Animals
All experimental procedures involving the use of animals were carried out in accordance with European Union (EU) Directive 2010/63/EU and under the approval of EMBL Animal Use Committee and Italian Ministry of Health License 541/2015-PR to C.G. All animal experiments complied with the ARRIVE guidelines. Adult mice (mus musculus) were singly housed in temperature and humidity-controlled cages with ad libitum access to food and water under a 12 h/12 h light-dark cycle. C57BL/6J mice were obtained from local EMBL colonies. Vglut2::ires-Cre and Vgat::ires-Cre mice (JAX stock no. 028863 and 028862) were used in heterozygous state. All mice were on a C57BL/6J congenic background. Female and male mice were used in this experiment (Table S1).
Viral production
pAAV-Ef1α::DIO-IGK-dAPEX2-KDEL and pAAV-Ef1α::COX4-dAPEX2 were purchased from Addgene as a gift from David Ginty (Addgene plasmid #117183; http://n2t.net/addgene:117183; RRID:Addgene_117183; Addgene plasmid # 117176 ; http://n2t.net/addgene:117176 ; RRID:Addgene_117176) and packed in chimeric serotype 1/2 AAV vectors by the EMBL Genetic & Viral Engineering Facility. Virus titers were 1.1x1014 and 1.3 x1014 vg/ml, respectively.
Stereotaxic surgeries
Mice were anesthetized with 5% isoflurane and subsequently head fixed in a stereotaxic frame (RWD Life Science) with body temperature maintained at 37°C. Anesthesia was sustained with 1 to 2% isoflurane and oxygen. The skull was exposed, cleaned with hydrogen peroxide (0.3% in ddH2O) and leveled. Craniotomy was performed with a handheld drill. AAV1/2-Ef1α::COX4-dAPEX2 was infused in the ACC (1.11, 1.71 and 1.41 mm anterior to bregma, 0.45 mm left to bregma and 2.0 mm ventral to the skull surface) and AAV1/2-Ef1α::DIO-IGK-dAPEX2-KDEL was infused in the dPAG (4.30 posterior to bregma, 1.50 left to bregma and 2.55 mm ventral to the skull surface, at a 20º lateral angle). Injections were unilateral and ~0.2 µl of virus was delivered in each injection site with a pulled glass capillary (intraMARK, 10-20 µm tip diameter, Blaubrand). After virus infusion the skin was sutured and saline solution and Carprofen (5 mg/kg) administered subcutaneously. Mice were allowed 8 weeks for viral expression.
Perfusion and sectioning
Tissue was prepared following a protocol described previously10 with minor modifications. Mice were anesthetized intraperitoneally with 2.5% Avertin (Sigma-Aldrich) and perfused transcardially with warm (~37°C) Ames medium (MilliporeSigma) with heparin (MilliporeSigma) followed by warm (~37°C) 2.5% glutaraldehyde (Electron Microscopy Sciences), 2% paraformaldehyde (Electron Microscopy Sciences) in cacodylate buffer (0.15 M sodium cacodylate [Electron Microscopy Sciences] and 0.04% CaCl2 [MilliporeSigma]). Ames medium was gassed with Carbogen for animals 7390, 7930, OBO-017266, and 8845. Brains were collected and post-fixed overnight at 4°C with the same fixative solution. Brains were washed with cacodylate buffer and embedded in 2% low melting point agarose prior to sectioning. ACC and PAG Coronal 100-150 µm sections were cut on a vibratome (Leica Microsystems) in cacodylate buffer.
DAB staining and bright field microscopy
ACC and PAG sections were washed 2 × 10 min with cacodylate buffer with 50 mM glycine (MilliporeSigma), 1 × 10 min with cacodylate buffer, and then incubated in 1 mL of DAB (MilliporeSigma; 0.3 mg/mL) in cacodylate buffer in the dark for 30 min at RT. Afterwards, 10 ul of 0.3% H2O2 (MilliporeSigma) in cacodylate buffer were added and let react for 1h in the dark at RT. Sections were placed in 3% glutaraldehyde in cacodylate buffer at 4°C overnight. PAG, DAB stained brain slices were washed1×10 min in cacodylate buffer, 1×10 min in 50 mM glycine in cacodylate buffer, and 2×10 min in cacodylate buffer. Then they were placed in a slide, freely floating in cacodylate buffer, covered with a coverglass and imaged with a Leica Thunder Imager Cell Culture microscope. 3-4 PAG slices per mouse containing satisfactory putative ER-dAPEX2 labels, or at ca. Bregma AP -4.30 in the case of control brains, were trimmed around PAG and selected for further staining. Remining PAG slices and ACC slices were washed in ddH2O, mounted with mowiol (Calbiochem) and imaged in an Olympus Slideview VS200 microscope.
Electron microscopy staining
Selected PAG tissue sections were processed following the rOTO protocol26 as described previously10. Slices were stained in 2% osmium tetroxide (Electron Microscopy Sciences) in cacodylate buffer for 1 h at room temperature, followed by 2.5% potassium ferrocyanide in cacodylate buffer for 1h, and washed 4 x 5 min in ddH2O. Subsequently, sections were incubated in 1% thiocarbohydrazide (Electron Microscopy Sciences) in ddH2O at 40°C for 15 min and washed 4 x 5 min in ddH2O. Samples were stained in 2% osmium tetroxide in ddH2O for 1 h at room temperature and washed 4 x 5 min in ddH2O. Sections were counterstained overnight with 1% uranyl acetate in 0.05 M sodium maleate (MilliporeSigma; pH 5.15; in samples 6 and 8, uranyl acetate was diluted in ddH2O). Sections were warmed to 50°C for 2 h in the uranyl acetate solution and then washed 4 x 5 min in ddH2O. Sections were dehydrated in 30%, 50%, 70%, 90% and 3x 100% ethanol, 5 min each, and in propylene oxide (Electron Microscopy Sciences) for 15 min. Once samples were dehydrated, they were infiltrated for 1 h each with 25%, 50%, 75% and 100% Durcupan ACM resin (Electron Microscopy Sciences). Next, samples were placed on a Durcupan block with a flat surface and covered with ACLAR® 33 C Film (Electron Microscopy Science). Samples were let polymerize for 72 h at 60°C. The majority of the reagents used for electron microscopy staining are highly toxic. Consulting the safety datasheets of all products is highly recommended.
Transmission electron microscopy (TEM)
Sections (70 nm) were cut from polymerized PAG samples using an ultramicrotome (Leica UC7). Sections were imaged without post-staining using a Philips CM120 Biotwin operated with an acceleration voltage of 100kV.
Focused ion beam electron microscopy (FIBSEM)
Flat-embedded PAG sections were glued to the lateral side of pre-polymerized blocks. Samples were trimmed to expose the region-of-interest (Leica UC7) using a 90º diamond trimming knife (Diatome cryotrim 90). Samples were mounted on a stub using silver epoxy resin (Ted Pella) with the sections perpendicular to the stub surface so that they were parallel to the milling beam. Samples were gold sputter coated (Quorum Q150RS) and FIBSEM imaging performed with a Zeiss Crossbeam 550 using Atlas3D (Fibics & Zeiss) for sample preparation and acquisition. Briefly, the surface above the region-of-interest (typically 50 x 50 µm) was protected with a platinum coat, deposited with 3 nA beam current. Auto-tuning lines were milled on the platinum surface and a carbon coat was deposited on top. Due to the flat embedding and perpendicular mounting of the section a short polishing step was enough to remove the thin layer of empty resin before exposing the embedded tissue. During the stack acquisition the milling was done with 1.5 nA beam current (at 30 kV). For imaging the SEM was operated at 1.5kV/700pA using an ESB detector (collector voltage 1100V). All stacks were acquired with a 10 nm isotropic voxel size. With these settings we acquired volumes of 15 x 30 x 30/55 µm.
Serial blockface electron microscopy (SBEM)
Flat-embedded PAG sections were mounted on a pin stub using silver conductive epoxy resin (section parallel to stub surface) and trimmed to the right size around the region-of-interest. The sample was then imaged with a Zeiss GeminiSEM 450 equipped with a Gatan 3View system and controlled by the open-source software package SBEMimage43. To reduce charging artifacts we used a focal charge compensation device (Zeiss). Images were taken at 1.5kV/300pA and 1.6 µs dwell time with a pixel size of 10 x 10 nm and 40 nm cutting thickness.
Image preprocessing
The image stacks were registered using the Fiji44 plugin “Linear Stack Alignment with SIFT”45 (transformation: translation) or AMST workflow46 and averaged in the z-axis to reduce noise and image size, producing a final voxel resolution of 10 x10 x 20 nm. This resolution was sufficient to identify synaptic vesicles and to trace axons in 3D. Lookup tables (LUT) were inverted and images were saved as a 8-bit tiff stack. Selected 2D planes were denoised using the Fiji plugin Noise2Void47 (https://imagej.net/plugins/n2v) for visualization purposes, but analysis was performed on raw data.
Data analysis and statistics
Aligned, scaled, and LUT inverted tiff stacks were opened using Fiji44 and a binary mask was imposed by thresholding pixel values using the threshold Fiji function. Pixel threshold was selected based on one manually seleted reference stained mitochondrion per sample. Mitochondria average pixel value was extracted by selecting one representative 2D plane for each analyzed mitochondrion, using the ROI manager function in Fiji44. Processes containing stained mitochondria and ER were traced and categorized manually. 3D reconstruction of selected processes was carried out using 3DMOD (http://bio3d.colorado.edu/imod/). Data visualization and statistics were done using custom Python scripts (Python Software Foundation). No statistical methods were used to predetermine sample sizes. Sample assignment was not randomized. Data collection and analysis were not performed blind to the conditions of the experiments.
Data availability
All FIBSEM datasets (samples 3 to control 8; Table S1) were deposited as tiff files after preprocessing in the Electron Microscopy Public Image Archive (EMPIAR; EMPIAR-10883)48 and uploaded to the interactive viewer MoBIE49 (https://github.com/mobie/mobie-viewer-fiji) with bookmarks directing the viewer to all figure captions. MoBIE projects can be visualized with the MoBIE Fiji44 plugin (https://imagej.net/plugins/mobie) under MoBIE>Open>Open Published MoBIE Project with the name ‘PAG-dAPEX2-FIBSEM’. TEM, SBEM, and raw FIBSEM data is available upon reasonable request.