An Endogenous Electron Spin Resonance (ESR) Signal Discriminates Nevi from Melanomas in Human Specimens: A Step Forward in Its Diagnostic Application

Given the specific melanin-associated paramagnetic features, the Electron Spin Resonance (ESR, called also Electron Paramagnetic Resonance, EPR) analysis has been proposed as a potential tool for non-invasive melanoma diagnosis. However, studies comparing human melanoma tissues to the most appropriate physiological counterpart (nevi) have not been performed, and ESR direct correlation with melanoma clinical features has never been investigated. ESR spectrum was obtained from melanoma and non-melanoma cell-cultures as well as mouse melanoma and non-melanoma tissues and an endogenous ESR signal (g = 2.005) was found in human melanoma cells and in primary melanoma tissues explanted from mice, while it was always absent in non-melanoma samples. These characteristics of the measured ESR signal strongly suggested its connection with melanin. Quantitative analyses were then performed on paraffin-embedded human melanoma and nevus sections, and validated on an independent larger validation set, for a total of 112 sections (52 melanomas, 60 nevi). The ESR signal was significantly higher in melanomas (p = 0.0002) and was significantly different between “Low Breslow’s and “High Breslow’s” depth melanomas (p<0.0001). A direct correlation between ESR signal and Breslow’s depth, expressed in millimetres, was found (R = 0.57; p<0.0001). The eu/pheomelanin ratio was found to be significantly different in melanomas “Low Breslow’s” vs melanomas “High Breslow’s” depth and in nevi vs melanomas “High Breslow’s depth”. Finally, ROC analysis using ESR data discriminated melanomas sections from nevi sections with up to 90% accuracy and p<0.0002. In the present study we report for the first time that ESR signal in human paraffin-embedded nevi is significantly lower than signal in human melanomas suggesting that spectrum variations may be related to qualitative melanin differences specifically occurring in melanoma cells. We therefore conclude that this ESR signal may represent a reliable marker for melanoma diagnosis in human histological sections.


Introduction
Skin melanoma is one of the most aggressive tumours in humans, showing high mortality at the metastatic stage, and increasing incidence worldwide. [1] Melanoma accounts for about 4% of skin cancers, causing however about 80% of skin cancerrelated deaths in western countries. Despite promising recent improvement at the therapeutic level, [2][3][4][5][6] surgical excision remains at this moment the most effective treatment at early stages, while therapeutic interventions have weak efficacy at advanced phases due to high metastatic potential and resistance to currently available therapies. [7].
Improving early diagnosis may therefore strongly affect melanoma-related mortality. Melanoma diagnosis routinely starts from non-invasive dermatoscopy-and epiluminescence-based skin inspection, to identify phenotypic features of the pigmented lesion. [8] Trained dermatologists still experience significant error-rate giving misdiagnosis and delay in treatment, and formal diagnosis still requires to be confirmed by histological analysis. [9] Hence, alternative non-invasive procedures are needed to improve the early non-invasive diagnostic accuracy.
Several reports indicate melanogenesis as a key process in the melanoma biology. [10,11] Melanin synthesis involves an oxidation/reduction reactions chain leading to the synthesis of final organic polymers. The intermediate free radicals formed within such process [12] give melanin paramagnetic properties. [13] Besides free radicals, melanin may also contain or interact with metal ions and paramagnetic gases (dioxygen, nitric oxide) which also contribute to its paramagnetic properties. [14] ESR spectroscopy is the technique of choice to detect and to investigate free radicals. As such, it has long been used in melanin basic research since melanin ESR signal is stable, [15][16][17] resistant to chemical degradation, [18] and different in eumelanin from pheomelanin. [19,20].
The main goal of the present study was to provide strong support to the use of ESR spectroscopy as a reliable diagnostic help in melanoma management. To this aim we identified an endogenous ESR signal (g = 2.005) in melanoma and nonmelanoma human cell lines, then investigated this signal in mouse melanoma tissues. Finally, we investigated ESR signal in human melanoma specimens compared to human nevus specimens. A specific ESR signal was found in melanoma human tissues, significantly different from the one recorded in nevus paraffinembedded specimens; ROC analysis showed that ESR signal is able to discriminate human melanoma sections from nevi, with very high accuracy.
Human adult low Calcium Temperature keratinocytes (HaCaT) were a kind gift of Dr. Pastore (IDI-IRRCS, Rome) [30] and were grown in DMEM medium with FBS 10% final concentration.
Cells were grown in 75 cm 2 flasks and media was changed every other day. Once a 75% confluence was reached, cells were trypsinized, harvested by centrifugation (10 minutes at 1500 RPM), washed and transferred into glass flat capillaries. ESR analysis was performed on intact cells; viability assay by trypan blue exclusion test always indicated at least 98% of live cells. Commercially available melanoma cells were used at maximum fifth passage, unless differently specified.

In vivo Mouse Melanoma Model
For in vivo mouse experiments, murine B16F10 melanoma cells were purchased from American Type Culture Collection (ATCC, Manassas, VA) (ATCC number CRL-6475) and grown in DMEM (Hyclone, Logan, UT) with 10% FCS (Hyclone, Logan, UT) [5]. Media were completed by the addition of glutamine (2 mM) and penicillin/streptomycin (50 U/ml-50 mg/ml) (Gibco, Carlsbad, CA). Cells were grown at 37uC with 5% CO 2 and subsequently injected in the dorsal skin of 20 weeks-old male C57BL/6 mice (number of animals = 5) according to a previously reported procedure [4]. Primary melanomas were removed 2 weeks after cell-inoculation and kept on ice in a PBS solution until ESR analysis.
Human Paraffin-embedded Tissue Sections 112 human paraffin embedded melanoma and nevus specimens were prepared as previously described [31,32]. Two experimental sets were analyzed: 26 paraffin-embedded specimens (40 microns slides) of human nevus or melanoma (13 nevi and 13 melanomas) were assigned to the ''Measuring set'' and were analyzed first. Then a second independent set of 86 paraffin-embedded specimens (47 nevi and 39 melanomas) was assigned to the ''Validation set'' and analyzed. ESR measurements were carried out according to the methodology reported below. Paraffin embedded slices were weighed and ESR data were all normalized accordingly. Slice of pure paraffin revealed no ESR signal.

Ethics Statement
The data were collected within a study approved by the local Ethic Committee (IDI IRCCS and San Carlo Hospital Ethical Committee Protocol, July 19th 2005; Reg. N. 154); written informed consent was obtained and all data were analyzed anonymously.

ESR Analysis
For ESR spectra on intact cells, signals were recorded using 80ml samples in flat glass capillaries (inner dimensions 0.464.0 mm) to optimize instrument sensitivity as previously described for cells in aqueous suspension. [33] Measurements on cells were performed at room temperature with an ESP300 X-band instrument (Bruker, Karlsruhe, Germany) equipped with a high sensitivity TM 110 -mode cavity. Spectra were measured over a 200 G range using 20 mW power, 2.0 G modulation, and a scan time of 42 s; 4 single scans were accumulated to improve the signal-to-noise ratio. Qualitative measurements of tissues and human paraffin-embedded sections were performed at room temperature in circular glass capillaries (inner diameter 1.10 mm) using the apparatus and experimental settings described above. Twenty four single scans were accumulated to improve the signalto-noise ratio.
Quantitative measurements of the samples belonging to the ''Measuring set'' and ''Validation set'' were carried out on a different instrument (Bruker Elexys E500 X-band, equipped with a super-high sensitivity probe head) [34,35]. Such measures were carried out over a 100 G range using 20 mW power, 3.0 G modulation, and a scan time of 42 s; 64 single scans were accumulated to improve the signal-to-noise ratio. The amplitude of the field modulation was preventively checked to be low enough to avoid detectable signal overmodulation.
The other experimental parameters have been set as follow: conversion time : 83.69 ms, time constant :163.84 ms, receiver gain 60 dB, number of points:1024. For selected samples signal saturation was checked to be reached above 60 mW microwave power.
The g value has been evaluated by means of an internal standard (DPPH). In details, DPPH was inserted in a very thin capillary. In turn, this capillary was inserted in the measuring test tube co-axially with the investigated samples.
ESR quantitative data were expressed both as peak-to-peak amplitude and as double integral intensity; linewidth of all samples was also measured.
In each sample of paraffin embedded samples, the ratio between the height of the major peak (a) and the height of a weak shoulder at lower field (g < 2.01) (b) has been measured. This ratio is reported to correlate in a linear manner with the proportion between eumelanin and pheomelanin monomers in a copolymer [18,20].

Statistical Analysis
For statistical analysis, the entire set of paraffin-embedded samples was divided in groups and subgroups, according to different parameters (diagnosis, sex, body location of lesions, Breslow's depth) ( Table 1). The statistical analyses were performed using the Graph-Pad Prism 5 software; D'Agostino and Pearson normality Test was performed and groups showing normal distribution were analyzed with T test, while groups showing not-normal distribution were analyzed by Mann-Whitney Test; two-tailed p,0.05 was chosen as significance threshold.
ANOVA analysis was carried out with the ''Bonferroni Multiple Comparison Test''; two-tailed p,0.01 was chosen as significance threshold. Spearman correlation was also performed to compute ESR signal amplitude correlation with Breslow's depth (expressed in mm) in all melanoma samples.
ROC analysis was also carried out to measure the ability to discriminate nevi from melanoma subgroups.

ESR Spectra in Melanoma Cell Cultures
Six human melanoma cell lines from primary and metastatic melanomas were analyzed. Three melanoma cell lines (SKMEL-110, SKMEL-28 and SKMEL-2) showed an ESR signal (g = 2.005) (Fig. 1A). Spectra obtained from SKMEL-110 and SKMEL-28 were stable and intense, by repeated measure of the ESR signal at hours of distance. On the contrary, spectra from SKMEL-2 showed a faint peak, which disappeared when the measure was repeated after one hour (data not shown). No signal was detected in three other melanoma cell lines analyzed (namely, SKMEL-3, SKMEL-31 and C32) (spectra not shown). Fig. 1A indicates that SKMEL-28 cell-line shows a remarkable ESR signal at the 5 th culture passage, while the same signal is lost at the 10 th culture passage.
Human endothelial cells (HUVEC), human keratinocytes (HaCaT) and human primary melanocytes were used as controls and did not show the ESR signal found in melanoma cells (Fig. 1B).

ESR Spectra in Fresh Samples of Primary Mouse Melanomas and Healthy Tissues
Freshly excised primary mouse melanomas were then collected from 5 different mice, previously inoculated subcutaneously with B16F10 cells (according to previously published protocol) [4]. ESR scanning was then carried out onto such samples under identical spectral conditions as reported for cultured cells. The analysis confirmed the presence of a strong ESR signal matching the one observed in melanoma cell lines. The signal was intense and stable when measured again at room temperature after 14 days of sample storage at 280uC (Fig. 2A). Liver, kidney and heart tissues taken from the same animals were used as controls, and a weak and broad ESR signal was recorded, different from the sharp signal found in mouse melanomas (Fig. 2B).

ESR Spectra in Paraffin-embedded Sections of Human Melanomas and Human Nevi
ESR spectra were then collected in human melanoma paraffinembedded specimens and in human nevus paraffin-embedded specimens (40 microns each), in order to perform more quantitative analyses and verify the hypothesis that ESR may help discriminate melanoma specimens from healthy controls.
A preliminary qualitative analysis of paraffin-embedded nevi and melanomas indicated that an ESR signal is present in human specimens, corresponding to the ESR signal observed in mouse melanoma tissues and that the signal is lower in nevi than in melanomas (Fig. 2C).
A quantitative analysis was then carried out on a group of 26 formalin-fixed paraffin-embedded blocks of human skin melanomas and nevi; this samples-group was named ''Measuring Set''.
To validate such analysis, an independent larger samples set (named ''Validation set'') of human melanomas and nevi was investigated (N = 86) using the same instrument and the same set up. Results shown in Fig. 3A (reported as mean 6 SEM) indicate similar data in the two sets, namely they indicate that nevi of the ''Measuring set'' show no significant difference vs nevi of the ''Validation set'', and melanomas of the ''Measuring set'' show no significant difference vs melanomas of the ''Validation set'', allowing us to conclude that two independent samples sets are not significantly different. Further, in either sets melanomas show a signal higher than nevi.
Therefore the two sets were combined in one all-inclusive set (named ''All set'') to achieve enough sample numerosity and to further analyze demographic and clinical features such as sex, body-location and tumour thickness (Table 1). Nevus and melanoma samples of the ''All set'' were divided in subgroups according to sex and lesion body location (''Trunk'', ''Limbs'' and ''Head and Neck''). Mann-Whitney Test revealed that in all subgroups (except ''Limbs'' location) a significantly different signal was found between nevi and melanomas (p#0.05). The superimposition of the selected peak of 8 nevi and 8 melanomas is reported in Figure S1.
An additional ANOVA analysis confirmed the highly significant difference of the melanomas ESR signal with ''High Breslow's depth'' vs nevi and melanomas ''Low Breslow'' (Fig. 4B).
All calculations reported in Fig. 3 and Fig. 4 were carried out on amplitudes values; each calculation has also been performed on double-integral values reaching almost superimposable results as compared to amplitudes (Fig. 5).
A correlation analysis by Spearman Test carried out in the 52 melanoma samples indicated a strongly significant correlation (R = 0.57; p,0.0001) between ESR signal amplitude and the corresponding Breslow's depth value expressed in millimetres. Similar results were observed using integral values (R = 0.42; p = 0.002).
The variation of the eumelanin/pheomelanin ratio (a/b) (see methods) was also investigated indicating a significant difference of ROC analysis was then carried out to test the ability of ESR signal to discriminate nevi from melanomas in paraffin-embedded sections. The computed area under the ROC curve quantifies the ability to discriminate controls from melanoma patients taking into account both sensitivity and specificity. A value of 1 indicates the ability to discriminate 100% of patients from controls and corresponds to a curve mostly left-shifted in the graph. According to such analysis, nevi were effectively discriminated from melanomas, showing a ROC area of 0.70 (corresponding to 70% accuracy; p,0.0001; Fig. 6A). Nevi were also effectively discriminated from melanomas ''High Breslow's depth'', with a ROC area of 0.81 (corresponding to 81% accuracy; p,0.0001; Fig. 6C). Melanomas ''Low Breslow'' were effectively discriminated from melanomas ''High Breslow'' with a ROC area of 0.86 (corresponding to 86% accuracy; p,0.0001; Fig. 6D). No significant discrimination was detected between nevi and melanomas ''Low Breslow'' (Fig. 6B).
These data were achieved using amplitude values, which have a low coefficient variation. When double integral values were used for ROC analyses, the ROC area was lower, while the statistical significance remained very high (p,0.005 in all cases) ( Figure S2).

Linewidth Measure
It has been reported that peak-to-peak amplitude is an indicative quantitative measure of the ESR spectra as effective as the corresponding double integral, provided linewidth remains constant. We therefore measured the linewidth of ESR peaks in all samples of the ''All set''; the values measured in nevi were much similar to melanomas (8.0 G vs 7.8 G, p = 0.52) as well as in melanomas ''High'' vs melanomas ''Low'' Breslow (7.8 G vs 7.9 G, p = 0.4); the entire nevi/melanomas population showed a mean linewidth of 7.9 G 60.15 SEM.

Discussion
Significant advances have been recently achieved on molecular melanoma markers [36][37][38]; however, clinical application to noninvasive melanoma diagnosis is still far to come. Early diagnosis is a key request in aggressive tumours such as melanoma; currently, trained dermatologists perform visual-and epiluminescence-based macroscopical inspection, and suspect melanoma is then confirmed by histological examination. ESR-based in situ analysis of pigmented skin lesions has been suggested as a possible noninvasive diagnostic tool, according to results obtained on melanoma cell lines and melanoma tissues [29,[39][40][41] and in animal models of melanoma [39,42]. As possible melanoma markers, stable melanin granule-connected [29] or cellular membrane-associated [43] or UV-induced short living melaninand oxygen-derived ESR signals have been proposed [44,45]. However, a statistical analysis of endogenous ESR signal on human melanomas and nevi samples collection has never been performed.
In the present study different melanoma cell lines are reported to exhibit endogenous ESR signal at g = 2.00560.001 in the spectrum obtained under specific physical settings while other melanoma cell lines do not present the ESR signal.
Such difference may be related to heterogenic features (e.g. primary or metastatic) and growth related properties (e.g. doubling time) of cultured melanoma cell lines, reflecting the clinical variability observed in melanoma patients [46]. Noteworthy in our experimental in vitro conditions the signal recorded in SKMEL-28 at the 5 th culture passage was lost at the 10 th culture passage. We hypothesize that this signal loss may be related to the cell senescence and modification of oxidative stress, processes intrinsically linked to melanogenesis [47].
The physical parameters of the measured ESR signal strongly suggested its connection with melanin [17,20,48]. Eu-and pheomelanogenesis occur under low-and high-GSH levels conditions, respectively [49,50], and changes of GSH levels and increase of oxidative cellular stress are typically observed in melanoma [51].
The presence/absence of ESR signal in different cellular systems may thus depend on factors including senescence level, oxidative stress level, melanogenesis and eu/pheomelanin proportion.
In human primary melanocytes (1.2610 6 cells) showing a dark cellular pellet, no ESR signal was recorded (Fig. 1B); on the contrary, melanoma cells SKMEL-28 and SKMEL-110 (1.0610 6 cells) show an intense ESR signal (Fig. 1A). Interestingly, previous data [52] report a clear ESR signal in melanocytes, in apparent contrast with our findings; however it should be noted that such signal was recorded on 100610 6 cells, i.e. a number of cells almost hundred times higher as compared to our experimental conditions, indicating that detection of an ESR signal in dark melanocytes strongly depends on the number of cells investigated.
It is well known that melanocytes visual pigmentation consistently correlates with eumelanin but not always with pheomelanin levels [53] and that some melanoma cell lines with high pheomelanin content has not visible pigmentation [54]. Consistently with all such considerations, it is reasonable to speculate that ESR analysis may detect qualitative alteration of the eu/ pheomelanin proportion, likely independent form the visual pigmentation and absolute eumelanin content.
The same ESR signal was specifically found in freshly drawn mouse melanoma-tissues as well as in paraffin-embedded primary human melanoma-tissues (Fig. 2) while it was absent in several mouse non-melanoma tissues (Fig. 2B). This signal appeared to be stable for at least 14 days upon freezing, suggesting that the conditions of preservation of the mouse tumours allowed successful measurements repeated in time.
ESR signal recorded in human paraffin-embedded nevi was markedly lower than ESR signal recorded in human melanoma sections (Fig. 2C), further supporting the hypothesis that, under our experimental conditions, endogenous ESR signal may be related to qualitative melanin differences occurring particularly in melanoma. [10,[55][56][57][58]. The fine analysis of ESR spectra may reveal the specific proportion of eumelanin and pheomelanin (a/b ratio). Comparison of the a/b ratio in melanomas ''Low Breslow'' and ''High Breslow'' and in nevi vs ''High Breslow'' melanomas in the present study indicated a significant difference between the copolymers composition in these patients groups.
It is therefore reasonable to hypothesize that melanin in melanoma cells undergoes a qualitative change associated with deregulation of eumelanin/pheomelanin ratio, leading to paramagnetic melanin-based radicals accumulation [59]. The significant increase of a/b ratio in melanomas ''High Breslow'' (Fig. 4C) also suggests a contribution of pheomelanin in melanoma [20]; this contribution increases with the progression of melanoma indicating an interesting field of future investigations.
A strong statistical difference of the ESR-signal was measured in human melanoma specimens versus nevus samples; this was observed in any subgroup analyzed (such as body location and sex) except in ''Limbs'' subgroup, indicating that in most cases ESR is consistently and significantly higher in melanomas than in nevi (Fig. 4A).
When all nevi were compared to ''Low Breslow'' melanomas and ''High Breslow'' melanomas, ANOVA analysis showed a significant difference as function of Breslow's depth (Fig. 4B) indicating that ESR analysis may discriminate nevi from melanomas as well as ''Low Breslow'' from ''High Breslow'' melanomas, while it is unable to discriminate nevi from melanomas ''Low Breslow''. Most interestingly Spearman's correlation test confirmed such observation, demonstrating a very significant positive correlation between ESR signal and Breslow's depth, computed with either amplitude and integral values. These observations prompted us to suggest a potential application of ESR-spectroscopy to melanoma diagnosis; such hypothesis was then verified by ROC analysis (Fig. 6), showing a strong and highly significant discriminating ability of ESR signal to identify melanomas from nevi.
ESR technique has been previously suggested for diagnosis and employed in melanoma research [41,42], however the present study is the first reporting a clear association of a specific ESR signal to a large number (n = 52) of human melanomas using a large number of healthy controls (n = 60 nevi). Furthermore, a different eu/pheomelanin ratio in nevi vs melanomas ''High Breslow'' has been shown here for the first time, strongly supporting that qualitative melanin changes may occur in nevi as compared to melanomas with worst prognosis.
The quantitative information of ESR spectra is usually expressed in arbitrary units by the integral intensity of the absorption signal. In the present study we report calculations carried out with both amplitudes and double-integrals, which are directly related, provided linewidth is constant. In the measurements performed in the present study no significant variation in linewidth was found for all samples. According to such calculations spectra amplitude was considered a good quantitative approxi- mation [12]. To further support this approximation, correlation of integrals with amplitude was computed in all spectra, giving a very high correlation coefficient (R = 0.89; p,0.0001).
Signal amplitude is the parameter directly measured by the instrument, is easy to be performed by all operators and is more reproducible than the integral calculated value. For these reasons we indicate amplitudes as an effective alternative to integrals, under our experimental conditions. Although a larger study is needed to further validate this observation in a multicenter study, the present investigation validates the hypothesis that ESR analysis may effectively discriminate human melanomas from human nevi supporting the routine histological diagnostic process.
We believe this study may stimulate further development of skin ESR scanners to open a novel path toward the early non-invasive melanoma diagnosis.

Author Contributions
Conceived and designed the experiments: EC LK JZP AF. Performed the experiments: EC GD GV MSA FP JZP. Analyzed the data: EC AF JZP GD. Contributed reagents/materials/analysis tools: FP. Wrote the paper: EC LK GD AF.