Introduction

Gene therapy has emerged as a promising approach to treat a spectrum of neurodegenerative disorders by delivering therapeutic cargoes to the central nervous system (CNS). Of all gene therapy vectors, adeno-associated virus (AAV) became the most powerful system for in vivo gene therapy, transducing a wide range of both dividing and nondividing cell types by different routes of administration with an impressive safety profile [1,2,3].

AAV is a small, single-stranded DNA parvovirus, initially identified in 1965 as a contaminant in adenovirus preparations [4]. In fact, it is one of the smallest viruses with a non-enveloped icosahedral capsid of about 25 nm in diameter, belonging to the Parvoviridae family and to the Dependovirus genus. AAVs package a 4.7 kilobase linear single-stranded DNA genome flanked by two inverted terminal repeats (ITRs) [5]. Recombinant versions of AAVs (rAAVs) can be tailored by the insertion of a gene of interest between the ITRs [6].

Based on multiple differences in the amino acid sequence of the capsid, there are 11 naturally occurring AAV serotypes and more than 100 variants have been isolated from human and non-human primate tissues [2, 7, 8]. Differences in the conformation of structurally variable regions are predicted to dictate that capsids of different strains have different antigenic properties and bind to different cell receptors, providing unique tissue tropisms and transduction efficiencies across different mammalian organs [9, 10]. The discovery that a particular serotype, AAV9, is able to cross the blood-brain barrier (BBB) has raised the possibility to use non-invasive delivery routes to achieve widespread CNS gene expression [11]. Since the first use of AAVs as a transduction vector [12, 13], preclinical studies, clinical trials and marketed AAV-based medicines continue to show promise and advance with cautious optimism [14,15,16,17].

In spite of the efforts committed to better understand the biology of AAVs, the lack of suitable imaging technologies to efficiently monitor their tropism in whole organs with single-cell resolution hinders the full understanding of AAV properties necessary for their best use as gene delivery vectors. The ability to image deep within tissues is limited by sample opacity. This is due to light absorption in tissues rich in endogenous chromophores (haemoglobin and myoglobin) and most notably, to the heterogeneity in the amount of scattering at the boundaries between different molecules, subcellular structures, membranes and cell populations [18,19,20,21]. In the mouse brain, opacity has limited light penetration depth up to 100–200 µm in standard laser-scanning confocal microscopy (LSCM) and up to 500–800 µm in two-photon excitation fluorescence microscopy (TPEFM). This issue seems to be even more problematic when brain samples are fixed to preserve detailed structures, decreasing the imaging depth to a maximum of 300 µm [22, 23].

As a common workaround, conventional histology studies require sectioning and reconstruction to observe deep cellular structures with the disadvantages of being time consuming, difficult to automate and prone to tissue imperfections and disruption of many structures [24, 25].

Alternatively, recent studies turned their efforts to render whole organs optically transparent by i) washing or decolorizing endogenous chromophores, ii) intentionally removing hydrophobic lipids, iii) mildly denaturing proteins and thus changing their hydration state and/or iv) matching the refractive indices (RI) of different mediums within cellular components [20, 26,27,28]. The idea first took root more than a century ago with Spalteholz and since then [29, 30], a remarkable number of protocols have been developed and can be classified according to their main mechanism as: organic solvent-based, aqueous-based and tissue transformation methods [23, 31,32,33]. Each method has its own advantages and drawbacks in terms of tissue transparency, tissue size change, morphological integrity at cellular and tissue level, antigen and fluorescence preservation, time, and cost efficiency.

As light travels unobstructed deep within tissues, it is possible to acquire three-dimensional (3D) images either by LSCM, TPEFM or light-sheet fluorescence microscopy (LSFM). While LSCM and TPEFM are slow point-scanning methods that have been used to image small samples, LSFM has re-emerged as a powerful tool to generate single-cell resolution images in wider and deeper areas of tissues at rapid acquisition speeds, along with reduced photo-bleaching and phototoxic effects [34,35,36,37]. For these reasons, since 2007, several LSFM techniques have become the method of choice to image large cleared samples [38, 39].

Having in mind the window of opportunities presented by the combination of tissue clearing techniques and LSFM, in the present study we set our goal to develop a new pipeline for the biodistribution analysis of the existing and novel serotype variants of rAAVs at a whole-brain level with single-cell resolution. As an example, the transduction efficiency of rAAV9 in the mouse brain following intravenous (IV) injection was assessed by (i) standard 30 µm brain sectioning and immunohistochemistry (IHC) staining, (ii) 1 mm brain sectioning, clearing and IHC staining, and (iii) whole brain hemisphere clearing and direct analysis.

Materials and methods

Viral vectors

rAAV9 viral vectors encoding enhanced green fluorescence protein (EGFP) (rAAV9–EGFP) under the control of cytomegalovirus (CMV) promoter were acquired from Addgene (Addgene viral prep #105530-AAV9-pAAV.CMV.PI.EGFP.WPRE.bGH from James M. Wilson; RRID: Addgene_105530; http://n2t.net/addgene:105530; Watertown, MA, USA).

Titer of rAAV preparations, expressed in viral genomes/µL (vg/µL), was determined by real-time PCR using the AAVpro™ Titration Kit Ver.2 (Takara Bio Inc, Shiga, Japan) and a StepOnePlus real-time PCR system (Applied Biosystems, Foster City, CA, USA), following the manufacturer’s instructions.

Animal handling

All experiments involving animals were carried out in compliance with the European Union Community directive (2010/63/EU) for the care and use of laboratory animals, transposed into the Portuguese law in 2013 (Decree Law 113/2013). Additionally, animal procedures were approved by the Responsible Organization for the Animals Welfare of the Faculty of Medicine and Center for Neuroscience and Cell Biology of the University of Coimbra licensed animal facility (International Animal Welfare Assurance number 520.000.000.2006). The researchers received adequate training (FELASA-certified course) and certification to perform the experiments from the Portuguese authorities (Direcção Geral de Alimentação e Veterinária, Lisbon, Portugal). The animals were housed in a temperature-controlled room, maintained on a 12-hour light/dark cycle. Food and water were provided ad libitum. All efforts were made to minimize suffering.

Neonatal intravenous injection

Wild-type neonatal mice were randomly used to evaluate the brain transduction efficiency of rAAV9–EGFP upon an IV injection and were generated in-house by mating male and female C57BL/6 mice obtained from Charles River Laboratories (Les Oncins, Saint Germain Nuelles, France). Pregnant mice were housed singly and monitored daily from embryonic day 17 to 21, with the least possible disturbance, to ensure that new-born pups could be dosed on postnatal day 2 (P2).

Six new-born mice were initially rested on a bed of ice for approximately 1 min for anesthetization. Viral preps containing a total of 1 × 1011 vg of rAAV9–EGFP, diluted in 50 µL of sterile phosphate buffered saline (PBS) 1x pH = 7.4 (Fisher BioReagents, Pittsburgh, PA, USA) and supplemented with 0,001% Pluronic F-68 100x (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), were manually injected into the facial vein of four neonatal mice using a 100 µL Hamilton syringe connected to a 30-gauge bevelled tip needle (Hamilton, Reno, NV, USA). A correct injection was verified by noting blanching of the vein. Two negative control animals were injected with 50 µL of PBS 1x supplemented with 0,001% Pluronic F-68 100x. After the injection, pups were allowed to recover in a warmed platform, identified by tattooing of the toes, carefully cleaned, rubbed with their original bedding to prevent rejection by the mother, and then returned to their original cage.

Tissue collection and preparation

Fifty-three days after IV administration of rAAVs (P55), mice were terminally anaesthetized by intraperitoneal administration of a mixture of ketamine (Clorketam 1000, Vétoquinol, Lure, France) and xylazine (Rompun, Bayer, Leverkusen, Germany) and transcardially perfused with cold PBS 1x (pH = 7.4), followed by a second perfusion with freshly prepared ice-cold 4% paraformaldehyde (PFA, Sigma-Aldrich, St. Louis, MO, USA). Perfused brains were excised and post-fixed in 4% PFA overnight at 4 °C, and then transferred to a 20% sucrose/PBS 1x solution for cryoprotection. Once the brains sank (approximately 48 h later), they were frozen and stored at −80 °C.

For standard IHC, serial sagittal sections were cut with a thickness of 30 µm, using a cryostat (CryoStar NX50, Thermo Scientific, Thermo Fisher Scientific) at −21 °C. For each animal, 96 sagittal sections of the left hemisphere were collected in anatomical series as free-floating sections in PBS 1x supplemented with 0.05% sodium azide (Sigma-Aldrich) and stored at 4 °C until further processing.

The right brain hemispheres were either processed intact for clearing or sectioned into 1 mm slices. For the latter, brain hemispheres were embedded in 4% low melting agarose cubes (Sigma-Aldrich) and 1 mm thick sagittal sections were obtained on a vibratome (VT1200S, Leica Biosystems, Wetzlar, Germany). The thick brain slices were subsequently processed for clearing and immunolabeling.

Standard fluorescence immunohistochemistry

Fluorescence immunohistochemical procedure was performed on eight sagittal 30 µm sections selected with an intersection distance of 240 µm for each animal. The protocol was initiated with a blocking and permeabilization step in 0.1% Triton X-100 (Sigma-Aldrich) containing 10% normal goat serum (NGS, Gibco) in PBS 1x for 1 h at room temperature. Then, free-floating sections were incubated overnight at 4 °C with the following primary antibodies: rabbit polyclonal anti-GFP antibody (1:1000, catalogue # A-6455, Invitrogen, Thermo Fisher Scientific), chicken polyclonal anti-GFP antibody (1:500, catalogue # ab13970, Abcam, Cambridge, United Kingdom), rabbit polyclonal anti-ionized calcium binding adaptor molecule 1 (Iba1) antibody (1:1000, catalogue # 019–19741, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan), mouse monoclonal anti-glial fibrillary acidic protein (GFAP) antibody (1:500, catalogue # IF03L, Merck Millipore, Burlington, MA, USA), rabbit polyclonal anti-neuronal nuclei (NeuN) antibody (1:1000, catalogue # ABN78, Merck Millipore), rabbit polyclonal anti-oligodendrocyte transcription factor 2 (Olig2) antibody (1:500, catalogue # AB9610, Merck Millipore) and mouse monoclonal anti-Calbindin (CALB) antibody (1:500, catalogue # 214011, Synaptic Systems, Goettingen, Germany). Following three washing steps in PBS 1x for 15 min, sections were incubated for 2 hours at room temperature with the respective secondary antibodies: goat polyclonal anti-rabbit antibody conjugated to Alexa Fluor 488 fluorophore (1:200, catalogue # A-11008, Invitrogen), goat polyclonal anti-chicken antibody conjugated to Alexa Fluor 488 fluorophore (1:200, catalogue # A-11039, Invitrogen), alpaca monoclonal anti-rabbit Nanobody conjugated to Alexa Fluor 568 fluorophore (1:500, catalogue # srbAF568-1, Chromotek, Planegg-Martinsried, Germany), alpaca monoclonal anti-mouse Nanobody conjugated to Alexa Fluor 647 fluorophore (1:500, catalogue # sms1AF647-1, Chromotek). Brain sections were washed three times in PBS 1x, mounted on gelatine-coated microscope slides and coverslipped on fluorescence mounting medium (S3023, Dako, Glostrup, Denmark).

Staining of sagittal brain sections subjected to EGFP IHC was visualized with a ZEISS Axio Scan.Z1 microscope (Carl Zeiss Microscopy GmbH, Jena, Germany), equipped with a Plan-Apochromat 20x/0.8 objective. Labelling with EGFP and the different cell markers was visualized in the cerebellum and hippocampus with a ZEISS Axio Imager Z2 (Carl Zeiss Microscopy GmbH), equipped with a Plan-Apochromat 20x/0.8 objective. Detailed images of specific regions in the cerebellum and hippocampus were generated by acquiring different z-stacks on an inverted confocal microscope ZEISS LSM 710 (Carl Zeiss Microscopy GmbH), equipped with a Plan-Apochromat 40x/1.4 Oil DIC objective.

Clearing and immunolabeling of thick brain sections

One-millimetre-thick mouse brain sections were generated on a vibratome and subsequently cleared and labelled with “Binaree Tissue Clearing Kit for Immunostaining” (SHBI-001, Binaree, Inc., Daegu, Republic of Korea) [40]. Brain sections were initially incubated in 2 mL of Fixing Solution at 4 °C in a gentle rolling motion for 24 hours (or until precipitation occurred). Samples were then transferred to 900 µL of Tissue Clearing Solution A and incubated at 37 °C/60 rpm for 4 days, followed by a washing step with 5 mL of Washing Solution at 37 °C/60 rpm for 12 hours. Brain sections were immersed in 900 µL of Tissue Clearing Solution B at 37 °C/60 rpm for another 2 days and then washed with 5 mL of Washing Solution at 37 °C/60 rpm for 12 hours. At this point, samples were incubated in 4 mL of a home-made Permeabilization and Blocking Solution A [PBS 1x containing 0.2% Triton X-100, 20% dimethyl sulfoxide (DMSO, PanReac AppliChem, Barcelona, Spain), 10% NGS] at 37 °C/60 rpm for 3 days and then labelled at 37 °C/60 rpm for 6 days with anti-CALB antibody (1:50), diluted in a home-made Permeabilization and Blocking Solution B (PBS 1x containing 1% Triton X-100, 5% DMSO, 10% NGS). Samples were washed with 2 mL of Washing Solution at 37 °C/60 rpm for 24 hours and incubated at 37 °C/60 rpm for 5 days with the secondary antibody alpaca monoclonal anti-mouse Nanobody conjugated to Alexa Fluor 647 fluorophore (1:100) diluted in the above-mentioned Permeabilization and Blocking Solution B. Finally, samples were washed with 2 mL of Washing Solution at 37 °C/60 rpm for 24 hours and then incubated in Mounting and Storage Solution for a minimum of 24 hours before imaging.

Samples were protected from light by wrapping the tubes in aluminium foil during the entire procedure. Volume changes were determined before and after clearing procedures by immersion of sections in water and subsequent measurement of water displacement.

Confocal images were acquired under a ZEISS LSM 710, Axio Examiner microscope (Carl Zeiss Microscopy GmbH), equipped with a W Plan-Apochromat 20x/1.0 DIC (UV) VIS–IR 75 mm objective. Since this objective is designed for dipping into water (RI = 1.33) or physiologic salt solutions, cleared brain slices fixed on a Petri dish were immersed in PBS 1x during the imaging procedure.

Clearing of whole brain hemispheres

Whole mouse brain hemispheres were cleared with the “Binaree Tissue Clearing Kit” (SHOC-001, Binaree, Inc.) [40], according to manufacturer’s instructions. At first, samples were incubated in 5 mL of Fixing Solution at 4 °C in a gentle rolling motion until precipitation occurred (approximately 24 h). Brain hemispheres were then transferred to 2 mL of Organ Clearing Solution and incubated at 37 °C/60 rpm for 36 hours, a step that was repeated with fresh Organ Clearing Solution for another 36 hours. Samples were then washed with 20 mL of Washing Solution at 4 °C for 12 hours, incubated in DraQ5 DNA Staining Solution (2.5 µM in PBS 1x, Invitrogen) at 37 °C/60 rpm for 12 hours and washed again with 20 mL of Washing Solution at 4 °C for another 12 hours. Lastly, brain hemispheres were incubated in 2 mL of Mounting and Storage Solution at 37 °C/60 rpm, until a high degree of transparency was achieved 36 hours later.

In order to preserve endogenous fluorescence, samples were protected from light with aluminium foil. Tissue volume changes were evaluated by measuring water displacement before and after clearing as mentioned above.

The cleared brain hemispheres were superglued to a sample holder and lowered into Mounting and Storage Solution that filled the imaging chamber of a ZEISS Lightsheet Z.1 microscope (Carl Zeiss Microscopy GmbH). Images were acquired with an EC Plan-Neofluar 5x/0.16 and a Clr Plan-Neofluar 20x/1.0 Corr, n = 1.45.

Image processing

The acquired images were initially processed with ZEN imaging software (version 3.1 blue edition, Carl Zeiss Microscopy GmbH) and 3D rendering was completed with arivis Vision 4D (version 3.3.0, arivis AG, Munich, Germany) and Imaris (version 9.5.1, Bitplane AG, Zurich, Switzerland) software tools.

The atlas-based analysis of the EGFP biodistribution was achieved by following the QUINT workflow, which comprises a set of tools developed in the framework of the EBRAINS research infrastructure, for automated quantification and spatial analysis [41]. The workflow was divided in 5 major steps. In step 1, 126 regularly interspaced planes of the brain hemisphere dataset were initially exported in PNG file format and selected for analysis. In step 2, the exported DraQ5 and EGFP datasets were processed using the Resize feature of the Nutil software (NeSys Utilities V0.7.0, Neuronal Systems Laboratory, Institute of Basic Medical Sciences, University of Oslo, Norway) to fit the input size requirements of QuickNII and ilastik software packages, respectively [42]. In step 3, DraQ5 images were registered to a 3D reference atlas (Allen Mouse Brain Atlas version 3 2017). To achieve this, an XML descriptor file was generated with the FileBuilder program, provided together with QuickNII (Version 2.2, Neural Systems Laboratory) [43]. Customized atlas maps matching the spatial orientation and proportions of the experimental sections were obtained in a semi-automated anchoring process by QuickNII. User-guided nonlinear refinements were then applied to each section with VisuAlign software (version 0.9, Neural Systems Laboratory). In step 4, the ilastik software (version 1.3.3 post3, Anna Kreshuk’s laboratory, European Molecular Biology Laboratory, Heidelberg, Germany) was used to segment the EGFP images [44]. The segmentation was performed using the Pixel Classification workflow to differentiate EGFP from background. Twenty-five images were used to train the classifiers that were subsequently applied to the whole-series. In step 5, the segmentation and registration information of each section image, were uploaded in the Quantifier feature of Nutil to extract quantitative information of EGFP expression in each region of the reference atlas. Colorized 3D point clouds of the segmented objects were plotted with the MeshView atlas viewer (Allen Adult Mouse Brain Reference atlas space version 3, Neural Systems Laboratory, https://www.nitrc.org/projects/meshview).

Quantitative analysis of the dendritic arbour of ten Purkinje cells in the cerebellum were performed on Imaris software after reconstruction.

Statistical analysis

GraphPad Prism 6 (version 6.01, GraphPad Software, San Diego, CA, USA) and R software (version 3.6.3, R Foundation for Statistical Computing, Vienna, Austria) were used to determine mean, standard error of mean (SEM) and to generate the presented graphs.

Results

In order to test our pipeline for the biodistribution analysis of rAAVs at a whole-brain level with single-cell resolution, wild-type C57BL/6 mice were injected in the facial vein on P2 with 1 × 1011 vg of rAAV9–EGFP or PBS, as shown in the experimental timeline (Fig. 1A). On P55, mice were terminally anesthetised and transcardially perfused with PBS, followed by a second perfusion with 4% PFA. As depicted in Fig. 1B, the collected brains were post-fixed in 4% PFA and then cryoprotected by immersion in 20% sucrose/PBS 1x for 48 hours before storage at −80 °C. The left hemispheres were processed by (i) standard IHC staining on 30 µm sections (i.e. immunolabelling with anti-EGFP and/or antibodies against cell markers) and imaged by epifluorescence microscopy and LSCM. The right brain hemispheres were either (ii) sectioned into 1 mm thick sections, cleared and immunolabeled with anti-CALB before imaging by LSCM, or (iii) directly processed by tissue clearing and imaged by LSFM.

Fig. 1: Schematic representation of the procedure for the evaluation of the potential of rAAVs to achieve a widespread biodistribution throughout the brain after an intravenous injection.
figure 1

A Experimental timeline. Neonatal mice were injected via facial vein with either PBS as a negative control, or 1 × 1011 vg of rAAV9 encoding EGFP under the control of CMV promoter. Fifty-three days after injection (P55), mice were sacrificed and transcardially perfused with cold PBS 1x followed by a second perfusion with freshly prepared ice-cold 4% PFA. B Excised brains were post-fixed in 4% PFA at 4 °C and then transferred to a 20% sucrose/PBS 1x solution. Once the brains sank, they were frozen and stored at −80 °C. Left hemisphere were processed by (i) standard IHC staining on 30 µm sections (i.e. immunolabelling with anti-EGFP and/or antibodies against cell markers), and the right brain hemispheres were either (ii) sectioned into 1 mm thick sections, cleared and immunolabeled with anti-CALB before imaging, or (iii) directly processed by tissue clearing. Sagittal brain sections processed by standard IHC (left hemispheres) were visualized by epifluorescence and LSCM. Right hemispheres were imaged either by LSCM (1 mm thick brain sections) or by LSFM (entire brain hemispheres).

Full size image

Standard IHC analysis reveals the preferential neuronal transduction of rAAV9 in the cerebellum and hippocampus upon an intravenous administration in neonatal mice

The biodistribution and tropism of rAAV9–EGFP were initially analysed by standard IHC (Fig. 2). As expected, after EGFP immunolabeling, animals injected with PBS displayed no EGFP fluorescence. The epifluorescence representative images acquired for the sagittal sections of animals intravenously injected with rAAV9–EGFP showed that, in these experimental conditions, rAAV9 was able to cross the BBB and transduce several brain regions including the olfactory bulb, cerebral cortex, medulla, pons and most notably, the hippocampus and cerebellum, shown in greater detail in Fig. 2.

Fig. 2: Evaluation of rAAV9 biodistribution throughout the brain following an intravenous injection.
figure 2

Immunofluorescence representative images showing the widespread EGFP expression (in green) throughout multiple structures of the brain, including the cerebellum, medulla, pons, hippocampus, cerebral cortex and olfactory bulb. Scale bars represent 2 mm for whole-brain images and 500 µm for the hippocampal and cerebellar detailed views.

Full size image

In light of the evidence that rAAV9–EGFP extensively transduced the hippocampus and cerebellum, we further characterized the cell-specific tropism of rAAV9–EGFP in these brain regions. For that purpose, 30 µm brain sections were simultaneously labelled with the following antibodies: anti-Iba1 (for microglia), anti-GFAP (for astrocytes), anti-Olig2 (for oligodendrocytes), anti-NeuN (for neurons) and anti-CALB (for Purkinje cells). For each cellular marker, epifluorescence representative images of the hippocampus (Fig. 3A) and cerebellum (Fig. 3B) are presented, as well as a more detailed 3D image analysis of the double-staining achieved by LSCM. Movies of the detailed views for the different markers in the hippocampus (Supplementary Movie 1) and cerebellum (Supplementary Movie 2) of animals injected with rAAV9–EGFP were also generated. These results revealed that intravenous administration of rAAV9–EGFP in neonatal mice achieved little to no transduction of microglia, astrocytes and oligodendrocytes in the hippocampus and cerebellum. However, an extensive colocalization of NeuN and EGFP was observed across the dentate gyrus and CA1-3 regions of the hippocampus (Fig. 3A). The co-labelling of EGFP and CALB further revealed the presence of EGFP signal in Purkinje cells (Fig. 3B). Altogether, these data demonstrated the preferential neuronal tropism of rAAV9–EGFP after an intravenous administration in neonatal mice, as previously described for this serotype [11, 45]. Sections of the animals injected with PBS were also labelled with the same antibodies, but no EGFP fluorescence was observed (Supplementary Fig. 1).

Fig. 3: Preferential cell tropism of rAAV9 in the hippocampus and cerebellum upon intravenous administration.
figure 3

A Immunofluorescence representative images showing a strong neuronal (NeuN) EGFP expression in the hippocampus and a poor to absent transduction of microglia (Iba1), astrocytes (GFAP) and oligodendrocytes (Olig2) (see Supplementary Movie 1 as well). B Cell-specificity in the cerebellum showing preferential tropism of rAAV9 for Purkinje cells (CALB) and little to no transduction of microglia (Iba1), astrocytes (GFAP), oligodendrocytes (Olig2) and other mature neuronal cells (NeuN) (see Supplementary Movie 2 as well). Arrows indicate regions of colocalization. Scale bars represent 500 µm for images of the whole hippocampus and cerebellum and 50 µm for the respective 3D detailed views.

Full size image

A simple aqueous-based clearing procedure allows the 3D imaging of transduced and immunolabelled cells in one-millimetre-thick samples

Since aqueous-based clearing methods display high biosafety and biocompatibility levels without the need for dedicated equipment [33], we subsequently evaluated the potential of a recently released clearing method of this family to recapitulate and transcend the results obtained by standard IHC. By giving optical access to large volumes of tissue while simultaneously immunolabeling proteins of interest, the “Binaree Tissue Clearing Kit for Immunostaining” could further elucidate the transduction efficiency and specific tropism of rAAV9. We reasoned that slicing the brain into 1 mm thick sections could achieve better results in terms of transparency and labelling by improving diffusion directly to the regions of interest. Figure 4A summarizes the protocol for clearing and immunolabeling 1 mm sections of animals injected with rAAV9–EGFP. Here, EGFP transduced cells were detected by endogenous fluorescence signal and Purkinje cells were labelled with anti-CALB antibody. Since antibody penetration into the tissue is a rate limiting step, we took advantage of secondary nanobodies due to their smaller molecular weight (15 KDa) when compared to conventional IgG antibodies (150 KDa). Additionally, we selected a fluorophore in the far-red spectrum to minimize the interference from tissue auto-fluorescence. Tissue transformation changes in multiple steps of the protocol can be observed in Fig. 4B. Of note, precipitation of sections occurred after 24 h in Fixing Solution and started to appear more transparent after Tissue Clearing Solution A, an event that was accompanied by tissue expansion. During the labelling procedure, samples turned opaque, but great transparency was achieved after immersion in Mounting and Storage Solution. The final incubation caused slight tissue shrinkage, as determined by measuring liquid displacement at the beginning and at the end of the procedure (Fig. 4C). The cleared sections were imaged by LSCM, allowing the simultaneous visualization of immunolabelled Purkinje cells (CALB) and endogenous EGFP fluorescence in transduced cells (Fig. 4D and Supplementary Movie 3). Even for small regions, LSCM is a slow scanning method that led to photobleaching events during image acquisition of our cleared sections. LSCM imaging revealed that both somata and dendritic arbours of immunolabelled Purkinje cells were visible on the surface of cleared tissues; however, only the somata was visible deep in the tissue. On the other hand, the endogenous EGFP fluorescence could be easily detected deep within the tissue, allowing the visualization of fine and detailed structures.

Fig. 4: LSCM imaging of thick brain sections cleared and labelled with “Binaree Tissue Clearing Kit for Immunostaining” for the evaluation of rAAV9 biodistribution and tropism in the cerebellum.
figure 4

A Schematic representation of the clearing and immunolabeling procedure of 1 mm mouse brain sections. Tissue clearing and antibody staining is achieved by simple immersion in different solutions. Antibody concentration and incubation periods should be optimized for each antibody. B Representative images of thick brain sections placed on a 0.5 cm grid, over the course of multiple stages of the procedure. C Sample volume was assessed before and after the clearing and immunolabeling procedure by measuring the liquid displacement. Data represent six mouse brain sections (1 mm thick) and are expressed as mean ± SEM. Statistical significance was evaluated with paired Student’s t-test (**p < 0.01). D The “Binaree Tissue Clearing Kit for Immunostaining” efficiently clears tissue and is compatible with virally delivered fluorescent proteins and antibody labelling. 3D imaging of 1 mm thick sections by LSCM revealed a strong colocalization between endogenous EGFP fluorescence and CALB labelling of Purkinje cells in the cerebellum of mice transduced with a rAAV9–EGFP (see Supplementary Movie 3 as well). Arrows indicate regions of colocalization. Scale bars represent 500 µm for the global image of the cerebellum and 50 µm for the detailed images.

Full size image

Widespread biodistribution of rAAV9 at a whole-organ level with single-cell resolution achieved by tissue clearing and LSFM

Having demonstrated the feasibility of this method to clear thick samples while simultaneously preserving tissue morphology and endogenous fluorescence, we turned our efforts to visualise the 3D biodistribution of rAAV9–EGFP at a whole-organ level. Whole mouse brain hemispheres were cleared following a protocol summarized in Fig. 5A using the “Binaree Tissue Clearing Kit”, which is better suited for endogenous fluorescence and DNA dyes. The appearance of brain hemispheres was monitored between incubations in different solutions (Fig. 5B). As before, samples precipitated after 24 h in Fixing Solution. After three days in Organ Clearing Solution, samples developed a gelatine-like consistency and slightly expanded. Cellular nuclei were stained by immersion in a DraQ5 solution, the excess was removed by a washing step and samples were finally transferred to the Mounting and Storage Solution. Sample volumes were measured by liquid displacement before and after clearing and a significant shrinkage was observed (Fig. 5C). The highly cleared brain hemispheres were mounted inside the image chamber of a ZEISS Lightsheet Z.1 microscope (Fig. 5A), which was subsequently filled with Mounting and Storage Solution (which has a RI between 1.45–1.48). As previously suggested in other studies [25], samples were kept in the imaging chamber for several hours before imaging in order to perfectly balance eventual temperature and RI mismatches, while simultaneously allowing the dispersion of air bubbles. It was possible to image the entire right brain hemisphere with single-cell resolution and observe the EGFP expression in several brain regions (Fig. 5D and Supplementary Movie 4). Detailed views of the cerebellum revealed the unique morphological features of Purkinje cells that had been sparsely transduced by rAAV9–EGFP. Throughout multiple lobules of the cerebellum, it was possible to observe the somata of Purkinje cells located in a thin layer (Purkinje cell layer). One or two primary dendrites extending from each soma, arborizing into a distinctive highly branched, planar, fan-shaped structure, were also visualized, as described in the literature [46,47,48,49]. Additionally, for multiple Purkinje cells, a single long axon was visualized by EGFP fluorescence leaving the respective cell body towards the granule cell layer (Fig. 5D and Supplementary Movie 4), a feature that could not be observed by standard IHC.

Fig. 5: Tissue clearing of whole brain hemispheres with “Binaree Tissue Clearing Kit”.
figure 5

A Diagram overview of the “Binaree Tissue Clearing Kit” protocol. Over the course of seven days, brain hemispheres were incubated with different solutions, rendering highly transparent samples that could be imaged on a ZEISS Lightsheet Z.1. B Appearance of a brain hemisphere across different stages of the clearing procedure. Brains were placed on a 0.5 cm grid. C Evaluation of sample volume before and after clearing by measurement of liquid displacement. Data represent five brain hemispheres and are expressed as mean ± SEM. Statistical significance was evaluated with paired Student’s t-test (****p < 0.0001). D Three-dimensional biodistribution analysis of rAAV9 at a whole-organ level with single-cell resolution. The “Binaree Tissue Clearing Kit” achieves highly transparent brain hemispheres while simultaneously preserving endogenous EGFP fluorescence (in green). Nuclei were stained with DraQ5 DNA stain (in red). ZEISS Lightsheet Z.1 enables fast image acquisition of a mouse brain hemisphere (top left) as well as several brain regions, including the cerebellum (top right) and Purkinje cells (bottom images) with particular detail (see Supplementary Movie 4 as well). Arrows 1 evidence single long axons leaving the respective cell bodies and arrows 2 illustrate highly branched and planar dendritic arbours. Scale bars represent 225 µm for the cerebellum (top right) and 90 µm for the detailed views of Purkinje cells (at the bottom).

Full size image

To unlock the full potential of this technique for viral vector biodistribution, we next extended data processing and analysis at the whole-brain level. Towards this goal, we applied the QUINT workflow for quantification and spatial analysis of EGFP expression, based on available 3D reference atlases [41,42,43,44]. Anatomical regions recognized by DraQ5 labelling were used to register 126 regularly interspaced planes of the brain hemisphere dataset (Fig. 5D) to a 3D reference mouse brain atlas. EGFP fluorescent signal was segmented using the Pixel Classification workflow of ilastik software. Finally, the generated atlas maps and segmentation images were used in the Nutil software to yield an EGFP quantification in each region of the reference atlas (Supplementary Fig. 2 and Supplementary Table 1). As an example, segmented pixels were grouped and colorized under ten customized brain regions and displayed in the online MeshView atlas viewer (Fig. 6A). The percentage of transduced area for each customized region is displayed in Fig. 6B and Supplementary Table 2. Under these experimental conditions, rAAVs were able to transduce the ventricular system (6,758% of the total area of this region), cerebellum (3,933% of the total area of this region), olfactory regions (2,474% of the total area of this region), cerebral cortex (2,382% of the total area of this region), fibre tracts (1,353% of the total area of this region), hippocampus (1,158% of the total area of this region), the midbrain, hindbrain and medulla (1,084% of the total area of these regions), thalamus (1,018% of the total area of this region), striatum and pallidum (0,438% of the total area of these regions), hypothalamus (0,037% of the total area of this region).

Fig. 6: Spatial analysis and quantification of EGFP fluorescence at a whole-organ level with the QUINT workflow.
figure 6

A As an example, the coordinates of each segmented object in 16 of the total image series were displayed in the MeshView brain atlas viewer. The point cloud is colorized according to defined regions of the mouse brain. B The percentage of transduced area for each brain region was measured with Nutil Quantifier in 126 planes of the brain hemisphere dataset.

Full size image

Reconstruction and quantitative analysis of dendrite arborization of EGFP-positive Purkinje cells

We further demonstrated the relevance of clearing entire mouse brain hemispheres and subsequent LSFM visualization, by quantitatively evaluating the morphology of ten AAV-positive Purkinje cells after reconstruction (Fig. 7A). A colour gradient was adopted to label dendritic segments according to the respective number of branching points in the shortest path from the soma (dendrite branch depth). As an example, the boxed Purkinje cell (PC 2) in Fig. 7B, highlights the extent of the dendritic arbour that would be lost if 30 µm sections had been generated and processed by standard IHC.

Fig. 7: Analysis of dendrite arborization of Purkinje cells by tissue clearing and LSFM.
figure 7

A Cleared brain hemispheres were imaged by LSFM, allowing the visualization of Purkinje cells expressing rAAV-derived EGFP (left panel). The representative Purkinje cells were reconstructed on Imaris software and are shown in different perspectives (middle and right panels), where different colours correspond to a different dendrite branch depth. Scale bars represent 20 µm. B Sagittal (left panels) and coronal (right panels) views of reconstructed Purkinje cell 2 (PC 2) are presented. The framed regions in sagittal views elucidate the extent of dendrite branch loss if 30 µm sections were generated and processed by standard IHC. Purkinje cells exhibit an elaborate monoplanar dendritic arbour on P55, as shown in coronal views. Scale bars represent 20 µm. C Schematic representation of a Purkinje cell (on the left) where different colours correspond to a different branch depth and the number of Sholl intersections is depicted by concentric spheres in black. The number of dendrite branching points (indicated by letters) and the number of dendrite terminal points (indicated by numbers) are illustrated on the right. D Quantitative analysis of the total number of dendrite branching points (left graph), dendrite terminal points (middle graph) and the sum of dendrite length (right graph) of ten reconstructed Purkinje cells (PC 1–10). Values are expressed as mean ± SEM. E Number of dendrite segments at each branch depth and (F) Sholl analysis of ten Purkinje cells. Values are expressed as mean ± SEM.

Full size image

Sholl analysis of dendritic arbours was conducted by drawing imaginary spheres centred at the soma with 5 µm intervals [50]. The number of dendritic segments intersected within each sphere was calculated. The dendrite branch depth as well as the number of Sholl intersections, branching points and terminal points, are illustrated in Fig. 7C. Interestingly, a higher number of dendrite branching points leads to a higher number of dendrite terminal points but not necessarily to a higher sum of dendrite length, as quantified in Fig. 7D. For instance, PC 8 and PC 4 display a similar sum of dendrite length, but the dendritic arbour of PC 8 had strikingly higher numbers of segments at depths 15–40 and a full branch depth of 51, while PC 4 exhibited lower numbers of dendrite segments at that depth range and a full branch depth of 85, as shown in Fig. 7E. The mean number of Sholl intersections (Fig. 7F) reached the highest levels from a radius of 60–190 µm. When compared to PC 4, we also observed that PC 8 had an overall higher number of dendritic segments, but lower number of edges and full branch level (Supplementary Fig. 3A). Branch level numbers are attributed at each branching point, according to the diameter of the originating segments. It is interesting to note that a group of Purkinje cells analysed here (PC 1–4 and 10) (Supplementary Fig. 3A) displayed a mean dendrite length similar to what has been reported by other studies using rAAVs and standard IHC procedures to label Purkinje cells of animals with the same approximate age [47]. The sum of dendrite area and volume were quantified as well (Supplementary Fig. 3A). To elucidate the dependence between different characteristics of dendritic arbours, a correlation plot was also generated (Supplementary Fig. 3B). Here it was possible to observe that the full branch depth is negatively correlated with the number of dendrite branching points, while a high number of dendrite branching points resulted in high numbers of dendrite segments, terminal points as well as higher sums of dendrite length, area and volume.

The employed aqueous-based clearing procedure preserves endogenous fluorescence

An important feature in clearing methods is the preservation of fluorescence, particularly if the image acquisition relies solely on endogenous fluorescent signals. Aiming to assess the preservation of EGFP signal with the “Binaree Tissue Clearing Kit” method, we applied the protocol summarized in Supplementary Fig. 4A in both cultured cells and brain sections. Briefly, the EGFP signal was assessed 48 h after transfection with the rAAV9–EGFP plasmid in HEK293T cells that were either cleared or non-cleared (Supplementary Fig. 4B). A total of 24 confocal images for each condition were acquired and quantified, revealing a fluorescence preservation of 71 ± 8.2% in cleared cells (n = 4 for each condition, Supplementary Fig. 4C). Similar results were observed by the fluorescence quantification in 60 µm brain sections. In this case, 6–14 cerebellar regions where imaged in each of the 4 brain sections before and after the clearing method, showing a fluorescence preservation of 67.25 ± 5.4% (Supplementary Fig. 4D, E). The comparable but slightly lower levels of endogenous fluorescence signal observed in 60 µm brain sections could be due to photobleaching events rising from the fact that the same sections were imaged before and after the clearing procedure.

Different clearing techniques provide different results in terms of sample opacity (and therefore imaging depth), preservation of tissue structure and fluorescence, suitability for immunolabeling and size variation changes (as summarized in Table 1). As evidenced on an application-oriented and global scheme of clearing methods (Supplementary Fig. 5), we showed that the “Binaree Tissue Clearing Kit for Immunostaining” combined with LSCM is better suited for the characterization of the cell specific tropism of rAAVs in thick slices of tissue, while the “Binaree Tissue Clearing Kit” offers the possibility to study the biodistribution of rAAVs at a whole-organ level by LSFM imaging.

Table 1 Overview of different tissue clearing techniques.
Full size table

Discussion

In the present study, we aimed to establish a new pipeline for the biodistribution analysis of rAAVs at a whole-brain level with single-cell resolution. As an example, the transduction efficiency of rAAV9 in the mouse brain following an intravenous administration was compared by standard IHC and 3D imaging of cleared brain sections and whole brain hemispheres.

As expected from previous works [11], in the current study, the intravenous administration of rAAV9–EGFP in neonatal mice led to a preferential neuronal tropism with little to no transduction of microglia, astrocytes and oligodendrocytes fifty-three days after the injection (P55), as determined by standard IHC.

A potential improvement over standard IHC performed on thin sections would be labelling and imaging larger blocks of brain tissue, allowing a faster and more reliable analysis of the biodistribution of rAAVs. To achieve that, 1 mm thick brain sections were labelled with anti-CALB antibody and cleared with the “Binaree Tissue Clearing Kit for Immunostaining”. LSCM imaging of the cerebellum revealed dendritic arbours and somata of Purkinje cells immunolabelled at the surface of cleared sections, suggesting that the epitopes necessary for antibody recognition were preserved. However, a decrease in the intensity of the labelling in deeper regions of the tissue was observed, probably due to the limiting role of passive diffusion of antibodies. This issue could potentially be bypassed by performing multiple rounds of labelling with freshly prepared antibody solutions over longer periods of time, by increasing antibody concentration, or with the advent of new labelling dyes and antibodies able to penetrate deeper within tissues. Alternatively, recently developed methods use external forces to enhance the transport of molecules into cleared tissues [51, 52], while others employ perfusion techniques to achieve whole-organ and whole-body labelling and clearing [53, 54]. The faint immunofluorescent signal in deep regions of the tissue was also accompanied with some autofluorescence, even though a fluorophore with emission peak in the far-red range was selected to label calbindin on cleared sections. To narrow this issue, some studies have employed glycine and heparin in their immunostaining protocols [26, 35, 37, 55, 56], a strategy that may also be implemented in future studies using the clearing protocol described here. Interestingly, the endogenous EGFP fluorescence of transduced Purkinje cells could be observed deep within the cleared sections, which prompted us to clear entire mouse brain hemispheres.

We found that the performed tissue clearing technique rendered highly cleared brain hemispheres while simultaneously preserving tissue morphology and endogenous EGFP fluorescence. During the clearing procedure, samples may turn slightly opaque and be temporarily swollen, but immersion in Mounting and Storage Solution leads to increased transparency and tissue shrinkage. In fact, shrinkage and expansion of cleared samples can be modulated by slightly changing the concentrations of different components that cause sample hydration such as urea, or dehydration such as sorbitol, glycerol, sucrose and fructose [18, 57, 58]. In a way, tissue shrinkage may be seen as an advantage since the volume to image and the respective generated data are smaller and high quality images may arise due to a larger portion of tissue fitting the optimal imaging region of a light-sheet microscope [33, 37, 59]. However, dehydration-driven shrinkage may also remove water molecules essential to maintain fluorescent proteins and fluorophores [25, 60]. On the other hand, tissue expansion alone can reduce light scattering and therefore increase transparency, a key feature used in expansion microscopy [57, 61]. Additionally, tissue expansion may also create more space for molecules and antibodies to diffuse. These are common artefacts of clearing methods that must be taken into account when measuring distances, but the proportion and general anatomy of structures are usually preserved, allowing comparison studies if the same clearing method is applied.

The biodistribution of endogenous EGFP fluorescence throughout the cleared brain hemispheres recapitulated in greater detail the observations made by standard IHC. In fact, by taking advantage of open-source software, it was possible to perform a quantitative analysis of viral vector biodistribution at a whole-organ level in a convenient semi-automated workflow. To further demonstrate the potential of this clearing strategy, the complexity of anatomical characteristics of Purkinje cell dendrites were also evaluated, showing that this method could be applied to quantitatively study the transduced Purkinje cells in specific lobules or even in the entire cerebellum, without physical sectioning.

Following an intravenous administration of rAAV9 in neonatal mice, in this study we observed an extensive EGFP expression within cerebellar Purkinje cells, which is in agreement with previous reports [11, 62]. This observation suggests that these vectors could be valuable tools not only for the elucidation of Purkinje cell development in vitro and in vivo [47, 48, 63], but also for research and preclinical studies of several diseases, including multiple forms of ataxia, Huntington’s disease, schizophrenia and autism spectrum disorders, where Purkinje cells are profoundly affected [64, 65]. Moreover, the clearing methodology could help to further characterize and monitor disease progression by circumventing standard IHC limitations.

Overall, the performed clearing technique offers several advantages over other clearing methods in terms of required time and equipment, transparency, and fluorescence preservation (as elucidated in Table 1), presenting itself as a promising tool when combined with LSFM. Although we focussed on the intravenous administration of a rAAV serotype 9 vector, this innovative methodology could have tremendous applications, possibly contributing to the characterization of the tissue tropism of other natural and engineered AAV serotypes through multiple routes of administration. Furthermore, it could prove its worth in the evaluation of new strategies using rAAVs to deliver potential therapeutic genes to the CNS.