Fingerprinting of skin cells by live cell Raman spectroscopy reveals melanoma cell heterogeneity and cell‐type‐specific responses to UVR

Abstract Raman spectroscopy is an emerging dermatological technique with the potential to discriminate biochemically between cell types in a label‐free and non‐invasive manner. Here, we use live single‐cell Raman spectroscopy and principal component analysis (PCA) to fingerprint mouse melanoblasts, melanocytes, keratinocytes and melanoma cells. We show the differences in their spectra are attributable to biomarkers in the melanin biosynthesis pathway and that melanoma cells are a heterogeneous population that sit on a trajectory between undifferentiated melanoblasts and differentiated melanocytes. We demonstrate the utility of Raman spectroscopy as a highly sensitive tool to probe the melanin biosynthesis pathway and its immediate response to ultraviolet (UV) irradiation revealing previously undescribed opposing responses to UVA and UVB irradiation in melanocytes. Finally, we identify melanocyte‐specific accumulation of β‐carotene correlated with a stabilisation of the UVR response in lipids and proteins consistent with a β‐carotene‐mediated photoprotective mechanism. In summary, our data show that Raman spectroscopy can be used to determine the differentiation status of cells of the melanocyte lineage and describe the immediate and temporal biochemical changes associated with UV exposure which differ depending on cell type, differentiation status and competence to synthesise melanin. Our work uniquely applies Raman spectroscopy to discriminate between cell types by biological function and differentiation status while they are growing in culture. In doing so, we demonstrate for the first time its utility as a tool with which to probe the melanin biosynthesis pathway.


| INTRODUC TI ON
Melanocytes are the pigment-producing cells found in the skin, hair and eyes. Their embryonic precursors are melanoblasts derived from a transient tissue known as the neural crest. 1 Melanocytes manufacture melanin from tyrosine and phenylalanine using the enzymes phenyalanine hydroxylase (PAH), tyrosinase (TYR), dopachrome tautomerase (DCT) and tyrosinase-related protein 1 and 2 (TYRP1 and TYRP2). 2 They use their dendrites to export melanincontaining melanosomes to keratinocytes which arrange them above their nuclei to protect from ultraviolet radiation (UVR) induced DNA damage. 3,4 Melanoma, the cancer of melanocytes, is the deadliest form of skin cancer with a rising worldwide incidence. 5,6 Melanoma risk is increased by exposure to UVR from the sun and from tanning beds. 7,8 Mutations in melanocytes arise from DNA damage caused either directly by UVB irradiation or indirectly following UVA irradiation through reactive oxygen (ROS) species via photosensitiser-mediated processes. 9 Solar UVR is composed of mainly UVA (320-400 nm wavelengths) with a lesser component of UVB (280-320 nm wavelengths) and a UVA/UVB ratio of about 20 depending on latitude and time of day. 10 Melanocytes respond differently to UVA and UVB. UVA induces immediate pigment darkening through photooxidation of melanin that can be observed within minutes of exposure. 11,12 UVB on the contrary induces epidermal melanocyte proliferation [13][14][15] and activation of melanogenic enzymes resulting in a delayed tanning response occurring 2-3 days postexposure. 12 UVA and UVB also cause ROS-mediated lipid peroxidation, [16][17][18] and widespread oxidative modification of proteins resulting in their proteasomal degradation. 19 Raman spectroscopy enables a non-invasive label-free analysis of the biochemical structures present within a sample. 20,21 Laser light is used to excite a sample and changes in scattering provide information about the molecular structures present. 22 Its utility has already been demonstrated in identifying and discriminating between the molecular components of skin, 23 and it has shown promise in the automated diagnosis of skin neoplasms in vivo including melanoma. 24,25 Although previous studies have examined normal and melanoma biopsy tissue and compared melanocytes and melanoma cells, 20,26 none have examined melanocyte function, differentiation status or their UV response. Here, we use live single-cell Raman spectroscopy and principal component analysis (PCA) to biochemically fingerprint melanoblasts, melanocytes, keratinocytes and melanoma cells. We identify the principal biomolecules that underlie these fingerprints and demonstrate that Raman spectroscopy is a highly sensitive method of probing the melanin biosynthesis pathway and its response to UVR.

| UV irradiation
Cells were seeded onto CaF 2 disks in 24-well plates using phenolfree media. Cells were irradiated with 100 KJ/m 2 UVA, 1000 J/m 2 UVA and UVB or 100 J/m 2 UVB and spectra acquired at 1, 3, 6, 16 and 24 h postirradiation as well as from an untreated control.
The UVA source consisted of an array of Philips TLR 36 W "blacklights" tubes. Wavelengths below 320 nm were filtered out using polyester film (No 130 clear; Lee Filters, UK, spectrally equivalent to Mylar). The spectral irradiance ranged from approximately 330-400 nm with a peak output at 365 nm and an intensity of around 60 Wm −2 . The UVB source consisted of an array of Phillips T140 UVB tubes, emitting a broad spectral irradiance ranging from 275 to 380 nm with peak output at 315 nm and an intensity of 5.8 Wm −2 . The UVA/UVB source consisted of an array of solarsimulating Q-Lab QUV UVA-340 tubes (Q-Lab, UK). Spectral irradiance ranges from 295 to 420 nm, with a peak output of 340 nm and an overall profile in the 295-365 nm range that is similar to noon summer solar output, and an intensity of 23 Wm −2 . Output spectra and intensities of all UVR sources were measured using a double monochromator spectrophotometer (model SR991-PC; Macam Photometrics, Livingston, UK). For UVA treatments, where the prolonged exposure times required could lead to heating of samples, tissue culture plates were placed on water-cooled metal plates.

| Raman spectroscopy
Raman spectra were acquired at 60× using a 785 nm laser (200 mW at source, ~10 mW at sample) at a spectral range of 600-1700 cm −1 using an InVia Raman microspectrometer (Renishaw plc, Gloucestershire, UK) equipped with an environmental chamber (Okolab, Ottaviano, NA, Italy) maintaining a temperature of 37°C in humidified air containing 5% (v/v) CO 2 . Collection was at 2 s with nine accumulations (18 s total) at 100% laser power. A total of 18 spectra were collected across three independent runs with spectra from six individual cells collected each run. Cells were selected randomly by moving diagonally along the sample and collecting spectra every 10th cell to ensure spread across the disk and that no overlap occurred. Data collected were baseline corrected using a polynomial method order 3, smoothed using

| Cell viability
Viability was measured using Cell Titre Glo (Promega). CaF 2 disks were transferred to a 24-well plate, along with control disks that had not been used for spectral acquisition. Diluted Cell Titre Glo was added to cells for 10 min before scraping and transfer to a white 96-well plate. Bioluminescence was measured as relative light units using a Perkin Elmer Wallac 1420 Victor2 microplate reader.

| Intracellular melanin measurements
Cells were UV-irradiated as described above. Subsequently, they were washed in PBS, trypsinised and counted, then pelleted at 1000 RPM for 3 min. 100 μl of melanin lysis buffer (90% 1 M NaOH and 10% DMSO) was added to each pellet before incubation at 80°C for 90 minutes. Lysed pellets were transferred to a 96-well plate and their absorbance read at 490 nm on a Perkin Elmer Wallac 1420 Victor2 microplate reader. A standard curve was obtained by diluting synthetic melanin (Sigma, Poole, UK) at 0-1 μg/ml to enable the final melanin concentration to be determined.

| Statistics
All statistical tests were performed in SPSS or using the 'R' statistics package (http://www.R-proje ct.org). The Raman data in Figures 2-4 were normalised individually for each peak across all cell types and conditions. A one-way analysis of variance (ANOVA) was performed across the six time points for each combination of cell type and UVA treatment followed by pairwise comparisons by Tukey's Honestly Significant Difference test (TukeyHSD). Therefore, both a significant ANOVA p value and a significant TukeyHSD p value are required for a significant pairwise difference to be considered. For the RT-qPCR data in Figures 2 and 3, data were analysed using SPSS. A one-way analysis of variance (ANOVA) was performed across the three time points for each cell type and UVR treatment; this was followed by pairwise Dunnett's post hoc tests. Significant differences between control and time were evaluated with *p values ≤ 0.05 as indicated on the graphs.  Table S1. Reactions were performed on a BioRad CFX RT-qPCR machine using the following parameters: 50°C for 2 min and 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Cycle threshold (C T ) values were calculated for each mRNA sample and compared to their respective actin control to determine gene expression changes using the comparative CT (2 −ΔΔC T ) method. 29

| Live single-cell Raman spectroscopy can discriminate between skin cell types and melanocyte differentiation status
We modelled keratinocytes, melanoblasts, melanocytes and melanoma cells using the COCA, melb-a, melan-a and B16F10 cell lines, respectively ( Figure S1a). [30][31][32][33] Live single-cell Raman spectroscopy was used to acquire individual spectra allowing us to establish their biochemical fingerprints. We confirmed viability of the cells after imaging using a CellTiter-Glo assay ( Figure S2a  were used to identify the principal Raman peaks responsible for the PCA groupings which we attributed to their likely biomarkers using the established literature (numbered 1-10 in Figure S1b and Table S1). Nine of the 10 peaks had been previously described and were consistent with the biochemical properties of the cell types examined including melanin biosynthesis (phenylalanine, tyrosine, melanin) photoprotection (β-carotene) or keratinocyte function (keratin/lipid) (Table S1). We identified a peak at 1333 cm −1 ( Figure   S1b) in melanocytes and melanoblasts not previously attributed to a known biomarker but consistent with the Raman spectra for the melanin precursor DOPA ( Figure S3a,b).
Undifferentiated melanoblasts and keratinocytes did not exhibit melanin peaks while the pigmented melanocytes and melanoma cells did ( Figure S1b). Phenylalanine was apparent in melanoblasts, keratinocytes and melanoma cells but not in melanocytes presumably because melanin synthesis depletes the phenylalanine precursor in melanocytes. We observed DOPA peaks in melanoblasts and keratinocytes consistent with previous reports that they can convert tyrosine to DOPA. 34-36

| Immediate and temporal effects on melanin biosynthesis pathways after UV irradiation detected by Raman spectroscopy
Melanin biosynthesis proceeds through a stepwise conversion of phenylalanine through tyrosine and DOPA. 37 As demonstrated above, melanocytes and melanoma cells are competent to perform all of these enzymatic steps while melanoblasts appear able to synthesise tyrosine but are unable to convert DOPA to melanin. 31 Although the enzymatic pathways are well delineated, the immediate and short-term biochemical response to UV exposure are still poorly understood. We therefore focused on the melanin biosynthesis pathway immediately after UV exposure by comparing log 2 fold change (Log2-FC) in mean Raman peak height (n = 18 spectra in all cases) from control for the biomarkers identified above (Figure 1 and Table S1) in response to irradiation from a UVA, UVB and a mixed UVA/UVB light source over a one to 24-h response period.

| Opposing effects of UVA and UVB irradiation on the melanin biosynthesis pathway in melanocytes
Melanocytes were the most responsive to UVR. We observed a rapid (within 1 h) and transient (up to 6 h) statistically significant increase in the phenylalanine peak and a rapid (within 1 h) and transient (up to 6 h) statistically significant decrease in the tyrosine peak after UVA exposure ( Figures 1A and S4a). In parallel, we observed

| Similar effects of UVA and UVB irradiation on the melanin biosynthesis pathway in melanoma cells
While melanocytes appear to respond differently to UVA vs UVB, melanoma cells demonstrated similar responses to the three light sources. We observed a rapid (within 1 h) and sustained (at least 24 h) statistically significant increase in the phenylalanine peak post-UVA irradiation ( Figure 2A) and a delayed (between 6 and 24 h) and sustained (at least 24 h) statistically significant increase in the phenylalanine peak post-UVB and UVA/UVB irradiation ( Figure 2B,C).
This was accompanied by a statistically significant decrease in the melanin peak between one and 24-h post-UVB and UVA/UVB exposure ( Figure 2B,C). The reduction in the melanin peak was the most pronounced after UVB treatment ( Figure 2B), and we observed a corresponding reduction in the gene expression of Tyrp1 at 16 and 24 h for UVB but not UVA suggesting a block in melanin synthesis ( Figure 2D-E) similar to the one observed in melanocytes (melan-a) after UVA treatment above. The changes in melanin levels detected by Raman after UVA irradiation were mirrored by similar trends in intracellular melanin measurement between 0 (control) and 1-3 h ( Figure 2F).

| Minimal effects of UVA and UVB irradiation on the melanin biosynthesis pathway in melanoblasts and keratinocytes
We observed a subdued UV response in the melanin biosynthesis pathway in melanoblasts. Following UVA exposure, we observed a delayed (between 16 and 24 h) statistically significant reduction in the phenylalanine and the tyrosine peaks ( Figure 3A

| Differentiated melanocytes but not melanoblasts, melanoma cells or keratinocytes demonstrate a photoprotective responses to UVR
Melanocytes have been shown to preferentially absorb β-carotene in culture. 39 Consistent with this, we observed a distinct β-carotene Raman peak only in melanocytes ( Figure S1b, Table S1). In response to UVR, we observed a statistically significant increase in this βcarotene peak (between one and 24 h post-UV) in response to UVA, UVB and mixed UVA/UVB light sources ( Figure 4B). We therefore hypothesised that β-carotene may be performing a cell-type-specific photoprotective function protecting proteins and lipids from oxidative damage. We investigated changes in the Raman peaks for lipids (1300 cm −1 peak) and amide I (protein peaks at 1650 cm −1 ) in our Raman spectra as biomarkers of oxidative damage ( Figure 4A-D).
We observed a rapid (1-3 h) and statistically significant reduction in the amide I peak in melanoblasts in response to UVA and UVB ( Figure 4A) which appeared absent or dampened in melanocytes ( Figure 4B). Furthermore, we observed a rapid (within 1 h) statistically significant reduction in the lipid peak in response to UVA  Figure 4C). Lipid levels have previously been shown to increase in human keratinocytes post-UVA and UVB dual irradiation. 40 Consistent with this, here we observed rapid (from 1-3 h) and sustained (at least 24 h) statistically significant increases in the lipid peak in response to UVA, UVB and our mixed UVA/UVB light source in keratinocytes ( Figure 4D).

| Melanoblasts, melanoma cells and keratinocytes but not melanocytes demonstrate changes in keratin levels on UV exposure
Raman peaks consistent with keratin were present in keratinocytes and melanoblasts. We observed a rapid (within 1 h) and sustained (up to 24 h) statistically significant increase in the Raman peak for Significance level for a one-way analysis of variance (ANOVA) is indicated at the end of each series (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001) keratin in keratinocytes after UVA treatment with similar rapid (1-3 h) and transient (<6 h) statistically significant increases after UVB and combined UVA/UVB treatments in these cells ( Figure 4D).
Melanoblasts displayed a significant reduction in keratin after individual UVA and UVB treatments but an immediate (from 1 h) and sustained (up to 24 h) statistically significant increase in keratin after exposure to the dual UVA/UVB light source ( Figure 4A). Melanoma cells but not melanocytes also exhibited a keratin peak and demonstrated a delayed (6-24 h) statistically significant increase after UVB treatment ( Figure 4C).

| DISCUSS ION
We demonstrate the utility of live Raman spectroscopy performed on cells grown in a controlled environmental chamber, as a unique tool to probe the melanin biosynthesis pathway and its immediate response to UVR and reveal rapid and opposing responses to UVA and UVB irradiation by melanocytes.

| Raman fingerprinting demonstrates that melanoma cells are phenotypically heterogeneous
Previous reports have discriminated between malignant melanoma cells and their wildtype counterparts using Raman spectroscopy. 20,26 Our Raman spectra appear consistent with those previously reported for melanocytes and melanomas cells, 22,26,41 while our spectra for melanoblasts and keratinocytes are consistent with their inability to synthesise melanin 31

| Rapid and opposing effects of UVA and UVB irradiation on the melanin biosynthesis pathway in melanocytes
We show here the utility of Raman spectroscopy to detect and sensitively measure changes in the melanin biosynthesis pathway in response to UVR over a smaller time-frame than is possible with conventional methods. UVA exposure has long been understood to result in immediate pigment darkening within a few hours of exposure by acting on pre-existing melanin and by de novo melanin synthesis. 45,46 The reduction in melanin post-UVA and accumulation of phenylalanine observed here may be consistent with a rapid tanning response resulting in the export of melanosomes from the cytosol and a subsequent block in synthesis because of their loss leading to accumulation of phenylalanine, consistent with reports UVA induces a redistribution of melanosomes in the skin. 47 However, we do not see the rapid accumulation of melanin overserved in human epidermal melanocytes. 45 UVB has long been demonstrated to cause a delayed tanning response mediated through DNA damage-induced de novo tyrosinase production. 48,49 We were surprised to see such a rapid and profound UVB-mediated response in the present study characterised by rapid and sustained increases in tyrosine, DOPA and melanin and reductions in phenylalanine suggesting that a previously overlooked mechanism independent of DNA damage acts immediately after UVB exposure in mouse melanocytes.

| Melanocytes uniquely harbour β -carotene with a possible role in photoprotection
The photoprotective effects of β-carotene have been attributed to its free radical/reactive oxygen species (ROS) scavenging abilities and direct absorption of UV at 400 nm. 50,51 We observe melanocytespecific accumulation of β-carotene consistent with previous reports 39 and report here that β-carotene levels increase significantly in response to UVA, UVB and a mixed UVA/UVB light source. The presence of β-carotene also appeared to correlate with an apparent stabilisation in the peaks of amide I and lipid in melanocytes, whereas these biomarkers appeared to change dramatically in response to UVR in the other cell types studied. Taken together, these finding suggest that melanocytes use β-carotene as part of a UVR protective mechanism shielding against UV-induced lipid peroxidation, DNA damage and protein oxidation as previously reported 51 .

| CON CLUS ION
We have defined a reference biochemical signature for each of the cell types studied that can be attributed to their biological function.
In doing so, we have demonstrated the utility of Raman spectroscopy as a tool with which to probe the melanin biosynthesis pathway with greater temporal resolution and sensitivity than conventional methods. We have used this approach to uncover hitherto undescribed immediate responses to UVA and UVB irradiation in melanocytes including a possible photoprotective role for β-carotene. The ability to probe biochemical status with high temporal resolution will be highly beneficial to continued research in the study of melanoma and the further development of sensitive screening and early detection methods.

This work was supported by a project grant from North West Cancer
Research (Grant CR1132).

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information can be found online in the Supporting Information section at the end of this article.