1.

Introduction

Over the last decade, optical spectroscopy has rapidly evolved as an in vivo tool for diagnosis and characterization of human skin. ^{1, 2, 3, 4, 5, 6} Like all biological tissues, the skin is composed of biomolecules with specific biophysical properties that are determined by their chemical structures and microscopic environments. The biophysical properties of cutaneous chromophores such as melanin, hemoglobin, collagen, keratin, and lipids can be studied noninvasively and in vivo using a variety of optical spectroscopic techniques such as diffuse reflectance, fluorescence, and Raman scattering. ^{2, 4, 7, 8, 9, 10, 11, 12, 13}

Melanins are ubiquitous pigments that largely determine skin, hair, and eye color. Functionally, melanins appear to play somewhat conflicting roles in terms of serving as natural sunscreens while potentially also initiating skin cancers.^{14, 15, 16} The unique photochemical and photobiological properties of melanins in the skin have been analyzed by optical spectroscopy. ^{8, 17, 18, 19, 20, 21} Among cutaneous chromophores, melanin is unusual in that its absorption decreases monotonically with increasing wavelengths from 300 to $1100\mathrm{nm}$ with an absence of discrete spectral bands or peaks for identifying its structural subunits.^{17, 20, 22} Nevertheless, some structural information about melanin can be derived from visible or near-infrared (NIR) absorption spectroscopy.^{13, 17} Spectroscopy that assesses the apparent absorption of tissue derived from reflectance measurements has been used as the basis for skin cancer diagnosis. In this instance, diagnosis is based on analyzing the absorbance of melanin, oxyhemoglobin $\left({\mathrm{HbO}}_2\right)$ , deoxyhemoglobin (Hb), and water $\left({\mathrm{H}}_2\mathrm{O}\right)$ from 400 to $1100\mathrm{nm}$ within suspicious skin lesions.^{1, 23, 24}

Fluorescence spectra of biologic tissues may often be more informative than optical absorption in terms of molecular structure and functional groups. Moreover, the fluorescence of specific tissue components is strongly influenced by the microenvironment.^{18, 19, 25} Under ultraviolet (UV) or short visible (VIS) light excitation melanin is a relatively low quantum yield fluorophore, and any fluorescence that it does emit is presumably attenuated by melanin’s broad spectral absorption.¹⁹ Because of this optical behavior under UV-VIS light excitation, melanin has previously been regarded as a nonfluorescent pigment that is distinct from the autofluorescent lipopigment family.^{25, 26} However, we have observed that the wavelength of excitation is important in obtaining fluorescence from melanin, which may explain why others have regarded melanin as a nonfluorescent pigment.

In this study, we extend the previous work of our group and others on UV/visible autofluorescence^{2, 3, 27} to the NIR domain by studying melanin samples in vitro as well as normal, sun-exposed, and diseased human skin in vivo under 785-nm laser excitation. Melanins within skin appendages such as hair and feathers were also evaluated. We have observed that synthetic and natural melanins exhibit significant fluorescence emission under near-infrared (NIR) light excitation at $785\mathrm{nm}$ , which is strikingly different from their apparent nonfluorescent behavior under UV-VIS light excitation.

2.

Materials and Methods

This study was approved by the Clinical Research Ethics Board of the University of British Columbia. NIR autofluorescence spectra were acquired from in vitro melanin as well as from hair, feathers, and in vivo human skin.

3.

Results

Solid-state in vitro melanins (Sepia and synthetic DOPA melanins) exhibit prominent fluorescence in the NIR range under $785\mathrm{nm}$ excitation [Fig. 1a ], with the spectrum being slightly higher in intensity at longer emission wavelengths. The two broadbands with peaks near 880 and $890\mathrm{nm}$ represent Raman scattering by melanin.²⁹ NIR-excited autofluorescence from black human hair, feline fur, and chicken feathers, which are replete with melanin [Figs. 1b, 1c, 1d], is similar to that of Sepia and synthetic melanin. White hair and white feathers, which presumably differ from their black counterparts only in terms of their relative absence of eumelanin, exhibit a markedly lower NIR autofluorescence signal.

Fig. 1

NIR autofluorescence spectra (820 to $920\mathrm{nm}$ ) obtained from: (a) Sepia and synthetic DOPA melanin; (b) human hair; (c) feline fur; and (d) chicken feathers. Excitation wavelength $785\mathrm{nm}$ ; light intensity $170\mathrm{mW}$ ; and integration time $1\mathrm{s}$ .

The visibly discernible differences in the degree of skin melanization between darker dorsal and lighter volar forearm skin of normal volunteers could be demonstrated by in vivo NIR autofluorescence and the corresponding autofluorescence difference spectrum [Fig. 2a ], with the dorsal forearm showing greater autofluorescence than that of the volar aspect in the 820- to 920-nm range ( $p<0.0001$ at all wavelengths). Consistent with this finding was the observation that the depigmented skin of vitiligo lesions had a much lower NIR autofluorescence intensity than that of the normally pigmented surrounding skin [Figs. 2b]. Furthermore, the spectra revealed the presence of several embedded Raman signals (e.g., 857, 871, 874, 885, 901, and $909\mathrm{nm}$ ) from both pigment-bearing (normal) and amelanotic (vitiligo) skin. These signals are due to cutaneous proteins and lipids,⁴ and their consistent presence in all skin measurements would be expected, since these components are unaffected by vitiligo.

Fig. 2

Paired mean NIR autofluorescence spectra (820 to $920\mathrm{nm}$ ) and the corresponding mean difference spectra obtained from: (a) dorsal and volar sites of the forearm [gray curve is the mean difference spectrum $\pm$ SD $(n=52)$ ] and skin lesions and their normal surrounding skin sites; (b) vitiligo $(n=4)$ ; (c) compound melanocytic nevus $(n=3)$ ; (d) nevus of Ota $(n=1)$ ; (e) superficial spreading melanoma $(n=3)$ ; and (f) postinflammatory hyperpigmentation $(n=1)$ . Excitation wavelength $785\mathrm{nm}$ ; light intensity $170\mathrm{mW}$ ; and integration time $1\mathrm{s}$ .

Dark-colored skin lesions due to abnormal hypermelanization (i.e., benign compound nevus, nevus of Ota, melanoma, and postinflammatory hyperpigmentation) all showed higher NIR autofluorescence when compared to the adjacent normal skin [Figs. 2c, 2d, 2e, 2f]. For heavy pigmented skin lesions, the difference spectra in Figs. 2c, 2d, 2e, 2f showed an increasing trend of intensity at longer wavelengths, which is in consistent with NIR fluorescence spectra observed from synthetic or extracted melanin samples and melanin-rich black hair or black feathers (Fig. 1). Therefore, melanin is an important fluorophore accounting for in vivo skin in the NIR range 820 to $920\mathrm{nm}$ .

In contrast to NIR-excited skin autofluorescence, visible autofluorescence is lower in the presence of cutaneous melanin, as in the case of a pigmented compound melanocytic nevus versus normal skin under blue light excitation at $437\mathrm{nm}$ [Fig. 3a ]. Similarly, normal skin had lower visible autofluorescence than depigmented patches of vitiligo [Fig. 3b].

Fig. 3

Paired autofluorescence spectra in the visible range (460 to $750\mathrm{nm}$ ) obtained from skin lesions and their normal surrounding skin site: (a) compound melanocytic nevus, and (b) vitiligo. Excitation wavelength $437\mathrm{nm}\pm 10\mathrm{nm}$ ; light intensity $9.5\mathrm{mW}$ ; and integration time $0.1\mathrm{s}$ .

4.

Discussion

Although the diagnostic application of cutaneous autofluorescence has been studied intensively over the last decade, most studies have focused on excitation wavelengths ranging between ultraviolet (UV) and shorter wavelength visible light. ^{2, 5, 12, 28, 30, 31, 32, 33, 34} A very limited number of autofluorescence studies have utilized longer red-to-NIR wavelengths, but these were mostly centered on examining tissue fluorescence in relation to porphyrins.^{35, 36, 37}

We have characterized the NIR autofluorescence properties of melanin obtained by chemical synthesis and tissue extraction, as well as through direct in vivo or in vitro measurements of skin, melanin-containing appendages (e.g., human hair, feline fur, and chicken feathers), and partially purified melanin. It has been shown by observation and histochemical methods that regardless of ethnicity, epidermal melanin content is significantly greater in chronically sun-exposed skin than it is in corresponding sun-protected skin sites (up to two-fold).^{15, 21} In this study, sun-exposed skin (dorsal forearm) exhibited significantly more NIR autofluorescence compared to sun-protected skin (volar forearm), confirming that NIR autofluorescence of epidermal melanin parallels the visible degree of pigmentation. The results clearly demonstrate the ability of melanin from a variety of biological sources to fluoresce positively under NIR excitation in a measurable fashion. In contrast, both visible absorption and fluorescence spectroscopy rely on the notion that the relative absence of a spectral signal implies the presence of melanin, and that the degree of suppression of a signal can be used to indirectly estimate the quantity of melanin present.^{7, 8} The ability to measure a “positive” optical signal in the presence of melanin within the NIR excitation domain thus provides a fundamentally different and more direct approach for in vitro and in vivo melanin detection.

That melanin exhibits strong NIR autofluorescence and Raman emission under NIR wavelength excitation is perhaps surprising, given the conventional acceptance that melanin was essentially nonfluorescent when measured in the UV and short visible wavelengths. The biophysical basis for melanin autofluorescence remains largely unknown. It has been suggested that it may be induced by partial oxidative breakdown of melanin, and therefore depends on structurally defective melanin polymers.^{19, 38} Our data show that prominent NIR autofluorescence emission appears to be a common feature of natural and synthetic melanins, and that melanin is one of the major fluorophores responsible for cutaneous NIR autofluorescence. The observation that the melanin-poor vitiligo lesions still exhibit significant NIR autofluorescence confirms that other fluorophores such as hemoglobin, porphyrin, collagen, etc. in the skin also contribute to cutaneous NIR autofluorescence.^{4, 29}

The NIR fluorescence line shape of white hair, white feather, and light colored skin tissue observed in this study is similar to that of other tissue types (e.g., nonskin tissues with no melanin content) reported in the literature.³⁶ The common feature of this fluorescence line shape is that the fluorescence intensity monotonically decreases with wavelength and shows no maxima. In comparison, the NIR fluorescence line shape of melanin (Fig. 1) is quite unique in that the fluorescence intensity increases with wavelength and appears to have a maximum. A complete understanding of these phenomena warrants further investigations.

In Fig. 2, all the difference spectra have positive values, confirming the significant contributions of melanin to in vivo skin tissue NIR fluorescence. However, one may note that in Fig. 2, not all the difference spectra follow the line shape of the pure melanin spectrum. The difference spectrum between normal and vitiligo (normal–vitiligo) does not show a maximum and is slightly titled toward the shorter wavelength end. We are building a tissue optics model and planning a Monte Carlo simulation approach to under this phenomenon. Tentatively, we have the following explanation for this phenomenon.

We use the “normal–vitiligo” difference spectrum as an example and assume that the only difference between vitiligo and its surrounding normal skin is melanin. Due to the lack of melanin absorption in vitiligo, excitation light can penetrate deeper down into the tissue than in normal skin; re-emitted fluorescence will also be easier to escape. Thus, when measuring vitiligo we are sampling a larger tissue volume and with higher detection efficiency than measuring normal skin. We are getting more nonmelanin-related fluorescence signal from vitiligo than from normal skin. The proportion of this extra nonmelanin related fluorescence signal should increase with wavelength, since longer wavelength fluorescence photons have higher chances to escape out of the tissue. On the other hand, melanin adds a strong fluorescence signal to the normal skin. Therefore, the line shape of the difference spectrum (normal–vitiligo) will not be exactly the same as the fluorescence spectrum of pure melanins. Instead, it will be determined by the balance between the extra melanin fluorescence signal generated in normal skin and the extra nonmelanin-related fluorescence signal detected from vitiligo. Similar considerations can also be applied to other spectral pairs shown in Fig. 2. Therefore, for heavy pigmented lesions [Figs. 2c and 2e], the difference spectra have a shape more close to that of the pure melanin, while for more lightly pigmented lesions [such as normal skin versus vitiligo in Fig. 2b], the difference spectra deviate significantly from a pure melanin spectral shape and the maximum may disappear. We expect that our Monte Carlo modeling, which accurately accounts for all the tissue optics effects on the difference spectra, will either confirm this consideration or slightly modify it.

We have observed two obvious characteristic Raman bands near 880 and $895\mathrm{nm}$ from melanin samples, black hair, and black feather (Fig. 1). For in vivo spectra of melanin-rich skin lesions shown in Fig. 2, these two Raman bands are not very obvious. They may have been masked, to a certain extent, by other strong Raman signals that originate from other biomolecules, such as proteins, lipids, porphyrin, etc., in skin.^{4, 29, 39} However, if we remove the fluorescence background from these in vivo spectra and concentrate on looking at the Raman signals, the melanin’s contribution to the total Raman signals becomes more evident (data not shown).

Changes in NIR autofluorescence spectra directly correspond to visible differences arising from melanin in pathological conditions such as vitiligo, malignant melanoma, benign compound melanocytic nevi, nevus of Ota, and inflammatory hyperpigmentation. Thus, NIR autofluorescence spectroscopy can potentially capture differences in melanins for pigmented skin lesions in vivo. It appears that the NIR autofluorescence technique may be particularly beneficial to the diagnosis of pigmented lesions, because these dark skin lesions can be lit up under NIR excitation. With UV/VIS excitation, melanin largely behaves as a photon absorber (Fig. 3), making visible fluorescence or reflectance measurements less informative. As previously mentioned, a “positive” NIR autofluorescence signal may be particularly useful for noninvasive analysis of lesions such as malignant melanoma, which can be difficult to differentiate from other pigmented but nonmelanin containing lesions using UV-VIS fluorescence or reflectance (i.e., apparent absorption) spectroscopic techniques.²⁴

5.

Conclusions

In aggregate, our results demonstrate that under NIR excitation, melanins exhibit prominent autofluorescence within the skin and its pigmented appendages, whereas with UV-visible excitation, any potential melanin fluorescence is essentially undetectable. Practical applications for melanin quantification and diagnostic evaluation of pigmented skin lesions are supported by this observation. Furthermore, it appears clear that the fluorescence properties of melanins are highly wavelength dependent, which in turn merits further detailed exploration.