2.1.

Tissue Samples

Skin tissue was provided through the Anatomical Gift Programme of the Royal College of Surgeons of Ireland (RCSI). Details of the cadaver samples available for this study are provided in Table 1. Skin samples were collected from 11 human cadavers. In each case, both dorsal (back) hand and proximal, medial thigh sections were employed, and each sample was measured in unprocessed, formalin-fixed, paraffin-embedded (FFPP), and subsequently dewaxed form. All samples are preserved at $-80°\mathrm{C}$ before cutting and analysis. Unprocessed cross-sections of 20 μm thickness were cut with a cryomicrotome (LeicaCM 1850 UV) and were stored at $-20°\mathrm{C}$ until used. Spectral profiling of hand and thigh sections, unprocessed and processed, of the samples indicated by shading in Table 1 are presented in detail.

Table 1

Human cadavers. Spectral profiling of hand and thigh sections, unprocessed and processed, of the samples indicated by shading are presented in detail.

The skin tissue was automatically processed (Leica Histokinette 2000) to wax blocks using four principle steps, as follows:

After wax impregnation, tissue was embedded and sliced into 20 μm sections using a microtome, mounted on a ${\mathrm{CaF}}_2$ substrate and dried. The samples were immersed in a series of baths consisting of two baths of xylene (British Drug Houses, Dorest, UK) for 5 and 4 min, respectively, two baths of Ethanol Absolute (Merck, Dorest, UK) for 3 and 2 min, and a final bath of Industrial Methylated spirits 95% (Lennox Dublin, Ireland) for 1 min.

2.2.

Cell Cultures

HaCaT cells are spontaneously immortalized human epithelial keratinocytes derived from adult skin, and have the characteristics of basal epidermal keratinocytes.⁴³ This cell line can be used as an in vitro model for highly proliferative epidermis.⁴⁴

HaCaT cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) F12 ($1:1$) medium (Sigma, Dorset, UK) containing 10% fetal calf serum (Gibco, Irvine, UK), 1% penicillin-streptomycin solution 1000 IU (Gibco, Irvine, UK), 2mM L-glutamine (Gibco, Irvine, UK), and $1\mathrm{\mu g}{\mathrm{mL}}^{-1}$ hydrocortisone (Sigma, Dorset, UK) in a humidified atmosphere containing 5% ${\mathrm{CO}}_2$ at 37°C. HaCaT cells were seeded at a concentration of $4\times {10}^4$ cells per substrate onto ${\mathrm{CaF}}_2$ substrate (Hellma Ltd., UK), previously sterilized using ethanol then dried in a laminar flow. All samples were incubated for 24 h at 37°C, 5% ${\mathrm{CO}}_2$ before measurement. Cells were measured live, in NaCl solution, using the LUMPlanF1, Olympus immersion objective.

2.3.

Biochemical Compounds

For comparison to tissue spectra, a number of biochemical compounds were analyzed by Raman spectroscopy. The samples were purchased from Sigma-Aldrich (Ireland). Samples of ceramide, sphingomyelin, and $\mathrm{L}\text{-}\alpha$-phosphatidylcholine (1,2-Diacyl-sn-glycero-3-phosphocholine, 99% from egg yolk) were dispersed in chloroform, and small amounts of material were drop cast onto ${\mathrm{CaF}}_2$ substrates.

2.4.

Raman Instrumentation

A Horiba Jobin-Yvon LabRAM HR800 spectrometer with an external 300 mW diode laser operating at 785 nm as source was used throughout this work. For the measurements, either a $\times 100$ objective (MPlanN, Olympus) or a $\times 100$ immersion objective (LUMPlanF1, Olympus) was employed, each providing a spatial resolution of $\sim 1\mathrm{\mu m}$ at the sample. The confocal hole was set at 100 μm for all measurements, the specified setting for confocal operation. The system was spectrally calibrated to the $520.7{\mathrm{cm}}^{-1}$ spectral line of silicon, and the intensity response function was corrected using the Standard Reference Material (SRM) No. 2243 of the National Institute of Standards, Boulder, Colorado, USA (NIST SRM 2243, 2242, 2241).⁴⁵ The Labram system is a confocal spectrometer that contains two interchangeable gratings (300 and $900\text{lines}/\mathrm{mm}$, respectively). In the following experiments, the $300\text{lines}/\mathrm{mm}$ grating was used, giving a spectral dispersion of $\sim 1.5{\mathrm{cm}}^{-1}$ per pixel. The detector used was a 16-bit dynamic range Peltier cooled CCD detector. A step size of 2 μm was employed for tissue mapping.

Tissue samples were measured under water immersion to minimize the spectral background, and spectra were recorded using the immersion objective (LUMPlanF1, Olympus).⁴⁶ All biochemical samples were recorded using the $\times 100$ objective (MPlanN, Olympus). Once all spectra were acquired, a background of substrate measured under identical conditions was subtracted. Minimal baseline correction, smoothing, and normalization were also performed in order to improve the quality of the acquired spectra.

2.5.

Data Analysis

The different data analysis steps were performed using Matlab (Mathworks, USA). Before statistical analysis, a Savitsky-Golay filter (5th order, 7 point) was applied to smooth any spurious peaks of the spectra and reference constituting the background signal.

K-means cluster analysis (KMCA) was employed to analyze the spectral variations in tissue. It is one of the simplest unsupervised learning algorithms that solves the well known clustering problem and is often used for spectral image analysis.¹⁹ In general, clustering is the partitioning of a data set into subsets (clusters) so that the differences between the data within each cluster are minimized, and the differences between clusters are maximized according to some defined distance measure. Using KMCA, the large amount of data in a spectral map can be reduced to mean spectra, and the spatial distribution can easily be visualized. The Raman data were used as inputs for KMCA. The clustering analysis algorithm was used to find groups of spectra with similar spectral characteristics (clusters), each one representing regions of the image with similar biochemical profiles. After KMCA, a different color is assigned to each cluster. Each grid element of the spectral map is then assigned the color of the particular cluster to which its spectrum belongs. In this way, a pseudo-color image of the skin sections is created to visualize the organization of the clusters in the original image. As the tissue mapping step size was 2 μm for all measurements, pixel size in the KMCA maps is $2\mathrm{\mu m}\times 2\mathrm{\mu m}$. The data range for all KMCA was limited to the fingerprint region, 400 to $1800{\mathrm{cm}}^{-1}$.

3.1.

Identification of the Different Skin Layers in Unprocessed Samples

Figure 1(a) shows an optical microscopic image of a 20 μm thick, unprocessed hand section of sample no 3. Visually, three different regions are apparent. Raman and FTIR spectroscopy, coupled with KMCA, can be used for the identification of different structures and classification of tumoral regions in tissue sections.^29,47 It may, therefore, be anticipated that distinct biochemical regions within the skin section can be identified. The maximum biochemical information is contained within the so-called fingerprint region of the spectrum (400 to $1800{\mathrm{cm}}^{-1}$) and therefore this region was initially analyzed. In an automated spectral map of the tissue sections, the spectra recorded on the edge of the tissue exhibit a high degree of variability due to the transition from the outer layer of tissue to the substrate. The variability between these spectra can interfere with the clustering analysis, resulting in the creation of distinct clusters. Best visualization and reproducibility of the different structures existing within the tissue was achieved by setting the number of clusters to five. In this way, the variability in the spectra obtained at the edge of the tissue is contained in distinct clusters and does not interfere with the identification of different structures present in the tissue. The spectra were assigned to the five different groups according to their similarities, and a color was attributed to each cluster. False color maps were constructed representing the partition of the different clusters in the tissue. The resulting image can be seen in Fig. 1(b).

Fig. 1

(a) Optical image of unprocessed hand tissue section; (b) K-means cluster analysis of Raman maps of unprocessed hand; (c) KMCA mean Raman spectra of unprocessed hand illustrating the differentiation of the superficial stratum corneum (SC; cluster $1=\mathrm{A}$), the intermediate epithelium (cluster $2=\mathrm{B}$), and the dermis (cluster $3=\mathrm{C}$); (d) Optical image of unprocessed thigh tissue section; (e) K-means cluster analysis of Raman map of unprocessed thigh; (f) KMCA mean Raman spectra of unprocessed thigh, illustrating the differentiation of the superficial SC (cluster $5=\mathrm{A}$), the underlying epithelium (cluster $3=\mathrm{B}$), and the dermis (cluster $4=\mathrm{C}$).

Three different structural layers can be found in the skin: the epidermis, the dermis, and hypodermis. The latter is too deep to be sampled in vivo using Raman microscopy and is not examined in this study. The epidermis forms the protective layer against the surrounding environment. The dermis provides a structural support to the skin, and the hypodermis is a connective tissue layer where fat is stored. The epidermis can also be subdivided into four different layers, starting with the stratum basale, adjacent to the dermis, containing mainly keratinocytes but also melanocytes responsible for the production of melanin, a pigment which protects against ultraviolet (UV) radiation. The second layer is called the stratum spinosum and is formed from dividing basale cells migrating towards the surface of the skin. The third layer is named the stratum granulosum and is characterised by anuclear cells. As a natural process of maturation, the cells flatten and lose their nucleus. Finally, in contact with the exterior environment, is the SC. The cells have reached the last stage of their maturation and are described as non-viable, cornified cells, called corneocytes. A fifth layer, the stratum lucidum, is found in the sole of the foot and palm of the hand and is therefore not present in our samples.

Three different regions of skin can be well differentiated spectroscopically, and Fig. 1(c) shows KMCA mean spectra related to the three regions of unprocessed hand derived from the subsection of Fig. 1(a) (indicated by the white rectangle). Of the other two clusters identified, the black (cluster 5) corresponds to the substrate, while the light blue (cluster 4) in Fig. 1(b) is not associated with specific skin structure and represents either a loss of layer integrity during the sectioning process or presence of debris on the substrate. Nevertheless, the specificity of the information contained in the spectra recorded allows discrimination of different skin layers using KMCA. Analysis of a thigh tissue section yields similar results, as shown in Fig. 1(d), 1(e), and 1(f). Notably, in both cases, the number of layers found is less than that of the skin anatomy described above. It is thus of interest to analyze the spectral features which differentiate these layers of skin, as well as skin from different anatomical sites and elucidate the biochemical origin of their differentiation.

3.2.

Characterization of the Dermal Layer

Figure 2(a) presents a comparison between the KMCA cluster 3 mean spectra for unprocessed hand and the equivalent cluster for thigh [see A and B in Fig. 2(a)] with pure collagen [see C in Fig. 2(a)]. It can be easily seen that the spectrum of the skin layer is dominated by collagen, and it is therefore identified as the dermis. This is due to the high composition of collagen in human dermis, constituting about 70% of the dry weight and 90% of the total protein content. There are about 20 types of collagen that exist in the body, but 80% of skin collagen is Type 1 and 15% is Type 3. The remaining 5% is thought to be predominately Type IV collagen.^48–50

Fig. 2

(a) KMCA mean spectra of unprocessed hand skin dermis cluster (cluster 3) (A), unprocessed thigh skin dermis cluster (cluster 4) (B), and spectrum of pure collagen (C); (b) KMCA mean spectra of the intermediate epithelial layer cluster (cluster 2) of unprocessed frozen hand skin (A), the intermediate epithelial layer cluster (cluster 3) of unprocessed thigh skin (B), and spectrum of pure melanin (C); (c) KMCA mean spectra of the SC cluster (cluster 1) of unprocessed hand (A), the SC cluster (cluster 5) of unprocessed thigh (B), spectra of Ceramide (C), Sphingomyelin (D), and $\mathrm{L}\text{-}\alpha$-phosphatidylcholine (E).

Assignments of most Raman bands of collagen have been made by Frushour and Koenig.⁵¹ Specific features in the collagen spectrum are two intense bands at $\sim 855$ and $\sim 938{\mathrm{cm}}^{-1}$. These bands originate from the amino acid side chain vibrations of proline and hydroproline as well as from a C–C stretching vibration of the collagen backbone. Proline and hydroxyproline make up about one fourth of the amino acids in collagen, a higher proportion than in most other proteins.⁴⁸ More generic protein bands present in the spectrum of the collagen and the dermis appear at $1452{\mathrm{cm}}^{-1}$, $1104{\mathrm{cm}}^{-1}$ due to CH deformation and C–N stretching, while features at 1004 and $1032{\mathrm{cm}}^{-1}$ are due to symmetric ring breathing and CH in-plane bending (phenylalanine).^52–56 These bands are also evident in the mean spectra of the other layers, but the collagen specific peaks are absent, and they are thus distinctive signatures of the dermal layer.

3.3.

Characterization of the Epidermal Layers

Overlaying the dermis, KMCA of the Raman maps identify two epidermal layers, although the anatomy of the epidermis divides it into four distinct layers. Analysis of the spectral information elucidates the origin of the subclassification.

Figure 2(b) displays a comparison between the KMCA cluster 2 mean spectrum of unprocessed hand (A) and the equivalent mean spectrum of unprocessed thigh (B) skin section corresponding to the intermediate epidermal layer with the spectrum of pure melanin (C). The pure melanin spectrum [see C in Fig. 2(b)] is dominated by the broad fluorescence emission spectrum, and no prominent Raman peaks are apparent. The spectra of the intermediate layers of unprocessed hand and thigh skin [see A and B in Fig. 2(b)] are a superposition of the fluorescence spectrum of melanin itself and the spectrum of the extracellular matrix present in the dermis of the skin. The basal layer of the epidermis is primarily composed of melanocytes, which are responsible for the production of melanin. Melanin is one of the most ubiquitous and biologically important pigments in the human body.⁵⁷ Although melanin is produced by melanocytes, the pigment accumulates in melanosomes which are transferred to the adjacent keratinocytes where they remain as granules. Thus, the melanin is distributed beyond the basal layer and can be found in the entire malpighian layer, a term that can be used to collectively describe the stratum spinosum and granulosum.^58,59 The organization and distribution of the melanin vary between individuals and in darkly pigmented skin, even the corneocytes can contain specks of melanin also described as “dust.” However the number of melanocytes is relatively constant.⁵⁷

Because of the pigmented nature, the cluster spectra show a strong background, potentially due to the fluorescence of the melanin, which is resonant at 785 nm,⁶⁰ although distinct Raman peaks can be clearly identified. As a result, the intensity of the spectrum is also exceptionally high. The presence of a strong background in the spectra does not allow to specifically discriminate the basal layer from the malpighian layer, and KMCA of the Raman map identifies the combination as an intermediate epithelial layer. An increase in the cluster number to 7 or 10 differentiates more “outlier” regions associated with the interface between sample and substrate, indicating that the variations in this region are larger than those in the intermediate layer, dominated by the melanin fluorescence. The strong background in the Raman spectrum makes it difficult to quantify the melanin levels based on Raman spectroscopy. Moreover, these layers, being mostly composed of keratinocytes, albeit at different stages of maturation, are more likely to be spectroscopically very similar.

Most of the Raman bands in the KMCA spectrum of cluster 1 for unprocessed hand and the equivalent cluster for thigh sections are in agreement with results obtained from FT-Raman measurements on isolated SC.^61–64 The SC spectra are dominated by contributions from keratin and lipids. Figure 2(c) represents a comparison between the Raman spectra of the KMCA cluster 1 for unprocessed hand (A) and the equivalent cluster for unprocessed thigh (B) with spectra of the pure lipids ceramide (C), sphingomyelin (D), and $\mathrm{L}-\alpha$-phosphatidylcholine (E). The position of the amide I band at $1655{\mathrm{cm}}^{-1}$ indicates that keratin in the human SC adopts predominantly an $\alpha$-helical conformation.^65,66 Additionally, the spectra of cluster 1 hand and the equivalent for thigh have prominent contributions of lipids, observed at 1064, 1085, and $1130{\mathrm{cm}}^{-1}$ due to chain C–C stretching, 1296 and $1302{\mathrm{cm}}^{-1}$ due to ${\mathrm{CH}}_2$ twisting, as well as 1446 and $1442{\mathrm{cm}}^{-1}$ due to CH scissoring. All bands can be compared to strong features in the Raman spectra of ceramide, sphingomyelin, or phosphatidylcholine. However, the most abundant classes of lipids present in the SC are ceramide, cholesterol, and fatty acids, wherein they play a pivotal role in the skin barrier function. The absence of the intense Raman features at 717, 1157, and $1526{\mathrm{cm}}^{-1}$ clearly highlights that sphingomyelin is absent from the SC. Similarly, the absence of a strong peak at $717{\mathrm{cm}}^{-1}$ in the spectrum of the skin seems to indicate that no contribution from the phospholipids can be observed. However the presence of a small feature at $1747{\mathrm{cm}}^{-1}$ remains identifiable. The exact lipidic composition of the SC remains difficult to evaluate. The presence of sebaceous lipids, composed of squalene, wax esters, and triglycerides, coating the skin surface can be a result of contamination during sample preparation.⁶⁷ However, in recent studies, the SC is commonly described as mostly composed of ceramides, fatty acids, and cholesterol although other lipids may be present in small quantities. Earlier studies indicated that phospholipids in the SC account for about 5% of the total lipids from samples taken from the legs or abdomen but smaller proportions in other locations such as the face.^68–70

The absence of characteristic features of carotenoids in the SC is also notable. The distribution of these powerful antioxidants in the skin was recently investigated in vivo using Raman microscopy based on the prominent Raman line at $1525{\mathrm{cm}}^{-1}$ ($\mathrm{C}=\mathrm{C}$ vibration).^71–73 It is assumed that they are degraded due to the oxidative stress induced by death.

Raman spectroscopy, coupled with KMCA, is clearly a powerful tool to differentiate the layers of skin, as well as layers from different anatomical sites, based on biochemical content. The images reveal significant differences between the morphology and thickness of the SC and basal layer of the two different anatomical sites. Using the false color images reconstructed from the KMCA, the dermis can be easily differentiated due to the high content of collagen, and therefore the delimitation between the dermis and the epidermis is easily discernable. The two remaining clusters are attributed to the SC and an intermediate combination of the basal and malpighian layers. Across the sections of unprocessed hand examined, the total thickness of the epidermis appears to be about $70\pm 15\mathrm{\mu m}$ whereas in the thigh sections it is within the range $35\pm 5\mathrm{\mu m}$. Although the thicknesses vary somewhat from sample to sample, these differences between sections of hand and thigh are consistent across samples from all 11 cadavers measured. The cluster identified as the top layer of the skin without any features of the melanin can be attributed to the SC, and was found to be respectively $40\pm 10\mathrm{\mu m}$ for the unprocessed hand and $25\pm 5\mathrm{\mu m}$ for the thigh.

Epidermal thickness is of considerable significance in dermatological research and a considerable amount of work has been done and reported, both in vivo^74,75 and ex vivo from biopsy samples,^76–78 defining the variations of epidermal thickness of different anatomical sites. The observations here are consistent with those expected for primarily sun-exposed sites (hand) and sun-protected sites (thigh). It has also been observed that the degree of variability on normally clothed body sites is less than that on the normally unclothed sites.⁷⁹ Variations in normal skin related to age and gender have also been reported.⁷⁷ Published values of the thickness of the viable layer (basal layer) and the SC have been variable. Moreover, the thickness of SC and other layers of the epidermis are known to be different from different anatomical sites. In recent studies, the total thickness has been found to be between 60 and 100 μm depending on the body sites considered, but more importantly the techniques employed for the measurements.^80,81 However, less is known about the biochemical differences, and Raman spectroscopic analysis can potentially shed further light on chemical differences between the skin layers, and between layers from different anatomical sites. Important for generalization of deductions is the consistency of the differentiation between specimens. Figure 3(a) and 3(b) compares the average spectra of the epidermis of unprocessed hand and thigh sections from different anatomical sites. The results are representative of all 11 cadaver measured samples.

Fig. 3

(a) A: Average Raman spectrum of the unprocessed hand SC; B: Average Raman spectrum of the dewaxed hand SC; C: Average Raman spectrum of the unprocessed thigh SC; D: Average Raman spectrum of the dewaxed thigh SC. (b) A: Average Raman spectrum of the unprocessed hand dermis; B: Average Raman spectrum of the dewaxed hand dermis; C: Average Raman spectrum of the unprocessed thigh dermis; D: Average Raman spectrum of the dewaxed thigh dermis.

The samples available for this study have been taken post-mortem from patients over 80 years (Table 1). The thickness of the different skin layers can be affected by aging or sun exposure and therefore can vary significantly from individual to individual. The observations made can thus not be considered to be representative of the whole population. However, as the samples have been taken from different locations (hand and thigh) for each cadaver, direct comparison between two samples from the same patient remains relevant.

Spectra of the SC and dermis of different individuals show strong similarities, but some structural and conformational differences are apparent, as illustrated in Fig. 3(a) and 3(b). Overall, the Raman signatures vary significantly and consistently according to body site (hand and thigh). Apparent shifts in the features at $1446/1442$ and $1296/1302{\mathrm{cm}}^{-1}$ are consistently observed from thigh to hand SC (Fig. 3). The relative intensity of the signal at 1296 and $1302{\mathrm{cm}}^{-1}$ with respect to the peak at $1266{\mathrm{cm}}^{-1}$ also varies. Similar differences are observed between unprocessed hand and thigh sections in the case of the intermediate epithelial layer [Fig. 2(b)] and the dermis [Fig. 3(b)]. The band located at $1302{\mathrm{cm}}^{-1}$ in thigh dermis [Fig. 3(b)] is assigned to ${\mathrm{CH}}_2$ twisting, and this is absent in hand dermis, while the Amide III band located at $1246{\mathrm{cm}}^{-1}$ in hand dermis is absent in thigh dermis. Figure 3 illustrates that these are regions of strong lipid contributions, however, and so, rather than spectral shifts, these characteristic differences are more likely due to differing contributions of lipids to the spectra relative to those of proteins.

Thigh SC and dermis also display a small peak at $1747{\mathrm{cm}}^{-1}$, which is absent from hand SC and dermis. This peak is evident in the spectrum of $\mathrm{L}\text{-}\alpha$-phosphatidylcholine [see E in Fig. 2(c)] and derives from the $\mathrm{C}=\mathrm{O}$ ester vibration, also present in triglycerides, which are not normally present in the SC. Triglycerides are esters derived from glycerol and three fatty acids, and their prominence in the thigh tissue may be evidence of higher levels of lipids and fatty acids.

3.4.

Study of the CH Region

Differences in lipidic content of the different layers can be more clearly visualized in the high wavenumber region of the Raman spectrum, in which the CH vibrations of the aliphatic chains feature strongly. Figure 4 presents a comparison between the average, baseline corrected, Raman spectra of unprocessed SC (A), unprocessed intermediate layer (B), unprocessed dermis (C) in the region from $\sim 2700$ to $\sim 3200{\mathrm{cm}}^{-1}$, compared with pure ceramide (D), sphingomyline (E), and $\mathrm{L}\text{-}\alpha$-phosphatidylcholine (F). The strong contributions of the lipids can be seen clearly in the spectra of all three layers of the frozen skin section, although their relative contributions vary. The spectra can be fitted with individual Gaussian/Lorentzian bands, and the results are summarized in Table 2. Spectral positions of the fitted peaks vary somewhat from spectrum to spectrum, and so are quoted $\pm 3{\mathrm{cm}}^{-1}$. Fitted peak intensities are reproducible within 10% and are normalized to the band at $\sim 2932{\mathrm{cm}}^{-1}$. The peaks in the high number region predominantly originate from CH, ${\mathrm{CH}}_2$, and ${\mathrm{CH}}_3$ stretching, and from literature, many can be assigned to lipids, the remainder being assigned to proteins.^34,35,82,83

Fig. 4

Average Raman spectra of high wavenumber region of unprocessed thigh SC (A), unprocessed intermediate epithelial layer (B), unprocessed dermis (C), ceramide (D), sphingomyline (E), and $\mathrm{L}\text{-}\alpha$-phosphatidylcholine (F).

Table 2

Fitted peak positions, relative intensities and assignments for the high wavenumber region of average Raman spectra of unprocessed skin sections.

In general, the SC is seen to be relatively rich in lipidic content, essential for the barrier function of the skin against external agents. The SC is made up of keratinized cells that are embedded in lipid matrices like bricks in mortar,⁸⁴ or more specifically corneocyte cells surrounded by a three-dimensional, multi-lamellar lipid domain.^85,86 As part of the cell maturation, the keratinocytes’ enzymes degrade the cellular viable components such as nucleus and other organelles. Therefore, corneocytes are anucleated cells, and no DNA or histones should be found in this layer of the skin. The major lipid components of the SC are ceramides (sphingosines and phytosphingosines), long-chain free fatty acids, and cholesterol.^87–90

The main lipids in the basal layer are phospholipids, cholesterol, and to a lesser extent triglycerides, which provide energy for metabolism. The lipids are subject to maturation, and it starts with the apparition of lamellar bodies in the stratum spinosum which contain phospholipids, sphingolipids, and cholesterol.⁹¹ The lamellar granules present in the stratum granulosum are particularly rich in glucosylceramides. These granules are released in the intercellular space of the SC where they will be converted into ceramides via hydrolysis by beta (b)-glucocerebrosidase.^92,93

Overall, it can be deduced that lipid/protein compositions are not uniform in the layers of hand and thigh of the same human cadavers. The sun-exposed hand skin is relatively low in lipidic content compared to the thigh. This is consistent with reports that a decrease in lipid content is associated with increased susceptibility to exogenous insults, which will naturally increase with age.⁹⁴ The observations may provide unique advantages in skin disease diagnosis, and this body-site difference needs to be factored into in vivo skin Raman assessment and disease diagnosis. However, it is noted that for both anatomical sites, the Raman spectra of the SC and dermis of skin sections are strongly influenced by the lipidic and extracellular protein content, rather than the cellular features, which may provide the most direct biochemical information on tissue pathology. Raman spectroscopy can be used, not only to elucidate the structure of the lipids involved in the SC barrier function (ceramides, cholesterol, and free fatty acids), but also to provide a direct insight on the conformational order and the lateral packing of these lipids. It has been shown that the barrier function is directly related to the compactness of the lipid structure and that the later can be affected by external insults such as UV radiation.^35,95–97

3.5.

Comparison with Processed FFPP Sections

For histological analysis, tissue samples are commonly preserved in paraffin wax.^36,37 Paraffin embedding facilitates tissue cutting, but also is commonly employed, worldwide, for archiving tissue samples. The availability of a wide range of pathologically characterized samples for study potentially enables extensive retrospective studies using spectroscopic and other techniques. However, the paraffin wax itself gives rise to strong Raman signals that overlap the molecular vibrations of the biomolecules of the samples, necessitating chemical removal for spectral analysis.³⁸ Although it has been demonstrated that the contributions can be removed digitally,⁹⁸ histological staining, considered a gold standard, necessitates chemical removal of the wax, and thus the procedure is employed here. Furthermore, although it has been demonstrated that dewaxing using hexane is more efficient than the commonly employed xylene, xylene is used here for consistency with clinical protocols.³⁸

Figure 5(a) shows an optical microscopic image of the dewaxed hand section of human skin of 20 μm thickness. As for the unprocessed sections, three different regions are visually apparent, the dermis at the bottom and the SC at the top, separated by an intermediate underlying epidermal layer. In contrast, the optical images of dewaxed hand and thigh skin sections in Fig. 5(a) and 5(d) indicate no major differences between the thicknesses of the total epidermis, which are $\sim 40\mu \text{m}$ in both thigh and hand. The SC also appears to have similar thicknesses in both locations, $\sim 20\mathrm{\mu m}$, despite the large differences observable in the unprocessed sections from the same cadavers. Through the processes of formalin-fixation, paraffin-embedding, and dewaxing, as well as cutting, the tissue sections likely undergo a considerable degree of chemical and physical changes, and it is important to consider the effects of the processing on the biochemical content and structure of the skin layers.

Fig. 5

(a) Optical image of dewaxed hand tissue section; (b) KMCA of Raman spectral map; (c) KMCA mean Raman spectra of processed hand illustrating the differentiation of the superficial SC (cluster $1=\mathrm{A}$), the intermediate epithelium (cluster $5=\mathrm{B}$), and the dermis (cluster $2=\mathrm{C}$); (d) Optical image of processed thigh tissue section; (e) KMCA of Raman map; (f) KMCA mean Raman spectra of processed thigh, illustrating the differentiation of the superficial SC (cluster $3=\mathrm{A}$), the intermediate epithelium (cluster $5=\mathrm{B}$), and the dermis (cluster $4=\mathrm{C}$).

Spectroscopically, the three regions of the sections are quite distinguishable, as illustrated by the KMCA mean spectra of Fig. 5(c), taken within the subsection of Fig. 5(a) indicated by the white rectangle. The KMCA map of Fig. 5(b) further supports classification of dermis, basal/malpighian layer, and SC. The mean spectra of the principle regions are plotted in Fig. 5(c). The mean spectrum of cluster 4 is similar to that of cluster 1. Similar results are observed for dewaxed thigh tissue sections and are shown in Fig. 5(d), 5(e), and 5(f). For both anatomical sites, at first glance, the mean spectra of the three identified regions are similar to the corresponding regions for the unprocessed skin sections, but closer examination reveals important differences.

Figure 6(a) presents a comparison between the spectra of KMCA cluster 1 of dewaxed skin hand SC (A) and the equivalent cluster of dewaxed thigh SC (B) with that of the HaCaT cell line (C), DNA (D), histone (D), and pure paraffin wax (F). Firstly, it is noted that the KMCA spectra of dewaxed epidermis of hand and thigh reveal little or no contributions from the paraffin wax, indicating that the dewaxing procedure has been effective. The characteristic bands of wax [see F in Fig. 6(a)] at 1062, 1131, 1296, and $1441{\mathrm{cm}}^{-1}$ are completely absent in the mean spectra of the epidermis of the dewaxed skin hand and thigh [see A and B in Fig. 6(a)]. Hence, the wax contribution of the epidermis has been completely removed by the dewaxing procedure. Notably, however, the mean spectra of dewaxed hand and thigh dermis are devoid of lipidic contributions. Both paraffin wax and biological lipids are long chain aliphatic molecules. The spectra of wax and pure lipids show remarkable resemblances [see F in Fig. 6(a); see C, D, E in Fig. 2(c)]. The Raman contributions of pure lipids at 1064, 1296, and $1442{\mathrm{cm}}^{-1}$ in cluster 1 of unprocessed skin hand and thigh [see A and B in Fig. 2(a)] are almost identical to those of pure wax in the wavenumber region of 1000 to $1700{\mathrm{cm}}^{-1}$. The results indicate, therefore, that while the dewaxing procedure is successful in completely removing the wax in the tissue, due to the molecular similarity, the process is also effective in the complete removal of the naturally occurring tissue lipids.

Fig. 6

(a) KMCA mean spectra of cluster 1 from dewaxed skin hand (A) cluster 3 from dewaxed thigh (B), Raman spectra of HaCaT cells (C), DNA (D), histone (E), and pure paraffin wax (F), (b) KMCA mean Raman spectra of cluster 5 from dewaxed skin hand (A), cluster 5 from dewaxed thigh (B), and spectrum of pure melanin (C), (c) KMCA mean Raman spectra of cluster 2 dewaxed skin hand (A) cluster 4 dewaxed thigh (B) pure collagen (C).

Despite minor differences, the spectra of the clusters corresponding to the dewaxed SC of hand and thigh [see A and B in Fig. 6(a)] show remarkable resemblances. Protein bands, such as those at 1450 and $1340{\mathrm{cm}}^{-1}$ due to protein CH deformation, $1255{\mathrm{cm}}^{-1}$ due to protein Amide III, and $756{\mathrm{cm}}^{-1}$ due to tryptophan, ring breathing modes are also present in the spectra of HaCaT cells [see C in Fig. 6(a)]. Notably, however, the signature bands of nucleic acids, visible in the spectrum of the HaCaT cells at 785 and $1580{\mathrm{cm}}^{-1}$, are absent in the mean spectrum of the SC.

Although the SC is rich in lipidic components, after dewaxing it is clear that most of the lipids have been lost, and the main molecules contributing to the spectra recorded are the proteins, contained within the corneocytes. Although the lipids of the SC are widely investigated and often considered as the most important component of the skin, the proteins are also subject to a certain degree of reorganization during the maturation of the keratinocytes from the deepest layers until they reach the surface of the skin. The large, insoluble protein named profilaggrin will be converted to filaggrin which is associated to the keratin filament in the deepest layers of the SC. The proteolysis of filaggrin converts it to its constituent amino acids, and amino-acid derivatives.^70,99 The final form is known as natural moisture factor (NMF), and it can represent up to 10% of the corneocyte dry weight. The complex composition of the NMF mixture can explain the presence of numerous features of the spectrum recorded from the SC. The pattern of the Raman spectra presented in Fig. 6(a) is consistent with data that can be found in the literature focusing on the study of the corneocyte maturation.⁹⁰ Moreover, the protein features are highlighted in the absence of the normally strong lipidic spectral features.

Figure 6(b) shows a comparison between cluster 5 of dewaxed hand skin and the equivalent cluster for thigh skin with pure melanin. As described in Sec. 3.3, both clusters [see A and B in Fig. 6(b)] can be attributed to a combination of the basal and malpighian layers of the human skin, both rich in melanin. Again, the intensities of the spectra are exceptionally high, and the spectra are dominated by the fluorescence spectrum of melanin. The spectra are similar to the corresponding spectra from the unprocessed tissue sections [see A and B in Fig. 2(b)]. However, no lipidic peaks are observable and, notably, no differences between the spectra of cluster 5 of dewaxed hand and the equivalent thigh spectrum are apparent [see A and B in Fig. 6(b)]. The process of waxing and dewaxing therefore substantially reduces the lipidic content of the intermediate epithelial layers, such that intermediate epithelial layers from different anatomical sites such as hand and thigh cannot be distinguished from each other.

Figure 6(c) presents a comparison between the KMCA spectra of cluster 2 of dewaxed hand dermis and the equivalent cluster for thigh [see A and B in Fig. 6(c)] with pure collagen [see C in Fig. 6(c)]. As is the case with the epidermis layers, no lipidic peaks are observed. In contrast to the spectra of dermal clusters of unprocessed hand and thigh [see A and B in Fig. 2(a)], the spectra look identical. The process of waxing and dewaxing appears to have removed all lipidic constituents, and therefore the dermis from different anatomical sites such as hand and thigh cannot be distinguished.

The Raman spectra of the stratrum corneum and dermis of the dewaxed skin sections from different anatomical regions and from different human cadavers are shown in Fig. 3(a) and 3(b). Remarkably, comparing the Raman spectra of dewaxed epidermis and dermis of human skin from three different cadavers, no significant differences can be found. As in the case of unprocessed sections, the results are representative of all 11 cadavers measured.

To confirm the depeletion of the lipidic content of the tissue, Fig. 7 shows the high wavenumber regions of the spectra of dewaxed human thigh SC, intermediate epithelial layer, and dermis. As for the unprocessed sections, the spectra were fitted with a series of Gaussian/Lorentzian bands, and the spectral positioning ($\pm 3{\mathrm{cm}}^{-1}$), intensities ($\pm 10\%$), and assignments are listed in Table 3.

Fig. 7

Average Raman spectra of the high wavenumber region of dewaxed human thigh section from SC (A), intermediate epithelial layer (B), dermis (C), and pure collagen (D).

Table 3

Fitted peak positions, intensities and assignments for high wavenumber region of average Raman spectra of dewaxed thigh skin section.

Comparison of the relative intensities of the lipid peaks of unprocessed and processed tissue sections of hand and thigh in Tables 2 and 3 confirms that the dewaxing procedure has significantly reduced the lipidic content and altered the tissue composition of the skin. Strikingly, however, the remaining biochemical structure of the skin layers is remarkably consistent across different cadavers and anatomical sites.

The lipidic content is the key to the permeability barrier function of skin and is of significant importance to the cosmetics industry and for transdermal drug delivery. Abnormalities in barrier function associated with lipid content are also associated with atopic dermatitis and other common cutaneous diseases.¹⁰⁰ UV radiation has been shown to effect the free fatty acid and triglyceride composition,¹⁰¹ and decrease in lipid content with age is associated with increased susceptibility to exogenous insults.⁹⁴ Therefore, a detailed knowledge of lipidic content, composition, and structure is critical to many studies of skin function, malfunction, and abnormality. Although archived tissue samples can serve as an important source for retrospective studies, this study demonstrates that, in their dewaxed form, no information related to their lipid content can be gleaned. Nevertheless, the study shows that the cellular content of the tissue and the protein extracellular matrix remain largely intact, and that the spectral profiles are enhanced by the removal of the lipidic content.