2.1.

Natural Biomaterial Scaffolds

The collagen scaffolds were purchased from BD Biosciences (Bedford, Massachusetts); Cat. No. 354613, BD 3-D collagen composite scaffold). The scaffold consists of type I and III collagens derived from bovine hide with a cylinder shape and a dimension of $\sim 5\mathrm{mm}$ $\left(\text{diameter}\right)\times 3\mathrm{mm}$ (height). The chitosan scaffolds were prepared by following the protocol published previously.²⁶ Briefly, chitosan powders with a molecular weight at $\sim 400\mathrm{K}$ (Koyo Chemical Co., Ltd., Japan) were dissolved in 1% acetic acid to make a 1% chitosan solution. The solution was then frozen in a $60\text{-}\mathrm{mm}$ Petri dish and lyophilized to form a sheet of porous scaffold. Scaffolds with a dimension of $\sim 5.5\mathrm{mm}$ $\left(\text{diameter}\right)\times 1.5\mathrm{mm}$ (height) were cut from the sheet and transferred to a 48-well plate. The microporous membrane of cellulose scaffold was a gift from Sartorius AG (Göttingen, Germany).²⁷ The membrane was approximately $250\mu \mathrm{m}$ in thickness and was cut into squares of $1\mathrm{cm}\times 1\mathrm{cm}$ prior to imaging. In conventional bright-light microscopy (Fig. 1 ), scaffolds were immersed in phosphate-buffered saline (PBS, Sigma, St. Louis, Missouri) or FocusClear solution (CelExplorer, Hsinchu, Taiwan) overnight prior to being imaged by a Nikon TS100 inverted microscope (Tokyo, Japan) to compare their light penetration.

Fig. 1

Optical clearing of collagen, chitosan, and cellulose scaffolds. (a) The gross image of the collagen scaffold with a diameter $\sim 5\mathrm{mm}$ and height $\sim 3\mathrm{mm}$ (left: side view; right: top view). The scaffold was immersed in saline (b) or FocusClear solution (c) prior to being imaged by a standard inverted microscope (Nikon TS100) from the top of the scaffold aiming at the bottom-left corner. Scale $\text{bar}=1\mathrm{mm}$ . (d) The gross image of the chitosan scaffold with a diameter $\sim 5.5\mathrm{mm}$ and height $\sim 1.5\mathrm{mm}$ (left: side view; right: top view). The scaffold was immersed in saline (e) or FocusClear solution (f) prior to imaging. Scale $\text{bar}=1\mathrm{mm}$ . (g) The gross image of the microporous cellulose membrane (cut into $1\mathrm{cm}\times 1\mathrm{cm}$ with a thickness of $250\mu \mathrm{m}$ ). The membrane was immersed in saline (h) or FocusClear solution (i), which reveals the polyester reinforcement fibers in the membrane after the cellulose matrix became transparent. Scale $\text{bar}=100\mu \mathrm{m}$ . In (b), (c), (e), and (f), the focal plane was on the background, the grid surface of a hemacytometer. In (h) and (i), the focal plane was on the cellulose membrane.

2.2.

Cell Culture and Preparation of the Fibroblast/Collagen and Osteoblast/Chitosan Constructs

Humane HT-1080 fibroblasts (ATCC, Manassas, Virginia) were stably transfected with a vector carrying an expression cassette of enhanced green fluorescent protein (EGFP) as previously described.²⁸ The same protocol was used to generate a stable line of human Saos-2 osteoblasts (ATCC) carrying an expression cassette of Ds-Red (Clontech, Mountain View, California). Both stable lines of fibroblasts and osteoblasts were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, $100\mathrm{U}∕\mathrm{mL}$ penicillin, $100\mu \mathrm{g}∕\mathrm{mL}$ streptomycin, and $0.25\mu \mathrm{g}∕\mathrm{mL}$ puromycin (selection pressure) at $37°\mathrm{C}$ in a humidified atmosphere of 5% $\mathrm{C}{\mathrm{O}}_2$ . The EGFP-labeled HT-1080 fibroblasts $(1\times {10}^6)$ were seeded onto the collagen scaffold, and the Ds-Red-labeled Saos-2 osteoblasts $(1.5\times {10}^6)$ were seeded onto the chitosan scaffold. Seeding was carried out in the wells of a 96-well plate. One day after the seeding, both the fibroblast/collagen and the osteoblast/chitosan constructs were transferred to the wells of a 24-well plate for a $7\text{-}\text{day}$ cell culture to allow cellular proliferation and infiltration into the scaffold.

2.3.

Quantification of the Optical-Clearing Effect

The heat-induced gel formation of bovine serum albumin (BSA; Sigma)^{29, 30} in the wells of a 96-well plate was used to quantify the optical clearing effect of different reagents. $50\mu \mathrm{l}$ of BSA in PBS $(350\mathrm{mg}∕\mathrm{ml})$ was transferred to one well of a 96-well plate. Afterward, the plate was carefully sealed and placed in a water bath at $60°\mathrm{C}$ for $2\mathrm{h}$ to induce the formation of a layer of BSA protein gel at the bottom of each well. $100\mu \mathrm{l}$ of different reagents including FocusClear, dimethyl sulfoxide (DMSO, Sigma), glycerol (Sigma), and PBS (negative control) were placed on top of the BSA gel for optical clearing for $2\mathrm{h}$ . Afterward, the percent transmittance of light at 351, 364, 488, 543, and $633\mathrm{nm}$ was measured via a microplate reader (SpectraMax M2e, Molecular Devices, Sunnyvale, California). Wells with $150\mu \mathrm{l}$ of water were used as the positive control. A schematic diagram of the setup is shown in Fig. 2a .

Fig. 2

(a) A schematic diagram of using a layer of light-scattering BSA aggregates and spectrophotometry to quantify the optical-clearing effect. (b) A gross image of a part of a 96-well plate with BSA layers immersed in saline, FocusClear, DMSO, or glycerol. Positive control: wells in water (without the BSA layer; far left). (c) Quantification of the optical-clearing effect. The percent transmittance of (b) was measured at 351, 364, 488, 543, and $633\mathrm{nm}$ via a microplate reader. Results are presented as $\text{mean}\pm \text{standard}$ deviation $(n=6)$ . (d) The effect of optical clearing on the increase in fluorescence emission from the FITC- or EGFP-doped BSA layer. Confocal microscopy was used to perform fluorescence excitation/emission $200\text{-}\mu \mathrm{m}$ deep into the BSA layer. A laser line at $488\mathrm{nm}$ was used for excitation. A long-pass $505\text{-}\mathrm{nm}$ filter was used to collect the fluorescence emission. Results are presented as $\text{mean}\pm \text{standard}$ deviation $(n=6)$ .

We used confocal microscopy (LSM 510, Carl Zeiss, Jena, Germany) to evaluate the impact of optical clearing on fluorescence detection. Prior to initiating the heat-induced gel formation of BSA, fluorescein isothiocyanate (FITC, $10\mu \mathrm{g}∕\mathrm{ml}$ , Sigma), or EGFP ( $10\mu \mathrm{g}∕\mathrm{ml}$ , BioVision, Mountain View, California) was dissolved in the BSA/PBS solution. After gel formation (in a BD Falcon 96-well imaging plate) and the subsequent optical clearing by FocusClear, DMSO, glycerol, DAKO aqueous mounting medium (Carpinteria, California), or PBS (negative control), confocal microscopy employing a $488\text{-}\mathrm{nm}$ argon laser and a Fluar $20\times$ objective (Carl Zeiss, $\mathrm{NA}=0.75$ ) was used to excite FITC and EGFP $200\mu \mathrm{m}$ under the gel surface. The fluorescence emission was collected using a long-pass $505\text{-}\mathrm{nm}$ filter. The “line” scanning pattern was used and the scanning $\text{speed}=8$ . The pixel intensity, ranging from 0 to 255, was set to be the mean value of four scans. The detection gain was carefully adjusted to avoid signal saturation. The Histogram function of the LSM 510 software (version 3.2, Carl Zeiss) was used to obtain the frequency distribution of pixels at different intensities, which subsequently was used to calculate the total intensity from the scanned area of $1024\times 1024\text{pixels}$ $(0.41\times 0.41\mu {\mathrm{m}}^2)$ .

2.4.

Sample Preparation for Imaging

The blank collagen and chitosan scaffolds were directly immersed in the FocusClear solution overnight prior to being imaged via confocal microscopy. The autofluorescence of collagen and chitosan induced by the $488\text{-}\mathrm{nm}$ laser was used to examine their scaffold structures.^{31, 32} The EGFP-fibroblast/collagen and Ds-Red-osteoblast/chitosan constructs were fixed with 4% paraformaldehyde for $40\min$ and permeabilized with 2% Triton X-100 for $2\mathrm{h}$ in PBS at room temperature. Prior to FocusClear immersion, samples of the EGFP-fibroblast/collagen constructs were stained with propidium iodide (PI, $62.5\mu \mathrm{g}∕\mathrm{ml}$ , Molecular Probes, Eugene, Oregon) overnight to label the nuclei. Samples of Ds-Red-osteoblast/chitosan constructs were directly immersed in the FocusClear solution without additional labeling. The microporous cellulose matrix, on the other hand, was stained with Calcofluor White M2R (Sigma) in PBS $(5\mathrm{mg}∕\mathrm{ml})$ for $20\min$ ,³³ rinsed in blank PBS, and followed by FocusClear immersion prior to confocal/two-photon microscopy.

2.5.

Imaging Settings

Confocal imaging of the FocusClear-immersed scaffolds/constructs was performed with a Zeiss LSM 510 confocal microscope (Carl Zeiss). For imaging the blank collagen and chitosan scaffolds, a $488\text{-}\mathrm{nm}$ argon laser was used to induce the autofluorescence; a long-pass $505\text{-}\mathrm{nm}$ filter was used to collect signals. The PI-stained EGFP-fibroblast/collagen constructs were excited at $543\mathrm{nm}$ (PI excitation, a helium/neon laser) and $488\mathrm{nm}$ (EGFP excitation). A long-pass $560\text{-}\mathrm{nm}$ filter and a band-pass $505\text{-}\text{to}530\text{-}\mathrm{nm}$ filter were used to collect signals from PI and EGFP, respectively. The Ds-Red-osteoblast/chitosan construct was excited at $543\mathrm{nm}$ (Ds-Red excitation) and $488\mathrm{nm}$ (to induce autofluorescence from chitosan). Band-pass filters of $560\text{to}615\mathrm{nm}$ and $505\text{to}530\mathrm{nm}$ were used to collect signals from Ds-Red and chitosan, respectively.

The Calcofluor-labeled cellulose matrix was excited at $364\mathrm{nm}$ with a UV laser, and the fluorescence emission was detected by a $385\text{-}\mathrm{nm}$ long-pass filter in confocal microscopy. In two-photon microscopy (performed by a Zeiss LSM510 META microscope), the labeled cellulose membrane was excited by an infrared laser at $820\mathrm{nm}$ (Chameleon-XR ultra-fast Ti-Sapphire laser, Coherent, Inc., Santa Clara, California), and the fluorescence emission was detected by a band-pass $390\text{-}\text{to}465\text{-}\mathrm{nm}$ filter.

A long-distance C-Apochromat $40\times$ water immersion objective (Carl Zeiss, $\mathrm{NA}=1.1$ ) was used to acquire the images in Figs. 3, 4a, 4b . A C-Apochromat $40\times$ water immersion objective (Carl Zeiss, $\mathrm{NA}=1.2$ ) was used to acquire the images in Figs. 5 and 6 . A Fluar $10\times$ objective (Carl Zeiss, $\mathrm{NA}=0.5$ ) was used to acquire the images in Figs. 4c, 4d, 7 . The voxel size $(x\mu \mathrm{m}\times y\mu \mathrm{m}\times z\mu \mathrm{m})$ was set at $(0.29\times 0.29\times 2.5)$ , $(0.41\times 0.41\times 3.5)$ , $(0.29\times 0.29\times 2)$ , $(0.82\times 0.82\times 4)$ , $(0.07\times 0.07\times 1)$ , $(0.06\times 0.06\times 1)$ , and $(0.06\times 0.06\times 0.5)$ for Figs. 3, 4a, 4b, 4c, 4d, 5a, 5b, 5c, 5d, 5e, 5f, 5g, 5h, 6, 7, respectively. Each confocal/two-photon micrograph consists of $1024\times 1024\text{pixels}$ at the $x∕y$ plane. The number of optical slices along the $z$ axis was at 101, 101, 151, 89, 41, 81, and 85 for Figs. 3, 4a, 4b, 4c, 4d, 5a, 5b, 5c, 5d, 5e, 5f, 5g, 5h, 6, 7, respectively. The “line” scanning pattern was used in each confocal/two-photon microscopy and the scanning $\text{speed}=8$ . The pixel intensity, ranging from 0 to 255, was set to be the mean value of four scans for confocal microscopy and two scans for two-photon microscopy.

Fig. 3

Three-dimensional imaging of the collagen scaffold. Stereo projection of the scaffold image stack creates a 3-D visualization of the imaged region. (a): top view; (b): side view. A $360\text{-}\mathrm{deg}$ panoramic presentation of the scanned region is shown in Videos 1. Dimensions of the scanned volume: $290\mu \mathrm{m}$ $\left(x\right)\times 290\mu \mathrm{m}$ $\left(y\right)\times 250\mu \mathrm{m}$ ( $z$ , depth).

Fig. 4

Three-dimensional imaging of the chitosan scaffold and osteoblast/chitosan construct. (a) and (b) Stereo projection of the scaffold image stack creates a 3-D visualization of the imaged region. (a) top view; (b) side view. Dimensions of the scanned volume: $290\mu \mathrm{m}$ $\left(x\right)\times 290\mu \mathrm{m}$ $\left(y\right)\times 300\mu \mathrm{m}$ ( $z$ , depth). (c) and (d) Stereo projection of the osteoblast/chitosan image stack. Human Saos-2 osteoblasts were labeled with Ds-Red. (c): top view; (d): side view. A $360\text{-}\mathrm{deg}$ panoramic presentation of (d) is shown in Videos 3. Dimensions of the scanned volume: $842\mu \mathrm{m}$ $\left(x\right)\times 842\mu \mathrm{m}$ $\left(y\right)\times 352\mu \mathrm{m}$ ( $z$ , depth).

Fig. 5

Optical sections of the confocal and two-photon microscopy of the microporous cellulose membrane with polyester reinforcement fibers. (a) to (d) Confocal micrographs of the Calcofluor-stained membrane at $\text{depth}=0$ (the surface), 10, 20, and $30\mu \mathrm{m}$ , respectively. $\text{Bar}=10\mu \mathrm{m}$ . (e) to (h) Two-photon micrographs of the membrane at $\text{depth}=20$ , 50, 70, and $80\mu \mathrm{m}$ , respectively. $\text{Bar}=10\mu \mathrm{m}$ . (i) An enlarged micrograph of the box shown in (h). Red, white, and yellow bars indicate the pores at the $5\text{-}\mu \mathrm{m}$ , $2\text{-}\mu \mathrm{m}$ , and $1\text{-}\mu \mathrm{m}$ range, respectively. (Color online only.)

Fig. 6

The exterior and interior details of the scanned 3-D region of the cellulose membrane. (a) Stereo projection of a set of two-photon micrograph stacks showing the exterior of the scanned region with a $\text{depth}=42\mu \mathrm{m}$ and an $x∕y$ plane area of $64\mu \mathrm{m}\times 64\mu \mathrm{m}$ . (b) Orthogonal section views of the image stack to reveal the interior configuration of the imaged membrane segment. The intersection of the three cross sections is indicated by $O$ .

Fig. 7

Penetrative imaging of the fibroblast/collagen construct. (a) to (i) Confocal micrographs at different depths of the fibroblast/collagen construct. The EGFP-labeled human HT-1080 fibroblasts grown in the collagen matrix were stained by propidium iodide (PI) to reveal the nuclei. The PI staining also served as a counterstain to visualize collagen matrices. The serial optical sections from $\text{depth}=0$ to $350\mu \mathrm{m}$ are shown in Videos 4. Scale $\text{bar}=100\mu \mathrm{m}$ . (j) and (k) Projection of the PI-stained nuclei and collagen matrices using the image stack shown in Videos 4 after excluding the EGFP signal. (j): top view; (k): side view. In (k), a cuboid was subtracted from the scanned volume to expose the interior domain of the construct. A $360\text{-}\mathrm{deg}$ panoramic presentation of (k) is shown in Videos 2. The arrow indicates the lack of image resolution on the $x∕z$ plane.

Video 1

A $360\text{-}\text{degree}$ panoramic presentation of the collagen scaffold (MPEG, $4.5\mathrm{MB}$ ).

Video 2

A $360\text{-}\text{degree}$ panoramic presentation of the volume-edited fibroblast/collagen construct shown in Figure 4k (MPEG, $4\mathrm{MB}$ ).

Video 3

A $360\text{-}\text{degree}$ panoramic presentation of the osteoblast/chitosan construct (MPEG, $4.9\mathrm{MB}$ ).

Video 4

Serial confocal micrographs from the surface to $\text{depth}=350\mu \mathrm{m}$ of the fibroblast/collagen construct (MPEG, $4.8\mathrm{MB}$ ).

2.6.

Postrecording Image Processing and Analysis

The voxel-based confocal micrographs were processed using the LSM 510 software (Carl Zeiss) and Amira 4.1.2 (Mercury, Chelmsford, Massachusetts) for projection and analysis. In Figs. 3, 4a, 4b, 4c, 4d, 6a, 7j, 7k, and Videos 1, 2, and 3, the Voltex module of Amira was used to project the image stacks and the Trackball function was used to adjust the projection angle. In Figs. 5a, 5b, 5c, 5d, 5e, 5f, 5g, 5h, 7a, 7b, 7c, 7d, 7e, 7f, 7g, 7h, 7i and Videos 4 and 5, the image stacks were displayed using the Ortho Slice function of Amira. The video was made via the Movie Maker function of Amira with the increase in display time in association with the depth of the optical section from the surface (beginning) to the interior domain of the construct (end). In Fig. 7k and Videos 2, the Volume Editing function of Amira was used to subtract a cuboid from the scanned volume to expose the interior domain of the sample. In Fig. 6b, the acquired image stack was displayed using the Orthogonal Section View function of the LSM 510 software to present the interior details of the scanned region.

Video 5

Serial confocal micrographs from the surface to $\text{depth}=40\mu \mathrm{m}$ of the microporous cellulose matrix (MPEG, $4.4\mathrm{MB}$ ).

3.1.

Optical Clearing of Collagen, Chitosan, and Cellulose Scaffolds

Collagen, chitosan, and cellulose are among the most common natural biomaterials used to fabricate tissue-engineering scaffolds for promoting tissue repair or regeneration. Major advantages of these three biomaterials are their abundance, their proven ability to support the growth/infiltration of a variety of cell types and tissues, and their relatively low immunogenicity. However, scaffolds of collagen, chitosan, and cellulose are intrinsically opaque [Figs. 1a, 1b, 1d, 1e, 1g, 1h]. This imposes a barrier to visualize the scaffold architecture in depth and avoids using optical microscopy to in situ characterize cellular behavior such as proliferation or differentiation in a 3-D environment. Traditionally, thick collagen, chitosan, and cellulose scaffolds as well as other biological/nonbiological samples are sectioned into thin slices (typically less than $10\mu \mathrm{m}$ per slice) to expose their interior domain for microscopic examination. However, the spatial information of the scaffold and tissue structure is often interrupted and/or lost during the sectioning process.

Previously, we have applied an optical-clearing solution, FocusClear, to increase the transparency of insect brains for mapping of neural circuits via confocal microscopy.^{22, 23, 24, 25} We found that a similar clearing effect was observed when the FocusClear solution was applied to the collagen, chitosan, and cellulose scaffolds [Figs. 1c, 1f, 1i]; the thicknesses of the collagen, chitosan, and cellulose matrices were $\sim 3\mathrm{mm}$ , $1.5\mathrm{mm}$ , and $250\mu \mathrm{m}$ , respectively. In particular, Fig. 1i shows that when the cellulose membrane was immersed in the FocusClear solution, the cellulose matrix became transparent, leading to a clear visualization of the polyester reinforcement fibers in the membrane.

3.2.

Quantification of the Optical-Clearing Effect and Its Influence on Fluorescence Detection

Tissues such as human and animal skin have been used to determine the optical-clearing effect of chemical reagents.^{9, 10, 11, 12} However, biological tissues are heterogeneous; they cannot be consistently labeled with fluorescent probes to study the influence of optical clearing on fluorescent signal detection. Alternatively, to quantify the optical-clearing effect, we used aggregates of bovine serum albumin (BSA, heat-induced gel formation,^{29, 30} $\sim \mathrm{1,560}\mu \mathrm{m}$ thick in this experiment) as a light-scattering medium and applied spectrophotometry to measure the transmittance of light across the protein layer [Fig. 2a]. Using this method, we compared the efficiency of optical clearing among different immersion solutions. Figure 2b shows apart of a 96-well plate with the BSA protein layers immersed in saline (control), FocusClear, and the common optical-clearing reagents DMSO and glycerol. Significantly, after $4\text{-}\mathrm{h}$ incubation, one can readily observe that wells with FocusClear were more transparent than those with DMSO or glycerol.

We quantify the optical-clearing effect by measuring the percent transmittance of light across the scattering protein layer. Five wavelengths, 351, 364, 488, 543, and $633\mathrm{nm}$ (common wavelengths of laser lines for confocal microscopy), were tested. Specifically, Fig. 2c shows that at $488\mathrm{nm}$ , the FocusClear immersion promoted percent transmittance to 48.18% from 1.04% of the saline immersion (control), or by 46.3-fold; a similar increase in transmittance was also observed at 543 and $633\mathrm{nm}$ . In addition, FocusClear was significantly better than DMSO and glycerol in achieving higher percent transmittance at 488, 543, and $633\mathrm{nm}$ in the visible light spectrum.

On the other hand, Fig. 2c also indicates that ultraviolet (UV) light at 351 and $364\mathrm{nm}$ suffered low-percent transmittance regardless of the reagent for immersion. It has been known that UV excitation has limited transmissivity in biological tissues due to Rayleigh scattering.³⁴ Our result indicates that UV radiation at 351 and $364\mathrm{nm}$ remained strongly scattered by the BSA protein layer despite the FocusClear, DMSO, or glycerol immersion.

To evaluate the impact of optical clearing on fluorescence detection, we next used the $488\text{-}\mathrm{nm}$ laser to excite fluorescent probes FITC and EGFP (they were dissolved in the light-scattering BSA medium) and measured the increase in fluorescence emission $200\mu \mathrm{m}$ into the BSA gel surface after the optical-clearing treatment. Figure 2d shows that the FocusClear immersion exhibited 17.2-fold and 6.8-fold increase in fluorescence emission from FITC and EGFP, respectively, in comparison to the saline immersion. Although glycerol and the DAKO aqueous mounting medium also improved fluorescence excitation/emission from the labeled BSA layer; the increase was less significant as compared with that of the FocusClear-mediated optical clearing.

3.3.

Three-Dimensional Imaging of the Collagen Scaffold

Taking advantage of the transparent collagen scaffold in FocusClear [Fig. 1c], we applied confocal microscopy to generate sharp 2-D fluorescence micrographs within the scaffold. Image stacks ranging from $\text{depth}=0\text{to}250\mu \mathrm{m}$ (101 optical slices, $2.5\mu \mathrm{m}$ per increment) were recorded. These digital, voxel-based confocal micrographs were processed using the 3-D Voltex projection module of the Amira software to generate a $360\text{-}\mathrm{deg}$ panorama of the scanned region (Fig. 3 and Videos 1). It is noteworthy that these confocal micrographs were acquired via detection of collagen’s autofluorescence induced by the $488\text{-}\mathrm{nm}$ argon laser.³¹ The result highlights the effectiveness of the FocusClear-mediated optical clearing, which minimized light scattering and allowed confocal microscopy for detection of the low-intensity autofluorescence from the collagen scaffold.

3.4.

Penetrative Imaging of the Fibroblast/Collagen Construct

We integrated optical clearing and confocal microscopy to visualize cell proliferation/infiltration in the collagen scaffold. Figures 7a, 7b, 7c, 7d, 7e, 7f, 7g, 7h, 7i are examples of confocal micrographs at different depths of the fibroblast/collagen scaffold $7\text{days}$ after cell seeding. These images show the supportive role of collagen fibers for adhesion and proliferation of fibroblasts in the construct. A video file of serial optical sections from the construct surface to $\text{depth}=350\mu \mathrm{m}$ is shown in Videos 4. Compared with the microtome-based image acquisition of a thick tissue, optical sectioning microscopy is noninvasive and capable of providing a continuous flow of image information to examine in situ a region of the tissue construct.

Figures 7j and 7k are projections of confocal micrographs that reveal the propidium iodide (PI)–stained nuclei and collagen fibers. Specifically, to expose the interior domain, we used the volume-editing function of the Amira software to subtract a cuboid from the scanned region [Fig. 7k]. Videos 2 is a $360\text{-}\mathrm{deg}$ panoramic presentation of the volume-edited construct. Noticeably, the image resolution of the $x∕z$ plane in Fig. 7k (indicated by the arrow) and the $x∕z$ and $y∕z$ planes in Videos 2 is lower than that of the $x∕y$ plane. The lack of $z$ axis resolution is an inherent limitation of the current optical and mechanical designs of confocal microscopy.

3.5.

Three-Dimensional Imaging of the Chitosan Scaffold

In addition to the protein scaffold formed by collagen matrices, we also applied confocal microscopy to examine the transparent chitosan scaffold [Fig. 1f] Figures 4a and 4b are 3-D projections of the confocal image stack showing the scanned chitosan scaffold from $\text{depth}=0$ (surface) to $300\mu \mathrm{m}$ . The image stack was obtained by detecting the autofluorescence induced by the $488\text{-}\mathrm{nm}$ argon laser,³² similar to what we had used in imaging the collagen scaffold. Significantly, for both the collagen and the chitosan scaffolds, their low-intensity autofluorescence was sufficient to allow in-depth structure visualization- [Figs. 4a and 4b], underlining the sensitivity of confocal microscopy when integrated with the FocusClear-mediated optical clearing.

3.6.

Penetrative Imaging of the Osteoblast/Chitosan Construct

Chitosan-based scaffolds have been widely studied in bone tissue engineering for in vitro and in vivo bone generation. The hydrophilic chitosan scaffolds allow the initial attachment and spreading of osteoblasts,² serving as a physical support to guide the growth of new bone tissues. To image the osteoblast/chitosan construct, we applied the penetrative confocal microscopy to visualize the adhesion and infiltration of human Saos-2 osteoblasts after they were seeded on the scaffold for $7\text{days}$ . Figures 4c and 4d and Videos 3 are projections of the confocal image stack showing that the osteoblasts attached to the chitosan scaffold and followed the scaffold structure in spreading. Unlike the conventional 2-D image analysis showing the infiltration only from a specific cut plane, Videos 3 provides a $360\text{-}\mathrm{deg}$ panoramic presentation of the scanned region. Specifically, we observed a few osteoblasts infiltrated into the scaffold at depth $\sim 250\mu \mathrm{m}$ $7\text{days}$ post-seeding.

3.7.

Confocal Microscopy of the Internal Matrix Structure of Microporous Cellulose Membranes

We next applied the penetrative imaging method to characterize cellulose scaffolds in the format of microporous membranes. A few studies have used confocal microscopy to examine the cellulose membrane structure.^{35, 36} Unfortunately, due to lack of dye labeling (cellulose has a much lower autofluorescence background than collagen or chitosan) and optical clearing of the membrane, the acquired micrographs have poor resolution and cannot clearly present the microporous structure of the cellulose matrix.

We combined Calcofluor (a cellulose-binding UV dye)³³ staining and FocusClear-mediated optical clearing [Fig. 1i] to improve the image resolution of confocal microscopy of cellulose matrices. Figures 5a, 5b, 5c, 5d are examples of confocal micrographs revealing the detailed cellulose membrane structure at the $\text{depth}=0$ (the surface), 10, 20, and $30\mu \mathrm{m}$ , respectively. The micrograph stack from $z$ $\text{axis}=0\text{to}40\mu \mathrm{m}$ are presented in Videos 5. It is noteworthy that the image resolution started to decline when the depth was close to $40\mu \mathrm{m}$ . The limitation is most likely caused by poor transmission of UV laser through the cellulose sample, thus leading to insufficient Calcofluor excitation.

3.8.

Two-Photon Optical Imaging Improves the Penetration Depth of Imagine the Cellulose Matrix

We next sought to improve the penetration depth using two-photon optical imaging with the long-wavelength $820\text{-}\mathrm{nm}$ Ti-Sapphire infrared laser to excite Calcofluor. In two-photon microscopy, excitation of fluorochrome is achieved by employing high photon density and flux to induce the absorption of two (or multiple) low-energy, infrared photons simultaneously, thus raising the fluorochrome into the excited state for the subsequent fluorescence emission.³⁷ Infrared radiations are scattered to a lesser degree than those of shorter wavelengths, therefore allowing deeper sample penetration, which is a benefit to high-resolution imaging.

Figures 5e, 5f, 5g, 5h show the detailed cellulose membrane structure at the depth $\left(z\text{axis}\right)=20\mu \mathrm{m}$ , $50\mu \mathrm{m}$ , $70\mu \mathrm{m}$ , and $80\mu \mathrm{m}$ , respectively. Significantly, use of the infrared laser-coupled two-photon microscopy increased the penetration depth to $80\mu \mathrm{m}$ from $40\mu \mathrm{m}$ of the UV laser-based confocal microscopy, or by 100%. Specifically, Fig. 7i shows that pores at the $1\text{-}\mu \mathrm{m}$ range can be identified at $z$ $\text{axis}=80\mu \mathrm{m}$ .

To present the exterior and interior details of the scanned 3-D region, the two-photon optical micrographs were further processed using the 3-D Voltex projection module offered by the Amira software [Fig. 6a] and the orthogonal section view function offered by the LSM 510 software [Fig. 6b]. Both Figs. 6a and 6b were derived from the same two-photon optical image stack with a $\text{depth}=42\mu \mathrm{m}$ and an $x∕y$ plane area of $64\mu \mathrm{m}\times 64\mu \mathrm{m}$ . It is noticeable that similar to confocal microscopy, two-photon optical imaging also suffers lower image resolution at the $x∕z$ and $y∕z$ planes than at the $x∕y$ plane. Nonetheless, despite the obvious need to improve the $z$ -axis resolution of the current optical sectioning microscopy, our new approach of combining optical clearing, confocal/two-photon imaging, and postrecording image processing and projection has extended the view of microscopic tissue scaffolds from planar images to an integrated 3-D presentation.