Vibrational microscopy and imaging of skin: from single cells to intact tissue
- First Online:
- Cite this article as:
- Zhang, G., Moore, D.J., Flach, C.R. et al. Anal Bioanal Chem (2007) 387: 1591. doi:10.1007/s00216-006-0852-0
Vibrational microscopy and imaging offer several advantages for a variety of dermatological applications, ranging from studies of isolated single cells (corneocytes) to characterization of endogenous components in intact tissue. Two applications are described to illustrate the power of these techniques for skin research. First, the feasibility of tracking structural alterations in the components of individual corneocytes is demonstrated. Two solvents, DMSO and chloroform/methanol, commonly used in dermatological research, are shown to induce large reversible alterations (α-helix to β-sheet) in the secondary structure of keratin in isolated corneocytes. Second, factor analysis of image planes acquired with confocal Raman microscopy to a depth of 70 μm in intact pigskin, demonstrates the delineation of specific skin regions. Two particular components that are difficult to identify by other means were observed in the epidermis. One small region was formed from a conformationally ordered lipid phase containing cholesterol. In addition, the presence of nucleated cells in the tissue (most likely keratinocytes) was revealed by the spectral signatures of the phosphodiester and cytosine moieties of cellular DNA.
KeywordsConfocal RamanInfrared imagingCorneocytesEpidermisProtein secondary structure
Technological developments in infrared and Raman microscopic imaging have permitted several new types of biomedical applications. As demonstrated elsewhere in this volume, the diagnosis of pathological states in a wide range of tissues is now feasible, and is being pursued vigorously in laboratories around the world. The vast volume of spectral data generated from imaging experiments permits sophisticated multivariate statistical analysis of data and enables the experimentalist to discern tissue regions where spectral signatures from pathological states differ from those of healthy tissue.
An important attribute of vibrational spectroscopy is the availability of spectra–structure correlations from many tissue components. This extensive background information can provide a useful supplement for biomedical diagnostics. By way of example, Boskey and her collaborators have characterized a new form of brittle bone disease (osteoporosis) based on structural changes in the collagen component of mineralized tissue . This pathological state differs from the more widely known forms of osteoporosis, which are characterized by altered hydroxyapatite levels/structure. The origin of the spectral changes in the collagen component of pathological tissue biopsies was traced to increased levels of collagen cross links and alterations in their spatial distribution. Evidently, the availability of such information facilitates the development of strategies for tracing the origin (i.e., genetic predisposition, external factors, etc.) of the pathology.
Skin is another tissue where the availability of spectra–structure correlations as well as the ability to detect spectral signals from exogenous constituents opens up new possibilities for vibrational microscopic imaging experiments relevant to pharmacology, physiology, and biochemistry [2–4]. In this article, we provide two examples depicting the interplay of vibrational microscopic imaging with spectra/structure correlations.
The first example uses both IR and Raman microscopy to evaluate structural changes in the keratin of single cells (corneocytes) induced by treatment with dimethyl sulfoxide (DMSO) or chloroform/methanol (C/M). These solvents have been widely used in dermatology for studies of permeation enhancement and for extracting lipid from the stratum corneum (SC), respectively. The characterization of DMSO-induced structural changes in keratin will aid in the determination of DMSO permeation pathways and in the evaluation of SC barrier recovery. Assessing C/M’s effect on keratin in isolated corneocytes including the reversibility of potential structural changes, will facilitate the delineation of solvent treatment effects from that of lipid removal.
In the second example, we demonstrate that confocal Raman microscopy along with multivariate (factor) analysis provides the detection of particular skin regions (stratum corneum, epidermis, dermis) as well as endogenous sub-areas (keratinocytes, lipid inclusions) in the epidermis. The availability of spectra–structure correlations permits molecular structure information to be deduced from either the spectra or from the factor loadings.
Materials and methods
Preparation of corneocytes
The Rutgers University Internal Review Board has approved all protocols. Corneocytes were collected by tape stripping (Sellotape, 3M Scotchguard) human forearm skin. The forearm was flushed with water for several minutes prior to tape stripping. The first two strips were discarded and corneocytes from the third tape strip were isolated as follows. Corneocytes were flushed from the tape with HPLC-grade hexane. The hexane/corneocyte suspension was sonicated for approximately 5 min to break up desmosomes and lipid cohesion between individual corneocytes. Corneocytes were then isolated using nitrocellulose membrane filters with a pore size of 0.22 μm (GE Osmonics Labstore, Minnetonka, MN) and resuspended with fresh hexane. This procedure was repeated three times. A small aliquot of this suspension was separately deposited on either CaF2 windows for transmission IR or gold-coated silicon substrates for Raman measurements. Samples were dried overnight under house vacuum prior to spectroscopic measurements. The remaining corneocytes were dried under N2 gas prior to treatment with DMSO or C/M.
Corneocytes were placed on substrates that were directly dipped in pure DMSO for about 1 min. Separate aliquots of harvested corneocytes were transferred to glass vials using 5 mL C/M (2:1 v/v). Non-covalently bound lipids were extracted for 68 h while stirring at room temperature in sealed vials. Corneocytes, collected by centrifugation, were further washed 3 times using fresh C/M before the final suspension was placed on the appropriate substrate. Samples were dried overnight under vacuum prior to spectroscopic measurements. No vibrational bands from residual solvent were observed in corneocyte spectra. Subsequently, corneocytes from both DMSO and C/M treatment were rehydrated on the solid substrates by exposure to 100% relative humidity overnight before a final set of spectra were acquired.
Intact porcine skin preparation
Skin biopsies from Yucatan white, hairless pigs were purchased from Sinclair Research Center, Inc (Columbia, MO). The stratum corneum of the biopsies was washed with water and samples were cut to a size of approximately 1 × 1 cm2, 0.5-cm-thick sections. Sections were placed, SC side up, into a milled brass cell, covered, and sealed with a microscope coverslip for confocal Raman measurements.
IR microspectroscopy of corneocytes
IR spectra were acquired with the PerkinElmer Spotlight system described elsewhere in detail . The instrument has imaging capabilities, but in this study the point mode was used. Since the typical diameter of corneocytes is about 40–60 μm, an aperture size of 40 × 40 μm2 was chosen to optimize the spectral quality. A total of 256 scans was acquired for each spectrum using 4 cm−1 spectral resolution. One level of zero-filling yielded data encoded every 2 cm−1. Approximately 50 corneocytes were examined from each sample preparation.
Raman spectra were acquired with a Kaiser Optical Systems Raman Microprobe. The instrument has been described in detail elsewhere [5, 6]. A 785-nm diode laser generates approximately 7–12 mW of single mode laser power at the sample with a spot size of about 2 μm using a ×100 objective. The backscattered radiation illuminates a near-IR CCD (ANDOR Technology, Model DU 401-BR-DD). Spectral coverage is from 100 to 3,450 cm−1 at spectral resolution of 4 cm−1. Following linearization, data are encoded every 0.3 cm−1. Spectra were acquired from 10–15 corneocytes from each sample preparation using an 80-s exposure time, 4 accumulations, and cosmic ray correction. Optical images of corneocytes were acquired with a Leica microscope (Model DMLP) coupled to the Kaiser system with a ×100 objective.
Confocal Raman spectra of intact pigskin were acquired using a 60-s exposure time, 2 accumulations, and cosmic ray correction. A ×100 oil immersion objective was used for the intact pigskin samples. To evaluate axial characteristics of the optical set-up used in these experiments, the following observations were made with the oil immersion lens using protocols previously published . The test sample consisted of five layers of a polymer laminate which incorporated two chemically identical layers (out of the five), each with a thickness of 7.5 μm separated by an intermediate layer of thickness 15 μm. This provides a center-to-center distance of 22.5 μm between the identical layers. Raman intensity profiles show a baseline-resolved width for each of the chemically identical layers of approximately 12–13 μm, a half-width of the intensity profile for each peak of about 8 μm, and a peak-to-peak distance of about 25 μm, in good agreement with the known test sample physical characteristics. The actual axial spatial resolution in the skin samples cannot be directly ascertained.
Grams/32 AI software version 6.0 (Thermo Galactic, Salem, NH) was used for processing individual Raman and IR spectra. Raman spectra were Fourier smoothed (80–90%) and 2-point baselines were applied to spectral regions of interest. This degree of Fourier smoothing could be applied without inducing spectral lineshape distortions. Confocal Raman spectra were analyzed (e.g., factor analysis) and Raman image planes were generated using ISys software version 3.1 (Spectral Dimensions, Inc. Olney, MD). Factor analysis was performed using the ISys score segregation routine. The analysis seeks to detect simple patterns in the relationships between observed variables in order to reduce the dimensionality of the data. Score segregation begins by normalizing PCA scores which may then be sharpened by raising them to a power specified by the acceleration parameter. The default value of 10 was used. Factor loadings are then calculated according to (S′S)−1*S′*X, unless S′S is not invertible, in which case iteration stops, and a factor is deleted before continuing.
Results and discussion
Solvent-induced conformational changes in the keratin of single cells
Although its mechanism of action is not well understood, DMSO has often been utilized as an efficient permeation enhancer for a wide variety of drugs, including antiviral agents, steroids, and antibiotics . It has been suggested to alter protein conformation in the stratum corneum (SC), the outermost layer of skin and the main barrier to permeability. Anigbogu et al.  monitored the protein (keratin) α-helix to β-sheet transition in hydrated SC. In addition, Caspers et al.  monitored protein structure following DMSO penetration into human SC in vivo by confocal Raman spectroscopy. A broadening of the Amide I contour was noted, although major new spectral features were not resolved. It is noted that DMSO concentrations differed in the two experiments.
Helix:sheet transitions are well documented for different types of keratin, and have been detected by IR spectroscopy. In particular, Oertel  used IR to observe similar DMSO-induced changes in human SC. The current work extends observation of this conformational transition to isolated human corneocytes. It is also noted that, based on the IR Amide I and II band contours, a fair amount of helical structure remains subsequent to DMSO treatment.
The Raman spectrum (Fig. 2b, middle trace) corroborates the β-sheet assignment with the appearance of a high frequency Amide I component at about 1,669 cm−1 and a strong Amide III component at approximately 1,240 cm−1. Both features are within frequency ranges assigned to antiparallel β-sheets [13, 14]. The observed changes in the Amide I region of the Raman spectrum are consistent with a previous report detailing the effects of a series of aqueous DMSO solutions on human SC . Comparisons between individual cells and the SC in the Amide III region are difficult due to overlapped bands from lipid components in the SC sample. In the current work, a second slightly weaker band is also observed for the DMSO-treated cells at approximately 1,270 cm−1. Bands in this region are usually assigned to α-helical structure [13, 14]. Generally, caution is required in associating characteristic frequency ranges with conformation for the Amide III mode given that the NH in-plane bend internal coordinate contributes to a number of modes in the 1,200–1,400 cm−1 region  and secondary-structure-dependent coupling occurs between the Amide III and the Cα–H bending vibration [16, 17]. This complexity, in addition to the inherent heterogeneity of corneocytes and variable lipid content of the samples examined herein, precludes an in-depth analysis of the Amide III contour in the current context.
The majority of the DMSO-induced secondary structure change is reversed upon rehydration of the corneocytes as demonstrated by comparison of the top and bottom spectra shown in Fig. 2a and b. Some residual broadening remains on the low frequency side of the Amide I band in the IR spectrum. The average full width at half height for the Amide I band in the IR spectra increased from 44 cm−1 for untreated corneocytes to 52 cm−1 following DMSO treatment and rehydration, indicating that some residual sheet structure may be present. An accurate measure of the Amide I bandwidth in the Raman spectra is precluded by the presence of a weak side chain vibration at approximately 1,608 cm−1. The general reversibility of the conformational change upon hydration was also observed by Oertel  for isolated SC.
C/M solutions are commonly used to extract lipids from isolated SC and intact skin samples, but solvent effects on the structure of the remaining constituents have not been detailed. The utility of vibrational spectroscopy for evaluation of keratin secondary structure after lipid removal is somewhat hampered by the native ceramide contribution to the Amide I and II band prior to solvent exposure. In SC samples it is therefore difficult to accurately access the contribution ceramides make to Amide I and II band intensities and position. In the current work, this concern is mostly overcome by acquiring spectra of isolated corneocytes in which the ceramide contribution to the Amide modes is substantially reduced.
Several changes are apparent in the IR and Raman spectra of corneocytes upon exposure to C/M (Fig. 2c and d, compare top to middle traces). Some of the solvent-induced alterations are similar to those just discussed for the DMSO-treated cells, e.g., the low (high) frequency Amide I component in the IR (Raman) spectrum at 1,626 cm−1 (1,670 cm−1) and the appearance of an Amide III band at 1,240 cm−1 in the Raman spectrum. These features provide strong evidence that C/M also induces antiparallel β-sheet formation. Differences are also evident when spectra from the two treatments are compared. Broadening on the high frequency side of the Amide I band in the Raman spectra is much more intense after C/M treatment vs. DMSO (compare the middle trace in Fig. 2d to b). Amide I Raman intensity in this region (ca. 1,685 cm−1) has been assigned to non-hydrogen-bonded disordered structures . The broadening in the Raman Amide I band is not completely reversible upon rehydration of the sample that had been treated with C/M, even though the intensity of the β-sheet component at about 1,670 cm−1 has diminished (Fig. 2d, bottom trace). In addition, upon rehydration, intensity remains in the Amide III band at approximately 1,240 cm−1, although this band also broadens. Finally, measurements of the Amide I bandwidth in IR spectra upon rehydration of several C/M-treated corneocytes are essentially the same (±2 cm−1) as widths measured for untreated cells. The close proximity of Amide I bands arising from helical and random coil structures makes it difficult to differentiate between these two forms. Considering both the Raman and IR spectral data after rehydration, it seems safe to conclude that the majority of keratin exists in a helical form with some disordered structure remaining.
The extent of the C/M-induced protein structural changes has been observed to vary among corneocytes. IR spectra of some cells with significant lipid bands (i.e., those characterized by relatively intense methylene stretching modes near 2,850 and 2,918 cm−1) still exhibit notable protein conformation changes (data not shown), while some cells from which the lipid has been essentially completely extracted do not show significant changes in protein secondary structure. It thus appears that the extent of lipid extraction is not directly related to C/M-induced protein conformational changes. Similar results have been noted in IR studies of DMSO-induced protein conformation changes in SC samples . Additional experiments under conditions of controlled hydration are needed to further explore possible origins of these conformational changes. One explanation for the α-helix to β-sheet structure change in keratin is that DMSO displaces bound water necessary for maintaining secondary structure . DMSO is known to form an association complex with water stronger than that formed between water molecules alone . It may be that C/M destabilizes keratin’s native secondary structure in a manner similar to DMSO, since the majority of the native structure is restored after both solvent treatments when the corneocytes are exposed to high relative humidity.
The importance of the current experiments lies in the fact, that in studies of supposedly native protein structures in the SC, C/M is often used to remove lipid components. As shown in the current work, this procedure is not innocuous, but can induce irreversible changes in protein structure in corneocytes.
Confocal Raman microscopy: delineation of skin regions from factor analysis
As demonstrated above, Raman microscopy complements the information available from IR. Although Raman scattering is inherently weaker than IR absorbance, the technique offers two major advantages for tissue studies. First, the spatial resolution of the method is inherently better than IR. For an excitation wavelength of 785 nm, Rayleigh’s rule for spatial resolution (ΔZ = 1.22λ/2(NA)) predicts a spatial resolution of approximately 0.5 μm for a wavelength (λ) of 785 nm and a numerical aperture (NA) of the ×100 objective of 0.95. In practice, as has been discussed in Materials and methods and elsewhere, the apparent depth and related measured optical properties are degraded by refractive index variations in the tissue [5, 18, 19]. Second, and equally important, spectra may be acquired in a confocal manner. Thus, physical sectioning of samples is not required to obtain molecular structure information from within the tissue. This evidently opens up the possibility for in vivo examination of skin, especially at the low laser powers (ca. 10 mW) used in these experiments. Puppels’ group has demonstrated the utility of confocal Raman determination of water levels in vivo in skin .
Finally, Factor 3 (Fig. 3b), whose score image reveals several small-localized pockets in the epidermis, contains one broad spectral feature centered at approximately 1,090 cm−1 that is not present in the other factor loadings. Based on both the position of this particular band and the spatial distribution of the factor score, the band is tentatively assigned to the phosphodiester mode of DNA [24, 25]. Since the identification of DNA is of interest, as it would reveal the location of nuclei within keratinocytes or other cells in the tissue, an additional strategy was employed to acquire better spectra from the epidermis.
The current experiments highlight some unique advantages of vibrational microscopic imaging. An understanding of the spectroscopy of the various tissue components permits insight into the molecular origins of the images generated from factor analysis. Thus, identification of the lipid pocket in Fig. 3c (Factor 2) and of the cell nuclei (as detected from the DNA spectral signatures) in Fig. 3c (Factor 3) and Fig. 5a were greatly facilitated by the availability of Raman spectra of lipids and of DNA, respectively. In addition, the availability of correlations between lipid spectra and chain conformation, permitted assignment of the lipid inclusions as having arisen from ordered lipid phases. The biological relevance of this observation remains to be determined. From a pharmacological perspective, the detection of cell nuclei (most likely keratinocytes) within intact skin permits co-localization experiments to begin to be designed. For example, for classes of drugs targeted to these cells, confocal Raman measurements will permit determination of whether the putative therapeutic agent reaches its intended target.
In addition, although all microscopic methods can obviously readily detect single cells, the ability of IR- or Raman-based methods to detect conformational changes within a particular component of a single isolated cell is unique. The example we have chosen in Fig. 2 is of practical importance in skin research, since solvents such as DMSO or chloroform/methanol are used for permeation enhancement and lipid extraction, respectively. In each case these solvents were found to be far from innocuous. Each induced large conformational changes in the cellular keratin. This fact had been known previously from spectra of intact SC, but two aspects of the current measurements are novel. The observation of IR data from single cells provides a sharper spectral assignment of the bands. We note for example that ceramides, a major component of the SC, have strong bands arising from the amide bond located in these lipids. These modes contribute significantly to the spectra of the intact SC and overlap protein Amide I and II vibrations; however, they are much reduced in relative intensity in spectra of cells. Thus, studies of the reversibility of the solvent-induced conformational changes benefit from the reduced intensity of ceramide interference.
Finally, it is feasible to begin to examine biochemical heterogeneity at a single cellular level. Issues such as the relationship of protein structure (either native or solvent-modified) to hydration levels or to the levels of natural moisturizing factors, in single cells may be profitably probed. In addition, changes in the spectra of cells from pathological states may be examined.
This work was generously supported by PHS grant GM 29864-25 to RM. Additional support from Rutgers University toward acquisition of instrumentation is also acknowledged.