Dorsal skinfold chamber IVM
All animal experiments were approved by the local authorities (LANUV, State Agency for Nature, Environment and Consumer Protection). Prior to surgery, mice were anesthetized, and the dorsal skin was shaved and chemically depilated. A titanium dorsal skinfold chamber frame (small dorsal kit, APJ trading Co Inc) was surgically attached to the disinfected skin, and a circular window area (~ 12 mm diameter) was prepared for microscopy. For further details, see Currie et al. (2022).
After analgesia and 24 h of recovery, the epidermis was marked with a stamp (handcrafted), subdivided into 20 fields, thereby introducing the first artificial landmark (Fig. 2a, b). Inflammation was induced by initiating the reverse passive Arthus reaction, as described by Currie et al. (2022). Prior to anesthesia, mice were i.v. injected with 100 µL cell culture graded bovine serum albumin (75 µg BSA/g body weight) in phosphate buffered saline (PBS). Anesthetized mice subsequently received an intradermal injection of 20 µL rabbit anti-BSA (30 µg/mouse, MP Biomedicals) antibody into the stamp-labeled area of the back skin. Live cell imaging was performed using a Zeiss LSM 880 microscope with Airyscan fast. For this procedure, the anesthetized mouse carrying a dorsal skinfold chamber was mounted onto a custom-designed stage. Visualization of blood vessels and platelets was achieved by systemically injecting LysMeGFP mice (fluorescently tagged myelomonocytic cells) with Dylight antibodies against CD31 (30 μg/mouse, eBioscience #16–0311-85) and GPIbβ (2 μg/mouse, #X649, Emfret Analytics). To study the interaction of these immune cells, intravital time-lapse videos were recorded for 90–120 min (Fig. 1 a, Supplementary Movie). The confocal analysis focused on one distinctive postcapillary venule displaying strong neutrophil diapedesis and subsequent single-platelet binding to the vessel wall. At the end of the experiment, the anatomy of the monitored vessel segment (ROI) and the surrounding vascular network was well documented in microscopic z-stacks, using different magnification levels (z-stacks at 20× and 10×). To create an additional reference point, a photograph of the laser position within the frame of the introduced stamp was taken (Fig. 2b).
Sample fixation and introduction of additional landmarks for CLEM orientation
After confocal IVM, mice were terminally anesthetized. The glass cover of the dorsal skinfold chamber was removed, and the skin was submerged into fixative for 2 h in a dark room (4% formaldehyde (FA) in PBS, prewarmed to 37 °C). All further steps were performed in an unlit room to prevent bleaching of fluorescence signals in the tissue. The dorsal skinfold chamber was separated from the body of the mouse, and the skin surface (epidermis) surrounding the stamp was carefully perforated with blood lancets and acupuncture needles to allow the penetration of fluids during further fixation and embedding. Next, the stamped area was punched out (12-mm dermatology skin punch), and trimmed to minimal rectangular size with a scalpel, while keeping the ROI centered. One corner of the four-sided sample was removed, serving as an additional landmark for better orientation of the specimen. The flattened tissue was glued onto a carbon gridded coverslip with letters for orientation (coating mask for finder grid; Leica, Austria) using a thin layer of 2% low-melting-point agarose, with the vascularized side of the skin sample directly facing the carbon layer (Fig. 2e, f). The agarose was solidified on ice, and the specimen was transferred to a 3-cm glass-bottomed dish (ibidi, Germany). Using the position of the stamp and the edge with the cutoff corner as landmarks, the ROI was relocated in the confocal microscope. The localization of the ROI with respect to the carbon letters on the gridded coverslip was documented in z-stacks at different magnification levels. To record special anatomic features of the surrounding tissue serving as additional landmarks, the anatomy of the retrieved ROI and the surrounding vascular network was captured in microscopic z-stacks, using different magnification levels (z-stacks at 20× and 10×). Finally, the distance in the z-direction between the vessel of interest and the coverslip was measured. The sample was then further fixed in 2% glutaraldehyde (GA), 2% formaldehyde, in 0.1 M cacodylate buffer (CB), pH 7.2, prewarmed to 37 °C.
Epon embedding in beam capsule
For electron microscopy analysis, the sample was further processed with an adapted protocol that is also suitable for SBF-SEM (Deerinck et al. 2010) (Goudarzi et al. 2017), as summarized in Table 1. Contrast enhancement was specifically achieved by keeping the specimen on a warm heating plate at 40 °C during all incubation times starting with postfixation reagents (osmium tetroxide OsO4, thiocarbohydrazide TCH, lead aspartate; OTOTO) and followed by several washing steps. Uranyl (UA) en bloc staining was performed at 4 °C overnight, followed by dehydration in ethanol and acetone at room temperature and subsequent incubation with increasing concentrations of epon. The last epon infiltration solutions were mixed with 0.07% (w/w) Ketjenblack (KB; TAAB Laboratories Equipment Ltd) to improve the conductivity of the resin, as required for SBF-SEM (Nguyen et al. 2016). Finally, a beam capsule was attached above the ROI. After resin polymerization, the gridded glass coverslip was blown off by sequential plunges in liquid nitrogen and hot water. Ideally, the carbon imprint remained on the sample surface. See Table 1 for a more detailed description of the individual steps for the protocol.
Table 1 Step-by-step protocol for optimized mouse skin fixation and SBF embedding Target sectioning
The carbon imprint of letters and crosses on the block surface, now preserved as a mirror picture, is indispensable for retracing the ROI during ultrathin sectioning (Fig. 3). Therefore, all steps modifying the block need to be documented carefully (see also Fig. 3b).
First, unlike standard protocols, the block was trimmed to a cubic and not pyramidal shape, so that the dimension of the sections will not increase with further cutting into the specimen. Using a diamond knife (Ultratrim 90, Diatome, Switzerland), the size of the specimen was reduced to maximal 1 × 1 mm2 in the x–y direction. One corner of the block was removed, creating a new reference point, as the carbon mask disappears with the first sections. The forward trimming and sectioning steps were all documented. At levels of interest, sections of 200 nm thickness were collected and stained with toluidine blue. Samples were inspected with a light microscope to identify structures within the vascular network that match the anatomical landmarks of the vasculature recorded by confocal fluorescence microscopy (z stacks). In the ROI, the overlay of all images (carbon grid, fluorescence image, light microscopy and electron microscopy image) verified that the target region was approached successfully.
Transmission electron microscopy
Thin sections of 60 nm of the ROI were collected on 1 slot-filmed copper grids, counterstained with lead, and imaged in the electron microscope (Tecnai 12-biotwin, Thermo Fisher Scientific Inc.). No counterstaining was needed if the samples were processed according to the OTOTO protocol. Additionally, consecutive serial sections of 60 nm or 200 nm thickness were collected and analyzed to provide more preliminary “volume” information in 3D (Fig. 4a, b). Representative images were taken with a 2k charge-coupled device (CCD) camera (Veleta, EMSIS, Münster, Germany) and arranged with Adobe Photoshop without further processing.
Serial block-face scanning electron microscopy
Preselected correlative samples were trimmed to a size of approximately 500 × 500 × 500 µm3 and mounted on an aluminum specimen holder using a two-component silver conductive epoxy adhesive in a ratio of 1.25:1 (Wanner et al. 2016). The samples were coated with a 30-nm thin layer of gold in a sputter coater (ACE EM 600, Leica, Germany). The trimmed sample was mounted in the sample holder of the SEM (JSM-7200F, JEOL, Japan) integrated ultramicrotome stage (3View2XP, Gatan, USA). To minimize the imaging time, the trimmed sample block was aligned such that the course of the blood vessel was as parallel to the knife edge as possible, based on the previous TEM analyses (Fig. 2). An additional helpful tool in aligning the sample block was the reference landmark (cut-off edge on one side of the square sample block).
Optimizing imaging parameters for SBF-SEM is highly dependent on the type of sample. To achieve the best contrast and resolution possible, both high accelerating voltages and high vacuum are desirable. However, this approach could increase the possible charge accumulation in regions of unstained and nonconductive biological tissue or empty resin areas. Whereas backscattered electrons are detected for the final image acquisition, secondary electrons are more prone to charging artifacts (Kim et al. 2008). Utilizing the secondary-electron detector unit within the microscope, charge accumulation manifested as intense brightness can therefore be detected more easily and precautions be taken. The sample investigated here proved to be stable under imaging parameters of 1.5 kV accelerating voltage, utilizing a 30 nm condenser aperture, HV conditions of 10 Pa, and a positive bias of 400 V. The pixel resolution of each ROI was set to 5 nm, in between ablation of 50-nm slices with a dwell time of 2 µs. The multi-ROI acquisition was controlled by Gatan Digital Micrograph software (version 3.32.2403.0). The same was used to convert the raw data (.dm4) into 16-bit TIFF files using a custom script command. Further processing of the datasets, including alignment, rendition, and extraction of subvolumes and segmentation, was handled using Microscopy Image Browser (version 2.7) (Belevich et al. 2016). To keep the considerable data size to a minimum, all datasets were converted to 8-bit TIFFs and binned to a factor of 2 or 4. In case high resolution was required, an unbinned subvolume of the ROI was isolated from the raw dataset. To keep imaging times short, the dwell times during acquisition can be reduced at the higher noise ratios. The noise can be removed via edge-preserving filters, e.g., anisotropic diffusion. Subsequent graylevel image histogram normalization can restore sufficient contrast for semiautomatic segmentation tools (e.g., watershed or global threshold) based on specific gray values. The final models and image stacks were exported for volumetric visualization in Amira 3D (version 2021.1, Thermo Fisher Scientific Inc.).