1. Introduction

Melanin is the most important pigment in human and is the determinant of our skin color. In human skin, melanin not only acts as a physical barrier [1] to protect epidermal keratinocytes and superficial dermal vessels from ultraviolet light irradiation, but also acts as a free-radical scavenger. Malfunctions of melanin production and distribution is a signature of skin pigmentary diseases [2] and a label-free melanin imaging tool is thus highly desired for pigmentary skin disease diagnosis and treatment assessment. Various label-free microscopy techniques have been developed for in vivo high-resolution imaging of melanin in human skin, including two-photon-excitation fluorescence microscopy (TPFM), fluorescent lifetime imaging (FLIM) [3,4], photoacoustic microscopy (PAM) [5,6], reflection confocal microscopy (RCM) [7,8], and third harmonic generation (THG) microscopy [9,10]. By using 1230 nm excitation source, strong THG contrasts provided by melanin can be observed in human skin in vivo [9]. Compared with FLIM, TPFM, and PAM, no light absorption is required for THG, while the third order nonlinearity of THG can provide superior signal-to-background ratio for deep tissue imaging [11] as compared with RCM. By operating at the high penetration wavelength around 1200nm-1300 nm window [11], THG microscopy is an ideal choice for label-free imaging of melanin. A recent study investigated the THG in melanin hydrocolloid solution [12]. A linear relationship between melanin concentration and THG intensity was reported, in sharp contrast to the expected nonlinear relationship. This previous study had not only concluded a three orders larger third-order hyperpolarizability of melanin hydrocolloids than that of water, but also indicated that the observed signal might not be originated from coherent resonance-enhanced THG [13–17], but from the incoherent high-order hyper-Rayleigh scattering process [18].

Since melanin hydrocolloid solution is very different from the skin tissue observed clinically, here we report our study on the origin of the melanin-enhanced third harmonic generation by using a live cell model. Under strong focusing conditions the Gouy phase shift prevents THG in isotropic media. Efficient THG occurs only in case of inhomogeneity in refractive index or third-order nonlinear susceptibility distributions [19]. Therefore the surrounding materials around melanin could be factors to affect THG strength. Since the goal of this study is to explore the origin of melanin-enhanced THG in human keratinocytes in vivo, natural environment such as a melanin-containing tissue or a melanin-generating cell line could be a better choice than artificial conditions such as melanin dispersed in agar or the melanin/glass interface. Different from the previous study, here by using the B16-F10 cell line, our study has indicated an initial nonlinear process where the THG intensity was resonance-enhanced according to the 3.5th power of melanin mass density (MMD). When the MMD is higher than 11 mg/ml, a transition from the resonance-enhanced THG to the high-order hyper-Rayleigh scattering process occurs. This saturation phenomenon of the virtual-transition-based nonlinear THG is attributed to the decoherence scattering by randomly-assembled multiple-melanosomes within the sub-femtoliter focal volume.

2. Experimental design

Eumelanin is the major component of human epidermal melanin, even in the lightest skin type, where it represents above 90% both in number of melanosomes [20] and in total mass of melanin [21]. However, even though eumelanin is the predominant melanin in the epidermis, pheomelanin may be the major melanin in cultured human melanocytes. The melanin profile changes when the cell is isolated from the epidermis. By contrast, the melanin content of mouse melanoma B16-F10 cell line was reported to be eumelanic. Eumelanin accounts for up to above 97% (wt./wt.) of melanin in mouse melanoma tissue and 95% (wt./wt.) of melanin in mouse melanoma cells [22]. Here we chose B16-F10 live cells for the study on melanin enhanced THG.

We chose the endogenous two-photon-excitation fluorescence (TPEF) signal of melanin as the melanin concentration indicator. Fluorescence spectra of organic substances, as characterized by their spectral shape, excitation spectrum, quantum yield and lifetimes, are inherent properties sensitive to both the excited-state characteristics of the fluorophore itself and the fluorophore matrix. In this study we attempted to find a proper combination of melanin and solvent to have the most similar fluorescence spectrum to the melanin in live B16-F10 cells.

Both natural melanin [23] and synthetic melanin [24] had been used in the literature as a model for eumelanin. Natural melanin and synthesis melanin were dispersed in the solvents respectively and these melanin samples were stored in the eppendorfs. These solutions were shaken for 10 seconds and bath sonicated for 30 minutes at 40°C. We used solvents which were frequently used in studying melanin optical properties, including NaOH [25], MilliQ water (ultrapure water) [26] DMSO [27], and NH₄OH [28].

Mouse B16-F10 melanoma cell line was purchased from the American Type Culture Collection and cultured in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum, 4 mmol/L L-glutamine, 100 µg/mL streptomycin, and 100 units/mL penicillin. To collect natural melanosomes, B16-F10 melanoma cells were cultured until that naked-eye observable melanin content was secreted by melanoma cells and expelled to the medium. We collected the culture media and stored them in centrifuge tubes at 4 ℃.

Nonlinear emission spectra were measured by using a femtosecond Cr:forsterite laser center at 1230 nm, mimicking the parameters of the previous THG skin imaging study [9]. The excitation beam was focused onto the specimen by using a water immersion objective (Olympus 40X/NA1.15). The nonlinear emission signal was collected by using the same objective and was directed into a spectrometer (Andor SR3031B). Figure 1(a) shows the collected nonlinear emission spectrum of the natural melanosomes collected from cultured B16-F10 cells. Besides THG at 410 nm and second harmonic generation at 615 nm, a broad TPEF from melanosome was observed to be centered around 690 nm. Based on the TPEF spectra of dispersed melanin in various solvents measured by using the same system, we found that the synthetic melanin dispersed in 1M NaOH solution provided the most similar TPEF spectrum (Fig. 1(b)). We consequently chose the combination of NaOH and synthetic melanin for further calibration.

 

Fig. 1. Nonlinear emission spectra excited by a femtosecond Cr:forsterite laser of (a) natural melanosomes collected from cultured B16-F10 cells, and of (b) the synthetic melanin dispersed in 1M NaOH solution.

Download Full Size | PDF

To study the relationship between THG intensity and melanin concentration, we configured a multi-modal THG and TPEF microscope. We utilized the TPEF signals to calibrate the melanin-enhanced THG signals in our study. This multi-modal microscope system is based on a Leica two-photon microscope (TCS SP5) integrating a femtosecond optical parametric oscillator (Coherent OPO) to take simultaneous epi-THG and epi-TPEF images of living B16-F10 cells. A live-cell chamber was coupled with the microscope. The environment in the chamber was controlled at 37.0°C, 5% CO₂. The excitation laser source was tuned to a center wavelength of 1230 nm and focused onto the sample using a water-immersion objective (Olympus 40X/NA1.15). Epi-TPEF and epi-THG were collected using the same objective, separated with a dichroic mirror and directed into two different photo-multiplier-tubes. The filters used for TPEF and THG were 650 long pass filter (Chroma) and D410/30x band-pass filters (BPF) (Chroma). The average power on sample was 25 mW. Every image was averaged over three frames. With line scan speed 400 Hz and the image dimension of 512 × 512 pixels, it costed 1.28 s to take one image. Therefore, for one final image, the overall acquisition time was 3.84 s.

3. Results and discussions

With the acquired multi-modal images, each pixel with its THG and TPEF values recorded was treated as one data point in the THG versus TPEF graph. Since melanin was contained in cytoplasm, proper selection of region of interest (ROI) is necessary for our study. Figure 2 shows the selection of the ROI. We collected 41 live cell images, 3,148,481 pixels in total, for statistical analysis. Figure 2(e) summarizes the statistical results. Assuming a positive correlation between the TPEF signals and melanin mass density, we could observe the relation between THG intensities and melanin mass densities. To evaluate the bleaching effect, even under >100s scanning time, stable values of TPEF signals were observed, indicating negligible bleaching effects.

 

Fig. 2. Simultaneous endogenous a. Epi-THG image and b. Epi-TPEF image of B16-F10 live cells. c. The corresponding bright field image. d. A demonstration of ROI selection. e. Statistics of 41 combined TPEF and THG images. Mean values and standard errors are shown. Error bars show standard error. Epi-THG and Epi-TPEF are represented by magenta and yellow pseudo-colors.

Download Full Size | PDF

To further correlate the TPEF pixel values to the melanin mass density (MMD), we acquired TPEF images of artificial melanin samples with different MMDs. Synthetic melanin were dispersed in 1M NaOH solutions. The imaging condition was the same as the studies on melanoma cells, except to be replaced with melanin solutions with different concentrations. At least three sets of data were recorded for each measurement to check the repeatability. We dropped 20µL of the melanin solution on the glass bottom dish which was exactly the same as used to culture B16-F10 cells. Images at several depths were recorded and THG signal here would help position the depth information since its strength would be significantly enhanced at the interface. Stable TPEF values can be observed when the laser focus was totally inside the solutions, while linear relationship can be observed between the MMD values and the TPEF pixel values, as shown in Fig. 3. As a result, the relation between THG strength and MMD can be obtained from Fig. 2(e) by replacing TPEF values with MMD values by using this linear relationship.

 

Fig. 3. Averaged TPEF pixel value of artificial melanin samples with different melanin mass densities. A strong linear relation (r-square = 0.99) between the measured TPEF intensity and melanin mass density can be found (shown as the orange line). The highly linear relationship verifies our assumption that TPEF is proportional to the melanin mass density and a simple linear equation $\textrm{TPEF} = 68 \times \textrm{MMD} + 391$ can be used to estimate the melanin mass density from TPEF pixel values.

Download Full Size | PDF

Our results indicated that when there is no melanin, THG signals exist. It is well-known that cell membrane and various cellular organelles can generate THG [29]. This represents the lowest THG value part of Fig. 2(e) (TPEF pixel value < 391) while these THG values are much lower than the THG values when melanin exists (TPEF pixel value >> 391). We then divided the measured THG pixel value by the average background value without melanin as a relative measure, called THG enhancement ratio (erTHG). Figure 4(a) shows the relation between erTHG and MMD. On the whole, the THG strength rises with the increase of melanin content present inside the focal volume. Melanin with a high local density over 50 mg/ml can enhance the THG background signal up to 25 times. The trend of enhancement is with two phases. For MMD below 11 mg/ml, erTHG grows intensely and nonlinearly, while with even higher MMD, a linear relation between the density and erTHG was observed.

 

Fig. 4. a. Statistics results illustrating the correlation between melanin mass density and THG enhancement ratio. b. A log-log plot of (a) with MMD below 11 mg/ml. The orange line is the linear regression. c. Data and the regression curves of two phases are plotted in the same graph. The intersection of two regression curves is at 11.00 mg/ml melanin mass density with a 5.93 THG enhanced ratio.

Download Full Size | PDF

Here we use a simple power-law behavior to formulate the relation between the erTHG and MMD in these two phases. By linear regression in the log-log plot, we extracted the values of power and coefficient. In phase 1, the THG enhancement ratio rises with 3.47 power of MMD (Fig. 4(b)), while in phase 2, the value of power is more linear-like as 0.95. The regression lines intersect at the point where MMD equals 11.0 mg/ml where the erTHG is 5.93. We thus constructed an empirical equation for the melanin-induced erTHG inside live cells as

(MMD < 11 mg/ml)

(MMD > 11 mg/ml)

Compared with previous result [12] where a linear relationship between melanin concentration and THG intensity was reported, it is noted that not only with a different excitation wavelength, the studied MMD range in this previous study was between 10-40 mg/ml. Similar to previous conclusion [12], when the melanin density was higher than 11 mg/ml, THG signals in a melanin containing cell is found to increase linearly with the MMD in our study. This linear relationship indicates the incoherent nature of the observed THG signal and supports the signal origin as the high-order hyper-Rayleigh scattering [12,18]. Different from previous study, our live cell study indicates an initial nonlinear process in Fig. 4(c) where the THG intensity was enhanced according to the 3.5th power of MMD, indicating the coherent nature of the THG signal and supporting the signal origin as the resonance-enhanced THG [13–17]. When the virtual level of THG matches a real level, nonlinear susceptibility of THG will be resonantly enhanced, resulting in strong THG signals. Due to the coherent nature of the THG electric field, summation of the electric field which is molecule-density-dependent will result in nonlinear increase of the THG power.

Under the excitation with a focused Gaussian beam, the resonance-enhanced coherent THG power ${P_{3\omega }}$ obeys the proportionality relation [30]

With the broadband excitation source around 1230 nm and using an objective lens with NA 1.15, the theoretical beam waist radius is 340 nm and half of the confocal parameter b is 787 nm. The laser focal volume can then be calculated to be not smaller than the theoretical limit of 0.38 fL. The threshold value of 11 mg/ml thus means that there will be at least 4.2fg melanin inside the focal volume. Inside our skin, melanin was confined inside the melanosomes. A previous study summarized the variation of melanosome areas in different skin types to be between 1.44-0.94 x10⁻² µm^-2 [31]. By converting the area information into volume and taking the density of the water for estimation, 11 mg/ml indicates the existence of at least 2.4-4.6 melanosomes inside the focal volume.

As studied and discussed in Ref. [18], in which hyper-Rayleigh scattering in the third harmonic generation was experimentally studied in silver island films and was attributed to the spatial inhomogeneity, the high order hyper-Rayleigh scattering of THG can be considered as the decoherence of the coherent THG signals. As a result, the incoherent THG power from different molecules adds up linearly. Similar to this previous case, in cells when the melanin concentration is high, we observed a transition from a nonlinear relationship into a linear relationship, a signature of high order hyper-Rayleigh scattering. With high concentration of inhomogeneously-distributed melanosomes as the randomly-assembled scatterers inside the focal volume [18], a transition from coherent THG into incoherent THG occurs. As a result, resonance-enhanced third-harmonic generation will then saturate into a high-order hyper-Rayleigh scattering process with an optical mechanism change.

4. Conclusion

In conclusion, as the dominant absorber in skin, melanin not only acts as a broad-band optical absorber but also provides the strongest THG nonlinearity in human skin. We studied how melanin affected the strength of THG in a melanoma cell line. A transition from the resonance enhancement with a real absorption level to the high order hyper-Rayleigh scattering was observed, with melanosomes serving as the scatters. Our observation provide evidences regarding how a virtual-transition-based nonlinear coherent process saturates with high density of scatters: it saturates into a linear high-order hyper-Rayleigh scattering process. If a potential method to retrieve the erTHG value in vivo can be developed, our reported finding might serve as a potential mechanism to realize in vivo melanin 3D imaging in human skin with absolute quantities and microscopic details.