2.1.

Human Organ Culture Wound Healing Model

Human skin specimens were obtained from reduction surgery in accordance with institutional protocols and used to generate acute wounds as previously described.²¹ A 3-mm biopsy punch was used to create an acute wound and skin specimens were maintained at the air-liquid interface with Dulbecco's modified Eagle's medium (DMEM) (BioWhittaker), antibiotic/antimycotic and fetal bovine serum (Gemini Bio-Products) at 37°C, 5% ${\mathrm{CO}}_2$, and 95% relative humidity. Specimens were collected for further analyses at zero, two, four, and six days postwounding.

2.2.

Gene Array Data Analysis

Microarray Suite 5.0 (Affymetrix) was used for data extraction and for further analysis. Data mining tool 3.0 (Affymetrix, Santa Clara, CA) and GeneSpring software 7.3.1 (Silicon Genetics, Redwood City, CA) were used for normalization, degree of change, and $p$-value calculations. Samples were normalized per chip to the 50th percentile and per gene to a median. Statistical comparisons of expression level between each condition were performed using ANOVA test. Only genes with a $p$-value less than 0.05 were considered to be statistically significant. Differential expressions of transcripts were determined by calculating the degree of change. Genes were considered regulated if the expression levels differed more than two-fold relative to healthy, unwounded skin.

2.3.

IR Imaging Experiments

For IR imaging, samples were flash frozen with liquid nitrogen and stabilized at $-{30}^{°}\mathrm{C}$ on a chuck within a Bright/Hacker 5030 Microtome (Bright Instrument Company, Huntington, UK/Hacker Instruments, Fairfield, NJ) oriented so that sections ($\sim 5\mu \text{m}$ thick) were cut perpendicular to the SC. Sections placed on ${\mathrm{CaF}}_2$ IR windows were generally mapped across both unwounded and wounded areas using a Perkin-Elmer Spotlight system (Waltham, MA) equipped with an essentially linear array ($16\times 1$ detector elements) mercury-cadmium-telluride detector. A high-precision XY sample stage permitted the collection of IR images (pixel area of $6.25{\mu \text{m}}^2$) over $\sim 0.5\times 0.5\mathrm{mm}$ sample areas. Spectral resolution was $8{\mathrm{cm}}^{-1}$. Overall, three sets of separate skin specimens were prepared and analyzed. Results shown are representative of all three sets.

2.4.

IR Data Analysis

IR spectra were analyzed using ISYS software version 3.1 (Malvern Instruments, LTD, UK). Image planes of spectral parameters and factor analysis scores were generated after linear baselines were applied over spectral regions of interest. Factor analysis, as detailed elsewhere,⁵ is a statistical multivariate technique that reduces the dimensionality of the data by detecting patterns in the relationships between observed variables. The approach begins with principal component analysis (PCA) followed by a score segregation routine that seeks transformations between the PCA loadings and Beer’s Law parameters. Factor analysis results in a set of factor loadings and factor scores that depict the correlations between the actual spectrum at each pixel and the factor loadings. The calculated factor loadings resemble typical spectra, although not of pure components. Usually three to six significant factors are observed in a wound healing dataset.

3.1.

IR Imaging Identifies Changes in Epidermal Lipid Composition During Wound Healing

A typical sequence of events in the current skin wound healing explant model is depicted in a series of visible micrographs from separate sections at various times (control, zero, two, four, and six days) following wounding, as shown in Fig. 1. In the unwounded control skin, the SC is clearly evident at the top of the micrograph. The image acquired immediately following wounding (day zero) shows both the presence of the residual SC at the top right as well as the wounded area on the left lacking the more dense SC tissue. The MET becomes visible covering some $\sim 3/4$ to $\sim 7/8$ of the surface from the right-hand to the left-central areas of the day two and day four images, respectively. Coulombe²² state that the MET arises from elongated keratinocytes that originate from the basal and suprabasal layers of the epidermis near the wound. In the current instance, the presence of the MET indicates that epithelization occurs in the explants and thus provides evidence for the validity of the model for biophysical studies. No uncovered area was found in the day six image, suggesting that the MET had covered the entire wounded area, i.e., the epithelization stage was complete.

Fig. 1

Visible images of microtomed wound healing sections. The control is a nonwounded section of human skin (left image) and the remaining images are marked according to a postwounding timeframe (left to right, day zero, two, four, and six). The 100-μm scale bar applies to all the images.

The visible micrographs in Fig. 1 do not contain molecular-level information and thus cannot be used to identify molecular constituents present in the image. IR microscopic analysis permits identification of particular molecular classes, and furthermore, permits elucidation of structural characteristics of particular constituents. Typical, single-pixel, raw IR spectra from different skin regions are overlaid in Fig. 2 with bands of interest marked. The high signal-to-noise levels in the spectra are immediately apparent. Subtle changes in several bands are observed comparing the spectra of different skin regions. Our analytical strategy was to apply factor analysis to the IR images within spectral regions known to be sensitive to particular aspects of molecular structure.

Fig. 2

Typical single-pixel IR spectra from various regions of microtomed human skin sections: stratum corneum (SC), migrating epithelial tongue (MET), viable epidermis (VE), and dermis as marked. Spectra of the SC, VE, and dermis were obtained from the day zero section (see Fig. 1) in the unwounded area and the spectrum of the MET was acquired from the day two section.

The sensitivity of several IR spectral regions to various aspects of skin, wound constituents, and particular elements of molecular structure are shown in Figs. 3Fig. 4Fig. 5Fig. 6 to 7. Figure 3(a) shows two factor loadings in the 1490 to $1720{\mathrm{cm}}^{-1}$ region, which encompasses the Amide I (mostly peptide bond $\mathrm{C}=\mathrm{O}$ stretch, frequency region 1620 to $1690{\mathrm{cm}}^{-1}$) and Amide II (mostly mixed N-H in-plane bend and $\mathrm{C}=\mathrm{N}$ stretch, frequency region 1510 to $1560{\mathrm{cm}}^{-1}$) vibrational modes. The factor loadings predominantly arise from keratin-rich and collagen-rich regions of the tissue. The Amide I spectral region shows substantial differences between these two proteins and provides a powerful means to distinguish between them in skin tissue. Collagen is characterized by a major Amide I peak near $1660{\mathrm{cm}}^{-1}$ and a shoulder near $1636{\mathrm{cm}}^{-1}$ while keratin is characterized by a more symmetric single band at $1652{\mathrm{cm}}^{-1}$. Significant differences in the Amide II patterns are also noted. Images of the factor scores for this spectral region in the day zero and day two samples help us to distinguish keratin in the epidermis and MET from the underlying dermis [Fig. 3(b), top and bottom, respectively]. The chemical identification of proteins known to be located at particular sites [see visible images in Fig. 3(c)], e.g., keratin in the MET, collagen in the dermis, provides a useful validation of the factor analysis method.

Fig. 3

Detection and IR characterization of the spatial distribution of keratin-rich and collagen-rich regions in day zero and day two ex vivo wound healing model human skin sections utilizing factor analysis conducted over the 1490 to $1720{\mathrm{cm}}^{-1}$ region. (a) Two factor loadings in the 1490 to $1720{\mathrm{cm}}^{-1}$ region. (b) IR score images for two factor loadings with color coding scale bar of scores as follows: $\text{red}>\text{yellow}>\text{blue}$. (c) Visible images of the two sample sections. The 100-μm scale bar applies to all the images.

Fig. 4

IR characterization of acyl chain conformational order (factor analysis conducted over the 2800 to $3000{\mathrm{cm}}^{-1}$ region) of a nonwounded control and wounded skin sections at time points zero, two, four, and six days postwounding. (a) Factor loadings of L1, ordered chains (red) and L2, disordered chains (black). (b) Score images of L1 and L2 for the control and day zero, two, four, and six sections as marked. The 100-μm scale bar applies to all the images.

Fig. 5

Images of univariate IR spectral parameters for the nonwounded control sample and wounded skin sections at time points zero, two, four, and six days postwounding. The integrated area of the symmetric ${\mathrm{CH}}_2$ stretching band (2840 to $2868{\mathrm{cm}}^{-1}$) is displayed in (a) and provides a semi-quantitative measure of the lipid acyl chain concentration. The integrated area of the high-frequency half of the lipid carbonyl (1744 to $1756{\mathrm{cm}}^{-1}$) is displayed in (b). The high-frequency half of the band was used to avoid interference from the incompletely resolved strong Amide I band (see Fig. 2). The 100-μm scale bar applies to all the images.

Fig. 6

Images of univariate IR spectral parameters for a day six postwounding section delineating (a) lipid acyl chain conformational order via the methylene symmetric stretching frequency, (b) relative lipid acyl chain concentration via the integrated area of the methylene symmetric stretching band (2840 to $2864{\mathrm{cm}}^{-1}$), and (c) relative lipid ester concentration via the integrated area of the $\mathrm{C}=\mathrm{O}$ stretching band (1728 to $1772{\mathrm{cm}}^{-1}$).

Fig. 7

IR characterization (factor analysis conducted over the 980 to $1480{\mathrm{cm}}^{-1}$ region) of nonwounded control and wounded skin samples at time points zero, two, four, and six days postwounding. (a) Factor loadings of F1 (blue), F2 (green), and F3 (red). (b) Score images of these three factors, showing spatial distribution of different chemical species for the control and day zero, two, four, and six sections as marked. The 100-μm scale bar applies to all the images.

Factor loadings and score images from the C-H stretching region (2800 to $3000{\mathrm{cm}}^{-1}$) calculated from the full set of skin sections (Fig. 1) are shown in Fig. 4(a) and 4(b), respectively. This spectral region was not analyzed in our prior study. The spectral patterns arise from relatively conformationally ordered (L1) or disordered (L2) lipid acyl chains. Acyl chains that are conformationally well-ordered, essentially all-trans, are characterized by methylene symmetric stretching ($\sim 2850{\mathrm{cm}}^{-1}$) and asymmetric stretching ($\sim 2920{\mathrm{cm}}^{-1}$) frequencies shifted down by 2 to $4{\mathrm{cm}}^{-1}$ from their counterparts in disordered chains. In our study, the methylene symmetric stretching mode ($v_{\mathrm{sym}}$ ${\mathrm{CH}}_2$) is shifted from $\sim 2854.5$ to $2852.7{\mathrm{cm}}^{-1}$ while the asymmetric stretching mode shifted from 2925.4 to $2922.3{\mathrm{cm}}^{-1}$, reflecting an increased degree of conformational order in the chains.

Concatenated images of factor scores generated from the loadings (L1 and L2) in the methylene stretching region are shown in Fig. 4(b) from the control skin and day zero, two, four, and six wound sections. The highest scores from the ordered lipid factor labeled L1 in the control and the period immediately following wounding (day zero) coincide with the SC as observed in the visual micrographs. The factor score images for L1 also confirm the identification of a small region of residual SC on the right side of the day two and four wound sections. In addition, slightly lower scores for factor L1 were observed in the nonwounded area of the viable epidermis (VE) (control, day zero and two) and throughout the MET (day two to six). Upon wounding, the spatial and temporal evolution of scores for the disordered lipid factor, L2, is notable. Scores are quite low throughout the entire area of the control skin sample and appear somewhat higher in the VE at day zero. The dark blue area limited to the top $\sim 10$ to 20 μm in the control and at day zero and two in the images represents very low scores for the disordered lipid factor (L2) in the SC as anticipated. In the day two image, small pockets with high scores, mostly in the area basal to the MET, were observed, prior to spreading throughout the basal area and leading edge of the MET on day four. The day six section displays high scores concentrated in the lower to middle regions of the newly formed epidermal layer.

A simple univariate measure was used to analyze the spatial distribution of the relative concentration of lipid in the dataset. The images shown in Fig. 5(a) were generated from the integrated area of the $v_{\mathrm{sym}}$ ${\mathrm{CH}}_2$ stretching band. The regions with high lipid content coincide with the ordered lipid located in the SC, as described above for Fig. 4(b) (L1 images) and with the disordered lipid (L2 images), especially the lipid pockets in the day two and day six sections. To aid in the identification of the chemical species of the disordered lipid, factor analysis was performed on the lipid $\mathrm{C}=\mathrm{O}$ (1720 to $1770{\mathrm{cm}}^{-1}$) stretching region (results not shown). Two major factors were observed, differing by about $2{\mathrm{cm}}^{-1}$ in carbonyl peak position. Both peak positions (1743 and $1745{\mathrm{cm}}^{-1}$) are characteristic of (lipid) ester carbonyls, although the origin of the difference is unknown. To further probe the spatial distribution of the lipid carbonyl content, the integrated area of this band was measured and the data is displayed in Fig. 5(b) for the wound healing dataset. The areas with the highest concentration of lipid carbonyl intensity overlap with regions of high lipid acyl chain concentration [Fig. 5(b)] in the day two, four, and six sections and notably with the disordered lipid pockets as mentioned above.

An alternative means to characterize the physical state of the lipid present in a healing wound from a similar skin specimen six days after wounding is presented in Fig. 6. We used a different manner of univariate measure to directly image lipid conformation via the methylene symmetric stretching frequency [Fig. 6(a)]. Herein, the 1 to $2{\mathrm{cm}}^{-1}$ increase due to chain disorder in $v_{\mathrm{sym}}$ ${\mathrm{CH}}_2$ comparing the SC to the VE is directly evident [Fig. 6(a)]. This type of univariate measure represents an average of the proportions of ordered and disordered lipid at a given location, whereas factor analysis provides separate images of ordered and disordered lipid species. In Fig. 6(a), lipids with disordered acyl chains, average $v_{\mathrm{sym}}$ ${\mathrm{CH}}_2$ observed at 2854 to $2854.5{\mathrm{cm}}^{-1}$, are observed throughout the MET and VE in the nonwounded area, along with the more ordered lipids of the SC ($v_{\mathrm{sym}}$ ${\mathrm{CH}}_2$ at $\sim 2852{\mathrm{cm}}^{-1}$).

A measure of the integrated area of the methylene symmetric stretching band, an approximate monitor of relative lipid acyl chain concentration, is displayed in Fig. 6(b). An equivalent plot for the lipid $\mathrm{C}=\mathrm{O}$ stretch is shown in Fig. 6(c). Comparison of Fig. 6(b) and 6(c) reveals that there is quite an overlap of area in the MET with both high methylene and carbonyl stretching intensities. Small pockets at the edge of the wound are observed as well. However, the integrated area of both stretching modes appears to be lower in the bulk of the nonwounded VE. As anticipated, high methylene intensity in the SC [Fig. 6(b)] is also evident in the image.

Further characterization of lipid, collagen, and keratin molecules is available through factor analysis of the 980 to $1480{\mathrm{cm}}^{-1}$ region (Fig. 7). We used the same wound specimens as shown in Fig. 1. The three factor loadings shown in Fig. 7(a) reveal distinct spectral features unique to different chemical species as deduced from their spatial distributions displayed in the score images [Fig. 7(b)]. High scores for factor loading 1 (F1) map to the regions previously assigned to disordered lipid [Fig. 4(b), L2] and a relatively higher ester $\mathrm{C}=\mathrm{O}$ content [Fig. 5(b)]. Correlating the three images, we tentatively assign the $1160{\mathrm{cm}}^{-1}$ band to the C-O single bond stretch of the ester group. The fact that the methylene bending mode, $\sim 1460{\mathrm{cm}}^{-1}$, is the highest-intensity band in F1 is also consistent with the presence of lipid in these regions. Thus, the proposition that disordered lipids are present in regions below and within the lower areas of the MET is reasonably justified. Score images for the remaining two factor loadings, F2 and F3, display discrete spatial distributions readily assigned to keratin in the epidermis and collagen in the dermis, respectively. In addition, several spectral features observed in the loadings along with factor analysis in the Amide I and II region (data not shown) are consistent with this assignment. For example, the intense band at $\sim 1400{\mathrm{cm}}^{-1}$ in F2 arises from the symmetric carboxylate stretch. Keratin has a larger percentage of glu and asp that contain carboxylate side chains than collagen, thus indicating its presence. The loading F3, arising from the collagen-rich dermis, is also confirmed by the presence of the $1338{\mathrm{cm}}^{-1}$ peak. This mode has been tentatively assigned as a ${\mathrm{CH}}_2$ wagging mode of the proline ring.²³

3.2.

Gene Expression Analyses Reveal Deregulation of Genes Involved in Lipid Metabolism

Microarray analyses were utilized to study gene expression in ex vivo wounded skin specimens two and four days postwounding compared to healthy control skin (Table 1). Changes were detected in genes involved in lipid metabolism. Fatty acid desaturases 1 and 3 (FADS1 and FADS3), enzymes that regulate unsaturation of fatty acids through the introduction of double bonds between defined carbons of the acyl chain, were induced over the four-day time period. In contrast, after an initial induction of stearoyl-CoA desaturase (SCD) at two days postwounding, expression decreased by day four. The principal product of SCD is oleic acid formed by desaturation of stearic acid. Furthermore, suppression of 3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1 (HMGCS1) was found at both time points tested. This enzyme catalyzes the reaction in which Acetyl-CoA condenses with acetoacetyl-CoA to form 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA), an intermediate in cholesterol synthesis, indicating deregulation of, and in this case, a decrease in cholesterol synthesis.

Table 1

List of lipid metabolism-related genes regulated two and four days after wounding.