The main equipment used to set up the technique was as follows: Vibratome (Leica VT 1200S); Variable Wattage Microwave (PELCO BioWave Pro 36,500–230); Ultramicrotome (Leica EM UC6); Diamond Knive (Diatome Histo #5961); Sputter Coater (Quorum Emitech SC7620); and Focused Ion Beam – Scanning Electron Microscope (FIB-SEM; Zeiss, CrossBeam 540).

0.1 M phosphate buffer solution (PB): the solution contains 2.65 g of sodium di-hydrogen phosphate 1-hydrate (PanReac #131965) and 14 g of di-potassium hydrogen phosphate (PanReac #121512) in 1 L of distilled H2O; pH 7.4.

Perfusion fixation solution: 4% paraformaldehyde (PFA; Electron Microscopy Sciences #15714-S) in PB. The solution must be prepared just before use in a fume hood.

First postfixation solution: 4% PFA in PB. Prepare just before use in a fume hood.

Sectioning solution: 10% sucrose (PanReac #57501) in PB.

Cryoprotection solution: 30% sucrose in PB.

Preincubation solution: 3% bovine serum albumin (BSA; Sigma #A4503-50G) in PB.

Primary antibody solution: rabbit anti-Vesicular GABA Transporter Antibody (VGAT; Synaptic Systems #131003; 1:2000) and 3% BSA in PB.

Positive control of primary antibody solution: rabbit anti-parvoalbumin (PV; ABCAM #AB11427; 1:1000) and 3% BSA in PB.

Secondary antibody solution: biotinylated goat anti-rabbit IgG antibody (Vector Laboratories #BA-1000; 1:200) and 3% BSA in PB.

Avidin-Biotin Complex (ABC)-based detection method: the solution contains 0.008% reagent A (Avidin; ABC Elite) and 0.008% reagent B (biotinylated HRP, ABC Elite) from the ABC kit (Vector Laboratories #PK-6100) in PB. Prepare 30 min before use.

Preincubation solution of 3,3′-Diaminobenzidine (DAB): 0.05% DAB (Sigma #D5905) in PB. Prepare immediately before use in a fume hood and protect from light. Filter with a syringe filter (Acrodisc 0.2 μm, #4612) before use.

Incubation solution of DAB: immediately before use, add 0.01% hydrogen peroxide (H2O2; Merck # 1.07209.1000) to the DAB solution described above and mix well.

Second postfixation solution: freshly prepared 4% PFA, 0.2% glutaraldehyde (GA; TAAB #G002), and 0.003% calcium chloride (CaCl2; Sigma #C-2661) in 0.1 M cacodylate buffer (Sigma #C0250). Prepare in a fume hood.

Microwave postfixation solution: freshly prepared 2% PFA, 2.5% GA, and 0.003% CaCl2 in 0.1 M cacodylate buffer. Prepare in a fume hood.

First osmium solution — prepared with or without potassium ferrocyanide: 1% OsO4 (Sigma #O5500), 0, 0.1% or 1% potassium ferrocyanide (Probus #23345) and 0.003% CaCl2 in 0.1 M cacodylate buffer. Always handle osmium and potassium ferrocyanide in a fume hood, with protective glasses and double gloves.

Second osmium solution: 1% OsO4 and 0.003% CaCl2 in 0.1 M cacodylate buffer. Always handle osmium in a fume hood, with protective glasses and double gloves.

Uranyl acetate solution for en bloc staining: the solution contains 1% uranyl acetate (Electron Microscopy Sciences #22400) in 50, 70, 90 and 100% ethanol. Filter with a syringe filter (Acrodisc 0.2 μm, #4612).

Silver paint (Electron Microscopy Sciences, #12630).

We used four adult female mice (C57BL/6, 8 weeks old) for the technique outlined in this study. Two of these mice were utilized to assess various concentrations of potassium ferrocyanide, with one mouse assigned to each condition. The remaining two mice were dedicated to VGAT validation — one with potassium ferrocyanide (0.1%) and the other without potassium ferrocyanide. All animal handling procedures were conducted in accordance with the guidelines for animal research outlined in the European Community Directive 2010/63/EU, and all procedures were approved by the Local Ethics Committee of the Spanish National Research Council (CSIC).

To begin the procedure, anesthetize the animals with an intraperitoneal injection of pentobarbital (40 mg/kg) and then intracardially perfuse with 100 mL of freshly prepared fixation solution (4% PFA in 0.1 M PB). Postfix the brains for 6 to 16 h (overnight) in the postfixation solution (4% PFA in 0.1 M PB). Then, cut the brains into sections (150 μm thick) using a vibratome and collect them in a sectioning solution (sucrose 10%, in 0.1 M PB) in 24-well flat-bottom plates.

This part of the procedure was carried out as follows: (1). Postfix the sections for 48 h at 4°C in the second postfixation solution (4% PFA, 0.2% GA and 0.003% CaCl2 in 0.1 M cacodylate buffer). (2) Wash the sections (three times, 10 min each) in 0.1 M cacodylate buffer. (3) Perform microwave postfixation by placing the sections in the microwave postfixation solution (2% PFA, 2.5% GA, and 0.003% CaCl2 in 0.1 M cacodylate buffer) for 1 min at 50°C using the variable wattage microwave at 150 W power. Carefully add the fixative using a plastic Pasteur pipette without agitating the sections to prevent curling or folding. This step should be conducted in a fume hood. (4) Wash the sections three times in 0.1 M cacodylate buffer, 10 min each wash. (5) Osmicate the sections for 1 h in the first osmium solution in a fume hood. Slowly add and remove the osmium solution using a plastic Pasteur pipette to avoid folding or breaking the sections. Note that during osmication, sections become brittle and should be handled with care, using a small spatula or weighing spoon. (6) Wash the sections three times in 0.1 M cacodylate buffer, 10 min each wash. (7) Osmicate the sections again for 1 h with the second osmium solution in a fume hood.

Using a variable wattage microwave at 50°C, 250 W power, dehydrate the sections in a series of uranyl acetate solutions, starting with 50% ethanol and continue with a solution of 1% uranyl acetate in increasing ethanol concentrations (50–70%-90–100%), finishing with absolute ethanol and clear three times in acetone (40 s each step). Embed the sections in Araldite, with a variable wattage microwave (under vacuum conditions at 70°C, 350 W power, 3 min each step), as follows: solution of 1 part Araldite, 1 part acetone — followed by a solution of 4 parts Araldite and 1 part acetone and finishing with a 3-step embedding with pure Araldite. Store the sections embedded in pure Araldite at 4°C for 8–16 h (overnight).

Temper the sections for 30 min at room temperature. Change the Araldite to a freshly prepared mixture and leave the sections to rest for three to 4 h. Flat-embed each section by placing them between two silicone coater slides covered with a transparent film for 48 h at 60°C. To ensure that the flat-embedding is homogenous, distribute small weights over the slide.

Once the resin has cured, the flat-embedded sections must be examined and photographed under an optical microscope to select the region of interest. Then, trim and glue the region of interest (in this case: the primary somatosensory cortex) with cyanoacrylate onto a blank Araldite block. Use a microtome and a diamond knife to obtain serial semithin sections, until reaching the tissue. Photograph the surface of the block to establish landmarks (such as blood vessels or other morphological features), which will later be used to precisely locate the area to be imaged with the FIB-SEM.

Once the region of interest has been selected in the Araldite block, mount it on an SEM specimen stub with a conductive carbon sticker (Electron Microscopy Sciences, #77825–09). To prevent charge build-up, the block must be covered with silver paint, except for the top surface. It is important not to cover or spill silver droplets on the upper surface of the block where the specimen is located. Conversely, the base of the block must be carefully painted to ensure electrical continuity between the Araldite block and the specimen stub. Allow the paint to dry for at least 24 h in a vacuum desiccator. Charge dissipation from the upper surface of the block is achieved by gold–palladium sputter-coating for 60 s. Carbon, gold alone, or other metals are also suitable for sputter coating, but care must be taken not to cover the specimen with a layer that is too thick as this might obscure surface details.

The surface of the block is then photographed with the SEM using the secondary electron detector. The landmarks in the section that were previously identified with the optical microscope (mainly small blood vessels) are also visible with the SEM, so the region of interest can be precisely located. A viewing trench is then excavated with the FIB using a 7 nA milling current, to provide visual access to the region that we plan to image. The front face of this trench must be located close enough to the target to allow its identification. The ion beam and the electron beam can be used simultaneously, so it is possible to monitor the progression of the trench as it is being excavated. As soon as we have identified our target, milling of the viewing trench is stopped. We then use a smaller FIB current (700 pA) to progressively mill the front face of the trench in steps of 20 nm. During each milling step, we remove 20 nm of material with the FIB, and then use the SEM to take a microphotograph of the freshly milled surface. In our equipment, the angle between the SEM and the FIB is 54°, so the angle of incidence of the SEM on the surface to be imaged is 36°, rather than perpendicular. The resulting perspective deformation is automatically corrected by the microscope software during acquisition (SmartSEM 6.02; Carl Zeiss Microscopy Ltd.), so no distortion is present in the final images.

Since the milling/imaging cycle can be fully automated, serial images of the target are obtained. We routinely use a milling step of 20 nm (equivalent to section thickness) and a resolution in the X-Y plane of 5 nm/pixel, so the actual voxel size is 5 nm × 5 nm × 20 nm (Merchán-Pérez et al., 2009). Other resolutions and milling steps can also be used, depending on the particular imaging needs, and the length of the series of sections can be selected according to the researcher’s needs.

Some drift may occur during the acquisition of the FIB-SEM image series. In this case, further alignment (registration) is required. For the registration process, programs such as FIJI, a distribution of ImageJ with preinstalled plugins for microscopy (Schindelin et al., 2012), can be used. We recommend setting a “rigid” registration protocol, with translation only allowed for the alignment, as this avoids deformation and rotation of individual images. The aligned stack of images is then visualized in EspINA software (Morales et al., 2011), which allows synaptic identification and segmentation through the original plane of section or the other two orthogonal planes (EspINA Interactive Neuron Analyzer, 2.9.12; https://cajalbbp.csic.es/espina-2).

A synapse is recognized according to well-established criteria (e.g., see Colonnier, 1981; Peters et al., 1991; Peters and Palay, 1996). The identification process involves confirming the presence of specific elements, including densities on the cytoplasmic faces in the pre-and postsynaptic membranes; synaptic vesicles in the presynaptic axon terminal adjacent to the presynaptic density; and a synaptic cleft. Generally, three types of structural units are employed for synapse identification (see Mayhew, 1996, for a review): terminal boutons, total apposition zones, and synaptic membrane densities. In this study, we primarily use synaptic membrane densities for synapse counting, especially when accompanied by synaptic vesicles near the presynaptic density, irrespective of the angle of section through which the synaptic junctions are viewed (i.e., whether a synaptic cleft is evident or not). Moreover, the identification of synapses relies on examining all serial sections where each individual synapse is visible. Additionally, utilizing EspINA software, the 3D course of the axons can be followed within stacks of sections to confirm the nature (AS or SS) of the synapses established in all their synaptic contacts.

We conducted pre-embedding immunocytochemical labeling of VGAT in fixed brain tissue sections and subsequently processed the tissue for FIB/SEM, as described above, with the following modifications to investigate the synaptic contacts established by VGAT-positive boutons. After vibratome sectioning, 150 μm-thick sections are cryoprotected using sucrose 30% in 0.1 M PB overnight. Permeabilize sections using liquid nitrogen. Place two to three brain sections in 5 mL Eppendorf tubes. Then, remove the remaining sucrose solution and ensure that the sections are distributed along the Eppendorf tube surface, clearly separated and stretched. Immerse the Eppendorf tubes in the liquid nitrogen solution for 2 or 3 s. Immediately after the cryopermeabilization, store the Eppendorf tubes at 4°C. The sections must have a white, opaque appearance. Once the sections return to their usual transparent appearance, slowly add 0.1 M PB (4°C) to the Eppendorf tube and store again at 4°C.

Carefully place the permeabilized section in 24-well flat-bottom plates filled with 0.1 M PB. Wash the sections (three times, 10 min each) in 0.1 M PB, under agitation. Then, pre-incubate the sections with the preincubation solution (3% BSA, in 0.1 M PB) for 2 h under agitation at room temperature. Incubate with the primary antibody incubation solution (rabbit anti-Vesicular GABA Transporter Antibody, in 3% BSA - 500 μL/section) for 48 h under agitation at 4°C. A positive control is recommended to exclude any possible miscoupling during the antibody reaction.

Allow the sections to temper for 10–15 min at room temperature. Wash the sections (three times, 10 min each) with 3% BSA, in 0.1 M PB. Incubate the sections with the secondary biotinylated antibody solution (biotinylated goat anti-rabbit IgG antibody, in 3% BSA), for 2 h under agitation at room temperature. Wash the sections (three times, 10 min each) with 0.1 M PB. Incubate the sections with the Avidin-Biotin Complex (ABC)-based detection kit to amplify the secondary antibody signal, for 1 h under agitation at room temperature. Wash three times in 0.1 M PB at room temperature, 10 min each time. In a fume hood, pre-incubate the sections in DAB solution without H2O2, protected from light. Next, incubate in DAB solution with H2O2 for 1 min. The sections will change to a brown, whiskey-like appearance, so visually monitor the color of the sections until the precipitate has reached the desired intensity. This can also be checked using an optical microscope. Stop the reaction by washing the sections three times (10 min each) in 0.1 M PB. Once the immunostaining is checked, follow the processing procedure for electron microscopy: postfixation and osmication are performed as described above, but adding 7% glucose (Merck #1.08337.0250 in the first and second osmium solutions to avoid excessive darkening of the sections). The first osmium solution contains 0.1% potassium ferrocyanide. Dehydration, en bloc staining and embedding are performed as described above.

To assess the impact of different concentrations of potassium ferrocyanide on the appearance of synaptic junctions, we conducted the study on layers II and III. A detailed examination of 201 slices (covering a total volume of 361.68 μm³) from an image stack treated with a concentration of 1% potassium ferrocyanide revealed excellent EM image quality, primarily due to clearly thickened membranes. However, identifying SS proved challenging as their thin PSD exhibited a thickness similar to the surrounding non-synaptic membranes. In the case of AS, they remained visible, but their PSD appeared thinner compared to the thicker surrounding membranes (Figure 1; Supplementary Figures S2.1A, S2.2A, S2.3A, S2.4A). Consequently, we explored lower concentrations of potassium ferrocyanide. We examined 299 slices (covering a total volume of 538.02 μm³) from an image stack treated with a concentration of 0.1% potassium ferrocyanide. As illustrated in Figure 2 (see also Supplementary Figures S2.1B, S2.2B, S2.3B, S2.4B), at a concentration of 0.1%, the quality of the EM images remained excellent, and AS and SS could be clearly distinguished through serial sections. In this image stack, we unambiguously identified 88 SS and 836 AS.

Next, brain sections that were labeled for VGAT and processed for EM using 0.1% potassium ferrocyanide were imaged using FIB-SEM to examine the morphology of the synaptic junctions established by VGAT-positive boutons — and to compare with the morphology of the synapses of unlabeled axon terminals. As has been shown previously (e.g., Takayama and Inoue, 2010), VGAT immunoreactivity was distributed across all layers of the mouse primary somatosensory cortex, where numerous stained puncta were scattered in the neuropil (Supplementary Figure S1A). In layers II to VI, especially in layer V, positive puncta were distributed both in the neuropil and around unlabeled somata and their proximal processes (Supplementary Figure S1B).

We conducted FIB/SEM analyses in the region that exhibited strong VGAT-immunoreactivity using correlative light-electron microscopy methods. A viewer trench was excavated with the FIB (Supplementary Figure S1C) to image the tissue within the penetration zone of the immunostaining. In this zone, VGAT-positive axon terminals filled with dark immunostained vesicles can be visualized (Figure 3). The intensity of the staining in these terminals decreases as the distance to the surface of the section increases (Figure 3B).

These VGAT-positive terminals established SS, while VGAT-negative axon terminals established AS (Figure 4).

Further verification of the morphology of synaptic contacts made by VGAT-positive boutons in the neuropil was obtained by examining the perisomatic innervation of pyramidal cells by VGAT-immunoreactive axon terminals, where it is known that only SS are established (DeFelipe and Fariñas, 1992). As shown in Figure 5, these perisomatic axon terminals clearly established SS, as expected, and were identical to those found in the neuropil (Figure 4).

We analyzed 266 serial images in the neuropil within the penetration zone of immunostaining, corresponding to 4,184 μm³, and identified 265 AS and 23 SS. All SS (8% of total synapses) were formed by VGAT-positive terminals, while all AS (92%) were established by VGAT-negative terminals (Figure 6).

This aligns with the findings of Turégano-López et al. (2021), who studied VGAT-positive boutons using FIB/SEM without the use of potassium ferrocyanide. As depicted in Figure 7, the morphology of the synaptic junctions formed by VGAT-positive boutons in this material (without potassium ferrocyanide) is similar to those observed in brain sections using 0.1% potassium ferrocyanide (Figure 4). Therefore, it is recommended to use the latter concentration of potassium ferrocyanide, i.e., 0.1%.