Animals
All experiments were conducted according to the National Institutes of Health guidelines for animal care and use and were approved by the Institutional Animal Care and Use Committee (IACUC), protocol number AUP-22-163, and the Institutional Biosafety Committee (IBC), protocol number BUA-R100, of the University of California, Irvine (UCI). All mice used for rabies in vivo testing have the same C57BL/6J background and were at least 8 weeks of age. Both male and female mice were used for the experiments. All strains used in this study were originally purchased from the Jackson Laboratory and breeding is maintained in Dr. Xiangmin Xu’s laboratory vivarium at UCI. Strains used in this study include: C57BL/6J (JAX stock#: 000664) and 5xFAD (MMRRC Strain #034848-JAX) (MODEL-AD at UCI), and B6 PV-Cre (JAX stock#: 017320).
Cloning and generation of recombinant rabies viruses from cDNA
The template for the initial cloning was pSADdeltaG-F3, a gift from Edward Callaway (Addgene plasmid # 32634) 2. The plasmid was digested with restriction endonuclease NheI, and inserts were cloned following ligation using T4 DNA ligase (New England Biolabs). Fluorescent reporters were amplified from the plasmids listed in Table S1. Inserts were amplified by PCR to add restriction sites to their termini and subcloned under a CMV promoter in a vector harboring the necessary added sequences (localization signals). Subclones were transfected into HeLa cells to validate the correct fluorescence and localization of the fluorescent proteins.
B7GG cells were plated in a 6-well plate. The following day they were transfected with 3 µg of plasmid harboring the RV genome, 1.5 µg of pcDNA-SADB19N, 750 ng of pcDNA-SADB19L, 750 ng of pcDNA-SADB19P and 500 ng of pcDNA-SADB19G using Lipofectamine 2000 (ThermoFisher, 11668019). These plasmids were a gift from Edward Callaway (Addgene plasmid # 32630, 32631, 32632, 32633) 14. After reaching confluency, cells were passaged and growth was carried out until >70% of the cells were infected. Virus supernatant was collected and filtered consistently every 2-3 days through a 50ml 0.45um vacuum filter (SE1M003M00, MilliporeSigma™). The filtered virus supernatant was stored in the -80°C until the amplification step described later.
In vitro characterization of recombinant rabies viruses
Initial fluorescence and localization controls were performed in HeLa cells. These cells were grown in Dulbecco’s Modified Eagle Medium (DMEM; Gibco, 12800017) supplemented with 1X Antibiotic Antimycotic solution (Omega Scientific, AA-40) and 10% NCS (Omega Scientific, NC-04). RVs were grown using B7GG cells 2 in DMEM (Gibco, 12800017) supplemented with 1X Antibiotic Antimycotic solution (Omega Scientific, AA-40) and 2% FBS (Omega Scientific, FB-12). Infected cells were grown until confluency and passage every 2 days. For imaging in vitro samples, cells were plated on glass coverslips and grown as previously described. Cells were fixed for 20 min in PBS + 3.7% formaldehyde, washed 3 times in PBS for 5 min, permeabilized 10 min in PBS + 0.2% Triton X-100, washed 3 times in PBS for 5 min, nuclei were stained with DAPI (4 µg/ml) for 2 min, and washed once with PBS before being mounted. Immunofluorescent staining was performed using antibodies against nanoluciferase (R & D systems; 965853) or Ferritin (ab69090; Abcam). Luminescence assays were performed using Nano-Glo Luciferase Assay (N1110; Promega) according to the manufacturer’s instructions.
Superior cervical ganglia (SCG) were isolated from embryonic day-16-17 Sprague-Dawley rat embryos (Charles River) and neurons were cultured in tri-chambers as described previously 45. Briefly, the SCG were incubated in 250 µg/ml of trypsin (Worthington Biochemicals) for 15 min. followed by 2x washing in HBSS buffer. Prior to plating, the ganglia were triturated in complete neuronal media using a fire-polished Pasteur pipette and then 2/3 of a ganglion was plated in the soma (S)- or compartments of the Teflon ring. The Teflon ring was placed within either a 35-mm plastic tissue culture dish or optical plastic dish (Ibidi) coated with 500 µg/ml of poly-DL-ornithine (Sigma Aldrich) diluted in borate buffer and 10 µg/ml of natural mouse laminin (Invitrogen). Neuronal culture medium consists of neurobasal medium (Gibco) supplemented with 100 ng/ml nerve growth factor 2.5S (Invitrogen), 2% B27 (Gibco), 2 mM glutamine (Invitrogen), penicillin, and streptomycin. Two days after plating, neuronal cultures were incubated with 1 mM of the antimitotic drug cytosine-D-arabinofuranoside (AraC; Sigma-Aldrich) for 2 days to kill dividing non-neuronal cell types. Neurons were cultured for 14-21 days prior to experiments. RV inoculum (104-105 pfu) was added in the S-compartments. Depending on the assay, infected axons or soma were imaged either at indicated hours post infection (hpi). Live-cell imaging was performed on a DMi8 inverted epifluorescence microscope (Leica) equipped with a DFC9000GT camera (Leica). Neuron cultures were kept in a humidified stage-top incubator (Tokai Hit) at 37°C with 5% CO2 during imaging. The movies and images were prepared using Leica imaging software and Adobe Photoshop.
Sytox Green Nucleic Acid Stain (Invitrogen S7020) was diluted 1:100 in DMEM, and 3 ul was added into the S compartments (final concentration is 500 nM). After 30 minutes incubation, media was replaced and S-compartments were imaged in phase, green and red channels. The ratio of green and red double positive neurons (dead RV infected) to red only neurons (RV infected) was calculated to determine the neurotoxicity.
Amplification, pseudotyping, and concentration of viral stocks
The protocols for rabies recovery, expansion, concentration, and titer have been described previously 14, 46.Virus supernatant collected from the transfection assay was used for the expansion of viral stocks. Recombinant RV was grown using B7GG cells in DMEM supplemented with 2% fetal bovine serum (FBS) (SH30070.03, Cytiva) in 35°C, >80% humidity, 5% CO2. B7GG cells were grown to 50% confluency in 10 tissue culture dishes (353025, Falcon™) and infected with virus supernatant. Three days post-infection, the cells were checked for fluorescence and supernatant was collected and filtered through a 0.45 μm sterile vacuum filter unit (S2HVU02RE, MilliporeSigma). The supernatant was collected every 2-3 days for up to 4 collections depending on cell health. The filtered supernatant was stored in the -80°C until pseudotyping and ultracentrifugation steps. Only supernatant collected from plates with >80% infection was used for ultracentrifugation to maximize the titer of the final concentrated stock for in vivo tests.
Recombinant RV was grown in BHK-EnvA cells (a gift from Edward Callaway) to produce rabies virions with an EnvA coat protein. BHK-EnvA cells were grown in DMEM supplemented with 2% FBS in 35°C, >80% humidity, 5% CO2. BHK-EnvA cells were grown to 50% confluency on 10 tissue culture dishes with grids (353025, Falcon™) and infected with virus supernatant. The cells were rinsed 3 times with 1x Dulbecco's phosphate-buffered saline (DPBS) (14190136, Gibco™) and trypsinized with pre-warmed 0.25% trypsin-EDTA (25200056, Gibco™) to remove any unpseudotyped virus 1 day post-infection. Three days post-infection, the cells were checked for fluorescence and virus supernatant was collected and filtered as stated previously in the amplification step. Only supernatant collected from plates with >60% infection was used for ultracentrifugation to ensure a high titer of the final concentrated stock for in vivo tests.
Both unpseudotyped and pseudotyped rabies stocks followed the same procedure for ultracentrifugation (Optima L-100XP, Beckman Coulter). Around 225 mL of frozen virus supernatant was thawed in a 35-37 °C water bath and centrifuged using SW32Ti rotor (Beckman Coulter) for 2 hrs at 20,000 RPM (70,000g) in 4 °C. This step was repeated twice to concentrate a total volume of 450 mL. The supernatant was discarded, and each pellet was resuspended using cold 1x Hanks’ Balanced Salt Solution (HBSS) (14175095, Gibco™). The viral suspensions were combined and overlaid on top of 2.5 mL of 20% sucrose/1x HBSS before ultracentrifugation using the SW55Ti for 2 hrs at 21,000 RPM (50,000g) in 4 °C. The final pellet was resuspended in 100 μL of cold HBSS, and the concentrated viral stock was stored in the -80°C prior to viral titer determination and in vivo testing.
HEK293T-TVA800 cells (a gift from Edward Callaway) were used to titer both unpseudotyped and EnvA pseudotyped rabies. Titration were performed using two different protocols: Using Chamber slides: A 8-well chamber slide (354118, Falcon™) is seeded at a density of 1.5 x 104 cells per well, and a ten-fold serial dilution is prepared from 10^3 to 10^9. Each well is infected with a dilution. Three days post-infection, the chamber slides are scanned using an inverted microscope. The images are stitched and manually counted to determine the number of infected cells. Generally, wells containing 10 to 150 fluorescent cells are counted. Titer calculations are performed following the existing protocol described in 14. Viral titers for unpseudotyped rabies viruses were consistently between 109-1010 infectious units (IU)/mL, while the viral titer for EnvA-pseudotyped rabies viruses were consistently 1 order lower, between 108-109 IU/mL. Using FACS: A 12-well plate (3513, Corning™) is seeded at a density of 1.5 x 105 cells per well. A serial dilution was prepared, and the cells were infected with applied virus volumes of 2.5 μL, 0.25 μL, 0.025 μL, 0.0025 μL, 0.00025 μL (Log 0, 1, 2, 3, 4 respectively). Three days post-infection, the cells were dissociated using pre-warmed 0.25% trypsin-EDTA (25200056, Gibco™) and fixed with cold 2% paraformaldehyde (PFA). The cells were filtered using a 5 mL round-bottom polystyrene test tube with cell strainer snap cap (352235 Falcon™). A flow cytometer (MA900, SONY) was used to analyze the percentage of infected cells for each log dilution. Generally, only logs 1, 2, and 3 are used for the calculation. Titer calculations were performed following the existing protocol described in Wickersham et al. 46. Viral titers for unpseudotyped rabies viruses were consistently between 109-1010 infectious units (IU)/mL, while the viral titer for EnvA-pseudotyped rabies viruses were consistently 1 log lower, between 108-109 IU/mL.
In vivo mouse viral injections
The general viral injection procedure has been previously described 10, 47. Both AAV and rabies virus were injected following the same procedure. The mice were anesthetized and a digital stereotaxic instrument guided by a digital atlas (Angle Two™ Stereotaxic for Mouse) was used to target the following coordinates: S1BF (primary somatosensory cortex, barrel field): anteroposterior (AP) = -0.70 mm, lateromedial (ML) = ±2.85 mm, and dorsoventral (DV) = -1.85 mm; dSUB (dorsal subiculum): AP = -3.40 mm, ML = ±1.96 mm, and DV = -1.67 mm; dCA1 (dorsal hippocampal CA1): AP = -1.94 mm, ML = ±1.40 mm; and DV = -1.35 mm; DLGN (dorsal lateral geniculate nucleus): AP = -2.30 mm, ML = ±2.14 mm, and DV = -2.75 mm; V1B (primary visual cortex, binocular area): AP = -3.40 mm, ML = ±2.70 mm, and DV = -1.40 mm; V1M (primary visual cortex, monocular area): AP = -3.40 mm, ML = ±1.96 mm, and DV = -1.00 mm. Virus was loaded into a pulled glass capillary (#BF100-50-10, Sutter Instrument) (tip inner diameter of 20-30 μm), and virus was injected using a pressure-controlled microinjection dispense system (picospritzer III, Parker Hannafin) at 20-30 nL/min with 10 ms pulse duration. A total volume of 200-400 nL was injected into each targeted region depending on the size of the targeted area.
Unpseudotyped SAD-B19-RVΔG recombinant viruses were directly injected into C57BL/6J background mice without the use of helper viruses, whereas a combination of AAVs expressing Cre, TVA, and G were used to transcomplement EnvA-pseudotyped SAD-B19-RVΔG to identify monosynaptic inputs. For the EM experiments, AAV2-retro-syn-Cre (1.57E+13) was injected into V1B while AAV8-DIO-TC66T-2A-mCherry-2A-oG (2.62E+13) was injected into DLGN and incubated for 21 days followed by the injection of EnvA-SAD-B19-RVΔG-emGFP-T2A-MLS-FTL into the DLGN. The mice were perfused after an additional 9 days.
Histology and immunochemical staining of mouse brain tissue
Mice were transcardially perfused with 1x phosphate buffered saline (PBS) followed by 4% paraformaldehyde (PFA), and their brains were dissected and post-fixed in 4% PFA for 24 hours. Samples were transferred to 30% sucrose/1x PBS. The brain tissue was frozen with dry ice and sectioned coronally into 30-μm-thick slices using a sliding microtome (SM2010R, Leica). Select brain sections expressing FTL were stained with Anti-Ferritin Light Chain antibody (ab69090, ABCAM, 1:500 dilution) followed by Cy™5 AffiniPure Donkey Anti-Rabbit IgG (H+L) (Code: 711-175-152, Jackson ImmunoResearch, 1:200 dilution). Brain sections expressing PSD95 were stained with PSD95 Rabbit mAb (A0131, ABcolonal, 1:500 dilution) followed by Cy™5 AffiniPure Donkey Anti-Rabbit IgG (H+L) (Code: 711-175-152, Jackson ImmunoResearch, 1:200 dilution). All sections except the ones stained with AmyloGlo were counterstained with 10 μM 4,6-diamidino-2-phenylindole (DAPI) and mounted for imaging. Amylo-Glo RTD amyloid Plaque Stain Reagent (Catalog: TR-300-AG, Biosensis) was used following the manufacturer’s protocol to visualize amyloid beta plaques, which are characteristic in the 5xFAD Alzheimer’s disease model mice. Sections stained with AmyloGlo were counterstained with 3 μM DRAQ7™ (Catalog: 7406, Cell Signaling Technology) and mounted for imaging. Fluoromount-G (Catalog: 0100-01, SouthernBiotech) was the mounting media used.
Imaging and analysis
Sections with sparse neuronal labeling were imaged using a confocal microscope (FV3000, Olympus) under 20x dry and 60x oil objective lens to characterize viral labeling in neurons. Tile scans were stitched and z-maximum projection (2D) images were generated using the Olympus confocal software. All files were exported as TIFF format. Brain sections expressing SAD-ΔG-1xNLS-tdtomato and SAD-ΔG-2xNLS were imaged using an automated slide scanning system (VS120, Olympus) under a 10x objective and were stitched and converted to TIFF format for automated counting analysis described later.
For mitochondria image processing and analysis, cell fluorescence intensity measurements were performed by using the Fiji-ImageJ software analysis tools. 2D images were preprocessed using the following commands: 1) “subtract background” to remove background noise; 2) “adjust threshold”; 3) “make into binary image”; and 4) perform analysis of mitochondria morphology. The binary images were used as the input for the “analyze particles” command, measuring for “area” and “perimeter". The area (μm2) was used for data reporting and statistical analysis. Particles less than 0.1 μm2 or larger than 2.4 μm2 were excluded for statistical analysis.
Neural circuit mapping and automated counting of infected neurons
To map the neural circuit in mouse hippocampal CA1, we used the following AAVs: AAV8-DIO-TC66T-2A-GFP-2A-oG (Salk Institute, CA, US, 5.06x1013 GC/ml) and pENN.AAV.CaMKII 0.4.Cre.SV40 (Addgene viral prep #105558-AAV1, 5.3x1013 GC/ml). pENN.AAV.CaMKII 0.4.Cre.SV40 was a gift from Dr. James M. Wilson. The AAV8-DIO-TC66T-2A-GFP-2A-oG was 1:2 diluted with phosphate-buffered saline (PBS). The pENN.AAV.CaMKII 0.4.Cre.SV40 was 1:4 diluted with PBS. These two diluted helper AAVs were then 1:1 mixed. Finally, 0.1 μl of the diluted AAV mixture was injected into the CA1 target region on day 1. After 21 days, the mice were injected with the rabies virus EnvA-SADΔG-RV-2xNLS-tdTomato (4.1x107 IU/ml, 0.4 μl) at the same injection site using pressure injection. The viruses were injected into the pyramidal layer of dorsal hippocampal CA1 using the following coordinates: anteroposterior (AP) -1.94 mm, mediolateral (ML) -1.40 mm, dorsoventral (DV) -1.35 mm, all values given relative to bregma.
To perform objective evaluation of the labeling quality, we built a custom-made multi-layer convolutional neural network (CNN) to perform automatic neuron detection with the captured microscopy images. We manually pick up 1512 neuron patches and 2348 background patches from 1xNLS red channel, 2xNLS red channel, and the reference green channel of 1xNLS data. Each extracted neuron or background patch has 56*56 pixel in size, and only contains the color channel in which the neurons are labeled. Z-score operation is taken to each patch for intensity normalization purpose, and pixels with negative intensity clipped to 0. In order to fit the network input requirement, the patches are converted to 3-color channels images in which the red channel contains the original patch.
The custom-made multi-layer CNN is built with the Deep Learning toolbox of MATLAB 2020b and contains 12 layers. The first layer is the input layer. Following layers contain 2 convolution layers with 14*14 and 7*7 weights, respectively. Each convolution layer accompanies by a batch normalization layer, a max pooling layer, and a ReLU activation layer. The last three layers contain a fully connected layer that generates two outputs corresponding to the neuron and background class, a SoftMax layer and a classification layer. Detailed explanation of the layers’ structure and features can be found in their corresponding studies 48, 49, 50, 51. During training, 900 neuron patches and 900 background patches are randomly selected as training data set, while the remaining data are served as validation data set. 10 training epochs are used and learning rate is set as 0.001. Training ends with validation accuracy at 99.53%. During actual neuron detection, each pixel in the microscopy image (except the 1-28 pixel close to each edge) will be used to construct a 56*56-pixel patch, with the pixel located at center, and have its intensity z-scored. The CNN will provide 0 if the patch is classified as background, and provide 1 if the patch is classified as neuron. In this way the image will be converted to a binary mask in which intra-neuron pixels are distinguished from background. Individual neurons will be isolated from the binary mask.
In vivo live imaging of neurons activity
To validate RV-GCaMP7f infected cells’ response to intracellular calcium level increase, cultured SCG cells were infected with the unpseudotyped SAD-B19-ΔG-RV-GCaMP7f virus. Two days post-infection, the SCG cells were imaged before and post 50 mM KCl treatment with a live imaging fluorescent microscope (Leica DMi8).
To further test in vivo monitoring of neuronal activities in awake animals, excitatory CA1 neurons in C57 mice were labeled with EnvA-pseudotyped RVΔG GCaMP7f. The mouse was injected with 0.1 μl of helper AAV mixture. The AAV mixture contained the 1: 1 mixture of AAV8-DIO-TC66T-2A-mCherry-2A-oG (titer: 3.42x1013 GC/ml) and AAV1-Camk-Cre (titer: 1.9x1012 GC/ml). 21 days after the AAV injection, 0.2 μl of EnvA-SAD-B19-ΔG-RV-GCaMP7f (titer: 1.2x109 IU/ml) was injected at hippocampal CA1. For both AAV and RV injection, the following coordinate: AP -2.06mm, ML 1.65mm, DV -1.47 mm was used. The mouse was then implanted with a cylindrical cannula on top of hippocampal CA1 following a published protocol 52 on the same day immediately after the RV injection. Briefly, the cortical tissue and a thin layer of corpus callosum above mouse CA1 were carefully removed, and a cannula sealed with a glass coverslip at the bottom was implanted on top of CA1. A head bar was further glued on the skull to allow for fixation on a two-photon microscope (Neurolabware, CA, US). Mouse received intramuscular dexamethasone injections before the surgery and subcutaneous carprofen daily for three days post-surgery. Around 9 days post-infection and surgery, the mouse was fully recovered and CA1 neurons were imaged when the mouse was awake. The two-photon imaging data was recorded and further analyzed using CNMFE-based custom-written MATLAB pipelines 53.
Correlated light and electron microscopy
For 3D correlative light, X-ray and electron microscopy in emGFP expressed mice, mice were anesthetized with an intraperitoneal injection of ketamine/xylazine and transcardially perfused with a brief flush of Ringer’s solution containing heparin and xylocaine, followed by approximately 50 mL of 0.1% glutaraldehyde / 4% paraformaldehyde in 0.15M sodium cacodylate buffer containing 2mM CaCl2 (CB, pH 7.4). The brain was post-fixed overnight on ice in the same fixative and 100 µm thick coronal sections. RV infected brain slices were collected and incubated in DRAQ5 (1:1000, Cell Signaling Technology) on ice for an hour. Confocal images of emGFP and DRAQ5 signals were collected on a confocal microscopy (Olympus FluoView1000; Olympus, Tokyo, Japan) with a 10X, 20X and 60X oil-immersion objective lens using 488 nm and 633 nm excitation.
Confocal imaged brain slices were prepared for MicroCT and SBEM as previously described 41. Briefly, immediately after confocal imaging, brain slices were fixed in 2.5% glutaraldehyde in 0.15M cacodylate buffer (CB, pH 7.4) at 4°C an hour. After removing the fixative, brain slices were washed with 0.15M CB and then placed into 2% OsO4/1.5% potassium ferrocyanide in 0.15 M CB containing 2 mM CaCl2 for 1 hours at room temperature (RT). After thorough washing in double distilled water (ddH2O), slices were placed into 0.05% thiocarbohydrazide for 30 min. Brain slices were again washed and then stained with 2% aqueous OsO4 for 30 min. Brain slices were washed and then placed into 2% aqueous uranyl acetate overnight at 4°C. Brain slices were washed with ddH2O at RT and then stained with 0.05% en bloc lead aspartate for 30 min at 60°C. Brain slices were washed with ddH2O and then dehydrated on ice in 50%, 70%, 90%, 100%, 100% ethanol solutions for 10 min at each step. Brain slices were then washed twice with dry acetone and then placed into 50:50 Durcupan ACM:acetone overnight. Brain slices were transferred to 100% Durcupan resin overnight. Brain slices were then flat embedded between glass slides coated with mould-release compound and left in an oven at 60°C for 72 h.
The MicroCT tilt series were collected using a Zeiss Xradia 510 Versa (Zeiss X-Ray Microscopy) operated at 50 kV (80 µA current) with a x20 magnification and 0.5331 µm pixel size. MicroCT volumes were generated from a tilt series of 3201 projections using XMReconstructor (Xradia). SBEMs were accomplished using Merlin or Gemini SEM (Zeiss, Oberkochen, Germany) equipped with a Gatan 3View system and a focal nitrogen gas injection setup. This system allowed the application of nitrogen gas precisely over the block‐face of ROI during imaging with high vacuum to maximize the SEM image resolution. Images were acquired in 2.5 kV accelerating voltage and 1 µsec dwell time; 4 nm XY pixels, 50 nm Z steps; raster size was 17k x 17k and Z dimension was 861 images. Volumes were collected using 40% nitrogen gas injection to samples under high vacuum. Once volumes were collected, the histograms for the slices throughout the volume stack were normalized to correct for drift in image intensity during acquisition. Digital micrograph files (.dm4) were normalized and then converted to MRC format. The stacks were converted to eight bit and volumes were manually traced for reconstruction using IMOD.
We used a custom software to align and merge the SBEM data with its corresponding 3D fluorescence counterpart into one single composite volume. It enabled us to easily track and refine landmark features shared between the two modalities while their correspondence was being progressively established.
RV infected brain slices were post-stained with only 2% OsO4 to add electron density to the ferritin precipitates for the ferritin validation. After dehydration and embedding in the Durcupan resin block, 100nm thick sections were prepared with a Leica Ultracut UCT ultramicrotome and Diatome Ultra 45° 4mm wet diamond knife. Sections were picked up with 100 mesh gilder copper grids (G50, Ted Pella, Inc).
Sections were loaded into a JEOL JEM 3100EF equipped with an in-column Omega Filter and operating at 200kV. Sections were pre-irradiated at a low magnification of 100x for about 30 mins to stabilize the sample and minimize contamination. All zero-loss images were acquired with a hardware bin of 4 X 4 pixels using a Ultrascan 4000 CCD detector from Gatan (Pleasanton, CA, USA). Electron energy loss spectrum (EELS) was used for the detection of iron showing the presence of Fe L2,3 edge at 708eV and 721eV.