A versatile vessel casting method for fine mapping of vascular networks using a hydrogel-based lipophilic dye solution

Summary

Efficient labeling of the vasculature is important for understanding the organization of vascular networks. Here, we propose VALID, a vessel-labeling method that enables visualization of vascular networks with tissue clearing and light-sheet microscopy. VALID transforms traditional lipophilic dye solution into hydrogel by introducing gelatin and restrains the dye aggregation, resulting in complete and uniform vessel-labeling patterns with high signal-to-background ratios. VALID also enhances the compatibility of lipophilic dyes with solvent-based tissue-clearing protocols, which was hard to achieve previously. Using VALID, we combined lipophilic dyes with solvent-based tissue-clearing techniques to perform 3D reconstructions of vasculature within mouse brain and spinal cord. We also employed VALID for 3D visualization and quantification of microvascular damage in a middle cerebral artery occlusion mouse model. VALID should provide a simple, cost-effective vessel-labeling protocol that would significantly widen the applications of lipophilic dyes in research on cerebrovascular complications.

Graphical abstract

Highlights

• •VALID is a method for casting vasculature using lipophilic dye hydrogel
• •VALID enables effective labeling for both large vessels and capillaries
• •VALID is compatible with 3D immunolabeling and tissue clearing
• •VALID enables 3D mapping and analysis of vasculature for health and disease

Motivation

3D reconstruction of the entire vasculature is fundamental to understanding vascular lesions during the progression of various neurovascular diseases. The combination of vessel-labeling, tissue-clearing, and optical imaging techniques allow 3D mapping of the vasculature of different tissues/organs. Current vessel-specific marker-based labeling methods can provide biological ground truth for blood vessels but are restricted by either insufficient signal-to-background ratios, incomplete labeling for large vessels, or high cost. Perfusion-based methods can serve as useful alternatives with relative high signal-to-background ratios; however, dye leakage has greatly limited their applications. Additionally, the tissue-clearing compatibility of these methods still needs to be improved. Here, we developed VALID, an efficient, easy-handling, and cost-effective protocol that enables complete and uniform labeling of different vasculatures with bright fluorescence.

Zhu et al. develop an efficient, easy-handling, and cost-effective vessel-labeling protocol with robust labeling capability and immunolabeling/tissue-clearing compatibility, termed VALID. VALID enables complete and uniform labeling of different vasculatures with bright fluorescence, as well as 3D imaging and analysis of vasculature under physiological and pathological conditions.

Introduction

The adult human body contains widespread blood vessels that form an exceptionally extensive and complex network essential for maintaining normal physiology in humans. Abnormal changes in vascular structures are closely associated with the occurrence and development of several diseases, such as stroke, diabetes, and Alzheimer’s disease.^1,2,3,4 Therefore, it is important to visualize entire vascular networks down to the capillary level for a better understanding of the underlying mechanisms of various vessel-associated diseases and for designing targeted therapy.

In recent years, a growing number of studies have examined the structural bases and remodeling of overall vascular networks under normal and pathological conditions using different imaging techniques.⁵ Traditional in vivo contrast-based angiography can resolve in vivo imaging of large arteries and veins in living larger animals and humans, which is indispensable in clinical diagnosis, but is limited by low image resolution and artifacts.^6,7,8 Vessel casting combined with traditional medical imaging tomography, like computed tomography (CT), can serve as powerful alternatives with higher resolution for imaging vasculature ex vivo.^9,10 However, potential limitations such as incomplete casting of vessel systems and insufficient resolution to distinguish capillaries or tissue distortions, have limited their applications.^11,12 Recently, great progress has been achieved in non-invasive imaging of the entire intact murine cranial vasculature ex vivo by employing low-density polymer vascular casting, iterative sample processing, and micro-CT imaging, providing a robust pipeline for 3D imaging and analysis of murine cranial vasculature.¹³

In addition to traditional medical imaging, the emergence of modern optical imaging techniques such as confocal, multi-photon, and light-sheet microscopies has provided powerful tools for high-resolution imaging of vascular networks down to the capillary level ex vivo.^14,15 However, the imaging depth remains limited because of the massive scattering and absorption of opaque samples.¹⁶ This problem has since been significantly resolved by the development of optical clearing techniques that render nearly all types of tissues transparent.^17,18 Combining vessel-labeling, tissue-clearing, and light-sheet imaging techniques creates a reliable procedure for 3D observation of vascular networks in various tissue types as well as the assessment of vascular remodeling after the occurrence of specific vessel-associated diseases.^3,19,20

Efficient and complete labeling of entire vascular networks with bright fluorescence is the first and most important step in the 3D imaging of vascular structures with tissue clearing and light-sheet microscopy. So far, various fluorescent vessel-labeling methods have been developed for labeling vascular structures, including transgenic mouse models that express fluorescent proteins in vascular endothelial cells²¹ and vessel specific markers (e.g., lectins and CD31 antibodies),^22,23 and filling of fluorescent dye solutions (e.g., Evans blue)²⁴ or fluorescent hydrogels (e.g., fluorescence-conjugated dextran).^25,26,27 These techniques have already been successfully used alongside different tissue-clearing protocols to label and visualize vasculature.

Despite the success of 3D vascular imaging with current vessel-labeling techniques, many challenging issues remain unresolved. Vascular labeling via transgenic mice is usually faint and discontinuous, and expression of fluorescent proteins within other structures is non-specific.^21,28 Additionally, the costs for using transgenic animals are rather high. Vessel specific markers, especially for antibodies, could provide reliable ground truth for vascular structures. However, limited antibody penetration depth is a common and troublesome problem when applied to large tissues. It always requires several weeks to months for passive diffusion for highly permeabilized tissue using qualified antibodies at high concentrations to achieve sufficient antibody penetration, which are somewhat expensive and time consuming.^29,30 As a result, some of these vessel specific markers, such as CD31 antibodies and tdTomato lectins, were directly injected into the mouse body via tail vein for vessel labeling in a quick manner.²³ However, this operation usually results in a low signal-to-background ratio and is less effective for labeling large vessels.³ Recently, a pioneer study developed Vessap, a vessel-labeling strategy based on the co-labeling of wheat germ agglutinin and Evans blue, could realize complete labeling of the brain vasculature.¹⁹ However, the combination of two fluorescent signals increased the difficulties in both labeling and image analysis procedures. In recent years, vessel labeling via perfusion of fluorescent dyes has become popular, and lipophilic dye, such as 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI), has proved to be an excellent candidate for labeling vessels due to its easy-to-use and cost-effective nature, as well as the bright fluorescent signals when it is inserted into the cell membrane.^31,32,33,34

However, vessel labeling through the perfusion of lipophilic dyes also has disadvantages that limit its applications. First, an aqueous solution is currently used to dilute lipophilic dyes for perfusion; however, most lipophilic dyes are highly hydrophobic, leading to quick aggregation of the dye molecules during perfusion. As a result, the precipitates easily occlude microvessels, causing dye leakage and heterogeneous labeling of vascular networks. Second, almost all available tissue-clearing methods with strong clearing capability are incompatible with lipophilic dyes because these techniques use detergents or solvents that can easily wash out lipophilic dyes. This approach can therefore only be used alongside several detergent-free and solvent-free clearing approaches, such as See Deep Brain (SeeDB),^35,36 ScaleS,³⁷ clearing-enhanced 3D (Ce3D),³⁸ and the M-xylylenediamine (MXDA)-based aqueous clearing system (MACS).³⁴ These problems limit the applications of this kind of vessel-labeling method in studying vessel-associated diseases using tissue clearing and should be addressed urgently.

In this study, we successfully tackle these two major challenges by developing a versatile vessel-casting method based on hydrogel-based lipophilic dye solution, termed vessel labeling via gelatin-based lipophilic dye solution (VALID). VALID transformed traditional lipophilic dye solution into hydrogel by introducing gelatin, successfully preventing the aggregation of the dye molecules for an extended period and dye leakage during perfusion. VALID achieves complete and uniform labeling of fine vascular structures in rodent brains and various organs at high signal-to-background ratios, as well as co-labeling with multiple kinds of antibodies through immunolabeling. More importantly, VALID greatly enhances the compatibility of lipophilic dyes with typical solvent-based clearing methods, enabling 3D reconstruction of whole-brain vasculature by combining lipophilic dyes with solvent-based tissue clearing for the first time to execute. This protocol allows 3D visualization and analysis of acute vascular lesions in a middle cerebral artery occlusion (MCAO) mouse model. We believe that VALID will serve as a versatile and reliable tool for studying various vascular dysfunctions and diseases from fundamental research to clinical investigation, such as microvascular injury and remodeling post-stroke.

Results

Development of the VALID protocol using a gelatin-based hydrogel solution to prevent the aggregation of lipophilic dye molecules

To achieve complete and reliable labeling of vasculatures, the aggregation of lipophilic dye molecules at physiological salt concentrations must be prevented at first. As one of the most common and easily accessible hydrogels, gelatin was proved to have great potential in preventing the aggregation of crystals or ions to stabilize heterogeneous suspensions.³⁹ Therefore, we introduced gelatin into lipophilic dye solution. Figure 1A shows the workflow for preparing lipophilic dye solutions with/without gelatin and comparing dye aggregation among these solutions. We observed the different working solutions 5 min, 30 min, and 24 h after preparation. As shown in Figure 1B, DiI aggregation in the traditional working solution became noticeable 30 min after preparation. In contrast, the working solution containing 2% and 4% gelatin showed excellent stability with no apparent DiI molecule aggregation even after 24 h storage. The vascular labeling protocol using this working solution of lipophilic dyes was termed VALID.

Figure 1.

The VALID protocol is effective in preventing dye aggregation

(A) Workflow for comparing dye aggregation of no-VALID and VALID labeling protocol.

(B) Bright-field images of the dye working solutions for no-VALID and VALID protocols at different times.

(C) Fluorescent images of the dye working solutions for no-VALID and VALID protocols at indicated time.

(D) Quantification of relative dye aggregation levels for no-VALID and VALID protocols (n = 6 samples for each protocol).

All values are presented as the mean ± SD

To prove that DiI molecules remain stable in the VALID solution, we performed continuous fluorescence imaging of the above three working solutions and quantitatively evaluated aggregation over time. As shown in Figures 1C and 1D, DiI aggregation in the traditional working solution actually began 5 min after preparation, with the aggregation level accelerating from min 5 to 10 and decelerating from min 10 to 30. On the contrary, the VALID solution bore nearly no aggregated dye molecules, and the aggregation level remained stable. After 24 h, the DiI molecules in the traditional working solution had aggregated into large crystals; however, the VALID solution still revealed no observable dye accumulation (Figure S1A). Additionally, we also revealed that the VALID solution would transform into stable, transparent hydrogel after cooling, indicating that the dye molecules dispersed uniformly within the hydrogel without obvious aggregation (Figure S1B).

VALID ensures complete and uniform labeling of vascular networks in various organs and is compatible with immunolabeling

We then labeled and imaged vascular networks labeled by solutions with/without VALID. We first labeled mouse brain vasculature and observed its vascular structures with a confocal microscope. As shown in Figure 2A, the vasculature labeled by no-VALID protocols was not uniform and contained bright spots in many typical brain regions, including the cortex, hippocampus, and cerebellum. On the contrary, VALID with both 2% and 4% gelatin showed complete and uniform labeling of brain vasculature and had no bright spots in the observed areas. Other lipophilic dyes, including 1,1′-dioctadecyl-3,3,3′3′-tetramethylindodicarbocyanine perchlorate (DiD) and 3,3′-dioctadecyloxacarbocyanine perchlorate (DiO), were also assessed using the VALID protocol. Similar to DiI, vasculatures labeled by DiD and DiO using no-VALID solution endured dye leakage and partial loss of vessel information in specific regions, while VALID ensured brain vasculature with satisfactory uniformity and completeness (Figure S2A).

Figure 2.

VALID improves dye leakage and loss of vessel information in typical organs

(A) Fluorescence images of overall vessel information in different regions in sagittal mouse brain slices labeled by no-VALID and VALID protocols using DiI.

(B) Fluorescence images of overall vessel information in mouse liver slices labeled by no-VALID and VALID protocols using DiI.

(C) Signal profiles of different lines marked in (B).

Besides the brain, typical mouse tissues and organs were also labeled and imaged. As shown in Figure 2B, the lumen of large vessels labeled without VALID was not effectively labeled with dye signals spread outside, indicating that the dye solution leaked out from the vessel and into the surrounding area. In contrast, VALID with 2% and 4% gelatin efficiently labeled liver vasculature in both small and large vessels with satisfactory uniformity. Additionally, the signal intensity of the large vessel lumen labeled using VALID (4% gel) was higher than that labeled using VALID (2% gel), implying that a higher concentration of gelatin has a better filling effect for large blood vessels in the liver (Figure 2C). Notably, VALID was capable for labeling the vascular structures at the capillary level within bones, such as mouse skull (Figure S2B). The VALID protocol was also effective for other organs, such as the mouse heart, spleen, intestine, kidney, etc. (Figure S2C). Additionally, VALID could also work in adult rat organs, like the brain (Figure S2D).

Next, we compared the labeling performance of our VALID protocol with several commonly used labeling methods, including CD31 antibodies (CD31), tdTomato lectins (lectin), wheat germ agglutinin (WGA), and Evans blue. All the assessed methods revealed similar and uniform labeling patterns on brain vasculature, except for Evans blue (Figure 3A). Evans blue was effective for vessels with relatively large size but lost a large part of the capillary information. We calculated the signal-to-background ratios of the fluorescence images for different approaches to evaluate the vessel-labeling efficiency of each method. As shown in Figure 3B, the signal-to-background ratio of VALID is significantly higher than other tested methods. In addition to mouse brain tissue, we also compared the labeling performance among these methods in different internal organs. As shown in Figures 3C and 3D, VALID could effectively highlight the vascular networks containing both large vessels and capillaries within mouse kidneys and livers. In contrast, labeling methods based on vessel specific markers, including CD31, lectin, and WGA, were less effective for labeling large vessels.

Figure 3.

Comparison of vessel-labeling performance of VALID with other vessel-labeling methods

(A) Fluorescent vascular images of mouse brain slices labeled using VALID (2% gel), VALID (4% gel), Evans blue, CD31 antibodies (CD31), tdTomato lectins (lectin), and WGA.

(B) Calculation of the signal-to-background ratios based on the maximum intensity projection (MIP) of vascular image stacks for each method (n = 6 samples for each labeling method).

(C) Fluorescent vascular images of mouse kidney slices labeled by each method.

(D) Fluorescent vascular images of mouse liver slices labeled by each method.

(E) Validation of labeling completeness of VALID with CD31 antibodies and lectins by passive immunolabeling (PI).

(F) Calculation of overlap ratios of vasculatures labeled using VALID (4% gel)/CD31 and VALID (4% gel)/lectin samples (n = 6 samples for each group).

All values are presented as the mean ± SD. Statistical significance in (B) (∗∗∗p < 0.001) was assessed using one-way ANOVA, followed by the Bonferroni post-hoc test.

Furthermore, we also validated the completeness of VALID-casted vascular structures. We co-labeled the blood vessels with commonly used lectin and CD31 antibodies both via passive immunolabeling post-VALID and intravenous injection pre-VALID. Vasculature labeled by VALID finely overlie vasculature stained with lectin and CD31 via two ways (Figures 3E and S3A). The overlap ratio between tested vessel specific marker-labeled vessels and VALID-casted vessels was close to 1 (Figures 3F and S3B). Additionally, we also tested the labeling accuracy of VALID on organs containing a large cavity, like the mouse heart. The dual-color images co-labeled by CD31 antibodies revealed that complete vascular structures within heart parenchyma were labeled without any additional information (Figure S3C). However, the dye hydrogel accumulation in the ventricle space was obvious in VALID-labeled (4% gel) heart tissues (Figure S2C), and this problem could be effectively resolved with a lower concentration of gelatin (Figure S3D). In a word, these results revealed that VALID did effectively label the true and complete vascular networks, and the hydrogel used did not obviously prevent other molecules to penetrate.

Then, we examined the compatibility of the VALID protocol with immunolabeling, which is considered a powerful tool for molecular labeling of various tissues. We employed multiple kinds of antibodies other than CD31, including parvalbumin (PV), glial fibrillary acidic protein (GFAP), tyrosine hydroxylase (TH), NeuN, ionized calcium binding adapter molecule 1 (Iba1), and microtubule-associated protein 2 (MAP2), and tested their compatibility with VALID. As shown in Figure 4A, all tested antibodies worked well on VALID-casted samples, revealing that VALID is also capable for labeling other cell types assisted by immunolabeling technique. Additionally, we also performed 3D immunolabeling on VALID-casted samples using NeuN and Iba1 antibodies (Figures 4B, 4C, S4A, and S4B). As shown in Figures 4D and S4C, the entire imaging depth was up to 600 μm, with both NeuN and blood vessels abundantly observable at different depths.

Figure 4.

Compatibility of VALID with 2D and 3D immunolabeling

(A) Fluorescence images of different antibodies co-labeled with VALID, including parvalbumin (PV), glial fibrillary acidic protein (GFAP), tyrosine hydroxylase (TH), NeuN, ionized calcium binding adapter molecule 1 (Iba1), and microtubule-associated protein 2 (MAP2).

(B and C) 3D reconstruction of the labeled NeuN-positive cells and VALID-casted vascular structures, respectively. Merged images shown in (B) and (C).

(D) Optical sections of reconstructed images shown in (C) at different imaging depths.

VALID enhances the compatibility of lipophilic dyes with commonly used solvent-based tissue clearing

We then investigated the compatibility of VALID-labeled fluorescence signals from lipophilic dyes in mouse brain with various combinations of solvent-based tissue-clearing protocols, including methanol (MeOH), ethanol (EtOH), tert-butanol (TB), and tetrahydrofuran (THF) for dehydration and dibenzyl ether (DBE) for refractive index matching (Figure 5A).

Figure 5.

VALID enhances the compatibilities of signals labeled using lipophilic dyes with various solvent-based tissue-clearing protocols and does not influence tissue transparency

(A) Experimental design.

(B) Fluorescence images of vascular structures labeled using no-VALID/VALID protocol before and after clearing with different solvent-based tissue-clearing protocols.

(C) The preservation ratio of vascular signals labeled using no-VALID/VALID after clearing with different solvent-based protocols (n = 6 samples for each combination of labeling/clearing method).

(D) The relative signal-to-background ratios of vascular structures after clearing with different solvent-based protocols (n = 6 samples for each combination of labeling/clearing method).

(E) Bright-field images of mouse brain samples with or without gelatin perfusion before and after clearing with different solvent-based protocols.

(F) The relative transparency of cleared samples shown in (E) (n = 3 samples for each group).

All values are presented as the mean ± SD. Statistical significance in (C) and (D) (∗∗∗p < 0.001) was assessed using one-way ANOVA, followed by the Bonferroni post-hoc test. Statistical significance in (F) (n.s., not significant, p > 0.05) was assessed using an independent-sample t test

As expected, due to elution by various organic solvents, fluorescence signals labeled by traditional protocols without VALID were largely lost after clearing with the four common solvent-clearing protocols. However, DiI signals labeled with VALID successfully restrained elution against organic solvents and were finely retained without apparent loss after clearing (Figure 5B). Quantification analyses also revealed that both the preservation ratio and relative signal-to-background ratio of VALID-labeled cerebrovascular structures using DiI were significantly superior to those of traditional protocols without VALID after clearing (Figures 5C and 5D). An identical degree of enhancement of compatibility by VALID was also observed when using DiD for vascular labeling (Figures S5A–S5C). We also compared the tissue-clearing compatibility of VALID with other vessel-labeling approaches under the same tissue-clearing protocols. As shown in Figures S5D and S5E, VALID showed more robust tissue-clearing compatibility over other tested methods.

To investigate the underlying mechanisms of how the vessels were casted by VALID, we imaged the VALID-casted mouse brain samples counterstained by DAPI under a confocal microscope with high magnification. As shown in Figure S6A, the lumens of large vessels are effectively labeled by filling with fluorescent hydrogels with DiI. However, we also noticed that the fluorescence signals in vascular walls, which could be recognized as the locations of endothelial cells by DAPI channels, were much higher than those in the lumen (Figures S6B and S6C). This indicated that a part of the DiI molecules might also be inserted into the plasma membranes of the endothelial cells because the signal of DiI molecules would be significantly enhanced after insertion into membranes. We speculated that simultaneous existence of the two bonding forms might render dye molecules more resistant to elution by organic solvents.

Additionally, the influence of this gelatin hydrogel on tissue transparency after clearing was also investigated. The clearing performances of the four selected clearing protocols for whole-brain samples perfused with or without 4% gelatin were observed and quantified (Figures 5E and 5F). The results showed that the addition of gelatin did not negatively affect tissue transparency for the entire the mouse brain.

VALID allows 3D visualization of vasculatures in different organs combined with different clearing protocols

After demonstrating VALID’s superior vessel-labeling ability and robust compatibility with tissue clearing, we achieved the 3D reconstruction of vascular systems using VALID-labeling and different tissue-clearing techniques. First, whole-brain vasculature labeled with VALID (4% gel) was cleared and imaged with the TB/DBE clearing protocol and light-sheet microscopy. As shown in Figure 6A and Video S1, entire cerebrovascular structures were finely visualized in 3D. The sagittal view displays a vascular distribution along the z axis, indicating that all vascular structures were excellently acquired throughout the entire imaging depth (Figure 6B). The detailed vascular networks in the cortex, hippocampus, thalamus, and cerebellum were also clearly identified (Figure 6C; Video S1). The blood vessels in the spinal cord were likewise finely perceived (Figure 6D; Video S2); both the central blood vessels and the surrounding capillaries could be identified (Figure 6E; Video S2).

Figure 6.

3D visualization of vasculatures using VALID and different tissue-clearing protocols

(A) 3D rendering of a VALID-casted vascular network (DiI) throughout an entire adult brain cleared with TB/DBE.

(B) Sagittal view of the reconstructed brain shown in (A).

(C) Detailed vasculature in the cortex, hippocampus, thalamus, and cerebellum.

(D) 3D visualization of VALID-casted vasculatures in the spinal cord cleared with THF/DBE.

(E) Magnification of the boxed regions in (D) reveals the large vessels and surrounded capillaries finely labeled and clearly visualized.

(F) Reconstructed vascular network of a VALID-casted adult mouse lung cleared with MACS.

(G) Magnification of the boxed regions in (F).

Additionally, we found that the VALID protocol possesses obvious superiority in labeling internal organs compared with traditional vessel-labeling methods, especially for mouse lung. As shown in Figure S7, no effective identification of blood vessels was found in the lung labeled using traditional intravenous injections of CD31 antibodies or lectins; however, the vascular structures within lung were effectively labeled by VALID. By combining VALID labeling and the aqueous-based MACS-clearing protocol, we reconstructed the fine vessel networks within mouse lungs. As shown in Figures 6F and 6G and Video S3, both the large vessels and capillaries were clearly visible.

VALID enables 3D analysis of detailed vascular networks in an ischemic brain

The 3D visualization and quantification of microvascular circuit remodeling in an ischemic brain are essential for evaluating recovery processes after stroke and determining potential treatment modalities. Here, we used the VALID protocol combined with solvent-based tissue clearing to scrutinize 3D pathological changes in brain vasculature after transient MCAO (Figure 7A).

Figure 7.

Visualization and quantification of pathological changes of cerebrovascular structures in ischemic mouse brains

(A) Experimental workflow for the labeling, clearing, imaging, and quantitative analysis of brain vasculature.

(B) Reconstruction of the vascular architecture in the brain reveals an obvious ischemic area with vessel information loss compared with the contralateral area.

(C) Magnification of non-ischemic and ischemic regions marked in (B).

(D) Representative images of the reconstruction and analysis of blood vessels by Filaments function in Imaris.

(E) Quantification of vessel length densities in non-ischemic and ischemic areas (n = 4 samples for each group).

(F) Statistical analysis of the distribution of vessel radius in non-ischemic and ischemic regions (n = 4 samples for each group).

All values are presented as the mean ± SD. Statistical significance in (E) and (F) (∗∗∗p < 0.001; ∗∗p < 0.01; n.s., not significant, p > 0.05) was assessed using an independent-sample t test.

Vascular structures in the ischemic brain were visualized 7 days post-stroke in 3D. As expected, the lateral striatum was recognized as the central ischemic region with vessel loss (Figures 7B and 7C). The reconstruction of images acquired at high magnification revealed a more substantial loss of vascular structures in the ischemic region compared with the non-ischemic contralateral region. Individual optical section images showed significant enhancement of tissue autofluorescence and loss of vessel information, particularly in microvessels (Figure 7C).

To further examine the vascular injury caused by ischemia, we thoroughly quantified vascular networks in ischemic and non-ischemic areas using the Imaris software (Figure 7D). Vessel length densities decreased drastically in ischemic regions, indicating that a considerable proportion of vessels were blocked and damaged (Figure 7E). Additionally, we calculated the distribution of vessel radii across ischemic and non-ischemic areas and found that the major form of vessel damage occurred in microvessels with radii less than 6 μm (Figure 7F). This finding was consistent with observations made in several previous reports.^40,41 These results suggest that VALID holds great potential for the overall investigation and quantification of vascular lesions in vessel-associated diseases.

Discussion

In recent years, interests in understanding 3D pathological changes in vascular structures under some pathological states has grown. The rapid development of optical imaging and tissue-clearing techniques has provided reliable tools for 3D imaging of vasculature down to the capillary level. The first step, which is the efficient and complete labeling of the entire vasculature, is fundamental to studying complex vascular networks. In this study, we developed VALID, a vessel-casting protocol for high-performance labeling and imaging of intact vascular networks. Gelatin was added to the aqueous-based lipophilic dye working solution and successfully restrained the aggregation of lipophilic dye molecules. This ensured much more complete and uniform labeling of vasculature than the traditional vessel specific marker-based labeling protocol. VALID could also be used alongside 2D and 3D immunolabeling for studying other cell types. Additionally, VALID significantly enhanced the compatibility of the lipophilic dye-labeled cerebrovascular information with common solvent-based tissue clearing. Using the VALID protocol, we first combined lipophilic dye labeling with solvent-based tissue clearing for the high-performance reconstruction of DiI-labeled vascular structures in whole mice brains and spinal cords. We also applied VALID to the 3D visualization and quantification of structural alterations in brain vasculature in a rodent MCAO model.

Fluorescent labeling of the vascular networks is closely related to the imaging quality of optical imaging tomography, such as confocal microscopy and light-sheet microscopy. Vessel specific markers, especially for antibodies, could provide reliable ground truth for vascular structures through passive immunolabeling procedures. However, limited penetration depth is a common and troublesome problem when applied to large tissues.⁴² Tail vein injection of these markers could accelerate the labeling procedure, but problems like incomplete labeling for large vessels and low labeling efficiency for injection via tail vein injection remain unresolved.³ In recent years, vessel labeling via perfusion of fluorescent dyes has become popular. Lipophilic dyes, such as DiI, have proved to be a potentially excellent candidate for labeling vessels due to their easy-to-use and cost-effective nature, as well as the bright fluorescent signals that are displayed when they are inserted into the cell membrane.⁴² However, lipophilic dyes also have major drawbacks remaining unresolved until now, namely severe dye aggregation and weak compatibility with tissue clearing. The developed VALID protocol successfully addresses the two major drawbacks, making lipophilic dyes ideal candidates for fluorescent labeling of vascular structures. VALID proves to have certain advantages over vessel specific marker-based vessel-labeling methods. First, VALID achieves a much higher signal-to-background ratio and shows enhanced tissue-clearing compatibility (Figures 3A, 3B, 5B–5D, S5D, and S5E). Second, VALID could label the complete vascular networks for both large vessels and capillaries, especially for superior ability in labeling large vessels (Figures 3C and 3D). Additionally, VALID not only effectively labels the vasculature but also enables simultaneous studying of other cell types with vascular networks assisted by immunolabeling techniques (Figure 4). This combination is expected to offer a powerful pipeline for investigation of the underlying mechanisms of certain vessel-associated diseases in biomedical studies.

Adding a certain proportion of gelatin into the lipophilic dye solution is the core of the VALID protocol. The introduction of gelatin not only effectively prevents dye aggregation but also greatly enhances the compatibility of the lipophilic dyes with common tissue-clearing methods. We have demonstrated that the solution will transform into stable, transparent hydrogel after cooling, and the dye molecules are uniformly spread within gelatin colloid (Figure S1B). Indeed, gelatin has been proved to have great potential in preventing the aggregation of crystals or ions to stabilize heterogeneous suspensions.^39,43 The VALID-labeled vasculatures display complete and uniform patterns without the dye leakage or missing vascular information usually encountered using traditional perfusion-based labeling protocols.

Notably, we discovered that VALID-labeled vascular information showed enhanced resistance to the elution caused by common tissue-clearing techniques. This is an important breakthrough because common lipophilic dyes can be easily washed out by organic solvents, resulting in huge restrictions for combining lipophilic dyes with tissue clearing to reconstruct the vasculature in 3D. This finding may greatly widen the applications of lipophilic dyes for vascular labeling in tissue-clearing research. To understand the underlying mechanism, we image the VALID-casted blood vessels under high magnification and find that the lumens of large vessels are effectively filled with fluorescent lipophilic dye hydrogels. Additionally, we also notice that part of the DiI molecules also inserted into the plasma membranes of the endothelial cells. Both the dye filling and molecular binding are expected to contribute to the enhanced compatibility with tissue clearing.

Several efforts have been made in the past to improve the vessel-labeling performances of lipophilic dyes. For example, Konno et al. employed neutral liposomes with lipophilic dyes to alleviate dye aggregation and achieve superior labeling performances.^44,45 However, acquiring the used neutral liposomes is a tall order because of the sparse marketing for them, the high cost, and the sensitivity and arguable incompatibility of neutral liposomes with organic solvents or detergents. CM-DiI has also been used recently as an alternative to DiI to achieve better compatibility.⁴⁶ CM-DiI is a type of modified DiI with an aldehyde-fixable group that adheres to both cellular membranes and proteins after fixation, enabling it to remain in the tissue after lipid extraction. However, experimentally, CM-DiI is only partially compatible with tissue-clearing methods using detergents or organic solvents,^34,47 such as CUBIC-L/R,⁴⁸ PACT,⁴⁹ and uDISCO.⁵⁰ Additionally, CM-DiI is nearly 100 times more expensive than the common DiI dye, which would make vasculature labeling using enormous amounts of CM-DiI dyes a luxury. Recently, other vessel-labeling methods based on fluorescent hydrogels were also developed for efficient and robust vascular labeling.^25,26,27 However, the remaining problems such as loss of antigenicity, leakage of fluorescent-conjugated dyes, and complexity in preparing acrylamide-based hydrogel prevent these methods from wide applicability. In contrast, the VALID protocol is an easy-to-use and cost-effective protocol with robust labeling capability and tissue-clearing compatibility, which should greatly widen the application of lipophilic dyes in research on cerebrovascular complications in the tissue-clearing field.

Although two representative concentrations of gelatin (e.g., 2% and 4%) are employed in the VALID protocol for good vascular labeling performances in rodent animals, other concentrations can also be used for effective vascular labeling. For example, lower concentrations (e.g., 0.5% and 1%) and higher concentrations (e.g., 6% and 8%) would work as well. It should be noted that adding a high concentration of gelatin would lead to increased viscosity of the solution and quicker polymerization, which would make perfusion more difficult. Higher concentrations of gelatin for the VALID protocol could be used potentially in studying large mammalian animals with larger blood vessels. Additionally, our results also revealed that high-concentration gelatin would lead to dye hydrogel accumulation in organs containing a large cavity, such as the heart, and a lower concentration of gelatin would alleviate this accumulation effect. However, this may still have negative effects on 3D reconstruction for image analysis.

Several previous studies also involved gelatin in different solutions for casting vessel, such as Indian ink and fluorescein isothiocyanate (FITC)-conjugated albumin.^51,52 However, both of them have non-negligible limitations. For example, Indian ink could not express fluorescence, therefore it is quite limited in current fluorescence-based optical microscopy. The FITC-conjugated albumin could label the blood vessels with fluorescence, but the image quality is limited due to the high background. In this study, we prove that a variety of lipophilic dyes are effective with gelatin, covering the major channels used in optical tomography within the visible spectrum, providing more choice for labeling vessels. The price of lipophilic dyes is also much cheaper than other dyes such as FITC-conjugated albumin. Notably, previous studies only employed gelatin with a concentration of 2%; we demonstrate that the optimal concentrations for different tissues/organs may vary. These findings may provide useful insights for researchers to acquire vessel maps with high quality from different experimental needs.

Additionally, vessel casting is also employed in traditional medical imaging for visualizing the vascular networks. Traditional medical imaging can realize dynamic monitoring of blood vessels under different conditions, which is its unique advantage that could hardly be replaced by optical imaging. Additionally, traditional medical imaging is more suitable for imaging bones. Recently, progress has been achieved in non-invasive imaging of the entire intact murine cranial vasculature ex vivo by employing low-density polymer vascular casting, iterative sample processing, and micro-CT imaging.¹³ As for optical imaging, the combination of vessel-labeling, tissue-clearing, and optical imaging techniques provide a powerful workflow for high-resolution imaging of vascular networks of specific tissues of interest; however, other tissues not of interest may be influenced or removed during sample preparation. These techniques are now in the stage of rapid development and are expected to overcome such limitations. We believe that traditional medical imaging techniques and optical imaging techniques are both indispensable for the development of clinical investigations.

In recent years, pioneer studies have paved the way for imaging the vessels within long bones using tissue-clearing and light-sheet imaging techniques.^21,53,54 Our studies reveal that the VALID protocol is also effective for labeling vascular networks at capillary level within bones, such as mouse skull. The skull vasculature can also be investigated in recent studies using traditional medical imaging techniques, such as micro-CT.¹³ Comparatively speaking, the main advantage of VALID is its high resolution and ability to simultaneously study the vasculature and other cell types assisted by immunolabeling. However, VALID needs to remove the skull from the mouse body and clear the samples, which may cause potential sample deformation, and micro-CT circumvents this limitation because it does not require tissue extraction and clearing. Choosing which pipelines depends on the different experimental needs.

In summary, VALID is an easy-to-use and cost-effective protocol with robust labeling capability and tissue-clearing compatibility. VALID could also work well with immunolabeling for simultaneous investigation of vasculature and different cell types. The developed VALID protocol will provide powerful tools in the tissue-clearing field for studying vasculature within different tissues. VALID is expected to provide a powerful tool for evaluating vascular lesions in pathological disease models.

Limitations of the study

This study has several potential limitations. First, while we proved that the addition of gelatin obviously enhances the compatibility of lipophilic-dye-labeled vasculature with organic solvents, we did not obtain positive results with commonly used detergents because the obtained gelatin hydrogel would be dissolved by the detergent solutions. Second, the compatibility of VALID labeling with solvent-based methods varied across different organs, presumably because the brain vasculature possesses more strict permeability due to the existence of the blood-brain barrier. Additionally, the mice labeled with VALID are not perfused with PFA for immediate fixation, and all tissues were post-fixed after harvesting for complete preservation of labeled fluorescence signals. However, this might lead to a failure to correctly capture the time-sensitive events in the tissue, such as expression of immediate-early genes (IEGs). These limitations might be resolved by selecting new hydrogels or optimizing the fixation procedure in future studies.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-GFAP primary antibody	Proteintech	Cat#16825-1-AP, RRID:AB_2109646
Goat anti-rabbit secondary antibody Alexa Fluor 488	Abcam	Cat#ab150077, RRID:AB_2630356
Anti-CD31 Alexa Fluor 647	BioLegend	Cat#102416, RRID:AB_493410
Anti-Tyrosine Hydroxylase Rabbit pAb	Servicebio	Cat#GB11181, RRID: AB_2921651
NeuN Polyclonal antibody	Proteintech	Cat#26975-1-AP, RRID:AB_2880708
Parvalbumin Polyclonal antibody	Proteintech	Cat#29312-1-AP, RRID: AB_2918280
CD31 Mouse Monoclonal antibody	Proteintech	Cat#66065-2-Ig, RRID: AB_2918476
MAP2 antibody	Proteintech	Cat# 17490-1-AP, RRID:AB_2137880
Iba1 antibody	Abcam	Cat# ab5076, RRID:AB_2224402
Chemicals, peptides, and recombinant proteins
PBS	Sigma-Aldrich	Cat#P3813
Paraformaldehyde	Sigma-Aldrich	Cat#158127
Pork skin gelatin	Sigma-Aldrich	Cat#V900863
DiI	Aladdin	Cat#D131225
DiO	Aladdin	Cat#D131213
DiD	Thermo Fisher Scientific	Cat#D7757
MeOH	Sinopharm Chemical Reagent Co. Ltd.	Cat#10014118
EtOH	Sinopharm Chemical Reagent Co. Ltd.	Cat#10009218
Tert-butanol	Sigma-Aldrich	Cat#360538
Tetrahydrofuran	Sigma-Aldrich	Cat#186562
DBE	Sigma-Aldrich	Cat#108014
MXDA	Tokyo Chemical Industry	Cat#D0127
Sorbitol	Sigma-Aldrich	Cat#85529
Triton X-100	Sigma-Aldrich	Cat#T8787
L. esculentum (Tomato) lectin	Vector Laboratories	Cat#DL1178
Sodium azide	Sigma-Aldrich	Cat#S2002
DAPI	Thermo Fisher Scientific	Cat# 62248
Wheat germ agglutinin, Alexa Fluor ™ 647 conjugate	Thermo Fisher Scientific	Cat# W32466
Evans Blue	Sigma-Aldrich	Cat# E2129
Experimental models: Organisms/strains
C57BL/6J	Jackson Laboratories	Cat# JAX:000664, RRID:IMSR_JAX:000664
Software and algorithms
Matlab	Mathworks	https://www.mathworks.com/products/matlab.html
Imaris	Bitplane	https://www.bitplane.com/imaris/imaris
Fiji	55	https://imagej.nih.gov/ij/
Other
Conical tubes (50 mL)	Corning	Cat#430828
Conical tubes (15 mL)	Corning	Cat#430790
Syringe pump	Leadfluid	Cat#TFD04
Peristaltic pump	Leadfluid	Cat#BT100L

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Prof. Dan Zhu (dawnzh@mail.hust.edu.cn).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact without restriction. Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact Prof. Dan Zhu (dawnzh@mail.hust.edu.cn).

Experimental model and subject details

Animals

Wild-type mice (C57BL/6J, 8–12 weeks old) were used in this study. Animals were housed in a specific-pathogen-free (SPF) animal house under a 12/12 h light/dark cycle with unrestricted access to food and water. All animal experiments were performed under the Experimental Animal Management Ordinance of Hubei Province, P. R. China, and the guidelines from the Huazhong University of Science and Technology and were approved by the Institutional Animal Ethics Committee of Huazhong University of Science and Technology.

MCAO mouse model

Adult (8–10 weeks) male C57BL/6J mice were used for the creation of the MCAO model. Surgery was performed as described previously.⁵⁶ Briefly, mice were anesthetized using 1% isoflurane, and body temperature was maintained at 37.0 ± 0.5°C with a heating pad. The left common carotid artery, external carotid artery, and internal carotid artery were then isolated. Next, a silicon resin-coated nylon monofilament was gently inserted from the common carotid artery to the internal carotid artery ending until reaching the origin of the left middle cerebral artery to occlude the middle cerebral artery. Blood supply was restored by withdrawing the monofilament 60 min after MCAO. After surgery, mice were placed in warm cages for recovery and subsequently returned to their cages. The mice were sacrificed seven days post-stroke.

Method details

Preparation of the dye solution

The lipophilic dyes stock solution was prepared by dissolving 60 mg of DiI powder (Aladdin, D131225) in 10 mL of 100% ethanol and stored in the dark at room temperature. The DiI working solution for the traditional protocols was prepared by adding 200 μL of the DiI stock solution to 10 mL of diluent (0.01 M PBS and 5% (w/v) glucose at a ratio of 1:4). The DiI working solution for VALID was prepared by adding 200 μL of the DiI stock solution to 10 mL of a gelatin-based solution (2% (w/w) or 4% (w/w) porcine skin gelatin type A (Sigma-Aldrich, V900863) dissolved in 0.01 M PBS or distilled water). The VALID solutions based on DiO or DiD were prepared using similar procedures. Note: the gelatin-based solution should be stored in warm conditions (40°C) until perfusion.

Vascular labeling

Vascular labeling using VALID

Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/20 mg/kg), then transcardially perfused with 0.01 M PBS to wash out the blood. Next, 10–15 mL of the VALID working solution was perfused under warm conditions to keep the solution liquid. After perfusion, mouse bodies were kept at 4°C overnight to solidify the gel. Mouse organs were then harvested and post-fixed in 4% PFA overnight. The mouse skull was carefully removed from the perfused mouse body to avoid unnecessary sample distortion. The dura was removed from the skull during the sample harvesting procedure. The harvested skull was then post-fixed in 4% PFA overnight.

Vascular labeling using traditional protocols without VALID

Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/20 mg/kg). Anesthetized mice were first perfused with 0.01 M PBS (Sigma, P3813), followed by 10–15 mL of the lipophilic dyes working solution at a rate of 1–2 mL/min. During perfusion with the lipophilic dyes solution, the color of mice’s ears, noses, and palms will turn slightly purple. Mouse organs were then harvested and post-fixed in 4% PFA overnight.

Vascular labeling using lectin

DyLight 649-conjugated L. esculentum (Tomato) lectin (LEL-D649, DL-1178, Vector Laboratories) was used to co-label vasculature alongside VALID. Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/20 mg/kg). LEL-D649 was diluted in saline to a concentration of 0.5 mg/mL and injected into anesthetized mouse body via tail vein (0.1 mL per mouse). After injection, the animals were placed in a warm cage for 30 min before perfusion. After labeling and perfusion, the desired organs were excised from the animal bodies and post-fixed overnight in 4% PFA at 4°C.

Vascular labeling using CD31 antibodies

Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/20 mg/kg). Alexa Fluor 647 anti-mouse CD31 antibody (CD31-A647, 102,416, BioLegend) was diluted in saline (10–15 μg) and injected into the anesthetized mouse body via tail vein (total volume of 200 μL). After injection, the animals were placed in a warm cage for 30 min before perfusion. After labeling and perfusion, the desired organs were excised from the animal bodies and post-fixed overnight in 4% PFA at 4°C.

Vascular labeling using WGA

Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/20 mg/kg). Alexa Fluor 647 conjugated wheat germ agglutinin (WGA-A647, W32466, Thermo fisher scientific) was diluted in saline (10–15 μg) and injected into the anesthetized mouse body via tail vein (total volume of 200 μL). After injection, the animals were placed in a warm cage for 30 min before perfusion. After labeling and perfusion, the desired organs were excised from the animal bodies and post-fixed overnight in 4% PFA at 4°C.

Vascular labeling using Evans blue

Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/20 mg/kg).The Evans blue solution (Sigma-Aldrich, E2129) (2% in PBS, 200 μL) was injected into the anesthetized mouse body via tail vein (total volume of 200 μL). After injection, the animals were placed in a warm cage for 30 min before perfusion. After labeling and perfusion, the desired organs were excised from the animal bodies and post-fixed overnight in 4% PFA at 4°C.

Immunolabeling

Fixed brain slices were blocked in 0.2% PBST (0.2 vol % Triton X-100 in PBS) containing 6% goat serum at 37°C for 3–6 h. The blocked samples were then subjected to primary antibodies in 1–2 mL 0.2% PBST containing 3% goat serum and 0.01% (w/v) sodium azide for 1–2 days. Next, the samples were washed with 10 mL of 0.2% PBST several times and incubated in secondary antibodies in 1–2 mL of 0.2% PBST containing 3% goat serum and 0.01% (w/v) sodium azide for 1 day. All operating steps were performed at room temperature and included gentle shaking.

For 3D immunolabeling, the samples were first pre-treated with a sequence of methanol solution: 20%, 40%, 60%, 80%, 100%, 80%, 60%, 40%, 20%, 1 h for each step, then washed with PBST for 1 h. Pre-treated samples were incubated in 0.2% PBST containing 20% DMSO and 0.3 M glycine at 37°C overnight, blocked in 0.2% PBST containing 10% DMSO and 6% goat serum at 37°C for 1 day, washed in PTwH (PBS–0.2% Tween 20 with 10 mg/mL heparin) overnight and then incubated with primary antibody dilutions in PTwH containing 5% DMSO, 3% goat serum and 0.01% (w/v) sodium azide at 37°C with gentle shaking on a shaker for 2–3 days. Samples were then washed for 1 day with PTwH and then incubated with secondary antibodies diluted in PTwH containing 3% goat serum and 0.01% (w/v) sodium azide at 37°C with gentle shaking for 2–3 days. Sections were finally washed in PTwH for 1 day and stored in PBS at 4°C prior to clearing.

Tissue clearing protocols

We used different tissue clearing protocols, including both solvent-based clearing methods and aqueous-based methods.

Solvent-based clearing protocols were performed as follows:

Tissue clearing based on MeOH and DBE was performed as previously reported with slight modifications.²⁹ In brief, fixed tissues are dehydrated using MeOH solutions at concentration gradients of 50%, 70%, 80%, 100%, 100%. The dehydrated tissues are then incubated in pure DBE for refractive index matching. The time for each dehydration step (50%–100% MeOH) is 4 h (tissue slices)/12 h (whole organs) and the time for refractive index matching with DBE is 4 h (tissue slices)/12 h (whole organs). All steps are conducted at room temperature, with gentle shaking.

Tissue clearing based on EtOH and DBE was performed as previously reported with slight modifications.⁵⁷ In brief, fixed tissues are dehydrated using EtOH solutions at concentration gradients of 50%, 70%, 80%, 100%, 100%. The dehydrated tissues are then incubated in pure DBE for refractive index matching. The time for each dehydration step (50%–100% EtOH) is 4 h (tissue slices)/12 h (whole organs) and the time for refractive index matching with DBE is 4 h (tissue slices)/12 h (whole organs). All steps are conducted at room temperature, with gentle shaking.

Tissue clearing based on TB and DBE was performed as previously reported with slight modifications.⁵⁰ In brief, fixed tissues are dehydrated using TB solutions at concentration gradients of 30%, 50%, 70%, 80%, 90%, 100%, 100%. The dehydrated tissues are then incubated in pure DBE for refractive index matching. The time for each dehydration step (50%–100% TB) is 4 h (tissue slices)/12 h (whole organs) and the time for refractive index matching with DBE is 4 h (tissue slices)/12 h (whole organs). All steps are conducted at 35°C, with gentle shaking.

Tissue clearing based on THF and DBE was performed as previously reported with slight modifications.⁵⁸ In brief, fixed tissues are dehydrated using THF solutions at concentration gradients of 50%, 70%, 80%, 100%, 100%. The dehydrated tissues are then incubated in pure DBE for refractive index matching. The time for each dehydration step (50%–100% THF) is 4 h (tissue slices)/12 h (whole organs) and the time for refractive index matching with DBE is 4 h (tissue slices)/12 h (whole organs). All steps are conducted at room temperature, with gentle shaking.

We also employed aqueous clearing methods, MACS. MACS was executed as demonstrated previously.³⁴ Briefly, fixed samples were serially incubated in MACS-R0, MACS-R1, and MACS-R2 solutions, with gentle shaking at room temperature. MACS-R0 was prepared by mixing 20% (vol/vol) MXDA with 15% (wt/vol) sorbitol (Sigma, 85529) in dH₂O; MACS-R1 was prepared by mixing 40% (vol/vol) MXDA with 30% (wt/vol) sorbitol dissolved in 1× PBS; MACS-R2 was prepared by mixing 40% (vol/vol) MXDA with 50% (wt/vol) sorbitol in dH₂O.

Imaging

Bright filed images were taken using a digital camera. Cleared brain and other samples were imaged with a light sheet microscope (LiToneXL, Light Innovation Technology, China) equipped with a 4× objective lens (NA = 0.28, working distance (WD) = 20 mm). Thin light sheets were illuminated from the four sides of samples, and a merged image was saved. A laser-scanning confocal microscope (LSM710, Zeiss, Germany) was used to image tissue slices; a 5× objective lens (FLUAR, NA = 0.25, WD = 12.5 mm), a 10× objective lens (FLUAR, NA = 0.5, WD = 2 mm), a 20× objective lens (FLUAR, NA = 0.8, WD = 0.55 mm) a 40× objective lens (FLUAR, NA = 1.4, WD = 0.13 mm) were utilized for imaging. The Zen 2011 SP2 software (Version 8.0.0.273, Carl Zeiss GmbH, Germany) was used to collect data.

Image data processing

All raw image data were collected in a lossless TIFF format (8-bit for confocal microscopy and 16-bit images for light-sheet microscopy). Processing and 3D rendering were executed on a Dell workstation with an 8 core Xeon processor, a 256 GB RAM, and an Nvidia Quadro P2000 graphics card. We used Imaris (Version 7.6, Bitplane AG) and Fiji (Version 1.51n) for 3D and 2D image visualizations, respectively. The 16-bit light sheet images were transformed into 8-bit images by Fiji to enable fast processing by other software, such as Imaris. The image stitching was performed as follows: The image acquired by Zeiss confocal microscope and LiToneXL light sheet microscope could be automatically stitched via the data acquisition software provided by the company. If manual stitching are needed, we recommended using the ‘Grid/Collection stitching’ plugin in Fiji. In brief, the ‘Grid: column-by-column’ mode was selected and the stitching order was depending on the mode of data acquisition (We used Down&Right). Then, the parameters such as grid size, tile overlap value, as well as directory for input and output files were manually determined based on the image data. Then the plugin would run and stitch the data with the selected parameters.

Quantification and statistical analysis

Quantifications

To evaluate the relative transparency of cleared samples using the bright field images, we estimated the average Michelson contrast (C) in different regions of each image, including brain regions and background regions. The average Michelson contrast (C) was calculated as follows:

$$\mathrm{C}=\frac{I_{\mathit{max}}+I_{\mathit{min}}}{I_{\mathit{max}}-I_{\mathit{min}}}$$

Where I_max indicates the mean gray value of selected rectangular regions of interest (ROIs) in the light parts of the images and I_min denotes the mean gray value of selected ROIs in the dark parts of the images. Transparency (T) was then calculated using the following equation:

$$\mathrm{T}=\frac{C_{brain}}{C_{background}}$$

Where C_brain indicates the average Michelson contrast of regions beneath the cleared brain and C_background denotes the average Michelson contrast of the background regions beyond the cleared brain.

To evaluate the efficiency of the proposed VALID protocol and other vessel-labeling methods, as well as the fluorescence compatibility of VALID with different solvent-based clearing protocols, overlap ratios and signal-to-background ratios were calculated. First, the maximum intensity projection (MIP) of image stacks from the surface of the tissue (50-μm-thickness) was obtained. Then the “threshold” function of the ImageJ software was used to extract the vascular signal from the MIP image, and the mean value of the signal area was determined as the average intensity of the fluorescence signal. The area without blood vessels in the MIP image was manually selected using the marking function, and the mean value of this area was calculated as the average intensity of the background. The signal-to-background ratio was defined as the mean signal intensity divided by the mean background intensity. The relative signal-to-background ratio was defined as the signal-to-background ratio after clearing divided by the signal-to-background ratio before clearing. The overlap ratio was calculated as the number of DiI-positive pixels divided by the number of CD31-positive/lectin-positive pixels. The preservation ratio of vascular information was calculated as the area of segmented vascular information after clearing divided by the area of segmented vascular information before clearing.

3D vascular reconstruction and quantification were performed by Imaris using a pipeline described in previous studies with slight modifications. In brief, the “Threshold (loops)” algorithm in filaments module was used to trace and quantify blood vessels. The parameter for filaments diameter was automatically set by the algorithm according to the voxel size. The voxel size of our data is 0.83 μm × 0.83 μm×2 μm. Threshold for extracting vascular information was manually adjusted to ensure the complete recognition of all blood vessels. After reconstruction of the vascular network, the local vessel density, average radius, and the vessel segment length were obtained using the “statistics-selection” in filaments module.

Statistical analysis

Data are presented as the mean ± SD and were analyzed using SPSS software (Version 22, IBM, USA) with 95% confidence interval. Sample sizes are indicated in the figure legends. For analysis of statistical significance, the normality of the data distribution in each experiment was checked using the Shapiro-Wilk test. The variance homogeneity for each group was evaluated by Levene’s test. p values were calculated using an independent-sample t test (two-sided) to compare data between two groups in Figures 5F, 7E and 7F. p values were calculated using one-way ANOVA followed by the Bonferroni post hoc test to compare data in Figures 3B, 5C, and 5D, S5B, S5C, and S5E. In this study, p < 0.05 was considered significant (∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001).

Published: February 8, 2023

Data and code availability

• •This paper does not report Standardized datatypes. All data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.