2.1.

Materials

Porcine biopsies from Yucatan white, hairless pigs were purchased from Sinclair Research Center, Inc. (Columbia, Missouri). Human abdominal skin specimens were obtained from dermatological offices (otherwise to be discarded) with informed consent and in accordance with institutional protocols. HSE tissues, in particular, Epiderm 200X (EpD200x) and Epiderm 200 (EpD200), were purchased from MatTek Corporation (Ashland, MA).

2.2.

IR Microscopic Imaging

Skin samples were fast frozen with liquid ${\mathrm{N}}_2$ to preserve sample integrity by substantially limiting ice damage. Frozen samples were fixed on a metal chuck stage of a Bright/Hacker 5030 Microtome (Bright Instrument Company, Huntington, UK; Hacker Instruments, Fairfield, New Jersey) and sealed in ice. Following 5 min of stabilization at about $-30°\mathrm{C}$, samples were microtomed to a thickness of 5 μm and carefully transferred to ${\text{C}\mathrm{aF}}_2$ windows. After 30 min of air drying, IR images were acquired with a Perkin-Elmer Spotlight 300 system equipped with an essentially linear array ($16\times 1$) of mercury-cadmium-telluride (MCT) detector elements. Pixel size was $6.25\times 6.25\mu \text{m}$ and the spectral resolution was $8{\mathrm{cm}}^{-1}$. Two specimens of each type of HSE were microtomed and three to five images were acquired for each specimen. Images of sections of excised human skin shown herein are typical of the many ($>20$) specimens examined.

2.3.

Confocal Raman Microscopy

Skin sample surfaces were cleaned several times with a cotton swab, inserted into a brass cell with the stratum corneum side up, and placed underneath a microscope coverslip that was pressed gently against the skin to ensure good optical contact. The cell-coverslip interface was sealed with clay to keep the moisture level constant.

Raman spectra were acquired with a Kaiser Optical Raman microprobe (Ann Arbor, Michigan) at a spectral resolution of $4{\mathrm{cm}}^{-1}$. A solid-state diode laser (785 nm) with a power of $\sim 7\mathrm{mW}$ at the sample was used as the excitation source and was focused to the desired depth with a $100\times$ oil immersion objective. Backscattered light from the sample was collected with a near IR-CCD (ANDOR Technology, Belfast, Ireland, UK, Model DU 401-BR-DD). Spectra were acquired using a 60-s single exposure time, three accumulations, and cosmic ray correction. Two specimens of each type of HSE were examined and the Raman images of excised human skin shown herein are typical of the many specimens examined as mentioned above.

2.4.

IR Spectra in Bulk Phases

Skin was treated with 0.5% trypsin solution at the air-liquid interface for one day. The stratum corneum was then peeled off, rinsed with distilled water, and transferred to a ZnSe window. The isolation of SC samples from EpD200X required a protocol slightly different from human and pig skin samples, since there is no apparent distinctive physical property difference between the SC and epidermal layer in the samples. These samples were thus treated in the trypsin solution for two days, at which time no additional physical change could be observed (i.e., sample thicknesses after this time remained constant), and no additional tissue could be removed from the viable epidermis side by scratching with tweezers.

The isolated HSE SC was then placed on a ZnSe window. SC samples were dried under house vacuum for a day or two. A second ZnSe window was placed on top and the sample was contained in a temperature-controlled IR cell (Harrick Scientific, Ossining, New York). IR spectra of the dried SC were acquired with a Nicolet 6700 spectrometer (Thermo Electron Corporation, Madison, Wisconsin) operating with a spectral resolution of $2{\mathrm{cm}}^{-1}$. Melting curves were generated from spectra acquired at temperature increments of $\sim 3°\mathrm{C}$ from 20°C to the highest desired value as noted in Fig. 1. Temperature was controlled using a circulating water bath and monitored with a thermocouple (Fisher Scientific Thermostat, Model 9001, Pittsburgh, Pennsylvania) embedded in the sample cell. All experiments were run in duplicate and typical data are shown.

2.5.

Factor Analysis of IR Images

ISys software (version 3.1.1 from Malvern Instruments, UK) was used for IR and Raman spectral analysis and production of images. Spectral factor loadings and image planes of factor analysis scores were generated after linear baselines were applied over spectral regions of interest. Factor analysis, as detailed elsewhere,²⁹ is a multivariate statistical technique that reduces the dimensionality of the data by detecting patterns in the relationships between observed variables. The algorithm begins with principal component analysis (PCA) followed by a score segregation routine that seeks transformations between the PCA loadings and Beer’s Law parameters in the IR or its equivalent in Raman. Factor analysis produces a set of factor loadings while factor scores depict correlations between the actual spectrum at each pixel and the factor loadings. Calculated factor loadings resemble typical spectra, although not of pure components. Thus, a factor loading will contain spectral features from chemical constituents (mostly skin lipids and proteins) that significantly contribute to the variance at sites with respective high factor scores. Three to six significant factors were observed from most data sets.

3.1.

Bulk Phase IR Spectroscopy

3.1.1.

Conformation-sensitive regions of lipid IR spectra

One obviously important criterion for evaluating the function of HSEs is the structure and persistence of a permeability barrier.¹⁰ Traditional (macroscopic) IR is used to compare SC lipid packing and chain conformational order in isolated human SC, porcine SC, and EpD200X. Two IR spectral regions sensitive to packing and structure are used for these purposes. The methylene rocking (715 to $735{\mathrm{cm}}^{-1}$) contour is exquisitely sensitive to the packing motif adopted by the chains. As shown in Fig. 1(a), for SC isolated from human skin at temperatures less than $\sim 30°\mathrm{C}$, the rocking mode frequency splits into a well-resolved doublet with peaks near 720 and $730{\mathrm{cm}}^{-1}$. The presence of the higher frequency component is reliably diagnostic for orthorhombic perpendicular subcell packing of the lipid chains.^31,32 With increasing temperature, the doublet progressively collapses to a single peak at $\sim 720{\mathrm{cm}}^{-1}$, reflecting the disappearance of the orthorhombic packing motif and its likely replacement by hexagonal packing arrangements or (at higher temperatures) disordered chains. The progressive disappearance of the orthorhombic phase is tracked as a function of temperature in Fig. 2(a), in which the $730{\mathrm{cm}}^{-1}$ band intensity has been normalized to the skin protein Amide II mode near $1550{\mathrm{cm}}^{-1}$.

Fig. 1

Stacked absorbance plots of the methylene rocking region (715 to $733{\mathrm{cm}}^{-1}$) from (a) isolated human stratum corneum as the temperature is progressively increased from 18°C to 104°C going from top to bottom; (b) isolated porcine stratum corneum spectra were acquired from 20°C to 72°C; and (c) barrier enhanced EpD200X stratum corneum from which spectra were acquired from 20°C to 72°C.

Fig. 2

(a) Temperature dependence of the area of the $729{\mathrm{cm}}^{-1}$ band arising from the orthorhombic phase in human SC as the temperature was increased as shown. The band intensity was normalized to the intensity of the Amide II mode in the spectrum. (b) Temperature dependence of the ${\mathrm{CH}}_2$ symmetric stretching frequency for human SC (•), EpD200X SC (○), and porcine SC (▾).

In contrast to human SC, porcine SC as shown in Fig. 1(b) displays a somewhat asymmetric rocking contour with a peak at $\sim 721{\mathrm{cm}}^{-1}$ and no detectable peak near $730{\mathrm{cm}}^{-1}$ (although the asymmetry may hint at the presence of a weak feature there at the very lowest temperatures). According to the correlations discussed above, the concentration of orthorhombically packed lipids in this tissue is diminished compared to human SC. Although the absence of an orthorhombic phase is not necessarily indicative of a substantially impaired barrier, certain functional assays in porcine skin indeed suggest a less efficient barrier than in human SC.^2–4 Results similar to porcine skin are noted [Fig. 1(c)] for EpD200X, including a slight inflection near $730{\mathrm{cm}}^{-1}$ in the lowest temperature spectrum again, possibly hinting at the presence of a very small proportion of orthorhombically packed structures.

The methylene symmetric (${\upsilon }_{\mathrm{sym}}{\mathrm{CH}}_2$, 2845 to $2855{\mathrm{cm}}^{-1}$) and asymmetric (${\upsilon }_{\mathrm{asym}}{\mathrm{CH}}_2$, 2914 to $2928{\mathrm{cm}}^{-1}$) stretching bands have also proven useful for characterization of lipid assemblies. As is well known,³³ the frequencies of these modes are sensitive primarily to changes in chain intramolecular conformational order (trans-gauche isomerization). As an example of the utility of this spectral region, the temperature dependence of ${\upsilon }_{\mathrm{sym}}{\mathrm{CH}}_2$ is plotted in Fig. 2(b) for SC isolated from human excised skin, porcine SC, and EpD200X. Two transitions are evident for the human SC sample. The $\text{hexagonal}\to \text{disordered}$ ($L_{\alpha }$) transition beginning at $\sim 70°\mathrm{C}$ and terminating at $\sim 95°\mathrm{C}$ to 100°C is accompanied by a frequency increase of $3.4{\mathrm{cm}}^{-1}$, from $\sim 2850$ to $2853.4{\mathrm{cm}}^{-1}$. This increase defines the progressive formation of a highly disordered lipid structure, with a final frequency similar to that observed in $L_{\alpha }$ phase of phospholipids.³³ The second broad transition from $\sim 20°\mathrm{C}$ to 40°C, detected by a significantly smaller frequency increase from $\sim 2849.2$ to $2849.7{\mathrm{cm}}^{-1}$, reflects the orthorhombic-to-hexagonal packing transition previously detailed from the rocking mode contour in Figs. 1(a) and 2(a). This small increase indicates that the chains remain conformationally ordered in both phases.

The situation is markedly different for the porcine SC and EpD200X samples compared to human SC. For the porcine and HSE samples, the main lipid $\text{order}\to \text{disorder}$ transition is noted [Fig. 2(b)] at much lower temperatures ($\sim 50°\mathrm{C}$ to 70°C) than in human SC. The lipid phase at low temperatures in porcine and in EpD200X SC is characterized by conformationally ordered chains but also by the absence of significant levels of orthorhombic subcell packing. While there is no direct IR marker for hexagonally packed phases in the SC, the transitions noted above are consistent with their occurrence. From a functional viewpoint, the permeability barriers presented by orthorhombic versus hexagonally packed phases are likely to be quite different. Orthorhombic phases are characterized by rigid chains locked into a highly ordered packing motif while hexagonal phases are suggested from NMR studies³⁴ of polyethylene to undergo rapid rotational motion about their long axes. One might anticipate that the barrier to permeability is enhanced in lipid lamellae containing orthorhombic phases, if all other factors are assumed equal. However, as shown by Ponec et al.,¹⁰ other factors (e.g., lipid composition) evidently play a role in barrier properties. It is nevertheless clear that the IR parameters described above provide a useful biophysical characterization of lipid packing and structure within the skin barrier.

3.2.

Confocal Raman Microscopic Determination of Skin Properties

To supplement skin barrier packing information, evaluation of spatial heterogeneity in lipid composition or chain order is possible via an approach whereby spatially resolved structural information may be generated. We illustrate the utility of confocal Raman microscopy to address this issue.

As noted earlier, confocal Raman microscopy permits acquisition of spectra with a spatial resolution of 1 to 2 μm in the lateral dimension and $\sim 3\mu \mathrm{m}$ in the direction perpendicular to the SC surface. An inherent experimental problem with this technology is the single-beam nature of the experiment. When carried out in an environment such as skin in which the refractive index changes unpredictably, relative concentration information cannot be directly inferred from peak intensities. The measured intensities must be corrected for scattering efficiency changes as the excitation source samples sites at increasing depths within the tissue. Our approach to this problem (Fig. 3) is based on the observation that in skin samples studied to date, the intensity of the Phe ring stretching vibration at $1004{\mathrm{cm}}^{-1}$ decays mostly monotonically with depth from the skin surface. The assumption required to utilize the $1004{\mathrm{cm}}^{-1}$ to normalize the scattering intensities is that the concentration of Phe residues from skin proteins is constant with depth.

Fig. 3

Development of a Raman intensity standard. The area of the $1004{\mathrm{cm}}^{-1}$ band arising from the ring breathing mode of protein Phe residues is plotted as a function of position for samples of full-thickness excised human skin, EpD200, and EpD200X. Spectra were acquired over a 2-D image plane every 4 μm to a depth of 60 μm as shown. The vertical scale bar is color coded for values of Raman scattering intensity as shown on the righthand panel.

With this assumption, the diminution in $1004{\mathrm{cm}}^{-1}$ peak intensity as the excitation laser is focused progressively deeper in the tissue is presumed to arise completely from scattering losses (i.e., the inability of Raman photons from deep in the tissue to reach the detector) rather than concentration variations. The type of data acquired routinely is shown in Fig. 3 for full-thickness excised human skin, EpD200, and barrier-enhanced EpD200X, respectively. The intensity of the $1004{\mathrm{cm}}^{-1}$ band as a function of depth shows similar trends in each case, i.e., diminution with depth is more or less monotonic, but the magnitude of the effect varies from sample to sample. The vertical depth to which the variation in scattering intensity is considered reasonable (i.e., is not affected greatly by noise) occurs at a depth of 50 μm for human skin and EpD200 and 40 μm for the EpD200X sample. We assume the more conservative (latter) value for the following.

The utility of confocal Raman microscopy for characterizing lipid domains in skin is depicted in Figs. 4 and 5. Typical spectra collected from the SC of excised human skin, EpD200, and EpD200x over the range 580 to 810 and 990 to $1150{\mathrm{cm}}^{-1}$ are shown in Fig. 4. Features of interest are (1) the $701{\mathrm{cm}}^{-1}$ band arising from cholesterol,²² (2) the $1004{\mathrm{cm}}^{-1}$ Phe band discussed previously, and (3) two chain $\mathrm{C}─\mathrm{C}$ stretching modes near 1063 and $1130{\mathrm{cm}}^{-1}$ whose origins³⁵ are as follows: The 1000 to $1150{\mathrm{cm}}^{-1}$ region contains skeletal vibrations in which alternate carbon atoms move in opposite directions. These “skeletal optical modes” appear in ordered lipid phases as strong features from the $k=0$ modes of an all-trans chain at $1130{\mathrm{cm}}^{-1}$ ($A_g$ symmetry in polyethylene) and $1060{\mathrm{cm}}^{-1}$ ($B_{1g}$ symmetry in polyethylene). In disordered phases a feature arises at 1080 to $1100{\mathrm{cm}}^{-1}$ from uncoupled $\mathrm{C}─\mathrm{C}$ stretching vibrations in chain segments containing gauche rotations. These features are labeled in Fig. 4.

Fig. 4

Raman spectral data from the SC of the samples indicated. Some assignments of the various spectral features are given. The intensities in the 600 to $810{\mathrm{cm}}^{-1}$ region have been multiplied by a factor of 3 compared to those in the 990 to $1150{\mathrm{cm}}^{-1}$ region.

Fig. 5

(a) Spatial distribution of the area ratio of the $701/1004{\mathrm{cm}}^{-1}$ bands for samples of full-thickness excised human skin, EpD200, and EpD200X as marked. This ratio provides a measure of the occurrence of lipid pockets enriched in cholesterol. See text for details. (b) Spatial distribution of the area ratio of the $1063/1004{\mathrm{cm}}^{-1}$ bands for samples of human skin, EpD200, and EpD200X as shown. This ratio provides a measure of the occurrence of ordered lipid phases, since the $1063{\mathrm{cm}}^{-1}$ band arises from a particular vibration of all-trans lipid chains.

From the measured band intensities, interesting structural information may be inferred. Thus, the peak height ratio $701/1004{\mathrm{cm}}^{-1}$ is used to monitor cholesterol-enriched domains in the epidermis. This parameter is imaged to a depth of 40 μm in Fig. 5(a) for excised full-thickness human skin, EpD200, and EpD200X. Substantial variations in the distribution of cholesterol-enriched lipid domains are noted between these materials and reveal the presence of many more cholesterol-enriched lipid pockets in the EpD200X samples. In contrast, excised human skin shows lower relative cholesterol levels generally and many fewer pockets. EpD200 reveals intermediate levels of these structural elements. We note that the delineation of the SC and viable epidermis (VE) boundary as shown in Fig. 5(a) pertains primarily to human skin. These two regions of the tissue are difficult to distinguish spectroscopically in the HSEs.

Further insight into domain formation is evident from plots in Fig. 5(b) of the $1063/1004{\mathrm{cm}}^{-1}$ peak intensity ratios. Significant differences are again observed in this parameter between the samples. Pockets of a relatively higher concentration of conformationally ordered lipid are observed between depths of 15 and 40 μm for the EpD200X and between 20 and 35 μm for the EpD200. In contrast, the confocal Raman image of full-thickness excised human skin shows less variability in the concentration range of ordered lipid, with the higher concentrations appearing throughout the SC. Quantitatively, the level of ordered lipids in the HSE pockets is approximately twice that observed in the human SC. From this experiment, the Raman imaging approach is shown to provide unique information about cholesterol-enriched pockets and the distribution of conformationally ordered lipid in both SC and epidermal regions.

3.2.1.

Comparative IR microscopic imaging of excised human skin and Epiderm 200X

Compared with confocal Raman measurements, IR imaging permits the convenient sampling of substantially larger areas, up to several ${\mathrm{mm}}^2$. This gain is partly due to worse spatial resolution at the longer wavelengths involved in the IR measurement compared with Raman microscopy. A disadvantage of IR is that confocal measurements are not possible due to the transmission-based nature of the experiment. On the positive side, spectral $S/N$ ratios are significantly higher in the IR compared with the Raman measurement.

The utility of IR imaging for comparison of full-thickness excised human skin with HSE is shown in Figs. 6 and 7 in which the results of factor analysis for EpD200X and human skin are depicted for two spectral regions, namely the $\mathrm{C}─\mathrm{H}$ stretching region (2800 to $3000{\mathrm{cm}}^{-1}$) in Fig. 6, and the 1180 to $1480{\mathrm{cm}}^{-1}$ region in Fig. 7. Primary IR spectral data from these samples were concatenated prior to the application of factor analysis. Six significant factor loadings (labeled F1 to F6 in the right panel of each figure) were observed in each spectral region. The score plots derived from each are labeled F1 to F6 on the left side of the lefthand figure in each instance. In Fig. 6, the spectral-like features depicted in the loadings are assigned as follows: $\sim 2850$ and $\sim 2920{\mathrm{cm}}^{-1}$, lipid methylene symmetric and asymmetric stretching modes; $\sim 2873{\mathrm{cm}}^{-1}$, protein symmetric ${\mathrm{CH}}_3$ stretching.

Fig. 6

Factor loadings and score images for the $\mathrm{C}─\mathrm{H}$ stretching region (2800 to $3000{\mathrm{cm}}^{-1}$) of full-thickness excised human skin and EpiD200X. The original IR imaging data were concatenated and factor analysis was then undertaken. Six significant loadings labeled F1 to F6 were extracted from the data and are depicted (overlaid) in the righthand panel. The scores from each loading are plotted in the lefthand panel as labeled for both human skin and the HSE.

Fig. 7

Factor loadings and score images for the 1180 to $1480{\mathrm{cm}}^{-1}$ region of full-thickness excised human skin and EpD200X. The same data set was used as in Fig. 6. Six significant loadings labeled F1 to F6 were extracted from the data and are depicted (overlaid) in the right hand panel. The scores from each loading are plotted in the lefthand panel as labeled for both human skin and the HSE.

As in the confocal Raman images, the main value of IR imaging arises because the spatial distribution of the chemical and structural information inherent in the spectra are exposed. As such, factor loadings can be used to identify the main chemical contributors to the variance at particular regions in the score plots of particular factors. The color coding schemes in Figs. 6 and 7 are used to define the relative contribution of each factor to a particular pixel in each plot. These are in the progressively diminishing sequence $\text{red}>\text{orange}>\text{yellow}>\text{green}>\text{blue}>\text{purple}$ depicted in the color bars adjacent to each image. We now consider the structural ramifications of the variances detected.

The factor F1 in Fig. 6 depicts a reasonably ordered lipid population (${\upsilon }_{\mathrm{sym}}{\mathrm{CH}}_2=2851.7{\mathrm{cm}}^{-1}$) with very little protein, as deduced from the low IR intensity of protein methyl stretching modes. The vertical dimension of the SC layer in the F1 factor score image perpendicular to the skin surface in each tissue is $\sim 3\text{pixels}$ ($\sim 20\mu \text{m}$), and corresponds to the anticipated thickness of this layer.

Factor loading F2 (Fig. 6), with high scores in human skin and no counterpart in EpD200X, arises from a protein-rich loading with little detectable lipid as deduced from the absence of a lipid feature near $2850{\mathrm{cm}}^{-1}$ and the presence of an intense feature at $2877{\mathrm{cm}}^{-1}$. The protein is readily identified as collagen both from the characteristic shape of its Amide I contour²³ (spectral data not shown here) and from the presence of the $1338{\mathrm{cm}}^{-1}$ band characteristic³⁶ of collagen 4-hydroxyproline (hyp) residues shown in factor loadings F3 and F4 (human skin) in Fig. 7. In this instance, the molecular structure information in the loading (F3 in Fig. 7) and the similarity in the score plots between F2 in Fig. 6 and F3 in Fig. 7 confirm the imaged species as collagen, mostly in the dermal region. Since EpD200X is created from keratinocytes, no collagen is anticipated in this HSE.

Two additional factors in Fig. 6, F3 and F4, depict conformationally ordered lipid chains with methylene frequencies of 2849.2 and 2851.2, respectively. These lipids are spatially distributed in the viable epidermal (VE) region of EpD200X, which is not normally considered to possess major barrier functions. No counterparts (with similar score images) are observed for the human skin. The spatial distribution of high scores for F3 and F4 in the upper VE of EpD200X is consistent with the confocal Raman image [Fig. 5(b), right] of the relative concentration of ordered lipid ($1063/1004{\mathrm{cm}}^{-1}$ peak intensity ratio).

In addition, a substantially disordered lipid population characterized by the factor loading F6 in Fig. 6 with a methylene symmetric stretching frequency of $2855.8{\mathrm{cm}}^{-1}$ is evident mostly in the lower region of the viable epidermis for both EpD200X and in the dermal and VE regions of human skin. The structural significance of this lipid is unclear, but presumably it is not likely to be involved in barrier functions.

Finally, we note the presence of very similar factor scores from F5 in Fig. 6 and F2 in Fig. 7, with suggestive spatial distributions. In human skin, the loading is concentrated in the VE. Examination of the $\mathrm{C}─\mathrm{H}$ spectral region (F5 in Fig. 6), indicates a mixture of lipid and protein content as shown by the presence of both 2850 and $2875{\mathrm{cm}}^{-1}$ spectral features, while the loading in the 1180 to $1475{\mathrm{cm}}^{-1}$ region (F2 in Fig. 7) shows no $1338{\mathrm{cm}}^{-1}$ feature, indicating the absence of collagen. The major protein contributor to the variance is thus identified as keratin. The complement to this in the HSE section appears to be present in factors F1 and F5 in Fig. 7, again with the absence of the collagen $1338{\mathrm{cm}}^{-1}$ band and the spatial distribution of high scores located in keratin-rich areas. Our prior studies of wound healing²³ suggest that this spectral region, in particular the intensity distribution in Amide III components, may be sensitive to different keratin types. In the HSE, the molecular origin of the high factor scores observed in the $\sim 10$-micron-thick region at the bottom of F5 in Fig. 6 and F2 in Fig. 7 is undetermined.

A drawback with factor analysis is the lack of a straightforward relationship between the intensity of a factor loading in a score image and the concentration of chemical constituents of the skin in the particular pixels of interest. Simple univariate analysis can overcome this issue as shown in Fig. 8 in which the spatial distribution of the lipid/protein ratio is presented as the $2852/2872{\mathrm{cm}}^{-1}$ peak height ratio. In excised human skin, the lipid content is maximized in the thin layer ($\sim 20\mu \text{m}$) corresponding to the SC, while in the EpD200X, the lipid layer is 70 to 100 μm thick and encompasses the entire epidermal layer.

Fig. 8

Examination of the location of excess lipid in the barrier enhanced EpD200X and in excised human skin using a simple univariate measure. The intensity ratio (IR) ($2852/2873{\mathrm{cm}}^{-1}$) is imaged with color coding as shown on the righthand panel. The parameter tracks the lipid/protein ratio in the 2-D image plane.

The power and detail available from IR imaging for the characterization of HSEs is evident from these figures. The “lipid-enhanced” EpD200X reveals substantial extra lipid (shown in factors F3 and F4 in Fig. 6, and in the univariate image in Fig. 8) located in all epidermal layers in addition to the anticipated ordered lipid layer of the SC (F1 in Fig. 6). However, the relevance of the expanded spatial distribution of ordered lipid in HSE to barrier function is an interesting, albeit currently unresolved issue.

4.1.

Macroscopic IR Experiments

As is evident from Figs. 1 and 2, various spectral regions have sensitivities to different elements of molecular structure and organization. Thus, orthorhombic chain packing motifs are easily identified from the IR methylene rocking contours in Fig. 1. In applications from this laboratory utilizing this spectral parameter, we have tracked the kinetics of orthorhombic phase reformation following thermal perturbation in both isolated SC³⁷ and in three-component SC model systems.³⁸ Boncheva et al. have used similar spectral parameters with ATR-IR spectroscopic sampling in vitro to acquire thermotropic information from the SC in full thickness skin.³⁹

The reliability of these IR spectra-structure correlations is established from comparison of our detection of orthorhombic phases in the SC with the x-ray diffraction study of Ponec et al.,¹⁰ who observed orthorhombic phases in native (human) SC at 30°C that disappeared by 45°C, consistent with our measurements on isolated human SC (Fig. 1). In addition, they obtained wide-angle and electron diffraction results consistent with chain hexagonal packing in reconstructed tissues. The latter observation is consistent with our observations in HSE of minimal orthorhombic structure. Thus the spectra-structure correlations utilized here are not only consistent with a myriad of prior IR studies on simpler samples such as alkanes or phospholipids but agree with the Bouwstra lab x-ray studies of samples possessing full biological complexity.

4.2.

Vibrational Imaging

Confocal Raman microscopy offers the significant advantages over other imaging approaches in that neither probe molecules nor substantial sample preparation is required. Either full-thickness skin or skin from which the dermal layer has been trimmed are each fully suitable. Confocal measurements may be acquired to a depth of $\sim 80\mu \text{m}$ with a “$Z$” resolution of $\sim 2$ to 3 μm. In contrast, IR absorption measurements preclude confocal applications and require skin samples that have been microtomed to the appropriate thickness. However, IR imaging provides some advantages for comparative characterization of skin sections. First, samples may be examined to any desired depth. As shown in Figs. 6 and 7, the conformational order of the SC lipids and disorder of the dermal lipids and the presence of collagen are easily detected. Also, as demonstrated in our initial studies of a healing wound,²³ the spatial distribution and activation of various keratin types is suggested from their distinctive characteristic spectral contours in the 1175 to $1450{\mathrm{cm}}^{-1}$ region.

4.3.

Vibrational Spectroscopy and Imaging as a Probe of HSE Structure

To date, the type of structural information available from various physical measurements on HSEs have been elegantly demonstrated in several studies by the Leiden group who used a combination of histology, electron microscopy, and small angle x-ray diffraction. For example, in their relatively early study¹⁰ of “human epidermis reconstructed on de-epidermized dermis,” they observed, among other features of interest, formation of lamellar bodies, extrusion of same at the stratum granulosum/SC interface and the presence of well-ordered lamellar lipid phases, with the SC lipids predominantly in a hexagonal sublattice. In a second investigation,¹⁴ tissue architecture and quality of the permeability barrier were tracked in commercially available reconstructed human skin models and compared to human tissue. The same laboratory has more recently compared⁴⁰ water distribution and NMF content in HSEs and human skin and showed that the SC water level in HSEs is regulated by factors in addition to NMF. This study is relevant to issues such as desquamation, a process controlled by enzymes dependent on pH and water levels.

The above well-designed investigations provide a useful basis for comparison of native tissue with HSEs. The vibrational spectroscopy and imaging approaches described in the current work provide additional useful elements of molecular structure information. Thus, it is not possible from EM or from imaging experiments with probe molecules to detect particular protein structures or alterations in same, to characterize the spatial distribution of disordered lipid phases, nor to elucidate the presence of cholesterol-containing domains. Finally, this laboratory has recently demonstrated that IR imaging, in which Beer’s law is obeyed, permits the estimation of the spatial distribution of the absolute concentration of exogenous materials in skin.⁴¹

In addition to the evaluation of endogenous skin constituents, IR and Raman imaging experiments permit characterization of perturbations in molecular structure induced by exogenous substances. Should drugs, permeation enhancers, or other exogenous skin additives alter the physical properties of the endogenous skin components, the inherent ability of IR and Raman spectroscopies to track molecular structure changes in response to the presence of exogenous substance confers unique advantages.

In the current proof-of-principle application, the lipid packing and chain conformational order differences observed between the isolated SC of human, porcine, and EpD200X, revealed in Figs. 1 and 2, provide a useful means to compare and contrast packing motifs in various samples. In addition, the confocal Raman data in Figs. 4 and 5 reveal the presence of extensive cholesterol-containing domains in the barrier-enhanced HSE compared to excised human skin. Finally, we learned from IR imaging (Figs. 6 and 7) that the spatial location of the enhanced lipid population in EpD200X was found to be throughout the epidermal region, leaving its relevance to skin barrier function unclear. As the level of structural information that can be acquired from vibrational imaging experiments develops, these measurements should generally provide a useful guide to the relationships between barrier structure and function within HSEs.

Although the use of IR and Raman imaging technologies to characterize reconstituted skin barriers are limited to date, there are studies in the literature suggesting that these approaches ought to be of general utility. For example, Bommannan et al.⁴² utilized nonimaging ATR-IR in combination with tape stripping to evaluate the disorder of the SC intercellular lipids as a function of depth. In another application, the Puppels lab has applied Raman microscopy toward the same end.²⁵ These structural parameters may obviously define sample-to-sample differences in barrier function and are easily detected with current instrumentation. The current work along with the aforementioned studies demonstrates that extension of these types of measurements to HSEs is quite feasible.