1.

Introduction

The molecular level determination of dermal and transdermal percutaneous absorption pathways is of interest in several areas of dermatological research and drug delivery research. For example, optimal therapeutic intervention may require delivery of drugs at particular concentrations to well-defined locations within the skin. In addition, knowledge of permeation pathways of penetration-enhancing molecules would provide a quantitative basis for understanding the thermodynamic and kinetic mechanisms of action of these substances.

Various research groups have probed lipid behavior and protein secondary structures in epidermis, as well as stratum corneum (SC) models, using infrared (IR) spectroscopy techniques.^{1 2 3 4} Furthermore, in-vivo and ex-vivo attenuated total reflectance (ATR) techniques have been applied to examine the outer layers of skin and thereby probe hydration, drug delivery, and percutaneous absorption.^{5 6 7} Several biomedical studies by the Canadian National Research Council Biodiagnostics group have demonstrated the ability of infrared (both mid-IR and near-IR) spectroscopy to differentiate pathological skin samples.^{8 9} However, two significant limitations of traditional IR sampling techniques are: 1. the requirement that samples be homogenized and that each spectrum is thus an average from a large sample area: or, if IR-microspectroscopy is used to avoid sample homogenization, 2. the problem of slow image accumulation in point-by-point IR microspectroscopy.

The commercial availability of mid-IR array detectors has overcome the limitation of traditional IR microscopy sampling. This has resulted in the introduction of IR imaging spectrometers and opened up a new field; mid-IR microspectroscopic imaging. Our current instrumentation consists of a 64×64 array of IR detectors imaged to the focal plane of an IR microscope. On each pass of the interferometer, a complete IR spectrum is collected from each array element, thus 4096 spectra are acquired in about the same time required to generate a single spectrum with a standard IR microscope. This approach constitutes true mid-IR spectroscopic imaging and results in a gain of about 1000 in time (to achieve equivalent signal to noise ratio) compared to point-by-point IR microscopy. Acquisition of the raw sample data (a matrix of 64×64 spectra) takes approximately 5 min. Initial applications of this technology to tissues have included studies of the hydroxyapatite and collagen constituents of bone,¹⁰ silicone in breast biopsies,¹¹ and the distribution of lipids and proteins in slices of brain tissue.¹² In addition, recent studies by Diem et al.; have revived the long-standing hope that IR microscopy will be useful in the field of cancer detection.¹³

The current study describes our initial applications of IR microspectroscopic imaging to monitor the permeation and tissue distribution of two known dermal penetration enhancers, propylene glycol (PG) and dimethylsulphoxide (DMSO) in porcine skin, and demonstrates the potential of this technique for fields such as transdermal and dermal drug delivery.

2.

Experimental Section

3.

Results

Figure 1 presents typical data from three spectral regions from an 8-μm-thick section cut perpendicular to the skin surface of a sample that had been treated with deuterated propylene glycol (DPG). The isotopic substitution of deuterium atoms into the exogenous molecule provides a useful spectroscopic handle, since the carbon-deuterium stretching modes appear in a spectral region not overlapped by bands from tissue components. Figure 1(a) shows the 1000 to 1800 cm⁻¹ region from a single spectrum (of the 4096 collected simultaneously). Several readily identifiable spectral features are observed as labeled on the figure. The intense bands near 1650 and 1550 cm⁻¹ arise primarily from the protein amide I (peptide bond C═O stretching vibrations) and amide II (mixed N─ H in-plane bending and C─ N stretch) vibrations, respectively. The feature near 1720 cm⁻¹ arises from ester C═O stretching modes of lipids such as triglycerides and phospholipids, which constitute the vast majority of lipids in the epidermis and dermis. We anticipate that in future studies it may be possible to image the distribution of distinct lipid classes by comparing images of ester, acid, and methylene modes across a biological section. Weaker bands from 1200 to 1500 cm⁻¹ are attributable mostly to protein components of the skin and arise from both backbone and amino acid side-chain vibrations. The high frequency region (2700 to 3150 cm⁻¹) shown in Fig. 1(b) (again from a single spectrum) contains C─ H stretching vibrations of CH ₂ and CH ₃ groups. The former occur primarily in lipid constituents of tissue, the latter in protein constituents. Two spectra are shown in Fig. 1(b). The top spectrum arises from a protein-rich region of the tissue section. The weak broadbands, marked in the spectrum at 2870 to 2880 cm⁻¹ and 2960 cm⁻¹ arise from stretching modes of the methyl group vibrations of the protein (amino acid) side chains. The bottom spectrum arises from a lipid-rich region of the tissue section. Bands near 2850 and 2920 cm⁻¹, strongest in the bottom spectrum, arise primarily from the symmetric and asymmetric CH ₂ stretching modes of the lipids.

Figure 1

(a) Typical IR spectrum of the 1000 to 1700 cm⁻¹ region of skin. The major spectral features arising from lipid and protein constituents are labeled.²⁰ (b) Typical C─ H stretching regions (2750 to 3150 cm⁻¹) from lipid-rich (bottom trace) and protein-rich (top trace) areas of a skin section. (c) Top trace: The C─ D stretching region (2000 to 2300 cm⁻¹) from a skin section containing deuterated propylene glycol. 25 spectra from the set of 4096 have been added to improve the signal-to-noise ratio. Bottom trace: The C─ D stretching region (2000 to 2300 cm⁻¹) from a pure sample of deuterated propylene glycol.

Figure 1(c) presents an expanded view (averaged from 25 pixels out of the 4096 spectra acquired) of the spectral region 2000 to 2300 cm⁻¹, which reveals the presence of the C─ D stretching modes from deuterated propylene glycol (CD ₃ CD(OH)CD ₂ OH) present in the skin (top spectrum). This assignment is confirmed by comparison with the spectrum of pure DPG shown at the bottom of the figure.

Figure 2 shows the spatial distribution of the protein, lipid, and DPG components in skin in four images generated from spectra acquired from a skin section cut perpendicular to the surface. A visible micrograph of each section is shown in the left-hand panel. The distance from the skin surface (in microns) is given on the right-hand side of the figure. The color scaling used is red>yellow>green>blue and corresponds to the relative integrated intensity of the given spectral parameter. Relative protein concentration was determined from the intensity of the amide II vibration near 1550 cm⁻¹, while the spatial distribution of polar lipids was estimated from the intensity of the ester C═O stretching mode near 1720 cm⁻¹. The spatial distribution of DPG is depicted in the third vertical panel and was monitored directly from the integrated intensity under the contour formed from the various CD, CD ₂, and CD ₃ stretching modes depicted in Fig. 1(c).

Figure 2

The spatial distribution of protein, deuterated propylene glycol, and lipid from a section of skin cut perpendicular to the surface. An optical micrograph is included in the left-hand column. Relative protein distribution is estimated from the intensity of the amide II mode; propylene glycol distribution is estimated from the area under all the various C─ D stretching vibrations; lipid distribution (phospholipids and triglycerides) is estimated from the ester C═O band intensity. The right-hand scale is the depth from the surface in microns. The vertical images overlap to varying extents as noted. The intensities are color coded as red>yellow>green>blue.

In Fig. 3, the spatial distribution of DMSO-d ₆ is depicted using the same general approach as for the DPG in Fig. 2. A series of visible and IR images, the latter corresponding to the spatial distributions of the protein, DMSO, and lipid components of four images were generated from spectra of a skin section cut perpendicular to the surface layer. The spatial distribution of the protein and lipid constituents were derived from the amide II and C═O intensities, respectively. The DMSO distribution was determined from the peak areas under the CD ₃ stretching vibrations. The DMSO distributions as depicted have been subjected to simple image processing methods to improve contrast, as shown in Fig. 4.

Figure 3

The spatial distribution of protein, deuterated DMSO, and lipid from a section of skin cut perpendicular to the surface. An optical micrograph of each imaged section is included as the left-hand column. Relative protein distribution is estimated from the intensity of the amide II mode; DMSO distribution is estimated from the area under the CD ₃ stretching vibrations (these images are generated after PCA analysis); lipid distribution (phospholipids and triglycerides) is estimated from the C═O band intensity. The right-hand scale is the depth from the surface in microns. The vertical images overlap to varying extents as noted. The intensities are color coded as red>yellow>green>blue.

Figure 4

The effect of image processing on the visualization of deuterated DMSO distribution in a skin section. Top left: Integrated intensity under the CD ₃ stretching vibrations. Top right: The histogram of intensity distribution. Bottom left: Integrated intensity under the CD ₃ stretching vibrations following application of principal component analysis and reconstruction of the image with the first four principal components. Bottom right: Histogram of intensity distribution in the enhanced image.

An original image of DMSO distribution in a region closest to the skin surface is shown in Fig. 4. This image clearly exhibits substantial noise, traced to the weakness of the original spectral features. This is best demonstrated in the data acquired from the top 50 μm of the image. This area in the field of view does not monitor the skin section, yet shows a pixilated distribution of intensities even though there cannot be any real DMSO intensity present. The absence of substantial numbers of pixels in the histogram of the raw image containing values of zero for the integrated intensity parameter (abscissa) is consistent with this image.

Significant improvement in image quality can be achieved with principal component analysis. This well-known technique yields the eigenvalues of the variance-covariance matrix, and defines components that are responsible for the observed spectral variance. To create an enhanced image, four principal components were retained (which accounted for 99.7 of the variance in the data) and the image was reconstructed with only these four components. The improved image quality is evident, and the histogram of the image shows the expected number of values with a spectral parameter close to zero.

4.

Discussion

For elucidation of the mode of action of specific penetration enhancers, as well as for a general understanding of dermal drug delivery mechanisms, it is essential to develop methods that permit direct observation of the spatial distribution of exogenous materials within the skin. The purpose of the current study was to determine the utility of IR imaging microspectroscopy for this purpose. The data clearly demonstrate that IR spectroscopic imaging provides the means to examine the relative concentration of the substance of interest with a spatial resolution approaching the mid-IR diffraction limit. In the current case, the pixel size of 6.3 μm provides the limiting resolution. We have chosen to examine DPG and DMSO-d ₆ to develop the IR imaging method for elucidation of penetration pathways in skin.

As is evident from an examination of Fig. 3, DMSO penetrates skin to a depth of at least 1 mm (under our experimental protocols), and reveals a spatial distribution that is coincident with the protein constituents of the skin. Particularly noteworthy in this regard is the observation in the second row of a dark object at a depth of ∼600 μm in the visible image (possibly a hair follicle), which is relatively protein rich and lipid poor. The DMSO is strongly partitioned into this anatomical feature.

Our results substantially extend the earlier work of Higo et al.;, who used attenuated total-reflectance (ATR-IR) spectroscopy to follow the penetration of 4-cyano-phenol (CP) across human SC.¹⁵ A comparison with tape stripping measurements allowed Higo et al.; to validate the equivalent of Beer’s Law for CP and, with some limitations, to determine the depth of the CP penetration. Our measurements permit near diffraction-limited spatial resolution of the relative concentration of the exogenous substance, as seen in Figs. 2 and 3 for DPG and DMSO, respectively. Since sections can be prepared either parallel or perpendicular to the skin surface, a reasonably comprehensive spectral mapping of molecular distribution can be obtained in a few images. The penetration of DMSO was mapped (Fig. 3) at depths exceeding 1 mm.

An advantage of the imaging IR approach is that in addition to mapping of exogenous substances in skin, the spatial relationships between the exogenous and endogenous components may be monitored and correlated. It is evident from Figs. 2 and 3 that both DPG and DMSO localize with the protein-rich components of the tissue and suggests that the penetration pathway of these molecules through the epidermis involves protein.

The current proof-of-principle studies clearly demonstrate that IR spectroscopic imaging technology will be a powerful tool in facilitating advances in several areas of dermatological and drug-delivery research. It has been suggested through a combined FT-IR/DSC study that some penetration enhancers, when present during heating processes, enhance an α-helix to β-sheet conversion in the keratin component of porcine SC.¹⁶ PG was found to be quite effective in this process. DSC spectroscopic techniques have also been used to evaluate the effects of DMSO on skin.¹⁷ Several studies have provided evidence that high concentrations of DMSO promote steady-state drug permeation, and delipidation and protein denaturation were suggested as being responsible for the penetration enhancing activity.¹⁸ With current IR imaging technology, it is feasible to examine intact skin sections for the spatial characterization of this process, and to seek relationships between the structural alteration and the relative concentration of the enhancers. Along similar lines, Clancy, Corish, and Corrigan suggested that DMSO caused extensive lipid extraction in addition to protein denaturation.¹⁹ The spatial distribution of these processes should also be readily elucidated with IR imaging. As this technology is further developed, our ability to generate images that map changes in protein secondary structure, or relative ratios of specific endogenous components, will provide a fast and direct method to characterize the molecular changes discussed before. Thus, for example, by mapping the intensity of two distinct components of the protein amide I mode, it will be possible to directly image the relative distribution of protein secondary structure changes across a sample. With these developments, IR spectroscopic imaging technology will move beyond a distribution mapping of components to a mapping of molecular structures and interactions across tissue sections.