The A375 (cat. no. CRL-1619) and A2058 (CRL-3601) melanoma cell lines were obtained from the American Type Culture Collection, whereas the SK-MEL-23 melanoma cell line was already available at the National Cancer Institute Pascale of Naples. A375, A2058 and SK-MEL-23 cells were cultured in a complete RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS), 2 mmol/l of L-glutamine, 100 UI of penicillin and 100 μg/ml streptomycin (all provided from Gibco; Thermo Fisher Scientific, Inc.). Each cell line was seeded in 100-mm cell-culture dishes (Qiagen GmbH) at the density of 1×10⁶ cells and cultured in a humidified incubator (5% CO₂) at 37°C for 48 h. Cell pellets were collected by scraping cell cultures in cold PBS 1X (Gibco; Thermo Fisher Scientific Inc.) and frozen at -80°C until analyses. Consecutive cohorts of 10 FFPE melanoma tissues (Age range: 35-75 years) and 10 FFPE nevi without atypical histological features (Age range: 18-55 years), as well as two serum samples from two melanoma patients that also provided FFPE tissues, were obtained from the National Cancer Institute 'Fondazione G. Pascale', Naples (Italy), using standard procedures. The FFPE tissues and serum samples were collected from January 2019 to May 2020 at the Melanoma Cancer Immunotherapy and Innovative Therapy Unit of the National Cancer Institute 'Fondazione 'G. Pascale' (Naples, Italy). The histopathological features of melanoma samples and sociodemographic characteristics of patients and healthy controls are reported in Table SI. The present study was conducted in accordance with the guidelines of the Declaration of Helsinki (seventh revision, 2013) and approved by the Institutional Review Board of the National Cancer Institute 'Fondazione 'G. Pascale' (Naples, Italy) (approval no. 33/17oss, approved on 10 January 2018). Informed consent was obtained from all subjects involved in the study. All experiments were conducted in duplicate.

Genomic DNA from A375, A2058 and SK-MEL-23 melanoma cell lines was extracted by using the PureLink Genomic DNA Mini Kit (cat. no. K1820-01; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. The cfDNA from serum samples was extracted according to a custom protocol, as previously described (30). Briefly, 1 ml of serum was treated with 240 μl of extraction solution (EDTA (250 mmol/l)/NaCl (750 mmol/l): 100 μl; sodium dodecyl sulfate (100 g/l): 100 μl; proteinase K (stock solution 20 mg/ml): 40 μl) and incubated at 56°C for 2 h. Subsequently, 200 μl of saturated 6M NaCl was added to the mixture to precipitate the proteins. The collected supernatant was mixed 1:1 with phenol-chloroform and incubated at RT for 5 min. The DNA solution was treated with an equal volume of absolute ethanol at −20°C overnight and centrifuged at 14,000 x g for 15 min at 4°C. The DNA pellet was washed with 70% ethanol and finally resuspended in 20 μl of RNAse/DNase-free water. Since the cfDNA amount was undetectable by Nanodrop-1000 (Thermo Fisher Scientific, Inc.), a fixed volume of 5 μl of cfDNA sample was used for downstream analyses. Genomic DNA from FFPE tissues (four sections with a thickness of 8 μm) was extracted using the QIAamp DNA FFPE Tissue kit (cat. no. 56404; Qiagen GmbH) and the deparaffinization solution (cat. no. 19093; Qiagen GmbH) according to the manufacturer's protocols. Nanodrop-1000 was used to assess the amount and quality of extracted DNA by evaluating the 260/280 nm ratio (~1.8 for pure DNA samples).

To perform bisulfite sequencing of SLC22A17 methDNA hotspot (chr14:23,821,229-23,821,230-Assembly: GRCh37/hg19) (Fig. S1A), 1.2 μg of genomic DNA from A2058, A375 and SK-MEL-23 cells were bisulfite-converted by using the EpiTect Plus DNA Bisulfite kit (cat. no. 59124; Qiagen GmbH) according to the manufacturer's protocol. The amplification of bisulfite-converted SLC22A17 target (chr14:23,821,176-23,821,349-Assembly: GRCh37/hg19) (Fig. S1A), whose unconverted sequence contains 15 CpG dinucleotides and 1 CCGG restriction site, was conducted preparing a reaction mix (20 μl) containing 100 ng of the bisulfite-converted DNA, 10 μl of the 2X ddPCR Supermix for Probes (No dUTP) (cat. no. 1863024; Bio-Rad Laboratories, Inc.), 10 μM (final concentration) of forward and reverse primers. The Bisulfite Primer Seeker (https://www.zymore-search.eu/pages/bisulfite-primer-seeker-accessed on 7th March 2022) was used to design the bisulfite primers. PCR thermal conditions and primer sequences are reported in Table I. PCR product was cleanup using the PureLink PCR Purification kit (cat. no. K310001; Thermo Fisher Scientific, Inc.) and sequenced with the Mix2Seq kit (Eurofins Genomics Germany GmbH) according to the manufacturer's protocols. The analysis of DNA sequences was performed by Chromas Lite software version 2.6.6 (https://technelysium.com.au/wp/chromas/) (accessed on 10th June 2022).

A sequence (276 bp) of SLC22A17 gene (chr14:23,821,170-23,821,445-Assembly: GRCh37/hg19), containing 3 CCGG restriction sites, was used to generate DNA unmethylated and methylated controls (Fig. S1B). Specifically, the SLC22A17 sequence was obtained by PCR amplification of A375 gDNA using the Taq DNA Polymerase, recombinant (5 U/μl) (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. Primers and PCR thermocycling conditions are included in Table I. The PCR product was purified using the PureLink PCR Purification kit (cat. no. K310001; Thermo Fisher Scientific Inc.) and quantified by NanoDrop-1000 (Thermo Fisher Scientific, Inc.). To obtain methylated control (100% of methylation), the amplified SLC22A17 sequence was treated with CpG methyltransferase (M.SssI) kit (cat. no. EM0821; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. Briefly, 200 ng of SLC22A17 PCR product was added to a reaction mixture containing 1 μl of M.SssI enzyme, 2 μl of M.SssI buffer 10X, 0.4 μl of SAM (S-adenosylmethionine) 50X and nuclease-free water to obtain a final volume of 20 μl. The unmethylated control (0% of methylation) was generated from SLC22A17 PCR product using the same methylation mixture without M.SssI enzyme. Both reactions were incubated at 37°C for 30 min and then stopped by heating at 65°C for 20 min. Finally, the methylated and unmethylated controls were purified by using the PureLink PCR Purification kit (cat. no. K310001; Thermo Fisher Scientific, Inc.) and quantified with Nanodrop-1000.

The SLC22A17 methylated (100%) and unmethylated (0%) CTRLs were mixed in different ratios (100, 75, 50, 25 and 0%) maintaining a constant total DNA concentration of 1.25×10⁻⁶ ng/μl.

To assess the digestion efficiency of HpaII and MspI in ddPCR-MSRE reaction and normalize the percentage of methylation of each methDNA target, methCTRL was generated by PCR amplification of a sequence of the fluorescent protein Clover, which contains 1 CCGG restriction site (Fig. S1C). Briefly, 10 ng of pcDNA3-Clover plasmid gently provided by Dr Michael Lin (Department of Bioengineering, Stanford University, Stanford, USA) (Addgene plasmid #40259; http://n2t.net/addgene:40259; RRID: Addgene_40259) was amplified using the Phusion High-Fidelity DNA Polymerase (2 U/μl) kit (cat. no. F-530XL; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. PCR thermal conditions and primer sequences are reported in Table I. The PCR product was subsequently treated with 1 μl DpnI (cat. no. FD1703; Thermo Fisher Scientific, Inc.) at 37°C for 15 min and then the enzyme was inactivated at 80°C for 20 min. Finally, the PCR reaction was purified using the PureLink PCR Purification kit (cat. no. K310001; Thermo Fisher Scientific, Inc.) and quantified with Nanodrop-1000.

MSRE digestion was performed on gDNA obtained from A375, SK-MEL-23 and A2058 cells. For each sample, three different reaction tubes (final volume 10 μl) were prepared by mixing 200 ng of gDNA, 10⁻⁶ ng/μl of methCTRL, 1X CutSmart Buffer (cat. no. B7204), and 20 UI of HpaII (cat. no. R0171S) for tube 1, 20 UI of MspI (cat. no. R0106S) for tube 2, and no enzyme for tube 3 (all the reagents were purchased from New England Biolabs). All the reaction tubes were incubated at 37°C for 1 h and stopped with Proteinase K (cat. no. EO0491; Thermo Fisher Scientific, Inc.) (final concentration 1 mg/ml) incubating the samples at 55°C for 30 min followed by an inactivation step at 95°C for 10 min. Following the standard MSRE digestion, 4 μl of RNase/DNase-free water molecular biology-grade was added to 1 μl of each digested sample for the downstream ddPCR amplification. In addition, to evaluate the effect of potential inhibitors (melanin) on the MSRE digestion, 20 ng instead of 200 ng of SK-MEL-23 gDNA were digested in 10 μl of final reaction volume according to the aforementioned standard MSRE protocol. Of note, 5 μl of these digested samples were directly added to the ddPCR mix for the downstream analysis.

To further evaluate the ddPCR efficiency in the amplification of targets digested by restriction enzymes and to evaluate the highest amount of the MSRE digestion mix that can be uploaded in the ddPCR mix, BamHI was used as a restriction enzyme that does not digest the SLC22A17 methDNA target. Briefly, 20 μl of digestion mix containing 20 UI of BamHI (cat. no. R0136S; New England Biolabs, Inc.), 1X CutSmart Buffer (cat. no. B7204; New England Biolabs, Inc.), 80 ng of A375 gDNA, and 4×10⁻⁶ ng of methCTRL was incubated at 37°C for 30 min, stopped with 1 mg/ml of Proteinase K (cat. no. EO0491; Thermo Fisher Scientific Inc.) at 55°C for 30 min followed by an inactivation step (95°C for 10 min). A total of 5 μl of the digested sample (20 ng) was used for the ddPCR amplification. As a control, 20 ng of A375 gDNA and 10⁻⁶ ng methCTRL in 5 μl of molecular biology-grade water were amplified in ddPCR.

The ddPCR amplification mix was prepared by mixing 11 μl of 2X ddPCR Supermix for Probes (no dUTP) (cat. no. 1863024; Bio-Rad Laboratories, Inc.), 5 μl of each sample, 450 nM of FAM probe (SLC22A17 target), 450 nM of HEX probe (methCTRL), 900 nM of forward and reverse primers for each target and RNase/DNase-free water molecular biology-grade up to a final volume of 22 μl (probe and primer sequences are reported in Table I).

Custom MSRE-ddPCR assay consists of one-tube reactions in which methylation-sensitive restriction enzymes (i.e. HpaII and MspI) directly digest the DNA targets in the ddPCR reaction mix. In particular, three different amplification mixes were prepared for each sample, containing HpaII, MspI and no enzyme as undigested control, respectively. The MspI mix was performed as an additional control to assess the enzymatic digestion efficiency. However, this optional control can be avoided because of the use of methCTRL. Briefly, each amplification mix (22 μl) was prepared by using 11 μl of 2X ddPCR Supermix for Probes (no dUTP) (cat. no. 1863024; Bio-Rad Laboratories, Inc.), 900 nM of forward/reverse primers and 450 nM of FAM/HEX probes for SLC22A17 target and methCTRL (probe and primer sequences are reported in Table I). Up to 20 ng of DNA sample and 10⁻⁶ ng of methCTRL (final volume 5 μl) were added to each MSRE-ddPCR mix along with 10 UI of restriction enzyme in the HpaII and MspI mix. All amplification mixes were incubated at 37°C for 30 min before droplet generation.

Droplet generation was performed by loading 20 μl of ddPCR amplification mix in DG8 Cartridges along with 70 μl of Droplet Generation Oil (cat. no. 1863005; Bio-Rad Laboratories, Inc.) within the sample and oil wells, respectively. Then the cartridge was covered with Gasket (cat. no. 1863009; Bio-Rad Laboratories, Inc.) and transferred into Droplet Generator QX100 (Bio-Rad Laboratories, Inc.) for droplet generation according to the manufacturer's instructions. The droplet mixture was recovered from the cartridge and transferred into a 96-well plate (cat. no. 12001925; Bio-Rad Laboratories, Inc.) to perform PCR amplification by the C1000 Touch Thermal Cycler (Bio-Rad Laboratories, Inc.) according to the manufacturer's protocol. The ddPCR thermal conditions are reported in Table I. Finally, the QX200 Droplet Reader (Bio-Rad Laboratories, Inc.) was used for droplet quantification. Absolute quantification (copies/μl) of DNA targets was retrieved using QuantaSoft software (version 1.7.4 (QuantaSoft). Amplitude thresholds were set manually by the operator on the basis of positive and negative droplet amplitudes. The fluorescence amplitude of droplets was also considered to evaluate the efficiency of the amplification reaction.

The methDNA percentage of the target genes was retrieved considering the ratio between the ddPCR absolute quantification of the methDNA target in HpaII and the undigested control mix for each sample (first term of Formula 1 and 2). Notably, this ratio is 1 when the methDNA target is fully methylated (100% of methylation), while it is 0 when the methDNA target is completely unmethylated (0% of methylation). (1) (2)

To overcome the bias in methDNA percentage estimation due to the inhibition of the enzymatic digestion, data normalization was performed by using an enzymatic digestion coefficient computed by the reciprocal ratio between the ddPCR absolute quantification of the methDNA target in MspI and undigested control mix (second term of Formula 1). Similarly, the enzymatic digestion coefficient was obtained by the reciprocal ratio between the methCTRL absolute quantification in HpaII and the undigested control mix avoiding the MspI mix (second term of Formula 2). Notably, the enzymatic digestion coefficient does not affect the methDNA percentage estimation when the enzymatic digestion is completed (enzymatic digestion coefficient=1). Conversely, the methDNA percentage is adjusted when the enzymatic digestion is partially inhibited.

Linear regression analysis was computed to evaluate the goodness of fit (R²) obtained from the MSRE-ddPCR analysis on scalar dilution samples of SLC22A17 methylated control. To test the sensitivity of MSRE-ddPCR, the horizontal best-fit lines through the mean of all the methDNA values, obtained from the serial dilutions of SK-MEL-23 and A375 gDNAs, were calculated using Non-linear fit analysis. Differential analysis of SLC22A17 methDNA in melanoma and nevi samples was evaluated by Mann-Whitney test. The difference between the comparing groups was reported as differences between the median levels of SLC22A17 methDNA. Receiver Operating Characteristic (ROC) curve analysis was performed to evaluate the performance and accuracy of the diagnostic test. P≤0.05 was considered to indicate a statistically significant difference. All analyses were executed using GraphPad Prism software (version 8.0.2) (Dotmatics).

To evaluate the potential interference between restriction enzymes and ddPCR chemistry in enzymatic digestion and amplification processes, MSRE-ddPCR protocol was performed analyzing DNA methylated (100%) and unmethylated (0%) controls obtained from the SLC22A17 sequence (chr14:23,821,170-23,821,445-Assembly: GRCh37/hg19). Specifically, 5×10⁻⁶ ng of each DNA methylated and unmethylated controls along with 10⁻⁶ ng of Methylation Internal Control (methCTRL) were added to HpaII, MspI and undigested mixes. methCTRL and SLC22A17 forward/reverse primers and probes were used for the amplification step.

As expected, the results revealed that the number of copies for the SLC22A17 methylated control within the HpaII mix was similar to the undigested sample (45.4±2.38 copies/μl and 42.5±1.7 copies/μl, respectively). Interestingly, the amplitude of SLC22A17 methylated control (FAM fluorescence) in the HpaII mix was similar to the undigested mix, demonstrating that the HpaII enzyme did not affect the ddPCR amplification efficiency (Fig. 1A). Moreover, no amplification was detected for SLC22A17 methylated control within the MspI mix. Similarly, the SLC22A17 unmethylated control showed no amplification within HpaII and MspI reactions compared with the undigested sample, suggesting that the enzymatic digestion efficiency was not affected by ddPCR chemistry (Fig. 1A).

Regarding methCTRL (HEX fluorescence), a similar amplification signal was observed in the undigested mix of SLC22A17 methylated and unmethylated samples (35.4±0.45 and 34.00±0.84 copies/μl, respectively), while a weak amplification signal was detected in both HpaII and MspI reactions (from 0.26 to 1.70 copies/μl) (Fig. 1B). The positive droplets of methCTRL in HpaII and MspI reactions are due to a not complete digestion that never reaches the theoretical 100% efficiency of HpaII and MspI in ddPCR mix. This background signal, detected in each MSRE-ddPCR mix, is removed by normalization procedures during the estimation of methDNA percentage.

To evaluate the accuracy of the MSRE-ddPCR method, linear regression analysis was performed comparing the expected and measured methDNA levels of each SLC22A17 methylated/unmethylated sample (100, 75, 50, 25 and 0%). The results demonstrated a high linear correlation of all expected/measured couples (R²=0.9925) indicating the reliability of the method at different percentages of methDNA target (Fig. 1C).

The MSRE-ddPCR assay was tested on gDNA extracted from different melanoma cell lines evaluating the levels of SLC22A17 methDNA hotspot (Fig. S1) selected in the present study. In particular, the SLC22A17 target was assessed on SK-MEL-23, A375 and A2058 cells (Fig. 2A and B).

The methylation levels of the SLC22A17 hotspot were higher in SK-MEL-23 (98.18±0.15%) compared with A375 (39.7±0.60%) and A2058 cells (1.16±0.19%) (Fig. 2A and B). Notably, the amplitude of FAM and HEX droplets was comparable to that observed in the setup experiment reported in Fig. 1.

The MSRE-ddPCR results agreed with those obtained from bisulfite sequencing of SLC22A17 hotspot in the same cell lines. In particular, the internal cytosine of CCGG within the SLC22A17 hotspot was unconverted in SK-MEL-23 (Fig. S2A), partially converted (~34% of methDNA) in A375 (Fig. S2B) and converted in A2058 (Fig. S2C), indicating that this hotspot was highly hypermethylated in SK-MEL-23, partially methylated in A375 and unmethylated in the A2058 cells. The absence of amplification signal in the MspI mix for the SLC22A17 target (Fig. 2A), as well as the slight signal detected for methCTRL in HpaII and MspI mix (Fig. 2B), indicated that the restriction enzymes were highly efficient in MSRE-ddPCR mix.

To assess the reliability of MSRE-ddPCR results, the methylation levels of the SLC22A17 hotspot were also evaluated by the standard MSRE procedure using HpaII and MspI enzymes as aforementioned. The methylation percentage computed through both methods was normalized by the methCTRL signal in HpaII mix and the amplification signal of each target in MspI mix (Table II).

The results indicated that the methylation levels of SLC22A17 target obtained with both MSRE-ddPCR and standard MSRE procedures were comparable showing only a slight reduction of the methDNA percentage estimation (<0,5%) when methCTRL normalization was applied (Table II). Interestingly, the standard MSRE assay was unable to quantify the SLC22A17 methylation hotspot in SK-MEL-23 cells since the SLC22A17 signal was relevant in the MspI mix (~90 copies/μl), suggesting that the MspI enzymatic digestion was inhibited by DNA sample contaminants (Fig. S3). It was probably due to the presence of melanin in the gDNA sample obtained from the highly melanin-pigmented SK-MEL-23 cells that appeared brown-pigmented during the DNA extraction procedure (data not shown). To test this hypothesis, the standard MSRE assay was performed on 20 ng of SK-MEL-23 gDNA (instead of 200 ng of gDNA) in 10 μl of final reaction volume, of which 5 μl were directly added to the ddPCR mix for the amplification. The results revealed an efficient digestion of MspI in SK-MEL-23 diluted gDNA, indicating that the high melanin content of SK-MEL-23 may reduce the efficiency of MspI (Fig. S3A and B). However, the detected amplitude for each diluted sample was significantly reduced compared with standard MSRE procedure due to the amount of the MSRE digestion mix (5 μl of diluted gDNA) used for the downstream analysis, demonstrating that MSRE mix affected the ddPCR amplification (Fig. S3A and B). Notably, the amount of SK-MEL-23 gDNA sample loaded in each standard MSRE mix (200 ng of gDNA in 10 μl final reaction volume) was ~7.5% of the reaction and ~0.75% for diluted gDNA sample (20 ng in 10 μl final reaction volume), whereas it was ~0.34% in each MSRE-ddPCR mix (20 ng of gDNA in 22 μl final reaction volume), considering an initial DNA concentration of 266.7 ng/μl. Therefore, using a lower quantity of DNA in each MSRE-ddPCR mix compared with the standard MSRE mix, the enzymatic inhibition was efficiently prevented.

To further evaluate the interference of the standard MSRE digestion mix on the downstream ddPCR amplification and to assess the highest amount of the MSRE digestion mix that may be processed by ddPCR, A375 gDNA was digested with BamHI using its standard reagents. This restriction enzyme was selected since it does not digest the SLC22A17 methDNA hotspot allowing to simulate the inhibition of ddPCR amplification by digestion components independently from the negative effect of MSRE digestion on the number of positive droplets. Specifically, 5 μl of BamHI reaction (1X CutSmart Buffer, 5 UI of BamHI) containing 20 ng of A375 gDNA and 10⁻⁶ ng of methCTRL were amplified in ddPCR to detect SLC22A17 target and methCTRL. The same amplification conditions were applied to the control sample consisting of 20 ng of A375 gDNA and 10⁻⁶ ng of methCTRL in 5 μl of molecular biology-grade water.

The results revealed that the SLC22A17 amplitude (FAM channel) was reduced by ~1,000 AU fluorescence in the BamHI mix compared with the control mix (Fig. S4A and C). Similarly, a significant reduction of methCTRL amplitude signal was observed in the BamHI reaction compared with the control (~5561 AU fluorescence for the control mix and ~4477 AU fluorescence for the BamHI mix) (Fig. S4B and D). The data confirmed the results obtained for SK-MEL-23 diluted MSRE mix (Fig. S3) indicating that the restriction enzyme reaction, including the MSRE reaction, consistently affected the ddPCR amplification. Therefore, the standard MSRE digestion of low-concentrated DNA samples cannot be analyzed by ddPCR because up to 5 μl of this reaction should be added in each ddPCR mix to reach the sensitivity threshold.

Linearity test of MSRE-ddPCR assay was performed evaluating the SLC22A17 methDNA target using a 2-fold serial dilution starting from 20 to 0.312 ng of gDNA obtained from SK-MEL-23 and A375 cells. The results suggested that the methylation percentage of the SLC22A17 hotspot, whose average values ranged from 88.29 to 95.49% for SK-MEL-23 and 37.65 to 44.20% for A375, demonstrated a slight deviation (3.6% for SK-MEL-23 and 3.27% for A375) from best-fit value (91.89% for SK-MEL-23 and 40.92% for A375) in both the gDNA dilutions. However, a significant increase in standard deviation (SD) was observed starting from 1.25 ng of gDNA, with the highest SD observed for samples diluted at 0.312 ng, suggesting that the accuracy of the MSRE-ddPCR was suitable until the concentration of 0.625 ng (Fig. 3).

MSRE-ddPCR assay was designed for the methylation analysis of low quantity and/or quality DNA samples, including gDNA from FFPE samples and cfDNA from different body fluids. To test the analytic performance of the MSRE-ddPCR, the SLC22A17 methDNA hotspot (Fig. S1B) was analyzed on a pilot cohort of FFPE melanoma tissues (n=10) and FFPE nevi (n=10). Moreover, to evaluate the adaptability of the MSRE-ddPCR for the analysis of cfDNA, two serum samples from melanoma patients were analyzed as a pilot test.

The plots indicated that the amplitude signal of both FAM and HEX channels was similar to the signal detected for melanoma cell lines (Fig. 4A and B). However, the droplet rain between positive and negative droplets was observed for SLC22A17 methDNA hotspot in the FFPE samples, indicating the presence of PCR inhibitors and DNA fragmentation (Fig. 4A and B). Moreover, these inhibitors slightly reduced the amplitude of the SLC22A17 methDNA hotspot and methCTRL in FFPE and cfDNA samples than melanoma cell lines. However, the amplitude difference did not influence the accuracy of target detection by ddPCR as demonstrated in Fig. S4.

The MSRE-ddPCR analysis revealed that the methylation levels of SLC22A17 were 21.04±0.90% for FFPE 1 and 33.43±1.88% for FFPE 2. Regarding the methylation analysis of serum cfDNA, the SLC22A17 hotspot was 9.74±0.38% for cfDNA 1 and 12.69±0.53% for cfDNA 2 (Fig. 4A and B). The results of SLC22A17 methDNA hotspot analysis on the melanoma and nevi cohorts revealed that the median percentage of methDNA was significantly (P≤0.01) higher in melanoma (39.37%) compared with healthy samples (21.78%) (Fig. 4C). Of note, despite the small number of analyzed samples, the ROC curve analysis demonstrated high sensitivity and specificity [Area Under Curve (AUC): 0.88; P≤0.01] of the SLC22A17 methDNA test as a diagnostic biomarker for melanoma (Fig. 4D).