2.1.

Animals

In this study, we used Thy1-GFP-M mice, CX3CR1-GFP mice, and C57BL/6J mice (6 to 12 weeks old). Mice were anesthetized with a mixture of 2% $\alpha$-chloralose and 10% urethane ($0.8\mathrm{mL}/100\mathrm{g}$) via intraperitoneal injection. Next, mice were transcardially perfused with 0.01 M phosphate-buffered saline (PBS) (P3813, Sigma-Aldrich) and then 4% paraformaldehyde (PFA) (158127, Sigma-Aldrich). Finally, the brains were dissected and postfixed overnight at 4°C in 4% PFA. Some brains were sliced into 1-mm-thick or 2-mm-thick coronal sections with a vibratome (Leica VT 1200s). The animal care and experimental protocols were in accordance with the Experimental Animal Management Ordinance of Hubei Province, China and have been approved by the Institutional Animal Ethics Committee of Huazhong University of Science and Technology.

2.2.

Chemicals and Reagents

In this study, we used the following chemicals: tetrahydrofuran (THF) (186562, Sigma-Aldrich), dibenzyl ether (DBE) (108014, Sigma-Aldrich), $N,N,N^{\prime },N^{\prime }$-tetrakis (2-hydroxypropyl) ethylenediamine (EDTP) (T0781, Tokyo Chemical Industry Co., Ltd.), DL-alpha-tocopherol (vitamin E) (A17039, Alfa Aesar), and triethylamine (80134318, Sinopharm Chemical Reagent Co. Ltd.).

For dehydration solutions, THF was mixed with ${\mathrm{dH}}_2\mathrm{O}$ according to the following concentration: 50% (v/v), 70% (v/v), 80% (v/v), and 100% (v/v). Triethylamine was added to adjust the pH to about 9.0. For clearing agents, the eFDISCO+ was prepared by adding 0.5% (w/v) EDTP to the DBE, and the vFDISCO+ was prepared by adding 0.5% (v/v) vitamin E to the DBE.

The column absorption chromatography with basic activated aluminum oxide (20001861, Sinopharm Chemical Reagent Co. Ltd.) was used to remove the peroxides produced in THF and DBE, as described in the literature.^19,20

2.3.

Measurement of Peroxide Concentration

The basic activated aluminum oxide was used to remove the peroxides in DBE. Then EDTP at concentrations of 0.1% (w/v), 0.5% (w/v), 1% (w/v), and 2% (w/v); vitamin E at concentrations of 0.1% (v/v), 0.5% (v/v), 1% (v/v), and 2% (v/v); and triethylamine at concentrations of 0.1% (v/v), 0.5% (v/v), 1% (v/v), and 2% (v/v) were added to 10 mL DBE in 50 mL centrifugal tubes (430829, Corning, USA) to prepare different refractive index (RI) matching solutions, respectively. Quantofix-25 strips (Macherey-Nagel, Germany) were used to measure the peroxide concentration of the reagent after 0, 9, 24, and 72 h. All tubes except the pure DBE group at LT were placed horizontally in a dark room at RT. The pure DBE group at LT was placed in a refrigerator at 4°C to 8°C.

2.4.

FDISCO+ Clearing Procedure

The 1-mm-thick brain sections were dehydrated with THF solutions at graded concentrations: 50% (v/v), 70% (v/v), 80% (v/v), and 100% (v/v) (twice), with 1 h each step at 4°C to 8°C. Then the dehydrated samples were transferred to the RI matching solutions for 30 min at RT.

For whole-brain clearing, the samples were dehydrated with 50% (v/v), 70% (v/v), 80% (v/v), and 100% (v/v) (thrice) THF solutions, with 12 h each step at 4°C to 8°C, followed by immersing in the RI matching agents for more than 3 h at RT.

During the clearing, the samples were covered with aluminum foil to ensure a dark environment, and all steps were performed with slight shaking. The cleared samples were stored at RT all the time.

2.5.

Fluorescence Labeling

For nuclei staining, the 1-mm-thick brain sections were incubated with 1% (w/v) propidium iodide (PI) (P1304MP, Life Technologies)/PBST (0.2% Triton X-100/PBS) for 24 h at RT with slight shaking. Then sections were washed several times with PBST for 6 h.

For vasculature labeling, $10\mu \mathrm{g}$ Alexa Fluor 647 conjugated anti-mouse CD31 antibody (CD31-A647) (102416, BioLegend) was used to label the vasculature in C57BL/6J mouse brain by caudal vein injection. The mouse was perfused 30 min after injection. The postfixed mouse brain was sliced into 1-mm-thick coronal sections.

2.6.

Measurement of Transparency

A digital camera (HDC-HS900GK) was used to acquire the bright-field images of samples. We used a visible-near-infrared optical fiber spectrometer (USB4000, Ocean Optics, USA) to produce a circular spot of light (diameter, 5 mm) to irradiate the samples immersed in the cleared medium. Then the transmitted light was measured on the other side of the samples. The light across only the clearing medium was measured as the blank value. Finally, the light transmittance was normalized by the blank value.

2.7.

Fluorescence Imaging

For brain slices, the cortical regions of the samples were imaged with a confocal fluorescence microscope (LSM 710, Zeiss, Germany) equipped with objectives of Fluar $5\times /0.25$ (dry; working distance, 12.5 mm) and Fluar $10\times /0.5$ (dry; working distance, 2.0 mm). Before and after clearing, the imaging parameters were kept the same when imaging the same regions.

For whole brains, samples were imaged with a customized Bessel light-sheet fluorescence microscope (LSFM) based on Olympus MVX10 and equipped with an sCMOS camera (Hamamatus Flash 4.0 V2), as described in the published literature.⁴⁵ In this study, 3 h to 4 h are needed for imaging per whole mouse brain.

2.8.

Data Analysis

The Fiji (Version 1.51n), MATLAB (Version 2014a, MathWorks), and Imaris (Version 7.6, Bitplane AG) were used for image processing and quantitative analysis of data. The analysis was derived from the literature.^44,46

For the quantification of GFP fluorescence in CX3CR1-GFP mouse brain, the maximum intensity projections (MIPs) of $25\mu \mathrm{m}$ thickness were acquired from each $z$ stack. We used Otsu algorithm for image thresholding segmentation and binarization.⁴⁷ Then we used MATLAB to extract the signal value and calculate the mean fluorescence intensity of the MIP images. The normalized fluorescence intensity was calculated by dividing the mean fluorescence intensity after $n$ ($n=3$, 6, 9) h by the value after 0 h. The processing is similar for quantification of PI signals, CD31-A647 signals, and GFP fluorescence for long-time preservation. And the normalized value of PI and CD31-A647 signals was described as the ratio of the mean fluorescence intensity before and after clearing, and the normalized GFP fluorescence intensity for long-time preservation was calculated by dividing the mean fluorescence intensity after $n$ ($n=3$, 7, 14) days by the mean fluorescence intensity after 0 day.

For the quantification of GFP fluorescence enhancement in Thy1-GFP-M mouse brain, we used Fiji to outline the same neurons in the cortical areas before and after clearing and measure the mean fluorescence intensity. The selective neurons must be located in the shallow position to avoid the influence of light scattering on the fluorescence intensity. The normalized mean fluorescence intensity was described by dividing the intensity of cleared group by the value of the uncleared group.

For the LSFM images, we used Fiji to transform the data to 8-bit images and then stitched the images using the Fiji plugins of Grid/Collection Stitching. Finally, we used Imaris to visualize the 3D-rendered images.

3.1.

Development of FDISCO+ by Inhibiting the Generation of Peroxides in DBE

The peroxide contaminations generated in DBE would quench the endogenous fluorescence severely even at a low concentration.¹⁹ FDISCO realizes the good preservation of fluorescence in DBE by controlling the temperatures at a low level.²⁹ In this work, we measured the peroxide concentration in DBE using Quantofix-25 strips. As shown in Figs. 1(a) and 1(b), the DBE at RT produced about $0.5\mathrm{mg}/\mathrm{L}$ peroxides in 9 h, whereas the DBE at LT did not produce peroxides in 72 h. This result indicates that LT can effectively reduce the generation of peroxides in DBE, facilitating fluorescence preservation. Hence, the samples cleared by FDISCO have to be kept at LT all the time to minimize fluorescence loss of fluorescence intensity for follow-up imaging, especially for the samples that needed repetitive imaging. However, it is not easy to achieve LT during imaging, as the imaging apparatus is usually under RT environment.

Fig. 1

Inhibiting the peroxide generation in DBE at RT preserves the GFP fluorescence. (a) The concentration of peroxides in DBE in different experimental conditions at 0, 9, 24, and 72 h, respectively. RT: the pure DBE was stored at RT; LT: the pure DBE was stored at LT; DBE (E): the DBE was added with EDTP and stored at RT; DBE (V): the DBE was added with vitamin E and stored at RT; and DBE (T): the DBE was added with triethylamine and stored at RT. (b) The quantification of peroxide concentration in (a). (c) 2-mm-thick C57BL/6J mouse brains were cleared in DBE with different concentrations of EDTP or vitamin E. (d) The cleared adult CX3CR1-GFP mouse brain slices were put in RI matching solutions at RT and were imaged at 0, 3, 6, and 9 h with the confocal microscope. (e) The decay curves of fluorescence intensity in (d) ($n=8$). All values are presented as the mean ± SD. Statistical significance in (e) (***$p<0.001$) was assessed by one-way ANOVA followed by the Bonferroni post hoc test.

To solve the conflict between fluorescence preservation and imaging temperature, we look for a modified protocol based on FDISCO to preserve endogenous fluorescence well at RT. First, we chose three reagents, including EDTP, vitamin E, and triethylamine and tested their antiperoxidation capability when added to DBE. As shown in Figs. 1(a) and 1(b), DBE added with EDTP at concentrations higher than 0.5% (w/v) did not produce peroxides for 72 h, indicating effective inhibition of the peroxide generation in DBE at RT. Additionally, for vitamin E, the concentration of 0.5% (v/v) in DBE also did not produce observable peroxides for 72 h. In contrast, DBE added with triethylamine at concentrations of 0.1% (v/v), 0.5% (v/v), 1% (v/v), and 2% (v/v) induced a faster generation of peroxides than the pure DBE group at RT, indicating that triethylamine could not prevent but accelerate the peroxide generation. These results indicate that the addition of proper concentration of EDTP and vitamin E has antiperoxidation capability on DBE. Further, we validated the influence of the addition of different concentrations of EDTP and vitamin E in DBE on tissue transparency. As shown in Fig. 1(c), the tissue transparency reduced slightly as the concentration of EDTP increased but was not affected by vitamin E with different concentrations. Considering both the oxidation resistance and the clearing transparency, 0.5% (w/v) of EDTP or 0.5% (v/v) of vitamin E was chosen as the final optimal concentration.

Then we used the cleared mouse brain slices (CX3CR1-GFP) to verify the fluorescence preserving capability of the DBE added with 0.5% (w/v) EDTP or 0.5% (v/v) vitamin E. The cleared samples were immersed in respective clearing agents at RT and imaged every 3 h. For the samples immersed in pure DBE, the fluorescence signal declined to 62% after 9 h. Although for the samples immersed in the DBE added with EDTP or vitamin E, the GFP signals were significant higher than in pure DBE and showed no noticeable decrease along with time due to inhibited peroxide generation, as shown in Figs. 1(d) and 1(e).

Based on the above results, we proposed an optimized clearing protocol for better fluorescence preservation under RT, termed FDISCO+. We devised two schemes: one using DBE containing 0.5% (w/v) EDTP as the clearing medium termed eFDISCO+, and the other using DBE containing 0.5% (v/v) vitamin E termed vFDISCO+.

3.2.

FDISCO+ Achieves Comparable Fluorescence Enhancement with FDISCO

To further investigate the fluorescence preserving capability of FDISCO+, we cleared the mouse brain samples labeled with different fluorescent probes. We imaged the same cortical areas of 1-mm-thick brain sections before and after clearing by FDISCO+ and FDISCO and quantified the intensity change of fluorescence signals. FDISCO+ showed excellent preservation for all tested fluorescence signals, including signals from endogenous GFP, PI, and CD31-A647, which revealed no obvious difference compared to the original FDISCO. The GFP fluorescence intensity of samples cleared by FDISCO+ increases to 175% to 185% compared to the fluorescence intensity before clearing, as shown in Figs. 2(a) and 2(b). The fluorescence intensity of PI signals after clearing increases to at least 210%, as shown in Figs. 2(c) and 2(d). In addition, FDISCO+ also significantly strengthens the CD31-A647 signals, as shown in Figs. 2(e) and 2(f).

Fig. 2

FDISCO+ achieves comparable fluorescence enhancement with FDISCO. (a) Confocal imaging of cortical neurons in adult Thy1-GFP-M mouse brains before and after clearing. (b) The normalized mean fluorescence in (a) ($n=16$). (c) Confocal imaging of PI-stained mouse brain sections before and after clearing. (d) The normalized mean fluorescence in (c) ($n=9$). (e) Confocal imaging of blood vessels stained with CD31-A647 antibody before and after clearing. (f) The normalized mean fluorescence in (e) ($n=9$). All values are presented as the mean ± SD. Statistical significance in (b), (d), and (f) (n.s., not significant) was assessed by one-way ANOVA followed by the Bonferroni post hoc test.

3.3.

FDISCO+ Allows Long-Time Preservation of GFP Fluorescence at RT

Further, we tested the ability of long-time fluorescence preservation of FDISCO+. We imaged the cortical neurons in Thy1-GFP-M mouse brains cleared by FDISCO+ and FDISCO at 0, 3, 7, and 14 days then quantified the fluorescence intensity changes normalized to 0 day.

Figure 3 shows the representative images and the quantitative data of intensity changes. The GFP fluorescence of brain blocks treated by FDISCO+ and FDISCO are preserved very well for at least two weeks compared to the pure DBE (RT) group. eFDISCO+ shows similar fluorescence preservation with FDISCO and better fluorescence preservation than vFDISCO+ after 7 days with a significant difference, indicating that eFDISCO+ is better in long-time fluorescence preservation than vFDISCO+.

Fig. 3

FDISCO+ allows long-time preservation of GFP fluorescence at RT. (a) Confocal imaging of cortical neurons in adult Thy1-GFP-M mouse brains cleared after 0, 3, 7, and 14 days using FDISCO+ and FDISCO methods. (b) The histogram of normalized mean fluorescence in (a) ($n=6$). All values are presented as the mean ± SD. Statistical significance in (b) (*$p<0.05$; **$p<0.01$; ***$p<0.001$) was assessed by one-way ANOVA followed by the Bonferroni post hoc test.

3.4.

FDISCO+ Shows Good Clearing Capability and Enables Visualization of the Neural Structures

Finally, we tested the clearing capability of FDISCO+ and displayed 3D visualization of the nervous system with LSFM imaging. Figures 4(a) and 4(b) show the bright-field images of whole brains before and after FDISCO+ and FDISCO clearing and the transmittance curves with wavelength. vFDISCO+ achieves similar transparency as FDISCO, and eFDISCO+ displays modest clearing ability for the mouse brain. Meanwhile, we imaged the Thy1-GFP-M mouse brain cleared by vFDISCO+ and performed 3D reconstruction from different views, as shown in Figs. 4(c) and 4(d) and Video 1. The fine neuronal structures in different brain regions, such as the cortex, caudate putamen, hippocampus, and midbrain, can be detected effectively, as shown in Figs. 4(e)–4(h). We also obtained the 3D reconstruction of neurons in half brain cleared by eFDISCO+, as shown in Video 2.

Fig. 4

FDISCO+ shows good clearing capability and enables visualization of neural structures. (a) Bright-field images of whole brains before and after clearing. (b) The transmittance curves of cleared mouse brains (C57BL/6J mice, 8 weeks old) ($n=3$). (c), (d) Images of the whole brain (Thy1-GFP-M, 6 weeks old) cleared by vFDISCO+. (e)–(h) The high-magnification images reveal fine structures of different brain areas in (c), including (e) the cortex, (f) caudate putamen, (g) hippocampus, and (h) midbrain. All values are presented as the mean ± SD (Video 1, MP4, 9821 kB [URL: https://doi.org/10.1117/1.NPh.8.3.035007.1]) and (Video 2, MP4, 7244 kB [URL: https://doi.org/10.1117/1.NPh.8.3.035007.2]).