2.1.

Animals

Pancreata from the female wild-type BALB/c mice (BioLASCO, Taipei, Taiwan) were used in insulin staining and tests of optical clearing. The tissues were also used in membrane and nuclear staining to examine the islet/pancreatic microstructure. Pancreata from the nestin–green fluorescent protein (GFP) transgenic mice were used in vessel painting to examine the islet microvasculature. In these transgenic animals, the GFP expression is under the control of the $5.8\text{-}\mathrm{kb}$ protomer and the $1.8\text{-}\mathrm{kb}$ second intron of the nestin gene, which encodes a type VI intermediate filament protein.²² Particularly, in the pancreas, this mouse line shows strong GFP expression in the exocrine acinar cells and, minimal, if any, GFP expression in the endocrine islet cells.²³ Overall, 15 and 7 wild-type BALB/c (36 islets) and nestin-GFP (32 islets) mice were used, respectively, to generate representative images. The care of the animals was consistent with Guidelines for Animal Experiments, National Tsing Hua University, Taiwan.

2.2.

Sample Preparation

Mouse pancreata were fixed by paraformaldehyde (4%) perfusion alone (used in BALB/c mice) or vessel painting of lipophilic dye DiD (4-chlorobenzene sulfonate salt, purchased from Invitrogen, Carlsbad, California, used in the nestin-GFP mice)^{18, 19} followed by paraformaldehyde perfusion prior to being harvested for examination. Once harvested, the tissues were fixed in paraformaldehyde (4%) for an additional $1\mathrm{h}$ . Mouse insulin staining was achieved by incubating the specimen with the primary rabbit-anti-mouse insulin antibody (1:50, Cell Signaling, Danvers, Massachusetts) and then revealing with Alexa Fluor 488-conjugated goat-anti-rabbit IgG (1:200, Invitrogen). Propidium iodide staining (PI staining, $25\mu \mathrm{g}∕\mathrm{ml}$ , Invitrogen) was performed at room temperature for $90\min$ to label the nuclei. DiD staining of membranes $(2\mu \mathrm{g}∕\mathrm{ml})$ was performed at room temperature overnight. Specimens from the DiD-perfused nestin-GFP mice (vessel painting) were stained with PI only. The labeled specimens were immersed in FocusClear solution (CelExplorer, Hsinchu, Taiwan) for optical clearing before being imaged via confocal microscopy.

2.3.

Imaging Settings

Confocal imaging of the optical-cleared specimens was performed with a Zeiss LSM 510 confocal microscope equipped with a $25\times$ objective LD Plan-Apochromat glycerine immersion lens or $40\times$ objective LD C-Apochromat water immersion lens. A $10\times$ objective Fluar was used to acquire gross images. The PI- and DiD- labeled samples were excited with helium/neon lasers at $543\mathrm{nm}$ and $633\mathrm{nm}$ , respectively. A bandpass $560\text{-}\text{to}615\text{-}\mathrm{nm}$ filter and a long-pass $650\text{-}\mathrm{nm}$ filter were used to collect signals from PI and DiD, respectively. A $488\text{-}\mathrm{nm}$ argon laser and a bandpass $505\text{-}\text{to}530\text{-}\mathrm{nm}$ filter were used for GFP excitation and fluorescence detection. In addition to fluorescence imaging, the Zeiss LSM 510 confocal microscope carries the function of transmission light imaging, which can be integrated with the confocal imaging when the detector of the transmitted light channel (ChD) is open. The imaging process is similar to viewing a thick, transparent tissue using a standard microscope with an additional function so that the focal path can be digitally adjusted in coordination with the confocal imaging. That is, while the detector of the confocal channel is collecting the signals from a single plane, the detector of the transmitted light channel is collecting the laser light traveling across the tissue.

Sample scanning was recorded with $1024\times 1024\text{pixels}$ of the $x∕y$ plane. The increment of the $z$ -axis optical section was $2.5\mu \mathrm{m}$ when $10\times$ objective lenses were used to acquire gross images. The increment was 1 or $1.5\mu \mathrm{m}$ when $25\times$ and $40\times$ objective lenses were used to acquire high-resolution images. The pixel intensity was set to be the mean of four scans, and the value ranged from 0 to 4095 ( $12\text{-bit}$ data) for the transmitted light imaging and from 0 to 255 ( $8\text{-bit}$ data) for the rest of the confocal imaging.

2.4.

Post-Recording Image Processing and Projection

The voxel-based confocal micrographs were processed with the use of the LSM 510 software (Version 3.2, Carl Zeiss) and the Amira 4.1.2 (Mercury, Chelmsford, Massachusetts) for projection and analysis. Typically, for a three-channel confocal microscopy (such as to collect signals from GFP, PI, and DiD), the size of the image data was $\sim 630\text{megabytes}$ ( $1024\times 1024\text{pixels}\times 200$ $\text{slices}\times 3$ channels of signals) for an imaging depth of $300\mu \mathrm{m}$ . The LSM image data were uploaded into Amira, which was operated under a Dell Precision T5400 workstation. We applied the Gaussian Filter function of Amira for noise reduction of the micrographs. In Videos 1 , 2 , and 5 the image stacks were displayed using the Ortho Slice function, and the video was made via the Movie Maker function with an increase in display time in association with the depth of the optical section. In Videos 2 to 4, 6 , and 7 , the Voltex module of Amira was used to project the image stacks. In addition, the Volume Editing function was used to subtract a cuboid(s) from the scanned volume to expose the interior domain of the specimen for visualization. In video preparation, the Demo Maker function was used to arrange the sequence of the image objects at different time intervals. The Camera Path function was used to adjust the projection angles and zoom-in and zoom-out movements of the 3-D images.

Video 1

Complete serial optical sections of the mouse pancreas shown in Figs. 2, 2, 2, 2, 2 $(\text{depth}=330\mu \mathrm{m})$ . The specimen was stained by DiD and propidium iodide to reveal the cellular membranes (green) and nuclei (red), respectively. Two islets were imaged in the scanned volume at depths of $\sim 180\mu \mathrm{m}$ and $\sim 270\mu \mathrm{m}$ (MPEG $4.9\mathrm{MB}$ ). (Color online only.) .

Video 2

High-resolution, stereo projection of the serial orthogonal views of the islet shown in Figs. 2, 2, 2. The specimen was stained by DiD and propidium iodide to reveal the cellular membranes (gray) and nuclei (red), respectively. Dimensions of the scanned volume: $175\mu \mathrm{m}$ $\left(x\right)\times 175\mu \mathrm{m}$ $\left(y\right)\times 122\mu \mathrm{m}$ ( $z$ , depth) (MPEG $4.9\mathrm{MB}$ ). (Color online only.) .

Video 4

3-D gross views of the mouse pancreatic vasculature. Lipophilic dye DiD perfusion was used to label blood vessels. The Volume Editing function of the Amira software was used to subtract signals from acini (green) and then vasculature (cyan) at the top corner of the imaged region to expose the interior domain for visualization. Dimensions of the scanned volume: $1300\mu \mathrm{m}$ $\left(x\right)\times 1300\mu \mathrm{m}$ $\left(y\right)\times 250\mu \mathrm{m}$ ( $z$ , depth). An $x∕y$ plane is shown at $z=125\mu \mathrm{m}$ to outline the tissue structure (MPEG $4.8\mathrm{MB}$ ). (Color online only.) .

Video 5

High-resolution, serial optical sections of the islet vasculature shown in Figs. 4, 4, 4, 4, 4 $(\text{depth}=225\mu \mathrm{m})$ . Fluorescent labeling of blood vessels (cyan) was done by cardiac perfusion of DiD, i.e., vessel painting. The pancreas from the nestin-GFP transgenic mouse was used in the imaging, which has strong nestin-GFP expression in the exocrine acinar cells (green). Propidium iodide staining was used to reveal the nuclei (red) (MPEG $5\mathrm{MB}$ ). (Color online only.) .

Video 6

Fly-through presentation of the islet vasculature shown in Fig. 4 using multiple projection angles and magnifications. The Camera Path function of the Amira software was used to adjust the viewing angles and the zoom-in and zoom-out movements. Dimensions of the scanned volume: $521\mu \mathrm{m}$ $\left(x\right)\times 521\mu \mathrm{m}$ $\left(y\right)\times 333\mu \mathrm{m}$ ( $z$ , depth). Two $x∕y$ planes are shown at $Z=135$ and $204\mu \mathrm{m}$ to outline the tissue structure. Two islets reside in the scanned volume (MPEG $4.9\mathrm{MB}$ ). .

Video 7

Fly-through presentation of the vasculature in the interior and exterior domains of an islet. The interior domain of the islet vasculature was assigned to yellow, using the Image Segmentation function of the Amira software and the ovoid islet surface as the segmentation boundary. Dimensions of the scanned volume: $369\mu \mathrm{m}$ $\left(x\right)\times 369\mu \mathrm{m}$ $\left(y\right)\times 207\mu \mathrm{m}$ ( $z$ , depth) (MPEG $5\mathrm{MB}$ ). (Color online only.) .

3.1.

Optical Clearing of the Mouse Pancreas

We applied an optical-clearing solution FocusClear (U.S. Patent No. 6472216)²⁴ to facilitate photon penetration across the mouse pancreatic tissue. The optical-clearing effect allowed the opaque specimen [Fig. 1, in saline] to become transparent when it was immersed in the FocusClear solution [Fig. 1]. In the transparent pancreas, we were able to visualize islets [Figure 1, insulin immunostaining] and the vascular structure [Figure 1, details described later] in situ. Traditionally, pancreata are sectioned into thin slices (e.g., $10\mu \mathrm{m}$ in thickness) to expose their interior domain for microscopic examination.¹ Here, the image area can be as large as $1000\mu \mathrm{m}$ $\left(x\right)\times 1000\mu \mathrm{m}$ $\left(y\right)\times 400\mu \mathrm{m}$ ( $z$ , depth). The thickness of the specimen can be in the $800\text{-}\mu \mathrm{m}$ range, given that separate visuals can be made from both sides of the specimen.

A significant feature of the FocusClear-mediated optical clearing is its reversibility. Figure 1 shows an islet residing in the transparent pancreas, which was acquired using the transmitted light channel of confocal microscopy. When we replaced FocusClear with saline [Figs. 1, 1, 1] and then repeated the optical-clearing immersion [Figs. 1, 1, 1], the optical-clearing effect was reproduced [Figs. 1, 1, 1, 1, 1, 1]. We measured the kinetics of this effect: Figure 1 shows a similar trend in all three tests, confirming the reversibility of the FocusClear-mediated optical clearing. Note that in sample preparation, paraformaldehyde was used to fix the specimen prior to the FocusClear immersion. Importantly, the size of the islet in the paraformaldehyde-fixed specimen remained the same, while the tissue optics changed during the test [Figs. 1, 1, 1, 1, 1, 1]—this is essential with the use of a new optical method for tissue examination. However, we would like to stress the finding by Goldstein that upon comparison of the freshly excised and the formalin-treated specimens, there was shrinkage after fixation.²⁵ Thus, there are still disparate observations in the size of tissue structures between in vivo and in vitro measurements.

3.2.

Penetrative Confocal Imaging of Islets of Langerhans in the Mouse Pancreas

Taking advantage of the transparent specimen, we next applied deep-tissue confocal microscopy to examine the fluorescent-dye-labeled structure. Figures 2, 2, 2, 2, 2 are examples of the confocal micrographs at different depths in the specimen, where the nuclei and membranes were revealed by propidium iodide (PI) and DiD (4-chlorobenzene sulfonate salt) staining, respectively. The complete serial optical sections, from the surface to $\text{depth}=330\mu \mathrm{m}$ , are shown in Video 1, which reveals two islets in the imaged region. Because the two islets are embedded in the exocrine acini, we used the Amira image software to digitally subtract two cuboids from the scanned volume to expose their locations [Figs. 2, 2, 2].

Fig. 2

Penetrative confocal imaging of islets of Langerhans in the mouse pancreas. (a) to (e) Confocal micrographs at different depths in the pancreatic tissue specimen. (a) and (b) The exocrine acinar tissue. (c) to (e) Islets of Langerhans in association with acinar cells, pancreatic duct (d), and blood vessels. Cells in the specimen were stained by DiD and PI to reveal the cellular membranes (green) and nuclei (red), respectively. Particularly in (e), the elongation of the endothelial cell nuclei along the vessel direction can be clearly identified. $\text{Bar}=50\mu \mathrm{m}$ . The complete serial optical sections, from the surface to $\text{depth}=330\mu \mathrm{m}$ , are shown in Video 1. (f) and (g) Stereo projections of the stack of confocal micrographs showing the two islets [(c) and (e)] embedded in the exocrine acini and linked with blood vessels. The DiD-stained cellular membranes are presented in gray scale for 3-D perspective views. In (f), a cuboid was digitally subtracted from the scanned volume to reveal the islet shown in (c). In (g), a second cuboid was subtracted from (f) to present both of the two islets shown in (c) and (e). (h) A clockwise, $90\text{-}\mathrm{deg}$ rotation of (g) was made to create a new viewing angle. Dimensions of the scanned volume: $290\mu \mathrm{m}$ $\left(x\right)\times 290\mu \mathrm{m}$ $\left(y\right)\times 330\mu \mathrm{m}$ ( $z$ , depth). (i) to (k) High-resolution, 3-D projections of an embedded islet in connection with its vascular network. In (i), the dotted circle indicates the location of the embedded islet. In (j), a cuboid was subtracted to expose the capillary (arrows). In (k), the fluorescent signals of the islet were digitally extracted from the original stack of confocal micrographs and then added into (j). Continuous orthogonal views of the islet and a $360\text{-}\mathrm{deg}$ panoramic presentation of (k) are shown in Videos 2 and 3, respectively. (Color online only.)

In a second example, Figs. 2, 2, 2, we used a higher magnification to examine the pancreatic islet structure. Video 2 uses stereo projection to display continuous orthogonal views of the imaged islet. Because there were no physical sections involved in sample preparation, we can digitally quantify the size of the islet by gathering the islet signals based on their locations in the micrograph and calculating the occupied voxels. For example, the size of the scanned volume shown in Figs. 2, 2, 2 is $175\times 175\times 122\mu {\mathrm{m}}^3$ , and the size of each voxel is $0.17\times 0.17\times 1.5\mu {\mathrm{m}}^3$ . By calculating the voxels occupied by the islet [Fig. 2], we estimate that the volume of the islet is $7.1\times {10}^5\mu {\mathrm{m}}^3$ in the specimen. A $360\text{-}\mathrm{deg}$ panoramic presentation of Fig. 2 is shown in Video 3. This approach allows an integral view and assessment of the islet structure, which cannot be done by the standard microtome-based 2-D analysis.

Video 3

$360\text{-}\mathrm{deg}$ panoramic presentation of Fig. 2. The voxels of the islet shown in Video 2 were digitally collected and merged with the signals shown in Fig. 2 (MPEG $5\mathrm{MB}$ ). .

3.3.

High-Resolution, 3-D Imaging of Islet Vasculature in the Mouse Pancreas

Vessel painting is a direct staining method that uses perfusion of lipophilic dye to label blood vessels in small experimental animals. Previously, this method has been used to visualize the vasculature of the retina, skin, lung, heart, and brain in mice, but with limited imaging depth.^{18, 19} Here, we tested the compatibility of vessel painting with our optical method to visualize the mouse islet/pancreatic vasculature in a 3-D fashion. Figures 3, 3, 3 show that the mouse tissues, including the pancreas, turned blue after perfusion with the lipophilic dye DiD. The labeled pancreatic tissue was then optically cleared by the FocusClear solution and underwent confocal microscopy for deep-tissue imaging. This approach led to a clear visualization of the vascular network, including the capillaries in the islet. Figures 3, 3, 3 are the gross views of the labeled pancreatic tissue, where there are three embedded islets, residing at $50\text{to}200\mu \mathrm{m}$ under the tissue surface. In Fig. 3, we used the Amira software to digitally subtract a cuboid of signals from acini to reveal the vasculature inside. In Fig. 3, we subtracted the vasculature signals on top of the cuboid (from $0\text{to}80\mu \mathrm{m}$ in depth) to reveal the locations of the three islets. Stereo projections of Figs. 3, 3, 3 are shown in Video 4.

Fig. 3

(a) to (c) Use of vessel painting, i.e., lipophilic dye DiD perfusion, to label blood vessels. (a) Prior to vessel painting and (b) post vessel painting of the nestin-GFP transgenic mouse. Arrows in (b) indicate the stained blue tissues/organs after the DiD perfusion. (c) An enlarged view of the harvested mouse pancreas (together with the spleen) after vessel painting. (d) to (f) 3-D gross views of the labeled pancreatic tissue. Signals of GFP expression (green) and DiD-labeled vasculature (cyan) are combined in (d). In (e), a cuboid $(\text{depth}=125\mu \mathrm{m})$ of GFP signals is subtracted from (d) to reveal the vasculature. In (f), a cuboid $(\text{depth}=80\mu \mathrm{m})$ of DiD signals is subtracted from (e) to reveal the locations of the three embedded islets, indicated by the white dotted circles. Dimensions of the scanned volume: $1300\mu \mathrm{m}$ $\left(x\right)\times 1300\mu \mathrm{m}$ $\left(y\right)\times 250\mu \mathrm{m}$ ( $z$ , depth). Video 4 provides additional projection angles and zoom-in and zoom-out movements of panels (d) to (f). In the 3-D projections, we added an $x∕y$ plane at $z=125\mu \mathrm{m}$ to outline the tissue structure. (Color online only.)

We next zoomed in to visualize the islet vasculature with high resolution. The optical sections at different depths of a typical islet are shown in Figs. 4, 4, 4, 4, 4 . Images from the transmitted light and fluorescence channels of confocal microscopy are presented on the left and right sides of the figure, respectively. Video 5 displays the serial optical sections of the scanned region. Specifically, on the left side of Fig. 4, large blood vessels can be seen using transmitted light imaging. On the right side, these vessels are also revealed by fluorescence imaging, confirming the success of vessel painting. Notably, the microvessels with a diameter of $\sim 5\mu \mathrm{m}$ are clearly visualized in the islet and acini.

Fig. 4

High-resolution, 3-D images of the islet vasculature in the mouse pancreas. (a) to (e) Micrographs of transmitted light (left) and fluorescence (right) imaging at different depths in the pancreatic tissue specimen. The vascular network (cyan) was labeled with lipophilic DiD by cardiac perfusion. PI staining was used to reveal the nuclei (red). We used the pancreas from the nestin-GFP transgenic mice (GFP expression driven by the nestin promoter) to highlight the exocrine acini (green) (Refs. 21, 22) This mouse line shows minimal, if any, GFP expression in the islet endocrine cells. In (e), large blood vessels can be seen in both the transmitted light (dotted lines indicate the approximate boundaries) and the fluorescence micrographs. The complete fluorescent serial optical sections of the imaged islet are shown in Video 5. (f) to (i) Separate and merged stereo projections of the blood vessels and the islet/pancreatic structure. (h) and (i) are projections from the top and bottom halves of the imaged region. A “landmark” pancreatic duct can be seen in (g), at the right side of the islet. The same duct can be seen in (h) and (i), yet some areas were blocked by the projection of blood vessels. Dimensions of the scanned volume: $369\mu \mathrm{m}$ $\left(x\right)\times 369\mu \mathrm{m}$ $\left(y\right)\times 225\mu \mathrm{m}$ ( $z$ , depth). (j) and (k) Representative 3-D projections of the islet vasculature in the mouse pancreas. Video 6 provides additional projection angles and zoom-in and zoom-out movements of panel (k). (Color online only.)

Figures 4 and 4 are stereo projections of the blood vessels and an orthogonal view of the islet shown in Figs. 4, 4, 4, 4, 4. The 3-D top view and bottom view of the islet vasculature, shown in Figs. 4 and 4, are created by using the Amira software to merge the signals shown in Figs. 4 and 4 and rotate the viewing angle to project the top and bottom halves of the imaged region. Figures 4 and 4 are two additional results of the high-resolution, 3-D vasculature imaging of the mouse pancreatic islets. A fly-through presentation of Fig. 4 is shown in Video 6. Furthermore, in Video 7, we applied the Image Segmentation function of the Amira software to highlight the vasculature in the interior and exterior domains of the islet. These results explore the feasibility of an in situ 3-D image comparison between the endocrine and exocrine parts of the pancreas, which could be used to examine the remodeling of microvessels during the progression of autoimmune diabetes.