ABSTRACT
Background
Skin cancer is the most common type of cancer worldwide, categorized into melanoma and non‐melanoma types. While genetics play a crucial role in the development of skin cancer, recent developments in metabolomics have revealed that circulating metabolites influence genetic risk profiles.
Methods
In this study, we utilized a bidirectional two‐sample Mendelian Randomization (MR) approach, analyzing metabolite datasets from the Canadian Longitudinal Study of Aging (CLSA) and skin cancer datasets from the UK Biobank. Replication analyses were performed with the 10th round of the FinnGen project. Instrumental variables (IVs) were chosen according to established MR criteria and assessed through the Wald ratio, IVW, MR‐Egger, and weighted median methods. Various sensitivity analyses were conducted utilizing Cochrane's Q, Egger's intercept, MR‐PRESSO, and leave‐one‐out (LOO) techniques.
Results
After meticulous analysis, we identified causal relationships between circulating metabolites and skin cancer risks. Specifically, levels of 1,2‐dilinoleoyl‐GPE (18:2/18:2), 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2), 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4), 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2), and Docosapentaenoate n3 DPA, the ratio of Arachidonate (20:4n6) to oleate to vaccenate (18:1), and Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) were linked to non‐melanoma skin cancer, whereas levels of trans‐urocanate were causally linked to melanoma of skin, and S‐adenosylhomocysteine (SAH) showed a putative association. It should be noted that docosapentaenoate n3 DPA and SAH were each supported by a single SNP and should be interpreted as hypothesis‐generating rather than definitive causal evidence. Reverse MR analysis highlighted robust causal relationships with 18 and 10 metabolites and non‐melanoma skin cancer and melanoma of skin, respectively.
Conclusion
These findings enhance our understanding of skin cancer etiology and risk factors, providing genetic evidence for hypothesis‐generating associations that support the development of targeted therapies and aid in the identification of potential biomarkers.
Keywords: melanoma, mendelian randomization, metabolites, skin cancer
1. Introduction
Skin cancer is globally recognized as the most prevalent type of cancer, broadly categorized into melanoma and non‐melanoma types. While melanoma is less common, it is the fifth most frequent type of new cancer case and is increasing annually by 2%–3% [1]. It originates from the melanocytes, the cells responsible for pigment production in the skin. In contrast, non‐melanoma skin cancers, which include basal cell carcinoma and squamous cell carcinoma, are more prevalent but generally less aggressive. Key risk factors for developing skin cancer include exposure to ultraviolet (UV) rays, genetic factors, fair complexion, and a history of sunburns or using tanning beds [2]. Despite extensive research into these factors, the molecular mechanisms underlying skin cancer development, especially how genetic factors interact with metabolic processes, are still not fully mapped out.
Metabolomics, the comprehensive study of metabolites within a biological system, has emerged as a crucial field in understanding disease pathogenesis. Metabolites are small molecules involved in metabolism and the products of cellular processes. They provide a snapshot of the physiological state and can reflect changes due to genetic, environmental, or lifestyle factors [3]. Recent research suggests that metabolites circulating in the blood might be indicators useful for early detection and prognosis of cancer [4]. Exploring how metabolic changes influence skin cancer could unlock novel pathways implicated in carcinogenesis and identify potential intervention points for prevention and treatment.
Although traditional observational studies have provided us with a wealth of data on the association between metabolites and skin cancers, these studies often face problems of confounding, selection bias, and reverse causality. Mendelian randomization (MR), which utilizes genetic variants as instrumental variables (IVs), has been proposed to determine the causal effects of exposures (like metabolite levels) on outcomes (such as skin cancer risk). MR operates on the principle that genetic variants are assorted randomly during meiosis and remain unaffected by the typical confounders seen in observational studies. Therefore, MR can offer more dependable causation evidence than conventional observational approaches.
The primary aim of this research is to investigate the causal links between circulating metabolites and different types of skin cancer (melanoma and non‐melanoma) using a bidirectional two‐sample MR method. This study aims to provide genetic evidence of how metabolic changes contribute to the development of skin cancer and to pinpoint potential metabolic markers for early detection and specific treatments.
2. Material and Methods
2.1. Overview
Our research complies with the guidelines for strengthening Mendelian randomization research reports (STROBE‐MR) [5]. The overall design is illustrated in Figure 1. The focus of our analysis is on the causal relationship between genetically predicted circulating metabolites and the risk of different skin cancer types. Our MR study is founded on three primary assumptions: The relevance assumption ensures IVs are strongly linked with the exposure (blood metabolites), demonstrating significant associations; The Independence Assumption confirms that the IVs operate free from confounders, with no direct effects on the outcome (skin cancer); the Exclusion Restriction Assumption maintains that the IVs influence the outcome exclusively through their impact on the exposure, without any alternative pathways. Employing a two‐sample MR allows us to address the limitations of assessing both exposure and outcome within the same cohort, thus boosting the statistical robustness of our findings. This method also reduces the susceptibility to the winner's curse bias seen in single‐sample MR analyses [6]. Additionally, utilizing bidirectional MR helps in accurately delineating the causal directions between the exposure and outcomes, effectively preventing issues of reverse causality [7].
FIGURE 1.
Flowchart of three hypotheses and corresponding analysis strategies.
2.2. GWAS Data Sources
We utilized the comprehensive circulating metabolites dataset from The Canadian Longitudinal Study of Aging (CLSA) as the exposure dataset for our study, with its summary data accessible in the GWAS catalog (accession numbers from GCST90199621 to GCST90201020) [8]. The metabolites were classified according to super‐pathways and sub‐pathways, respectively (Table S1). For our primary analysis on the outcomes of melanoma and non‐melanoma skin cancers, we employed summary GWAS data from the UK Biobank [9]. The UK Biobank dataset encompasses a diverse population across the UK, ensuring there is no overlap with the Canadian CLSA dataset. To confirm our findings, we also used summary GWAS data from the FinnGen project, a comprehensive nationwide GWAS meta‐analysis incorporating data from nine Finnish biobanks10 [10]. Further details on these datasets are available in the Supporting Information.
2.3. IVs Selection
In alignment with the foundational principles of MR, our selection of IVs adheres to three essential assumptions. For the first assumption, we adopted a strict p‐value threshold of less than 5*10−8 to identify SNPs linked to skin cancer and circulating metabolites. Using the 1000 Genomes Project data on a European population, we excluded any SNPs within a 1000 kb radius that showed linkage disequilibrium (LD) with an R2 exceeding 0.01. To address the second assumption, we evaluated horizontal pleiotropy through the Egger intercept and the MR‐PRESSO test and performed leave‐one‐out (LOO) analysis to detect any outliers. For the third assumption, we excluded any SNPs that were directly associated with the outcomes, applying the same stringent p‐value criterion.
2.4. MR Analysis
Initially, we harmonized the SNP data across the exposure and outcome datasets, eliminating any palindromic sequences. For phenotypes linked to a single SNP, we applied the Wald ratio method. In cases involving two SNPs or when heterogeneity tests returned a p‐value of ≥ 0.05, we used the fixed effects inverse‐variance weighted (IVW) method. For phenotypes with more than three SNPs and the heterogeneity test showing a p‐value < 0.05, we implemented the random effects IVW method as the main analytical approach. Additionally, to ensure the reliability of our causal inferences, we used the MR Egger and weighted median methods alongside IVW as supplementary methods. For phenotypes instrumented by only one or two SNPs, the Wald ratio or fixed‐effects IVW was used as the primary estimate; in these cases, assessments of heterogeneity (Cochrane's Q), horizontal pleiotropy (Egger's intercept, MR‐PRESSO), and leave‐one‐out stability could not be performed. Consequently, associations derived from such limited instruments are considered preliminary and hypothesis‐generating.
2.5. Heterogeneity Analysis and Directional Test
Cochrane's Q test was used to assess heterogeneity among SNPs, with a p < 0.05 signifying heterogeneity. Based on this, we pinpointed candidate metabolites using specific criteria: (i) uniformity in effects across three MR methods; (ii) no evidence of horizontal pleiotropy; (iii) absence of outliers in LOO analysis. Moreover, the Steiger test was employed to determine if reverse causation influenced the observed causal relationships. If SNPs contribute more significantly to the genetic risk of skin cancer than to metabolite levels (Steiger p > 0.05), this indicates a possible bias in the causal direction. Additionally, reverse MR analysis was performed to explore how cancer might influence metabolite levels.
2.6. Statistical Analysis
All statistical computations were carried out using R software (version 4.3.2), employing the packages TwoSampleMR, MendelianRandomization, ieugwasr, mr. raps, meta, and plinkbin. The robustness of IVs was evaluated using the F statistic and the variance explained ratio (R2), with only IVs displaying an F statistic greater than 10 being considered strong enough for inclusion in our analyses. Odds ratios (OR) are utilized to report the findings from MR, with p‐values adjusted via the False Discovery Rate (FDR) approach, and a significance level set at 0.05. We conducted meta‐analyses using both common and random effects models to synthesize findings across various datasets. Heterogeneity was quantified using the restricted maximum likelihood method indicated by Tau2, while the I2 statistic was used to quantify the proportion of total variation attributable to heterogeneity.
3. Results
3.1. Characteristics of the Included IVs
Following efficacy validation, we identified 3587 SNPs, with each phenotype associated with 1 to 38 SNPs. The mean R2 was 1.40%, with individual values spanning from 0.36% to 23.26%. The mean F‐statistic across these SNPs was 116.87, ranging from 29.71 to 2297.79, confirming that all SNPs serve as robust instruments for MR analysis (F‐statistic > 10, Table S2).
3.2. Main MR Analysis Results
In our initial analysis using data from the UK Biobank, we established causal links between the genetic markers of 11 circulating metabolites and the risk of non‐melanoma skin cancer (Table S3). Additionally, causal associations were detected for 9 metabolites with melanoma risk, following adjustments for sub‐pathways (Table S4). Further analyses were performed using MR‐Egger and weighted median approaches for metabolites represented by three or more IVs to validate the consistency of our causal conclusions. These analyses confirmed that the impact of these metabolites on skin cancers remained consistent across various methods (Table S5).
3.3. Sensitivity Analysis
The vast majority of the causal estimates did not show any heterogeneity or pleiotropy, except for the effect of 4‐guanidinobutanoate levels on the risk of non‐melanoma skin cancer (Table S6), which was verified by funnel plots and scatter plots (Figure S1~S4). MR PRESSO also did not find any outliers in the datasets, indicating good robustness of the results. In the LOO analysis, we did not observe any shift in the causality between metabolites and skin cancers after excluding SNPs (Figures S5 and S6), suggesting that MR results were not driven by any outliers. Through the Steiger test, we further validated the causal direction of these metabolites to skin cancer.
The majority of our causality analyses revealed no signs of heterogeneity or pleiotropy, with the notable exception of the influence of 4‐guanidinobutanoate levels on non‐melanoma skin cancer risk (Table S6). The results were corroborated using funnel plots and scatter plots (Figure S1 to S4). Additionally, the MR PRESSO tool confirmed the absence of outliers in our datasets, enhancing the reliability of our results. In LOO analysis, the removal of SNPs did not alter the causal relationship between metabolites and skin cancers (Figures S5 and S6), indicating that our MR findings were not influenced by any specific outliers. The Steiger test further substantiated the direction of causality from these metabolites towards skin cancer.
3.4. Replication Analysis
We performed a replication analysis using datasets from the FinnGen project. This analysis confirmed the causal links between non‐melanoma skin cancer and various metabolites, including 1,2‐dilinoleoyl‐GPE (18:2/18:2), 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2), 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4), 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2). Additionally, the ratios of arachidonate (20:4n6) to oleate to vaccenate (18:1), and Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) were also linked to non‐melanoma skin cancer. Docosapentaenoate n3 DPA showed a putative association based on a single‐SNP Wald ratio estimate; pleiotropy and heterogeneity tests were not computable. The replication also verified the relationship between trans‐urocanate and melanoma. S‐adenosylhomocysteine (SAH) showed a putative association based on a single‐SNP Wald ratio estimate; pleiotropy and heterogeneity assessments were not computable (Table S7, Figures 2, 3).
FIGURE 2.
Meta‐analysis of forward MR results of UKB and Finngen datasets of non‐melanoma skin cancer. (A.) 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels. (B) 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels. (C) 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels. (D) Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) ratio. (E) 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels. (F) Arachidonate (20:4n6) to oleate to vaccenate (18:1) ratio. (G) Docosapentaenoate n3 DPA levels (single‐SNP Wald ratio; hypothesis‐generating).
FIGURE 3.
Meta‐analysis of forward MR results of UKB and Finngen datasets of skin melanoma. (A) Trans‐urocanate levels. (B) S‐adenosylhomocysteine (SAH) levels (single‐SNP Wald ratio; hypothesis‐generating).
3.5. Reverse MR Analysis
In our reverse analysis, we utilized 63 SNPs, as detailed in Table S8. Analysis of data from the UK Biobank revealed causal associations between 92 metabolites and non‐melanoma skin cancer and 74 metabolites with melanoma, as shown in Tables S9 and S10. However, inconsistencies were observed in the effects of skin cancers on certain metabolites across different MR methods. We then intersected the findings from the UK Biobank with those from the FinnGen datasets, eventually identifying 18 metabolites linked to non‐melanoma skin cancer and 9 linked to melanoma that demonstrated consistent causal relationships (Figures 4, 5, Table S11).
FIGURE 4.
Meta‐analysis of reverse MR results of UKB and Finngen datasets of non‐melanoma skin cancer. (A) 1‐(1‐enyl‐palmitoyl)‐2‐linoleoyl‐GPC (p‐16:0/18:2) levels. (B) 1‐(1‐enyl‐stearoyl)‐2‐linoleoyl‐GPE (p‐18:0/18:2) levels. (C) 1‐linolenoyl‐GPC (18:3) levels. (D) 2‐ketocaprylate levels. (E) 4‐hydroxychlorothalonil levels. (F) 5‐oxoproline levels. (G) Caproate (6:0) levels. (H) Indolepropionate levels. (I) N‐palmitoyl‐sphinganine (d18:0/16:0) levels. (J) Perfluorooctanoate (PFOA) levels. (K) Salicylate to taurocholate ratio. (L) X‐21258 levels. (M) Phosphate to 5‐oxoproline ratio. (N) Phosphate to 2′‐deoxyuridine ratio. (O) N‐carbamoylalanine levels. (P) Methylsuccinoylcarnitine levels. (Q) Kynurenine levels. (R) N‐acetyl‐2‐aminoadipate levels.
FIGURE 5.
Meta‐analysis of reverse MR results of UKB and Finngen datasets of skin melanoma. (A) 3‐(4‐hydroxyphenyl) lactate levels. (B) 3‐ethylcatechol sulfate (2) levels. (C) 4‐hydroxy‐2‐oxoglutaric acid levels. (D) Alanine to asparagine ratio. (E) N‐lactoyl phenylalanine levels. (F) N‐methylhydroxyproline levels. (G) X‐21834 levels. (H) Nonanoylcarnitine (C9) levels. (I) Gamma‐glutamylglutamine levels.
4. Discussion
Through comprehensive analysis, we identified several metabolites with strong causal associations to skin cancer. Among the seven metabolites ultimately identified to be associated with non‐melanoma skin cancer, three are glycerophosphatidylcholine (GPC), phosphatidylcholine has long been considered a key metabolite in cancer and GPC is its core pathway [11, 12]. As mentioned above, UV is a clear risk factor for skin cancer and Jeffrey B Travers et al. demonstrated that endogenous GPC acts as a mediator in UV‐induced activation of peroxisome proliferator‐activated receptor γ (PPAR γ) to regulate the inflammation and cytokine release in epidermal cells [13]. UV can also directly induce the production of platelet activating factor (PAF) in the skin, promoting GPC oxidation and further regulating cell apoptosis and photocarcinogenic effects [14, 15, 16]. In addition to its direct effect on epithelial cells, GPC may also be involved in shaping the immune microenvironment of tumors. Xiaoyu Yang et al. illustrated that GPC is the major metabolite of human preputial fibroblasts altered by UV damage [17]. Langni Liu et al. further found that UV induces keratinocytes to produce microvesicular particles containing high levels of GPC, which in turn induces systemic immunosuppression that contributes to photocarcinogenesis [18]. Notably, 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) and 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) exhibited protective effects while 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) exhibited inducing effects, which may be due to the fact that the former two contain linoleic acid with anti‐inflammatory and antioxidant properties. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4), on the other hand, has pro‐inflammatory properties due to the presence of arachidonic acid. Besides, 1,2‐Dilinoleoyl‐GPE (18:2/18:2) is a glycerophosphoethanolamine (GPE) lipid involved in the composition and fluidity of cell membranes. The presence of linoleic acid (18:2) in this metabolite suggests a role for polyunsaturated fatty acids (PUFAs) in skin cancer pathogenesis. PUFAs are known for their anti‐inflammatory, cell growth and differentiation regulation, and oxidative stress alleviating properties, which can play a role in cancer prevention [19]. Given the protective role of PUFAs, the levels of 1,2‐dilinoleoyl‐GPE may indicate altered lipid metabolism and oxidative stress contributing to non‐melanoma skin cancer development.
It is important to contextualize our circulating metabolite findings within the established biology of cutaneous melanogenesis. Melanin synthesis—specifically the balance between photoprotective eumelanin and pro‐oxidant pheomelanin—and associated pigmentary pathways are fundamental determinants of melanocyte function, UV sensitivity, and skin cancer risk [20, 21, 22, 23]. For instance, pheomelanin has been shown to promote oxidative stress and photocarcinogenesis, whereas eumelanin confers relative photoprotection. Our Mendelian randomization framework captures systemic metabolic associations, but we acknowledge that local pigment biochemistry—including tyrosine metabolism, melanin intermediate pathways, and follicular melanocyte activity—may act as critical effect modifiers that are not fully reflected in plasma‐based GWAS instruments. Future studies integrating skin‐specific eQTL and metabolite data are needed to bridge this gap.
When it comes to skin melanoma, we found that trans‐urocanate and SAH may play an important role in its development. Trans‐urocanate is a metabolite derived from histidine, widely present in the skin, and serves as a natural UV filter [24]. Beyond its direct photoprotective role, trans‐urocanate functions as a photometabolite with immunomodulatory properties, suppressing immune responses and alleviating inflammation [25]. Skin inflammation is closely related to melanoma occurrence, suggesting that trans‐urocanate may influence melanoma risk through UV‐induced immunomodulation [26, 27]. However, it is important to recognize that trans‐urocanate represents only one component of a broader landscape of UVR‐driven metabolic alterations. UV radiation induces diverse photometabolites and photoproducts—including reactive oxygen species, DNA photoproducts (e.g., cyclobutane pyrimidine dimers), and lipid peroxidation products—that regulate immune responses, oxidative stress, and DNA damage repair [28, 29, 30]. These cutaneous photometabolic pathways are particularly relevant to non‐melanoma skin cancer pathogenesis. Our circulating metabolite findings should therefore be interpreted as complementary to, rather than replacing, local photometabolic research. SAH is a byproduct of methyl transfer from S‐adenosylmethionine and a negative regulator of methylation reactions [31]. However, given that this association rests on a single genetic instrument (Wald ratio), we refrain from extensive mechanistic interpretation and present it as a hypothesis‐generating finding warranting further investigation with multi‐SNP instruments or experimental validation. Although previous studies have provided numerous clues about the association and potential mechanisms of these metabolites with skin cancer, their results are often ambiguous [32]. These ambiguous findings can be attributed to the fact that observational studies involving serum metabolites are highly sensitive to confounding factors such as environmental effects, dietary patterns, and pathophysiological states, and our findings based on MR provide strong genetic evidence.
Forward MR analyses reveal the origins and risk factors of diseases, offering genetic insights crucial for developing therapeutic interventions. Conversely, reverse MR analyses help in deciphering disease pathways, potentially indicating new diagnostic or prognostic indicators. Through detailed analysis, we have identified significant circulating metabolites that are causally linked to skin cancer. Beyond the GPC and GPE derivatives identified in the forward analysis, certain amino acid metabolites like N‐carbamoylalanine and kynurenine are also causally connected to non‐melanoma skin cancer. N‐carbamoylalanine, for instance, is involved in synthetic lethality in colorectal cancer cells [33]. Research by Skaidre Jankovskaja et al. suggests that the tryptophan/kynurenine ratio could act as a non‐invasive biomarker for skin cancer [34]. The metabolites associated with melanoma significantly differ from those linked to non‐melanoma skin cancer, reflecting distinct metabolic phenotypes. These metabolites affect processes such as cellular proliferation, oxidative stress responses, and tissue remodeling. Given their absence in previous observational studies, further investigations are required to explore their mechanisms and confirm their potential as melanoma biomarkers.
Our findings are grounded in Mendel's laws of genetics, leveraging the principle that alleles randomly assigned at conception are less likely to be influenced by environmental confounders, thus enabling more effective control over confounding variables. Unlike observational studies that struggle to distinguish between exposure and outcome, MR studies can provide evidence of causality and circumvent reverse causation issues. However, our study also has limitations: firstly, both our exposure and outcome data are derived from European populations, lacking concrete evidence to affirm the same associations exist across other ethnic groups; secondly, some of the causal relationships are mediated by only one or two SNPs, making sensitivity analysis challenging and thus weakening the reliability of our findings—this constraint applies specifically to docosapentaenoate n3 DPA (non‐melanoma skin cancer) and S‐adenosylhomocysteine (melanoma), which are presented as putative associations rather than established causal relationships; thirdly, our study relies exclusively on circulating (systemic) metabolites, whereas many key photometabolites and pigment‐related intermediates are generated locally in the skin and may not be adequately reflected in plasma‐based GWAS datasets [35]. This creates a significant translational gap: the metabolite‐cancer associations we report may reflect systemic predisposition or distant metabolic effects rather than direct cutaneous mechanisms. Validation using skin tissue‐specific metabolomics, single‐cell metabolite profiling, and spatial transcriptomics will be essential to bridge this gap.
5. Conclusion
Our bidirectional two‐sample MR study has uncovered potential causal relationships and a prioritized set of candidate correlations between circulating metabolites and the risk of both melanoma and non‐melanoma skin cancers. These findings enhance our understanding of the metabolic alterations associated with skin cancer and highlight hypothesis‐generating biomarkers and therapeutic targets. Continued research in this area holds promise for improving the early detection, prevention, and treatment of skin cancer, ultimately contributing to better patient outcomes.
Author Contributions
Jiangfan Yu: conceptualization, methodology, software, data curation, investigation, validation, writing – review and editing, writing – original draft, formal analysis. Hongyan Gong: supervision, data curation, software. Zhuotong Zeng: supervision, project administration, resources. Rong Xiao: conceptualization, methodology, project administration, resources, visualization.
Funding
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figure S1: Funnel plot of MR heterogeneity between circulating metabolites and non‐melanoma skin cancer. A. 4‐guanidinobutanoate levels; B. 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels; C. 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels; D. Nervonoylcarnitine (C24:1) levels; E. 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels; F. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels; G. Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) [2] to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) [1] ratio.
Figure S2: Funnel plot of MR heterogeneity between circulating metabolites and melanoma of skin. A. 4‐guanidinobutanoate levels; B. 5alpha‐androstan‐3beta, 17beta‐diol monosulfate (2) levels; C. Andro steroid monosulfate C19H28O6S (1) levels; D. 16a‐hydroxy DHEA 3‐sulfate levels; E. Trans‐4‐hydroxyproline levels; F. Trans‐urocanate levels; G. X‐12100 levels.
Figure S3: Scatter plot of MR‐Egger intercept between circulating metabolites and non‐melanoma skin cancer. A. 4‐guanidinobutanoate levels; B. 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels; C. 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels; D. Nervonoylcarnitine (C24:1) levels; E. 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels; F. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels; G. Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) [2] to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) [1].
Figure S4: Scatter plot of MR‐Egger intercept between circulating metabolites and melanoma of skin. A. 4‐guanidinobutanoate levels; B. 5alpha‐androstan‐3beta,17beta‐diol monosulfate (2) levels; C. Andro steroid monosulfate C19H28O6S (1) levels; D. 16a‐hydroxy DHEA 3‐sulfate levels; E. Trans‐4‐hydroxyproline levels; F. Trans‐urocanate levels; G. X‐12100 levels.
Figure S5: LOO analysis for MR between circulating metabolites and non‐melanoma skin cancer. A. 4‐guanidinobutanoate levels; B. 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels; C. 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels; D. Nervonoylcarnitine (C24:1) levels; E. 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels; F. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels; G. Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) [2] to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) [1].
Figure S6: LOO analysis for MR between circulating metabolites and melanoma of skin. A. 4‐guanidinobutanoate levels; B. 5alpha‐androstan‐3beta,17beta‐diol monosulfate (2) levels; C. Andro steroid monosulfate C19H28O6S (1) levels; D. 16a‐hydroxy DHEA 3‐sulfate levels; E. Trans‐4‐hydroxyproline levels; F. Trans‐urocanate levels; G. X‐12100 levels.
Table S1: Serum metabolites classification.
Table S2: Included IVs for MR analysis.
Table S3: MR analysis for non‐melanoma skin cancer and serum metabolites.
Table S4: MR analysis for melanoma and serum metabolites.
Table S5: Summary for four main MR method.
Table S6: Sensitivity analysis.
Table S7: Validation of MR results with Finngen datasets.
Table S8: Included IVs for reverse MR analysis.
Table S9: Reverse MR analysis for non‐melanoma skin cancer.
Table S10: Reverse MR analysis for melanoma of skin.
Table S11: Validation of reverse MR results with Finngen datasets.
Contributor Information
Zhuotong Zeng, Email: zengzhuotong@csu.edu.cn.
Rong Xiao, Email: xiaorong65@csu.edu.cn.
Data Availability Statement
The summary‐level GWAS data for circulating metabolites from the Canadian Longitudinal Study of Aging (CLSA) are available in the GWAS catalog under accession numbers GCST90199621 to GCST90201020. Summary GWAS data for melanoma and non‐melanoma skin cancers were obtained from the UK Biobank. Replication summary GWAS data were obtained from the FinnGen project.
References
- 1. Siegel R. L., Giaquinto A. N., and Jemal A., “Cancer Statistics, 2024,” CA: A Cancer Journal for Clinicians 74 (2024): 12–49. [DOI] [PubMed] [Google Scholar]
- 2. Perez M., Abisaad J. A., Rojas K. D., Marchetti M. A., and Jaimes N., “Skin Cancer: Primary, Secondary, and Tertiary Prevention. Part I,” Journal of the American Academy of Dermatology 87 (2022): 255–268. [DOI] [PubMed] [Google Scholar]
- 3. Hui S., Cowan A. J., Zeng X., et al., “Quantitative Fluxomics of Circulating Metabolites,” Cell Metabolism 32 (2020): 676–688. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Hagyousif Y. A., Sharaf B. M., Zenati R. A., et al., “Skin Cancer Metabolic Profile Assessed by Different Analytical Platforms,” International Journal of Molecular Sciences 24 (2023): 1604. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Skrivankova V. W., Richmond R. C., Woolf B. A. R., et al., “Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization: The STROBE‐MR Statement,” JAMA 326 (2021): 1614–1621. [DOI] [PubMed] [Google Scholar]
- 6. Forde A., Hemani G., and Ferguson J., “Review and Further Developments in Statistical Corrections for Winner's Curse in Genetic Association Studies,” PLoS Genetics 19 (2023): e1010546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Davey Smith G. and Hemani G., “Mendelian Randomization: Genetic Anchors for Causal Inference in Epidemiological Studies,” Human Molecular Genetics 23 (2014): R89–R98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Chen Y., Lu T., Pettersson‐Kymmer U., et al., “Genomic Atlas of the Plasma Metabolome Prioritizes Metabolites Implicated in Human Diseases,” Nature Genetics 55 (2023): 44–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Sudlow C., Gallacher J., Allen N., et al., “UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age,” PLoS Medicine 12 (2015): e1001779. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Kurki M. I., Karjalainen J., Palta P., et al., “FinnGen Provides Genetic Insights From a Well‐Phenotyped Isolated Population,” Nature 613 (2023): 508–518. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Glunde K., Penet M. F., Jiang L., Jacobs M. A., and Bhujwalla Z. M., “Choline Metabolism‐Based Molecular Diagnosis of Cancer: An Update,” Expert Review of Molecular Diagnostics 15 (2015): 735–747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Sonkar K., Ayyappan V., Tressler C. M., et al., “Focus on the Glycerophosphocholine Pathway in Choline Phospholipid Metabolism of Cancer,” NMR in Biomedicine 32 (2019): e4112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Zhang Q., Southall M. D., Mezsick S. M., et al., “Epidermal Peroxisome Proliferator‐Activated Receptor Gamma as a Target for Ultraviolet B Radiation,” Journal of Biological Chemistry 280 (2005): 73–79. [DOI] [PubMed] [Google Scholar]
- 14. Travers J. B., Berry D., Yao Y., Yi Q., Konger R. L., and Travers J. B., “Ultraviolet B Radiation of Human Skin Generates Platelet‐Activating Factor Receptor Agonists,” Photochemistry and Photobiology 86 (2010): 949–954. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Barber L. A., Spandau D. F., Rathman S. C., et al., “Expression of the Platelet‐Activating Factor Receptor Results in Enhanced Ultraviolet B Radiation‐Induced Apoptosis in a Human Epidermal Cell Line,” Journal of Biological Chemistry 273 (1998): 18891–18897. [DOI] [PubMed] [Google Scholar]
- 16. Sahu R. P., DaSilva S. C., Rashid B., et al., “Mice Lacking Epidermal PPARγ Exhibit a Marked Augmentation in Photocarcinogenesis Associated With Increased UVB‐Induced Apoptosis, Inflammation and Barrier Dysfunction,” International Journal of Cancer 131 (2012): E1055–E1066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Yang X., Wang J., Wang H., Li X., He C., and Liu L., “Metabolomics Study of Fibroblasts Damaged by UVB and BaP,” Scientific Reports 11 (2021): 11176. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Liu L., Awoyemi A. A., Fahy K. E., et al., “Keratinocyte‐Derived Microvesicle Particles Mediate Ultraviolet B Radiation‐Induced Systemic Immunosuppression,” Journal of Clinical Investigation 131 (2021): e144963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Miura K., Way M., Jiyad Z., et al., “Omega‐3 Fatty Acid Intake and Decreased Risk of Skin Cancer in Organ Transplant Recipients,” European Journal of Nutrition 60 (2021): 1897–1905. [DOI] [PubMed] [Google Scholar]
- 20. Slominski A., Tobin D. J., Shibahara S., and Wortsman J., “Melanin Pigmentation in Mammalian Skin and Its Hormonal Regulation,” Physiological Reviews 84 (2004): 1155–1228. [DOI] [PubMed] [Google Scholar]
- 21. Slominski A., Zmijewski M. A., and Pawelek J., “L‐Tyrosine and L‐Dihydroxyphenylalanine as Hormone‐Like Regulators of Melanocyte Functions,” Pigment Cell & Melanoma Research 25 (2012): 14–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Slominski R. M., Sarna T., Płonka P. M., Raman C., Brożyna A. A., and Slominski A. T., “Melanoma, Melanin, and Melanogenesis: The Yin and Yang Relationship,” Frontiers in Oncology 12 (2022): 842496. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Slominski R. M., Kim T. K., Janjetovic Z., et al., “Malignant Melanoma: An Overview, New Perspectives, and Vitamin D Signaling,” Cancers 16 (2024): 2262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Brosnan M. E. and Brosnan J. T., “Histidine Metabolism and Function,” Journal of Nutrition 150 (2020): 2570s–2575s. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Uksila J., Laihia J. K., and Jansén C. T., “Trans‐Urocanic Acid, a Natural Epidermal Constituent, Inhibits Human Natural Killer Cell Activity in Vitro,” Experimental Dermatology 3 (1994): 61–65. [DOI] [PubMed] [Google Scholar]
- 26. Hart P. H., Grimbaldeston M. A., and Finlay‐Jones J. J., “Sunlight, Immunosuppression and Skin Cancer: Role of Histamine and Mast Cells,” Clinical and Experimental Pharmacology & Physiology 28 (2001): 1–8. [DOI] [PubMed] [Google Scholar]
- 27. De Fabo E. C., Webber L. J., Ulman E. A., and Broemeling L. D., “Dietary L‐Histidine Regulates Murine Skin Levels of Trans‐Urocanic Acid, an Immune‐Regulating Photoreceptor, With an Unanticipated Modulation: Potential Relevance to Skin Cancer,” Journal of Nutrition 127 (1997): 2158–2164. [DOI] [PubMed] [Google Scholar]
- 28. Slominski R. M., Chen J. Y., Raman C., and Slominski A. T., “Photo‐Neuro‐Immuno‐Endocrinology: How the Ultraviolet Radiation Regulates the Body, Brain, and Immune System,” Proceedings of the National Academy of Sciences of the United States of America 121 (2024): e2308374121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Slominski R. M., Raman C., Jetten A. M., and Slominski A. T., “Neuro‐Immuno‐Endocrinology of the Skin: How Environment Regulates Body Homeostasis,” Nature Reviews. Endocrinology 21 (2025): 495–509. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Slominski A. T., Zmijewski M. A., Plonka P. M., Szaflarski J. P., and Paus R., “How UV Light Touches the Brain and Endocrine System Through Skin, and Why,” Endocrinology 159 (2018): 1992–2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Sah V. R., Karlsson J., Jespersen H., et al., “Epigenetic Therapy to Enhance Therapeutic Effects of PD‐1 Inhibition in Therapy‐Resistant Melanoma,” Melanoma Research 32 (2022): 241–248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Worrede A., Douglass S. M., and Weeraratna A. T., “The Dark Side of Daylight: Photoaging and the Tumor Microenvironment in Melanoma Progression,” Journal of Clinical Investigation 131 (2021): e143763. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Ahmad A., Prakash R., Khan M. S., et al., “N‐Carbamoyl Alanine‐Mediated Selective Targeting for CHEK2‐Null Colorectal Cancer,” ACS Omega 7 (2022): 13095–13101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Jankovskaja S., Engblom J., Rezeli M., Marko‐Varga G., Ruzgas T., and Björklund S., “Non‐Invasive Skin Sampling of Tryptophan/Kynurenine Ratio in Vitro Towards a Skin Cancer Biomarker,” Scientific Reports 11 (2021): 678. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Slominski R. M., Raman C., Chen J. Y., and Slominski A. T., “How Cancer Hijacks the Body's Homeostasis Through the Neuroendocrine System,” Trends in Neurosciences 46 (2023): 263–275. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Funnel plot of MR heterogeneity between circulating metabolites and non‐melanoma skin cancer. A. 4‐guanidinobutanoate levels; B. 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels; C. 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels; D. Nervonoylcarnitine (C24:1) levels; E. 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels; F. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels; G. Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) [2] to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) [1] ratio.
Figure S2: Funnel plot of MR heterogeneity between circulating metabolites and melanoma of skin. A. 4‐guanidinobutanoate levels; B. 5alpha‐androstan‐3beta, 17beta‐diol monosulfate (2) levels; C. Andro steroid monosulfate C19H28O6S (1) levels; D. 16a‐hydroxy DHEA 3‐sulfate levels; E. Trans‐4‐hydroxyproline levels; F. Trans‐urocanate levels; G. X‐12100 levels.
Figure S3: Scatter plot of MR‐Egger intercept between circulating metabolites and non‐melanoma skin cancer. A. 4‐guanidinobutanoate levels; B. 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels; C. 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels; D. Nervonoylcarnitine (C24:1) levels; E. 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels; F. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels; G. Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) [2] to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) [1].
Figure S4: Scatter plot of MR‐Egger intercept between circulating metabolites and melanoma of skin. A. 4‐guanidinobutanoate levels; B. 5alpha‐androstan‐3beta,17beta‐diol monosulfate (2) levels; C. Andro steroid monosulfate C19H28O6S (1) levels; D. 16a‐hydroxy DHEA 3‐sulfate levels; E. Trans‐4‐hydroxyproline levels; F. Trans‐urocanate levels; G. X‐12100 levels.
Figure S5: LOO analysis for MR between circulating metabolites and non‐melanoma skin cancer. A. 4‐guanidinobutanoate levels; B. 1‐stearoyl‐2‐linoleoyl‐gpc (18:0/18:2) levels; C. 1,2‐dilinoleoyl‐GPE (18:2/18:2) levels; D. Nervonoylcarnitine (C24:1) levels; E. 1‐palmitoyl‐2‐linoleoyl‐gpc (16:0/18:2) levels; F. 1‐stearoyl‐2‐arachidonoyl‐gpc (18:0/20:4) levels; G. Oleoyl‐linoleoyl‐glycerol (18:1 to 18:2) [2] to linoleoyl‐arachidonoyl‐glycerol (18:2 to 20:4) [1].
Figure S6: LOO analysis for MR between circulating metabolites and melanoma of skin. A. 4‐guanidinobutanoate levels; B. 5alpha‐androstan‐3beta,17beta‐diol monosulfate (2) levels; C. Andro steroid monosulfate C19H28O6S (1) levels; D. 16a‐hydroxy DHEA 3‐sulfate levels; E. Trans‐4‐hydroxyproline levels; F. Trans‐urocanate levels; G. X‐12100 levels.
Table S1: Serum metabolites classification.
Table S2: Included IVs for MR analysis.
Table S3: MR analysis for non‐melanoma skin cancer and serum metabolites.
Table S4: MR analysis for melanoma and serum metabolites.
Table S5: Summary for four main MR method.
Table S6: Sensitivity analysis.
Table S7: Validation of MR results with Finngen datasets.
Table S8: Included IVs for reverse MR analysis.
Table S9: Reverse MR analysis for non‐melanoma skin cancer.
Table S10: Reverse MR analysis for melanoma of skin.
Table S11: Validation of reverse MR results with Finngen datasets.
Data Availability Statement
The summary‐level GWAS data for circulating metabolites from the Canadian Longitudinal Study of Aging (CLSA) are available in the GWAS catalog under accession numbers GCST90199621 to GCST90201020. Summary GWAS data for melanoma and non‐melanoma skin cancers were obtained from the UK Biobank. Replication summary GWAS data were obtained from the FinnGen project.