Global CNS correction in a large brain model of human alpha-mannosidosis by intravascular gene therapy

See Rahim and Gissen (doi:10.1093/brain/awaa189) for a scientific commentary on this article.

Intravascular delivery of adeno-associated virus into the brain has proven limited in large animals with gyrencephalic brains. Yoon et al. show that intracarotid injection of an AAVhu.32 vector – as opposed to the more commonly used AAV9 – corrects global brain pathology in the alpha-mannosidosis cat, extending lifespan.

Abstract

Intravascular injection of certain adeno-associated virus vector serotypes can cross the blood–brain barrier to deliver a gene into the CNS. However, gene distribution has been much more limited within the brains of large animals compared to rodents, rendering this approach suboptimal for treatment of the global brain lesions present in most human neurogenetic diseases. The most commonly used serotype in animal and human studies is 9, which also has the property of being transported via axonal pathways to distal neurons. A small number of other serotypes share this property, three of which were tested intravenously in mice compared to 9. Serotype hu.11 transduced fewer cells in the brain than 9, rh8 was similar to 9, but hu.32 mediated substantially greater transduction than the others throughout the mouse brain. To evaluate the potential for therapeutic application of the hu.32 serotype in a gyrencephalic brain of larger mammals, a hu.32 vector expressing the green fluorescent protein reporter gene was evaluated in the cat. Transduction was widely distributed in the cat brain, including in the cerebral cortex, an important target since mental retardation is an important component of many of the human neurogenetic diseases. The therapeutic potential of a hu.32 serotype vector was evaluated in the cat homologue of the human lysosomal storage disease alpha-mannosidosis, which has globally distributed lysosomal storage lesions in the brain. Treated alpha-mannosidosis cats had reduced severity of neurological signs and extended life spans compared to untreated cats. The extent of therapy was dose dependent and intra-arterial injection was more effective than intravenous delivery. Pre-mortem, non-invasive magnetic resonance spectroscopy and diffusion tensor imaging detected differences between the low and high doses, and showed normalization of grey and white matter imaging parameters at the higher dose. The imaging analysis was corroborated by post-mortem histological analysis, which showed reversal of histopathology throughout the brain with the high dose, intra-arterial treatment. The hu.32 serotype would appear to provide a significant advantage for effective treatment of the gyrencephalic brain by systemic adeno-associated virus delivery in human neurological diseases with widespread brain lesions.

See Rahim and Gissen (doi:10.1093/brain/awaa189) for a scientific commentary on this article.

Introduction

A large number of single gene disorders affect the CNS, most of which are caused by deficiencies of specific proteins in metabolic pathways (Scriver et al., 2001; Pierson and Wolfe, 2005). Most neurogenetic diseases produce lesions throughout the brain due to the metabolic processes being shared by all cells or by all cells of a specific type (Simonato et al., 2013). Somatic gene transfer has the potential to permanently correct the underlying metabolic deficiency by transferring a normal copy of a defective gene into a patient’s own cells. However, in the brain, the diseased cells require widespread gene delivery to correct the globally distributed lesions (Castle et al., 2016). Serotype 9 of adeno-associated virus (AAV9) has been widely used for experimental treatment of CNS diseases due to its ability to cross the blood–brain barrier after systemic intravascular injection. In adult mice, AAV9 transduction occurs throughout the brain and this property has also been shown with serotypes rh.8, rh.10, AS, PHP.B, and Anc80L65 (Duque et al., 2009; Foust et al., 2009; Gray et al., 2011; Yang et al., 2014; Choudhury et al., 2016; Deverman et al., 2016; Hudry et al., 2018). This strategy provides a potential means of therapeutic gene delivery to the CNS by a single non-invasive injection and has shown efficacy in mouse models of CNS disorders (Foust et al., 2010; Valori et al., 2010; Dominguez et al., 2011; Fu et al., 2011; Ruzo et al., 2012; Ahmed et al., 2013).

In contrast, the distribution AAV in the CNS of large mammals is much more restricted after systemic AAV injection. Gene transduction of cells is concentrated in the lower brain and spinal cord and relatively few cells are transduced in the forebrain (Duque et al., 2009; Bevan et al., 2011; Gray et al., 2011; Samaranch et al., 2012; Yang et al., 2014; Gurda et al., 2016; Mendell et al., 2017). Although this has made AAV9 useful for translational gene therapy in diseases affecting the spinal cord and motor neurons (Foust et al., 2010; Valori et al., 2010; Bevan et al., 2011; Dominguez et al., 2011; Mendell et al., 2017), it is inadequate for treating the whole brain (Gurda et al., 2016). The cerebral cortex of large mammals is much larger and more complex than in rodents (Duque et al., 2009; Bevan et al., 2011; Gray et al., 2011; Samaranch et al., 2012; Yang et al., 2014; Gurda et al., 2016; Matsuzaki et al., 2018) due to the gyrencephalic cortical structure compared to the lissencephalic rodent brain. The cerebrum is extensively affected in many neurogenetic diseases, most of which are characterized by severe intellectual disability. This problem was illustrated recently when a synthetic serotype that was selected in vivo for extensive transduction in the mouse brain (PHP-B) (Deverman et al., 2016) showed very little transduction in monkey brains (Hordeaux et al., 2018; Matsuzaki et al., 2018).

When AAV9 is injected directly into the brain parenchyma, the vector virion and genome can be transported to distal sites via axonal pathways (Cearley and Wolfe, 2006, 2007). This property is shared by a limited number of other serotypes (Cearley et al., 2008), suggesting they may also be able to cross the blood–brain barrier after intravascular injection. In this study, we evaluated the transduction characteristics of an AAV belonging to the same clade as AAV9, AAVhu.32. Intravenous injection into adult mice showed that AAVhu.32 resulted in significantly higher transduction in the brain at the same dose than the other transportable serotypes. To test the AAVhu.32-GFP vector in a large brain mammal, it was injected into normal cats, which resulted in widespread distribution of vector genomes in many brain regions including the cerebral cortex. The therapeutic efficacy of AAVhu.32 for global brain disease was evaluated in the cat model of alpha-mannosidosis (AMD), which has prominent lysosomal storage lesions in nearly all cells of the brain. Systemic delivery of AAVhu.32 expressing feline alpha-mannosidase (MANB) was compared by arterial versus venous delivery and by dose. High dose delivery via the carotid artery resulted in complete histological correction of the brain, improved clinical measures, and increased lifespan.

Global transduction throughout the gyrencephalic cerebral cortex in a large animal species indicates that AAVhu.32 may mediate widespread brain transduction in humans. AAVhu.32 may thus be useful for gene delivery in many brain diseases that produce disseminated lesions, particularly in the cerebral cortex, such as gene replacement for inherited neurogenetic diseases as well as for acquired disorders where an appropriate molecular modification strategy is available.

Materials and methods

Plasmid and AAV production

The gene transfer cis-plasmid containing enhanced green fluorescent protein (eGFP) or feline alpha-mannosidase (fMANB) gene were expressed from the minimal human GUSB promoter in the pZac vector that included an SV40 splice donor/acceptor signal and the bovine growth hormone polyadenylation signal, as described (Cearley and Wolfe, 2006, 2007; Cearley et al., 2008; Husain et al., 2009). Recombinant AAVrh.8, AAVhu.32, AAVhu.11, and AAV9 vectors (Gao et al., 2004; Cearley et al., 2008) were packaged by the University of Pennsylvania Vector Core by triple transfection of HEK293 cells with the AAV cis-plasmid, AAV trans-plasmid containing AAV rep and cap genes, and adenovirus helper plasmid. Vectors were purified using iodixanol gradient ultracentrifugation, and the titres were determined by real-time PCR (Lock et al., 2010).

Animals and vector injections

All animal care and procedures were in accordance with the Institutional Animal Care and Use Committee at the Children’s Hospital of Philadelphia (mice) and the University of Pennsylvania (cats). Normal adult BALB/c mice were injected into the tail vein with 200 μl of vector at 2.9 × 10¹² viral genomes (vg) total. Normal and AMD affected cats were produced in the breeding colony of the University of Pennsylvania School of Veterinary Medicine by carrier-to-carrier breeding. Intracarotid injections of cats were performed under anaesthesia using a cut-down procedure to access the common carotid artery. The injected doses in cats were 6 × 10¹³ vg/kg for the GFP vector and 2.9 × 10¹³ (low dose) and 4.8 × 10¹³ (high dose) vg/kg for the fMANB vectors. The surgeon, vivarium staff, operators of magnetic resonance acquisition and analyses, and acquisition and analysis of clinical testing were all blind to the dose received.

MANB enzymatic activity assay

MANB enzymatic activity was measured in serum, CSF or brain tissue as described (Vite et al., 2001, 2005; Yoon et al., 2016). Tissues were homogenized in saline containing 0.2% Triton™ X-100, and MANB activity was determined using the substrate 4-methylumbelliferyl α-d-mannopyranoside in 0.1 M sodium citrate buffer (pH 4.4) at 37°C for 1 h. Protein was quantified using bicinchoninic acid (BCA) assay.

Tissue collection and preparation

Mice were euthanized 4 weeks post-injection and transcardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA). Normal cats injected with the GFP vector were euthanized 6 weeks post-injection using an overdose of intravenous barbiturates and transcardially perfused with 0.9% cold saline. Tissues were drop-fixed in 4% PFA for 48 h. The diseased and treated cats were allowed to live as long as possible and were euthanized when humane end point criteria were reached according to the IACUC protocol. Euthanasia allowed these brains to be perfused with PBS and immediately transferred to PFA in order to preserve brain morphology.

For cryosectioning, brains were cryoprotected in 30% sucrose, embedded in O.C.T. solution (Sakura) and cryosectioned at 20 µm (Leica Microsystems). For immunohistochemical staining, tissues were embedded in 2% agarose and sectioned coronally at 50 μm on a vibratome (Leica VT1000S, Leica). For histology, tissues were paraffin embedded and stained with haematoxylin and eosin or Luxol fast blue.

For serum collection, whole blood was allowed to clot for 30 min at room temperature then centrifuged at 1000g for 15 min. The supernatant was then aspirated and stored at −80°C. CSF was collected using a 22-gauge spinal needle from the cerebello-medullary cistern and stored at −80°C.

Immunohistochemistry

GFP-positive cells were labelled and phenotyped using standard immunohistochemistry. Free-floating sections were permeabilized and immunoblocked for 30 min in 4% goat or donkey serum in PBS-T (PBS containing 0.3% Triton™ X-100). The sections were then incubated overnight at 4°C with the following primary antibodies: rabbit anti-GFP (1:1000, A11122, Molecular Probes), mouse anti-NeuN (1:500, MAB377, Millipore), chicken anti-GFAP (1:1000, AB5541, Millipore) and mouse anti-APC (1:100, OP80, Millipore). After three washes in PBS-T, sections were incubated with the appropriate fluorescently labelled secondary antibodies (1:250; Alexa 488 and Alexa 594, Molecular Probes) in PBS-T for 45 min. After removal of the secondary antibodies and further washes in PBS-T, the sections were mounted onto glass slides and coverslipped with VECTASHIELD Mounting Medium (Vector Laboratories).

For DAB immunohistochemistry, blocking and primary antibody incubations were done as described above. Sections were washed in PBS-T and incubated with the appropriate biotinylated secondary antibodies (goat anti-rabbit, anti-mouse or anti-chicken, 1:250, Vector Laboratories) for 45 min followed by PBS-T washes. The antibody binding was visualized using VECTASTAIN Elite ABC reagent and 3,3′-diaminobenzidine (DAB) substrate kit for peroxidase (Vector Laboratories). Sections were then mounted onto glass slides, dehydrated and mounted in Cytoseal™ 60 mounting medium (Richard Allen Scientific) with glass coverslips.

Images were visualized using a Leica AF6000 LX microscope (Leica) and acquired using a DFC 360FX or DFC 425 digital camera (Leica). GFP expressing cells were quantified in mouse brain hemisections at every 1–1.5 mm region. Images were converted to greyscale and the identical threshold was applied. The number of cells in the sections over the set threshold was counted by particle analysis using ImageJ software (NIH, Bethesda, MD). In the cat the percentage GFP+ cells was determined by DAB+ cells/haematoxylin and eosin-positive nuclei/mm². The investigator was blinded to the experimental groups when assessing outcomes.

Real time polymerase chain reaction

Quantitative real time PCR was used to determine the viral genome copies present in the brain. Genomic DNA was extracted from two to three sections of each transverse brain block and the vector genome copies were quantified separately at each transverse level. Copies of fMANB vector genome were quantified using PowerUp™ SYBR Green Master Mix (Thermo Fisher). Triplicate samples derived from each DNA pool were used for quantification. The primer sequences were as follows: forward: 5′-GGG AGG AGC ATG GCT AGA CA-3′, reverse: 5′-AAC AAC AGA TGG CTG GCA ACT-3′.

Statistical analysis

Unpaired two-tailed Student’s t-test and one-way ANOVA were used, where applicable, to determine mean differences between groups. Data are reported as means ± SEM unless otherwise stated.

MRI

All cats underwent MRI on a 3T Tim Trio whole body magnetic resonance scanner (Siemens) equipped with a single-channel 11-cm internal diameter transmit-receive birdcage coil (M2M). The anatomical imaging protocol included a 3-plane scout localizer to determine orientation and position of the brain. Additionally, axial T₂-weighted images (repetition time/echo time = 2500/60 ms); section thickness = 2 mm; number of slices = 20; number of excitations = 1; and axial T₁-weighted images (repetition time/echo time = 650/27 ms) were acquired. Physiological monitoring including pulse oximetry and vital signs (oxygen saturation and heart rate) were recorded before and during the entire scanning period.

Proton magnetic resonance spectroscopic imaging

Single slice 2D multivoxel proton magnetic resonance spectroscopic imaging (¹H-MRS) was performed using a spin echo (point-resolved spectroscopy) sequence with water suppression using a chemical shift selective saturation (CHESS) sequence. Sequence parameters included: repetition time/echo time = 2500/30 ms, number of excitations = 16, field of view = 55 × 55 mm², slice thickness = 7 mm, bandwidth = 1500 Hz, matrix size = 16 × 16. The typical voxel size was 3.4 × 3.4 × 7 mm³. The volume of interest was selected to include the different cortical and thalamic regions, avoiding the scalp. Four outer volume saturation slabs (30-mm thick) were placed outside the volume of interest to suppress Lip signals from the scalp. The dataset was acquired using elliptical k-space sampling with weighted phase encoding to reduce the acquisition time. Acquisition time for ¹H-MRS sequence was 6:22 min. Manual shimming was performed to achieve an optimal value of full-width half-maximum of the water signal. In general, a shimming of <20 Hz was achieved on the magnitude signal of the water resonance. Water unsuppressed ¹H-MRS spectra were also acquired to use the water signal for computing relative metabolite ratios.

Diffusion tensor imaging

Diffusion tensor imaging (DTI) data were acquired using 30 non-collinear/non-coplanar directions with a single-shot spin-echo, echo-planar read-out sequence. The sequence parameters were as follows: repetition time/echo time = 3000/75 ms, number of excitations = 8, field of view = 80 × 80 mm², matrix size = 64 × 64, in-plane resolution = 1.25 × 1.25 mm²; slice thickness = 2 mm; b = 0, 1000 s/mm², number of slices = 20 covering the whole brain.

Data analysis

The ¹H-MRS data were processed off-line on a Leonardo workstation using the Syngo software (Siemens). For each spectrum, the acquired ¹H-MRS signal (free induction decay) was zero-filled (2048 datapoints), smoothed (Hanning filter, width 200 ms), and Fourier-transformed. This was followed by phase (zero and first order polynomial) and baseline correction for optimal linear frequency dependence. Peak areas of oligosaccharide (3.4–3.9 ppm) and unsuppressed water (4.8 ppm) resonances were determined from each voxel encompassing the bilateral cortical and thalami regions. The fitting of the individual peak can be optimized by adjusting the chemical shift, amplitude, and line-width interactively. Peak areas of oligosaccharide were normalized with respect to unsuppressed water signal for each voxel. The metabolite ratios for oligosaccharide/water from 5 to 10 voxels that encompassed cortex and four to six voxels that encompassed thalamus were computed. The mean and standard deviation (SD) values of metabolite ratios of oligosaccharide/water were reported from each region.

DTI data were processed and analysed using DTI studio software (Johns Hopkins School of Medicine, www.mristudio.org). Pixel-wise fractional anisotropy (FA) maps were computed. FA colour maps were used to draw the regions of interest from white matter regions (corpus callosum, external and internal capsule). These regions of interest were translated on FA maps to compute FA values since FA was the only in vivo DTI parameter at 3 T that statistically distinguished normal from untreated AMD brains (Kumar et al., 2016).

Statistical analysis

From ¹H MRSI data, metabolite ratios of oligosaccharide/water from cortical and thalamic regions were compared across different groups of animals (normal, untreated AMD, high dose treatment, and low dose treatment) by one-way ANOVA with a post hoc Bonferroni test. Similarly, FA values as obtained from DTI data from corpus callosum, external capsule and internal capsule were compared across different groups of animals. All data analyses were performed using a statistical tool (SPSS for Windows, version 18.0; SPSS Inc., Chicago, III, USA).

Data availability

The data generated or analysed during the current study are available from the corresponding author on reasonable request.

Results

Neuron-transportable AAV serotypes mediate widespread brain transduction after systemic injection of adult mice

Four of the AAV serotypes that are capable of neuronal transport (AAVs 9, hu.11, rh.8, pi.2 and hu.32) (Cearley et al., 2008) were compared for distribution of transduction in the brain after intravenous injection into adult mice. AAVhu.32 resulted in a significantly greater number of transduced cells, which were distributed throughout the brain from the olfactory bulb to the rostral end of cerebellum, predominantly in the grey matter (Fig. 1). Hu.32 also mediated the highest transduction in the cerebellum (Supplementary Fig. 1A). AAVrh.8 resulted in similar distribution and density of transduction as AAV9, while AAVhu.11 transduced very few cells. Serotype pi.2 could not be tested due to inability to generate a sufficiently high titre vector. The transduced cells were exclusively neurons by morphology and dual immunofluorescent staining with antibodies against GFP and the neuronal marker, NeuN (Supplementary Fig. 1B).

Figure 1.

Comparison of mouse brain transduction following intravenous delivery of transportable AAV vectors. (A) Intravenous injection of 2.9 × 10¹² vg total (1.4 × 10¹⁴ vg/kg) of AAV9, hu.11, rh.8 and hu.32 expressing eGFP in adult mice resulted in GFP expression throughout the brain 4 weeks post-injection (n = 3 mice for each group). Scale bar = 500 μm. (B) The amount of transduction was quantified by counting the number of GFP-positive objects throughout the brain in sections at distances from Bregma as shown. Horizontal lines represent means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

Pattern of CNS transduction by systemic delivery of AAVhu.32 in a large brain mammal

Substantial differences in the distribution of transduced cells in the CNS after systemic injection of AAV vectors have been observed between rodents and larger mammals, with most of the transduction in large animals occurring in the lower brain and spinal cord, and with relatively small amounts present in the cerebral cortex (Duque et al., 2009; Bevan et al., 2011; Gray et al., 2011; Samaranch et al., 2012; Yang et al., 2014; Gurda et al., 2016; Matsuzaki et al., 2018). However, in most human neurodegenerative diseases the cerebral cortex has extensive pathology. Large mammals have a significantly greater proportion of total brain volume in the cerebral cortex due to the gyrencephalic structure compared to the lissencephalic structure of rodent brains (Simonato et al., 2013). The distribution of AAVhu.32 transduction was analysed in normal cats using the eGFP reporter gene driven by the pan-cellular GUSB promoter (Husain et al., 2009). The vector was injected into the carotid artery to allow it to pass through the brain before circulating to the rest of the body in order to maximize the amount of virus available in the target organ. All of the cats recovered well after the procedure and were euthanized 6–8 weeks later.

Vector transduction was analysed throughout the brain by immunohistochemistry for GFP in coronal sections taken at intervals along the rostral-caudal axis. In the cats, most of the AAVhu.32 transduced cells had neuronal morphologies and were found in the grey matter (Fig. 2), with smaller numbers in white matter regions and the spinal cord. The cerebral cortex, caudate nucleus, putamen, hippocampus and midbrain contained the largest density of positive cells (Fig 2B). The density of transduced cells was variable in different areas of the brain, ranging between approximately 0.9% and 4.8%. Dual immunofluorescent staining for GFP and NeuN confirmed the neuronal transduction (Fig. 3). There was little to no co-localization between GFP and either the astrocyte marker (GFAP) or oligodendrocyte (APC) marker (Supplementary Fig. 2).

Figure 2.

Intracarotid injection of AAVhu.32 in cats results in widespread transduction in the brain. Three 6-week-old cats were injected with 6 × 10¹³ vg/kg of AAVhu.32-eGFP into the carotid artery. Vector transduction was analysed by immunohistochemistry for eGFP at 12 weeks of age. (A) Locations of the six brain sections analysed in each cat. The numbers indicate the position of the coronal sections shown in B. (C) Negative control brain section with no primary antibody. (D) Image from spinal cord; right panel is magnification of the square area in the left panel. Scale bars = 500 μm (B); 600 μm (D left panel); 60 μm (D right panel).

Figure 3.

Neuronal transduction in the cat brain by carotid injection of AAVhu.32. The neuronal phenotype of the transduced cells in the brain was verified by dual immunofluorescent staining with antibodies against GFP and the neuron marker NeuN in the striatum, cortex and hippocampus. Images in the right-hand columns of GFP and merge are magnifications of the boxed area in the adjacent images on the left. Scale bars = 100 μm (left column); 50 μm (right column).

Clinical improvement in alpha-mannosidosis cats after intracarotid delivery of AAVhu.32

To evaluate the clinical effect of the widespread delivery, AMD affected cats were injected with AAVhu.32-fMANB into the carotid artery at 4–6 weeks of age. One group received the same dose of the therapeutic vector as was used for eGFP (2.9 × 10¹³ vg/kg) (low dose) and another group was injected with a 60% higher dose (4.8–5.0 × 10¹³ vg/kg) (high dose). Serum and CSF samples were collected before and after injection and assayed for MANB enzymatic activity in individual animals (Fig. 4A and B). The mean serum level of MANB activity in the treated cats was above the background level of untreated AMD cats by 3 months post-treatment, with the high dose group achieving a significant level more quickly (Fig. 4A) (the probability could not be determined at >6 months as untreated cats do not live that long). Although CSF levels were above background in the high dose group at all time points, the difference only reached P < 0.05 in the 1–2 months post-injection period (Fig. 4B).

Figure 4.

Clinical improvement in AMD cats following intracarotid injection of AAVhu.32. AMD cats were treated at 4–6 weeks of age with a single intracarotid injection of AAVhu.32 expressing fMANB at 2.9 × 10¹³ vg/kg (low dose, n = 3) or 4.8-5.0 × 10¹³ vg/kg (high dose, n = 3). Serum and CSF samples were collected before and after injection and assayed for MANB enzymatic activity in individual cats. (A) Serum MANB activity in the high dose group was elevated above untreated cats from 1 to 2 months and the low dose group beginning from 3 to 4 months post-injection and they remained elevated until the end point. (B) CSF MANB activity was elevated in the high dose treated animals by 1–2 months post-injection and remained elevated above untreated cats until the end point but declined over time. (C) All of the treated AMD cats (n = 6) lived longer than any of the untreated cats (n = 18) (P < 0.0001) and the high-dose group lived significantly longer than the low-dose group. (D) Monthly neurological examinations showed high dose treated cats had a delayed onset of whole-body tremor but not the onset of ataxia. At 18 weeks of age, both the whole-body tremor and truncal ataxia were less severe for both treatment groups compared to untreated cats. Values at each time point represent means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Untreated AMD affected cats show progressive worsening of neurological signs including whole body tremor and truncal ataxia starting at 5 weeks of age (Vite et al., 2001, 2005; Yoon et al., 2016) and were euthanized between 8 and 27 weeks of age due to their deteriorating condition (mean lifespan 17.1 ± 1.3) (Fig. 4C). In contrast, all of the carotid-treated cats lived longer than untreated AMD cats and the high dose group lived significantly longer than the low dose group (Fig. 4C). The low dose cats reached their end stage between 31 and 40 weeks of age (mean 36.0 ± 2.6), and the high dose cats between 56 and 67 weeks of age (mean 60.3 ± 3.0) (Fig. 4C). In comparison, the intravenously treated cats at the low dose only lived 18.2 ± 5.7 weeks (P = 0.76 compared to untreated) and the high dose group lived 36.8 ± 9.3 weeks (P < 0.01), similar to the low dose carotid treated group.

The high dose carotid treated group had a delayed onset of whole body tremor but the onset of ataxia was not delayed (Fig. 4D). By 18 weeks of age, however, the severities of whole body tremor and truncal ataxia were less in both treatment groups compared to untreated AMD cats (Fig. 4D).

Non-invasive imaging detects dose differences and predicts post-mortem findings

MRS can measure the abnormally accumulated mannose-rich oligosaccharides in grey matter regions in the living AMD cat brain (Kumar et al., 2016), and in vivo changes in white matter areas can be assessed by DTI and other magnetic resonance analyses (Vite et al., 2005, 2008). Our previous studies provided baseline measurements of age-matched AMD and normal cats against which the effect of treatment for the brain was assessed. In MRS, there was a significant reduction of the major abnormal mannose-containing oligosaccharide peak after high dose treatment but only partial reduction with the low dose (Fig. 5A), which corresponded to the differences in histology (Fig. 5B).

Figure 5.

Non-invasive MRS imaging show AAV dose-dependent correction in AMD cats. AMD cats were treated at 4–6 weeks of age with a single intra-carotid injection of AAVhu.32 expressing fMANB at 2.9 × 10¹³ vg/kg (low dose, n = 3) or 4.8 × 10¹³ vg/kg (high dose, n = 3). (A) ¹H-MRS at 3 T from thalamus shows correction by AAVhu.32-fMANB at the high dose. Large arrows point to the major peak of mannose-rich oligosaccharides (MS); small arrows to the minor peak (Magnitsky et al., 2010). (B) Representative haematoxylin and eosin stained brain sections of normal, untreated, and high and low dose of AAVhu.32-fMANB-treated cats. Storage lesions present in the cerebral cortex of untreated animals are reduced in AAVhu.32-fMANB treated animals. While the low dose treatment produced partial storage reduction, the high dose treatment resulted in complete correction of the storage lesions. Age at analysis: untreated 19 weeks (end of life), normal 38 weeks, low dose 37 weeks and high dose 56 weeks.

Quantitative analysis of normal, untreated, and both high and low dose treated cats by MRS and DTI indicate that the high dose treated brains were normalized (Table 1), with P-values between 0.0001 and <0.05. The quantitative changes in MRS and DTI in the low dose group showed partial reductions in MRS and DTI (P-values 0.08–0.18) consistent with the partial histological correction. The MRS data are specific for the mannose-containing oligosaccharides that accumulate in this disease (Magnitsky et al., 2010). The DTI data show correction of the leukodystrophy (Vite et al., 2001, 2005, 2008), which was corroborated by the improvement in white matter areas by Luxol fast blue staining (Supplementary Fig. 3A) and DTI images (Supplementary Fig. 3B).

Table 1.

MRS and DTI from AAV-treated cats at two different doses

Genotype	Vector	Titre	Age weeks	n	MRS Ms/H2O ratio (×10−3)		DTI Fractional anisotropy
					Cerebral cortex	Thalamus	External capsule	Internal capsule	Corpus callosum
Normal	None	NA	–	3	0.14 (0.06)	0.13 (0.05)	0.66 (0.05)	0.68 (0.05)	0.29 (0.04)
AMD	None	NA	–	5	1.96 (0.34)	2.11 (0.26)	0.44 (0.04)	0.46 (0.05)	0.21 (0.03)
Treated high-dose	AAVhu.32. fMANB	4.8 × 10−13	16	4	0.15* (0.02) P < 0.0001	0.13* (0.02) P < 0.0001	0.57* (0.03) P = 0.01	0.56* (0.03) P = 0.01	0.26* (0.02) P = 0.002
			30–34	3	0.17* (0.06) P < 0.0001	0.16* (0.04) P < 0.0001	0.55* (0.06) P = 0.03	0.55* (0.07) P = 0.04	0.27* (0.03) P = 0.003
			41–54	3	0.18* (0.06) P < 0.0001	0.19* (0.05) P < 0.0001	0.56* (0.05) P = 0.03	0.56* (0.08) P = 0.04	0.27* (0.05) P = 0.004
Treated low-dose	AAVhu.32. fMANB	2.9 × 10−13	16–18	3	1.57 (0.24) P = 0.09	1.83 (0.09) P = 0.12	0.48 (0.05) P = 0.12	0.50 (0.03) P = 0.08	0.22 (0.02) P = 0.27
			23	3	1.62 (0.32) P = 0.11	1.88 (0.15) P = 0.14	0.49 (0.04) P = 0.11	0.52 (0.05) P = 0.09	0.23 (0.04) P = 0.29
			37	1	1.60	1.91	0.46	0.49	0.23

MRS is expressed as ratios of areas under the curve (integration) of metabolites to water for the oligosaccharide peaks (±SD) as shown in Magnitsky et al. (2010). The DTI measure used is fractional anisotropy (±SD) as shown in Vite et al. (2008) and Kumar et al. (2016). NA = not applicable.

^* T-test between untreated and treated groups P ≤ 0.05.

MANB enzymatic activity in treated alpha-mannosidosis cat brains and peripheral organs

MANB enzymatic activity was measured in post-mortem coronal brain sections taken at intervals along the rostral-caudal axis as shown in Fig. 2A. MANB activities above untreated brain sections were detected throughout the brains of all the intracarotid injected AMD cats (Fig. 6A), with the high dose group having ∼6–11% of normal MANB activity in different slices of brain. MANB enzymatic activity was also detected in peripheral organs of treated AMD cats, with higher levels of activity in the liver, kidney, heart and spleen of the high dose treated cats compared to untreated cats (Fig. 6B). Vector genome copy numbers per diploid genome for the high dose group were higher than the low dose group in all brain regions, despite considerable variance in individual samples (Supplementary Fig. 4). Intravenous injections of the same doses did not result in significant MANB activity in the brain above background and was low in the peripheral organs (Supplementary Fig. 5).

Figure 6.

MANB enzymatic activity in treated AMD cat brain and peripheral organs. (A) MANB enzymatic activity in treated and untreated brains (n = 3 for high dose, n = 3 for low dose, n = 4 for untreated) was measured in coronal sections taken at intervals along the rostral-caudal axis (as shown in Fig. 2A). The mean activity was expressed as a percentage of normal. The transverse slices along the rostral-caudal axis show the wide spatial distribution within the brains of animals treated with high dose AAVhu.32-fMANB. (B) A significant increase in MANB enzymatic activity was measured in the liver, kidney, heart and spleen of high dose treated AMD cats. Horizontal lines represent means + SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

Correction of lysosomal storage lesions in treated alpha-mannosidosis cat brains

In the untreated brain, cytoplasmic vacuolation and distension of cells, caused by lysosomal storage of the undegraded oligosaccharides, are present throughout the brain. Although there were variations between individual cats, haematoxylin and eosin stained sections of intracarotid treated cats showed resolution of lysosomal storage vacuoles throughout the brain (Fig. 7). In the low dose treated brains, the histological reduction was incomplete, whereas in the high dose group the histology was indistinguishable from the normal brain samples. The differences between high and low dose responses were in complete concordance with the quantitative in vivo measurements by MRS in grey matter and DTI in white matter regions of the brain. The clinical assessments, the magnetic resonance analyses, and the histological evaluation were performed blind to knowledge of the dose. The intravenously injected animals showed only partial storage correction in the high dose group and minimal reduction in the low dose group (Supplementary Fig. 6).

Figure 7.

Correction of lysosomal storage lesions in brain tissue of treated AMD cats. Representative haematoxylin and eosin stained brain sections of normal, untreated, and high and low dose AAVhu.32-fMANB-treated cats. Only partial storage reduction was seen in the low dose treated cats, while the high dose treatment group had complete correction of lysosomal storage vacuoles throughout the brain. Scale bar = 60 μm.

Serum chemistry in treated alpha-mannosidosis cats

Clinical chemistry assays were performed on serum samples pre- and post-injection in all AMD cats. The serum chemistry values post-injection were all within or near the control range of 26 age-matched normal cats from our colony (Supplementary Table 1). Thus, there was no indication for any liver, renal or other toxicity from the vector injections in the cats following systemic injection of AAVhu.32-fMANB.

Discussion

Intravascular delivery of AAV into the brain is clinically relevant for a number of diseases affecting the CNS as it would allow gene transfer with a minimally invasive procedure. In mice, AAV serotypes 9, rh.8, rh.10 and Anc80L65, as well as cap proteins selected for increased entry in the brain such as AS and PHP.B, can mediate CNS gene delivery when administered intravascularly (Duque et al., 2009; Foust et al., 2009; Gray et al., 2011; Yang et al., 2014; Choudhury et al., 2016; Deverman et al., 2016; Hudry et al., 2018). However, when large animals are injected systemically with these AAV serotypes, only low amounts of forebrain transduction occur (Duque et al., 2009; Bevan et al., 2011; Gray et al., 2011; Samaranch et al., 2012; Yang et al., 2014; Gurda et al., 2016; Hordeaux et al., 2018; Matsuzaki et al., 2018). The brains of large mammals are much more similar in structure to human brains than rodent brains, having a gyrencephalic cerebral cortex.

In large brain mammals, relatively little gene transfer has been seen with systemic AAV9 or AAVrh.10 in the cerebral cortex, which will be a critical target region in many human brain diseases. Intravenous injections of AAV9 or AAVrh.10 in monkeys, cats and pigs transduce mostly glial cells and the distribution is primarily in the spinal cord and lower motor neurons (Duque et al., 2009; Bevan et al., 2011; Gray et al., 2011; Yang et al., 2014). Although some transduced cells have been found in the cerebral cortex, they are far too few to be effective for correction of the whole brain. In monkeys, intravenous versus intra-arterial injections of AAV9 have been done in a very small number of animals but in the data reported to date, the routes resulted in similar levels of transduction in the brain, and substantially more brain cells are transduced by intra-CSF infusion (Simonato et al., 2013; Gurda et al., 2016; Yoon et al., 2016). In contrast, as shown here, intracarotid injection of AAVhu.32 is capable of transducing cells widely distributed in the forebrain of a large brain model of disseminated neurogenetic disease.

The different patterns of transduction between AAVhu.32 and AAV9 suggest that they target different cellular components. Serotypes hu.32 and 9 are closely related in DNA sequence (Gao et al., 2004), both being assigned to AAV clade F, and have similar transduction patterns when directly injected into adult mouse brains (Cearley et al., 2008). However, after intravascular delivery in the large animals, the predominant distribution of AAVhu.32 transduction shifts from lower brain and spinal cord seen with AAV9 to forebrain structures, and from glial cells to neurons. There are 12 differences in amino acid sequence between serotypes 9 and hu32 and all of them are located at the amino terminal end of the capsid protein, before the first loop (NCBI database; Gao et al., 2004; Agbandje-McKenna and Kleinschmidt, 2011) (Supplementary Fig. 7). In addition, the serotypes selected for this study based on their shared property of axonal transport of vector genomes (9, hu.32, hu.11, rh8) all retain the galactose binding motif of 9 (N470, D271, N272, Y446, W503) (Bell et al., 2012) and none of them have the sequences for HS binding sites of AAV2 that when mutated allow expanded intraparenchymal diffusion (Tordo et al 2018).

The density of transduction in the mouse brains was relatively evenly dispersed (Fig. 1). In contrast in the cat brain, the distribution of the individual transduced cells was more irregular, with some areas having little transduction, some areas having numerous single transduced neurons, and other areas where there are clusters of positive cells (Fig. 2). Nevertheless, this mediated complete correction of the storage lesions. Complete correction can only occur if the normal enzyme reaches all cells in sufficient quantities to restore the metabolism. The distribution of transduced cells throughout the volume of the brain indicates that the total enzyme delivered from all of the transduced cells can reach all of the affected cells. This is due to a combination of lysosomal enzyme distribution mechanisms (Simonato et al., 2013): (i) the classic cross-correction mechanism in which a lysosomal enzyme is secreted from a vector-corrected cell into the extracellular space and is endocytosed by neighbouring cells via the cell surface mannose-6-phosphate receptor; (ii) lysosomal enzymes can be transported to distal sites within axonal pathways (Passini et al., 2002); and (iii) with certain serotypes of AAV, the vector genome itself can be transported to distal sites and then produce enzyme for secretion and further cross-correction around the distal sites (Cearley and Wolfe, 2006). Indeed, the transportability of certain AAV serotypes (Cearley et al., 2008) was the rationale for selecting the vectors we tested. The combination of these effects is that enzyme production occurs from transduced cells distributed in three dimensions throughout the volume of the brain.

The high dose treatment resulted in the highest levels of MANB in brain tissue (Fig. 6) and variable but overall highest levels of MANB activity in CSF and serum (Fig. 4A, B and Supplementary Fig. 8). This resulted in the essentially complete correction of histology, which is congruent with the normalization of the MRS measurements. Consistent with this finding is that very low densities of transduction in a mouse model of another lysosomal storage disease can correct the whole brain when the transduced cells are widely dispersed within the brain parenchyma (Cearley and Wolfe, 2007). The proof that the density of transduction achieved with the high-dose carotid injection is sufficient resides in the complete resolution of the storage lesions and is further supported by the dose responsiveness.

The mechanism of entry into the CNS is not known; however, direct intraparenchymal injection of AAVhu.32-GFP into the mouse brain results in transduction of neurons, astrocytes and oligodendrocytes (Cearley et al., 2008) whereas intravascular injection of AAVhu.32 resulted in predominately neuronal transduction. Since hu.32 was selected for testing due to its ability to be transported in neurons, this suggests that the vector is somehow accessing CNS neurons after crossing the blood–brain barrier.

Another notable aspect of the present study is that a single-stranded AAV genome was used. Most large animal studies have used self-complementary (scAAV) vectors because they have higher transduction efficiency than traditional single-stranded vectors (McCarty et al., 2003; McCarty, 2008; Gray et al., 2011). However, the packaging capacity of the scAAVs is approximately half that of conventional single-stranded (ss) AAVs, limiting their use for therapeutic gene cDNAs that are too large for the scAAV configuration. The MANB cDNA coding region is ∼3.0 kb, which by itself is too large for a scAAV vector. In addition, AAV vectors require a minimum of a promoter, a splice donor-acceptor, and a polyA addition sequence. In the present study, a single stranded AAV genome packaged in the AAVhu.32 cap vector was able to achieve robust, widespread transduction of the brain allowing transfer of the MANB cDNA. The larger packaging capacity of this vector would also enable use of cell-type-specific promoters, which are often large and thus of limited use in the scAAV design. These properties could greatly expand the repertoire of CNS diseases amenable to gene therapy as well as expand the number of experimental uses for gene transfer in the brain.

No adverse clinical effects were observed in any of the animals following intravascular injection of AAVhu.32. Others have reported transient rises in ALT levels, inflammation and immune responses following intravascular or intrathecal GFP injection (Gray et al., 2011; Samaranch et al., 2014). All the GFP-injected animals appeared clinically normal, and serum chemistry revealed no indication of liver, renal, or other toxicity following therapeutic vector injection in the AMD cats. Pre-existing AAV9 neutralizing antibodies have been associated with low CNS transduction after systemic injection in non-human primates (Gray et al., 2011; Samaranch et al., 2012). Although there is no assay to detect anti-hu.32 antibodies, the finding of widespread transduction of the brain in all of the vector-injected animals indicates there were no pre-existing antibodies to AAVhu.32. This is consistent with epidemiological data that shows a low prevalence of neutralizing antibodies against AAV serotypes other than AAV2 (Calcedo et al., 2009; Boutin et al., 2010; Calcedo et al., 2011). Furthermore, the prevalence of anti-AAV antibodies in infants and young children is low (Erles et al., 1999; Calcedo et al., 2011), when it would be most desirable to treat neurogenetic diseases before the pathology is fully developed. The absence of adverse effects in the AAVhu.32 vector-injected animals suggests that intravascular injection of AAVhu.32 may be effective for gene delivery into the brain clinically.

Consistent with the predominant distribution of AAV9 to the lower brain and spinal cord, AAV9 has been effective for treating a motor neuron disease clinically (Mendell et al., 2017). However, it was not effective in changing the brain disease in the only large animal model (dog) of a lysosomal storage disease that has been evaluated by systemic delivery (Gurda et al., 2016). In contrast, systemic AAVhu.32 delivery of the alpha-mannosidase cDNA in AMD cats mediated reversal of storage lesions throughout the brain. The levels of MANB produced in the brain by the high dose intracarotid infusion of AAVhu.32 was greater than by either direct intra-parenchymal injections or CSF infusions of the same vector genome in the AMD cat model (Vite et al., 2005; Yoon et al., 2016). The intra-arterial injection in the present study was more effective clinically at the same dose than intravenous injections. This occurred even though the arterial injections were done post-symptomatically (∼6 weeks; Vite et al., 2001) because they required cut-down surgery and anaesthesia, while the intravenous injections could be done at an earlier pre-symptomatic age since they were by percutaneous injection. In general, early treatments for lysosomal storage diseases are more effective because of the progressive nature of the diseases; however, most human lysosomal storage diseases are not diagnosed until after symptoms appear, thus the intra-arterial treatment at an older age in the AMD cats is a more realistic model for treatment of human patients. The clinical parameters of the treated cats were significantly improved compared to untreated cats. Furthermore, pre-mortem non-invasive MRS and DTI imaging showed dose-responsive resolution of substrate accumulation, which correlated with the extent of histological correction.

The ability of AAVhu.32 to transduce cells throughout the cerebral cortex in a large mammalian gyrencephalic brain indicates that it may be effective for treating human brain diseases since the scale up is by body weight. Many human brain diseases that may be treatable by gene therapy have extensive pathological changes in the cerebral cortex (Simonato et al., 2013). Treatment will depend on having a disease-specific molecular target for modification, but in most lysosomal storage diseases, gene replacement would be used. The AAVhu.32 vector also provides a means to deliver genes into neurons in the cerebrum in higher mammals for experimental manipulations, such as optogenetics or gene editing, without the confounding effects of neurosurgery. The predominant neuronal transduction, the wide distribution within the cerebrum, and the capacity to accommodate larger coding sequences and expression control elements of AAVhu.32 provide a realistic reagent for treatment of the whole brain in human neurogenetic and potentially other CNS diseases where widespread neuronal gene delivery is needed.

Glossary

AAV =

adeno-associated virus

AMD =

alpha-mannosidosis

DTI =

Diffusion tensor imaging

(f)MANB =

(feline) alpha-mannosidase

MRS =

magnetic resonance spectroscopy