Intracerebroventricular Administration of AAV9-PHP.B SYN1-EmGFP Induces Widespread Transgene Expression in the Mouse and Monkey Central Nervous System

Abstract

Viral vectors made from adeno-associated virus (AAV) have emerged as preferred tools in basic and translational neuroscience research to introduce or modify genetic material in cells of interest. The use of viral vectors is particularly attractive in nontransgenic species, such as nonhuman primates. Injection of AAV solutions into the cerebrospinal fluid is an effective method to achieve a broad distribution of a transgene in the central nervous system. In this study, we conducted injections of AAV9-PHP.B, a recently described AAV capsid mutant, in the lateral ventricle of mice and rhesus macaques. To enhance the expression of the transgene (the tag protein emerald green fluorescent protein [EmGFP]), we used a gene promoter that confers high neuron-specific expression of the transgene, the human synapsin 1 (SYN1) promoter. The efficacy of the viral vector was first tested in mice. Our results show that intracerebroventricular injections of AAV9-PHP.B SYN1-EmGFP-woodchuck hepatitis virus posttranscriptional regulatory element resulted in neuronal EmGFP expression throughout the mice and monkey brains. We have provided a thorough characterization of the brain regions expressing EmGFP in both species. EmGFP was observed in neuronal cell bodies over the whole cerebral cortex and in the cerebellum, as well as in some subcortical regions, including the striatum and hippocampus. We also observed densely labeled neuropil in areas known to receive projections from these regions. Double fluorescence studies demonstrated that EmGFP was expressed by several types of neurons throughout the mouse and monkey brain. Our results demonstrate that a single injection in the lateral ventricle is an efficient method to obtain transgene expression in many cortical and subcortical regions, obviating the need of multiple intraparenchymal injections to cover large brain areas. The use of intraventricular injections of AAV9-PHP.B SYN1-EmGFP could provide a powerful approach to transduce widespread areas of the brain and may contribute to further development of methods to genetically target-specific populations of neurons.

INTRODUCTION

In basic and translational neuroscience research, introduction or modifications of genetic material in nontransgenic species is commonly accomplished through viral vectors. Vectors made from adeno-associated virus (AAV) have emerged as the preferential tool for genetic manipulations, given the long-lasting and robust expression of the transgene, low toxicity, and the fact that the genetic material remains episomal.^1–3

Most frequently, the AAV is delivered through direct injection of the viral vector solution in the brain (intraparenchymal).⁴ This method has the advantage that it bypasses the blood–brain barrier (BBB) and that a specific brain region can be targeted by stereotaxic, magnetic resonance imaging (MRI), or electrophysiologic guidance methods. However, intraparenchymal injections of AAV solutions may result in limited diffusion in the brain tissue^1,5 and may be off-target, particularly when injections aim at small brain structures. Although the limited spread of the solution (a few mm in most cases⁶) is likely enough to cover large brain areas in rodents, this represents a significant problem in animals with large-sized brains, such as human and nonhuman primates (NHPs).

An alternative approach to achieve a more extensive distribution of the virus solution in the central nervous system (CNS) is to deliver the AAVs into the cerebrospinal fluid (CSF), by injecting in the cerebral ventricles (intracerebroventricular, ICV), in the cisterna magna (intracisterna magna, ICM), or in the subarachnoid space of the lumbar spinal cord (intrathecal). Previous studies reported broad expression of the transgene throughout the brain and/or spinal cord of large animal models such as NHPs following ICV, ICM, or intrathecal injections of AAVs.^7–16 ICV injections are already used in clinical trials in humans (reviewed in ref.¹⁷).

Recently, novel AAV capsid mutants that more efficiently transduce neurons have been developed using Cre recombinase-dependent selection methods. One of the mutants generated, AAV9-PHP.B, effectively crosses the BBB in adult mice after intravenous injections.^18,19 However, this result was not replicated in NHPs.^20,21 In rhesus macaques, ICV injections of AAV9-PHP.B resulted in widespread transduction throughout the CNS, but the transgene expression was mainly confined to neurons and glia in the cerebral cortex.¹⁹ In this, and previous studies in which AAVs were delivered into the CSF of primates, the expression of the transgene was driven by a ubiquitous promoter (such as cytomegalovirus or chicken beta-actin).¹⁹

One of the main advantages of genetic-based approaches is the ability to target the transgene expression to a specific cell type. For example, the use of a neuronal promoter could be highly advantageous to restrict the expression to neurons over other brain cells. To address this issue, we performed ICV injections of AAV9-PHP.B in two adult rhesus macaques, using a gene promoter that confers high neuron-specific expression of the transgene, the human synapsin 1 (SYN1),^22,23 to drive the expression of emerald green fluorescent protein (EmGFP) in neurons. We first tested the approach in mice to evaluate the efficacy of the viral vector in a small animal model before scaling the approach to NHPs. In both species, we found significant expression of EmGFP throughout the cerebral cortex and in several subcortical regions. Our results suggest that the use of ICV injections of AAV9-PHP.B with a neuronal promoter may be a suitable approach to induce widespread transgene expression in various populations of cortical and subcortical neurons in the primate brain.

MATERIALS AND METHODS

Animals

All animal experiments were carried out in accordance with federal guidelines (NRC, 2011), and the United States Public Health Service Policy on the Humane Care and Use of Laboratory Animals (revised 2015). Three mice (mixed sex C57BL/6J, stock no: 000664; The Jackson Laboratory, Bar Harbor, Maine) and two adult rhesus monkeys (MR322, male and MR325, female; both were 4 years old and 7.6 and 5.36 kg, respectively, at the time of virus injection) from the Yerkes Primate Center colony were used in this study. Before the project assignment, blood samples from the two monkeys were analyzed by the Yerkes Virology core to ensure that they were sero-negative for AAV9. The housing, feeding, and experimental conditions used in these studies followed the guidelines for animal use and welfare set by the National Institutes of Health (National Research Council), and have been approved by University of British Columbia and Emory University Institutional Animal Care and Use Committees (IACUC).

Cloning and virus production

One recombinant AAV was used in this study. The SYN1 promoter sequence was chosen for ubiquitous expression in neurons.^22,24,25 The 469 bp continuous human sequence, plus 8 bp restriction enzyme sites on each end was DNA synthesized (GenScript, Inc., Piscataway, NJ) and cloned by exchanging the promoter from pEMS2113 (version 1)²⁶ with AscI and FseI, to produce plasmid pEMS2155. The final plasmid included: an intron (optimized chimeric; 173 bp; Promega, Madison, MI),²⁷ NotI flanked EmGFP (720 bp)²⁸; AsiSI flanked woodchuck hepatitis virus posttranscriptional regulatory element (WPRE) mut6 (587 bp)²⁹; and SV40 polyA (222 bp) (Promega) sequences.

The plasmid was propagated in Escherichia coli SURE cells (Agilent Technologies, Santa Clara, CA). DNA was prepared by QIAgen Spin MiniPrep Kit (QIAgen, Germantown, MD); plasmid was confirmed free of rearrangements by AhdI digest, inverted terminal repeats were verified by SmaI digest, and cloning sites verified by sequencing. The plasmid was then sent to the University of Pennsylvania Vector Core (Philadelphia, PA) for large-scale DNA amplification using the EndoFree Plasmid Mega Kit (QIAgen, Hilden, Germany). Quality control on the plasmid preparation was carried out by SmaI, PvuII, and SnaBI digests, and confirmed plasmid was packaged into AAV9-PHP.B capsid.^18,30,31 Both the cloning (pEMS2156) and virus genome (pEMS2155) plasmids with the SYN1 gene have been deposited at Addgene (Watertown, MA).

Surgical procedures

Mice

For intraventricular injections in mice, animals were anesthetized with 4% isoflurane with 1% oxygen and then fixed in a stereotaxic apparatus. A single injection each of buprenorphine (0.1 mg/kg s.c.) and bupivacaine (1 mg/kg s.c. infiltrated under the skin at the injection site) were administered for prophylactic analgesia. The hair on the head was shaved and the area cleaned with hibitane and alcohol before making a small incision in the skin to expose the skull. Mice received unilateral injections of virus solution into the lateral ventricle using coordinates of −0.3 mm posterior, +1.1 mm lateral, and −3.5 mm ventral from Bregma.³²

The virus solutions (AAV9-PHP.B SYN1-EmGFP-WPRE, 3.0 × 10¹¹ granule cells [GC]/mouse, in a total volume of 10 μL) were injected through small burred holes in the skull using a 10 μL Hamilton syringe over 2 min, at a rate of 5 μL/min, and the syringe was left in place for an additional 5 min to allow diffusion of the virus solution. The needle was then slowly removed from the brain and the incision sutured. During the surgical procedure and recovery from anesthesia, the body temperature of the mice was monitored and controlled using a heat pad. The animals were monitored every 15 min until they woke up, at which point they were transferred to a housing room. Mice were monitored for body weight and general health twice per day for 72 h postsurgery.

Monkeys

Each monkey underwent an MRI scan on the day of the surgical procedure before being brought to the surgery suite. For this, the animal was tranquilized with ketamine (10 mg/kg, i.m.) and brought to the Yerkes MRI Center where it was intubated and anesthetized with isoflurane (1–3%). The head of the animal was then fixed in a stereotaxic frame and brought to the scanner. During the procedure (∼1 h long), the isoflurane anesthesia was administered and monitored by the MRI anesthesia technician with help of the veterinary staff, as needed. Once the MRI scan was complete, the animal remained anesthetized with isoflurane in the stereotaxic frame and was brought to the preoperative room in the surgery suite. The head of the animal was shaved and the skin overlying the site of the operation was cleaned with betadine and alcohol.

After retracting the skin and exposing the skull, a small hole (3–5 mm in diameter) was drilled in the skull, and a Hamilton microsyringe was lowered in the brain under stereotaxic guidance to the lateral ventricle (+18 mm anterior to the interaural line, 2 mm from midline, and −17.5 mm ventral from the cortical surface). Once the target region was reached, the syringe was left in place for 5 min before the beginning of the injection. A total volume of 3 mL of virus solutions (AAV9-PHP.B SYN1-EmGFP-WPRE) was delivered in the lateral ventricle over a period of 3 min. Monkey MR322 received 3.0 × 10¹³ GC (3.95 × 10¹² GC/kg), whereas monkey MR325 received 1.0 × 10¹³ GC (1.87 × 10¹² GC/kg). The volume of viral solution delivered is similar to what was used in a previous study doing intraventricular infusions of virus solutions.¹³

The virus solution was diluted to the final titer using artificial CSF (CMA Microdialysis, Kista, Sweden). The injection was conducted manually, at an approximate rate of 1 mL/min. The needle was left in situ for 10 min before being slowly withdrawn. Once the injections were completed, the surgical site was cleaned with sterile saline and the skin was sutured. At the end of the procedure, the animals received postsurgical analgesics (buprenorphine or banamine, continued up to 3 days postsurgery) before being brought to their home cage where their behavior and health status were monitored daily by the Yerkes veterinarian staff.

Immunohistochemical procedures

Mice

Brains were collected 28 days postinjection. Animals were injected with 100 μL heparin, anaesthetized by intraperitoneal injection of 0.5 mg/g avertin (2,2,2 tribromoethanol), and then terminally perfused through the ascending aorta with cold 3% paraformaldehyde in 0.1 M phosphate buffer at a flow rate of 5 mL/min for 10 min.

After perfusion, the brains were kept within the skull for 24 h in perfusion buffer. Subsequently, the brains were removed from the skull and kept at 4°C in phosphate-buffered saline (PBS; 0.01 M, pH 7.4) until further processing. Before sectioning, the tissue was equilibrated in 30% sucrose in PBS for 24–48 h at 4°C, after which it was embedded in Optimal Cutting Temperature compound (Thermo Fisher, Waltham, MA), frozen on dry ice, and then cut on a cryostat into 25 μm thick coronal sections. Sections were collected floating into PBS supplemented with 0.01% sodium azide.

Immunofluorescence

After mounting on glass slides (Fisherbrand Superfrost Plus; Thermo Fisher), sections were permeabilized in 0.1% Triton X-100 in PBS, blocked with 5% normal donkey serum, then processed for double immunofluorescence staining according to procedures used in our previous studies. Sections were incubated at 4°C for 12–20 h in cocktail of primary antibodies. Information about the commercial sources and research resource identifier numbers of the various primary antibodies used in this study is given in Table 1. Immunofluorescence detection was achieved by a 1-h incubation at room temperature with secondary antibodies specific to the primary antibodies used (Table 1).

Table 1.

Sources and concentrations of antibodies

Antibody	Vendor	Catalog no.	Antibody registry no.	Studies	Concentration used
Primary
Chicken anti-GFP	Thermo Fisher Scientific (Waltham, MA)	A-11122	AB_221569	Mice	1:500
Rabbit anti-GFP	Thermo Fisher Scientific	A11122	AB_221569	Monkey	1:5000
Mouse anti-tyrosine hydroxylase	Millipore Sigma (St Louis, MO)	MAB318	AB_2201528	Monkeys, mice	1:1000
Rabbit anti-DARPP32	Cell Signaling Technology (Danvers, MA)	2306	AB_823479	Monkeys, mice	1:1000
Rabbit anti-calbindin	Millipore Sigma	AB1778	AB_2068336	Mice	1:500
Rabbit anti-parvalbumin	Millipore Sigma	PC255L	AB_2173906	Mice	1:500
Rat anti-serotonin	Millipore Sigma	MAB352	AB_94865	Monkeys	1:500
Mouse anti-NeuN	Millipore Sigma	MAB377	AB_2298772	Monkeys, mice	1:2000
Mouse anti-GABA	Millipore Sigma	A-2052	AB_477652	Monkeys	1:40,000
Secondary
Donkey anti-rabbit Alexa 594	Thermo Fisher Scientific	A-21207	AB_141637	Mice	1:1000
Donkey anti-mouse Alexa 594	Thermo Fisher Scientific	A-21203	AB_141633	Mice	1:1000
Goat anti-chicken Alexa 488	Thermo Fisher Scientific	A-11039	AB_142924	Mice	1:1000
Goat anti-rabbit Biotinylated	Vector Laboratories (Burlingame, CA)	BA-1000	AB_2313606	Monkey	1:200
Donkey anti-chicken Fluorescein	Jackson Immunoresearch (West Grove, PA)	703-095-155	AB_2340356	Monkey	1:100
Donkey anti-rabbit Fluorescein	Jackson Immunoresearch	711-095-152	AB_2315776	Monkey	1:100
Donkey anti-mouse Rhodamine RedX	Jackson Immunoresearch	715-295-150	AB_2340831	Monkey	1:100
Donkey anti-rabbit Rhodamine RedX	Jackson Immunoresearch	711-295-152	AB_2340613	Monkey	1:100

GFP, green fluorescent protein.

All sections were then incubated for 5 min in 4′,6-diamidino-2-phenylindole (DAPI; 1:10 000; Millipore Sigma, St Louis, MO) at room temperature. Sections were cover-slipped using Depex fluorescence mounting medium (Electron Microscopy Sciences). Images were captured using an Olympus BX61 Fluorescence and Transmittance Wide Field Microscope. The images were compiled using Photoshop CC 2018 software (Adobe Systems, San Jose, CA).

Monkeys

After 29- and 32-day survival (MR322 and MR325, respectively), the animals were deeply anesthetized with pentobarbital (100 mg/kg) and perfusion-fixed with ∼300 mL of cold oxygenated Ringer solution followed by 2.5 L of a mixture of 4% paraformaldehyde and 0.1% glutaraldehyde in phosphate buffer (0.01 M, pH 7.4). The brains were then removed from the skull, blocked in the stereotaxic plane, and cut in 50 μm thick serial sections with a freezing microtome. All sections were collected in PBS before being processed for immunohistochemistry or transferred in an anti-freeze solution and stored at −20°C until further processing.

Immunoperoxidase

One of 10 sections collected throughout the whole brain of each monkey were processed for GFP immunostaining. The tissue was first placed in a sodium borohydride (1% PBS) solution for 20 min. After thorough washes in PBS, sections were put into a preincubation solution (10% normal goat serum, 1% bovine serum albumin [BSA], and 0.3% Triton X-100) for 1 h at room temperature, before being incubated in the GFP antibody solution (Table 1) for 24 h at room temperature. Sections were then washed in PBS and placed for 90 min at room temperature in a secondary antibody solution consisting of 1% normal goat serum, 1% BSA, biotinylated goat anti-rabbit antibodies (Table 1), and 0.3% Triton.

Sections were then washed thoroughly in PBS before being incubated in an avidin–biotinylated complex solution (Vector Laboratories, Burlingame, CA) for 90 min. After this incubation, the tissue was rinsed twice with PBS and once with tris(hydroxymethyl)aminomethane (0.05 M, pH 7.6). Thereafter, the sections were placed in 0.025% 3-3-diaminobenzidine tetrahydrochloride (Millipore Sigma), 0.01 M Imidazole (Fisher Scientific), and 0.006% H₂O₂ for 10 min. The sections were then mounted on slides, covered, and digitized with an Aperio Scanscope CS system (Leica, Buffalo Grove, IL). Control sections were incubated in solutions from which the primary antibody was omitted (Supplementary Fig. S1).

Immunofluorescence

To characterize the phenotype of GFP-immunoreactive neurons in specific regions of the monkey (MR322) brain, double immunofluorescence procedures were used. Equivalent regions of the brain were selected in mice and monkeys. The double-labeling procedures used here are similar to those described in our previous studies.³³ In brief, sections were incubated for 24 h in a mixture of anti-GFP antibodies and each of the primary antibodies listed in Table 1, which was followed by exposure to the appropriate secondary fluorescent antibodies (Table 1). The sections were then mounted with Vectashield (Vector Laboratories) and examined using a confocal laser scanning microscope (Leica DM5500B) equipped with a CCD camera (Orca R2; Hamamatsu).

Analysis of material

Quantification of double-labeling in mouse

The number of cells co-labeled for GFP and a marker of interest was quantified using StereoInvestigator software (MBF Bioscience, Williston, VT). The region of interest (ROI) was outlined at 4 × magnification on two sections (four hemispheres) from each mouse. For the substantia nigra (SN), the ROI was outlined based on tyrosine hydroxylase (TH) immunoreactivity; for the striatum (Str) and cortex, the ROI was delineated using anatomical features visible with the DAPI counterstain. The sampling grid and counting frame sizes were set over the ROI as follows: Str—sampling grid 400 × 400 μm, counting frame 100 × 100 μm; cortex—sampling grid 400 × 400 μm, counting frame 100 × 100 μm; SN—sampling grid 100 × 100 μm, counting frame 50 × 50 μm.

Under 40 × magnification, GFP-positive cells were marked with a digital marker at each sampling site. The filter was subsequently switched to the red channel, and a second marker was placed on cells that were also positive for the marker of interest (DARPP32 in the Str, parvalbumin [PV] in the cortex, TH in the SN). In the cortex, GFP-positive pyramidal neurons were excluded from the analysis as they are known to be PV negative. The percentage of double-labeled cells for each mouse was calculated as the total number of GFP-positive cells that also expressed immunoreactivity for the marker of interest over the total counts of GFP-immunoreactive cells in the different ROIs. Results are reported as mean ± standard deviation (SD).

Estimation of the number of NeuN-positive cells that are also GFP positive in the mouse cortex was performed using a similar methodology, with adjustments to account for the large number of cells to be counted. Instead of two, three coronal sections per mouse were counted, covering portions of the motor, somatosensory, auditory, and entorhinal cortex areas. Sections were chosen at ∼1.8, 2.4, and 3.0 mm caudal from Bregma. The sampling grid was increased to 500 × 500 μm, and the counting frame was set to 50 × 50 μm. The percentage of double-labeled cells for each mouse was calculated as the total number of NeuN-positive cells that were also GFP immunoreactive. Results are reported as mean ± SD.

Immunoperoxidase-stained monkey material

The relative distribution and abundance of GFP-labeled cells and axonal processes were described in eight representative coronal sections distributed across the rostrocaudal extent of the monkey brain. For this, the digitized images were analyzed at 2 × to 5 × , and neuronal cell bodies were counted using the counter tool in Aperio ImageScope (v12.3.2.8013; Leica). Using a graphics editor (CorelDraw 2019 and 2020; Corel Corp., Ottawa, ON, Canada), each 10 neuronal cell bodies were represented as one circle in the corresponding brain region in digital copies of a rhesus monkey brain atlas.³⁴

To represent in a semi-quantitative manner the presence of GFP-labeled neuropil elements (which were interpreted as dendrites and/or putative axonal terminals), images acquired at 0.6 × using ImageScope were first converted to pseudo color images using the lookup tables function in ImageJ. These images were then overlapped to the digital copies of the brain atlas in CorelDraw, and colored areas were manually delineated in CorelDraw. Shades of gray were used to represent areas with intense, moderate, and mild GFP-labeled neuropil.

Quantification of double-immunofluorescence labeling in monkeys

Sections from monkey MR322 were double-immunostained for NeuN (a marker of neurons) and GFP to determine the proportion of total cortical neurons that expressed GFP in a subset of transduced cortical regions (premotor cortex, primary motor cortex, superior parietal, and inferotemporal cortices). Four sections containing these cortical regions were used. The proportion was calculated as number of GFP-positive cells/number of NeuN-positive cells.

Double-immunostained sections from MR322 were also used to determine the chemical phenotype of specific populations of GFP-immunoreactive neurons in various brain regions. In monkeys, the relative prevalence of randomly encountered GFP-containing neurons that expressed immunoreactivity for the marker under study (Table 1) was calculated for the different brain regions examined (caudate nucleus [CD]-DARPP32; cortex-gamma aminobutyric acid [GABA]; SN-TH; and dorsal raphe [DR]-serotonin [5HT]). In most brain regions, GFP-positive cells were randomly chosen, except in the cerebral cortex where nonpyramidal neurons were specifically looked at for their colocalization with GABA. A total of two to four slides per brain region were analyzed in MR322.

From these slides, the total number of GFP-positive neurons examined in each brain region was as follows: 563 in CD, 189 in cortex, 37 in SN, 129 in DR. For each of those brain regions, the relative percentage of GFP-immunostained neurons that displayed immunoreactivity for the neurotransmitter marker under study was calculated. Results are reported as mean ± SD.

RESULTS

In both mice and monkeys, the ICV injections of AAV9-PHP.B SYN1-EmGFP resulted in neuronal GFP expression throughout the whole brain. In mice, the GFP labeling was localized with immunofluorescence, whereas in monkeys both immunoperoxidase and immunofluorescence methods were used.

There was no difference in the pattern of GFP labeling displayed with either method in the monkey brain (not shown). In both species, the GFP expression was largely confined to neuronal structures (see hereunder). In addition to neuronal cell bodies and dendrites, various brain regions were also enriched in GFP-immunoreactive axonal- and terminal-like profiles (Figs. 2–5; Supplementary Figs. S3 and S4). These axons and terminals likely originated from specific populations of GFP-positive neurons. In both species, double immunofluorescence was used to determine the chemical phenotype of GFP-containing neurons in specific brain regions (Figs. 1 and 6).

Figure 2.

General pattern of expression of SYN1-EmGFP after ICV injection of AAV9-PHP.B in the monkey lateral ventricle. Schematics of coronal sections through the rostrocaudal axis of the rhesus monkey brain to map the distribution and relative density of GFP-immunoreactive neuronal cell bodies and axonal processes in the two monkeys (MR322 and MR325) that received ICV injections of AAV9-PHP.B SYN1-EmGFP-WPRE. For each animal, the schematics are based on GFP-immunostained sections from the hemisphere ipsilateral to the ICV injection. The interaural level of each section, based on the stereotaxic atlas of Paxinos et al.,³⁴ is indicated in the lower right corner of the schematics. The relative abundance of GFP-positive neuronal cell bodies is indicated in the maps as green circles, while areas with GFP-positive neuropil are represented with orange and brown shadings. Color images are available online.

Figure 3.

Expression of EmGFP in monkey PFC and forebrain. (A–C) Low-power views of GFP-immunoreactive neuronal cell bodies and axonal processes in the PFC and basal forebrain structures of monkey MR322. (D, E) Medium- and high-power views of labeled neurons in the PFC. Both pyramidal (black arrows) and nonpyramidal (non-filled arrows) cells were immunostained in deep and superficial cortical layers, respectively (one in A). (F) Dense aggregates of immunoreactive neuronal cell bodies within a strongly immunostained neuropil in the periventricular region of the dorsomedial CD. (G) Immunoreactive axonal processes in the PU. (H) Dense immunostained neuropil and a large number of immunoreactive neuronal cell bodies in the BLA and BMA regions of the amygdala. (I) Immunoreactive neurons in the NBM/SI region. (J) Labeled axonal processes (arrows) and a few immunoreactive cell bodies (marked with circles) in the dorsal part of the GPe. Scale bar shown in (A) represents 5 mm in (A–C); 1 mm in (F–J); 0.5 mm in (D); 0.1 mm in (E). BLA, basolateral; BMA, basomedial; CD, caudate nucleus; GPe, external globus pallidus; NBM, nucleus basalis of Meynert; PFC, prefrontal cortex; PU, putamen; sep, septum; SI, substantia innominata. Color images are available online.

Figure 4.

Expression of EmGFP in monkey midbrain. (A) Low-power view of GFP-immunostained neuronal elements at the level of the midbrain in monkey MR322. (B) Bundle of immunostained axons traveling through the IC. Note paucity of labeling in the neighboring RTN and VP thalamic nuclei. (C) Patches of GFP-immunostained axons and terminals with sparsely distributed immunoreactive neuronal cell bodies (open circles) throughout the medial VLm of the thalamus. (D) GFP-immunoreactive cell bodies in superficial layers of the LGN. (E, F) Low- and high-power views of GFP labeling throughout the Hipp. Rich plexuses of immunoreactive pyramidal cell bodies and dendritic processes within a dense immunostained neuropil are found throughout CA1-3 regions (F). In contrast, the DG is almost completely devoid of immunoreactivity. (G, H) GFP immunoreactivity in the SNc and SNr. In both regions, the GFP labeling is largely confined to rich plexuses of axons and terminals with few immunoreactive neuronal cell bodies (H). (I) Strong labeling of pyramidal (arrows) and nonpyramidal (boxed area) cells in the insular somatosensory cortices. Scale bar shown in (A) represents 10 mm in (A), 5 mm in (G), 3 mm in (C–E); 2 mm in (B); 1.5 mm in (H, I); and 1 mm in (F). DG, dentate gyrus; IC, internal capsule; LGN, lateral geniculate nucleus; RTN, reticular thalamic nucleus; SNc, substantia nigra pars compacta; SNr, substantia nigra reticulata; VLm, ventrolateral nucleus; VP, ventroposterior. Color images are available online.

Figure 5.

Expression of EmGFP in the monkey brainstem and Cer. (A) Low-power view of GFP immunostaining in the upper brainstem in monkey MR322. (B–D) High-power views of GFP-immunoreactive cell bodies and fibers in the PAG (B), SC (C), PBG (D) and PPN (D). The inset in (B) is a high-power view of the area in the rectangle showing GFP-positive periventricular ependymal cells (arrows). (E) Low-power view of GFP labeling in the monkey upper medulla. The DR and LDT nuclei contain labeled cell bodies, whereas the Py display dense axonal labeling. (F) High-power views of GFP-labeled cells in the DR, the LDT, and the LC. Note the stronger neuropil immunoreactivity in the DR compared with other regions surrounding the 4v. (G) GFP-immunoreactive cell bodies in the POI surrounding the IO devoid of GFP immunolabeling. (H) GFP immunostaining in the Cer and RF at the level of the central medulla. Areas of strong cellular labeling at this level include the CoN and the cerebellar cortex. Py are enriched in GFP-labeled axons. Sparsely distributed neuronal cell bodies are found throughout the RF. (I) Dense clusters of GFP-positive GC and Pf within the cerebellar cortex. The dendrites of some PC are also seen. (J) High-power view of the dense cell body and fibers labeling in the CoN. (K) GFP labeling in the lower medulla at the level of the PYx. Some cellular labeling can be found in the SpT and Cu/Gr nuclei. The PYx is enriched in GFP-labeled axons. (L) Clusters of GFP-positive PC and their dendritic processes that extend in the Mol of the cerebellar cortex. Scale bar shown in (A) represents 10 mm in (A, K); 5 mm in (E, H); 2 mm in (I, J, L); 1 mm in (B, D, F, G); 0.7 mm in (C). 4v, 4th ventricle; AQ, aqueduct; CoN, cochlear nucleus; Cu/Gr, cuneate/gracile; DR, dorsal raphe; GC, granule cells; IO, inferior olive; LC, locus coeruleus; LDT, laterodorsal tegmental; PAG, periaqueductal gray; PBG, parabigeminal nucleus; POI, peri-olivary nucleus; PPN, pedunculopontine nucleus; RF, reticular formation; SC, superior colliculus; SpT, spinal trigeminal; Py, pyramids; PYx, pyramidal decussation. Color images are available online.

Figure 1.

Widespread expression of SYN1-EmGFP after ICV injection of AAV9-PHP.B into the lateral ventricle of mice. (A) Low-power view of GFP immunostaining in the mouse forebrain at the level of the rostral Str. (B) Higher magnification of the area outlined by box 1 in (A), demonstrating GFP-positive axonal bundles projecting from the cortex. (C–E) Double immunostaining of cells in the Str (box 2 in A) for GFP and DARPP32 (a marker for medium spiny neurons) demonstrates a high degree of co-labeled cells (arrowheads in E). (F) Low-power view of GFP immunostaining at the level of the rostral Hipp and thalamus. (G) Higher magnification of the area outlined by box 1 in (F), demonstrating GFP immunoreactivity in the so and the sr of the CA1 region. Few pyramidal neuron cell bodies were GFP-positive. (H–J) Double immunostaining of cells in the S1 cortex (box 2 in F) for GFP and PV demonstrates the GABAergic phenotype of many nonpyramidal transduced cells (arrowheads in J). Large GFP-positive pyramidal neurons are also apparent (arrows in J). (K) Low-power view of GFP immunostaining in the mouse midbrain. (L) Higher magnification of area outlined in the box in (K), demonstrating the SN labeled with TH (a marker of catecholaminergic cells, red) and GFP (green). (M–O) Double immunostaining of cells in the SN for GFP and TH demonstrates the catecholaminergic phenotype of a subset of GFP-positive cells in this region (arrowheads in O). (P) Low-power view of GFP immunostaining in the Cer and brainstem. (Q) Higher magnification of the area outlined in the box in (P), demonstrating sparse GFP immunoreactivity in the Gl, the Mol, and the PCL of the cerebellar cortex. (R–T) Double immunostaining of cells for GFP and CB (a marker of PC, red) demonstrates occasional co-labeled cells (arrowheads in T). Scale bar is 1 mm in (A, F, K, and P); scale bar is 50 μm in (R, S, and T); scale bar is 100 μm in all other panels. All images are representative of at least two sections each from three mice where results were similar in all images examined. Boxed areas are representative of where higher magnification images were taken, but not all images shown are from the same section. DAPI (blue) is used as a counterstain in lower magnification panels and overview images. AAV9, adeno-associated virus 9; AC, anterior commissure; CB, calbindin; CC, corpus callosum; Cer, cerebellum; DAPI, 4′,6-diamidino-2-phenylindole; DARPP32, dopamine and cAMP-regulated phospho = protein Mr 32 kDa; EmGFP, emerald GFP; GABA, gamma aminobutyric acid; GFP, green fluorescent protein; Gl, granule cell layer; Hipp, hippocampus; ICV, intracerebroventricular; Mol, molecular layer; PC, Purkinje cells; PCL, Purkinje cell layer; PV, parvalbumin; SN, substantia nigra; so, stratum oriens; sr, stratum radiatum; Str, striatum; SYN1, synapsin 1; TH, tyrosine hydroxylase; ZI, zona incerta. Color images are available online.

Figure 6.

GFP colocalizes with various types of neuronal markers in monkey. Confocal images of double immunostaining for GFP and various neurotransmitter markers to characterize the chemical phenotype of some GFP-immunoreactive neurons throughout the monkey brain. All images are from monkey MR322. (A, A′, A″) Colocalization of GFP and DARPP32, a marker of striatal projection neurons, in the CD. (B, B′, B″) Colocalization of GFP and GABA in non-pyramidal cells in the cerebral cortex. (C, C′, C″) Colocalization of GFP and TH in the SN. (D, D′, D″) Colocalization of GFP and 5-HT in the DR. In (A–D) arrowheads point to examples of double-labeled neurons. Scale bar in (A) represents 30 μm in (A, B) rows, 20 μm in rows (C, D). 5-HT, serotonin. Color images are available online.

GFP immunostaining in mice

As shown in Fig. 1, a single unilateral ICV injection of AAV9-PHP.B SYN1-EmGFP in the mouse lateral ventricle led to widespread GFP expression throughout the whole mouse brain. The general pattern of staining was similar for the three mice used in this study. At the level of the injection site, there was no significant contamination of the overlying cerebral cortex by the virus spread along the injection track, indicating that the bulk of the solution was delivered into the lateral ventricle. Some evidence of stronger GFP expression in the ipsilateral hemisphere compared with the contralateral hemisphere was apparent in structures nearest the injection site, including the caudal Str and rostral hippocampus (Hipp).

Although most brain regions contained a significant number of GFP-immunostained neurons, some areas, such as the cerebral cortex, the Str, the Hipp, the thalamus, the SN and various periventricular regions in the pons and medulla were particularly enriched in transduced neurons (Fig. 1). At the cortical level, numerous GFP-positive pyramidal and nonpyramidal cells were found in all layers. Although most cortical regions contained labeled neurons, the cortical labeling was more profuse in frontal cortices (motor, prefrontal, limbic, cingulate, etc.) than in other cortical areas (Fig. 1A, F, K). The prevalence of GFP-transduced neurons in several cortical regions was estimated with double fluorescence using antibodies against NeuN, a neuronal marker (see Methods section). On average, 17.66 ± 1.05% NeuN-positive neurons expressed GFP immunoreactivity (Supplementary Fig. S2).

The GABAergic phenotype of nonpyramidal cortical cells (presumed interneurons) was demonstrated with double immunofluorescence using immunoreactivity for PV (Fig. 1H–J). We found that 41.7 ± 4% of total nonpyramidal GFP-containing cells were also PV-positive. Strong cellular labeling was also found in the Str (Fig. 1A) where neurons were distributed throughout the full mediolateral and dorsoventral extent of the structure.

Rostrocaudally, GFP-positive neurons were most abundant in the precommissural Str. The majority of labeled cells displayed the morphology of medium spiny striatal projections neurons, which was confirmed with colocalization experiments showing that most (79 ± 3.5%) of GFP-positive striatal cells expressed DARPP32 immunoreactivity, a known marker of striatal spiny projection neurons (Fig. 1C–E). In addition to immunoreactive cells, the dorsomedial Str also contained GFP-immunostained bundles of axons, which likely arose from transduced cortical projection neurons (Fig. 1B). In the hippocampal formation, the neuropil of both the stratum radiatum (sr) and the stratum oriens (so) was strongly immunoreactive for GFP across all CA regions.

A small number of GFP-immunostained cell bodies were apparent in the tightly packed layer of pyramidal neurons in the CA3 region, whereas none was seen in CA1 (Fig. 1G). The dentate gyrus (DG) was less intensely stained, containing only a few GFP-positive neurons (Fig. 1G). At the level of the midbrain, GFP-immunostained neurons abounded in the caudal thalamic nuclei, the superior and inferior colliculi, the zona incerta, the SN, and the ventral tegmental area (Fig. 1K). Double immunofluorescence experiments were conducted to determine if nigral dopaminergic neurons expressed GFP. We found that 41.8 ± 10% of GFP-containing neurons coexpressed TH indicating their dopaminergic phenotype (Fig. 1L–O).

In the pons and medulla, widespread clusters of GFP-immunostained neurons were found in the reticular formation (RF), various cranial nerve nuclei, raphe, pedunculopontine nucleus, cochlear nucleus (CoN), and periolivary region, among other nuclei (Fig. 1P). The cerebellum (Cer) contained some GFP-labeled cells in the granular cell layer and the molecular layer (Fig. 1Q). Occasional weakly labeled Purkinje cell (PC) bodies, which co-labeled with the PC marker calbindin, were also observed (Fig. 1R–T).

GFP immunostaining in monkey

GFP-positive elements were found throughout the whole brain in both monkeys used in this study. The GFP labeling was found in neuronal cell bodies, but it also abounded in axonal and terminal-like profiles, which originated from the various populations of transduced neurons. We did not find GFP labeling in putative astrocytes or other glial cells, except for sparse staining of astrocytes in the white matter region immediately dorsal to the injection site in the lateral ventricle (not shown). Although similarities in staining pattern and distribution were common between the two animals, some differences were noticed, as illustrated in Fig. 2 (see also Supplementary Fig. S3).

Monkey MR322 (that received 3.95 × 10¹² GC/kg) showed more abundant GFP-positive neurons than monkey MR325 (that received 1.87 × 10¹² GC/kg). In addition, some areas in which the neuropil was GFP-positive in MR322, such as the amygdala, Hipp, and superior colliculus (SC), only contained sparse labeling in MR325. The transduction in the spinal cord and the dorsal root ganglia was not analyzed in either monkey. In control experiments, brain sections of both monkeys incubated without the primary GFP antibody showed no labeling (Supplementary Fig. S1).

Figures 3–5 illustrate examples of GFP-immunostained elements in various brain regions of monkey MR322, as revealed with the immunoperoxidase approach, whereas Fig. 6 shows results of double-labeling immunofluorescence experiments that were conducted to characterize the chemical phenotype of specific populations of GFP-transduced neurons. Supplementary Figure S3 shows representative examples of GFP immunostaining in monkey MR325.

As performed for the mouse material, the prevalence of GFP-transduced neurons in several cortical regions was estimated using double-labeling experiments with antibodies against NeuN and GFP. This quantification was carried out in monkey MR322. We found that 5.6 ± 3.2% of NeuN-positive neurons coexpressed GFP immunoreactivity (Supplementary Fig. S2).

ICV injection sites and animal overall health

In the two monkeys, the unilateral injections of the viral vector solution were delivered at the same location in the lateral ventricle. There was no evidence of structural damage or abnormal enlargement of the lateral ventricles at the injection sites and throughout the whole ventricular space. Neither animal displayed any behavioral changes or unusual signs of distress after the surgical procedures and during the 4-week survival period. Although the injection was delivered into the left lateral ventricle, the GFP expression was largely similar in both hemispheres.

Cerebral cortex

Overall, the whole cerebral cortex contained GFP-immunoreactive pyramidal and nonpyramidal cell bodies that spread through all cortical layers, but as shown in Fig. 2, the density of positive cortical neurons was higher in MR322 than MR325. In both monkeys, the dendritic arbor of GFP-containing pyramidal cells was extensively labeled allowing to visualize the full extent of their basal and apical dendrites (Figs. 3D, E and 4I; Supplementary Fig. S3D).

Many nonpyramidal neurons also displayed GFP immunoreactivity throughout the cerebral cortex. In general, they were most abundant in superficial cortical layers (Figs. 3D and 4I). Double-labeling experiments revealed that a large subset (60.3 ± 13%; n = 189) of nonpyramidal GFP-positive neurons were also positive for GABA, indicating their interneuronal phenotype (Fig. 6B). In addition to labeled cell bodies, some cortical areas also contained a significant number of axonal and terminal-like profiles, some of which most likely originating from local axon collaterals and corticocortical connections of GFP-positive cortical neurons.

Striatum and globus pallidus

In the Str, strong GFP immunoreactivity was found along the periventricular region of the head and body of the CD (Fig. 3B, C, F). As shown in Fig. 3F, these neurons were distributed within a densely labeled neuropil (see also Supplementary Fig. S4B). Colocalization studies with DARPP32 revealed that 80.2 ± 5% of GFP-positive cells (n = 563) in the CD belong to the population of striatal medium spiny projection neurons (Fig. 6A). In contrast to these periventricular striatal regions, the putamen and the lateral part of the CD contained far less GFP-immunoreactive cell bodies. However, they were invaded by rich plexuses of labeled axon- and terminal-like profiles (Fig. 3G; Supplementary Figure S3A, B; Supplementary Fig. S4A, B) that likely originated from GFP-transduced corticostriatal neurons.

In the internal and external globus pallidus, the cellular staining was sparse (Fig. 3J; Supplementary Fig. S3A, C). Most of the GFP immunoreactivity was associated with axonal processes that travel through the dorsal third of both pallidal segments, a pattern reminiscent of the location of striatopallidal projections from the CD,³⁵ suggesting that the main source of this pallidal labeling are the GFP-containing neurons in the medial part of the CD (Fig. 3J; Supplementary Fig. S3C).

Basal forebrain and amygdala

In both monkeys, the amygdala contained GFP-immunostained neurons, but the prevalence of labeled cells was significantly higher in MR322 than MR325. In both animals, labeled neurons were mainly found in the basolateral (BLA) and basomedial (BMA) nuclei, although the ventral cortical amygdaloid region also displayed moderate neuronal staining (Fig. 3C). In all amygdala regions, GFP-labeled neurons displayed the morphology of glutamatergic projection cells, although some interneurons may have also been transduced. In addition to labeled cell bodies, both the BLA and BMA were also enriched in labeled axonal and terminal profiles (Fig. 3H).

In both monkeys, some neurons in the nucleus basalis of Meynert that lay within the ventral most region of the substantia innominata also displayed GFP immunoreactivity (Fig. 3I; Supplementary Fig. S3A, H). The septum was also enriched in labeled neurons within a strongly labeled neuropil (Fig. 3B).

Thalamus and hypothalamus

In general, thalamic nuclei did not exhibit strong cellular GFP immunostaining in both monkeys, except for the anterior nuclei that lay along the ventral border of the lateral ventricles (Fig. 4A). Both the anteroventral and anterodorsal nuclei displayed strong cellular and neuropil GFP immunostaining. In other thalamic nuclei (e.g., medial ventrolateral nucleus, mediodorsal nucleus), a few GFP-immunoreactive neuronal cell bodies and dense patches of labeled axonal- and terminal-like profiles were found (Fig. 4A, C, Supplementary Figure S3E, F; Supplementary Fig. S4C, D). Other thalamic nuclei contained sparse or no GFP-immunostained elements (Figs. 2 and 4A, B). In the lateral geniculate nucleus, GFP-immunoreactive cell bodies were mainly confined to the superficial layers, with only a few scattered cells in deeper layers (Fig. 4D).

In the hypothalamus, the paraventricular nucleus was most enriched in neuronal cell bodies. Although less abundant, the periventricular part of the ventromedial and dorsomedial nuclei also contained positive neurons. A moderate neuropil of labeled axons and terminals invaded the entire hypothalamus.

Hippocampus

The Hipp was strongly immunostained in monkey MR322 (Fig. 2). Cell body and axonal-like labeling was found throughout the whole hippocampal formation, except in the DG, which was almost completely devoid of immunoreactivity (Fig. 4E, F). In monkey MR325, only sparse cell body labeling was observed (Fig. 2). In CA1, CA2, and CA3 regions, the cell bodies and extensive parts of the dendritic arbors of a large number of pyramidal-shaped neurons were immunoreactive (Fig. 4F). Although in lower number than pyramidal cells, other nonpyramidal neuronal perikarya were also found. In addition to positive cell bodies, all CA regions also contained dense meshworks of axon- and terminal-like profiles that filled the entire extent of each region (Fig. 4E, F).

Midbrain

At the level of the midbrain of MR322, cell body labeling abounded in the SC, the external cortex of the inferior colliculi, the parabigeminal nucleus (PBG), the periaqueductal gray (PAG), the microcellular tegmental nucleus, the pontine nuclei (PN), the laterodorsal tegmental (LDT) nucleus, and the DR. Although to a lower prevalence, GFP-immunoreactive neuronal perikarya were occasionally found in the SN, ventral tegmental area, and the RF (Figs. 4A, G, H and 5A–D). In the PAG, SC, and PBG, labeled neurons lay within a dense immunoreactive neuropil made up of dendritic, terminal-, and axonal-like processes (Fig. 5B–D; Supplementary Fig. S4F).

In the SN, very few GFP-immunoreactive cell bodies were found, but rich clusters of terminal-like profiles invaded the substantia nigra reticulata (SNr) and the lower tier of the substantia nigra pars compacta (SNc) (Figs. 2 and 4G, H; Supplementary Fig. S4E). Although the exact source of these terminals was not determined in this study, GFP-containing neurons in the CD are the most likely source of this nigral innervation.³⁵ Large clusters of GFP-labeled axons also invaded the cerebral peduncle, further indicating strong GFP expression in corticospinal axons. Colocalization studies indicated that 68.8 ± 13% of GFP-positive neurons in the SN (n = 37) displayed TH immunoreactivity, thereby confirming the dopaminergic phenotype of some of these neurons (Fig. 6C). TH-negative neurons may be GABAergic SNr neurons. In MR325, the GFP labeling at the level of the midbrain was far less intense, except for the rich neuropil staining in the SN (Supplementary Fig. S3E, G).

Pons and medulla

In the pons and medulla of MR322, regions most enriched in GFP-containing neuronal cell bodies laid along the border of the 4th ventricle and cerebral aqueduct including the LDT and DR (Fig. 5E, F). In the DR region, 30.4 ± 1.8% of GFP-positive cells (n = 129) examined in double immunostained sections coexpressed 5HT immunoreactivity (Fig. 6D; Supplementary Fig. S4G). The CoN, PN, periolivary nucleus, and the spinal trigeminal nucleus also harbored large groups of labeled cells (Fig. 5G, H, J). In addition to these cell bodies-enriched regions, the pyramids were heavily stained with large bundles of GFP-positive axons, confirming the extensive labeling of the corticospinal tract (Fig. 5E, H, K). In monkey MR325, GFP-positive cell body and neuropil labeling was sparse in all regions (Supplementary Fig. S3I, K).

Cerebellum

Dense clusters of GC and Purkinje neurons displayed strong GFP immunoreactivity (Fig. 5I, K, L) throughout the whole cerebellar cortex of MR322. In addition, rich networks of GFP-immunostained parallel fibers, originating from the granule cell layer, cut across the dendritic tree of labeled Purkinje neurons in molecular layer (Fig. 5I). Deep cerebellar nuclei contained a few sparsely distributed GFP-labeled neurons. A similar pattern of cerebellar labeling was found in MR325, but to a significantly lower extent (Supplementary Fig. S3J).

DISCUSSION

Single intraventricular injections of AAV9-PHP.B SYN1-EmGFP in mice and monkeys resulted in distributed transgene expression throughout the brain. In both species, the GFP-positive neurons were observed mostly in cortical regions, but also in some subcortical nuclei in the basal forebrain, midbrain, and upper brainstem, as well as in the Cer. As expected for transgene expression driven by the SYN1 promoter, the vast majority of GFP-containing cells were neurons, of which the chemical phenotype was characterized in various brain regions in mice and monkeys. In both species, but particularly in monkeys, periventricular subcortical regions harbored a larger density of GFP-positive neurons than deep structures away from the ventricular spaces. Abundant labeling was also seen in putative axonal fibers and terminals that likely originated from GFP-positive cell bodies in cortical or subcortical regions.

The GFP expression in subcortical structures was more frequently found in nuclei located along the ventricles or subarachnoid spaces, as could be expected after diffusion of the virus solution through the CSF. However, GFP-positive cell bodies were also found in regions not immediately adjacent to the CSF spaces, such as the amygdala. We cannot discard the possibility that the presence of GFP-labeled cells in the amygdala of MR322 was the result of retrograde transport of the virus from cortical or subcortical regions with abundant GFP expression. For example, some of the GFP-containing neurons in BLA could have been labeled by retrograde transport from the CD, which is known to receive inputs from BLA.^35,36 However, in the SNc, another major source of inputs to the Str, the neuronal GFP expression was very sparse. Furthermore, the amygdala of MR325 was almost completely devoid of GFP-containing neurons, despite strong GFP expression in the CD. Together, these data suggest that the strong GFP labeling in BLA of MR322 cannot be accounted for solely by retrograde transduction from the CD.

The low transduction of deep subcortical structures distant from the ventricles in the monkey brain was likely owing to the limited amount of virus that spread through the ventricular walls into the brain parenchyma. Consistent with this possibility, ependymal cells along the ventricles expressed strong GFP immunoreactivity (as shown in inset of Fig. 5B).

Our results are in general agreement with previous reports of AAV injections in the CSF in NHPs, which reported widespread expression of the transgene throughout the brain.^{7,8,10–12,15,16,19,37,38} However, in contrast to the previous studies, we rarely found expression of GFP in non-neuronal cells, suggesting that the use of SYN1 as neuronal-specific promoter allowed to selectively target neurons. The GFP expression in non-neuronal cells was limited to ependymal cells lining the ventricles or to astrocytes in the white matter dorsal to the injection sites. One possible explanation for this non-neuronal transduction is that the SYN1 promoter could drive expression in non-neuronal cells at high virus titers,³⁹ as is the case in white matter next to the injection site and in the ventricular spaces.

A recent study reported abundant transgene expression throughout the brain after intrathecal injections of AAV9-PHP.B in monkeys.¹⁹ However, these authors observed transgene expression not only in neurons, but also abundantly in glial cells, a difference likely explained by the fact that the GFP expression was driven by the ubiquitous promoter CAG in this study, whereas we used the selective neuronal promoter SYN1. Other distinctions that might have contributed to differences in GFP expression between our study and this previous report include the use of different variants of GFP (EGFP⁴⁰ used by Liguore et al. vs. EmGFP⁴¹ in our study); inclusion of WPRE in our plasmid, but not in the one used by Liguore et al., different sources of AAVs, and different locations of the intra-CSF injections (ICM injections in Liguore et al. vs. ICV injections in our study). There was also a difference in the dose and volume infused: Liguore et al. injected 1 × 10¹² GC/kg (total dose 2 × 10¹² GC) in 2 mL, whereas we injected 3.95 × 10¹² and 1.87 × 10¹² GC/kg (total doses 3 × 10¹³ and 1 × 10¹³ GC) for monkeys MR322 and MR325, respectively, in a 3 mL volume. Finally, differences in the results between the two studies could be, at least in part, owing to capsid-promoter interactions. It was recently demonstrated that the capsid of AAVs interacts with the promoter to strongly influence the gene expression in specific cell types⁴² and even particular neuronal pathways.⁴³

The AAV9-PHP.B mutant capsid was originally selected in mice.¹⁸ In contrast to observations in rodents,^18,19,44 AAV9-PHP.B shows limited ability to cross the BBB after IV injections in NHPs,²¹ but as we have demonstrated, the AAV9-PHP.B serotype very efficiently transduced neurons in NHP when the virus solution was delivered through intraventricular injections and a neuronal specific promoter was used. It remains to be determined, however, if AAV9-PHP.B is more efficacious than its parent serotype AAV9 to transduce neurons after intraventricular administration, a question that was not addressed in this study.

Despite the widespread GFP expression across various cortical regions described in our study, the proportion of GFP-positive neurons over the total population of cortical cells (based on double labeling with the neuronal marker NeuN) was relatively low (∼17% and 5% in mice and monkey, respectively). Thus, although AAV9-PHP.B SYN1 may be a useful tool to achieve widespread neuronal transduction across the whole primate brain, future studies are needed to systematically test and optimize parameters (such as titer and volume) to achieve a higher transduction efficiency in monkeys.

It is noteworthy that not all neuron types were equally targeted by the SYN1 promoter, as suggested by our double-labeling studies. The proportion of GFP-positive cells that were also positive for the various neurochemical markers used is variable (e.g., ∼80% for DARPP32 in both mice and monkey, and ∼60–40% for GABA or PV neurons in monkeys and mice, respectively). Our results suggest that the selectivity of the SYN1 promoter for specific neuron types could also be species dependent. Although the general pattern of transgene expression was similar in both mice and monkeys, some differences were observed. For example, although only scarce PC were transduced in the mouse Cer (Fig. 1T), rich clusters of Purkinje neurons were GFP-positive in monkeys (Fig. 5). It is also possible that the unequal targeting of specific neurons resulted from the tropism of the AAV9-PHP.B virus, or a combination of this virus and the SYN1 promoter.

When we compared the relative density of transgene expression between the two monkeys, GFP-positive neurons in some regions (e.g., cortex and amygdala) were more abundant in the animal that received a higher dose of virus (MR322), perhaps indicating a trend toward a dose effect of the virus solution. However, studies with a larger number of animals and using a wider range of virus doses are needed to test this possibility.

The time course of transgene expression after AAV injections remains, to the best of our knowledge, poorly defined. Studies using in vivo methods, such as intensity of fluorescence or PET imaging, to monitor transgene expression suggest that the transgene expression levels begin to plateau ∼50 days after injection.^24,45 Thus, we cannot rule out that a survival period longer than 4 weeks would have resulted in a larger number of GFP-expressing neurons than what was found in our study. However, it is also possible that a longer survival time could increase potential toxicity of EmGFP.³⁷

In summary, we propose that the combination of AAV9-PHP.B and the SYN1 promoter may optimize neuronal expression in NHPs. This strategy could be useful in studies to express a transgene across cortical regions (e.g., for studies of cortical neurodegeneration), or in combinatory approaches, in which a recombinase enzyme (e.g., Cre) packaged in AAV9-PHP.B SYN1-EmGFP is injected ICV, and a second Cre-dependent transgene is injected at targeted locations. ICV injections of AAV9-PHP.B could also be used to broadly express a chemogenetic receptor (such as designer receptors activated exclusively by designer drugs [DREADDs]⁴⁶ or pharmacologically selective actuator molecules [PSAMs]⁴⁷) combined with injections of the corresponding actuators in targeted locations to regionally activate a population of neurons. This method could also be advantageous in optogenetic experiments.⁴⁸

One could use AAV9-PHP.B to express opsins in many areas of the cortex, and subsequently restrict the light delivery to a delimited area to achieve neuronal modulation in a cortical sub-region. Furthermore, ICV injections of AAV9-PHP.B, in combination with neuronal-specific promoters or enhancers,^26,49,50 could lead to a stronger and more abundant expression of a transgene in selective neurons after a single ICV injection, minimizing the need of repeated intracerebral injections. Finally, the development of enhanced versions of the AAV capsid, with improved efficiency to enter the primate brain from the ventricles, could further enable widespread and profuse transduction.

From a clinical application point of view, the ability to obtain widespread expression of a transgene after a single intraventricular infusion could be advantageous (e.g., to help treat brain diseases characterized by broad expression of misfolded proteins). Although our data indicate that further optimization is needed to increase the proportion of cells transduced with this technique, they offer a useful platform to identify the optimal parameters to achieve this goal.

AUTHORS' CONTRIBUTIONS

E.M.S., Y.S., and B.R.L designed the experiments, T.L.P, A.H. A.J.K, G.L; Y.S, and A.G. performed experiments; T.L.P, Y.S., K.R., and D.C. analyzed data; A.G., Y.S., E.M.S., and T.L.P. wrote the article, all authors reviewed and approved the article.

AUTHOR DISCLOSURE

No competing financial interests exist.

FUNDING INFORMATION

This work was supported by the Weston Brain Institute and the NIH/ORIP base grant (P51 OD011132) of the Yerkes National Primate Research Center. D.C. was supported by NIH training Grant T32-GM008602.

SUPPLEMENTARY MATERIAL

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4