1. Introduction

Analysis of complex structure of the brain is crucial to our understanding of how the brain works. The brain contains a variety of types of neurons, glial cells and complex blood vessels. These structures work together to keep brain functioning [1]. Multiple structures are affected in brain diseases [2,3]. For instance, Alzheimer’s disease (AD) destroys neurons and blood vessels, affecting the morphology and distribution patterns of glial cells, and is often accompanied by the presence of age plaques and tangles [4]. Therefore, multi-structure simultaneous labeling and fine imaging is very important in studying such diseases. It is helpful for researchers to understand the spatial position and morphological characteristics of different structures and the relationship between them to explore the complex neural circuitry of the brain and the pathogenesis of brain diseases.

Fluorescence labeling strategies play a pivotal role in understanding of the multiple structures, and the connections among different structures of the complex nervous system [5,6]. There are various types of fluorescent dyes with more flexible labeling approach. For instance, DyLight [7], Alexa [8], and fluorescein isothiocyanate (FITC) [9], all contain different fluorescent colors. Therefore, multicolor simultaneous labeling could be realized [6,9,10]. On the other hand, these dyes are compatible with a variety of labeling methods both in vivo or in vitro. For example, we combine fluorescent dyes with neuroanatomical tracer Cholera toxin subunit B (CTB) or Phaseolus vulgaris leuco-agglutinin (PHAL) to achieve nerve tracking [11], with lectin to label blood vessel by perfusion or caudal vein injection [9], and with secondary antibodies in immunohistochemistry to stain and localize specific structures like neurons, fibers and blood vessels. Moreover, fluorescent dyes and fluorescent proteins are also commonly used for multiple structure labeling in the same tissue [9,12–14].

Multiple structures and brain regions efficiently work together to keep the brain functioning. It is necessary to explore the architecture on a whole-brain scale. Resin embedding [15–18] and tissue clearing [9,19,20] in combination with advanced microscopy imaging are two typical techniques that can be used to perform high-resolution 3D reconstruction of whole organs. Tissue clearing requires the use of various chemical reagents and an extremely time-consuming process, it is difficult to acquire fine information labeled with fluorescent proteins and dyes in large-scale samples [9,20,21]. Fluorescence micro-optical sectioning tomography (fMOST) combines microscopic imaging with continuous micro-thickness sectioning of the resin embedded whole samples has been widely used to acquired long distance projections information with single axon resolution [16,22]. The samples for three-dimensional continuous imaging need to be embedded with resin to obtain enough hardness for continuous micron-thickness cutting to achieve high resolution imaging.

In the last few years, researchers have made a lot of efforts to improve the plastic embedding method to achieve higher fluorescence retention and lower background noises for various fluorescent proteins, including green fluorescent protein (GFP), blue fluorescent protein (BFP), tdTomato and mCherry (popular red fluorescent protein) labeled samples [17,23–26]. With the use of high-resolution imaging, fine complex structures could be visualized for quantitative studies in the whole brain. Recent advances in these tools enabled mapping neuronal circuits labeled with fluorescent proteins to reveal long-range connections between different brain areas [15,18,27].

Since the fluorescent dyes could be easily quenched by chemical reagents during resin embedding, the existing methods has been preliminarily verified that resin could be used for immunolabeled tissues. However, it has not been used for the reconstruction of fluorescent dyes labeling structures due to the low signal to noise ratio [24]. These drawbacks limited the application of multiple fluorescent dyes labeling biological samples in whole brain imaging with submicron resolution. To date, retaining the intensity of the fluorescent dyes in plastic embedding of large samples is a technical challenge that need to be overcome.

Here, we identified the key factors affecting fluorescent dyes in large sample resin embedding process and modified the glycol methacrylate (GMA) embedding method for a variety of fluorescence labeled samples. We demonstrated the ability of this optimized embedding method to preserve the fluorescent signals of multiple fluorescent dyes and proteins. Neurons, blood vessels and senile plaques labeled by different fluorescent dyes could be clearly detected after embedded. Furthermore, by using high-resolution imaging system, we simultaneously acquired the distribution of vasoactive intestinal polypeptide (VIP) neurons labeled with tdTomato and blood vessels labeled with BSA-FITC within three days. Quantitative analysis revealed the unique distribution patterns and morphological characteristics of neurons and blood vessels in the cortex. These results demonstrated that the GMA-T has great potential to map the anatomical structure of large-scale organs with multi-color fluorescence labeling.

2. Materials and methods

2.1 Animals

Five 8-week-old C57BL/6J male mice, an 8-week-old Thy1-GFP M-line transgenic mouse, a 10-month-old 5×FAD transgenic mouse and five 8-week-old VIP-ires-cre: Ai14 mice were used. The VIP-ires-Cre mice and Cre-reporter expressing tdTomato Ai14 mice (007903) were obtained from the Jackson lab. VIP-ires-cre: Ai14 mice were generated by crossing male VIP-ires-Cre mice with female tdTomato Cre reporter Ai14 mice. All animal experiments were approved by the Institutional Animal Ethics Committee of Huazhong University of Science and Technology.

2.2 Perfusion

The mice were anesthetized with 10% (wt/vol) Urethane (sigma, U2500) and 2% (wt/vol) chloral hydrate (sigma,23100) mixture (1ml/100g body weight) via intraperitoneal injection. Then, the animal was perfused with 0.01M Phosphate-Buffered Saline (PBS) at 37 °C in 7 min. After another 7 min, the perfusion buffer was replaced by 4% Paraformaldehyde (PFA) in PBS (4 °C or keep on ice). Finally, the brain was removed from the skull and post-fixed in 4% PFA at 4 °C for 24h.

2.3 Fluorescence labeling

Virus injection.

To obtain samples expressing blue fluorescent protein (BFP), a C57 mice were injected with 100 nL of AAV-DIO -BFP (4.2× 10¹² gc/ml) into the primary motor cortex (M1) to express BFP. 21 days after injection, the mice were perfused. Then, 100 µm brain slices were sectioned using a vibratome (Leica VT 1200S, Germany)

Immunostaining.

A C57 mouse brains were sectioned at 70 µm thickness on a microtome. Brain Sections were washed three times for 15 min in 0.1 M PBS at room temperature (RT), then permeabilized using 0. 3% Triton-X 100 in PBS, blocked with 5% bovine serum albumin (BSA, Sigma) and 0.3% Triton in PBS, and subsequently incubated with anti-Choline Acetyltransferase (Chat) (1:500, goat, Millipore, AB144P) and anti-parvalbumin (PV) (1:1000, mouse, Millipore, MAB1572), respectively, at 4 °C overnight. The brain sections were washed three times for 15 min in 0.1 M PBS at RT and incubated with secondary antibodies (1:1000; Alexa Flour 594, Rb-Anti-Gt; 647, Gt-Anti-Mouse) for 1 h respectively at 37 °C. After the washing steps, the sections were mounted and coverslipped with Fluoro-gel. The representative images were acquired with a Zeiss 710 LSM confocal microscope (Zeiss, Jena, Germany).

Labeling of the blood vessels.

A VIP-ires-cre: Ai14 mouse received an intravenous (i.v.) injection of 50 µg of lectin-DyLight 488 (Vector lab, DL-1174), a C57 mouse received an intravenous (i.v.) injection of 50 µg of lectin-DyLight 594 (Vector lab, DL-1177). 20 minutes after injection, the mice were perfused. In addition, a VIP-ires-cre: Ai14 mouse and two C57 mice were perfused firstly with 50 ml of PBS and then with 50 ml of PFA. This was followed by perfusion with 5 ml of a fluorescent gel perfusate with the body of the mouse tilted 30° head down. Then, the mice were submerged in ice water to rapidly cool and solidify the gel. The brains were carefully extracted after 30 min of cooling. Fluorescent gel solutions contained porcine skin gelatin type A (Sigma-Aldrich, G1890) and fluorescein-labeled-albumin (Sigma, A9771). A 10% (w/v) solution of gelatine was prepared in boiling PBS and allowed to cool to <50 °C. Then it was combined with 1% (w/v) albumin-FITC (BSA-FITC), kept at 40 °C with stirring before perfusion.

2.4 Resin embedding

Embedding procedures were performed on brain slices and whole brain. For the brain slices at 100 µm thickness, slices were dehydrated in a graded ethanol series (50%, 70% and 95% ethanol), changing from one concentration to the next every 5 min at 4 °C. After dehydration, the brain slices were immersed in a graded glycol methacrylate (GMA) series (Ted Pella, USA), including 0.2% Sudan black B (SBB, Sigma-Aldrich, 4197-25-5) or 0.3%-5% True BlackTM (TB, Biotiom, 23007) (70%, 85%, and 100% GMA for 15 min each, 100% GMA overnight and pre-polymerization GMA solution for 24 h at 4 °C). Finally, the brain slices were embedded in a vacuum oven at 35 °C for 24 h.

There are two main steps in GMA-T embedding processing for whole brain (Fig. 1(A)). Step1: for optimal embedding and good morphology, tissue was fixed before processing. Step2: this process was carried out by immersing tissues in a series ethanol and resin solutions, background fluorescence inhibitor TB was added in the resin. Specifically, brains were dehydrated in a graded ethanol series every 1 h at 4 °C and immersed in a graded GMA-T series for 2 h each and 100% GMA-T overnight at 4 °C. Subsequently, the samples were impregnated in a pre-polymerization GMA-T solution for 3 days at 4 °C. Finally, the whole brains were embedded in a vacuum oven at 35 °C for 24 h. The 100% GMA-T solution comprised 52.5 g of A solution (glycol methacrylate, GMA), 2.8 g of deionized water, 44.1 g of B solution (n-butyl methacrylate, BMA), 0.7 mL TB and 0.8 g of 2, 2’-Azobis-(2,4-dimethylvaleronitrile) (ABVN) as an initiator. The graded ethanol series were prepared from absolute ethyl alcohol and distilled water. The 70% and 85% GMA solutions (wt/wt) were prepared from 95% ethanol and 100% GMA.

 

Fig. 1. Main factors affecting fluorescent dyes in plastic embedding. (A) The resin embedding process for whole-brain and the fluorescence signal of DyLight dyes before and after embedding. Scale bar: 20 µm. (B) The effect of organic reagents, polymerization temperature and background inhibitor on fluorescent dye. Projection thickness was 30 µm. Scale bar: 20 µm. (C) The absorption spectrum of TB and SBB. (D) The fluorescence emission spectra of TB and SBB.

Download Full Size | PDF

2.5 Imaging

Commercial confocal microscope imaging was performed with a Zeiss LSM 710 system. BFP was excited with an excitation wavelength of 405 nm. DyLight 488, Alexa 488, and GFP were excited with an excitation wavelength of 488 nm. tdTomato was excited with an excitation wavelength of 514 nm. DyLight 594 and DANIR-8c were excited with an excitation wavelength of 594 nm. Alexa 647 were excited with an excitation wavelength of 647 nm. Images were obtained using a 20× dry objective with 0.8 NA. The resulting images were analyzed with ImageJ (NIH) software. Whole-brain imaging was performed with Brain-wide positioning system (BPS) [17]. Briefly, GMA-T embedded whole brain sample was fixed in the anterior-posterior (A-P) direction on the stage. The resin-embedded sample was sectioned at 2 µm thickness and imaged every plane. Microtome removed the superficial layer, and revealed a smooth surface, then subsequently images were acquired through mosaic scanning in each layer. The sectioning-imaging cycles were repeated to acquire whole-brain imaging data at a voxel resolution of 0.32 µm×0.32 µm×2 µm requiring 3 days to achieve the final 3D reconstruction.

2.6 Image preprocessing

The brain-wide images preprocessing for mosaic stitching used in this study has been previously described in detail [17]. Briefly, the mosaics of each coronal section were stitched to obtain an entire section. Image preprocessing was performed in C++ and optimized in parallel using the Intel MPI Library (v3.2.2.006, Intel, Santa Clara, CA, USA). All full coronal sections were saved for the final 3D reconstruction.

2.7 Fluorescence intensity quantitative analysis

The fluorescence intensity was measured using ImageJ software. First, we usd the oval-selection tool, and the grey values of all pixels were selected in the object region. Then, we used the histogram tool, and the mean fluorescence intensity of the oval area was measured as the fluorescence intensity of the object region. Next, the mean fluorescence intensity was used as the gray value (pixel value), and the normalized fluorescence intensity was obtained by normalizing the gray value between 0 and 255.

2.8 Reconstruction and data analysis

To ensure the accuracy of the analysis results, three skilled persons performed back-to-back manual segmentation and counting. We then manually checked to avoid missed or repeated identification. We segmented the primary motor cortex (M1), somatosensory (S1) and the primary visual (V1) in the mice brain according to Allen CCF3 [28] by manual validation and amendment in Amira software (v6.1.1, Mercury Computer Systems, San Diego, CA, USA). All the segmentation results were saved as 3D-TIFF files. Cell counts included all fully visible neurons within M1, S1 and V1 with Manual Cell Counter plugin in the ImageJ (Version 1.48, NIH, United States), while the volume of the cells was calculated with the Voxel Counter plugin in the ImageJ. Blood vessels were reconstructed and measured with the Imaris (Version 7.6, Bitplane AG) software. The vascular networks were identified with the FilamentTracer function in Imaris. Three skilled persons performed back-to-back manual validation and amendment to ensure that proper vessel connections were made. The measurements include vessel diameter, vascular length was generated and saved as excel sheets. We used the filament editor module of Amira software in 3D by a human-machine interaction to reconstruct the morphology of neurons. We continuously loaded data blocks along soma and fibres into Amira. We assigned initial and terminal points of fibres, and then Amira automatically calculated the path that the fibre passed along between these two points. Repeat this procedure until the whole neural morphology was reconstructed. Back-to-back manual reconstructions and validation done by three skilled persons. In most cases, manual counting and segmentation by skilled persons was considered as the ground truth. Automatic analysis usually calculates the two parameters of sensitivity and specificity to evaluate the accuracy of automatic counting and segmentation by compared with skilled person identified manually. Therefore, the data obtained from the manual counting and segmentation were reliable. Graphs and statistical analysis were created with GraphPad Prism 8.4.0 (Graphpad Software, CA, USA). All result values were given as means ± SEM. P-values ≤0.05 were considered significant.

3. Results

3.1 Increases the signal-to-noise ratio of fluorescent dyes labeled samples with TB

Resin embedding includes two major stages with multiple steps in each part [23]. The first stage is the pretreatment of samples: the perfused brain needs to be post-fixed in 4% PFA and rinsed with PBS (the first row in Fig. 1(A)). The second stage is the dehydration and resin embedding of samples. Ethanol, GMA resin and background fluorescent inhibitors are used for sample processing, and subsequently, polymerization happens in a vacuum oven (the second line in Fig. 1(A)). The vascular branches labeled with DyLight 594 were clearly visible after post-fixed (the fluorescence image in the first row of Fig. 1(A)). We found that the second step (from dehydration to polymerization) had a great effect on the fluorescence intensity of the DyLight 594.The morphology of the blood vessel was not visible after resin embedding (the fluorescence image in the second row of Fig. 1(A)).

We investigated the effects of processing of step 2 (Fig. 1(A)) include chemical reagents (ethanol and resin), background inhibitor (SBB) and polymerization temperature (35°C, the lowest polymerization temperature of GMA resin [23]) on the fluorescence retention respectively, as shown in Fig. 1(B).

Following post-fixation with 4% PFA for 24 h, the Dylight 594 labeled brain was cut into 100-µm-thick slices. The brain slices were divided into three groups, the first group was immersed in a graded ethanol series (50%, 70% and 95% ethanol) and resin series (70%, 85%, and 100% GMA, changing from one concentration to the next every 5 min), the second group was placed in the oven at the polymerization temperature (at 35 °C), the third group was immersed in SBB solution (black powder diluted with 75% ethanol). Before and after treatment, the same area of brain slices (the arrow showed in picture) was imaged via confocal microscopy using the same parameter. The figures in the first row showed the florescent intensity of DyLight 594 and the vascular structure on brain slice before treatment, while the second row showed the fluorescent intensity after treatment. The vascular structures were still clearly visible after the treatment of ethanol, resin and the polymerization temperature (the arrow showed in the first and third column of Fig. 1(B)). However, after the treatment of the background fluorescence inhibitor, SBB, the fluorescence intensity of DyLight 594 decreased, vascular structures could not be visualized clearly as the arrow showed in second column of Fig. 1(B). Then, we confirmed that in existing resin-embedding technique SBB effectively decreased the intensity of dye.

The absorption spectrum (Fig. 1(C)) and the histochemical investigations [29,30] of TB indicated that TB could inhibit auto-fluorescence in multiple wavelengths, which showed the similar absorption characteristics as SBB. In previous studies, as the background inhibitor, SBB was used in multicolor fluorescent proteins labeled whole-brain embedding methods [17,23,24]. These results demonstrate that SBB could effectively reduce auto-fluorescence at multi-excitation-wavelength. As the absorption spectrum of SBB shown in Fig. 1(C), there is a large absorption peak at 500 to 700 nm. These results showed that TB not only could reduce auto-fluorescence in immunostaining on brain slices [29,30], also in embedding processes. Furthermore, no auto-fluorescence was observed in TB solution (Fig. 1(D)). We speculated that TB could be used to replace the background fluorescence inhibitor SBB in resin embedding method. However, TB combined with the whole-brain plastic embedding methods have not been reported in previous literature. Since TB is a background fluorescence inhibitor, the higher the concentration of TB, the better the inhibition effect on the background fluorescence in a certain concentration range but may reduce the preservation of fluorescence intensity. Therefore, it is necessary to select an appropriate concentration of TB to keep a balance between inhibiting the background fluorescence and retaining the fluorescence signal to improve the signal-to-noise ratio of multiple fluorescence images.

To quantify the effect of TB on fluorescent dyes and background fluorescence, we calculated the intensity of DyLight dyes in labeled blood vessels of mouse brain (the first row of Fig. 2(A) and (B)). Following post fixation with 4% PFA for 24 h, the brain was cut into 100-µm-thick slices. Then, we added different concentrations of TB to the dehydration agents. Before and after embedding in resin with different concentrations of TB, we imaged the brain slices via confocal microscopy and analyzed the fluorescence intensity with ImageJ software.

 

Fig. 2. The effect of different concentrations of TB on DyLight dyes and auto-fluorescence. (A, B) Effects of different concentrations of TB on DyLight594 and DyLight488 fluorescence signal. The projection thicknesses: 30 µm, Scale bar: 20 µm. (C, D) Fluorescent preservation ratios of DyLight594 and DyLight488 treated with TB at different concentrations. The effect of different concentrations of TB on auto-fluorescence. (n = 15, 19, 15, 16 respectively. Error bars represent SD. One-way ANOVA followed by Tukey’s post hoc tests. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.001) (E) Effects of TB on auto-fluorescence of the brain in the red channel. (F) The auto-fluorescence reduction rate in the red channel after staining with TB. (n=3, each column. Error bars represent SD. One-way ANOVA followed by Tukey’s post hoc tests) (G) 0.7% TB reduced auto-fluorescence in the green and blue channels. Scale bar: 200 µm.

Download Full Size | PDF

The results showed that the higher-concentration TB group showed not only reduced tissue auto-fluorescence (Fig. 2(E)-(G)) but also reduced DyLight dyes intensity (the second row of Fig. 2(A) and (B)). In the 0.3%, 0.5%, and 0.7% TB groups, the DyLight594 preservation rate was 85.36 ± 3.3%, 78.8 ± 4.0%, and 79.0 ± 4.9%, respectively, and the DyLight488 preservation rate was 86.115.5%, 83.1 ± 16.47%, and 82.5 ± 13.0%, respectively. In the 1% TB group, the DyLight594 and 488 dyes preservation rates were 75.8 ± 3.3% and 78.9 ± 14.4%, respectively, and in ≥2% TB groups, the DyLight dyes preservation rate was lower than 50% (Fig. 2(C) and (D)). The results indicated that higher concentrations of TB effectively decreased the intensity of DyLight dyes, especially for concentrations above 2%.

We then tested the effect of TB on auto-fluorescence. The mouse brain shows strong auto-fluorescence under the red, green and blue channels (the first row of Fig. 2(E) and (G)). The auto-fluorescence decreased with higher concentration of TB. In red channels of the 0.3% and 0.5% TB groups, the auto-fluorescence decreased by 69.8 ± 3.0% and 83.52 ± 6.0%, and in the 0.7% TB group, the auto-fluorescence decreased by 90.8%±4.7%. In ≥1% TB groups, it decreased ≥95%. Similarly, the mouse brain slices show lower auto-fluorescence under the green and blue channels after 0.7% TB staining (Fig. 2(G)). Therefore, 0.7% TB effectively retained over 80% of the intensity of DyLight dyes and inhibited over 90% of the background fluorescence.

To improve the signal-to-noise ratio of the fluorescent dye labeled samples, we replaced SBB with 0.7% TB in resin embedding and developed a new resin embedding method, GMA-T (Glycol methacrylate with TB). The embedding process is indicated in Fig. 1(A).

3.2 Fluorescence preservation with multiple fluorescence labeling strategies in the GMA-T method

To demonstrate that the GMA-T embedding method could preserve the structure of multiple fluorescent dye labels, we embedded mouse brain samples with multiple fluorescent dye labeling strategies. The 100-µm-thick brain slices were dehydrated in a graded series of ethanol solutions for 5 min, subsequently infiltrated in a graded series of GMA-TB solutions for 15 min, and then placed in the prepolymer of GMA-TB solutions. After 6 h, the brain slices were embedded in an oven at 35 °C. Including blood vessels labeled by lectin DyLight dyes through the tail vein and amyloid beta-protein (Aβ) staining with DANIR-8c fluorescent probe in 5XFAD mice [4], specific types of neurons with immunohistochemical staining.

The results showed that the structure of blood vessels and Aβ could be clearly displayed after GMA-T embedded (the larger arrow in Fig. 3(A)). The fluorescence preservation rate of DyLight488 dye was 99.4 ± 6.3%, DyLight594 dye was 99.0 ± 5.0% and DANIR-8c fluorescent probe was 81.0 ± 6.2% (Fig. 3(B)). Blood vessel walls could also be seen in single fluorescence images, which was a benefit from the retention of fluorescence of DyLight594 dye and the effective inhibition of the background noises (the larger arrow in Fig. 3(C)). Aβ labeled with DANIR-8c deposits in brain (the arrow showed in Fig. 3(C)) and the morphological characteristics of Aβ deposition around the blood vessel could also be clearly displayed (the lager arrow showed in Fig. 3(C)). The results showed that the GMA-T embedding method could preserve the fluorescence dyes with low background noises, which meant that the position and morphology of the labeled structures in the brain could be clearly visualized.

 

Fig. 3. The preservation of DyLight and DANIR-8c after GMA-T embedding. (A) DyLight 488 and DANIR-8c fluorescence signal after GMA-T embedding. The projection thicknesses: 30 µm. Scale bar: 20 µm. (B) Fluorescent preservation ratios of three fluorescent dyes after GMA-T embedded. (n=14, 18, 6) (C) Single fluorescence imaging of DyLight 488 and DANIR-8c after GMA-T embedded. Scale bar: 20 µm.

Download Full Size | PDF

Then, we tested the GMA-T method in samples labeled with Alexa dyes, which combined with secondary antibodies. Here, Alexa 594 labeled chat neuron in VDB, and Alexa 647 labeled PV neurons in cortex as shown in Fig. 4 A and B. The enlarged representation showed the neuronal morphology was still visible after the GMA-T embedding (the arrow showed in Fig. 4). The tiny fibers could also be preserved as the asterisk showed in Fig. 4. The results indicated that the GMA-T embedding was suitable for multiple fluorescent dyes labeled samples with multiple labeling methods.

 

Fig. 4. The preservation of Alexa dyes combined with second antibody in IHC after the GMA-T embedding. (A) The fluorescence signal of chat neuron labeled with Alexa Fluor 594. (B) The fluorescence signal of PV neuron labeled with Alexa Fluor 647. The projection thicknesses: 20 µm. Scale bar: 20 µm.

Download Full Size | PDF

In addition, we tested the GMA-T method in samples labeled with GFP, tdTomato, and BFP. Here, GFP labeled the axons of pyramidal neuron within Thy1-GFP transgenic mice, tdTomato labeled the VIP neurons within VIP-Cre: Ai14 mice, and BFP labeled the RV-infected neuronal nuclei in the motor cortex. The intensity of fluorescent proteins could be preserved after the GMA-T embedding. As indicated by the arrow in Fig. 5(B), after the GMA-T embedding, GFP and tdTomato labeled axons and branches and BFP-labeled neuronal nuclei were clearly observed. Therefore, fine structures by multiple labeling strategies with fluorescent markers could be preserved in the nervous system after GMA-T embedded.

 

Fig. 5. The preservation of fluorescent proteins after GMA-T embedding. (A) Comparison of the fluorescence intensity before and after the GMA-T embedding. (B) Fine structures labeled with fluorescent proteins after GMA-T embedding. The projection thicknesses: 20 µm. Scale bar: 10 µm, 5 µm, 10 µm from left to right.

Download Full Size | PDF

3.3 Simultaneous acquisition of the fine structures of neurons and blood vessels in the same whole brain

To confirm whether the GMA-T method is suitable for precise multiple fluorescent imaging in the whole brain, we embedded 3 whole brains labeled with two fluorescent strategies, tdTomato labeled all VIP neurons and BSA-FITC specifically labeled the blood vessels. VIP-cre/tdTomato and BSA-FITC are highly specific for VIP-express neurons [31] and blood vessels [9], respectively. In addition, the intensity of fluorescent dyes and proteins could be preserved after the GMA-T embedding. In combination with the precise whole-brain imaging system [17], tdTomato and BSA-FITC signals were obtained simultaneously at single-neuron resolution in the whole brain (Fig. 6(A)). The continuous 3D fluorescence images allowed us to quantify the exact distribution patterns and morphological characteristics of neurons and blood vessels in specific brain regions as well as their relationships. In this study, magnified view of the M1 on coronal sections (Fig. 6(B)) in whole brain images showed the distribution of neurons and blood vessels at different layers. Fluorescence images showed high signal-to-noise ratio, structural features of neurons and blood vessels were clear observed (Fig. 6(C)). Morphology of neurons and vessels could be clearly distinguished. Fluorescence signal could be extracted to reconstruct the accurate morphological of blood vessels and neurons (Fig. 6(D) and (E)).

 

Fig. 6. Whole mouse-brain neuronal and vascular imaging. (A) Whole brain 3D imaging of tdTomato-expressing neuron and BSA-FITC expressing the blood vessels acquired with fMOST. The projection thickness was 100 µm. (B) Enlarged image of boxed area in a showing A. Show VIP neurons and vessels in different layers of cortex. (C) 3D presentation of B. (D) Overlap between imaging(green) and automatic (yellow) segmentation. (E) Reconstructed VIP neurons in the M1 exhibited unipolar neuron (green line), bipolar neuron (yellow line) and multi polar neuron (rose red line) patterns. Scale bar: 200 µm. (F) Numbers of VIP neurons and vascular density in different layers of cortex. (G). Soma volume of VIP and diameter of the vessel in different layers of cortex. (I) Soma volume of VIP neuron in cortex. (H) Statistical results of three morphological VIP neurons in different layers of cortex.

Download Full Size | PDF

We quantified the distribution and morphology of VIP neurons and vessels in the cortex, including M1, S1, and V1 as shown in Fig. 6(F)-(H). VIP neurons were found throughout all layers in the cortex, but these neurons had a heterogenous distribution pattern. We calculated the proportions of VIP neurons and blood vessels in each layer (Fig. 6(F)). The number of VIP neurons was the lowest in layer I compared with all other layers, which contained less than 6% of VIP neurons. The number of VIP neurons was the highest in layer II/III, which contained over 60% of all layers. From this peak, the number of VIP neurons further decreased in the deeper layer (less than 20% in layer IV, V and VI respectively). However, the distribution of blood vessels in all layers was relatively uniform while the highest density was in layer Vin M1 (29% of all layers) and in layer IV in S1 (23% of all layers) and in layer IV in V1 (24% of all layers) (Fig. 6(F)). Layer VI showed the lowest blood vessels density (23%, 18% and 17% in M1, S1 and V1 respectively).

VIP neuronal cell bodies had different volumes in different layers as shown in Fig. 6(G). In S1 and V1, the volume of cell bodies was smaller in layer II/III (265.5 µm3 and 359.5 µm3 respectively), and their volumes increased as the depth increased in S1 (409 µm3, 422.5 µm3 and 435.0 µm3 in layer IV, V and VI). However, there was no significant difference in all layers in M1, 385 µm3 in layer I, 420.5 µm3 in layer II/III, 411 µm3 in layer V and 439.5 µm3 in layer VI. We also analyzed the diameter of blood vessels. The results showed that in M1 and V1, the diameter was the biggest in layer I (4.5 µm in M1 and 4.6 µm in V1) and smallest in layer IV/V(3.0 µm in M1 and 3.1 µm in V1). In S1, the diameter was the biggest in layer II (5.8 µm) and the smallest in layer I (2.6 µm), and the diameter decreased as the depth increased (Fig. 6(G)).

Additionally, we quantitatively analyzed the morphology of individual VIP neurons in the cortex as shown in Fig. 6(H). VIP neurons had similar morphological characteristics in the same layer of different areas in the cortex, but not in different layers. Each layer of the cortex contained unipolar, bipolar, and multipolar VIP positive neurons. The proportion of unipolar neurons was the smallest in each layer (less than 15%). On the contrary, the proportion of bipolar neurons was the largest (∼50%), especially in layer II/III. Multipolar neurons were mainly distributed in the deep layers (Fig. 6(H)).

These results were consistent with the results of electrophysiology of VIP neurons from layers IV-VI in the barrel cortex. Indicated that the VIP neurons of different layers might participate in the formation of different neural circuits and perform different functions. However, there was no clear evidence supporting a correlation between the blood vessels and VIP neurons.

Our results demonstrated that the GMA-T method was suitable for embedding multiple fluorescent strategies labeled biological samples for precise whole-organ imaging and 3D reconstruction of multiple structures.

4. Discussion and conclusion

Fluorescence labeling has been widely used in the study of complex brain structures. With 3D imaging technology, we could observe axons and dendrites of neurons, blood vessels, and pathological structures labeled with multiple fluorescent dyes and proteins. It is helpful to simultaneously visualize multiple structures with clear spatial and temporal scales and location information.

Tissue embedding helps maintain the ultrastructure of the sample and produce ultrathin sections for optical imaging to perform high-resolution 3D reconstruction of large-scale tissues. High fluorescence intensity of multiple fluorescent dyes and proteins is essential for simultaneous acquisition of multidimensional structural information. Paraffin-embedding is a commonly used embedding technique. However, various chemical reagents were used in the paraffin-embedding, which reduced the fluorescent signals intensity [32–34]. In addition, high temperature used in the paraffin embedding further results in the quenching of fluorescent signals. Although, optimized paraffin embedding method was explored for GFP labeled sample, it has poor fluorescent and is difficult for visualization of fine structures, especially for large-volume samples [33]. With lower process temperature than paraffin-embedding, resin embedding was modified to acquire detailed structural information on multicolor fluorescent protein-labeled samples [16,17,26,27]. But has not been used for the high-resolution 3D reconstruction of fluorescent dyes labeling structures due to the lower preservation of dyes. In this study, we found that SBB was the key factor affecting fluorescent dyes in resin embedding methods for large volume biological tissues.

However, background inhibitors were necessary for fluorescence imaging of thick biological samples because both the tissue auto-fluorescence and the fluorescence at the bottom of the sample would interfere with the target fluorescent signals. TB as background inhibitor was better than SBB for preserving the intensity of fluorescent dyes. In addition, TB could not only reduce auto-fluorescence produced by lipid but also blacken the whole tissue and reduce the interference of deep fluorescence on the surface fluorescence signal. These effects of TB, combining with fast wide microscopes, helped obtain the fine structures of large tissues. Hence, the blue-black background fluorescent inhibitors used in resin embedding, such as TB, were indispensable for cutting and surface imaging of thick tissue blocks. We improved the signal-to-noise ratio of multiple fluorescence labeled samples by quantifying the effects of TB concentration on the intensity of DyLight dyes and auto-fluorescence of the tissue and defining the range of TB concentration. Then, we established a resin embedding method, GMA-T, which was suitable for large volume biological tissues labeled with multiple fluorescent dyes and proteins. We found that structural details in brain tissue labeled by various fluorescence (at different wavelengths) could be well preserved in GMA-T embedded samples.

GMA-T embedding method work well with various fluorescent dyes and proteins. It is also compatible with above 5 types of fluorescent signals. We verified the validity of GMA-T embedding method by embedding brain slices labeled with multiple fluorescent makers here: DyLight 488 and 594, Alexa 594 and 647, BSA-FITC, DANIR-8c/594, BFP, GFP and tdTomato. All fluorescent dyes and proteins showed good compatibility with GMA-T method.

Technically, GMA-T can work well with the new fluorescent dyes (such as CF and ATTO that performs higher intensity) labeled tissues. Considering fluorescence intensities vary with different types of dyes, the concentration of TB could be flexibly selected to achieve a good balance between fluorescence retention and background fluorescence inhibition. According to the experimental results in this paper, the recommended concentration is 0.7%-1%.

Furthermore, we embedded 3 whole brains labeled with two fluorescent strategies, tdTomato labeled VIP neurons and BSA-FITC specifically labeled the blood vessels. In combination with the whole-brain imaging system, the distribution and morphological characteristics of VIP neurons and blood vessels could be clearly visualized and quantified. These results showed that the GMA-T method performed well in the preservation of information of fine structures of large biological tissues.

Generally, in vivo fluorescent dye labeling had several potential advantages for monkey and human brain tissues. For these fluorescent dyes labeled monkey brains or human brain tissues, fine three-dimensional structures could be obtained by combining the GMA-T embedding method and continuous three-dimensional imaging systems.

Hence, the GMA-T method provided a reliable technology choice for the study in brainsmatics and has a wide application prospect in complex brain structure research.