2.1.

Mice, Allergens, and Epicutaneous Sensitization

In the in vivo study of imaging the mouse ear, nude mice were used as our animal model. The nude mouse is a genetic mutant that suffers from an agenesis of the thymus, and (thus is incapable of mounting most types of immune responses. The skin phenotype of the mouse is that hair shafts fail to penetrate the epidermis, resulting in a lack of body hair. The experimental protocols were performed as previously described.²⁵ In the AD study, $6\text{-}\text{to}10\text{-}\text{week}$ -old female BALB/c or C57BL/6 mice were obtained from the Animal Center of the College of Medicine, National Taiwan University. AD is associated with increased Th2-type inflammation in acute skin lesions, but chronic AD results from the switching to a Th1-type inflammation.³ It is known that C57BL/6 mice are a strain with a propensity for Th1 responses while BALB/c mice are a strain with a propensity for Th2 responses.²⁶ We use fixed skin of C57BL/6 mice for the evaluation of chronic skin lesions in vitro, and observed the fresh skin of the BALB/c mice for the assessment of acute skin lesions in vivo. All procedures performed on the mice were approved by The Animal Care and Use Committee of National Taiwan University Hospital. The protein antigens used in this study were ovalbumin (OVA) $(100\mathrm{mg}∕\mathrm{ml})$ and staphylococcal enterotoxin (SEB) $(0.4\mu \mathrm{g}∕\mu \mathrm{l})$ (Sigma Chemical Company, Staint Louis, Missouri). The antigen solution was prepared in distilled water. Epicotanesus (EC) sensitization of mice was performed as previously described.²⁷ In brief, $20\mu \mathrm{L}$ of OVA and $10\mu \mathrm{L}$ of SEB were first added to the paper disk insert of a Finn Chamber (Epitest Company, Tuusula, Finland), and the disk was applied to a $1\text{-}{\mathrm{cm}}^2$ patch of sterile gauze, which was then applied to the skin on the lateral side of their abdomen, then fixed and secured with an elastic bandage to ensure skin contact and to prevent oral ingestion. Three groups of C57BL/6 mice were divided according to the antigen type. Group 1 was OVA group, group 2 was $\mathrm{OVA}+\mathrm{SEB}$ group, and group 3 was the negative control group with application of phosphate buffered saline (PBS), which was used for comparison. Three mice were used in each group. The courses of immunization were for $5\text{days}$ .

2.2.

Histological Analysis

Skin specimens (lesional skin) obtained from patched areas after the patch was removed at the end of sensitization were used for histological examination. The skin specimens over the other side of the abdomen (nonlesional skin) were also obtained for comparison. Multiple $4\text{-}\mathrm{mm}$ punch biopsies were performed, and skin specimens were fixed in 10% formaldehyde overnight, transferred to 30% sucrose, and embedded into optimal cutting temperature (OCT) formulation for frozen horizontal sectioning. The specimens were finally stained with hematoxylin and eosin (HE) for histological analysis. After 10% formaldehyde fixation, some specimens were observed under the stage of microscopy.

2.3.

Cr:Forsterite Laser-Based Multimodality Nonlinear Microscopy System

The imaging system used for our experiments was developed in the Ultrafast Optics Laboratory of the Graduate Institute of Photonics and Optoelectronics, National Taiwan University (Fig. 1 ).^{21, 22} The Cr:forsterite laser-based multimodality nonlinear microscopy system was adapted from a FluoView 300 (FV300) scanning unit (Olympus, Tokyo, Japan) combined with a BX51 microscope (Olympus, Tokyo, Japan). The light source was a home-built Cr:forsterite fs laser centered at $1230\mathrm{nm}$ with a $130\text{-}\mathrm{fs}$ pulse width and a $110\text{-}\mathrm{MHz}$ repetition rate. The infrared (IR) beam exiting the laser was first shaped by a telescope and then directed into the Olympus FV300 scanning unit and the Olympus BX51 microscope. Real-time $x\text{-}y$ scanning was accomplished with a pair of galvanometer mirrors inside the scanning unit. The specimen was mounted on a fixed stage within the microscope. An IR water-immersion objective of working distance $2\mathrm{mm}$ (LUMplanFL/IR 60×/NA 0.9, Olympus, Tokyo, Japan) was used to focus the laser beam into the observed skin specimen. The average illumination power applied onto the surface of the specimen was $100\mathrm{mW}$ . Typically, a mean power of $100\mathrm{mW}$ with a $110\text{-}\mathrm{MHz}$ repetition rate was chosen, which corresponds to a pulse energy of $0.91\mathrm{nJ}$ . For in vivo observation, the laser beam exiting the objective was scanned across the specimen, and the backward-propagating multimodal nonlinear signals from the specimen were collected with the same water-immersion objective. The microscope objective employed to focus the laser beam can also act as the collection lens to collect the backward propagation SHG, THG, and multiphoton excited fluorescence. These signals were deflected and discriminated with the $865\text{-}\mathrm{nm}$ dichroic mirror (865dcxru, Chroma Technology, Rockingham, (Vermont) and transmitted through a color filter (SKG5, CVI Laser optics, Albuquerque, (New Mexico), which ensures filtering out the fundamental laser light at $1230\text{-}\mathrm{nm}$ wavelength. Different optical harmonic signals were separated with another beamsplitter (BS 1, 490DRXR, Chroma Technology, Rockingham, (Vermont). The backward-propagating harmonic signals were detected by different photomultiplier tubes (PMTs) (Harmamatsu R4220P for THG and Harmamatsu R943-02 for SHG, Bridgewater, Westport, (Connecticut) with 410- and $615\text{-}\mathrm{nm}$ narrowband interference filters (D410/30× and D615/10×, respectively, Chroma Technology) in front, which were synchronized with the galvanometer mirrors. The backward-propagating TPF signals, which were not deflected with dichroic mirrors, were separated with another beamsplitter (BS 2, 865dcxru, Chroma Technology) and detected by another PMT (Harmamatsu R928P). The scanning rate of the FV300 scanning unit was $1000\text{lines}∕\mathrm{s}$ corresponding to two frames per second for a $512\times 512\text{pixels}$ resolution. Spatial resolution was $\approx 500\mathrm{nm}$ and $\approx 410\mathrm{nm}$ for SHG and THG signals in the $x\text{-}y$ plane, respectively. Resolution was on the order of $1\mu \mathrm{m}$ in the $x\text{-}z$ plane.²² Optical observations were made as described.

Fig. 1

The experimental setup of the Cr:forsterite laser-based multimodality nonlinear microscopy system. A live mouse mounted on a fixed stage within the microscope ready for acquiring images is shown. A thermal blanket maintaining the body temperature of the test mouse was used during images acquisition. BS, beam splitter; CF, color filter; DM, dichroic mirror; PMT, photomultiplier tube; THG, third-harmonic generation; SHG, second harmonic generation; TPF, two-photon fluorescence. (Color online only.)

2.4.

Determination of Hyperkeratosis of Epidermis

We measured the $x\text{-}z$ -sectioned nonlinear images in fixed skin flaps with intact dermis of C57BL/6 mice. Purple was used as a pseudocolor for the THG signal from SC and epidermis. The skin layers of SC and epidermis were determined as previously described.²¹ Linear lengths of the thicknesses of SC and epidermis on the $x\text{-}z$ plane images were determined using ImageJ (National Institutes of Health, Bethesda, Maryland). Five fixed skin flaps were taken from individual mice of each group (control, OVA, and $\mathrm{OVA}+\mathrm{SEB}$ ) to generate a total of 20 $x\text{-}z$ plan images for each group.

2.5.

Quantitative Analysis of Two-Photo Fluorescence and Third-Harmonic Generation from Epidermis and Second-Harmonic Generation from Dermis

We use the same $x\text{-}z$ -sectioned nonlinear images taken from fixed skin samples as described before for quantitative analysis of nonlinear signals. Yellow was used as the pseudocolor for the TPF signal from SC. Green was used as the pseudocolor for the SHG signal from the collagen within the dermis. The skin layers of dermis were determined as previously described.²¹ In brief, the optical sectioning image was saved as uncompressed $8\text{-}\text{bit}$ grayscale tiff files and analyzed using the ImageJ software. The signal level was converted to a numerical value using a grayscale from 0 (white) to 255 (black). The background noise was subtracted from all the acquired values. To account for variations in the grayscale intensity due to possible nonuniformity of intact skin tissue, mean gray values (MGVs, average gray value within the region of interest) were measured to determine the levels of TPF and/or THG signals and SHG signals, which reflect the distribution of SC and the distribution of collagen fibers, respectively. The ratio of MGV was defined as: (the MGV of the lesional skin)/(the MGV of the paired nonlesional skin).

2.6.

In Vivo Imaging of Murine Atopic Dermatitis Model

To compare the severity of dermal fibrosis of individual BALB/c mice, we acquired stacks of planar projection ( $x\text{-}y$ plane) images with a depth interval of $1.5\mu \mathrm{m}$ starting from the epidermodermal junction at an average depth of $z=30\mu \mathrm{m}$ untill $z=60\mu \mathrm{m}$ from the sample surface in vivo from the control and the OVA group, respectively. Serial images at different tissue depths separated by $1.5\mu \mathrm{m}$ from individual mice were obtained for analysis. The mean gray value of SHG within the dermis of each planar projection image was determined from three different skin areas of each BALB/c mouse using the same analysis protocol of ImageJ as described before.

2.7.

Study of Skin Specimen from Atopic Dermatitis Patients

Skin biopsy was performed from the lesional skin in the left forearm of three female volunteer with AD. After biopsy, the skin specimens were fixed in 10% formaldehyde overnight, transferred to 30% sucrose, and embedded into OCT formulation for frozen horizontal sectioning. The specimens were finally stained with HE for histological analysis. Some samples of specimens after 10% formaldehyde fixation were observed under the stage of multimodal nonlinear microscopy. Stacks of $x\text{-}y$ images with a depth interval of $1.5\mu \mathrm{m}$ starting from the skin surface were acquired. The simultaneous mean gray intensity of THG and TPF of serial images were also determined. The Declaration of Helsinki protocols were followed and patients gave their written, informed consent. Our study was approved by the Institutional Review Board of National Taiwan University Hospital.

2.8.

Statistical Analysis

Data are expressed as $\text{mean}\pm \text{standard}$ deviation, unless otherwise specified. The comparison of linear lengths of SC and epidermis and mean gray value among groups were statistically analyzed with a nonparametric Mann-Whitney $U$ test. We compared the mean gray value within the dermis in vivo between the control and the OVA group with analysis of variance (ANOVA). Spearman’s rank correlation test was used to test for correlation between TPF and THG. All statistical analyses were performed with Prism software (Graph Pad, San Diego, California). A $P$ value of less than 0.05 was considered to indicate a statistically significant difference.

3.1.

Multimodal Images Depicting Detailed Morphology of Layered Skin Tissues, Subcutaneous Tissues, and Cartilage Tissues in the Nude Mouse Ear In Vivo

THG signals are sensitive to optical inhomogeneities of local tissue,²⁸ hence THG imaging can be used for general morphological assessment. As for SHG, collagen fibrils are known to provide excellent contrast in biological SHG microscopy.²⁹ SHG imaging has recently emerged as a novel approach for studying the pathophysiological remodeling processes of the collagen matrix.³⁰ It is thus highly desirable to take advantage of the least invasive Cr:forsterite fs laser-based multimodality nonlinear microscopy to in vivo map morphological features and tissue structures exhibiting nonlinear optical characteristics.

Figure 2 show the representative multimodal $x\text{-}y$ plane images acquired at different tissue depths $\left(z\right)$ in the ear from the nude mice in vivo. The individual shape of densely packed epidermal cells and transitions between epidermal layers could be clearly resolved through the THG modality. The skin surface of the nude mice reveals the stratum corneum with the typical morphological features of hexagonal shapes of corneocytes. The dark round region exhibiting the nucleus inside the cell body of epidermal cells was easily identified due to the strong THG contrast with laminated organelles,¹⁶ consistent with the observation of neurons and epithelial cell of oral mucosa using THG microscopy.^{31, 32} The basal cells are smaller in size compared to the corneocyte. The deeper tissue sections reveal fibrous structures of collagen shown by the SHG signals, which are indicative of the dermis layer. The vascular networks accommodated by the dermis and/or subcutaneous tissue are positively identified through THG by the blood vessels filled with red blood cells (erythrocytes). The adipose tissue or subcutaneous fat composed of adipocytes are clearly revealed by the THG signals with the typically spherical shapes of adipocytes. Below the subcutaneous tissue, THG further observed honeycomb-like structures with hexagonal arrangement. The morphology of this layer is consistent with the structure of cartilage tissue below the dermis. Collectively, each distinct layer of the skin and supporting tissue of nude mouse ear can be clearly resolved by the THG and SHG signals.

Fig. 2

In vivo multimodal $x\text{-}y$ plane optical sectioning images at various depths of the ear of a nude mouse. THG signals (purple) reveal morphological features of epidermal tissue layers, subcutaneous tissue (including blood vessel and fat tissue), and the cartilage tissue. The presence of a cartilage tissue layer below the subcutaneous tissue is characteristic of the ear. SHG signals (green) depict the distribution of collagen fibers within dermis. Scale bar: $20\mu \mathrm{m}$ . (Color online only.)

3.2.

Multimodal Images Revealing Morphological Features of Layered Tissue in the Epidermis and Local Collagen Structures in the Dermis

To determine whether we could identify the histological features in the skin lesions of AD with the Cr:forsterite laser-based multimodality nonlinear microscopy, we induced a murine AD model by epicutaneous sensitization with OVA.²⁷ Multimodal images and emission spectra were acquired for the excitation wavelength ${\lambda }_{\mathrm{ex}}$ , equivalent to $1230\mathrm{nm}$ from the excised skin lesions after fixation. Because we aimed to take advantage of the capacity of Cr:forsterite laser-based multimodality nonlinear microscopy to evaluate the morphological and structural changes of skin lesions of AD, we performed the observation only into the depth of dermis (about $z=60\text{to}90\mu \mathrm{m}$ ). The representative multimodal $x\text{-}y$ plane images acquired at different tissue depths $\left(z\right)$ into the fixed skin samples from the control and the OVA-treated C57BL/6 mice are shown in Fig. 3 (a). We observed many “streaks” of strong THG signals depicting the general morphology of stratum corneum (SC) and epidermal cells. Figure 3(a) also depicts the local aggregations of some fibril that correspond to collagen fibers (for example, $z=60\mu \mathrm{m}$ ) through the occurrence of the SHG signals.

Fig. 3

Multimodal nonlinear images of fixed skin from control and OVA-treated C57BL/6 mice (a) Multimodal optical $x\text{-}y$ plane sectioning images at various depths. THG signals (purple) reveal general morphology of SC and shape of densely packed epidermal cells. The dark round region indicates the nucleus inside the cell body of epidermal cells. SHG signals (green) map local aggregations of collagen fibers. Some TPF signals (yellow) within the SC (arrow) were noted in the skin of OVA-treated mice. Scale bar: $20\mu \mathrm{m}$ . Data are representative images of three mice in each group. THG, third-harmonic generation; SHG, second-harmonic generation; TPF, two-photon fluorescence; sc, stratum corneum; ec, epidermal cells; n, nucleus of epidermal cells; dp, dermal papilla; cb, collagen bundle. (b) Emission spectra obtained from the surface $(z=0)$ for $1230\text{-}\mathrm{nm}$ excitation in the OVA-treated mice. Counts are in arbitrary units. (Color online only.)

A typical emission spectrum acquired from the skin surface is shown in Fig. 3b. We confirmed the nonlinear signal sources by verifying that the emission spectra of THG signals were with a peak at one third the excitation wavelength $\left(410\mathrm{nm}\right)$ and those of SHG signals were with a peak at half the excitation wavelength $\left(615\mathrm{nm}\right)$ [Fig. 3b]. Some fluorescence signals were noted within SC [ $z=0$ , Fig. 3a]. The fluorescence signals occurred much more frequently in the fixed skin of OVA-treated mice. The two-photon autofluorescence (TPF) signals we observed had a central emission wavelength around $680\mathrm{nm}$ [Fig. 3b].

3.3.

Comparing Optical Sectioning Images to Conventional Microscopic Histology

To ascertain that the histological features observed under the Cr:forsterite laser-based nonlinear microscope were comparable with the conventional histological findings, we compared the nonlinear optically sectioned images with the conventional histologically sectioned images. To evaluate the potential of our nonlinear microscope to assess the histological changes under different disease states, we also introduced another group of murine AD models sensitized to staphylococcal enterotoxin B (SEB), which is known to contribute to the exacerbation of AD.³³ The representative $x\text{-}z$ plane images from the fixed skins of three groups of C57BL/6 mice are shown in Fig. 4 . These typical pathological changes in the skin lesions of AD can be clearly demonstrated using a conventional microscope [Fig. 4a], consistent with the histopathological features reported previously.³⁴ We found that sensitization with $\mathrm{OVA}+\mathrm{SEB}$ induced the most prominent thickening of the SC and epidermis and a remarkable fibrosis in the dermis, indicating a more severe disease status. With the nonlinear microscopy, we observed that after sensitization with $\mathrm{OVA}+\mathrm{SEB}$ , SC exhibited the strongest THG signals, while dermis displayed considerable local aggregations of collagen fibers depicted through the SHG signals [Fig. 4b]. Our results demonstrate similar morphological and structural changes in optically sectioned images of intact skins compared with standard histological images of the same tissue. The epidermis in both treated groups (OVA and $\mathrm{OVA}+\mathrm{SEB}$ ) exhibited multiple cellular layers. It was also noticed that the intensities of the THG signals within the epidermal layers were higher in both treated groups. These findings are also comparable with the histological findings, confirming that the increased cellular infiltration within the epidermis can be readily assessed by THG imaging.

Fig. 4

Comparison of hematoxylin-eosin stained histological images with multimodal nonlinear images of fixed skin from control, OVA-treated mice, and $\mathrm{OVA}+\mathrm{SEB}$ -treated C57BL/6 mice. (a) Representative H&E-stained histological images of fixed skin including the whole skin layer. (b) Representative multimodal nonlinear $x\text{-}z$ plane images of fixed skin. The dark round region indicates the nucleus inside the cell body of epidermal cells. Notice the increased cellular layers within the epidermis of OVA- and $\mathrm{OVA}+\mathrm{SEB}$ -treated mice groups. sc, stratum corneum; ec, epidermal cells; n, nucleus of epidermal cells; cb, collagen bundle. Scale bar: $20\mu \mathrm{m}$ . Data are representative images of three mice in each group. (Color online only.)

No significant SHG signals were detected in the layers of epidermis, consistent with previous reports.^{10, 21} As shown in Fig. 2, SHG mainly reflects the collagen distribution in dermis. Even though THG is contributed from optical inhomogeneity, Fig. 2 also shows that THG signals are strong in stratum corneum, basal cells, red blood cells, the surface of adipocytes, and cartilage tissues. It is thus reasonable to expect weaker THG signals in the dermis layer, while SHG is a much better modality to identify the collagen matrix distribution. In the $x\text{-}z$ images shown in Fig. 4, the THG signals appear dark in the dermis, layer also due to the limited dynamic range of any single image, as the THG intensities in the deeper layer were not normalized according to the signal degradation, which is our case. This dynamic range issue is much less serious when viewing the image in the $x\text{-}y$ plane, as shown in Figs. 2 and 3. It is also important to notice that we did not take images deeper than $90\mu \mathrm{m}$ from the sample surface. As for signal degradation, there were two main contributions for the THG signal degradation: 1. the collagen fibrils in the dermis are known to be highly scattering and could decrease the detected THG intensity,²² and 2. we had shown that optical signal degradation in the studied human skin specimens is dominated by the distortion-induced broadening of the point spread function (PSF).³⁵ This signal degradation made the SHG/THG signals from the deep layers too weak to show in the same $x\text{-}z$ image with the shallow layer due to limited dynamic range of any single image. Even limited by the dynamic range of one $x\text{-}z$ image, these unprocessed nonlinear images have enough depth penetration capability for the required analysis.

3.4.

Epidermal Hyperkeratosis Revealed by Third-Harmonic Generation Signals

To determine whether we could assess the severity of AD in a quantitative manner, we measured the thickness of epidermis (taken from $z=0\mu \mathrm{m}$ to the epidermodermal junction) and SC from the $x\text{-}z$ plane images of fixed skins in lesions of the murine AD models, and compared those with the conventional histological images. A total of 200 measurements (SC and epidermis, respectively) were obtained from each group. For histological sections, results were calculated from 40 conventional HE images at $\times 400$ of each group. We detected significant differences in the thickness of epidermis and SC between three groups using both the nonlinear microscope [Fig. 5a ] and by using conventional histological methods [Fig. 5b] (see Table 1 ). Sensitization with $\mathrm{OVA}+\mathrm{SEB}$ induced the greatest thickening of the SC and epidermis.

Fig. 5

Determination of the thickness of stratum corneum and epidermis in fixed skin of control, OVA-treated, and $\mathrm{OVA}+\mathrm{SEB}$ -treated C57BL/6 mice groups using THG signals. (a) Thickness of epidermis and SC obtained by measuring the optical sectioning images in the $x\text{-}z$ plane. (b) Thickness of epidermis and SC obtained from the conventional histological images. Each data point represents one measurement of the thickness of epidermis and SC, respectively, from control (◇), OVA (▽), and $\mathrm{OVA}+\mathrm{SEB}$ (△) groups Horizontal lines indicate the mean.

Table 1

Measurement of the thickness of epidermis and stratum corneum by histological and optical sections in control, OVA-treated, and OVA+SEB -treated mice group. Mean±S.D(μm) , and n=number of measurement.

3.5.

Two-Photon Fluorescence Levels Correlate with Third-Harmonic Generation Levels within Thickened Stratum Corneum

Our observation of certain TPF signals within the SC that exhibited markedly prominent THG signals (Fig. 3) led us to conclude that the occurrence of these TPF signals was most likely related to the thickening of SC associated with allergen sensitization. We measured the levels of the TPF signals from the $x\text{-}z$ plane nonlinear optical images of fixed skin using ImageJ, and compared them between three groups. A total of 40 measurements were obtained for each group. We found the mean gray intensities of the TPF signals for OVA- and $\mathrm{OVA}+\mathrm{SEB}$ -treated groups ( $7.58\pm 1.52$ and $8.43\pm 1.48$ , respectively) were greater than those of the control group $(6.83\pm 1.22)$ , indicating that the intensity of the TPF signals was related to the severity of AD [Fig 6 (a)]. The representative TPF and THG images acquired from $x\text{-}z$ and $x\text{-}y$ planes of the fixed skin from the three groups are shown in Figs. 6b, 6c, 6d. We found that TPF imaging is characterized by the fluorescence pattern within the SC. The TPF signals exhibit diffuse patches of yellow pseudo color in the images of the $\mathrm{OVA}+\mathrm{SEB}$ -treated mice. Notably, there seemed to be a close correlation between the distribution of TPF signals and THG signals. We then analyzed the magnitude of the TPF and THG signals from the $x\text{-}y$ plane images acquired at different tissue depths $\left(z\right)$ into the fixed skin samples using ImageJ. We found a positive correlation between the TPF signals and the THG signals within the control, OVA, and $\mathrm{OVA}+\mathrm{SEB}$ groups [Figs. 6b, 6c, 6d]. Notably, the correlation coefficient (Spearman $\rho$ ) of the OVA- and $\mathrm{OVA}+\mathrm{SEB}$ -treated groups reveals that TPF signals and the THG signals from the SC are more closely correlated than the control groups. Our previous observations confirmed that the general distribution of SC could be revealed by strong THG signals (Fig. 3). Hyperkeratosis is defined by thickening of the skin caused by an increased thickness of the SC.³⁶ We had demonstrated that epicutaneous sensitization with allergens induces thickening of the SC and epidermis (Fig. 5). In lesional skin of AD, there is increased hyperkeratosis of the epidermis, which is less prominent within normal skin tissue. We found the relationships between these two signals are similar in both lesion and normal tissue, but the association is less prominent within normal skin tissue. Collectively, our findings suggest that increased autofluorescence/TPF within the SC is closely associated with the hyperkeratosis of skin (revealed by increased THG signals), which in turn is a manifestation of more severe AD.

Fig. 6

Correlation between the TPF signal and THG signal in thickened stratum corneum in skin lesions of murine AD models. We measured the intensity levels of TPF and THG signals by determining the mean gray value (MGV) of optical sectioning images ( $8\text{-}\text{bit}$ grayscale tiff files) of the fixed skin from three groups. (a) Comparison of MGV of the TPF signals on fixed lesional skin (left) and ratio of the MGV of TPF on fixed lesional skin (LS) to that on fixed nonlesional skin (NS) (right). The box extends from the 25th percentile to the 75th percentile of all measurements. Horizontal line indicates the median. (b), (c), and (d) Simultaneous TPF and $\mathrm{THG}+\mathrm{SHG}$ images from $x\text{-}z$ and $x\text{-}y$ planes of fixed skin from mice of (b) control, (c) OVA-treated, and (d) $\mathrm{OVA}+\mathrm{SEB}$ -treated (d) C57BL/6 mice. The relationships between the mean gray value of TPF and that of THG on the same optical sectioning image at different tissue depths of epidermis were analyzed by Spearman correlation coefficient $\left(\rho \right)$ , and the trends of change between the mean gray value of TPF and that of THG with increasing tissue depth of epidermis is shown. sc, stratum corneum; ec, epidermal cells; n, nucleus of epidermal cells; cb, collagen bundle. Scale bar: $20\mu \mathrm{m}$ . (Color online only.)

3.6.

Extent of Dermal Fibrosis Revealed by Increased Second-Harmonic Generation Signals

SHG signal intensity reflects the distribution of connected tissues, composed mainly of type-1 collagen fibers, in the dermis layer.²¹ To determine whether we could assess the severity of AD by quantifying the extent of dermal fibrosis, we measured and compared the intensities of SHG signals from $x\text{-}z$ plane images of fixed skin from three groups. A total of 100 measurements were obtained from each group. The comparisons of the intensities of SHG signals among the three groups are shown in Figs. 7a and 7b . We found sensitization with $\mathrm{OVA}+\mathrm{SEB}$ induced the greatest enhancements of SHG signals, implying a direct correlation between the degree of fibrosis and severity of AD. The comparisons of representative $x\text{-}y$ plane SHG images at $z=60\mu \mathrm{m}$ from the fixed skin of three different mice are shown in Fig. 7c. We observed that the collagen fibers in the control group were fine bundles that lie in haphazard arrangement, while in the OVA and $\mathrm{OVA}+\mathrm{SEB}$ group, the dense bundles of collagen were organized in an interlacing fashion admixed with some hypertrophic collagen fibers [Fig. 7c]. Our findings supported that the intensity of the SHG signal is strongly dependent on the collagen fibril thickness.³⁵ The increased distribution of hypertrophic collagen fibers was consistent with our finding that sensitization with allergens induced increased levels of SHG signals.

Fig. 7

Correlation between SHG intensities and the extent of fibrosis in fixed skin of control, OVA-treated mice, and $\mathrm{OVA}+\mathrm{SEB}$ -treated C57BL/6 mice group. We measured the intensities of SHG signals by determining the mean gray value (MGV) of optical sectioning images ( $8\text{-}\text{bit}$ grayscale tiff files) from the $x\text{-}z$ plane of fixed skin from three groups. (a) Comparison of MGV of SHG on fixed lesional skin and (b) ratio of the MGV of SHG on fixed lesional skin (LS) to that on fixed nonlesional skin (NS). The box extends from the 25th percentile to the 75th percentile of all measurements. Horizontal line indicates the median. (c) Representative $x\text{-}y$ plane SHG images of fixed skin from control, OVA-treated mice, and $\mathrm{OVA}+\mathrm{SEB}$ -treated C57BL/6 mice group. SHG signals (green) map different distribution patterns of collagen fibers. The dense collagen fibers organize in interlacing fashion in the OVA group, while hypertrophic collagen bundles display aggregated structure in the $\mathrm{OVA}+\mathrm{SEB}$ groups. The number (#) indicates different individual mice. Scale bar: $20\mu \mathrm{m}$ . Data are representative images of each of three mice in each group. (Color online only.)

3.7.

In Vivo Multimodal Histology of Mice Skin

The high efficiency of the various endogenous nonlinear optical signals (THG, TPF, and SHG) involved show that real-time multimodal imaging lends itself to facile acquisition and analysis of excised tissue in a noninvasive manner. Because most of our analyses were performed in vitro, we also sought to test the possibility of the safe application of Cr:forsterite laser-based nonlinear microscopy in vivo. To determine whether we could replicate the previous findings on equivalent skin lesions elicited in C57BL/6 mice, we replaced mouse strain C57BL/6 with BALB/c mice and induced AD by sensitization with OVA. We then acquired the serial images at depth intervals of $1.5\mu \mathrm{m}$ from the skin on the abdomen of anesthetized mice using nonlinear microscopy in vivo. In vivo movies of stacks of $x\text{-}y$ plane images in mice skin from control and OVA-treated groups were obtained (supplementary files: Videos 1 and 2, respectively). We observed the same histological morphology of epidermis and structural details of dermis as that of in vitro experiments in each group. Some strong fluorescence signals were noted within the SC of the epidermis from the OVA-treated group, but not within the control ones. We measured the levels of the TPF signals using ImageJ and compared them between control and OVA-treated groups. Again, we found the mean gray intensities of the TPF signals for OVA-treated groups were significantly greater than those of the control group ( $14.13\pm 1.48$ versus $7.04\pm 0.54$ , $P<0.0001$ ), indicating that the intensity of the TPF signals was related to AD in vivo. Moreover, to determine whether we could evaluate the serial changes of dermal fibrosis at different tissue depths of dermis, we measured the intensities of SHG signals from $x\text{-}y$ plane images of skin lesions from both mice groups in vivo. We found that the changes in the intensities of SHG acquired from the epidermodermal junctions of three different skin areas of individual mice were significantly different between control and OVA-treated groups [Fig. 8a ]. The representative serial $x\text{-}y$ plane images at different depths are shown in Fig. 8b, indicating that the increased intensities of SHG signals were directly proportional to increased amounts of dense collagen fibers at various tissue depths of the dermis of OVA-sensitized mice. Representative H and E-stained histological images of $x\text{-}y$ planes at various depths of fixed skin confirmed that the dense bundles of collagen were organized in an interlacing fashion in the OVA-treated group [Fig. 8c].

Fig. 8

In vivo assessment of dermal fibrosis in individual BALB/c mice. (a) The change of dermal fibrosis represented by the increased mean gray intensity of SHG from dermis of individual mice. The mean gray value of SHG was measured from $x\text{-}y$ plane images from three different abdominal skins of individual mice in vivo. Values are $\text{mean}\pm \text{SD}$ . The comparisons between control and OVA-treated BALB/c mice groups were analyzed by ANOVA. (b) Three representative SHG images at depths of 15, 22.5, and $30\mu \mathrm{m}$ from the epidermodermal junction, respectively, from both groups. In the OVA group, the amounts of collagen fibers (represent by SHG signal in green) increase proportionately with the tissue depth of dermis. Scale bar: $20\mu \mathrm{m}$ . Data are representative of three mice in each group. (c) Representative HE-stained histological images of $x\text{-}y$ plane at various depths of fixed skin from control and OVA-treated BALB/c mice. sc, stratum corneum; ec, epidermal cells; n, nucleus of epidermal cells; dp, dermal papilla; cb, collagen bundle. (Color online only.)

Video 1

In vivo movies of depth-resolved $x\text{-}y$ plane sectioning images from the abdominal skin of control BALB/c mice. THG signals in purple pseudocolors reveal the general morphology of SC and epidermis. SHG signals in green pseudocolors map local aggregations of collagen fibers within dermis. The optical depth difference between adjacent images is $1.5\mu \mathrm{m}$ . Image size: $80\times 80\mu \mathrm{m}$ . Video is representative of three mice (QuickTime, 3 MB).

Video 2

In vivo movies of depth-resolved $x\text{-}y$ plane sectioning images from the abdominal skin of OVA-treated BALB/c mice. In comparison with Video 1, thickened SC of epidermis and hypertrophic collagen fibers within dermis were clearly demonstrated. Note that the bright spot (yellow pseudocolor) in the right upper field indicates the occurrence of two-photon autofluorescence (TPF) signals within the thickened SC. The optical depth difference between adjacent images is $1.5\mu \mathrm{m}$ . Image size: $80\times 80\mu \mathrm{m}$ . Video is representative of three mice (QuickTime, 2 MB).

3.8.

Preclinical Application in Atopic Dermatitis Patients

To ensure the reproducibility of our previous findings in a murine AD model, optical sectioning images of fixed skin from human AD patients were acquired. We included one female volunteer who was $26\text{years}$ old and diagnosed with AD for $3\text{years}$ , and who provided a biopsy specimen from the AD lesion over her left forearm extensor side. The representative optical sectioning and conventional histological images from the $x\text{-}z$ plane are shown in Fig. 9a . We observed that increased THG signals reflect the markedly thickened SC, which corresponds to epidermal hyperkeratosis in the HE staining image [Fig. 9a]. The thickness of the SC in the female forearm was found to be $10\text{to}15\mu \mathrm{m}$ .¹⁰ Both types of images showed that the thickness of the SC in the forearm of the female patient with AD was around $20\text{to}40\mu \mathrm{m}$ , supporting the fact that increased thickness of SC (hyperkeratosis) is a characteristic feature of chronic AD.⁴ These data also illustrate that the feasibility of determining epidermal thickness and SC in human AD studies using nonlinear microscopy is comparable to that of conventional microscopy. The representative THG-SHG $x\text{-}y$ plane images acquired at different tissue depths $\left(z\right)$ in the fixed skin sample are shown in Fig. 9b. We observed that serial changes in the morphological features of SC, stratum spinosum, and stratum basale could be clearly recognized. Additionally, we observed increased TPF within the thickened THG signal representing SC [Fig. 9a]. The dark round region indicates the nucleus inside the cell body of epidermal cells due to the strong THG contrast with laminated organelles.¹⁶ We validated the nonlinear signal origins by verifying that the emission spectra of THG, SHG, and TPF were centered at 410, 615, and $680\mathrm{nm}$ , respectively [Fig. 9c]. We tried to verify the origin of TPF by measuring the emission spectrum acquired from pure human keratin solution (K0253, Sigma Aldrich), but no measurable fluorescence from pure keratin solution was detected. Many protein molecules are expressed in particular locations in the SC.³⁶ Our results indicated that the measured TPF might be contributed from other endogenous molecules associated with the process of keratinization/hyperkaratosis.

Fig. 9

Comparison of optical sectioning image with HE-stained histological images of fixed skin lesions from an AD patient. (a) The optical nonlinear sectioning image from two different $x\text{-}z$ planes are shown (white arrow). The morphological features of SC in the epidermis are clearly demonstrated by THG signals in purple pseudocolor. The dark round region indicates the nucleus inside the cell body of epidermal cells. The autofluorescence is located within the thickened SC (yellow arrow). (b) Serial $x\text{-}y$ plane optical sectioning images from fixed skin lesions of an AD patient. THG signals (purple) reveal general morphology of SC and transitions between epidermal layers; SHG signals (green) map local aggregations of collagen fibers. sc, stratum corneum; ss, stratum spinosum; sb, stratum basale; ec, epidermal cells; n, nucleus of epidermal cells; dp, dermal papilla; cb, collagen bundle. Scale bar: $20\mu \mathrm{m}$ . (c) Emission spectra obtained from the surface $(z=0)$ of fixed skin from an AD patient for $1230\text{-}\mathrm{nm}$ excitation in an AD patient. Counts were in arbitrary unit. (Color online only.)

To investigate the possible association of TPF with hyperkeratosis, we further enrolled two additional female volunteers who were 18 [Figs. 10a, 10b, 10c ] and $16\text{years}$ old [Figs. 10(d), 10(e), and 10(f)], respectively. The representative simultaneous THG and TPF images acquired from the $x\text{-}y$ plane of the same position of fixed skin are shown in Figs. 10a and 10d, respectively. We found an excellent correlation between the localization and distribution of THG and TPF signals. We then applied the same analysis technique as in the murine AD model for assessing the intensities of TPF and THG signals from $x\text{-}y$ plane images. We found significantly positive correlations between the intensities of TPF and THG signals from the fixed skin of the two AD patients [Figs. 10b, 10c, 10e, 10f]. Thus we were able to show that the changes in TPF intensity with hyperkeratosis in the skin lesion of AD patients in humans are compatible with our mouse model studies.

Fig. 10

Increase in intensity of TPF signal within the thickened stratum corneum in skin lesions of AD patients. (a) and (d) Simultaneous $x\text{-}y$ plane TPF and THG images of fixed skin from two AD patients. (b) and (e) Correlation between the mean gray value of TPF with that of THG analyzed by Spearman correlation coefficient $\left(\rho \right)$ . (c) and (f) Trends of change between the mean gray value of TPF and that of THG with increasing tissue depth of epidermis. Scale bar: $20\mu \mathrm{m}$ . (Color online only.)