5.1 Animals
All animal experimental procedures were performed according to animal experimental guidelines of the Experimental Animal Management Ordinance of Hubei Province, China, and the recommendations from the Huazhong University of Science and Technology, and were approved by the Institutional Animal Ethics Committee of Huazhong University of Science and Technology. 8-week-old female Thy-1-EGFP, Thy-1-YFP and Cx3cr1EGFP/+ mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA) and housed and bred in the Wuhan National Laboratory for Optoelectronics in a normal cycle (12 h light/dark). 8-week-old female wild-type BALB/c and C57 mice were supplied by the Wuhan University Center for Animal Experiment (Wuhan, China). Transgenic mice expressing fluorescent proteins under the influence of the Thy1 promoter (Thy1-EGFP and Thy1-YFP) were used for dendritic spine imaging, and those expressing the enhanced green fluorescent protein in microglia (Cx3cr1EGFP/+) were used for microglia imaging. The wild-type BALB/c mice were used for non-stained cerebral blood flow/blood oxygen imaging and FITC-stained vasculature fluorescence imaging. The wild-type C57 mice were used for GCaMP6s-stained calcium imaging.
5.2 Reagents for TIS
The TIS window was established using 3 reagents: S1 (a saturated supernatant solution of 75% (vol/vol) ethanol (Sinopharm, China) and urea (Sinopharm, China) at room temperature), S2 (a high-concentration sodium dodecylbenzenesulfonate solution prepared by mixing a 0.7 M NaOH solution (Aladdin, China) with dodecylbenzenesulfonic acid (Aladdin, China) at a volume-mass ratio of 24:5), and S3 (a UV-curable adhesive, ergo 8500, Kisling, Switzerland). The refractive indices (RI) of S1, S2, and S3 are 1.399 ± 0.0008 and 1.364 ± 0.0003, and 1.513 ± 0.0016, respectively.
5.3 Establishment of the TIS window
Mice were anesthetized before the TIS window establishment. For longitudinal studies, mice were anesthetized with isoflurane (1.5%) in air. For the TBI study, mice were anesthetized with a mixture of 2% α-chloralose and 10% urethane (8 mL/kg). As shown in Additional file 1: Fig. S16a-b, to establish the TIS window, a midline incision was made on the scalp in the direction of the sagittal suture. A holder with a 1 cm diameter hole at the center was glued to the skull, and the mouse was immobilized on a custom-built plate. Next, S1 was applied to the exposed skull for 10 min to dissociate the collagen in the skull. S1 was then removed, and S2 was smeared on the skull for 5 min to eliminate lipids in the skull. S2 removal was followed by the application of a small quantity of S3 to the skull for RI matching; S3 provided the exposed skull with a thin layer for cover, and a 1-mm coverslip was placed on the surface of S3. Lastly, the skull area was irradiated with a UV LED for 2 min, solidifying S3. Thus, the TIS window was established (Additional file 1: Fig. S15c) and used for cortical observation either immediately or later for long-term scrutiny. To measure the thickness of the cured S3, after establishment of TIS window, the cured S3 was carefully removed, along with the cover slide, from the skull, and the total thickness of S3 and cover slide was measured to be 230 μm by a vernier caliper. Then, the thickness of the cured S3 was calculated as 230 μm-150 μm (the thickness of the cover slide) = 80 μm.
As shown in Additional file 1: Fig. S15d, the TIS window provides a large field of view without craniotomy and, therefore, will not cause the immune response that is often induced by removing a large chunk of the skull. While the thinned-skull window also steers clear of inflammation, it only provides a much smaller field of view than the TIS window.
5.4 In vitro evaluation of the TIS window technique
Fresh skull samples (0.4 × 0.4 mm2) excised from mice were positioned in front of a spectrometer (USB4000-VIS–NIR, Ocean Inslight, USA) to evaluate the transmittance enhancement of the skull by the TIS window technique. A broad-spectrum white light source (HL-2000, Mikropack, Germany) was used to measure the transmittance spectrum (400–1000 nm) of the intact skull. The skulls were treated gradually following the steps used to establish the TIS window, and changes in the transmittance spectrum were recorded (Additional file 1: Fig. S1a).
Fresh skulls from mice were cut in half down the middle to macroscopically test the imaging quality through the TIS window. The left parts were immersed in S1, S2, and S3, followed by a 2-min UV irradiation, and the right parts were used as control. Lastly, two parts of the skulls were placed on grid paper and photographed with a camera (Additional file 1: Fig. S1b).
Fresh mice skull samples (0.4 × 0.4 mm2) were put on a 1951 United States Air Force (USAF) resolution test target and treated with S1, S2, S3, and UV irradiation in sequence. A camera (AxioCamHRc, Zeiss, Germany) attached to a stereomicroscope (Zeiss Axio Zoom. V16, Zeiss, Germany) was used to capture white-light images of the skulls before and after the application of the TIS technique to quantificationally determine the increase in imaging resolution (Additional file 1: Fig. S1c).
5.5 Dual-modal imaging system for blood flow/blood oxygen imaging
A home-built dual-modal optical imaging system[54] was used to capture blood flow and blood oxygen distribution in cortical vasculature to evaluate the efficacy of the TIS window in vivo. The dual system consisted of LSCI and hyperspectral imaging (HSI). These two imaging modes share a stereoscopic microscope, and each has a light source and a detection system. For LSCI, a He–Ne laser (632.8 nm, 3 mW) beam passing through an adjustable optical attenuator was used to illuminate the areas of interest after expansion. A charge coupled device (CCD) camera (Pixefly, PCO GmBH, Germany) mounted on a stereomicroscope recorded a sequence of raw speckle images. For HSI, a ring-like LED light with a polarizer was used for illumination. A liquid crystal tunable filter (LCTF, Perkin Elmer, USA) was placed before a CCD camera in another channel to split wavelengths. The laser speckle temporal contrast[55] and multiple linear regression analysis methods[56] were used to calculate blood flow and blood oxygen, respectively.
The contrast-to-noise ratio (CNR) was used to assess the imaging quality of LSCI and was defined by the following equation [57]:
$$\mathrm{CNR}=\frac{|{C}_{vessel}-{C}_{background}|}{\sqrt{{f}_{vessel}{\sigma }_{vessel}^{2}+{f}_{background}{\sigma }_{background}^{2}}}$$
where \({C}_{vessel}\) and \({C}_{background}\) were the mean contrast values of the blood flow and background, respectively; \({\sigma }_{vessel}^{2}\) and \({\sigma }_{background}^{2}\) were the variances in the contrast values of the blood flow and background, respectively; \({f}_{vessel}\) and \({f}_{background}\) were the fractions of pixels classified as blood flow and background in all the selected pixels in the speckle contrast image, respectively.
5.6 TBI modeling and in vivo immune cell observation
8-week-old female BALB/c mice were anesthetized, and their scalps were cut open to expose the skull surface. Body temperature was monitored continuously with a rectal probe and maintained at 37 ± 0.5 °C. The weight-drop device consisted of a fixed plastic guide tube, 20 cm in length and centered above the skull (− 1.0, − 2.0 mm to the bregma). A 25 g weight (diameter: 3 mm) was dropped by its gravity through the guide tube. Holding the weight with one hand immediately after the initial impact prevented a rebound impact.
After TBI, 200 μL (0.05 mg/mL) of Alexa Fluor 488 anti-mouse Ly-6G (Cat# 127626, Biolegend, USA) was injected intravenously to stain the neutrophils, followed by the TIS window establishment. A commercial fluorescence magnification microscope (Zeiss Axio Zoom. V16, numerical aperture (NA) = 0.25, Germany) was then used to continuously record the movement of immune cells on the bilateral cortex scale. The exposure time was set as 4 s per frame. After 4–6 h of monitoring, mice were sacrificed before they had the chance to wake up.
5.7 Two-photon microscopy for cerebrovascular and neural structural imaging
EGFP-expressed neuronal dendritic spines/axons/cell bodies in mice (Thy-1-EGFP) and FITC-dextran-labeled cerebral vasculature were imaged at 920 nm using a two-photon microscope (Ultima 2p; Bruker, USA) with a Ti:sapphire laser (80 MHz, Chameleon, Coherent, USA). Mice mounted with holders on their exposed skulls were placed under the microscope, and image stacks were captured utilizing a water-immersed objective (20 × , NA = 1.00, working distance = 2 mm, Olympus). The laser power after the objective lens ranged from 10 mW (imaging at the surface) to 100 mW (imaging above a 600 μm-depth). The image stack was acquired with a depth interval of 5 μm, and pixel dwell time was 2 μs. The description of imaging depth in the Results section was from the surface of the cortex. Bright-field images of the superficial cortical vasculature were used to relocate and reimage regions of interest. While the bright-field imaging quality before skull optical clearing was limited, the low-resolution vasculature map was used for relocation.
5.8 Two-photon microscopy for calcium imaging in awake mice
Two-photon calcium imaging was performed in the somatosensory cortex region in awake mice through the TIS window. First, a small hole was drilled into the skull above S1HL (− 0.82 mm AP, 1.5 mm ML from bregma)) using a cranial drill. Second, a glass micropipette and a PicoSprizer III (Parker) were utilized to inject virus-containing GCaMP6s (rAAV-hSyn-Gcamp6s-WPRE-hGH pA, 0.3 μL) into the cortex at an angle of 45° and depth of 0.3 mm. Next, the skull hole was filled up with bone wax, and the TIS window was established.
Two weeks later, upon the full expression of the virus, mice heads with the TIS window were immobilized under the two-photon microscope for imaging, with mice free to move their pawa about. Neural calcium imaging was then conducted to record electric stimuli activities 100 μm below pia using 2PM with 30 fps. The electro-stimulation of the front paw was performed on the ipsilateral side of the recorded sensorimotor cortex as described previously [58]. The isolated pulse stimulator (AM-system, Model 2100) was used for front-paw electrical stimulation (0.2 mA, 3 Hz, biphasic ±). The stimulus included 8 trials, each trial comprising a 3 s baseline period, 3 s stimulation events, and a 24 s convalescence period for a total of 30 s.
5.9 Three-photon microscopy imaging
Mice (Thy-1-YFP) with YFP-expressing neuronal dendritic spines/axons/cell bodies were imaged for examination of the suitability of the TIS window application for long-term 3PM monitoring. We used a three-photon microscope (Ultima 2p plus; Bruker, USA) equipped with a water-immersed objective (20 × , NA = 1.00, working distance = 2 mm, Olympus) and a 1300-nm fs laser (400 kHz, 50 fs) from a noncollinear optical parametric amplifier (Spirit-NOPA-VISIR, Spectra-Physics, USA) pumped by a regenerative amplifier (Spirit-16, Spectra-Physics, USA). Mice mounted with holders on their exposed skulls were placed under the microscope. The laser power after the objective lens ranged from 5 mW (imaging at the surface) to 50 mW (imaging above an 800 μm-depth). The image stack was acquired at a depth interval of 5 μm, and pixel dwell time was 2 μs; 4 frames were averaged. The description of imaging depth in the Results section was from the surface of the cortex.
5.10 Segmentation
Neutrophils were segmented using a simple semantic segmentation network based on dilated convolutions. The network contained four blocks of convolution, batch normalization, and ReLU layers. The filter size for each convolutional layer was 3 × 3. The calculated dice score of a softmax layer and a dice classification layer was considered the loss function. The optimizer was Adam, the epoch was 150, the initial learning rate was set to 10–3, and the minibatch size was 20.
For ground truth establishment, all datasets were labeled manually with MITK [59] based on semi-automated thresholding. The ratio of the training set to the test set was 9:1.
5.11 Cell tracking
Cells were tracked using the TrackMate plug-in in ImageJ [60] based on the segmentation results.
5.12 Safety of the TIS window
In vivo and ex vivo examinations were performed to explore the activation of microglia cells, the expression of GFAP, and the distribution of neutrophils in the cortex through the TIS window.
The TIS window was established on the left side of the skull of Cx3cr1-GFP mice in which GFP was expressed in microglia. The mice were then placed under the two-photon microscope for 1 h of microglia monitoring. Next, the same region of interest was re-observed in vivo 48 h after the TIS window establishment because microglia activity usually reaches maximum levels 48 h after a craniotomy. After imaging, the mice were perfused and fixed with PBS and 4% paraformaldehyde (PFA, Sigma-Aldrich, USA), and their brains were subsequently extracted, placed in 4% PFA overnight for post-fixation, sliced (100 μm), and imaged with a confocal microscope (Zeiss, Germany).
Similarly, to evaluate GFAP expression, the same steps outlined above were performed, but this time the brain slices were immunohistochemically stained with anti-GFAP antibodies 10 days after the TIS window establishment and imaged with the same confocal microscope since astrocyte activity reaches a maximum at a certain time point after a craniotomy.
Given that the longest observation in this investigation lasted 21 days, microglia and GFAP expression monitoring in brain slices were conducted 21 days after the TIS window establishment.
Then, brain slices were H&E-stained and imaged using an upright microscope system (Nikon, Japan) to determine whether neutrophils penetrated the parenchyma, a sign of an inflammatory response.
Finally, the laser damage threshold through TIS window was evaluated by the two-photon microscope. The evaluation was performed by two steps. Step 1: Firstly, a two-photon image of neural dendrites was taken with normal laser power (10 mW after the objective). Secondly, turned the focus down to 700 μm deep and the laser power was set as the maximum of the laser (250 mW after the objective), then continuously scanned for 10 cycles. Thirdly, returned to the surface and took a two-photon image of the same area to evaluate if it would damage the dendrites. Step 2: used different laser power (10, 50, 100, 150, 200 mW after the objective) to focus on the brain surface and performed 10 cycles of scanning for each power to evaluate how much laser power would cause injury of the dendrites.
5.13 Data quantification
All imaging data were analyzed using the Image J software developed by the National Institutes of Health (Bethesda, Maryland). 3D image reconstruction was performed on the Imaris (Bitplane).