re-engineering of AAV5’s capsid-SIA interactions holds significant promise for enhancing CNS gene transfer.

Here, we describe the development of a next-generation AAV5-based capsid variant (AAV5neo, harboring the M569V A-site mutation) with altered SIA dependency in vitro. The objective of the preclinical studies was to assess the efficacy and biodistribution of AAV5neo in small and large animals following intraparenchymal injection. We found enhanced brain-wide vector biodistribution and improved transduction efficiency at lower doses than the parental AAV5 vector.

Results

AAV5neo is functional in vitro with cell type-dependent reduction of transduction efficiency

To determine whether the AAV5neo capsid had altered transduction capability compared to the parental AAV5, different cell lines (COS-1, HEK293T, SH-SY5Y, and Huh7 cells) were exposed to either AAV5 or AAV5neo at a multiplicity of infection (MOI) of 1 × 10⁵ and 1 × 10⁶ genome copies (gc)/cell (Figure 1A; Figure S1A). Transduction efficiency was quantified by measuring nanoLuciferase (nLuc) transgene activity 48 h post transduction. AAV5neo transduced all tested cell lines, albeit at lower levels than the parental capsid. Most striking differences in transduction efficiency were observed in the kidney cell lines COS-1 and HEK293T (Figure 1A), in which AAV5neo displayed 28-fold and 18-fold lower nLuc reporter gene activity, respectively. However, in Huh7 (Figure S1A) and SH-SY5Y cells, nLuc activity was reduced 5- to 6-fold when compared to AAV5.

Figure 1.

Transduction of cell lines and organoids with AAV5 and AAV5neo at a multiplicity of infection of 1 × 10⁴, 1 × 10⁵, or 1 × 10⁶ gc/cell expressing nLuc or eGFP as reporter transgenes in the presence/absence of neuraminidase A and N-acetylneuraminic acid

(A) Transduction assay in COS-1, HEK293T, and SH-SY5Y cell lines. N = 3 repeats per experimental condition, shown as mean nLuc activity. nLuc activity was measured 48 h post transduction. (B) Transduction of HEK293T cells using eGFP as a reporter transgene. Photomicrographs were taken using bright-field illumination; fluorescence microscopy was used to visualize eGFP expression. Scale bars: 0.3 mm (C) Flow cytometry analysis of dissociated organoids transduced with AAV5 or AAV5neo with eGFP-positive cells quantified over 3 weeks. N = 2 repeats were performed per experimental condition. (D) Transduction of HEK293T cells either untreated or treated with 10 units per well of neuraminidase A for 2 h prior to AAV addition. Transduction efficiency was measured via nLuc activity 48 h post transduction. N = 3 repeats per experimental condition; values are shown as percentage nLuc activity normalized to untreated cells. (E) Transduction of Huh7 cells treated with an anti-AAVR or an isotype control antibody (5 mg/mL) for 1 h prior to transduction with AAV. Transduction efficiency was measured via nLuc activity 48 h post transduction. N = 3 repeats per experimental condition; values are shown as percentage nLuc activity normalized to isotype-treated cells. (F) COS-1 cells were transduced with AAV pre-incubated with 0–25 mM Neu5Ac. Transduction efficiency was measured via nLuc activity 48 h post transduction. N = 3 repeats per experimental condition; values are shown as percentage nLuc activity normalized to untreated (0 mM Neu5Ac) cells. eGFP, enhanced green fluorescent protein; FITC, fluorescein isothiocyanate; gc, genome copies; MOI, multiplicity of infection; Neu5Ac, N-acetylneuraminic acid; nLuc, NanoLuciferase; AAV, recombinant adeno-associated virus; RLU, relative light units.

To confirm the phenotype observed in vitro, HEK293T cells were transduced with either AAV5 or AAV5neo using enhanced green fluorescent protein (eGFP) as a reporter transgene. Cells were analyzed 48 h post transduction for transgene expression via fluorescence microscopy (Figure 1B) and flow cytometry (Figure S1B). Depending on MOI, AAV5neo showed a 2- to 6-fold reduction in the number of eGFP-positive cells, which was also accompanied by a reduction in mean fluorescence intensity, confirming initial nLuc findings.

To determine whether the reduction in transduction efficiency resulted from the short (48 h) post-transduction period, brain organoids were transduced (MOI of 5 × 10⁵ gc/cell for 3 days), maintained in culture for 3 weeks, and evaluated via flow cytometry (Figure 1C). Over the observed time period, an increase in eGFP-positive cells was observed for both AAV5 and AAV5neo.

However, AAV5neo transduction resulted in lower absolute numbers of eGFP-positive cells than the parental vector. Overall, AAV5neo retains the capability to transduce cells in vitro but with lower efficiency compared to AAV5 in all tested conditions and cell types.

AAV5neo transduction in vitro is SIA independent but requires AAVR interaction

Based on the reduced transduction efficiency in vitro together with the AAV5 mutation studies conducted by Afione et al.,³⁴ we evaluated SIA dependency of AAV5neo. Specifically, we determined the effect of enzymatic cleavage of linear and branched non-reducing terminal SIA residues from glycoproteins and glycopeptides on the transduction efficiency of AAV5 or AAV5neo. HEK293T cells were incubated with sialidase (neuraminidase A), AAV was added, and 48 h post transduction, nLuc activity was measured (Figure 1D). Untreated cells were transduced as a reference for total transduction efficiency. When treated with sialidase, cells transduced with AAV5 showed a ∼70% reduction in nLuc reporter activity compared with the untreated condition. In contrast, the sialidase treatment did not reduce the reporter gene activity of AAV5neo. Additional experiments were conducted to determine whether AAV pan-receptor AAVR (KIAA0319L) interaction^37,38 was altered by the AAV5neo mutation. HEK293T cells were incubated with either a mouse-derived polyclonal anti-AAVR antibody or an isotype control antibody, and reporter gene activity (nLuc) was measured 48 h post transduction (Figure 1E). An ∼80% reduction of transduction efficiency was observed for both capsids when AAVR was blocked with an anti-AAVR antibody, demonstrating that AAV5neo and the parental AAV5 require AAVR interaction for efficient transduction. This finding was corroborated by transient transfection-mediated overexpression of AAVR in HEK293T and HeLa cells (Figure S1C), in which overexpression enhanced AAV5neo transduction (HEK293T ∼10-fold and HeLa ∼50-fold).

N-acetylneuraminic acid inhibits AAV5 but not AAV5neo

Having demonstrated that AAV5neo transduction is SIA independent, we hypothesized that increasing concentrations of the SIA, N-acetylneuraminic acid (Neu5Ac), might compete with the cellular attachment of AAV5 but not AAV5neo. Both capsids were incubated with increasing concentrations of Neu5Ac prior to addition to cells, and nLuc activity was measured 48 h post transduction (Figure 1F). Neu5Ac inhibited AAV5 cellular transduction in a dose-dependent manner, with ∼30% reporter activity reduction at low concentrations (2.5 mM Neu5Ac) and up to 90% reduction at the highest (25 mM Neu5Ac) concentration. Conversely, AAV5neo transduction was not reduced at any tested concentration but increased relatively at concentrations of 10 and 25 mM Neu5Ac. To determine whether the Neu5Ac-mediated inhibition of AAV5 transduction was more pronounced at lower MOIs, a range of MOIs (1 × 10³–1 × 10⁶ gc/cell) were preincubated with 25 mM Neu5Ac prior to addition to cells (Figure S1D). For reference, transduction activity of untreated virus at the same MOI range was assessed. The inhibition of AAV5 transduction efficiency was more pronounced at lower MOI, reducing nLuc activity at MOI 1 × 10³ and 1 × 10⁴ by ∼80%. At higher MOIs of 1 × 10⁵ and 1 × 10⁶, reductions of 70% and 50%, respectively, were observed.

The SIA binding-deficient AAV5neo capsid enhances intraparenchymal transgene delivery to the murine brain

Previous studies demonstrated that modulating the SIA dependence of other AAV serotypes influenced CNS tropism in murine models.³⁹ Given that AAV5neo transduction was not inhibited by Neu5Ac, we speculated that AAV5neo might demonstrate an altered transduction phenotype in vivo.

C57BL/6J mice (n = 9 per group) were bilaterally injected into the striatum with AAV5-CBA-eGFP or AAV5neo-CBA-eGFP constructs at identical infusion rates and a total volume of 6 μL containing either 1.0 × 10⁹ (low dose), 1.0 × 10¹⁰ (mid-dose), or 1.0 × 10¹¹ (high dose) gc. Coordinates of the AAV deposit site and accuracy of the injection are illustrated in Figure S2A. Four weeks post injection, one brain hemisphere was collected for immunohistochemical detection of transgenic eGFP; the other hemisphere was used for molecular analysis by quantitative PCR (qPCR). Strikingly, fluorescent immunolabeling of sagittal brain sections of animals injected with AAV5neo showed a larger dose-dependent eGFP-positive area than AAV5 (Figures 2A and 2B; Figure S2B). At the low dose, transgene expression was limited to the hippocampus and striatum for both AAV5 and AAV5neo with an average of 1.4% and 2.9% eGFP-positive areas, respectively (Figure 2B; Figure S2C). At the mid- and high dose, AAV5neo treatment resulted in an average 2-fold (p < 0.0001) increase in eGFP-positive area when compared to AAV5 (9.4% versus 21.1% at mid-dose and 23.2% versus 43.2% at high dose). At the high dose, AAV5neo displayed significantly (p < 0.0001) larger eGFP-positive areas in the cortex, hippocampus, striatum, thalamus, and hypothalamus (Figure S2C).

Figure 2.

Mice received AAV5 or AAV5neo via intrastriatal injections

Enhanced green fluorescent protein (eGFP) transgene expression in the brain was detected and quantified by immunofluorescent staining. Vector DNA and transgene mRNA were measured by real-time quantitative PCR quantification. (A) Immunofluorescent detection of eGFP transgene expression in sagittal sections from mice treated with AAV5 (top row) and AAV5neo (bottom row) at a dose of 1 × 10¹¹ gc/brain. Images are representative for n = 6 injected animals. eGFP-positive staining is seen as green signal. Scale indicated per zoom-in tile. (B) Average eGFP-positive area quantification of all analyzed regions (for individual region of interest, see Figure S2C) for low-, intermediate-, and high-dose administration of either AAV5 or AAV5neo. Values shown as mean percentage eGFP-positive area with data points of individual animals (n = 6). Significance levels of statistical tests are indicated (one-way ANOVA). (C) Quantification of vector genome copies in 5 regions of interest by qPCR. Values are the mean of vector copies/μg DNA with data points of individual animals (n = 9 for treatment groups, n = 5 for vehicle). (D) Quantification of eGFP mRNA copies in 5 regions of interest by RT-qPCR. Values are the mean of eGFP mRNA copies/μg RNA with data points of individual animals (n = 9 for treatment groups, n = 5 for vehicle). Significance levels of statistical tests are indicated (unpaired t test). eGFP, enhanced green fluorescent protein; GC, genome copies; AAV, recombinant adeno-associated virus; qPCR, quantitative polymerase chain reaction; ROI, region of interest.

The remaining hemisphere was used for the quantification of vector genomes and eGFP mRNA levels in the rostral and caudal cortex, striatum, thalamus, and cerebellum. Vector DNA levels in AAV5 and AAV5neo-injected mice were not significantly different in any brain region analyzed (Figure 2C). However, the quantification of eGFP mRNA demonstrated a significant (p < 0.01) increase in eGFP mRNA in the rostral cortex of AAV5neo-treated animals in the mid- and high-dose groups (Figure 2D). The other brain areas investigated showed a small relative increase in eGFP mRNA levels, but differences were not statistically significant.

The AAV5neo capsid enhances intraparenchymal transgene delivery to the minipig brain

Libechov minipigs (n = 3 per group) were bilaterally injected into the putamen as described by Evers et al. with 1 × 10¹² gc per hemisphere of AAV5-CBA-eGFP or AAV5neo-CBA-eGFP, and transgene expression was localized (4 weeks post administration) in brain sections using cryo-immunohistochemistry. Although identical doses were given to both cohorts, AAV5neo-injected animals showed more intense immunohistochemical staining (Figure 3A), suggesting higher levels of eGFP expression. The eGFP-positive area in AAV5neo-treated animals was, on average, ∼3-fold larger than AAV5 (Figure 3B). To determine which cell types were transduced, tissue punches from the caudate and white matter of the frontal lobe from AAV5neo-injected animals were used for single-nuclei RNA (snRNA) sequencing (Figure 3C). Overall, eGFP was predominantly expressed in oligodendrocytes and neurons. In tissue with higher neuron density (caudate), 47.5% of eGFP-positive cells were neurons, 25% oligodendrocytes, 7% astrocytes, and the remaining 21% were classified as other cell types. In frontal lobe white matter, the majority (83%) of eGFP-expressing cells were oligodendrocytes, with 10% unclassified cells and the remaining fraction of cells equally divided between neurons (3.6%) and astrocytes (3.6%).

Figure 3.

Minipigs were intraparenchymally injected with AAV5 or AAV5neo using identical surgical procedures for both groups

Enhanced green fluorescent protein (eGFP) transgene expression in coronal minipig brain sections was detected and quantified by cryo-immunohistochemistry. Different cell types of eGFP-expressing cells were quantified by single-nuclei RNA sequencing. Vector DNA and eGFP mRNA estimated by real-time quantitative PCR quantification in tissue punches across AAV5 and AAV5neo-injected minipig brains. (A) Immunohistochemical detection of eGFP in frozen brain sections of either AAV5 (top row) or AAV5neo (bottom row)-injected animals at a dose of 2 × 10¹² gc/brain. Scale indicated per tile. (B) Quantification of eGFP signal of whole coronal section area shown as DAB staining-positive area. Values shown as mean percentage eGFP-positive area with data points of individual animals (n = 3). (C) Cell type distribution of eGFP-positive cells in brain punches, taken from frontal lobe white matter or caudate (pooled from n = 3 animals) of AAV5neo-injected animals, subjected to single-nuclei RNA sequencing. Values indicate percentage of designated cell types within eGFP-positive cell population. (D) Quantification of vector genome copies in 22 regions of interest by qPCR. Values are the mean of vector copies/μg DNA with data points of individual animals (n = 3). Significance levels of statistical tests are indicated (two-way ANOVA). (E) Quantification of eGFP mRNA copies in 22 regions of interest by RT-qPCR. Values are the mean of eGFP mRNA copies/μg RNA with data points of individual animals (n = 3). Significance levels of statistical tests are indicated (two-way ANOVA). DAB, diamino benzidine; eGFP, enhanced green fluorescent protein; IBA1; LLOQ, lower limit of quantification; PFC, prefrontal cortex; AAV, recombinant adeno-associated virus; ROI, region of interest; VPL, ventral posterolateral nucleus.

Tissue punches (n = 22) were taken throughout the brain according to punching scheme shown in Figure S3, and the copy number of vector genomes was determined per μg of total DNA by qPCR (Figure 3D). Vector copy numbers in AAV5neo-injected brains were consistently higher when compared with AAV5. In punches located proximally (7, caudate; 8, putamen; and 11, putamen) to the injection sites, vector copies were, on average, 27-fold (p = 0.01), 75-fold (p < 0.0001), and 35-fold (p < 0.0001) higher, respectively. In samples distal to the injection site, a less-pronounced difference of ∼6-fold higher vector copies was observed. AAV5neo-injected brains contained, on average, 11-fold more vector genomes than AAV5-injected brains when all brain areas sampled were combined. No vector genomes were detected in liver samples taken from AAV5 or AAV5neo-treated animals (data not shown).

Quantification of transgene expression using quantitative reverse-transcription PCR (RT-qPCR) (Figure 3E) showed higher eGFP expression in punches proximal to the injection sites, for example, 19-fold (ns), 62-fold (p = 0.037), and 150-fold (p = 0.0005) higher in punches 7, 8, and 11. Overall, eGFP transcripts were 12-fold more abundant in AAV5neo-injected punches when compared with AAV5, although brain regions more distal to the injection site had lower eGFP expression.

Intraparenchymal delivery of a therapeutic transgene via the AAV5neo capsid in non-human primates showed improved potency

Using a therapeutically designed transgene encoding a CBA-microRNA (miRNA) targeting the alpha-synuclein gene (SNCA), non-human primates (NHPs; Macaca fascicularis) (n = 3 per group) were bilaterally injected into the putamen under magnetic resonance image (MRI) guidance (Figure S4A) with either 1.1 × 10¹² (low dose), 4.8 × 10¹² (mid-dose), or 2.3 × 10¹³ (high dose) gc of AAV5 or 2.0 × 10¹² (low-intermediate) gc of AAV5neo and sacrificed 8 weeks later. The low-intermediate dose of AAV5neo was 1.8-fold higher and 2.4- and 11.5-fold lower than the low, mid, and high doses of AAV5, respectively. In line with the minipig study, increases in vector DNA levels in different brain structures were observed when comparing AAV5neo with AAV5 (Figure 4A). The AAV5neo-treated putamen (punch 9b; Figure S4B) showed a ∼60-fold greater level of vector DNA when compared with low-dose AAV5 treatment. Overall, the levels of transduction achieved with a low-intermediate dose of AAV5neo were similar to the transduction observed with the mid and high doses of AAV5. Furthermore, in the thalamus (punch 26b; Figure S4B), vector DNA levels of the AAV5neo group were ∼37-fold higher compared to AAV5 low-dose group and comparable to the levels observed in the AAV5 mid-dose group. A similar trend was observed in the globus pallidus (punch 22b; Figure S4B); however, observed differences in vector DNA levels between treatments were not as pronounced as in the putamen and thalamus.

Figure 4.

Intraparenchymal injection of non-human primates (Macaca fascicularis) with 1.1 × 10¹² low-dose, 4.8 × 10¹² mid-dose, or 2.3 × 10¹³ high-dose genome copies of AAV5 or 2.0 × 10¹² low-intermediate dose genome copies of AAV5neo

Vector genome copies were measured by qPCR. miRNA (miSNCA) targeting SNCA transcript and SNCA transcript were quantified by RT-qPCR. Relative abundance of miSNCA among the total miRNA pool was quantified by small RNA sequencing. SNCA protein levels were measured using a Meso Scale Discovery (MSD) assay. (A) Quantification of vector genome copies in the putamen, globus pallidus, and thalamus by qPCR. Values are the mean of vector copies/μg DNA with data points of individual animals (n = 3). (B) Transgenic miRNA (miSNCA) quantification by RT-qPCR in the putamen, globus pallidus, and thalamus. Values are the mean of miSNCA copies/μg RNA with data points of individual animals (n = 3). (C) Quantification of relative abundance of miSNCA among the pool of total miRNA by small RNA sequencing. RNA was isolated from tissue punches of the putamen. Values are the mean percentage of miSNCA among total miRNA with data points of individual animals (n = 3). (D) Cell type distribution of miSNCA positive cells in brain punches, taken from putamen (pooled from n = 3 animals) of AAV5neo-injected animals, subjected to single-nuclei RNA sequencing. Values indicate percentage of designated cell types within miSNCA-positive cell population. (E) Quantification of SNCA mRNA by RT-qPCR. Values are the mean percentage levels of SNCA target mRNA expressed as a percentage of formulation buffer-injected control animal SNCA mRNA levels with data points of individual animals (n = 3). (F) Quantification of SNCA protein by MSD. Values are the mean percentage levels of SNCA protein expressed as a percentage of formulation buffer-injected control animal SNCA protein levels with data points of individual animals (n = 3). GC, genome copies; miSNCA, microRNA targeting synuclein alpha; MSD, Meso Scale Discovery; NHP, non-human primate; AAV, recombinant adeno-associated virus.

AAV5neo-transduced cells express high levels of therapeutic transgene

RT-qPCR was used to determine whether the AAV constructs would express their therapeutic miSNCA payload (Figure 4B). Like the results obtained for vector DNA quantification, AAV5neo expressed similar levels of miRNA as the mid- and high-dose AAV5. Small RNA sequencing was conducted on RNA isolated from the putamen (punch 9b; Figure S4B) to quantify the mature transgenic miSNCA among the total population of miRNAs (Figure 4C). AAV5neo treatment consistently resulted in a greater fraction of transgenic miSNCA among the total pool of miRNAs when compared with the three tested doses of AAV5. AAV5neo contributed 3.3% transgenic miSNCA to the total miRNA, while the highest dose of AAV5 only reached ∼2.5%.

AAV5neo predominantly transduces neurons and astrocytes in the NHP putamen

Brain punches of AAV5neo-injected animals, specifically sampled from the putamen, were used for snRNA sequencing to determine which cell types were transduced (Figure 4D). In line with the minipig snRNA sequencing results, AAV5neo delivered the therapeutic transgene into the putamen predominantly to neurons (54.9%), followed by astrocytes (24.8%) and other cell types (15.8%) and, to a lesser extent, oligodendrocytes (4.5%).

miRNA expressed in AAV5neo-transduced cells shows target engagement and causes effective protein lowering

To investigate whether miSNCA expression would lower SNCA mRNA and, in turn, protein levels, RT-qPCR analysis using primers specific for SNCA transcript was performed. As observed for vector DNA and miRNA levels, across the three tested regions of interest, the use of a low-intermediate dose of AAV5neo strongly reduced SNCA mRNA levels (60%–95%), similarly to the reduction achieved using higher doses of AAV5 (Figure 4E). This reduction in SNCA mRNA levels directly translated to a 35%–70% decrease in alpha-synuclein protein (Figure 4F).

3D biodistribution mapping confirms that intraparenchymal delivery using AAV5neo is more effective

To give a global view of AAV5neo biodistribution compared to AAV5, three-dimensional biodistribution mapping was performed on mice bilaterally injected in the striatum with 1.0 × 10¹⁰ gc/brain (mid dose) of AAV5neo or AAV5 (n = 3 per group). Despite the intrastriatal injection, AAV5neo exhibits more widespread transduction in murine brains, particularly reaching the cortex (Figure 5A). In contrast, AAV5 resulted in lower general eGFP expression and more restricted three-dimensional biodistribution. The quantification of eGFP-positive cells per volume of brain tissue showed a significant increase of transduced cells in the cerebral cortex when comparing AAV5neo with AAV5 (1,477 cells/mm³ versus 720 cells/mm³) (Figures 5B and 5C). The thalamus displayed comparable median eGFP-positive cell densities. At the injection site, both AAV5 and AAV5neo displayed the highest density of transduced cells, with AAV5neo having a 1.3-fold (ns) higher median density than AAV5 (4,677 cells/mm³ versus 3,525 cells/mm³).

Figure 5.

3D biodistribution mapping of eGFP transgene expression delivered by AAV5 or AAV5neo at a dose of 1 × 10¹⁰ gc/brain

(A) 3D brain reconstruction through light sheet microscopy imaging detecting fluorescently labeled eGFP (green) in AAV5 (top row) or AAV5neo (bottom row)-injected murine brains. Scale bars, 1.5 mm. (B) Heatmap representation of eGFP-positive cells with color-coding indicating cell densities (eGFP-positive cells/mm³) mapped to Allen Brain Atlas brain sections in coronal plane (top row), transverse plane (center row), and sagittal plane (bottom row). (C) Density quantification (cells/mm³) of eGFP-positive cells in the cerebral cortex, striatum, thalamus, and hippocampal formation. Boxplots show median with quartiles of eGFP-positive cell densities (n = 3). Significance levels of statistical tests are indicated (unpaired t test). AAV, adeno-associated virus.

Discussion

Viral infection is a complex, multi-step process. For many viruses, initial attachment to the cell surface dictates tissue tropism. For AAVs, glycans such as heparan sulfate proteoglycans (HSPGs) or SIA serve as low-affinity, “promiscuous attachment factors,” sequestering and concentrating AAV at the plasma membrane to facilitate initial host cell contact.^32,40 Several AAV serotypes, including AAV1, AAV4, AAV5, and AAV6, interact with SIA. AAV5 specifically recognizes N-linked (α2-3) SIA.^34,41 Following this initial interaction, endocytic uptake, intracellular trafficking, and nuclear entry are orchestrated, in part, by the universal pan-AAV receptor (AAVR).³⁸ Notably, AAV5’s interaction with AAVR is thought to be independent of glycan attachment.⁴⁰

Early studies demonstrated that modulating AAV-glycan interactions affect neurotropism, infectivity, and biodistribution. For example, blocking AAV-HSPG interactions using heparin co-injection improved AAV2 transduction efficiency and vector spread in the murine brain.^42,43 Similarly, mutagenesis of two HSPG-interacting residues in AAV2 significantly altered CNS transduction profiles in NHPs.³⁰ While HSPG modulation has been extensively studied, the impact of manipulating AAV5-SIA binding on in vivo CNS transduction is not fully characterized.

SIA is a key component of brain glycoproteins and glycolipids, playing an essential role in neural development, communication, and cognition (for review, see Schnaar et al.).⁴⁴ Beyond its structural contributions, SIA is a crucial regulator of microglial activation and brain immunity, influencing immune homeostasis (for review, see Liao et al.).⁴⁵ Perturbations in SIA balance may disrupt innate immune regulation, potentially leading to inflammation and aberrant immune responses. This regulation occurs through key receptors such as integrin alpha M and SIA-binding immunoglobulin-type lectins, which influence microglial activation and neuroimmune signaling.⁴⁶ Given the importance of these pathways in neuroprotection, disturbances in brain SIA homeostasis could have implications for brain innate immunity, neuroinflammation, and disease pathogenesis.

Wild-type AAV5 was originally isolated from human skin.⁴⁷ The human brain has the highest SIA concentration of any organ,⁴⁴ yet wild-type AAV5 likely evolved in environments with lower SIA concentrations, resulting in suboptimal SIA affinity for the effective transduction of CNS tissues. X-ray crystallography studies identified two SIA-binding regions on the AAV5 capsid: the “A site,” located at the 3-fold symmetry axis, and the “B site,” situated beneath the βHI loop near the 5-fold symmetry axis.³⁴ The “A site” regulates SIA-dependent transduction, with the M569V mutation abrogating or reducing transduction across multiple cell lines and in murine salivary gland tissues.³⁴

In our in vitro studies, AAV5neo (an AAV5 M569V mutant) exhibited ∼10-fold lower reporter gene expression compared to AAV5 when tested in several cell lines. Similarly, long-term organoid cultures transduced with AAV5neo showed a ∼5-fold reduction in transgene-positive cells compared to AAV5. Notably, the AAV5neo transgene expression in vitro was negatively affected by pre-treatment of the cells with an AAVR-blocking antibody but not by the pre-treatment with neuraminidase. This suggests that the introduced M569V mutation affects the viral attachment (through SIA binding) but not the canonical AAVR-mediated entry. Overexpression of AAVR in the transduced cells enhanced AAV5neo transgene expression, further corroborating the AAVR dependency of the mutant capsid.

Interestingly, while loss of SIA-dependent transduction has been associated with reduced AAV5 efficacy in vivo,³⁴ our findings suggest the opposite for direct brain administration. AAV5neo showed superior CNS transduction when injected directly into the brain parenchyma while performing comparably with AAV5 when delivered via other routes (data not shown). We hypothesize that high Neu5Ac (the dominant SIA form in the brain) concentrations may restrict AAV5 access to AAVR, thereby limiting transduction efficiency. The estimated human brain SIA concentration is ∼3 mM based on wet weight calculations, and our studies showed that 2.5–5 mM SIA (Neu5Ac) reduces parental AAV5 transduction in vitro by 25%–40%, supporting the idea that SIA abundance influences CNS transduction. Indeed, murine studies conducted using engineered AAV1 variants and wild-type AAV1 have advanced the notion that limiting SIA dependence could improve in vivo AAV1 CNS transduction.³⁹

Minipig and NHP studies confirmed superior CNS transduction and vector spread by AAV5neo following direct brain injection. Minipigs receiving intraparenchymal AAV5neo-eGFP displayed a 3-fold larger transgene-positive area and up to 150-fold higher eGFP expression in the putamen and caudate than AAV5. Similarly, in NHPs, AAV5neo achieved comparable vector DNA levels in the putamen despite being administered at ∼10-fold lower doses than AAV5. Additionally, AAV5neo mediated enhanced miRNA expression and more efficient target mRNA suppression, demonstrating its improved functional transduction efficiency in the CNS of NHP models.

When an AAV drug suspension is locally injected into a soft viscoelastic tissue, such as the brain, it distributes less uniformly than it would in a liquid-filled organ (e.g., blood or cerebrospinal fluid).⁴⁸ Instead, convection and diffusion contribute to a gradient-like distribution, with the highest viral concentration remaining near the injection site.⁴⁹ The brain’s extracellular space (∼20% of total volume) allows limited movement of viral particles influenced by molecular size, charge, and tissue resistance.⁵⁰ Since distal brain areas are exposed to a lower MOI than at the injection site, we expect more sensitivity to SIA abundance, thereby limiting the spread of AAV5-delivered transgenes but not the SIA-insensitive AAV5neo-delivered transgenes. This results in more efficient transduction and increased vector dissemination. Additionally, the inherent retrograde transport of AAV5-based capsids may further facilitate the enhanced distribution of AAV5neo.²²

Given that SIA abundance is relatively higher in humans than in other mammals, particularly Neu5Ac, we anticipate that the benefits of AAV5neo will be more pronounced in human applications than in the animal models evaluated. The improved transduction efficacy and three-dimensional distribution of AAV5neo may be leveraged to minimize vector dose and the number of injections required.

Given the role of SIAs in neuroimmune regulation, understanding how AAV5neo may improve the safety and efficacy of AAV5-based gene therapy will be crucial. Future studies should therefore investigate AAV5neo’s impact on innate immunity, including microglial activation and cytokine signaling.

Conclusions

Here, we have shown that AAV5neo, a modified AAV5 vector engineered with a single amino acid substitution, demonstrates significantly improved CNS biodistribution and enhanced potency compared to the parental AAV5 vector. AAV5neo exhibits altered glycan-binding properties that reduce its affinity for SIA. Limiting SIA binding likely enables more effective dissemination within the SIA-rich brain parenchyma while leaving the interaction with AAVR unperturbed. AAV5neo demonstrated a ∼10× increase in transduction efficiency in murine, minipig, and NHP models compared to AAV5. 3D biodistribution mapping confirmed the enhanced dissemination of AAV5neo, which transduced regions beyond the injection site, reaching deep brain structures and cortical areas. We conclude that the improved efficacy and biodistribution position AAV5neo as a promising candidate for human CNS gene therapy applications.

Materials and methods

AAV5neo plasmid generation

The pRep-CapAAV5neo expression plasmid was generated by site-directed mutagenesis PCR using Primer Fw 5′-CAGGTGGCCACCAACAAC-3′ and Primer Rv 5′-CCCGCCGACGTTGTACG-3′ using the helper virus-free AAV-5 Rep-Cap plasmid (Cell Biolabs Inc., San Diego, CA) as a PCR template.

The amplicon was agarose gel purified and kinase-ligase-DpnI treated using the Q5 site-directed mutagenesis kit (New England Biolabs, Ipswich, MA). Circularized plasmid DNA was propagated in NEB stable E. coli (New England Biolabs, Ipswich, MA), and the identity of isolated plasmid DNA was confirmed by Sanger sequencing (Macrogen, Amsterdam, the Netherlands) and restriction-digestion analysis.

AAV production of AAV5 and AAV5neo constructs expressing nLuc and eGFP reporter genes for in vitro studies

The parental AAV5 and AAV5neo vectors were produced via polyethylenimine (PEI)-mediated transient triple transfection of HEK293T cells (Sigma-Aldrich, St. Louis, MO). In brief, HEK293T cells were seeded in 15 cm dishes 24 h pre-transfection and transfected 1 day later with pHelper, pRepCapAAV5 or pRepCapAAV5neo, and pITR-nLuc or pITR-eGFP. Cells were harvested 72 h post transfection, and cell pellets were lysed with 1× AAV lysis buffer (Lonza, Basel, Switzerland). AAV was affinity purified from cleared lysate using AVB Sepharose HP resin (Cytiva, Marlborough, MA). Viral titers were determined using Taqman real-time PCR (Thermo Fisher Scientific, Waltham, MA). Capsid purity and viral protein stoichiometry were evaluated via SDS PAGE analysis using 4%–20% Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad, Hercules, CA).

Production of AAV5 and AAV5neo expressing eGFP or miSNCA transgenes for in vivo studies

The AAV5 and AAV5neo vectors were produced (Sirion Biotech GmbH, Gräfelfing, Germany) via PEI-mediated transient triple transfection of HEK293T cells (Sigma-Aldrich, St. Louis, MO). In brief, HEK293T cells were cultured in 10-stack cell factories and transfected with pHelper, pRepCapAAV5 or pRepCapAAV5neo, and pITR-CBA-eGFP. AAV was affinity purified from cell lysate and PEG-precipitated culture supernatant using Poros CaptureSelect AAV-X (Thermo Fisher Scientific, Waltham, MA) 48 h post transfection. Full capsids were enriched by iodixanol gradient centrifugation, concentrated using Amicon ULTRA 15 (Merck Millipore, Burlington, MA) centrifugal units, and diluted in PBS and 0.001% Pluronic F-68 (Thermo Fisher Scientific, Waltham, MA). Viral titer was evaluated via ITR-specific qPCR and total particle count using an AAV5-specific ELISA (PROGEN Biotechnik GmbH, Heidelberg, Germany). Capsid purity and viral protein stoichiometry were evaluated via SDS PAGE analysis. Endotoxin levels were determined using Endosafe nexgen-PTS (Charles River Laboratories, Wilmington, MA).

AAV5 or AAV5neo vectors encoding the CBA-miSNCA cassette for NHP dosing were generated by a baculovirus-based AAV production system (uniQure, Amsterdam, the Netherlands) as previously described.¹⁰ The ITR-flanked CBA-miSNCA expression cassettes were inserted in a recombinant baculovirus vector by homologous recombination, and clones were selected by plaque purification. The recombinant baculovirus containing the cassettes was further amplified, and clones were screened for optimal production and stability. To generate AAV5 or AAV5neo, Sf+ cells were infected with recombinant baculoviruses encoding rep for replication and packaging, and cap-5 or cap-5neo for the AAV capsid and the CBA-miSNCA expression cassette.

After viral particle assembly, purification was performed with AVB Sepharose high-performance affinity resin (Cytiva, Marlborough, MA, United States) using AKTA Explorer purification system (Cytiva, Marlborough, MA, United States). Viral particles were formulated in phosphate-trehalose buffer, and viral titer was evaluated via promoter-specific qPCR and total particle count using an AAV5-specific ELISA (PROGEN Biotechnik GmbH, Heidelberg, Germany). Capsid purity and viral protein stoichiometry were evaluated via SDS PAGE analysis. Endotoxin levels were determined using Endosafe nexgen-PTS (Charles River Laboratories, Wilmington, MA, United States).

In vitro transduction

Luciferase reporter gene expression assay

HEK293T (Sigma-Aldrich, St. Louis, MO), Huh7 (FUJIFILM Wako Chemicals Europe GmbH, Neuss, Germany), and COS-1 cells (ATCC, Manassas, VA, United States) were cultured in Gibco Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific, Waltham, MA) supplemented with 10% Gibco FBS (Thermo Fisher Scientific, Waltham, MA) (DMEM-FBS) and seeded at 1 × 10⁵ cells per well in 48-well culture plates 24 h prior to transduction. SH-SY5Y cells (Sigma-Aldrich, St. Louis, MO, United States) were cultured in a 1:1 mixture of Gibco modified Eagle’s medium (MEM) (Thermo Fisher Scientific, Waltham, MA, United States) and Ham’s F-12 Nutrient Mix + GlutaMAX-I (Thermo Fisher Scientific, Waltham, MA, United States), supplemented with 15% FBS, Gibco sodium pyruvate (Thermo Fisher Scientific, Waltham, MA, United States), and Gibco MEM non-essential amino acids (Thermo Fisher Scientific, Waltham, MA, United States), and seeded at 1 × 10⁵ cells per well in 48-well culture plates.

The seeding medium was aspirated, and cells were infected in their respective culture medium with adenovirus 5 (Ad5) at an MOI of 30 and either AAV5-nLuc or AAV5neo-nLuc at an MOI of 1 × 10⁵ or 1 × 10⁶ gc per cell. Cells were incubated for 48 h at 37°C/5% CO₂. Following incubation, nLuc activity was measured using the Nano-Glo Luciferase Assay System (Promega, Madison, WI, United States) in GloMax Explorer Multimode Microplate Reader (Promega, Madison, WI, United States).

Fluorescence microscopy

Cells for in vitro experiments were imaged in bright field and FITC channels using an Evos M5000 (Thermo Fisher Scientific, Waltham, MA) microscope using 10× magnification objective.

Flow cytometric analysis

HEK293T cells were seeded at 1 × 10⁵ cells per well in 48-well culture plates 24 h prior to transduction and similarly infected with Ad5 at an MOI of 30 and either AAV5-eGFP or AAV5neo-eGFP at an MOI of 1 × 10⁵ or 1 × 10⁶ gc/cell. Cells were incubated for 48 h at 37°C/5% CO₂ prior to harvesting by washing with 1× Gibco DPBS (Thermo Fisher Scientific, Waltham, MA) and trypsinizing with Gibco Trypsin-EDTA (Thermo Fisher Scientific, Waltham, MA). Cells were stained with LIVE/DEAD Fixable Aqua Stain (Invitrogen, Waltham, MA) prior to flow cytometry analysis using an LSR Fortessa flow cytometer (BD Biosciences, Franklin Lakes, NJ, United States) to quantify eGFP expression in the viable cell population.

Dorsal forebrain brain organoid transduction assay

STEMdiff (STEMCELL Technologies, Vancouver, Canada) dorsal forebrain organoids were generated from human induced pluripotent stem cell according to an adapted STEMdiff dorsal forebrain organoid differentiation kit protocol. Representative forebrain organoids were enzymatically and mechanically dissociated, and cell numbers were determined via flow cytometry. An average of 3.5 × 10⁵ cells per organoid was determined for subsequent MOI calculations. Two organoids per 48-well plate well and per sampling point were cultured in neural organoid basal medium 2 (STEMCELL Technologies, Vancouver, Canada). AAV5-eGFP or AAV5neo-eGFP was added at an MOI of 5 × 10⁵ gc/cell and incubated for 3 days at 37°C/5% CO₂ and 110 rpm orbital shaking. Neural organoid basal medium was refreshed every 3 days. Two organoids per treatment were harvested for analysis. In brief, organoids were transferred to non-coated low attachment 48-well plates, washed twice with DPBS, and treated for 30 min with Gibco Hank’s balanced salt solution (Thermo Fisher Scientific, Waltham, MA, United States) supplemented with 250 U/mL DNaseI (Thermo Fisher Scientific, Waltham, MA, United States) and 30 U/mL papain (Sigma-Aldrich, St. Louis, MO). During enzymatic treatment, organoids were mechanically dissociated. Enzymatic digestion was stopped by adding MEM with GlutaMAX-I (Thermo Fisher Scientific, Waltham, MA, United States) supplemented with 10% Gibco FBS (Thermo Fisher Scientific, Waltham, MA, United States). Cell suspensions were homogenized via passage through a cell strainer. Cells were washed and stained with LIVE/DEAD Fixable Aqua Stain (Invitrogen, Waltham, MA, United States) prior to flow cytometry analysis for live-cell eGFP expression.

Neuraminidase A transduction assay

HEK293T cells were seeded in DMEM-FBS at 1 × 10⁵ cells per well in 48-well culture plates 24 h prior to transduction. Seeding medium was replaced with DMEM-FBS containing Ad5 at an MOI of 30 and incubated for 2 h at 37°C/5% CO₂. Ad5-containing medium was then replaced with 1× Glycobuffer I (New England Biolabs, Ipswich, MA, United States) supplemented with 10U per well neuraminidase A (New England Biolabs, Ipswich, MA, United States) and incubated for 2 h at 37°C/5% CO₂. Neuraminidase A was then replaced with DMEM-FBS containing AAV5 or AAV5neo at an MOI of 1 × 10⁵ gc per cell and incubated for 48 h at 37°C/5% CO₂. Following incubation, nLuc activity was measured using the Nano-Glo Luciferase Assay System (Promega, Madison, WI, United States) in GloMax Explorer Multimode Microplate Reader (Promega, Madison, WI, United States).

Anti-AAVR antibody transduction assay

Huh7 cells (FUJIFILM Wako Chemicals Europe GmbH, Neuss, Germany) were seeded in DMEM-FBS at 5 × 10⁴ cells per well in 48-well culture plates 24 h prior to transduction. Seeding medium was replaced with DMEM-FBS supplemented with either 5 mg/mL anti-AAVR antibody (ab105385; Abcam, Cambridge, United Kingdom) or 5 mg/mL mouse anti-human IgG isotype control (MT91/145; Mabtech, Nacka Strand, Sweden) and incubated at 4°C for 1 h. Antibody-containing medium was replaced with DMEM-FBS containing Ad5 at an MOI of 30 and either AAV5 or AAV5neo at an MOI of 1 × 10⁵ GC per cell. Cells were incubated for 48 h 37°C/5% CO₂. nLuc activity was measured using the Nano-Glo Luciferase Assay System (Promega, Madison, WI, United States) in GloMax Explorer Multimode Microplate Reader (Promega, Madison, WI, United States).

Neu5Ac competition assay

COS-1 cells (ATCC, Manassas, VA, United States) were seeded in DMEM-FBS at 3 × 10⁴ cells per well in 96-well culture plates 24 h prior to performing the experiment. AAV5 or AAV5neo was added at an MOI of 1 × 10⁴ GC/cell in DMEM-FBS supplemented with 0, 2.5, 5, 10, or 25 mM Neu5Ac (Sigma-Aldrich, St. Louis, MO, United States). Alternatively, DMEM-FBS, either containing 25 mM Neu5Ac or no Neu5Ac, was used to dilute AAV to an MOI of 1 × 10³, 1 × 10⁴, 1 × 10⁵, or 1 × 10⁶ gc/cell and incubated for 3 h at 37°C with 135 rpm orbital shaking. In parallel, COS-1 cells were infected with Ad5 at an MOI of 30 for 3 h at 37°C. After Ad5 removal, DMEM-FBS containing AAV and Neu5Ac was added to COS-1 cells and incubated for 48 h at 37°C/5% CO_2. Subsequently, nLuc activity was measured using the Nano-Glo Luciferase Assay System (Promega, Madison, WI) in GloMax Explorer Multimode Microplate Reader (Promega, Madison, WI, United States).

Transient transfection-mediated overexpression of AAVR in cell lines

AAVR ([KIAA0319L], Homo sapiens) was cloned into a pUC-based plasmid and transfected into HeLa cells using Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, USA) and a reverse transfection protocol. 48 h post transfection, AAV5neo was added at an MOI of 1 × 10⁵ gc/cell, and nLuc activity measured 48 h post transduction using the Nano-Glo Luciferase Assay System (Promega, Madison, Wisconsin) in GloMax Explorer Multimode Microplate Reader (Promega, Madison, WI, United States).

In vivo transduction

Mice Mus musculus

All experiments were conducted in accordance with the Dutch Competent Authority and approved by the local Animal Welfare Body, following the European Directive 2010/63/EU on the protection of animals used for scientific purposes. C57BL/6J mice (male, 6–8 weeks) were obtained from Charles River Laboratories (Charles River Laboratories, Sulzfeld, Germany). Animals were identified via RFID tags (Digitail tags, Somark Innovations, San Diego, United States) subcutaneously injected under anesthesia in the tail base prior to intraparenchymal administration.

Libechov minipigs Sus scrofa

All experiments were carried out according to the guidelines for the care and use of experimental animals and approved by the State Veterinary Administration of the Czech Republic. The study was carried out in accordance with the Institutional Animal Care and Use Committee of the Institute of Animal Physiology and Genetics, according to current Czech regulations.

NHPs Macaca fascicularis

Studies complied with all applicable sections of the current version of the Final Rules of the Animal Welfare Act Regulations (9 CFR), and the Guide for the Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council, 8th edition.

Procedures used in these studies have been designed with the consideration of the well-being of the animals. Whenever possible, efforts to avoid or minimize discomfort, distress, or pain were implemented.

Intraparenchymal delivery of AAV5 and AAV5neo constructs expressing eGFP in Mus musculus C57BL/6J mice

Animals were anesthetized with Isoflutek 1,000 mg/g (Laboratorios Karizoo, Barcelona, Spain) in oxygen (induction at 5% and maintenance at 2%–3%) using a RAS-4 Rodent Anesthesia System (Revvity, Waltham, MA, United States). Animals were placed in an automated RWD71000, RWD stereotaxic device (RWD Life Science Co., Ltd., Shenzhen, China), and the head was leveled with a maximal height difference between bregma and lambda of 0.1 mm. AAV5 CBA-eGFP or AAV5neo CBA-eGFP was injected bilaterally (3 μL) into the striatum (total volume of 6 μL containing 1.0 × 10⁹ [low dose], 1.0 × 10¹⁰ [mid dose], or 1.0 × 10¹¹ [high dose] gc per brain) at a flowrate of 0.2 μL/min. The drug substance was bilaterally injected at coordinates AP 0.8 mm/ML ±1.8 mm/DV 3.2 mm as indicated in Figure S2A. Mice (n = 9 per treatment) were injected with AAV5 CBA-eGFP or AAV5neo CBA-eGFP constructs. Negative control (n = 4) animals were injected with 3 μL PBS+0.001% pluronic per hemisphere. Tissues were collected 4 weeks post procedure. Mice were anesthetized with Isoflutek 1,000 mg/g (Laboratorios Karizoo, Barcelona, Spain) in oxygen (5% induction, 2%–3% maintenance) and transcardially perfused with ice-cold saline. Brain and liver tissues were extracted and either post hoc fixed in 4% paraformaldehyde (PFA) followed by storage in 70% ethanol for subsequent immunohistochemical analysis or snap-frozen in liquid nitrogen for molecular analysis.

3D AAV biodistribution mapping

Mice were intrastriatally injected with 1 × 10¹⁰ gc per brain (mid-dose) of AAV5 or AAV5neo (n = 3 per group) following the procedure described previously. Four weeks after dosing, animals were transcardially perfused with ice-cold 1× PBS supplemented with 10 U/mL heparin until the fluids ran clear, followed by ice-cold 4% PFA. The brains were extracted, and the samples were post fixed in 4% PFA for 24 h and transferred to PBS with 0.02% sodium azide. Whole-mouse brain clearing and imaging were performed at LifeCanvas Technologies (Cambridge, MA, United States).

Fixed whole brains were prepared using proprietary SHIELD technology to preserve protein before being cleared and immunolabeled with eGFP, NeuN, and GFAP antibodies using SmartBatch + LifeCanvas Technologies (Cambridge, MA, USA). Labeled brains were imaged using a SmartSPIM light sheet microscope LifeCanvas Technologies (Cambridge, MA, USA). Images were subsequently reconstructed using the ImarisViewer 10.2 Software Package (Oxford Instruments, Abingdon, United Kingdom). ImageJ software (NIH, Bethesda, MD, United States) was used for image processing, analysis, and quantification.

Immunohistochemical detection of eGFP in murine brain sections

Mouse brain tissues were dehydrated in graded ethanol and embedded in paraffin blocks. Sagittal sections (4 μm thickness) were prepared by microtome sectioning, deparaffinized, and treated for antigen retrieval at 95°C in Tris-EDTA (pH 9.0). Tissue sections and appropriate controls were blocked and stained using a Leica Bond RX system (Leica Biosystems, Nussloch, Germany) with (1:500) rabbit anti-GFP antibodies (Ab290, Abcam, Cambridge, United Kingdom) and (1:1,000) donkey anti-rabbit-Alexa Fluor 647 (A31573, Thermo Fisher Scientific, Waltham, MA, United States) and counterstained with DAPI (Abcam, Cambridge, United Kingdom). eGFP immunofluorescence images were acquired in Alexa Fluor 647 channel using a Zeiss AxioScan Z1 Slide Scanner (Carl Zeiss, Oberkochen, Germany) with a Plan-Apochromat 20× objective and a 24-bit RGB camera (200 μs exposure) using the ZEN Blue Software v.2.6 (Carl Zeiss, Oberkochen, Germany). Image analysis was performed using HALO Software v.4.0 (IndicaLabs, Albuquerque, NM, United States) with an AI-based tissue detection classifier. Brain regions were segmented based on eGFP fluorescence with reference to the Allen Mouse Brain Atlas. The following regions were manually annotated: cortex, hippocampus, thalamus, striatum, hypothalamus, midbrain, brainstem, cerebellum, and olfactory bulb. Quantitative analysis of immunohistochemistry (IHC)-stained sections was conducted using Area Quantification FL v.2.3.4 (IndicaLabs, Albuquerque, NM, USA). DAPI counterstaining was used for nuclei detection. Results are expressed as eGFP-positive area (μm²) within the regions of interest or per brain section.

Quantification of vector genome copy numbers in murine samples

Murine brain tissues were dissected into respective anatomical areas and homogenized using the cryoPREP tissue disruption system (Covaris, Woburn, MA, USA) and MatrixD lysing matrix tubes (MP Biomedicals, Santa Ana, CA, USA) into AllPrep kit lysis buffer. DNA and RNA were subsequently isolated using the AllPrep DNA/RNA mini kit (QIAGEN, Venlo, the Netherlands) as per the manufacturer’s instructions. Isolated nucleic acids were quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). DNA concentrations of samples were normalized. Vector genomes were quantified by detecting transgenic promoter sequences (CMV enhancer Fw: 5′-AGT AAC GCC AAT AGG GAC TTTC-3′; CMV enhancer Rv: 5′-GGC GTA CTT GGC ATA TGA TAC A-3′; CMV enhancer Probe: 5′-TTA CGG TAA ACT GCC CAC TTG GCA-3′) using TaqMan Real-Time PCR (Thermo Fisher Scientific, Waltham, MA, USA). Vector genome copy numbers were calculated by plasmid standard line and expressed as copies per μg of total DNA.

eGFP mRNA quantification

RNA concentrations were normalized and treated with DNase I (Thermo Fisher Scientific, Waltham, Massachusetts), to remove residual contaminating DNA. cDNA was synthesized from total RNA using Maxima Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA, USA).

Primer/probe combinations specific for the eGFP transgene sequence (eGFP Fw: 5′-AGC AAA GAC CC CAAC GAG AA-3′; eGFP Rv: 5′-GCG GCG GTCA CGA ACT C-3′; eGFP probe: 5′-CGC GAT CAC ATG GTC CTG CT-3′) were used to quantify transgene amplification in a TaqMan qPCR (Thermo Fisher Scientific, Waltham, MA, USA) on Quantstudio 5 real-time PCR system (Applied Biosystems, Waltham, MA, USA). The amount of transgene transcript was calculated from a plasmid standard curve. Results were reported as transcript copies per μg of total RNA.

Intraparenchymal injection of Sus scrofa Libechov minipigs with AAV5 and AAV5neo constructs expressing eGFP

Male minipigs 6–9 months were anesthetized, and surgical procedures were carried out as previously described.¹⁰ Animals (n = 3 per group) were injected with a total dose of 2 × 10¹² gc/brain of AAV5 CBA-eGFP or AAV5neo CBA-eGFP. One untreated animal acted as a negative control. Two injections of AAV particles were administered to each animal, with the injection rate increasing every minute until reaching the final volume of 50 μL per putamen (first min: 1 μL/min, second min: 2 μL/min, etc.). The skull was closed with bone graft and cement, periosteum stitching, and skin by suturing. Perioperative as well as postoperative antibiotic and analgesic were administrated. Four weeks after dosing the animals were anesthetized by TKX mixture and euthanized by an overdose of thiopental. Immediately after overdosing, the pigs were transcardially perfused with 20 L of ice-cold PBS. Brain samples for molecular analysis were taken from one hemisphere as 4-mm punches from 4-mm-thick coronal brain slabs (punching scheme Figure S3) and were immediately snap-frozen in liquid nitrogen and stored at −80°C. The remaining hemisphere was cut into 4-mm slabs and post-fixed in 4% PFA (pH 7.4) in PBS for 24 h.

Cryo-immunohistochemical analysis of eGFP expression in minipig brain sections

Post-fixed brain slabs were transferred into 30% sucrose solution containing 0.01% sodium azide and left until saturation. Well-saturated slabs were subsequently frozen in a cryostat (CM1950; Leica Biosystems), sectioned into 20-μm-thick coronal sections, and mounted onto large microscope slides (76 × 51 × 1 mm). The endogenous peroxidase activity was blocked with a solution of 0.3% hydrogen peroxide in methanol for 20 min, and the brain sections were immunostained using a rabbit primary antibody raised against GFP (1:1,000, ab6556; Abcam, Cambridge, United Kingdom), a biotinylated donkey anti-rabbit secondary antibody (1:400, RPN 1004V; GE Healthcare Life Sciences), and a tertiary avidin-peroxidase complex (1:400, A3151; Sigma-Aldrich, St. Louis, MO, United States), which was visualized by a 5-min incubation with 3,3-diaminobenzidine solution (4170; Kementec Solutions, Taastrup, Denmark). The sections were dehydrated and mounted with DePeX (Thermo Fisher Scientific, Waltham, MA, United States). Images were acquired using a histological scanner (Virtual Slide Microscope VS120-5 fluorescence; Olympus, Tokyo, Japan), and quantitative analysis of IHC-stained brain sections was performed using Fiji ImageJ distribution (https://fiji.sc/). The sections were scaled, calibrated, converted to 8-bit, and thresholded using default (GFP positivity) as well as triangle (for whole section) threshold. Manually assigned ROIs were used for striatum selection in GFP-stained sections. Results are expressed as eGFP-positive area (% of total area) per brain section.

Molecular analysis of vector DNA and transgene mRNA

Brain punches were taken according to the punching scheme (Figure S3). Deep-frozen brain punches were homogenized using the cryoPREP tissue disruption system (Covaris, Woburn, MA, United States) and MatrixD lysing matrix tubes (MP Biomedicals, Santa Ana, CA, United States) into AllPrep kit lysis buffer. DNA and RNA were subsequently isolated using the AllPrep DNA/RNA mini kit (QIAGEN, Venlo, the Netherlands) as per the manufacturer’s instructions. Isolated nucleic acids were quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). DNA concentrations of samples were normalized. Vector genomes were quantified by detecting transgenic promoter sequences (CMV enhancer Fw: 5′-AGT AAC GCC AAT AGG GAC TTTC-3′; CMV enhancer Rv: 5′-GGC GTA CTT GGC ATA TGA TAC A-3′; CMV enhancer Probe: 5′-TTA CGG TAA ACT GCC CAC TTG GCA-3′) using TaqMan Real-Time PCR (Thermo Fisher Scientific, Waltham, MA, United States). Vector genome copy numbers were calculated by plasmid standard line extrapolation and expressed as copies per μg of total DNA. Transgene mRNA transcripts were quantified following RNA extraction and quantification as previously described. After normalization of RNA concentration, RNA samples were treated with DNase I (Thermo Fisher Scientific, Waltham, MA, United States), to remove residual contaminating DNA. cDNA was synthesized from total RNA using Maxima Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA). Primer/probe combinations specific for the eGFP transgene sequence (eGFP Fw: 5′-AGC AAA GAC CC CAAC GAG AA-3′; eGFP Rv: 5′-GCG GCG GTCA CGA ACT C-3′; eGFP probe: 5′-CGC GAT CAC ATG GTC CTG CT-3′) were used to quantify transgene amplification in a TaqMan qPCR (Thermo Fisher Scientific, Waltham, MA) on Quantstudio 5 real-time PCR system (Applied Biosystems, Waltham, MA). The amount of transgene transcript was calculated from a plasmid standard curve. Results were reported as transcript copies per μg of total RNA.

snRNA RNA sequencing in minipig and NHP brain samples

Single-nuclei sequencing was conducted by SingulOmics Corporation (New York, NY, United States) on AAV5neo-eGFP-injected frozen minipig brain tissue punches obtained from the frontal lobe white matter (punch 56) and caudate nucleus (punch 80; Figure S3), following standard procedures. Brain punches from NHP injected with AAV5neo-miSNCA were obtained from the putamen (punch 21a; Figure S4B). Briefly, nuclei from these regions were pooled from three animals to yield around 5,000 nuclei per sample. 3′ single-cell gene expression libraries using Next GEM v.3.1 (10× Genomics, Pleasanton, CA, United States) were generated using the Chromium system (10× Genomics, Pleasanton, CA, United States). Following library preparation, sequencing was performed on the Illumina NovaSeq platform (Illumina, San Diego, CA, United States), resulting in approximately 400 million paired-end 150-base reads per sample. The minipig sequencing data were subsequently analyzed against the minipig reference genome (GCA_000003025.6), incorporating the eGFP transgene. The NHP sequencing data were subsequently analyzed against the Macaca fascicularis reference genome (GCF_000364345.1), incorporating the miSNCA transgene.

Intraparenchymal delivery of AAV5 and AAV5neo constructs expressing miSNCA in Macaca fascicularis NHPs

Animals were anesthetized and maintained per test facility SOP of Northern Biomedical Research, Inc. (Spring Lake, MI, United States). Dose administration was performed using real-time MRI-guided injections targeting the putamen (bilateral) using an occipital approach. Each group (n = 3) was injected with either 1.1 × 10¹² (low dose), 4.8 × 10¹² (mid dose), or 2.3 × 10¹³ (high dose) gc of AAV5 or 2.0 × 10¹² (low-intermediate) gc of AAV5neo. One animal was injected with phosphate-trehalose formulation buffer acting as vehicle control. An incision was made over the calvarium and a craniotomy over the putamen. An anatomical MRI scan was performed to identify the infusion site, and the cannula guide was aligned to the target infusion site. Trajectories and target depth were confirmed by MRI, and the actual target location and target coordinates were recorded. MRIs acquired during the treatment were retrospectively analyzed by Brainlab Automatic Image Registration software (Brainlab, Munich, Germany) to estimate gadoteridol distribution within and outside the target region and to assess follow-up imaging changes within the infused region (Figure S4A). A dose volume of 150 μL was administered using convection-enhanced delivery via a calibrated MRI-compatible PHD 2000 Infuse/Withdraw infusion pump (Harvard Apparatus, Holliston, MA, United States). Animals were euthanized 8 weeks post surgery.

Vector DNA quantification in NHP brain punches

Vector DNA quantification was performed according to NBS (Northern Biomolecular Services, Kalamazoo, Michigan) validated SOP 0027-0002-B. Brain tissue punches obtained according to punching scheme (Figure S4B) were weighed upon collection and frozen on dry ice. Tissues were homogenized using a Precellys tissue homogenizer (Bertin Technologies SAS, Montigny-le-Bretonneux, France). Homogenized tissues were lysed, and lysates destined for DNA extraction were treated with RNase A according to the NBS SOP. Lysates were loaded on QIASymphony DNA kit and processed using a QIASymphony instrument (QIAGEN, Venlo, the Netherlands) for automated DNA isolation. Isolated DNA samples were analyzed by NanoDrop (Thermo Fisher Scientific, Waltham, MA, United States) for concentration and purity. Target vector genome copy numbers were quantified using a validated TaqMan qPCR assay (Primer Fw: 5′-CAT CGG CGC TAT GCT TCC T-3′; Primer Rv: 5′-GTG TCT CAT CGC AAA CTT ACA GTA TAT G-3′, TaqMan Probe: 5′- 6FAM- TGA AAT CCC AGC CAG GG-MGB-NFQ-3′; Applied Biosciences, Waltham, MA, United States) on a QuantStudio 7 Flex Real-Time PCR system. The target DNA copy number in each sample qPCR reaction was interpolated from the included linearized plasmid-based standard curve.

RT-qPCR quantification of transgene microRNA (miSNCA) expression in NHP brain samples

Stem-loop RT-qPCR assay for absolute quantification of vector-derived miSNCA miRNA was performed according to NBS (Northern Biomolecular Services, Kalamazoo, MI, United States) validated SOP 0027-0004-C. Brain tissue punches obtained according to punching scheme (Figure S4B) were weighed upon collection and frozen on dry ice. Tissues were homogenized using a Precellys tissue homogenizer (Bertin Technologies SAS, Montigny-le-Bretonneux, France). Homogenized tissues were lysed and lysates destined for RNA isolation were chloroform extracted according to NBS SOP. Lysates were loaded on QIASymphony RNA kit and processed using a QIASymphony instrument (QIAGEN, Venlo, the Netherlands) for automated RNA isolation. Isolated RNA samples were analyzed by NanoDrop (Thermo Fisher Scientific, Waltham, MA, United States) for concentration and purity. cDNA was synthesized from normalized RNA using MultiScribe Reverse Transcriptase (Invitrogen, Waltham, MA, United States) and used in a TaqMan custom small stem-loop RNA RT-qPCR assay (primer/probe kit product no.: 4919392; Applied Biosciences, Waltham, MA, United States) for absolute quantitation of vector-derived miSNCA miRNA according to the NBS SOP.

Small RNA sequencing in NHP brain samples

Small RNA sequencing of total RNA samples acquired from putamen brain punches was conducted (Psomagen, Rockville, MD, United States) following standard procedures. In summary, small RNA libraries were generated using the NEXTFlex Small RNA Sequencing Kit v.4 (Bioo Scientific Corporation, Austin, TX, United States) according to the manufacturer’s guidelines. Subsequently, the small RNA libraries underwent sequencing on the Illumina NovaSeq 6000 SP platform (Illumina Inc., San Diego, CA, United States) with a single flow cell, producing 50-base pair paired-end reads. Analysis of miRNA expression and processing was performed using CLC Genomics Workbench 22.0.2 (QIAGEN, Venlo, the Netherlands). The small RNA reads were trimmed and annotated against the miRNA human database (miRbase v22) and the reference sequences of the pri-miSNCA constructs. IsomiRs were extracted using CLC Genomics Workbench 22.0.2 (QIAGEN, Venlo, the Netherlands), and the processing lengths and proportion of miSNCA expression in the total pool of endogenous miRNAs were determined through custom code in R/RStudio 2022.2.0 (RStudio, Boston, MA, United States).

RT-qPCR quantification of SNCA expression in NHP brain samples

Tissue processing, RNA extraction and quantification, and cDNA synthesis followed NBS SOP as described previously. cDNA samples were analyzed for NHP endogenous SNCA (primer/probe kit product no.: 4351368; Applied Biosciences, Waltham, MA, United States) and hypoxanthine phosphoribosyltransferase 1 (HPRT1; Primer Fw: 5′-TGA GGA TTT GGA AAG GGT GTT T-3′, Primer Rv: 5′-CCT TCA TCA CAT CTC GAG CAA G-3′, TaqMan Probe: 5′-VIC – TTC CTC ATG GAC TAA TTA TGG ACA GGA CTG AAC G – QSY-3′, Applied Biosciences, Waltham, MA, United States) mRNA copy numbers using a qualified one-step duplex RT-qPCR method. Prior to calculating group means, the sample results were normalized to 1 μg of RNA input.

Alpha-synuclein protein quantification

Tissue lysates for protein measurements were generated using TissueLyser III (QIAGEN, Venlo, the Netherlands). Briefly, each brain punch was added into 350 μL of RIPA buffer (Sigma-Aldrich, St. Louis, MO, United States) containing Pierce phosphatase and protease inhibitors (Thermo Fisher Scientific, Waltham, MA, United States) in Matrix D tubes (MP Biomedicals, Santa Ana, CA, United States). The tissue was then lysed in TissueLyser III (QIAGEN, Venlo, the Netherlands) twice for 30 s at 30 Hz. The lysate was cleared by centrifugation, and total protein concentration was determined by standard line extrapolation using Pierce BCA Protein Assay Reagent (Thermo Fisher Scientific, Waltham, MA, United States). For determination of alpha synuclein levels, the tissue lysates were normalized to 125 μg/mL total protein. The samples were measured in duplicates following manufacturer’s instructions (Meso Scale Discovery [MSD], Rockville, MD, United States). In summary, streptavidin-coated plates were coated with Biotin anti-human α-synuclein antibody. Then, the detection antibody (Sulfo-tag anti-human α-synuclein) mixed with sample or alpha synuclein standard was added to the coated plate. The plate was read on Mesoscale Quickplex SQ 120 plate reader after addition of 150 μL/well of Read Buffer T. A standard curve was generated using alpha synuclein calibrator between 2.61 and 10.700 pg/mL. The concentration of alpha synuclein in each sample was then extrapolated from the standard curve.

Statistical methods

For two groups, data were analyzed using Student’s t test or ordinary one-way ANOVA to determine statistical significances between samples. For multiple comparison data, two-way ANOVA with Šidák correction (α = 0.05) were applied. Where applicable, p values are listed in text or figures.

Data and code availability

Materials and reagents commercially procured will be shared on request. Reagents such as the AAV vectors generated by authors will not be shared unless a memorandum of understanding is signed between the research sponsoring institution and the requester. Methods are available to disclose, and all data associated with this study are present within the manuscript.

Acknowledgments

This study was funded by uniQure biopharma B.V., Amsterdam, the Netherlands. We thank Renata Baptista for providing forebrain organoids and related protocols. We are grateful to Elena Santidrián Yebra-Pimentel, Bas Bosma, Tom van der Zon, Rudy de Laat, Shrijana Tripathi, and former colleagues, Lyubomir Momchev and Hendrina Wattimury, for their technical contributions. We also thank Noortje Heidinga and Soraya Smit for their contributions to the in vivo mouse studies. We further acknowledge former colleagues Lieke Paerels, Lukas K. Schwarz, Diewertje Bink, and Elena de Miguel for their work and scientific input on histology and imaging. Special thanks goes to members of the Research and Preclinical Development team at uniQure biopharma, especially Nick Pearson, as well as former colleagues Astrid Vallès, Melvin Evers, and Liesbeth Heijink for their valuable discussions.

We thank Štefan Juhás and Zdeňka Ellederová of the Pigmod Center (Institute of Animal Physiology and Genetics CAS, Czech Republic) for conducting the in vivo minipig studies. We also acknowledge Brian Nguyen and LifeCanvas for the AAV biodistribution analysis in the murine model. Medical writing and editorial support were provided by Julia Jenkins of GK Pharmacomm Ltd. and funded by uniQure biopharma B.V., in accordance with Good Publication Practice (GPP3) guidelines (http://www.ismpp.org/gpp3). The graphical abstract was created with BioRender.com.

Author contributions

S.N.K., E.A.S., M.B., and A.k. conceived and designed the experiments. M.L.v.d.B., R.v.W., G.S., M.W., T.M.H., R.P., and A.k. gave scientific input. Experiments were performed by S.N.K., E.A.S., M.B., C.v.R., G.B., M.W., S.J.H.W., and C.W.H., and data were analyzed by S.N.K., R.D.V.S.M., S.A.B., R.v.W., M.W., and A.k. Preparation of reagents, materials, and analytical tools was performed by S.N.K., R.D.V.S.M., S.A.B., and R.v.W. The manuscript was prepared and written by S.N.K. and A.k. with input from all authors. All authors provided critical revisions of the manuscript, provided final approval prior to submission, and agreed to be accountable for the work.

Declaration of interests

S.N.K. and A.k. are employees of uniQure, hold uniQure stock, and are named inventors on related patent applications. E.A.S., M.B., R.D.V.S.M., M.L.v.d.B., S.A.B., C.v.R., R.v.W., G.S., G.B., M.W., S.J.H.W., W.H., T.M.H., and R.P. are also employees of uniQure and hold uniQure stock.