2.1.

Patients

This study was conducted in agreement with the Helsinki declaration and performed under the guidelines of the research ethics committee of the ‘‘Centre de Recherche du Centre Hospitalier Universitaire de Québec.” All patients were given adequate information to provide written consent. Pathologic samples came from ex vivo psoriatic plaques from three patients deceased from heart attack or respiratory failure (thereafter identified as PHSI, PHSII, and PHSIII; 56 yr-old male, 48 yr-old female, and 73 yr-old female, respectively). Normal samples came from several women’s breast reduction surgeries (thereafter identified as NHSI, NHSII, NHSIII, NHSIV, and NHSV; 37, 46, 39, 44, and 48 yr-old females, respectively). The information about the patients is summarized in Table 1. Biopsies were cut in small pieces and then embedded in an optimum cutting temperature compound (Tissue-Tek* O.C.T. Compound, Sakura* Finetek) and frozen for further analysis.

Table 1

Information about the patients.

2.2.

Histological Analyses

Five-micrometer-thick cryosections were cut and stained using Masson’s Trichrome. The staining allowed distinguishing the different layers of the samples (SC, LE, and dermis).

2.3.

ATR-IR analyses

IR spectra of the SC of full-thickness skin samples were obtained using a golden gate single reflection ATR system (Specac, Pleasantville, New York) fitted with a diamond crystal. This method probes the first micrometers of the surface of the sample that is in contact with the ATR crystal. Spectra were recorded with a Nicolet Magna 850 Fourier transform spectrometer (Thermo-Nicolet, Madison, Wisconsin) equipped with a narrow-band mercury–cadmium–telluride detector and a germanium-coated KBr beam splitter. At each temperature, a total of 128 interferograms were acquired, coadded and Fourier transformed using a Happ–Genzel apodization function, to give a spectral resolution of $4{\mathrm{cm}}^{-1}$ in the spectral range 4000 to $750{\mathrm{cm}}^{-1}$. All data were processed with Grams 8.0 software (Galactic Industries Corp., Salem, Massachusetts). The spectral region corresponding to the ${\mathrm{CH}}_2$ stretching vibrations was baseline-corrected using a cubic function, and the peak position was determined using the center of gravity at the top 10% of the bands.

2.4.

Infrared microspectroscopy analyses

Ten micrometer thick cryosections of the samples were placed on reflective-coated microscope slides (Low-e MirrIR, Kevley Technologies, Chesterland, Ohio). IR microscopy images were recorded using an Agilent 620 IR microscope equipped with a liquid-nitrogen cooled $32\times 32$ focal plane array detector and a motorized stage. Each of these 1024 detector elements measured $5.5\mu \mathrm{m}\times 5.5\mu \mathrm{m}$, which corresponds to the pixel size in the IR images. This microscope assembly was attached to an Agilent 660 spectrometer. Images were recorded by stacking individual 1024 pixel images recorded by the displacement of the microscope motorized stage. All images were acquired in reflectance mode with each spectrum originating from a detector element being recorded at a $4{\mathrm{cm}}^{-1}$ spectral resolution and a 4000 to $1000{\mathrm{cm}}^{-1}$ spectral range. For each detector element, 128 interferograms were typically coadded, Fourier transformed, and ratioed over a background recorded from a blank region of the slide. All images were computed using Resolutions Pro software (version 5.2.0.846) while spectra were analyzed using Grams 9.1 (Thermo Fisher, Waltham, Massachusetts). All spectra are presented without any normalization procedure. The region of the spectra between 3150 and $2750{\mathrm{cm}}^{-1}$ was baseline corrected using a cubic function.

2.5.

Raman Microspectroscopy Analyses

The cryotomed transverse $20\mu \mathrm{m}$-thick sections of the samples were put on glass microscope slides. All sections were then analyzed without any further modification.

Raman spectra were recorded using a LabRam HR-800 spectrometer (Horiba Jobin-Yvon, Villeneuve d’Ascq, France) coupled to an Olympus BX-30 fixed stage microscope, equipped with an open electrode Peltier-cooled CCD detector ($1024\times 256$ pixels) (Andor Technology, Belfast, Northern Ireland). An MPlan $100\times$ objective (0.90 NA) (Olympus) and an internal He–Ne laser emitting at 633 nm ($4\text{-}{\mu \mathrm{m}}^2$ illumination spot) were used for the data collection. The confocal hole and the entrance slit of the monochromator ($600\text{lines}/\mathrm{mm}$ grating) were fixed at 400 and $100\mu \mathrm{m}$, respectively. All spectra were the result of three acquisitions of 60 s each. Spectra were recorded for two wavelength windows that are 700 to $1800{\mathrm{cm}}^{-1}$ (fingerprint region) and 2500 to $3200{\mathrm{cm}}^{-1}$ (high-frequency region).

Ten spectra were recorded in each characteristic layer of the samples (SC, LE, and dermis) and were analyzed using GRAMS/AI 8.0 (Thermo Galactic, Salem, New Hampshire). A cubic function was used on each spectral region for the baseline correction to get rid of the fluorescence background. All spectra were normalized according to different criteria as a function of the spectral feature of interest: in the fingerprint region, spectra were normalized with respect to the amide I band ($1650{\mathrm{cm}}^{-1}$) intensity while in the high-frequency region, spectra were normalized using the intensity of the band located at $2930{\mathrm{cm}}^{-1}$.

2.6.

Sample Variability

For a given psoriatic sample (PHSI, PHSII, and PHSIII), three sections per sample (three sections for PHSI, three sections for PHSII, and three sections for PHSIII) were analyzed in IR and Raman microspectroscopies. For a given normal skin sample, in IR microspectroscopy, one section per sample was analyzed (NHSI, NHSII, and NHSIII); in Raman microspectroscopy, three or more sections per sample were analyzed (three sections for patient NHSIII, five sections for patient NHSIV, and four sections for patient NHSV). Information about the use of the samples is summarized in Table 2. The sections were not taken consecutively, but rather with a random distance between them (between 10 and $100\mu \mathrm{m}$). Despite the natural variability observed in the molecular structure and composition that occurs in biological samples, our data (not shown) demonstrate that the trends in IR and Raman spectra of NHS are similar throughout all investigated samples in terms of peak occurrence, position and shape. By contrast, PHS samples lead to various types of spectra from sample to sample, thus exhibiting less clear general trends. This spectral disparity most likely arises due to the variability in terms of type of psoriasis, medical history of the patient (intake of medicine or topical treatment application), donor site, and so on.

Table 2

Use of the samples.

3.1.

Histological Analysis

Figure 1 displays histology images of NHSI and PHSI. The histology image of NHS shows three well defined characteristic layers, i.e., the dermis in blue, the LE in dark red, and the SC in light red. The boundaries between the different layers are sharp and regular and at this scale, the layers seem globally homogeneous. In PHS, it is still possible to distinguish three layers: dermis in blue, LE in purple, and SC in dark red. However, the layer boundaries are somewhat blurred. Moreover, some inclusions seem to be embedded in the LE (an example is evidenced by the white arrow). These inclusions seem to contain some collagen fibers and few cells. According to the literature, these inclusions could be attributed to edema.³³ The overall structure of PHS appears to be less organized than that of NHS.

Fig. 1

Histology images of: (a) normal human skin (NHSI) and (b) psoriatic human skin (PHSI) ($\text{scale bar}=100\mu \mathrm{m}$)

3.2.

ATR-IR

The lipid chain order of the SC as a function of temperature measured through ATR-IR spectroscopy is displayed in Figure 2. This figure shows the temperature dependence of the frequency of the ${\mathrm{CH}}_2$ symmetric stretching mode associated with the lipid chains, measured in the temperature range 28°C to 80°C for both NHS and PHS samples. It is well-known that the position of the band attributed to the symmetric ${\mathrm{CH}}_2$ stretching mode is sensitive to the lipid chain order. More specifically, it allows distinguishing lipid molecule packing from lipid chain conformational changes. In fact, lipid molecule packing geometry alterations (usually arising from solid–solid phase transitions) are inferred if cooperative changes appear below $\sim 2850{\mathrm{cm}}^{-1}$, while the introduction of gauche conformers into the lipid acyl chains appears at higher frequency values.⁹ Then higher frequencies of the ${\mathrm{CH}}_2$ stretching bands are generally characteristic of a high content of gauche conformers (disordered state), while bands at lower frequencies are associated with the predominance of trans conformers (organized state).

Fig. 2

Thermotropism curves of the stratum corneum of the three PHS samples (I; squares, II; triangles, III; diamonds) compared to NHS control (circles; after³⁴).

For NHS, a transition is observed between $\sim 35°\mathrm{C}$ and 45°C, and is assigned to a solid–solid (orthorhombic → hexagonal) phase interconversion.^9,24,25 The frequency of the symmetric ${\mathrm{CH}}_2$ stretching mode then keeps rising with temperature showing that the lipid chains become less ordered. In the case of PHS, the position of the symmetric ${\mathrm{CH}}_2$ stretching mode is already higher than $2850{\mathrm{cm}}^{-1}$ at 28°C, indicating that lipid chains in PHS are in a disordered state near room temperature. The frequency of the symmetric ${\mathrm{CH}}_2$ stretching mode also increases with temperature, showing that the lipid chains become more disordered. For the whole range of temperatures investigated, the frequency of the ${\mathrm{CH}}_2$ symmetric stretching mode is always higher than that of the NHS indicating that overall, the lipid chains of the SC of PHS are more disordered than those of the SC of NHS whatever the temperature.

3.3.

Infrared microspectroscopy

Figure 3 displays (a) optical and IR images of PHSI that were computed to highlight the presence of (b) lipids, (c) collagen, and (d) lipid-to-protein ratio. Figure 3(b) is generated by representing, on a color scale, the area under the peak centered at $\sim 2850{\mathrm{cm}}^{-1}$ (between the two minima surrounding the peak). This image indicates that the lipids are mainly present in the SC. The lipid map distribution also allows distinguishing the LE and the dermis, showing that some lipids are present in the LE whereas they are almost absent in the dermis. The comparison of typical spectra of the SC of PHS and NHS [Figs. 3(e) and 3(f)] further shows that the ${\mathrm{CH}}_2$ stretching region of NHS is dominated by lipids (bands at 2850 and $2920{\mathrm{cm}}^{-1}$) whereas bands due to proteins are more present for PHS (bands at 2870 and $2950{\mathrm{cm}}^{-1}$). Therefore, the SC of the PHS seems to contain fewer lipids as compared to NHS. This is further confirmed by Fig. 3(b) which displays a poorly defined lipid-rich SC layer with no clear discontinuity with the LE. This is clearly different from the lipid distribution in NHS.³¹

Fig. 3

(a) Microscopic optical view of a PHSI section ($15\times$ objective) and corresponding infrared (IR) images showing (b) the lipid distribution, (c) the collagen distribution, and (d) the protein-to-lipid ratio ($1\text{pixel}=5.5\mu \mathrm{m}$). The red dotted line indicates a high content/ratio, whereas the green solid line color indicates a low content/ratio. (e) Superimposition of spectra of the stratum corneum for NHSI and PHSI, and (f) ${\mathrm{CH}}_2$ stretching mode region.

The image generated in Fig. 3(c), computed from the integrated area of the $1342{\mathrm{cm}}^{-1}$ feature of collagen (${\mathrm{CH}}_2$ and ${\mathrm{CH}}_3$ wagging and deformation and CN stretching mode), allows localizing collagen in the dermis, thus creating a clear separation between this protein-rich layer on one hand, and the SC and LE on the other hand. In addition, Fig. 3(d) displays the protein-to-lipid content throughout the sample as calculated from the ratio of the integrated area of the amide I peak (centered at $1650{\mathrm{cm}}^{-1}$, between the two minima surrounding the peak) to that of the $2850{\mathrm{cm}}^{-1}$ peak. This image shows that the dermis is a protein-rich layer. It is noteworthy that the images of lipids, collagen, and protein-to-lipid ratio distributions are similar for all of the PHS samples (I, II, and III), showing interpenetration between layers. Some of the PHS images (Fig. 4) allowed visualizing the elongation of dermal papillae. By contrast, NHS images display well-separated layers.³¹

Fig. 4

(i) Microscopic optical views ($15\times$ objective) and corresponding IR images showing (ii) the lipid distribution, (iii) the collagen distribution, and (iv) the protein-to-lipid ratio ($1\text{pixel}=5.5\mu \mathrm{m}$) for: (a) PHSI (b) PHSII, and (c) PHSIII.

The inclusions embedded in the LE of PHS are easily distinguishable in these images [black arrows, Figs. 3(b)–3(d)]. The core of these inclusions contains some lipids (but in fewer amounts with respect to LE) and their periphery is rich in proteins. Since this ring-shaped region seems to be rich in collagen, it is consistent with the fact that the inclusion could be edema. As a matter of fact, edemas are part of the papillar layer of the dermis and can detach and migrate from it.³³ The structure of these inclusions needs to be further investigated.

Figure 5 compares the lipid order maps of NHS and PHS as plotted from the position of the center of gravity of the peak corresponding to the methylene symmetric stretching mode, at approximately $2850{\mathrm{cm}}^{-1}$. Figure 5(b) reveals that the frequency of this mode is significantly lower in the SC relative to the LE for NHS, indicating that the lipid acyl chains are fairly more ordered in the SC layer, in agreement with previously published data.³¹ The images obtained from either of the PHS samples [example provided in Fig. 5(d)] do not allow consistent distinguishing of the SC from the LE with a lower definition of the skin layer boundaries in terms of lipid organization.

Fig. 5

Microscopic view ($15\times$ objective) of: (a) NHSI and (c) PHSI, and the corresponding IR images across skin sections: (b) NHSI and (d) PHSI, computed from the position of the center of gravity of the peak corresponding to the ${\mathrm{CH}}_2$ symmetric stretching mode [color scale from low (green) to high (red) frequencies].

Focusing on the SC, three random spectra per analyzed sections of PHS samples (I, II, and III) and NHS (I, II, and III) have been taken in the SC. The frequency of the methylene symmetric stretching mode has been measured in each spectrum. The results, summarized in Table 3, clearly evidence that higher frequencies (by about $2{\mathrm{cm}}^{-1}$) are measured in psoriatic samples. This result shows that lipids are less ordered in psoriatic skin, in agreement with the ATR data.

Table 3

Average frequency of the peak corresponding to the CH2 symmetric stretching mode for normal and psoriatic human skin samples.

3.4.

Raman Microspectroscopy

Further characterization of the lipid distribution and organization was performed using Raman microspectroscopy. Figure 6(a) displays a microscopic view of a section of PHSII. For comparison, a histological image is presented in Fig. 6(b). Red dots in Fig. 6(a) highlight locations where spectra were recorded. In the case of NHS, spectra taken in a given skin layer (either SC, LE or dermis) were similar to each other in terms of peak occurrence and intensities and can be easily coadded to generate an averaged spectrum for this layer. However, in the case of PHS, the spectra recorded from a given layer exhibit several different patterns, showing that the skin is heterogeneous. Spectra had then to be averaged per group of similar spectra. For instance, some of the spectra recorded from the SC of PHS are similar to those obtained for NHS. Consequently, this series of spectra were averaged and the resulting spectrum is referred as to group 1. However, other spectra of SC of PHS display very similar features to each other, although being significantly different from those of the SC of NHS, and are identified as being part of group 2. Finally, other spectra cannot be grouped together because each of them displays its own spectral profile. These latter spectra will not be discussed any further.

Fig. 6

(a) Optical image of a PHSII section ($10\times$ objective). The red dots correspond to the locations where the Raman spectra were recorded prior being averaged (scale bar $20\mu \mathrm{m}$); (b) histological image of a PHSII section (scale bar $100\mu \mathrm{m}$); and (c) Raman spectra of the stratum corneum for NHSIII (see text for details) and PHSII-group 1 in the fingerprint (left) and $\mathrm{C}─\mathrm{H}$ stretching (right) regions.

The averaged spectrum from group 1 is shown in Fig. 6(c) as compared with NHS in the fingerprint and high-frequency regions. It exhibits a sharp band at $1296{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$ twisting mode) which is characteristic of the presence of lipids. However, in the case of PHS, the intensity of this peak is lower and broader than the one of the NHS spectrum [Fig. 6(c)], suggesting that there are fewer lipids in PHS than in NHS. The broadening of this peak further indicates that the lipids are less ordered in PHS. This is confirmed by the ${\mathrm{CH}}_2$ asymmetric stretching mode at $2880{\mathrm{cm}}^{-1}$, which appears as a shoulder in the case of PHS whereas it is sharp and more intense in the case of NHS. These Raman data overall support the IR observations.

Previous works demonstrated the usefulness of the frequency domain ranging from 1000 to $1150{\mathrm{cm}}^{-1}$ to study lipid chain order. Indeed, the vibrational modes at 1060 and $1130{\mathrm{cm}}^{-1}$ correspond to the $\mathrm{C}─\mathrm{C}$ skeletal stretching vibration of lipids in trans conformation (ordered phase), whereas the vibrational mode at $1085{\mathrm{cm}}^{-1}$ is due to the $\mathrm{C}─\mathrm{C}$ stretching vibration of lipids in gauche conformation (disordered phase).²⁶ The lipid packing can thus be semiquantitatively evaluated using the $I_{1060}/I_{1080}$ and $I_{1130}/I_{1080}$ intensity ratios.³² These ratios are presented for both NHS and PHS (from averaged spectra from group 1 of PHSI, PHSII, and PHSIII) in Fig. 7. On average, the values of both ratios are lower in PHS than in NHS, confirming that the lipid chains of the SC are less ordered for PHS than for NHS.

Fig. 7

Histograms of the (a) $I_{1130/I1085}$ and (b) $I_{1060/I1085}$ ratios for the different specimens.

Figure 8(a) displays the superimposition of group 1 and group 2 averaged Raman spectra taken from the SC of a section of PHSIII. As previously mentioned, the red (group 1) spectrum is similar to the averaged spectrum recorded from the SC of NHS in terms of peak occurrence and overall profile. However, the spectrum of group 2 exhibits particular features, such as lower intensities of the bands at 1003 and $1342{\mathrm{cm}}^{-1}$ due to proteins (phenylalanine residue and skeletal protein vibrations, respectively³⁵) with respect to the lipid band at $1296{\mathrm{cm}}^{-1}$, illustrating that the protein content decreased markedly as compared to lipids.³⁶ This is further supported by the intensity increase of the peak at $2850{\mathrm{cm}}^{-1}$ and the neighboring broad features at $\sim 2900$ to $2860{\mathrm{cm}}^{-1}$ that are due to ${\mathrm{CH}}_2$ stretching modes of lipids [Fig. 8(a), inset].

Fig. 8

(a) Superimposition of group 1 and group 2 averaged Raman spectra from the stratum corneum of a single section of PHSIII—the inset shows the $\mathrm{C}─\mathrm{H}$ stretching region in detail, (b) superimposition of two types of IR spectrum identified in the stratum corneum of a single section of PHSIII, (c) third type of IR spectrum identified in stratum corneum of a single section of PHSIII, and (d) schematic representation of the organization of the stratum corneum of PHSIII (top, the grid represents IR pixels) as deduced from microscopic Raman-IR (bottom, squares represent IR pixels) and optical¹² observations.

Performing a pixel-by-pixel analysis of the IR images previously recorded further supports the heterogeneous nature of the PHS samples. As a matter of fact, IR spectra recorded from similar locations as Raman spectra from group 1 and group 2 [Fig. 8(a) and 8(b)] also display very distinct features. Again, the typical IR spectrum from group 1 exhibits features resembling those of the SC of NHS while the IR protein and lipid spectral contributions observed in spectra from group 2 behave similarly to what was observed in their Raman counterpart [e.g., decrease of the intensities of the protein amide A at $\sim 3250{\mathrm{cm}}^{-1}$, amide I band at $\sim 1650{\mathrm{cm}}^{-1}$, and amide II band at $\sim 1550{\mathrm{cm}}^{-1}$, and increase of the absorbance of the lipid spectral contribution at $\sim 2850{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$ stretching) and $\sim 1744{\mathrm{cm}}^{-1}$ ($\mathrm{C}═\mathrm{O}$ stretching mode)]. It is noteworthy that such spectra did not originate from isolated pixels but were rather observable from several neighboring pixels covering an area of about 500 to $1500{\mu \mathrm{m}}^2$ (not shown). The SC of PHS thus exhibits significant heterogeneity in lipid/protein composition, some SC regions being characterized by “normal” regions (group 1 spectra) and others by “lipid-rich” regions (group 2 spectra) as shown in Fig. 8(d).

Using scanning electron microscopy, Ghadially et al.¹² showed that in chronic psoriasis, the most common type of this skin disease, the number of lipid lamellar bodies appeared normal. Mature lamellar basic unit structures were also of normal dimensions. However, a paucity of intercellular lamellae was observed throughout the intercellular interstices and the mature patterns of lipid bilayers usually observed in the upper SC were not obvious.¹² Last but not least, lipid droplets of about $10\mu \mathrm{m}$ and less in diameter were observed in the SC ¹². Considering the spatial resolution of both vibrational techniques that is approx. $4{\mu \mathrm{m}}^2$ for Raman and $25{\mu \mathrm{m}}^2$ for IR, the Raman spectra of group 2 and the IR spectra probably correspond to probed areas represented by the overlap of part of these lipids droplets and part of “normal” (protein-rich) SC domains, as represented in Fig. 8(d). The origin of these lipid droplets remains to be elucidated.

As shown in Fig. 8(c), a third type of IR spectra has been recorded. These typical spectra are peculiar in that they are reminiscent of an almost pure lipid region (${\mathrm{CH}}_2$ stretching vibrations, $\mathrm{C}═\mathrm{O}$ stretching mode near $1745{\mathrm{cm}}^{-1}$) with very few proteins (absence of amide bands). Such spectra were observed mainly from an almost continuous layer of about one pixel thick located at the top of the “lipid-rich” domains. This observation is schematized in Fig. 8(d) (dark green thin layer). Solving the origin/attribution of this pure lipid layer requires a deeper investigation.

Only a few studies on skin microspectroscopy were made using human tissues. This work presents the great advantage of bringing information that is of prime interest for pharmacological research, which aims to develop drug formulations adapted to human skin composition and structure. That said, working with human tissues also brings several concerns, most likely due to logistic issues. For instance, the various pathologists who made the biopsies on cadavers did not take care of removing samples of noninvolved skin and consequently, the present study did not allow analyzing with specific internal standards. In addition, scientific research is not the main concern when a human being’s death happens. For this reason, cadaver skin samples could take up to one month from the time of death until the analyses. Other issues such as the difficulty of retrieving a sufficient number of samples of human origin to perform appropriate statistical analyses, the different sites where the biopsies were taken, possible contaminations from human sebum, and histology artifacts in the SC area, count among the unpredictable circumstances that may have influenced the interpretation of the data presented herein.