In malignant melanoma, CNVs are associated with tumor development and progression. Melanomas with a poor prognosis show significantly more frequent genetic abnormalities and a significantly higher number of CNVs have been reported compared to melanoma cases with a good prognosis26.At the beginning, we looked at copy number changes in genes CDK4, CDKN2A and CDKN2B. Although primarily associated with familial melanoma, CNVs in these genes have been observed in all melanoma types6,7,8,9,27,28,29,30,31,32,33,34,35. Our results (on FFPE tissue samples) showed that copy number changes in these genes were present only at later stages of the disease (Clark stage IV and V), although Rákosy and colleagues30 showed that deletion of the 9p21 region (CDKN2A/B) is also present in early stages of melanoma. This circumstance could stem from the smaller sample size obtained from patients in stages II and III. It could be due to rapid tumor growth and progression36 or late diagnosis of the disease. Xavier et al. (2016)37 investigated the reasons for the delay in the diagnosis of MM. The main component of the delay was related to the patient. Only 31.3% of patients knew that melanoma was a serious skin tumor and most thought that the pigmented lesion was not important. In the CDK4 gene, we detected amplification only. For CDKN2A and CDKN2B genes, we primarily detected deletion, but in one sample (MM30), we observed amplification. Although the CDKN2A gene is predominantly deleted in MM, amplification can also occur30. In neither case, we did not detect a combination of CDKN2A/B deletion and CD4 amplification in the tissue sample, confirming that these two phenomena are mutually exclusive38.Although several studies have identified copy number changes of CDK genes in tissue samples39,40,41, there is limited information on their capture from the plasma of MM patients. Thus, we decided to analyze CNVs in these genes in circulating ctDNA from MM patients and determine whether these aberrations may serve as a suitable biomarker for the noninvasive diagnosis of melanoma patients.The measured values were compared with the SPAST reference gene. Using this gene, we were also able to quantify the total circulating DNA in the blood of the patients. Compared to healthy plasma, we detected a 2.9- (Clask II); 1.7- (Clark III); 4.2- (Clark IV); and 1.9-fold (Clark V) increase in circulating DNA in the tumor plasma. Although such quantification cannot define the ratio of cfDNA to ctDNA, it can indicate increased levels of circulating DNA in the blood. These can be elevated in various physiological and clinical conditions in addition to cancer, such as myocardial infarction42, trauma43, during pregnancy44,45, or in post-transplant patients derived from a donor organ46,47. In cancer, higher levels of circulating DNA (cfDNA) are associated with a greater disease burden, more advanced disease stage and a higher number of metastatic sites48. Measurement of circulating DNA levels has potential use in response to immunotherapy. Rising levels of DNA from tumor cells targeted by immunotherapy may indicate treatment success49.Copy number analysis of CDK genes from the plasma of the patients did show deletion in the CDKN2A/B genes and amplification in the CDK4 gene compared to healthy samples, but these changes were not statistically significant. We also found no significant difference between pre-operative and post-operative plasma, which was collected approximately 3 months after tumor removal, ensuring that elevated levels of circulating DNA due to the surgical procedure were not present. Thus, we concluded that these CNV biomarkers are not suitable for liquid biopsy and decided to identify new ones that would be more suitable for non-invasive diagnosis from blood. For this purpose, we chose low-coverage whole-genome sequencing (lcWGS).When comparing several methods, it is WisecondorX that has produced the best results in terms of variance, distributional assumptions and basic ability to detect true variation20.Various genome-wide studies have defined genes/chromosomal regions in which frequent deletions/amplifications associated with MM have been observed6,7,8,9, but information regarding the CNV profile of ctDNA isolated from MM patients is limited. Therefore, we set out to identify the most common copy number changes from plasma samples of MM patients. In previous studies, frequent CNV changes in MM tissue samples were located, for example, on chromosomes 7q, 4q, 8q, 6p, 6q, 10q, 11q, 12q, 20q, 22q6,29,33,50. Our results pointed to the most affected chromosomal regions (whether deletions or amplifications) 6q, 4p, 13q, 18q, 20q and chromosomes 10, 15, 16, 17, 20 or 22, which partially corresponds with the observed results presented in the previous literature. Although CNVs were detected in a high number of protein coding genes, when filtering the genes present in patients only (4 and more), we reduced it to 47 chromosomal regions mostly located in or around the telomeric or centromeric regions. Furthermore, CNVs in CDKN2A, CDKN2B, and CDK4 were not in the filtered gene set, despite their presence in one-third of the tissue samples. This discrepancy indicates a potentially low concordance between tissue and plasma samples, underscoring the necessity for computational tools like WisecondorX, when analyzing liquid biopsy samples. Such programs are essential for generating reliable CNV references from non-cancerous samples, enabling the precise identification of significant CNV alterations in cancer plasma.In general, chromosomal breaks and rearrangements are more frequent near centromeres and telomeres, contributing to the formation of CNVs, also, these regions play crucial roles in maintaining genomic stability and regulating gene expression. Alterations in these regions can disrupt essential functions, potentially promoting cancer development51,52. On the other hand, centromeres and telomeres contain highly repeated sequences, making them challenging for accurate mapping and alignment of short-read sequencing technologies resulting in possible false positive results when using shallow sequencing methods53. That is why we decided to verify the lcWGS results using ddPCR. We chose 4 genes (KIF25, E2F1, DIP2C, TFG) that were present in at least 4 tumor patients, were not present in the healthy population, were detected outside of patients 18, 19, 30, 34, and 38, and have the potential to serve as therapeutic targets22,23 or diagnostic and predisposing biomarkers24,25. When searching these genes in the GDC (Genomic Data Commons) data portal of National Cancer Institute54, in project ID: TCGA-SKCM where they identified CNV in 467 melanoma samples, deletion predominated and was detected in 50.54% of samples for gene KIF25 and in 48.39% samples for gene DIP2C. Predominantly, gain of gene E2F1 was detected in 38.54% samples and in case of TFG, gain was captured in 13.92% and loss in 14.56% of melanoma samples.KIF25 is one of the members of the kinesin-like protein family, which are microtubule-dependent molecular motors and thus play an important role in the processes of intracellular transport and cell division. The KIF25 gene (Cytoband: 6q27) has been associated with various tumor types, including breast cancer22, osteosarcoma55, and malignant mesothelioma56. Groth-Pedersen et alia55 showed that inhibition of KIF25 expression in human osteosarcoma cells by U2OS significantly reduced their proliferation. In breast cancer, Zou and colleagues22 found out that estrogen induces the expression of KIF genes, including KIF25, and they also pointed out the possible deregulation of the kinase family by the ANCCA (AAA nuclear coregulator cancer associated) coregulator. Although in the TCGA database54 it was shown that deletion of this gene was predominant, other studies show that the overexpression of this gene22,55 can cause cell proliferation. That means that we can assume that amplification that was detected by lcWGS in three of our MM samples, may cause increased proliferation of tumor cells, in contrast, deletion detected in four samples, may serve as a cell defense mechanism against neoplastic-driven proliferation.E2F1 is a transcriptional activator that plays a major role in cell cycle control under physiological and pathological conditions. E2F1 protein acts in cooperation with pRB and CDKs57. CNV alterations in the E2F1 gene (Cytoband: 20q11) or its increased expression have been localized in several tumor types, renal cell carcinoma58, prostate cancer59 or hepatocellular carcinoma60. Amplification of the E2F1 gene has been observed in melanoma cell lines (compared to normal melanocytes) and also in melanoma metastases localized in lymph nodes. In addition, Western blot analysis demonstrated increased levels of E2F1 protein in 8 of 9 melanoma cell lines compared to normal melanocytes61. In our samples, amplification was detected in only 1 sample, with predominant deletion in 5 ctDNA samples. This does not correlate neither with results from other studies nor with TCGA database54,59,60,61. This gene may serve as another potential therapeutic target. Inhibition of the E2F1 protein using the small molecule inhibitor HLM006474 can induce cell death in melanocytic tumor cells resistant to BRAF inhibitors23. Moreover, CNV alterations in this gene may serve as a predisposing biomarker for skin cancer25.The DIP2C gene, a member of the disco-interacting protein homolog 2 family, is highly expressed in the brain and influences its development and function62. A global transcriptome study in DIP2B-deficient mice (DIP2B is a paralog of DIP2C) suggested that this gene plays multiple roles in cell proliferation, migration, and apoptosis63. Loss of DIP2C can affect DNA methylation and changes in gene expression, cellular senescence, and epithelial-mesenchymal transition in cancer cells64. Aberrations of the DIP2C gene (Cytoband: 10p15.3) have been detected in breast cancer24, prostate cancer65 or spitzoid melanoma66. Liu and colleagues (2023)65 found that exosomal miR-375 (tumor-derived) specifically targets the DIP2C gene, thereby regulating the Wnt signaling pathway and promoting osteoblastic metastasis and prostate cancer progression. The deletion of this gene, which was detected by lcWGS in five (5/6) of our melanoma samples, confirms the results from the studies of Li et al. (2017)24 and Liu et al. (2023)65 which showed a decrease in the expression of this particular gene in tumor cells, also our results are consistent with TCGA (TCGA-SKCM) database54 where deletion was detected in 48.39% of melanoma samples. Interestingly, in the study by Larson and colleagues64 gene expression profiling revealed 780 genes for which expression levels were affected by loss of DIP2C, including the CDKN2A gene encoding a tumor suppressor that is frequently deleted in MM.The Trk-fusion gene (TFG) was amplified in four melanoma patient plasma samples according to our lcWGS results, but digital PCR also indicated a possible deletion. Although the function of this gene is currently poorly characterized, it is thought to be involved in the NF-kB and MAPK signaling pathways67. One of the key links of TFG to cancer is through fusions of this gene with others. Such fusions can generate new proteins, oncoproteins, which can interfere with normal cell cycle regulation and lead to neoplastic transformation of the cell68,69,70,71,72. Endoh et al. (2012)73 showed that the expression level of TFG in prostate tumor tissues was higher than in non-tumor tissues in 63.9% of cases. Targeted inhibition of TFG by specific silencing RNAs led to reduced proliferation of PC3 tumor cells and induction of premature senescence. The increased expression correlated with our results, which indicated amplification of this gene. As TFG is primarily involved in the NF-kB (important in the inhibition of apoptosis and treatment resistance in melanomas) and MAPK (activation of this signaling pathway occurs in approximately 70% of melanomas) pathways, aberrations in this gene may be considered important in the tumorigenesis of melanomas.Overall, in the context of individual gene examination, result consistency between lcWGS and ddPCR and cross-referencing with existing studies and TCGA database, DIP2C demonstrated a potential for further analysis as non-invasive CNV biomarker of malignant melanoma.When comparing lcWGS and ddPCR, the results from ddPCR were identical to those from lcWGS in 54% for evaluation model 1 and in 46% for evaluation model 2. We therefore propose to use model 1, due to the higher similarity with the NGS results, since the NGS analysis compared the results with the overall genetic profiles and in case of ddPCR, only one reference gene was used. Comparison of plasma vs. FFPE tissue (both methods, both scoring models) showed similarity ranging from 29% up to 63%, although the comparison of tissue vs. liquid biopsy samples, especially for quantitative changes, may yield discrepancies due to variations in sample types, collection methods, differences in the stability of genetic material, and the presence of cfDNA from non-tumor cells in liquid biopsy samples. The low percentage of similarity of results may be due to several limitations of both methods.Biological limitations of lcWGS are caused by naturally highly fragmented genomes of oncological patients18, which produce short sequenced reads for mapping to the reference genome. Modeling of copy number profile is further complicated by a low tumor fraction, proportion of ctDNA in the total cfDNA of tumor patients, especially in those at a lower stage of the disease or in patients with slow-growing tumors74. The presence of cfDNA from normal cells, such as hematopoietic or endothelial cells, introduces background noise, hindering the detection and quantification of tumor-specific copy number alterations and potentially resulting in biased representations of the tumor’s CNV profile. More technical limitation of CNVs is selected bin size associated with the accuracy of CNVs length and positions in general which can detect only approximate borders of CNVs rounded up to selected bin size. In other words, we can say inside which bins there are CNVs, but not exactly where inside the bin. In addition, when we overlap these positions with gene locations on the genome, it can produce false matches and it would be appropriate also to consider the length of overlapping with genes. Use of WisecondorX is also highly dependent on the selection of healthy reference samples. The more samples we use with health status in mind, the better we can model a healthy population and filter out CNVs correlated with non-cancerous diseases in the population.The limitations of ddPCR in this case represent the introduction and optimization of the ddPCR CNV assays themselves. For liquid biopsy samples, they can present challenges, mainly due to the selection of an adequate reference gene(s) to which the copy number of the target gene will be compared. Although SPAST was shown to be the most suitable reference gene from tumor tissue MM samples and healthy skin samples (based on our MLPA results; data not shown) and was used as the only reference gene in non-invasive cfDNA analysis, inconsistent results when compared with lcWGS results suggest that in case of detection CNVs from circulating DNA, the inclusion of multiple reference genes would help to obtain a better and more balanced reference value. Considering the importance of CNVs in human genetic diversity and their association with multiple complex disorders75, it is critical to pay great attention to the selection of multiple reference genes for ddPCR analysis. Ideally, these reference genes should have a single copy in the genome and their copy number should not vary between healthy individuals and cancer patients. For example, Ma et al. (2023)76 identified AGO1, AP3B1, MKL2, and RPP30 as suitable reference genes for CNV analysis by ddPCR because their copy numbers were not altered in either tumor or non-tumor samples. However, this research was only about genomic DNA, not circulating DNA. Another helpful point would be to determine the ctDNA fraction from the patient’s total cfDNA. To do so, genetic or epigenetic profiling of tumors using next-generation sequencing is recommended77,78,79,80. Another challenge is the limited sample volume, which does not allow us to analyze a large number of genes simultaneously. Therefore, a more effective alternative could be the use of ddPCR with TaqMan probes, which enables multiplex analysis. The use of multiple reference genes could improve the accuracy of CNV detection by reducing the influence of variations in individual reference genes81.Both methods, lcWGS and ddPCR, have their own advantages and limitations in CNV detection. Therefore, the combination of these methods appears to be beneficial for the identification of new diagnostic biomarkers and the subsequent creation of dPCR assays that can be used for early detection of the disease, monitoring its course or response to treatment. Thus, such an integrative approach provides more comprehensive information for potential use in the non-invasive diagnosis of CNV tumor changes.
