Graphical abstract graphic file with name fx1.jpg [41]Open in a new tab Highlights * • The serum metabolome differs between patients with melanoma and healthy individuals * • Six serum metabolites can predict the diagnosis of melanoma * • Muramic acid is a particularly powerful diagnostic biomarker across melanoma stages __________________________________________________________________ Morsy et al. present an untargeted metabolomics analysis of patients with melanoma, revealing a distinct metabolome composition compared to healthy control individuals. Six specific serum metabolites can accurately diagnose melanoma in a clinical setting. Particularly, the lead biomarker muramic acid represents a very promising accuracy across all melanoma stages. Introduction Melanoma represents one of the most deadly cancers worldwide, and not only incidence rates but also mortality rates keep increasing.[42]^1^,[43]^2^,[44]^3 The further rise in incidence seems to be caused primarily by the increased exposure to UV irradiation, which is one of the biggest risk factors for melanoma.[45]^4 Notably, melanoma is not only frequent in elder patients but has also become one of the most common cancers in young adults.[46]^5 Projected estimates suggest that by 2040, the global burden of melanoma might reach about 500,000 newly diagnosed cases and about 100,000 melanoma-related deaths per year.[47]^6 Such numbers highlight the urgency of highly accurate and easily measurable biomarkers in daily clinical practice, since early detection and treatment of melanoma are critical to improve the prognosis for patients.[48]^3 Currently available biomarkers are particularly related to the correct histopathologic diagnosis as well as to staging, prognosis, and monitoring of therapeutic efficacy of specific therapies. Among those biomarkers are approaches based on the detection of circulating tumor cells, serum nucleic acid levels, gene and protein expression levels and patterns, as well as genetic variations.[49]^7 However, reliable biomarkers for the diagnosis of malignant melanoma are not available. The currently used serum levels of lactate dehydrogenase or S-100 protein are particularly useful only for monitoring disease course or recurrence but not for diagnostic purposes.[50]^8^,[51]^9 The microbiome has recently gained enormous attention in cancer research, and some studies suggested that distinct microbiome patterns might serve as biomarkers for the prediction of response to cancer immunotherapies[52]^10^,[53]^11^,[54]^12 or the onset of immune-related side effects.[55]^8 However, to date, neither distinct microbiome compositions nor specific bacterial strains play a role as biomarkers in the clinical standard of care, and the observed associations are mostly not suited to predict the diagnosis of melanoma in patients. However, the observed associations between particular microbiomes of patients with melanoma and therapy outcomes are strong, and it thus seems plausible that the microbiome functionality might represent a useful tool for biomarker detection. One of the most important functional aspects of the microbiome is its metabolic activity. Therefore, the metabolome might not only be important for disease and therapy outcomes in patients with melanoma but might also serve as a potent biomarker in such cases. Notably, current knowledge about the serum metabolome in patients with melanoma is scarce. Most studies addressing the role of the metabolome in melanoma focused on melanoma cells or melanoma tissue[56]^13^,[57]^14^,[58]^15^,[59]^16^,[60]^17 or even on the fecal metabolome.[61]^18^,[62]^19^,[63]^20 Only one study provided data on the serum metabolome in patients with melanoma.[64]^21 Although these authors applied targeted metabolomics, they, nevertheless, described a significant difference between patients with melanoma and healthy individuals, when analyzing 181 metabolites in the serum of 26 patients with melanoma stage III or IV.[65]^21 Furthermore, in this small and unvalidated cohort, the authors also identified distinct metabolites that might serve as biomarkers for diagnosis of advanced melanoma.[66]^21 Here, by performing untargeted serum metabolomics analysis, we demonstrate a significant difference in the composition of the serum metabolome between patients with melanoma and healthy individuals and provide a complete overview of the serum metabolome in patients with melanoma. In addition, by conducting our analysis in both an exploratory and an independent validation cohort, we were able to identify six serum metabolites as clinically relevant biomarkers for the diagnosis of advanced melanoma, while our lead candidate even exhibited a very strong biomarker performance in early-stage melanoma. Results Untargeted serum metabolome analysis reveals a distinct metabolome composition in patients with melanoma compared to healthy individuals To determine the complete serum metabolome composition of patients with melanoma, we prospectively collected serum samples from an exploratory cohort of 87 patients with melanoma as well as from an independent validation cohort of 37 additional patients with melanoma featuring late-stage tumors. Samples were collected at baseline just before the initiation of cancer immunotherapy. In addition, we studied serum samples from 18 healthy control individuals in the same age range as a control group. Patient characteristics are shown in [67]Table S1. The principal component analysis (PCA) of the exploratory cohort (87 patients with melanoma) and 18 control individuals demonstrated a clear separation between the two patient groups ([68]Figure S1A). Similarly, the partial least squares discriminant analysis (PLS-DA), which was used to visualize the variance between samples, demonstrated a clear separation between samples from healthy controls and patients with melanoma ([69]Figure 1A). This distinct separation indicates that the serum metabolome of patients with melanoma strongly differs from that of healthy individuals, suggesting that the serum metabolome can potentially identify patients with melanoma and thus be useful for diagnostic purposes. This distinct separation was further corroborated by the heatmap of the 50 most prominent metabolites, which visually presents the hierarchical clustering using the Euclidean method for distance calculation and Ward’s linkage for clustering ([70]Figure 1B). While some overlap is observed due to the broader set of metabolites, which further separate the patients into two different melanoma subgroups, this analysis reinforces the overall distinction in metabolic profiles associated with melanoma. Out of the 1,466 detected metabolites, 840 significantly differed between the two groups ([71]Table S2). Furthermore, 35 metabolites showed fold changes above 2 or below 0.5, indicating a substantially different abundance in patients with melanoma and healthy controls. Among them, 23 metabolites showed higher serum levels in patients with melanoma, including dihydrolipoate, N-(1-deoxy-1-fructosyl)glycine, and muramic acid, while 12 metabolites showed lower serum levels compared to healthy controls, including ascorbate, octanoic acid, oxoglutarate, and fumarate ([72]Figure S1B; [73]Table S3). Figure 1. [74]Figure 1 [75]Open in a new tab Multivariate analysis and metabolite profiling demonstrate striking differences in the serum metabolome between healthy control individuals and patients with melanoma (A) Partial least squares discriminant analysis (PLS-DA) showing a clear separation between 87 patients with melanoma and 18 healthy controls based on the 1,466 metabolites detected. (B) Heatmap showing the top 50 significant metabolite changes across melanoma and control individuals. (C) Bar graph showing top 25 affected metabolite sets after performing metabolite set enrichment analysis (MSEA). (D) Pathway enrichment analysis showing the most affected pathways ranked by the impact of the metabolites in these pathways. Functional serum metabolome analysis reveals an enrichment of lipids, fatty acyls, and organic acids in patients with melanoma We next aimed to identify the metabolite groups that are most prominently regulated in the serum of patients with melanoma when compared to healthy control individuals. For this purpose, the metabolite set enrichment analysis (MSEA) was performed against the “main” dataset, and a generalized linear model was used during the quantitative enrichment analysis. The 840 significantly altered metabolites were also used in MSEA and pathway analysis, highlighting key metabolic pathways affected in patients with melanoma. We found a distinct enrichment of specific classes of metabolites, including lipids and lipid-like molecules, fatty acyls, organic acids, as well as glycerophospholipids and sterol lipids, in the serum of patients with melanoma ([76]Figure 1C). Performing a pathway enrichment analysis, we detected that the most important metabolic pathways enriched in the serum of patients with melanoma were ubiquinone and other terpenoid-quinone biosynthesis, alpha-linolenic acid metabolism, and amino acid metabolism ([77]Figure 1D). The validation cohort confirms distinct serum metabolome profiles for patients with melanoma and healthy controls After demonstrating the pronounced difference between the serum metabolome of patients with melanoma and that of healthy controls, we next aimed to confirm our findings in an independent validation cohort consisting of 37 additional patients with melanoma. Again, serum was collected from patients with melanoma at baseline before the initial start of cancer immunotherapy. Further, we again measured the same serum samples from our control cohort consisting of 18 healthy individuals. In this validation cohort, we were able to corroborate our previous findings from the exploratory cohort ([78]Figure 1). We again detected a clear separation between the serum samples from healthy controls and patients with malignant melanoma, in both the PCA and PLS-DA ([79]Figures 2A and [80]S1C), as well as in the heatmap presenting the 50 most prominent metabolites ([81]Figure 2B). However, in this smaller validation cohort, we did not detect the previously observed further separation of patients with melanoma into two distinct subgroups, albeit at least a trend toward such discrimination was noticeable. The volcano plot further demonstrated significantly up- and downregulated metabolites ([82]Figure S1D). The functional metabolome analysis of our validation cohort fully confirmed the results of the MSEA of the exploratory cohorts. In our validation cohort, we again detected a clear enrichment of lipids and lipid-like molecules, fatty acyls, and organic acids ([83]Figure 2C). However, pathway enrichment analysis revealed different subtypes of amino acid metabolisms as the most affected metabolic pathways in the serum of patients with melanoma ([84]Figure 2D). Figure 2. [85]Figure 2 [86]Open in a new tab Serum metabolome data from a validation cohort consisting of 37 patients with melanoma confirm the difference between healthy control individuals and patients with melanoma (A) Partial least squares discriminant analysis (PLS-DA) showing a clear separation between 37 patients with melanoma and 18 healthy controls based on the 1,466 metabolites detected. (B) Heatmap showing the top 50 significant metabolite changes across melanoma and control individuals. (C) Bar graph showing the top 25 affected metabolite sets after performing metabolite set enrichment analysis (MSEA). (D) Pathway enrichment analysis showing the most affected pathways ranked by the impact of the metabolites in these pathways. Specific serum metabolites may serve as clinically relevant biomarkers for the diagnosis of patients with melanoma After validating our observations from the exploratory cohort in the validation cohort, we next compared the serum metabolome data of both cohorts and aimed to identify specific metabolites that could serve as clinically relevant biomarkers for the diagnosis of melanoma. Although the cohorts were analyzed separately, we generated a combined heatmap illustrating the top 50 most prominent metabolites identified in both cohorts to visualize their collective discriminatory potential, which confirmed the clear separation between samples from the patients with melanoma and the samples of the healthy controls ([87]Figure 3A). By performing receiver operating characteristic (ROC) curve and area under the curve (AUC) analysis, we evaluated the performance of the top metabolites as biomarkers, and only metabolites with an AUC greater than 0.95 in each cohort were considered as strong candidates for biomarker selection. With this approach, we detected six serum metabolites that might serve as powerful biomarkers for the diagnosis of melanoma in clinical practice. Particularly, we found that six metabolites with false discovery rate less than 0.05, namely 1-[(5-amino-5-carboxypentyl)amino]-1-deoxyfructose (fold change [FC]: 0.80 in melanoma and 1.06 in validation), dihydrolipoate (FC: 1.69 and 2.43), N-(1-deoxy-1-fructosyl)leucine (FC: 1.26 and 2.58), N-(1-deoxy-1-fructosyl)valine (FC: 0.97 and 2.23), N-(1-deoxy-1-fructosyl)glycine (FC: 1.58 and 0.87), and muramic acid (FC: 1.15 and 1.98), exhibited an AUC of at least 0.9544 in both the exploratory and the validation cohort ([88]Figures 3B–3G). The best performance was detected for the bacterial wall component muramic acid, which featured AUCs of 0.9936 and 1.0 for the diagnosis of melanoma in the exploratory cohort and in the validation cohort, respectively. Notably, all six biomarkers were strongly upregulated in the serum of patients with melanoma when compared to healthy controls in both the exploratory as well as the validation cohort. Figure 3. [89]Figure 3 [90]Open in a new tab Serum metabolomics analysis identifies potential serum biomarkers for the diagnosis of patients with melanoma (A) Heatmap of the top 50 identified biomarkers with an AUC greater than 0.95 across both the exploratory and validation cohorts. (B–G) Intensity levels of the top 6 selected metabolites (muramic acid, N-(1-deoxy-1-fructosyl)leucine, N-(1-deoxy-1-fructosyl)valine, N-(1-deoxy-1-fructosyl)methionine, 1-[(5-amino-5-carboxypentyl)amino]-1-deoxyfructose, and dihydrolipoate) and their corresponding ROC curves. Data are represented as mean ± SD. The selected biomarkers exhibited a notable AUC cutoff of 0.9, indicating their high discriminatory capacity in distinguishing melanoma cases. These biomarkers were overlapped across both the exploratory and validation cohorts, affirming their consistent identification and reliability in differentiating melanoma cases. Moreover, quantitative analysis showcased uniform intensity measurement for these biomarkers across various samples, emphasizing their stability and potential as reliable indicators for melanoma detection or prognosis. The tested biomarkers were not affected by the age or gender of the respective patients ([91]Figures S1E, S1F, and [92]S2A). Also, there was no obvious contribution of clinical characteristics, such as tumor stage, BRAF mutation status, onset of immune-related adverse events, or therapy response to the clustering ([93]Figure S2B). Expanding diagnostic biomarkers: Serum metabolites validate melanoma across all stages We broadened our study to validate the biomarker aspect of the identified molecules also in patients with early-stage melanoma. Thus, we included 22 patients with stage 1 melanoma, 67 patients with stage 2 melanoma, and 14 control subjects ([94]Table S4). Our findings indicate that the metabolomics profile of patients with early-stage melanoma again differs from that of control individuals, as demonstrated by both PLS-DA and hierarchical clustering of altered metabolites ([95]Figures 4A and 4B). Subsequently, we analyzed the enriched metabolites and discovered that they were comparable to those in later stages, with carboxylic acids, fatty acyls, and organooxygen compounds identified as the most highly enriched metabolite groups ([96]Figure S3A). Furthermore, amino acid metabolism emerged as the primary pathway category enriched in our pathway analysis, suggesting that this metabolic process plays a role in disease progression from its early stages ([97]Figure S3B). Consequently, leucine and valine derivatives may serve, among others, as effective functional biomarkers for the early diagnosis of melanoma ([98]Figures S4A and S4B). Finally, we assessed the AUC of our lead biomarker candidate, namely muramic acid, and found that it features an AUC of 0.964 in patients with stage I melanoma as well as of 0.908 in patients with stage II melanoma. In addition, we evaluated its level across three independent cohorts of patients with inflammatory bowel disease (IBD), colorectal cancer (CRC), psoriasis, and squamous cell skin cancer. Of note, we did not observe a significant alteration of muramic acid levels in any of those three cohorts when compared to the levels detected in the serum of healthy individuals ([99]Figure S4C). Figure 4. [100]Figure 4 [101]Open in a new tab Serum metabolome data from a cohort consisting of patients with stage I and stage II melanoma further confirm the difference between healthy control individuals and patients with melanoma (A) Partial least squares discriminant analysis (PLS-DA) showing a clear separation among 22 patients with stage I melanoma, 67 patients with stage II melanoma, and 14 healthy controls based on the 1,170 metabolites detected. (B–D) (B) Heatmap showing the top 50 significant metabolite changes across melanoma and control individuals. Intensity levels of muramic acid in patients with (C) stage I and (D) stage II melanoma and the corresponding ROC curves. Data are shown as boxplots with median, interquartile range, and 1.5× IQR whiskers; individual observations are overlaid with jitter. Discussion We uncovered potentially clinically relevant biomarkers for the diagnosis of melanoma using untargeted metabolomics to generate serum metabolite profiles of two independent cohorts of patients with melanoma. Overall, we uncovered significant alterations in lipid metabolism, organic acid pathways, and amino acid synthesis metabolic pathways between patients with melanoma and healthy controls. These pathways correspond nicely with the known pathophysiological mechanisms involved in the development of melanoma,[102]^22^,[103]^23 and they were consistently modified across both cohorts. Alterations in lipid metabolism were a prominent feature in our analysis, as evidenced by the enrichment of lipid-related metabolites. This is in line with previously reported data showing that melanoma cells reprogram their lipid metabolism to support rapid proliferation and metastasis.[104]^24 Moreover, increased lipid synthesis and uptake, as observed in our cohorts, are hallmarks of cancer cell metabolism, maintaining membrane fluidity and energy storage.[105]^25^,[106]^26 Our findings also highlighted significant changes in organic acid metabolism. Cancer cells, including melanoma cells, often exhibit altered energy metabolisms, particularly increased glycolysis and alterations in the tricarboxylic acid cycle. This metabolic reprogramming supports rapid cell growth and survival under various conditions, as corroborated by studies demonstrating modifications in amino acid and organic acid metabolism in melanoma.[107]^27 Specific pathway enrichments, such as ascorbate and alternate metabolism as well as D-glutamine and D-glutamate metabolism, in our analysis, may reflect the adaptation of melanoma cells to oxidative stress and an increased demand for amino acids for protein synthesis. Thus, our findings resonate with the research highlighting the importance of amino acid metabolism in cancers such as melanoma, which supports rapid growth and adaptation to the tumor microenvironment.[108]^28 Thus, identifying molecules related to such pathways as biomarkers for melanoma makes sense in a way that this might represent the presence of the malignant diseases/melanoma tumor in the individual. Based on these findings, we anticipate that the melanoma cells are frequently the origin of those molecules. This might also explain the altered levels of those molecules in patients with melanoma when compared to healthy individuals. Muramic acid is an amino sugar acid and as such a bacterial wall/structural component and is part of the bacterial peptidoglycan. Thus, it originates from bacteria. However, indeed, it is less obvious why such bacterial wall component is present in such a high abundance in the serum of patients with melanoma. Of note, it has been demonstrated that simply the presence of the melanoma tumor might increase intestinal barrier defects,[109]^29 thus promoting the occurrence of bacterial components in the patient’s serum.[110]^30 On a functional level, muramic acid as a bacterial antigen might impact on, particularly, adaptive immune responses but potentially also affect cancer cell proliferation, cell death, or immunogenicity. However, of note, muramic acid is not specific to distinct bacterial species; thus, it might not indicate the presence of specific bacterial species contributing to melanoma development but rather indicate somehow a general mechanism of barrier leakage occurring in patients with melanoma. Nevertheless, one must admit that a simple intestinal barrier breakdown is not overly likely to be the cause of muramic acid in the serum of patients with melanoma, since muramic acid levels in patients with IBD or CRC, both being diseases that classically feature severe intestinal barrier disruption, are not altered when compared to that of healthy individuals. This suggests that there might indeed be a melanoma-specific mechanism and/or function of muramic acid. Gaining further functional understanding of the potential mechanistic role of muramic acid with respect to the pathogenesis of melanoma and its impact on the immune system will be part of our further research lines. The consistent separation between melanoma and control groups in our PLS-DA mirrors the distinct metabolic profiles often observed in patients with cancer compared to healthy individuals.[111]^13^,[112]^31 This distinction in metabolic profiles is pivotal for the identification of potential biomarkers for early detection and monitoring of melanoma, as previous metabolomics studies suggested for various cancers.[113]^32 Bayci et al. had previously identified serum-based biomarkers for melanoma using targeted metabolomics.[114]^21 However, in contrast to our data, by performing targeted metabolomics, they clearly limited their potential targets. Further, they analyzed a single cohort only consisting of 26 patients with melanoma. Thus, our data, generated by performing untargeted serum metabolomics, seem more relevant for the clinical setting, particularly since we additionally validated our findings from our much larger cohort of patients with melanoma in an independent validation cohort. Although the identification of stage-specific biomarkers is important, our current sample sizes within individual stages were not sufficient to allow for statistically robust stratified analyses. To address this limitation, we focused on identifying biomarkers that were consistently altered across cohorts and stages. At the clinical level, our data provide evidence for a number of highly promising biomarkers for the diagnosis of melanoma. By featuring an AUC of at least 0.9544 in both the exploratory as well as the validation cohort, six of our identified biomarkers are able to accurately diagnose melanoma in a clinical setting. Particularly our lead biomarker, namely muramic acid, represents a very promising accuracy across all melanoma stages. Considering their AUC, they might be clearly superior to existing biomarkers such as S-100B or lactate dehydrogenase (LDH) in this regard. The discriminative ability of serum S-100B to identify disease relapse (pooled area under the ROC [AUROC] 78.64 [95% confidence interval (CI) 70.28; 87.01]) was significantly greater than the discriminative ability of serum LDH (AUROC 64.41 [95% CI 56.05; 7278]) (p = 0.013). Both are far lower than metabolite biomarkers described herein.[115]^33 Our serum biomarker could tremendously enhance the current diagnostic standards. First, our data suggest that the accuracy of our biomarker is outstandingly high, which will significantly enhance the accuracy of diagnosis and thus the detection rate of melanoma. Secondly, by performing an easy serum test, the screening using our biomarker can be easily performed across the population (at risk, but also not at risk) even by primary care physicians. Thus, the screening rates for melanoma would tremendously increase. Thirdly, the serum test would indicate the presence of melanoma. Thus, a positive serum test requires effort by the treating physician/dermatologist to identify the melanoma. Current screening strategies are dependent on visually detecting the melanoma, e.g., by dermoscopy. However, it can happen that, e.g., on the head or other regions of the body where visibility of the skin is restricted (e.g., by hairs, etc.), melanoma is not detected. Our biomarker would overcome this problem, since a positive test would require an extensive search for the melanoma until it has indeed been detected and found. Fourthly, several naevi/skin lesions can macroscopically look similar to melanomas, and frequently, histology needs to be taken to determine whether melanoma is present or not. Here, our biomarkers would tremendously help, since a positive serum screening test would support the decision-making process of the treating physician to indeed take out the suspicious lesion. Thus, overall, there are many factors explaining how our serum biomarker would tremendously enhance current diagnostic standards and improve patient care. Overall, our data provide a comprehensive description of the entire serum metabolome composition in patients with melanoma. They demonstrate that patients with melanoma have a distinctly different metabolomics serum profile compared to healthy controls and that specific serum metabolites may serve as tumor markers for melanoma diagnosis, which could be of significant translational importance for patient care. We acknowledge that the identification of metabolites was based on accurate mass and database matching, which does not allow for definitive structural confirmation, especially in distinguishing isomers. Therefore, a limitation of this work is the absence of orthogonal validation using complementary analytical techniques. Future studies should aim to validate the identity and diagnostic relevance of the key metabolites identified here using targeted approaches such as liquid chromatography-tandem mass spectrometry or NMR spectroscopy. These methods would enable unambiguous confirmation of metabolite structures and provide quantitative measurements, thereby enhancing the translational applicability of our findings. We have highlighted this as an important direction for follow-up investigations. While muramic acid levels clearly distinguished patients with melanoma from healthy controls in our main analysis, we observed no statistically significant differences between healthy individuals and patients with other diseases in these independent cohorts. Those findings suggest that muramic acid is not broadly altered across unrelated inflammatory or neoplastic conditions, thereby supporting its potential specificity for melanoma. In conclusion, our study corroborates and extends the current understanding of melanoma metabolism, emphasizing the potential of metabolomics in unraveling the complex metabolic alterations associated with melanoma. These insights pave the way for the exploration of therapeutic targets and biomarkers, contributing to the evolving landscape of melanoma research and treatment. Limitations of the study While our study identifies six promising serum metabolite biomarkers for melanoma through an exploratory and validation cohort design, several limitations must be considered regarding the clinical utility and generalizability of our findings. First, the metabolomics profiling was performed using an Agilent QTOF 6550 mass spectrometer, which offers high-resolution data but cannot determine absolute metabolite concentrations. As a result, fold changes rather than absolute values were used for biomarker selection, which may limit direct clinical translation or threshold-based diagnostic development. Second, all serum samples were collected exclusively from the Zurich region. This geographic constraint may restrict the demographic and environmental diversity of the cohort. Expanding the study to include a more globally representative population is essential to ensure broader applicability and to develop robust, standardized biomarker panels suitable for global use. Finally, although we employed a two-phase discovery validation design, the control cohort was relatively small (n = 18). Further validation in larger, multi-center studies will be crucial to confirm the diagnostic performance, specificity, and clinical utility of the proposed biomarkers. Resource availability Lead contact Requests for further information, resources, and reagents should be directed to and will be fulfilled by the lead contact, Michael Scharl (michael.scharl@usz.ch). Materials availability This study did not generate new, unique reagents. Data and code availability * • The datasets generated and/or analyzed during the current study are available from Zenodo: [116]https://doi.org/10.5281/zenodo.15845162. Due to intellectual property considerations and patent-related restrictions, the data are not publicly available; however, they can be made accessible upon reasonable request and with permission from the corresponding author. * • The code was deposited in GitHub and is mentioned in the [117]key resources table. * • Any additional information required to reanalyze the data reported in this work is available from the [118]lead contact upon request. Acknowledgments