It is challenging to simultaneously obtain both structural and chemical information from intact skin tissue. Structural information is typically visualized by employing imaging techniques that detect reflected light, such as confocal laser scanning microscopy,12.–3 multiphoton fluorescence microscopy,4 and optical coherence tomography.5 Imaging by the techniques listed above typically does not provide specific information about chemical composition. A select number of autofluorescent species in skin tissue can be investigated without tissue preparation.6 Exogenous fluorescent probes can be added, which provide indirect information about the skin chemistry.7,8 With these two approaches, fluorescence imaging can provide chemical information for selected applications. Infrared and Raman spectroscopic imaging techniques give an abundance of chemical information from a multidimensional spectral data set,9,10 but visualizing spatial distributions can require extensive data processing.11,12
Stimulated Raman scattering (SRS) microscopy is a label-free methodology for acquiring three-dimensional (3-D) structural information with a high degree of molecular specificity in real time.1314.–15 Skin tissue contains multiple components that produce strong Raman responses in the high wavenumber region of the vibrational spectrum, including proteins, lipids, and water.16 In the two-laser SRS system, the wavelengths are tuned to generate signal at frequencies corresponding to the chemical components within the tissue. The nonlinear coherent excitation enhances the signal over conventional Raman spectroscopy. Rapid scanning of the lasers in the focal plane generates chemical images with submicron resolution at video rate. Step-wise movement of the focal plane creates an image stack for visualization of the 3-D structure of the skin. Image contrast is superior compared to coherent anti-Stokes Raman spectroscopy (CARS) due to the lack of background interference from autofluorescence or nonresonant signals.13 An additional advantage over CARS is the nearly linear dependence of the SRS signal with concentration. This allows direct interpretation of the SRS vibrational spectrum and obviates the need for postacquisition processing to visualize the chemical distribution.
The architecture of porcine epidermis has been visualized using two-photon confocal microscopy.17 Porcine skin has been shown to have morphological and functional similarities to human skin.18,19 The top layer of the skin, the stratum corneum, is composed of layers of cells called corneocytes. The corneocytes are observed to group together in clusters typically 100 to 250 μm across the surface. The cell clusters are separated from one another by microanatomical structures, termed canyons, which are invaginations or microfolds of stratum corneum cell layers. These SC canyons often reach the viable cells of the epidermis and may potentially provide delivery pathways for topically applied material.
SRS imaging is utilized in this work to probe the chemical nature of skin cluster and canyon microstructure by visualizing the distributions of proteins, lipids, and water throughout the cell cluster and canyon regions. This powerful spectroscopic technique provides the unique ability to obtain detailed 3-D chemical images of the microscopic features of the layers of the skin. Image stacks were collected at three vibrational frequencies that approximately correspond to the Raman shifts of the three predominant skin chemical classes (Table 1). The spectroscopic information inherent in SRS imaging allows display of the spatial distribution of water across and through the skin layers, and identifies lipid-rich regions separating the water-rich cell clusters. The explicit chemical nature of the SRS signal also provides the opportunity to track the spreading of topically applied materials in native skin in real time without disruption. We investigate the diffusion of oleic acid, often used as a skin penetration enhancer.20 Applying a fully deuterated molecule allows selection of a unique wavelength for imaging this material. Overlaying the oleic acid and skin images shows how the material can reach deeper layers.
Wavelengths and Raman assignments for stimulated Raman scattering (SRS) imaging. The four wavelengths selected with the pump probe for this study and the corresponding Raman shifts, vibrational mode assignments, and the primary chemical species contributing to the SRS signal are shown.
|Pump beam wavelength (nm)||Raman shift (cm−1)||Vibrational mode||Band assignment|
|810.6||2950||ν(CH3)||Protein and lipid|
Methods and Materials
Back and belly skin from Yucatan white, hairless pigs was received from Sinclair Bioresources (Auxvasse, Missouri). All tissues were harvested only as waste byproducts from animals aged three to six months. Tissues were only rinsed with water prior to shipment and were not frozen. A Padgett Electric Dermatome (Plainsboro, New Jersey) was used to isolate the top 500 μm of the skin. Pieces of the dermatomed skin, , were placed between a glass microscope slide and No. 1.5 glass cover slip (VWR, Radnor, Pennsylvania) for SRS transmission microscopy. (Sigma-Aldrich, St. Louis, Missouri) was applied directly to the skin surface using a positive displacement pipette (Gilson, Middleton, Wisconsin) to deliver 2 μL onto a treatment area.
SRS Microscopic Imaging
The details of SRS have been previously described.13,21 The skin sample is mounted on the stage of a modified upright laser scanning microscope (BX81, Olympus, Pittsburgh, Pennsylvania) configured for transmission illumination. Two overlapping laser beams are focused on the sample with either a UPLSAPO20X air objective or a UPLSAPO60XW water immersion objective (Olympus). The 1064-nm Stokes beam source was an laser (picoTRAIN, High-Q, Watertown, Massachusetts). Part of the 1064-nm output is used to pump a tunable optical parametric oscillator (Levante Emerald, APE-Berlin, Berlin, Germany). The pump beam was tuned to 816.0, 810.6, or 781.3 nm for skin imaging, corresponding to Raman shifts of ca. 2850, 2950, or , respectively. The main contributors from skin at these vibrational frequencies can be assigned to vibrational modes from lipids, proteins, and water22 (Table 1). The pump laser was tuned to 869.2 nm to image the vibrational mode at after topical application of , ensuring that only the applied material would be visible in the SRS image stack. Image integration time was typically 4 to . The beams were scanned in two dimensions with a pair of galvanometer mirrors (Olympus fluosview FV-300). The transmitted light was collected with a 1.4-NA oil condenser lens (Nikon, Melville, New York) and detected by a large-area photodiode (FDS1010, Thorlabs, Newton, New Jersey). Images were collected parallel to the skin surface. Image stacks were obtained through sample stage movements controlled by the microscope software, with a step size of 2 μm over depth ranges of 60 to 100 μm. An image stack at a single wavelength can be acquired in 30 to 60 s.
SRS Image Analysis
The three associated image stacks (lipid, protein, and water) were combined into a four-dimensional (4-D) array (X,Y,Z, wavelength) in MATLAB® (version 2009a, MathWorks, Natick, Massachusetts). The region of interest (ROI) (cell cluster and/or canyon) was designated on the water image using a rectangular area selection tool that records the pixel location values for all selected points. Depth profiles for each pixel location in the ROI are extracted from the 4-D array for each wavelength. The water profiles for clusters and canyons were computed using the method previously reported for in vivo confocal Raman spectroscopic measurements.23 Each water profile was normalized against the corresponding protein depth profile with a calibration factor of 1. The SRS technique minimizes background response [typically of the peak intensity15], but a small profile is present in the image response. The background signal is neither spectroscopic nor resonant in origin. The magnitude of this response will change slightly at different regions of the Raman spectrum, but since the wavelengths we are using are very near to each other, the ratio method used for analysis will cancel out background effects. Additionally, the method proposed by Caspers et al. is, at best, an approximation due to their use of overlapping spectral bands for protein and water.23 A pure response is, therefore, never achieved, so any calculation for the constant (R) can only ever yield an approximate value because it is dependent on the calculated values for water and protein. We used the same equation but replaced with the arbitrary value of . This maintains the same profile shape but changes the absolute scale. Depth profile plots are presented as relative water content. The water content approximation, therefore, was modeled after the Caspers method as , where and are the intensities from the water and protein images, respectively, and . and represent single-wavelength intensities from the SRS images, whereas the confocal in vivo Raman spectroscopic method employs integrated wavelength ranges. The water content values reported here are averaged over all pixels in the ROI. Lipid content values are also generated by a similar protocol, with the lipid intensity L replacing water intensity in the above equation.
Cross-sections calculated across canyon regions were generated by selecting a rectangular ROI of the water image stack and averaging across the planes to form a single water depth profile. The resulting images are displayed using a relative normalized scale from 0 to 255, corresponding to the gray-scale range of the original images. The red false color indicates high water content and the blue false color represents low water content. The overlaid 3-D surface contour images of and the skin protein microstructure were generated in Amira (version 5.4, Visage Imaging, San Diego, California). False colors were selected to maximize contrast between the two images in the 3-D rendering.
An example of the cluster and canyon structure of porcine skin is shown in an SRS water () image collected near the skin surface at magnification (Fig. 1). In SRS images, the intensity of each pixel corresponds to the amount of the chemical component(s) with a Raman signal at that vibrational frequency. In this data, there is some contribution in each of the single-wavelength images to components other than the labeled chemical component. Every effort was made to minimize overlap with the other component images so as to maximize image purity. The idea here is not to present absolute quantitation of the component content from the images, but rather to show that differences in chemical information can be obtained in a relatively straightforward manner. The water () image shows a clear distinction between high-intensity water-rich cellular regions (cell clusters), such as region 2 in Fig. 1, and low-intensity water-deficient canyons separating the clusters (region 3). Multiple cell clusters and canyons are observed at this lower magnification. More detailed examination of the changes in skin microstructure as a function of depth can be accomplished by collecting multiple image stacks at magnification. The unique and varying architecture of the epidermal layers is represented in SRS images selected from different depths in three image stacks acquired at the higher magnification (Fig. 2). The three columns show images taken at the frequencies corresponding to protein (), lipid (), and water (), while the three rows portray depths of 6, 18, and 28 μm below the surface. Near the surface of the skin, the corneocytes aggregate into clusters ranging from 100 to 250 μm wide [Figs. 2(a) to 2(c)]. Lipid, protein, and water content appear to be distributed across the visible areas, as evidenced by the approximately uniform SRS intensities for each of the three frequencies. The dark spots in these images (marked “D” in the first row of images) are optical shadows from air pockets between the skin surface and coverslip. The corneocyte clusters are separated by invaginations 10 to 25 μm wide, features previously designated as canyons.17
The transition from the stratum corneum into the viable epidermis is discernible in images collected 18 μm below the surface [Figs. 2(d) to 2(f)]. At this depth, the distinct boundaries between the clusters are visible, such as between the arrows in Fig. 2(e). The average canyon width in these images is . The signal intensity in the protein and lipid images is highest along the canyon walls, as evidenced by the bright lines outlining the canyons [Figs. 2(d) and 2(e)]. Fine lines are visible between the canyons, as can be seen between the arrows in the lipid () image [Fig. 2(e)]. These are the edges of stratum corneum corneocytes following the contours of the canyon, as has previously been proposed.17 The water signal is significantly weaker in the canyons than in the cluster regions. In the cluster regions, nuclei are visible as dark ovals in the protein and lipid images [Figs. 2(d) and 2(e)], and as bright ovals in the water image [Fig. 2(f)]. Around this depth, the distinctive polygonal shape of the cells is revealed as the lipid-rich cell wall membranes appear in the lipid () image [Fig. 2(e)]. The chemical specificity of SRS imaging allows these cell walls to be visualized, whereas they are often missed in other optical techniques, such as autofluorescence.13,17
Looking deeper into the epidermis at 28 μm below the surface, the canyons become progressively narrower [Figs. 2(g) to 2(i)]. The water () image is dark between the cell clusters, indicating any water that may be present in the canyons is below the limit of detection. The cell walls are brightly illuminated in the lipid (), showing the cell shape becoming rounder [Fig. 2(h)]. The density of visible nuclei has increased in this layer.
Measured SRS signals are proportional to molecular concentration, providing a distinct advantage over nonlinear techniques, such as CARS.8 This linear dependence is a requirement for quantitative examination of the spatial distribution of chemical species within skin substructures. The relative water content spanning one intercluster canyon was averaged through all images of the water () depth stack at each pixel location from the pixel region designated as region 1 in the water image (Fig. 1). The average water content values are plotted as a false color map, distinctly showing the boundary between the water-rich cells (displayed in red) and the water-free canyon (in blue) (Fig. 3). The red false color indicates high water content and the blue false color represents low water content. This visual representation can be used to measure the distance between the two clusters. For this example, the widest gap—the spacing between the two bright red regions—is , corresponding to a width of . The water map gives an indication of the varying canyon width as a function of depth. The weaker water signals shown in yellow and green correspond to narrowing of the canyon at a depth of . The water content of each pixel is averaged through the entire stack. The pixels shown in yellow and light blue are located in areas that are open canyons near the surface, but viable cells deeper in the skin.
This technique provides detailed depth information about the hydration of skin and the distribution of water throughout the substructures. The mean water signal intensity was calculated throughout the image depth stack for the cluster and canyon features designated as regions 2 and 3 in Fig. 1. Water depth profiles are generated from the intensities in the protein () and water () images, as described above. These depth profiles further demonstrate that the water content of the skin resides predominantly within the cellular areas and not in the canyons [Fig. 4(a)]. The distribution of water within cell clusters, shown here as a function of depth, is in agreement with water profiles measured using in vivo confocal Raman spectroscopy in human skin.10 In healthy skin, the water content is in the stratum corneum. There is a sharp rise in water content located at the boundary between the stratum corneum and viable cells, eventually reaching a plateau of 60 to 70%. We observed the same sigmoidal profile of water as a function of depth in SRS image data [Fig. 4(b)] as seen in single-point in vivo confocal Raman microspectroscopy data.
This same analytical approach can be employed to generate lipid depth profiles. Using the same ROI 2 and 3 from Fig. 1, evaluation of the lipid () image stack shows that there is a significant accumulation of lipid material near the top of the canyon [Fig. 4(c)]. These canyon lipids are in a discrete volume, as evidenced by the sharp maximum in the lipid profile, in contrast to the relatively even distribution in the cell clusters. Lipid levels in all areas start to decline 15 μm below the skin surface [Fig. 4(d)]. It is important to consider that the signal at is not exclusively from skin lipids and may include some contribution from protein as well.
Canyons can potentially play a role in the delivery of materials into the viable skin cells depending upon the compactness, thickness, and ultrastructure of stratum corneum that forms the canyon walls. We are beginning to explore possible penetration pathways by evaluating the diffusion of topically applied to the porcine skin surface. Figure 5 shows a 3-D space model rendered from two SRS image stacks collected after application of oleic acid. The canyon and cluster structure of the skin is colored orange in the 3-D image obtained from the protein signal in the SRS stack. A separate oleic acid image stack was obtained for the C-D stretching mode at ca. 25 min after treatment. The oleic acid surface contour map is superimposed in white against the canyon structure, showing the deuterated material residing in the canyon and coating the canyon walls. This example shows penetration of materials deep into the canyons in a short period of time.
The quantitative spectroscopic response of SRS enables extraction of relative component concentration profiles directly from the image depth stacks. This can be employed not only to understand the chemical composition of skin features, but also to follow delivery and penetration of actives in real time. SRS imaging provided an improved understanding of the skin architecture first observed with two-photon microscopy.17 We have been able to build upon those classifications by identifying the chemical nature of canyons and cell clusters. Canyons are low in water content and occasionally contain lipidaceous material capping the top of the canyon close to the skin surface. The oleic acid example shows that the material can be located within skin substructures at different depths. Transdermal diffusion is typically measured by collecting all the materials passing through a thick () piece of skin in a Franz cell. Stratum corneum and epidermal penetration have typically been assessed by removing layers of skin by tape-stripping and conducting quantitative analysis of the individual tapes.20,24 An apparent drawback of the layer-by-layer approach exposed here is that material sitting deep in the canyon might be considered “penetrated” active. In contrast, SRS shows that the oleic acid is primarily present in the canyons and has not entered into the cells in the cluster regions of the skin after 25 min of exposure. SRS microscopy, thus, provides the unique opportunity to investigate penetration pathways of actives at high spatial resolution by exploiting the intrinsic contrast of the active.
SRS microscopy provides a unique view of the chemical organization of the skin, rapidly and noninvasively assessing the layers of intact tissue. These experiments were conducted with a priori knowledge of the tissue, which enabled selection of distinct, representative wavelengths for imaging the key components of the skin chemical structure. When interpreting the images, one must remember that these spectroscopic bands are not unique for a single chemical species. For example, the wavelength assigned to protein does contain a small contribution from lipid content. Multivariate analysis is preferred over single-wavelength evaluation, as shown previously.11,25,26 The speed of the hardware to perform rapid wavelength changes is a limiting factor toward implementation of multivariate analysis. Development of the instrumentation will progress toward fast multiwavelength spectroscopy when a large spectral window can be scanned, opening new areas of exploration. The expanded wavelength domain will enhance the ability to identify and distinguish components with similar chemical structures and overlapping spectroscopic signatures, such as distinct lipid components present within the skin.
We thank Paul Baker for creating the three-dimensional renderings, and Fa-Ke Lu and Minbiao Ji for their technical support with the stimulated Raman scattering system. The work conducted at Harvard is funded by Unilever.
M. Rajadhyakshaet al., “In vivo confocal scanning laser microscopy of human skin: melanin provides strong contrast,” J. Invest. Dermatol. 104(6), 946–952 (1995).JIDEAE0022-202Xhttp://dx.doi.org/10.1111/jid.1995.104.issue-6Google Scholar
K. Sauermannet al., “Age related changes of human skin investigated with histometric measurement by confocal laser scanning microscopy in vivo,” Skin Res. Technol. 8(1), 52–56 (2002).0909-752Xhttp://dx.doi.org/10.1046/j.0909-752x.2001.10297.xGoogle Scholar
S. Neerkenet al., “Characterization of age-related effects in human skin: a comparative study that applies confocal laser scanning microscopy and optical coherence tomography,” J. Biomed. Opt. 9(2), 274–281 (2004).JBOPFO1083-3668http://dx.doi.org/10.1117/1.1645795Google Scholar
K. König, “Multiphoton microscopy in life sciences,” J. Microsc. 200(2), 83–104 (2000).JMICAR0022-2720http://dx.doi.org/10.1046/j.1365-2818.2000.00738.xGoogle Scholar
T. GambichlerV. Jædicke, “Optical coherence tomography in dermatology: technical and clinical aspects,” Arch. Dermatol. Res. 303(7), 457–473 (2011).ADREDL0340-3696http://dx.doi.org/10.1007/s00403-011-1152-xGoogle Scholar
C. L. Evanset al., “Chemical imaging of tissue in vivo with video-rate coherent anti-Stokes Raman scattering microscopy,” Proc. Natl. Acad. Sci. U. S. A. 102(46), 16807–16812 (2005).1091-6490http://dx.doi.org/10.1073/pnas.0508282102Google Scholar
C. L. EvansX. S. Xie, “Coherent anti-Stokes Raman scattering microscopy: chemical imaging for biology and medicine,” Annu. Rev. Anal. Chem. 1, 883–909 (2008).ARACFU1936-1327http://dx.doi.org/10.1146/annurev.anchem.1.031207.112754Google Scholar
W. Minet al., “Coherent nonlinear optical imaging: beyond fluorescence microscopy,” Annu. Rev. Phys. Chem. 62, 507–530 (2011).ARPLAP0066-426Xhttp://dx.doi.org/10.1146/annurev.physchem.012809.103512Google Scholar
H. U. GremlichB. Yan, Eds., Infrared and Raman Spectroscopy of Biological Materials, Marcel Dekker Inc., New York (2001).Google Scholar
P. J. CaspersG. W. LucassenG. J. Puppels, “Combined in vivo confocal Raman spectroscopy and confocal microscopy of human skin,” Biophys. J. 85(1), 572–580 (2003).BIOJAU0006-3495http://dx.doi.org/10.1016/S0006-3495(03)74501-9Google Scholar
J. J. AndrewT. M. Hancewicz, “Rapid analysis of Raman image data using two-way multivariate curve resolution,” Appl. Spectrosc. 52(6), 790–796 (1998).APSPA40003-7028http://dx.doi.org/10.1366/0003702981944472Google Scholar
T. M. HancewiczJ. H. Wang, “Discriminant image resolution: a novel multivariate image analysis method utilizing a spatial classification constraint in addition to bilinear nonnegativity,” Chemom. Intell. Lab. Syst. 77(1), 18–31 (2005).CILSEN0169-7439http://dx.doi.org/10.1016/j.chemolab.2004.07.013Google Scholar
C. W. Freudigeret al., “Label-free biomedical imaging with high sensitivity by stimulated Raman scattering microscopy,” Science 322(5909), 1857–1861 (2008).SCIEAS0036-8075http://dx.doi.org/10.1126/science.1165758Google Scholar
L. Chritet al., “In vivo chemical investigation of human skin using a confocal Raman fiber optic microprobe,” J. Biomed. Opt. 10(4), 044007 (2005).JBOPFO1083-3668http://dx.doi.org/10.1117/1.2003747Google Scholar
D. C. CarrerC. VermehrenandL. A. Bagatolli, “Pig skin structure and transdermal delivery of liposomes: a two photon microscopy study,” J. Control. Release 132(1), 12–20 (2008).JCREEC0168-3659http://dx.doi.org/10.1016/j.jconrel.2008.08.006Google Scholar
G. A. SimonH. I. Maibach, “The pig as an experimental animal model of percutaneous permeation in man: qualitative and quantitative observations—an overview,” Skin Pharmacol. Appl. Skin Physiol. 13(5), 229–234 (2000).SPAPFF1422-2868http://dx.doi.org/10.1159/000029928Google Scholar
A. M. BarberoH. F. Frasch, “Pig and guinea pig skin as surrogates for human in vitro penetration studies: a quantitative review,” Toxicol. Vitro 23(1), 1–13 (2009).TIVIEQhttp://dx.doi.org/10.1016/j.tiv.2008.10.008Google Scholar
A. Naiket al., “Mechanism of oleic acid-induced skin penetration enhancement in vivo in humans,” J. Control. Release 37(3), 299–306 (1995).JCREEC0168-3659http://dx.doi.org/10.1016/0168-3659(95)00088-7Google Scholar
M. B. J. Roeffaerset al., “Label-free imaging of biomolecules in food products using stimulated Raman microscopy,” J. Biomed Opt. 16(2), 021118 (2011).JBOPFO1083-3668http://dx.doi.org/10.1117/1.3516591Google Scholar
B. W. BarryH. G. M. EdwardsA. C. Williams, “Fourier transform Raman and infrared vibrational study of human skin: assignment of spectral bands,” J. Raman Spectrosc. 23(11), 641–645 (1992).JRSPAF0377-0486http://dx.doi.org/10.1002/jrs.v23:11Google Scholar
P. J. Casperset al., “In vivo confocal Raman microspectroscopy of the skin: noninvasive determination of molecular concentration profiles,” J. Invest. Dermatol. 116(3), 434–442 (2001).JIDEAE0022-202Xhttp://dx.doi.org/10.1046/j.1523-1747.2001.01258.xGoogle Scholar
A. Cappielloet al., “Application of liquid chromatography-direct-electron ionization-MS in an in vitro dermal absorption study: quantitative determination of trans-cinnamaldehyde,” Anal. Chem. 83(22), 8537–8542 (2011).ANCHAM0003-2700http://dx.doi.org/10.1021/ac201839xGoogle Scholar
C. Xiaoet al., “Uncertainties in depth determination and comparison of multivariate with univariate analysis in confocal Raman studies of a laminated polymer and skin,” Appl. Spectrosc. 58(4), 382–389 (2004).APSPA40003-7028http://dx.doi.org/10.1366/000370204773580202Google Scholar