1. Introduction

Melanin is perhaps the most characteristic chromophore of human skin and as everyday life experience confirms, it can greatly affect skin color in ways that may have profound psychological and sociological implications. Besides this obvious manifestation though, several aspects regarding melanin structure, function, and biological role remain relatively unexplored and not fully understood. The most widely accepted function of melanin in skin is that of photoprotection [1–3]. However, several other functions seem plausible and have been proposed; these include free radical scavenging, photosensitizing, thermoregulation, metal-ion chelation, and drug binding, among others [2–3].

The controversy surrounding melanin function is further enhanced by our incomplete knowledge regarding its structure. Melanin appears to be a polymer of specific known monomers but that’s about where our definitive knowledge ends [4–6]. Moreover, recent accumulating evidence supports the proposition that there may not be a definite polymer structure at all; melanin rather appears to be composed of a variety of oligomeric components put together in various random ways, making up a final structure which is thus characterized to a significant extent by structural and chemical disorder [6–8].

In skin, melanin appears in two forms: eumelanin and pheomelanin. Eumelanin is composed of 5,6-dihydroxyindole (DHI) and 5,6-dihydroxyindole-2-carboxylic acid (DHICA) as well as their various redox forms. These compounds are thought to form oligomeric protomolecules composed of 4-5 basic units which introduce chemical diversity on a primary level [7–8]. The oligomeric units may then combine to form eumelanin, thus introducing the element of structural diversity on a secondary level. In a similar manner, pheomelanin is composed of various types of benzothiazine units; these also combine in unclear ways to form the final polymer structure of pheomelanin [5]. Eumelanin and pheomelanin, although chemically distinct in terms of their monomer constituents, may be simultaneously present in skin, in a mixed form. Evidence suggests that in the presence of cysteine, pheomelanin synthesis is favored in vivo over eumelanin synthesis. Following cysteine depletion, eumelanin formation proceeds with eumelanin deposition on the surface of the preexisting pheomelanin particles [5].

One way to investigate melanin structure is through the study of its optical properties. The optical properties of melanin have been extensively investigated in vitro (for a review, see [9]). Eumelanin appears brown-black while pheomelanin appears brown-red which indicates reduced absorption in the red region of the visible spectrum. The absorption spectrum of both melanin types exhibits a broad dependence on the wavelength extending from the UV to the visible and well into the NIR regions of the spectrum. This behavior is very uncharacteristic of a biomolecule and indicates a complex, heterogeneous underlying structure. Moreover, the absorption spectrum of melanin exhibits a characteristic exponential dependence on the wavelength as a plentitude of in vitro studies confirms [9]. In contrast to the numerous in vitro published studies though, information on the in vivo optical properties of melanin is scarce [9–10]. This is probably due to the fact that in vivo optical measurements are confounded by absorption due to other biomolecules and scattering effects due to various morphological microstructures present in skin.

Study of melanin optical properties in vivo is very important. This is because valuable information may be obtained in this way regarding melanin structure and function in its native environment i.e. within living human skin. In addition, such information may also be very important and useful for the characterization of various pigmented skin lesions. Thus, in this article, we focus on the in vivo study of the of skin melanin optical properties, non-invasively, using diffuse reflectance spectroscopy. We employ a simple but robust, recently developed model, which enables quantitative determination of skin scattering and absorption properties. We use this model to confirm the exponential nature of melanin absorption in vivo and to identify an uncertainty in the melanin absorption spectrum which reflects underlying disorder in structure and chemical composition.

2. Methods

2.1 Instrumentation

The experimental setup included a compact CCD spectrophotometer (Ocean Optics, USB2000). Light delivery and collection to and from the skin was by means of a fiber optic probe (Ocean Optics, R200-7) consisting of six 200 µm core diameter optical fibers for delivery of light and a single 200 µm core optical fiber for light collection. The six illumination fibers were arranged around the central collection fiber in a circular manner. Illumination was provided by a tungsten-halogen light source (Ocean Optics, HL-2000). All spectra were referenced to a calibration spectrum measured on a diffuse reflectance standard (Ocean Optics, WS-1). Spectra were collected in the 450-1000 nm range with a spectral resolution of approximately 1.5 nm and typical signal to noise ratio greater than 100:1.

2.2 In vivo data

Data were collected on the skin of 7 adult volunteers with Fitzpatrick skin type III. A total of 41 melanocytic nevi were studied with typically a few diffuse reflectance spectra measured at different locations on each nevus, depending on nevus size. Typically, 4–8 nevi were studied on each volunteer i.e. nevi were evenly distributed among volunteers. In all cases, spectra were measured by gently placing the probe in perpendicular contact with the skin, in a consistent and repeatable manner, with no pressure applied to the skin. Great care was taken such that the geometry for spectral measurements was reproducible and that there were no probe pressure artifacts. All nevi studied were clinically identified as macular (flat) acquired melanocytic nevi with diameter smaller than 5 mm. Nevi were not examined histologically. This was because of the initial and exploratory nature of this study, the fact that melanocytic nevi could be easily identified clinically, and the fact that all the adult healthy volunteers participating in the study did not wish to have their nevi removed.

2.3 Diffuse reflectance model

Diffuse reflectance spectra, R(λ), were modeled using a recently developed analytical model [11]:

with μ_a(λ) and μ′_s(λ) the absorption and reduced scattering coefficients of skin respectively and k ₁=0.025 mm^-1, k ₂=0.057 two constants which depend on the specific probe geometry. The coefficients μ_a(λ) and μ′_s(λ) were expressed as follows:

Here, c_Hb1 is the oxyhemoglobin concentration, c_Hb2 is the deoxyhemoglobin concentration, c_m is the melanin concentration, c_w is the water concentration, and c ₁ is a parameter related to the effective light scatterer size in skin [12]; ε_Hb1(λ), ε_Hb2(λ), and ε_w(λ) are the absorption spectra of oxyhemoglobin, deoxyhemoglobin, and water respectively [11]. The remaining parameters are λ₀=λ₁=400 nm and λ₂=800 nm. The melanin absorption spectrum in Eq. (2) is assumed to exhibit an exponential dependence on the wavelength with k_m the corresponding exponential decay constant [9]

The reduced scattering coefficient is assumed to exhibit a linear dependence on the wavelength, an assumption dictated by the experimental data. Skin reflectance, whenever it is mainly due to scattering (i.e. minimal absorption effects), generally exhibits a linear dependence on wavelength when measured through a fiber optic probe like the one employed in this study. This is usually most evident in the 600–900 nm range [11–12], but appears to hold true down to at least 450 nm, as spectra on vitiligo lesions (minimal presence of melanin) indicate (unpublished data). This linear dependence of the reflectance on the wavelength is a direct consequence of the use of a fiber optic probe which introduces a linear dependence of the reflectance on the reduced scattering coefficient [11]. Hence, the reduced scattering coefficient must exhibit a linear dependence on the wavelength in the 450–900 nm range. Although this is clearly an approximation, especially over such a wide wavelength range, it is also supported by Mie theory predictions for ensembles of spherical scatterers with a Gaussian distribution in size [12].

It is important to note that we have not observed any signs for a sharp increase in μ′_s in the 450–600 nm range. This feature, often attributed to Rayleigh scattering [13], has been observed in skin samples in vitro and it could be related to in vitro artifacts or it could be due to scattering in deeper layers of the dermis, which were not probed in this study. In addition, the reduced scattering coefficient is very often described by a power law dependence on the wavelength for various biological tissues [14]. Nevertheless, we have found based on this and other previous studies [11–12] that the linear dependence on the wavelength is a very good approximation for the description of the reduced scattering coefficient in the superficial layers of the skin examined in this study, in vivo.

2.4 Data analysis

Data were analyzed by fitting the experimental spectra to Eq. (1) by taking into account Eqs. (2) and (3). Thus, fitting produced a set of seven parameters characteristic of each skin spectrum: c_Hb1, c_Hb2, c_m, k_m, c₁, μ′_s(λ_min), and c_w. Fitting was performed using the Levenberg-Marquardt minimization method offered by the MERLIN optimization environment [15]. This method is generally very robust for minimizing functions expressed as the sum of squares. In all cases, there were no bounds imposed on the parameters and fitting produced positive meaningful parameter values. In general, the model parameters were in reasonable agreement with previous similar data reported in the literature, including our own previous investigations. In addition to automated minimization, a manual investigation of the minimum was performed by varying the model parameter values. In all cases, it was found that the minimization software identified the relevant global minimum and no spurious minima were identified. This finding is reasonable because even though the number of model parameters is not very small, unique determination of the global minimum may be achieved due to the distinctly different spectral dependences of hemoglobin and melanin absorption.

3. Results and discussion

Figure 1 shows typical diffuse reflectance spectra measured on two representative melanocytic nevi together with the corresponding model fits (red lines). Visual inspection of Fig. 1 reveals that the model describes the data remarkably well. Note the spectral features near 540 and 580 nm indicative of oxyhemoglobin absorption and the small dip around 960 nm due to water absorption. The general reduction in reflectance intensity observed from longer to shorter wavelengths is due to melanin absorption i.e. melanin is the dominant chromophore responsible for the reflectance lineshape. The excellent model fits confirm the exponential dependence of melanin absorption on the wavelength, in vivo.

Most importantly, k_m values were found to follow a Gaussian distribution (Fig. 2) characteristic of the underlying random nature of a biological variable (Shapiro-Wilk test, p>0.05). No correlation was observed for any other pair of the seven model parameters. This is an indication that determination of k_m was not affected by any of the remaining parameters. Thus, the remaining model parameters are not presented in further detail here. Determination of k_m was with typical accuracy better than 5%. This error was mainly related to the fitting procedure.

The observed variation in the absorption spectrum of melanin is consistent with recent accumulating evidence in support of a chemical and structural disorder picture for the melanin polymer [6–8]. In other words, melanin is not a biological chromophore with a well defined molecular structure (and hence a well defined absorption spectrum) such as hemoglobin, for example; the variation in k_m which follows a normal distribution serves as indirect evidence for that fact. This finding opens the way for further investigations. For example, it would be very interesting to study various pigmented skin lesions and different skin types; possibly other melanin types such as those found in the eye and the brain as well.

 

Fig. 1. Diffuse reflectance spectra measured on two representative melanocytic nevi, from two different volunteers. Red lines indicate model fits.

Download Full Size | PDF

 

Fig. 2. Distribution of melanin absorption exponential decay parameter, k_m, for all melanocytic nevi skin sites studied. The red line indicates a Gaussian distribution fit to the data, centered at k_m=5.3 with width equal to 1.1 (FWHM). To produce this figure, k_m values were binned with a bin width equal to 0.25.

Download Full Size | PDF

It would also be interesting to further investigate this observed variation in k_m and better pinpoint its origins. The variations observed maybe further studied and confirmed on a microscopic level through the use of a more accurate and specific technique, e.g. confocal microscopy. Then, variations in melanin composition in terms of the melanin/pheomelanin ratio can be probed by studying the absorption spectrum of melanin. Increased pheomelanin relative concentration would lead to melanin with lighter color and therefore higher values of k_m. Another way the observed k_m values can be affected is by means of the DHI/DHICA ratio. It is known that DHICA produces brown eumelanin as opposed to the black DHI-rich eumelanin [16]. Hence, DHI-rich melanin would be characterized by lower k_m values. Most importantly, variations in the structure of melanin itself could be probed in terms of the way they affect k_m together with potential interaction with other biomolecules, within the melanosome, such as proteins.

It is also quite interesting to put our in vivo results in perspective with the in vitro results reported in the literature. The main and basic similarity is that melanin absorption exhibits an exponential dependence on wavelength, both in vivo and in vitro. This appears to be a universal feature of the spectra and calls for a more detailed analysis and investigation of this feature’s origin. Another basic observation is the fact that k_m values, for in vivo melanin in the melanocytic nevi of subjects with Fitzpatrick skin type III, appear to be higher compared to those reported for various types of melanins in in vitro studies [9]. This is reasonable and perhaps consistent with brown melanin present in the nevi examined in the present study, as opposed to synthetic black melanin or even black melanin found in the skin of subjects of African descent (Fitzpatrick skin type VI). We expect that future studies will provide a clear and consistent picture of the optical properties of melanin in vivo among various skin types and also among various pigmented skin lesions.