Skip to main content
Download PDF

Higher polygenic risk for melanoma is associated with improved survival in a high ultraviolet radiation setting

Journal of Translational Medicine volume 20, Article number: 403 (2022) Cite this article

Abstract

Background

The role of germline genetic factors in determining survival from cutaneous melanoma (CM) is not well understood.

Objective

To perform a genome-wide association study (GWAS) meta-analysis of melanoma-specific survival (MSS), and test whether a CM-susceptibility polygenic risk score (PRS) is associated with MSS.

Methods

We conducted two Cox proportional-hazard GWAS of MSS using data from the Melanoma Institute Australia, a high ultraviolet (UV) radiation setting (MIA; 5,762 patients with melanoma; 800 melanoma deaths) and UK Biobank (UKB: 5,220 patients with melanoma; 241 melanoma deaths), and combined them in a fixed-effects meta-analysis. Significant (P < 5 × 10–8) results were investigated in the Leeds Melanoma Cohort (LMC; 1,947 patients with melanoma; 370 melanoma deaths). We also developed a CM-susceptibility PRS using a large independent GWAS meta-analysis (23,913 cases, 342,870 controls). The PRS was tested for an association with MSS in the MIA and UKB cohorts.

Results

Two loci were significantly associated with MSS in the meta-analysis of MIA and UKB with lead SNPs rs41309643 (G allele frequency 1.6%, HR = 2.09, 95%CI = 1.61–2.71, P = 2.08 × 10–8) on chromosome 1, and rs75682113 (C allele frequency 1.8%, HR = 2.38, 95%CI = 1.77–3.21, P = 1.07 × 10–8) on chromosome 7. While neither SNP replicated in the LMC, rs75682113 was significantly associated in the combined discovery and replication sets. After adjusting for age at diagnosis, sex and the first ten principal components, a one standard deviation increase in the CM-susceptibility PRS was associated with improved MSS in the discovery meta-analysis (HR = 0.88, 95% CI = 0.83–0.94, P = 6.93 × 10–5; I2 = 88%). However, this was only driven by the high UV setting cohort (MIA HR = 0.84, 95% CI = 0.78–0.90).

Conclusion

We found two loci potentially associated with MSS. Increased genetic susceptibility to develop CM is associated with improved MSS in a high UV setting.

Introduction

Cutaneous melanoma (CM) is the third most common skin cancer and is responsible for over 1,300 deaths in Australia annually [1] and more than 7,000 deaths in the United States of America (USA) [2]. While survival rates have been improving since 2013, likely due to advances in immunotherapies and BRAF-targeted therapies, management of CM remains a major public health burden, with an annual cost of over AUD 200 million in Australia and USD 24 billion in the US [3, 4].

CM-susceptibility is driven by host factors including skin pigmentation and number of naevi, as well as environmental factors, most importantly exposure to ultraviolet radiation [5,6,7,8,9]. Germline genetic factors can influence the risk of developing CM through modification of these host risk factors, and other biological pathways; genome‐wide association studies (GWAS) have identified over 50 CM-susceptibility loci [10].

Although there are well known prognostic factors for melanoma-specific survival (MSS) including primary tumour thickness, ulceration, mitotic rate, melanoma type, anatomical site and the stage of the tumour at diagnosis [11, 12], the role of host genetic factors in MSS is not well understood. Death of a relative from CM is associated with poorer MSS, raising the possibility that germline genetic factors influence survival [13]. Higher naevus count has been associated with improved survival [14]. Naevus count is strongly influenced by germline genetics [15, 16], and is the strongest risk factor for the development of melanoma [17], suggesting germline genetic risk for CM may also impact survival. Telomere length is another biological pathway to high genetic CM-susceptibility [18] and may also influence MSS [19].

A powerful approach to test whether germline genetic risk for a given disease or trait (e.g. risk for CM) influences another trait (e.g. MSS) is to combine individual genetic effects in a polygenic risk score (PRS). Death from all causes has been associated with the joint effect of PRSs associated with risk of a range of diseases (e.g. coronary artery disease, pancreatic cancer, and lung cancer) or associated with mortality risk factors (e.g. cholesterol, sleep duration) [20], suggesting that germline risk for development of a disease can help predict outcomes. However, it is not known whether a genetic predisposition to CM influences melanoma outcomes.

To explore these two questions, we first aimed to identify germline genetic factors that influence MSS by performing a large-scale GWAS of MSS. Following this we assessed whether a PRS for CM-susceptibility (referred to as PRS_susceptibility) was associated with MSS.

Methods and materials

Genome-wide association studies of melanoma-specific survival

Discovery cohort 1: Melanoma Institute Australia

Samples for this cohort were derived from the Melanoma Institute Australia (MIA) Biospecimen Bank (protocol HREC/10/RPAH/530) and patient information from the MIA Research Database (protocol HREC/11/RPAH/444). With written, informed consent, patients with histo-pathological confirmed CM cases managed at MIA, Sydney, Australia were identified from this Biospecimen Bank and Database. Participants’ clinical and biospecimen data were captured and prospectively collected follow- up for outcomes including death due to melanoma. MIA study protocols were approved by the Sydney Local Health District Ethics Review Committee, Royal Prince Alfred Hospital, Camperdown, Australia. Participants were genotyped in phases using the Oncoarray in 2014 and 2016, and the Global Screening Array in 2018 (Illumina, San Diego).

Full details of the GWAS data cleaning quality control for both MIA datasets have been previously reported [10, 21]. Briefly, for Oncoarray genotyped samples, individuals were removed based on high genotype missingness (> 3%), extreme heterozygosity (± 0.05 from the mean), being related to other samples (identified by descent pihat > 0.15), or were more than 6 standard deviations (SDs) from the means of principal components (PCs) 1 and 2 of a European reference population [10]. In addition, single nucleotide polymorphisms (SNPs) were removed if they were missing more than 3% of their calls, had a minor allele frequency (MAF) < 0.01, or their Hardy–Weinberg equilibrium (HWE) P-value was less than 5 × 10−10 for patients with melanoma or less than 5 × 10−4 in CM-free individuals in Landi et al., [10]. Individuals genotyped on the Global Screening Array were removed due to high genotype missingness (> 5%), non-European ancestry or relatedness (as above), and SNPs were excluded due to a low MAF (< 0.01), high missingness (> 5%), HWE P < 1 × 10–6, or a low GenTrain score (< 0.6) [21]. The cleaned genotyped data were batched by their genotyping array (Oncoarray and Global Screening Array) and imputed to the Haplotype Reference Consortium (v1) panel using the University of Michigan imputation server [22].

For this study, the primary endpoint was death from melanoma which was ascertained through MIA clinical records and linkage to Australian Cancer Registries (including the New South Wales Cancer Registry), electoral rolls, and the Birth and Death Register. This analysis was restricted to 5,672 participants of European ancestry diagnosed with CM. For participants with multiple CM, the first primary CM was used to define the start point. MSS survival time (in years) was defined as the duration between the date of diagnosis of the (first) primary CM, and the date of death due to melanoma. Patients were censored on the last day of follow-up or when they died of non-melanoma causes.

Discovery cohort 2: UK Biobank

UK Biobank (UKB) is a large population-based cohort of approximately 500,000 adult participants (40–70 years at recruitment) recruited with informed consent from the United Kingdom between 2006 and 2010. Participants were followed up for disease outcomes including death from melanoma. Details on participant recruitment, phenotype measurement and genotyping have been published elsewhere [23, 24]. In brief, participants were genotyped using the UK Biobank Axiom Array and the UK BiLEVE Axiom Array (Affymetrix Inc, California, USA) and imputed using the Haplotype Reference Consortium and UK10K reference panels. The study was approved by the United Kingdom’s National North West Multi-Centre Research Ethics Committee. For this present study, we included 5,220 participants of European ancestry with histo-pathologically confirmed invasive CM based on the International Classification of Diseases (ICD) 10 (UKB data field 40,006) and 9 (data field 40,013) and ICD for Oncology, 3rd edition codes (data field 40,011) for melanoma. Participants were then filtered for missingness (< 3%), relatedness (identity by descent pihat < 0.2), and population ancestry outliers (from the European reference). The primary endpoint was MSS which was ascertained through linkage of the participant records with Cancer Registries, electoral rolls, and the Birth and Death Register in the UK.

Replication cohort: Leeds Melanoma Cohort

The Leeds Melanoma Cohort (LMC) is a population-based cohort of 2,184 participants diagnosed with incident melanoma between September 2000 and December 2012 and residing in Yorkshire and the North of England [25]. Details on the recruitment, follow-up and phenotype/genotype data processing have been published previously [25, 26, 27]. In brief, for two periods (September 2000–December 2001, and July 2003 to December 2005) recruitment was restricted to patients with a primary tumour thickness of > 0.75 mm, while all patients with invasive melanoma were invited to participate between January 2002 and June 2003, and between January 2006–2012. Melanoma survival information was collected by direct communication with patients and their families, clinical records and from national registers.

Melanoma diagnoses were clinico-histopathologically confirmed through data linkage with the Northern and Yorkshire Cancer Registry and Information Service. Samples were genotyped using the Infinium HumanOmniExpressExome array (Illumina San Diego, CA, USA). After genetic quality control procedures (filtering for missingness, relatedness, and population outliers), this present study was restricted to 1,947 participants with genetic and phenotype data, and consent. Ethical approval for research involving the LMC was obtained from the Northern and Yorkshire Research Ethics Committee, and all participants provided written informed consent.

SNPs with MAF < 0.03, control Hardy–Weinberg equilibrium (HWE) P < 10–4 or missingness > 0.03 were excluded, as were any individuals with call rates < 0.97, identified as first degree relatives and/or European outliers by principal components analysis. Samples were imputed using the Haplotype Reference Consortium panel at the University of Michigan imputation server [22] and variants with an imputation quality score < 0.5 or MAF < 0.0001 were discarded.

Statistical analysis: genome-wide association study of melanoma-specific survival

First, we conducted two GWAS of MSS in the MIA cohort (5762 patients with melanoma and 800 melanoma-specific deaths) and in UKB cohort (5220 melanoma patients and 241 melanoma-specific deaths). Using Cox proportional-hazard modelling, hazard ratios (HRs) were computed using PLINK 1.9 [28] and the R survival package [29]. In both the MIA and UKB analyses, we adjusted for age, sex and the first ten PCs; in the MIA cohort we also adjusted for genotyping batch. We defined survival time as the duration between the date of diagnosis of the (first) primary CM and the date of death due to melanoma, measured in years. Other participants were censored on the last day of follow-up or when they died of any other cause. Analysis was restricted to participants of European ancestry and SNPs with MAF > 0.5%, and an imputation quality score > 0.5.

Next, we conducted a meta-analysis for both GWAS (N = 10,982 and 1,041 melanoma deaths) using a fixed-effects inverse-variance weighted model in METAL [30]. In addition, measures of heterogeneity (such as I2) were computed. Lead genome-wide significant (P < 5 × 10–8) SNPs independent at linkage disequilibrium (LD) r2 < 0.1 were identified using FUMA v1.3.6a (https://fuma.ctglab.nl/) [31].

Lead SNPs were tested for replication in the LMC (N = 1,947 patients with melanoma and 370 melanomas-specific deaths). The replication p-value threshold was set to 0.05. Next, we conducted a fixed- and random- effects inverse-variance meta-analysis of the lead SNPs from all three sets (MIA, UKB and LMC) using METAL [30]. For the two lead SNPs the nearest gene, and any significant expression quantitative trait loci (eQTLs) were identified using FUMA v1.3.6a [31].

Cutaneous melanoma polygenic risk score

Cutaneous melanoma risk discovery cohorts and GWAS meta-analysis

As the three MSS GWAS cohorts contributed to the discovery CM-susceptibility GWAS meta-analysis [10], and overlap between datasets used to generate, optimise or test PRS can lead to overfitting and other biases [32], we re-analysed the CM-susceptibility GWAS meta-analysis excluding the three MSS GWAS datasets. We further excluded the QSkin Sun and Health Study cohort to use as an independent data set to validate the generated PRSs. Details on recruitment, case definitions, genotyping, quality control, imputation approaches and ethical approvals for each cohort have been extensively described before [10]. The updated meta-analysis consisted of 23,913 cases, and 342,870 controls of European ancestry from Europe, Australia and the United States of America (USA) (Additional file 2: Table 1).

Table 1 Characteristics of Melanoma Institute Australia, UK Biobank and Leeds Melanoma Cohorts
Full size table

With the exception of the self-reported 23andMe, Inc. dataset, all CM cases were histopathologically confirmed; previous work has shown that 23andMe cases are very similar to the confirmed cases: the susceptibility loci show very similar effects in the self-reported and confirmed CM cases [10]. Each study was approved by the human research ethics committee at their respective institution, and all participants provided written informed consent. Specifically, for 23andMe, participants provided written informed consent and participated in the research online, under a protocol approved by the external AAHRPP-accredited IRB, Ethical & Independent Review Services.

Only SNPs with an imputation quality score > 0.5 were included, and a fixed-effects inverse variance weighted meta-analysis of log odds ratios (ORs) was performed using PLINK 1.9 [28]. Next, we selected 6,342,711 non-ambiguous, autosomal, bi-allelic GWAS meta-analysis SNPs with a MAF > 1% that were present in the validation (QSkin) and target (MIA and UKB) cohorts, and in the LD reference panel.

CM PRS validation cohort: The QSkin Sun and Health Study cohort

The QSkin Sun and Health Study (QSkin) cohort is a population-based cohort comprising over 43,000 adult participants recruited from Queensland, Australia. Detailed information on participant recruitment, phenotype measurement, genotyping and quality control measures have been published elsewhere [10, 33]. In summary, 18,087 participants were genotyped using the Global Screening Array [Illumina, San Diego, USA], and individuals were removed if they had non-European ancestry (6 s.d from the mean of PC1 and PC2 of 1000 Genomes European samples), were related to another participant (one from each pair removed if identity by descent pihat value > 0.1875), or had high genotype missingness (> 3%). SNPs were also removed due to HWE violations (P < 1 × 10–6), a low GenTrain score (< 0.6), or a low MAF (< 0.01). Cleaned genotype data were imputed to the haplotype reference consortium (v1) panel using the University of Michigan imputation server [22].

The Human Research Ethics Committee of QIMR Berghofer Medical Research Institute, Brisbane, Australia approved the study protocol and all participants provided written informed consent. We selected 16,708 participants (1,285 histopathologically confirmed CM cases and 15,423 controls) of European ancestry. CM cases were ascertained through data linkage with the Queensland Cancer Registry as well as assessing histopathology reports from pathology laboratories in Queensland.

Generation of the cutaneous melanoma polygenic risk score models

We used the CM-susceptibility GWAS data (generated above) and an LD reference panel of 2,000 unrelated individuals of European ancestry from UKB, to generate 30 PRS_susceptibility models at 1 megabase (Mb), 2 Mb, 3 Mb, 4 Mb and 5 Mb of LD radii each with varying fractions of causal SNPs i.e. 1 (F0), 0.1 (F1), 0.01 (F2), 0.001 (F3), 0.0001 (F4), and 0.00002 (F5). For this analysis we used LDpred, a Bayesian method that utilises all SNPs in the discovery GWAS (here CM-susceptibility GWAS), and their LD information, to derive LD-adjusted effect estimates (log ORs) for the trait (here CM-susceptibility) [34].

Validation of the cutaneous melanoma polygenic risk score in QSkin cohort

Next, we used the QSkin validation cohort to select the optimally performing PRS. Next, for each model we computed scores for 16,708 individuals (1,285 melanoma cases and 15,423 controls) in the QSkin Cohort using the LDpred-adjusted effect sizes (log ORs) and the imputed allelic dosages using PLINK 1.9 [28]. Then we computed and used Nagelkerke’s R2 [35] to select the optimally performing PRS_susceptibility model by comparing the model fit for CM risk ~ PRS_susceptibility + age + sex + 10 PCs, and a null model (CM risk ~ age + sex + 10 PCs) using the PredictABEL R package [36]. Model performances are presented in Fig. 1, and the best performing PRS model was used in the subsequent analyses.

Fig. 1
figure 1

Cutaneous melanoma polygenic risk score model performance in the validation cohort (QSkin). The x-axis represents the different melanoma polygenic risk score (PRS) modelling varying fractions of causal SNPs, 1 (F0), 0.1 (F1), 0.01 (F2), 0.001 (F3), 0.0001 (F4) and 0.00002 (F5), and differing linkage disequilibrium (LD) radii, 1 megabase (Mb), 2 Mb, 3 Mb, 4 Mb and 5 Mb. The y-axis represents Nagelkerke's R2 (%) for each of the 30 PRS models. The horizontal dashed black line highlights the optimal model (F3 and 5 Mb) with the highest Nagelkerke’s R2 of 7.03%

Full size image

Testing for association between cutaneous melanoma polygenic risk scores and melanoma-specific survival

The best performing PRS_susceptibility model was applied to the MIA and UKB cohorts using imputed allelic dosages and PLINK 1.9. The PRS was normalised to have a mean of 0 and an SD of 1 and tested for association with MSS in a Cox proportional hazard model adjusted for age at diagnosis, sex and the first ten PCs using the survival package in R [29]. We further calculated the MSS HR and 95% confidence interval (CI) per SD increase in the PRS_susceptibility. Next, we conducted a fixed- and random- effects inverse-variance meta-analysis to compute the pooled HR and 95% CI using the meta R package [37]. We then tested for association between MSS and the same PRS_susceptibility in the LMC, adjusting for the same covariates. Finally, we meta-analysed the MIA, UKB and LMC results.

Sensitivity analyses for polygenic susceptibility to melanoma and melanoma-specific survival

Pigmentation and naevus count loci are major biological pathways for CM-susceptibility [10, 38]. We further explored whether any association between the PRS_susceptibility and MSS was driven by SNPs associated with pigmentation and/or naevi pathways (Additional file 1). In addition, we generated PRSs for pigmentation (PRSP), naevus count (PRSN) and telomere length (PRSTL) and tested whether they were associated with MSS (Additional file 1). To rule out the possibility of thin or slow-growing melanomas influencing the PRS-survival association, we explored the potential influence of tumour stage, thickness and lead-time bias on any associations (Additional file 1). To test, and adjust, for the presence of an index-event bias we used the method of Dudbridge et al., 2019 [39] as implemented in the indexevent R package (https://rdrr.io/github/DudbridgeLab/indexevent/man/indexevent.html). The MIA survival GWAS, CM risk GWAS meta-analysis, and LD panel (5000 random individuals from the UK Biobank) were filtered to a common set of SNPs, and then pruned to LD independence (LD r2 < 0.01) in plink v1.9b6.26, leaving 102,286 SNPs. As per Dudbridge et al., 2019, the log(HR) effect sizes from the MIA and UKB survival GWAS were regressed onto the log(OR) effect sizes from the CM risk GWAS meta-analysis, and SCIMEX (1000 simulations) was used to adjust the regression slope for regression dilution to calculate the correction factor, and its variance, which was used to adjust the survival GWAS results. The impact of index-bias on the association between polygenic risk for melanoma, and survival was estimated as done in Howe et al. [40]. Briefly, using the LD pruned set used to calculate the correction factor above, 57 SNPs with a CM risk P < 5 x 10-8 and independent (LD r2 <  0.01) were selected, and the inverse-variance weighted method (implemented in the R MendelianRandomization package [41]) was used to estimate the association between polygenic risk for melanoma and survival before and after the correction.

Results

Baseline characteristics of the melanoma survival cohorts

This analysis was restricted to 5,762 melanoma patients in the MIA cohort, 5,220 in the UKB cohort, and 1,947 in the LMC. Summary data on mean age at diagnosis, sex, duration of follow up and the number of melanoma-specific deaths are presented in Table 1.

Genome-wide significant genetic variants for melanoma-specific survival

A MSS GWAS meta-analysis of the MIA and UKB cohorts identified two independent genome-wide significant (P < 5 × 10–8) loci (Table 2, and Additional file 1: Figure S1); rs41309643 (P = 2.08 × 10–8) on chromosome 1 (1q42.13) and rs75682113 (P = 1.07 × 10–8) on chromosome 7 (7p14.1) (Table 2). However, neither SNP was replicated in the LMC (rs41309643 P = 0.679 and rs75682113 P = 0.411 (Table 2, and Additional file 2: Table S2). Following the meta-analysis of all three cohorts, rs41309643 was no longer formally significant at P < 5 × 10–8 (HR = 1.83, 95% CI = 1.45–2.30, P = 3.21 × 10–7) with high heterogeneity metrics (Table 2). rs75682113 remained genome-wide significant with no significant evidence of heterogeneity (C-allele HR = 2.23, 95% CI = 1.68–2.95, P = 2.13 × 10–8; Table 2).

Table 2 Genetic variants for melanoma-specific survival in the discovery cohorts (MIA + UKB) and replication cohort (LMC)
Full size table

Rs41309643 on chromosome 1 is an intron of the PSEN2 gene and is associated with the expression of the Coenzyme Q8A (COQ8A) (formerly ADCK3) gene in blood. COQ8A is induced by p53 in response to DNA damage and inhibition of COQ8A counteracts p53-induced apoptosis [42]. rs75682113 on chromosome 7 is in an intron of the Succinyl-Coa:Glutarate-Coa Transferase (SUGCT) gene. This SNP has not been reported as an eQTL for any genes. Independent variants in the SUGCT gene have been associated with glutaric aciduria type 3 disease susceptibility [43].

The optimal cutaneous melanoma susceptibility polygenic risk score model

Of the thirty PRS tested, the model with the F3 causal fraction (0.001) and a 5 Mb LD radius performed best, with a Nagelkerke’s R2 of 7.02% (Fig. 1), and was used in all subsequent analyses.

Association of polygenic susceptibility to melanoma and melanoma-specific survival

After adjusting for age at diagnosis, sex and the first ten PCs, a one SD increase in the PRS_susceptibility was associated with improved MSS in a fixed-effects meta-analysis of MIA and UKB cohorts (HR = 0.88, 95% CI = 0.83–0.94, P = 6.93 × 10–5). However, the association between the PRS_susceptibility and MSS was highly heterogeneous across the two studies (I2 = 87.7%, 95% CI = 52.4–96.8%), with the finding mainly driven by the high UV setting cohort (MIA HR = 0.84, 95% CI = 0.78–0.90). Although not statistically significant, the magnitude and direction for the random effects model was also consistent with the fixed-effects results (fixed effects model HR = 0.92, 95% CI = 0.75–1.13, P = 0.43). The inverse association between polygenic susceptibility to melanoma and MSS persisted after excluding genomic regions associated with naevus count (fixed-effects HR = 0.91, 95% CI = 0.86–0.97, P = 0.0038; random-effects HR = 0.93, 95% CI = 0.83–1.03, P = 0.16) and pigmentation (fixed-effects HR = 0.91, 95% CI = 0.850–0.97, P = 0.0023; random-effects HR = 0.93 95% CI = 0.82–1.06, P = 0.26). The association between polygenic risk for melanoma and MSS was not replicated (P > 0.05) in the LMC; however, the directions of the effect estimates were consistent (Fig. 2).

Fig. 2
figure 2

Association of polygenic risk for cutaneous melanoma and melanoma-specific survival. All models were adjusted for age, sex and the first 10 PCs and additionally genotype batch effects in the MIA analysis. HRs were estimated using Cox proportional-hazards models. The full model refers to the PRS_susceptibility (CM PRS), while for the remaining two models the PRS_susceptibility respectively excluded SNPs in the naevus count and pigmentation pathways. MIA- Melanoma Institute Australia, UKB—United Kingdom Biobank, LMC—Leeds Melanoma Cohort, IVW- Inverse variance weighted methods, Het- heterogeneity, HR- hazard ratio. CI- confidence interval

Full size image

In a meta-analysis of the three cohorts a one SD increase in the PRS_susceptibility was still associated with improved MSS (fixed-effects HR = 0.90, 95% CI = 0.85–0.95, P = 6.35 × 10–5; random-effects HR = 0.93, 95% CI = 0.83–1.04, P = 0.20), even after excluding naevus and pigmentation loci (Fig. 2). There was substantial heterogeneity across the three studies (I2 = 78.7%, 95% CI = 31.6–93.4%). Sensitivity analyses showed that the skin colour PRS was also associated with improved MSS (PRSP; HR = 0.90, 95% CI = 0.85–0.96, P = 1.1 × 10–3), while the naevus count PRS also provided suggestive evidence (PRSN; HR = 0.95, 95%CI = 0.89–1.02, P = 0.179) (Additional file 1: Figure S5).

Influence from melanoma prognostic factors, lead-time, and index-event biases in the MIA Cohort

In the MIA cohort the PRS_susceptibility remained associated with improved survival after excluding participants with melanoma in-situ, and those with an unknown stage (HR = 0.84, 95% CI = 0.78–0.90, P = 2.15 × 10–6). In addition, the association was consistent even after adjusting for age, sex, 10 PCs, AJCC 2010 Stage, and primary tumour thickness (HR = 0.84, 95% CI = 0.78—0.91, P = 1.90 × 10–5) (Table 3). There was also no evidence for interaction by the tumour stage or tumour thickness (Table 3). In a stratified analysis, there was no evidence that the association between the PRS and MSS differed by tumour stage (Fig. 3a) and primary tumour thickness at diagnosis (Fig. 3b). The PRSTL was suggestive but not significantly associated with MSS in the MIA cohort (PRSTL; HR = 0.90, 95% CI = 0.64–1.27, P = 0.5504). After excluding the first two years of follow-up (following diagnosis), there was no evidence of lead-time bias (survival bias) (HR = 0.84, 95% CI = 0.77–0.91, P = 4.03 × 10–5). We applied the index-event bias proposed in Dudbridge et al. [39] to the MIA survival GWAS results; the results for both individual SNPs, and estimates of the association of polygenic risk and survival, were essentially unchanged before and after the correction [data not shown].

Table 3 Testing for an interaction between the polygenic susceptibility to melanoma and survival prognostic factors in the MIA Cohort
Full size table
Fig. 3
figure 3

Stratified analysis of the PRS and MSS association by the AAJC Stage and primary tumour thickness in the MIA Cohort. The y-axis represents the hazard ratio for MSS per standard deviation (SD) increase in the PRS. Error bars show the 95% confidence interval of the HR. The x-axis shows the strata for tumour stage and thickness at diagnosis of melanoma. The dashed grey line represents a null effect at a hazard ratio of 1. Panel 3a shows the association between the CM PRS and MSS stratified by the AJCC 2010 tumour stage, after controlling for age at diagnosis, sex, the first 10 ancestral components and genotype batch effects. Stage I/II included 4493 participants and 427 melanoma deaths, while stage III/IV included 789 participants and 242 melanoma deaths. Panel 3b reports the association of the CM PRS and MSS stratified by the primary tumour thickness, after controlling for age at diagnosis, sex, the first 10 ancestral components and genotype batch effects. The thin (< = 1 mm) stratum included 1,898 participants and 122 melanoma deaths, while the thick (> 1 mm) stratum included 3,384 participants and 547 melanoma deaths

Full size image

Discussion

In this study, we performed the largest GWAS for MSS to date using data from Australia and the United Kingdom and potentially have identified two independent, novel, genome-wide significant (P < 5 × 10–8) loci for MSS at 1q42.13 and 7p14.1. While the two loci did not formally replicate in an independent cohort, the confidence intervals (particularly for rs75682113) in the replication set overlap the estimate from the discovery cohorts. Confirmation of these two loci will require replication in larger cohorts. rs75682113 is particularly promising as it was genome-wide significant (P < 5 × 10–8) in our meta-analysis of the discovery and replication samples.

In addition, we report evidence that increased genetic susceptibility for CM, as measured by a one SD increase in a PRS_susceptibility, was significantly associated with improved MSS. However, caution is required as the result was primarily driven by a strong association in the MIA cohort. Genetic susceptibility to CM is primarily driven by loci in the pigmentation and naevus count pathways [44]. HRs for PRS_susceptibility and MSS were slightly attenuated (but still with a significant association) when we removed SNPs in either pathway. In turn PRS designed specifically for these traits were also associated (though not significantly for naevus count) with MSS. In addition, the PRS for telomere length (another pathway to both CM susceptibility and survival) was not significantly associated with MSS in our sensitivity analysis. These pathway-analysis results suggest that if genetic propensity to CM is associated with improved survival it is not simply due to pigmentation, nevus count or telomere length.

However, this study suggests that if there is a true association, its magnitude may differ across populations, presumably due to environmental factors such as high UV (e.g. in Australia) and other effects. Firstly, the MIA and UKB meta-analysis results did not replicate in the LMC. Secondly, the high heterogeneity metrics (e.g. I2) indicate that the effect sizes may not be consistent across the three studies, with a very strong result in the MIA cohort (a high UV setting) and weaker associations in the UK samples (Table 1 and Additional file 2: Table S2). Although the fixed-effects model shows a strong statistically significant association, the results are not significant for the random-effects model even when they are of a similar magnitude. The observed heterogeneity may be due to differences in recruitment, where the MIA cohort recruitment was from clinics as opposed to the population-based UKB and LMC. It is also possible that the strong inverse result in Australia is influenced by overdiagnosis for melanoma [45]. It is estimated that 54% of all melanomas and 15% invasive melanomas in Australia are over-diagnosed [46]. Thus, patients may be diagnosed with non-lethal melanoma and subsequently exhibit improved survival. However, recent evidence suggests that regular skin checks (which may lead to overdiagnosis for melanoma) are not associated with MSS [47]. Since sun exposure or high UV exposure is associated with improved MSS [48, 49], it is also possible that differences in high or long-term sun and ultraviolet-radiation exposure in Australia are in part responsible for the heterogeneity.

Analyses of outcomes such as survival necessitate the inclusion of people based on having the index disease, which can introduce an index event bias [39]. This can potentially lead to spurious associations between disease risk factors (such as SNPs associated with risk of disease) with survival. The application of a recent method to identify and adjust for this bias in our data did not meaningfully change the results, similar to what has been observed for previous studies [40].

In a more detailed analysis in the MIA cohort, our study suggests that this inverse association is consistent even after further adjusting for (and testing for interaction with) strong predictors of MSS like tumour stage and primary tumour thickness at diagnosis. The stratified analysis shows that the association is not modified by primary tumour thickness or stage. Thus, if replicated in additional cohorts, a CM-susceptibility PRS is potentially an independent prognostic factor for MSS.

To our knowledge, while no prior study has examined the association of a CM susceptibility PRS and survival outcome, similar inverse relationships have been found in other cancers e.g. higher breast cancer PRSs and better breast cancer prognosis/characteristics [50, 51]. Also, a follicular lymphoma PRS was associated with improved overall survival among women in a population in the USA [52]. BRCA1/2 mutations which increase breast cancer risk were associated with better overall survival among triple-negative breast cancer women [53]. A CAD PRS was inversely associated with all-cause mortality (OR = 0.91; 95% CI = 0.85–0.98), and ischaemic stroke (OR = 0.78; 95% CI = 0.67–0.90) in CAD patients [40].

The mechanisms underlying this inverse association are unclear. Particularly for MSS, it could be that a higher genetic risk for CM leads to thin melanomas or slow-growing melanomas that are less lethal [54, 55, 56], and respond better to treatment. However, detailed analysis in the MIA cohort showed no difference in survival for both thin and thick tumour categories. In addition, after excluding the initial two years of follow up, the results were consistent, suggesting there is no survival/ lead-time bias.

As noted in our study above, higher nevus counts may be associated with a lower chance of dying from melanoma [14]. It is possible however that those with large numbers of naevi are subjected to increased screening, which may lead to overdiagnosis and greater survival relative to those with fewer moles [57]. However, as already indicated, increased screening is not associated with MSS [47].

Another possible mechanism could be via gene-environment interaction, where those at highest genetic risk of CM benefit more from treatment (e.g. immunotherapy), as it is the case for those at high genetic risk for coronary artery disease (CAD) and treatment benefits from PCSK9 inhibitors in the FOURIER and ODYSSEY OUTCOMES trials [58, 59].

This study presents new insights that highlight the potential clinical utility of PRS_susceptibility for profiling and monitoring patients for melanoma outcomes following diagnosis during the “melanoma follow-up care program” [60, 61]. In combination with other prognostic factors, it could be used to guide patient care e.g. counselling on modification of mortality-related non-genetic behaviours and lifestyle factors, or guide the direction of patient-specific treatment to help improve survival after diagnosis. It may also be useful for the stratification of patients while recruiting into clinical trials evaluating melanoma treatment and outcomes.

Conclusions

In a GWAS meta-analysis of MSS, we identified two novel loci potentially associated with survival from cutaneous melanoma, both of which contain candidate genes linked to tumour progression; however, replication in large independent cohorts is required. In line with observations in other cancers and complex diseases, increased germline genetic susceptibility for CM was strongly but heterogeneously associated with improved MSS especially in a high UV setting. If validated, a PRS_susceptibility could be used to predict melanoma outcomes after diagnosis and profile patients for personalised care.

Availability of data and materials

CM GWAS summary statistics used to generate the LDPred PRSs can be accessed as indicated by Landi et al. 2020. Underlying data for the cohorts used in the paper are available through application to the respective cohorts; UKB (http://www.ukbiobank.ac.uk/wp-content/uploads/2012/09/Access-Procedures-2011-1.pdf); MIA (https://www.melanoma.org.au/research/collaborate-on-research-with-mia/); Q-Skin (By application to Q-Skin Principal Investigator David Whiteman David.Whiteman@qimrberghofer.edu.au).

Abbreviations

AJCC:

American Joint Committee on Cancer

CM:

Cutaneous melanoma

GWAS:

Genome-wide association studies

HR:

Hazard ratio

ICD:

International Classification of Diseases

Mb:

Megabase

LD:

Linkage disequilibrium

MIA:

Melanoma Institute Australia

MSS:

Melanoma-specific survival

MAF:

Minor allele frequency

PRS:

Polygenic risk score

PCs:

Principal components

QSkin:

QSkin Sun and Health Study

SNP:

Single nucleotide polymorphism

SD:

Standard deviation

UKB:

UK Biobank

R2 :

Variance

LMC:

Leeds Melanoma Cohort

CI:

Confidence interval (CI)

References

  1. Cancer Australia. Melanoma of the skin statistics. Cancer Australia. 2019 [cited 2020 Mar 18]. https://melanoma.canceraustralia.gov.au/statistics

  2. NCI. Cancer Stat Facts: Melanoma of the Skin. National Cancer Institute. 2021 [cited 2022 Jan 20]. https://seer.cancer.gov/statfacts/html/melan.html

  3. Elliott TM, Whiteman DC, Olsen CM, Gordon LG. Estimated healthcare costs of melanoma in Australia over 3 years post-diagnosis. Appl Health Econ Health Policy. 2017;15:805–16.

    Article  PubMed  Google Scholar 

  4. Zaorsky NG, Khunsriraksakul C, Acri SL, Liu DJ, Ba DM, Lin JL, et al. Medical service use and charges for cancer care in 2018 for privately insured patients younger than 65 years in the US. JAMA Netw Open. 2021;4: e2127784.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Gandini S, Sera F, Cattaruzza MS, Pasquini P, Abeni D, Boyle P, et al. Meta-analysis of risk factors for cutaneous melanoma: I. Common and atypical naevi. Eur J Cancer. 2005;41:28–44.

    Article  PubMed  Google Scholar 

  6. Gandini S, Sera F, Cattaruzza MS, Pasquini P, Zanetti R, Masini C, et al. Meta-analysis of risk factors for cutaneous melanoma: III. Family history, actinic damage and phenotypic factors. Eur J Cancer. 2005;41:2040–59.

    Article  PubMed  Google Scholar 

  7. Mitra D, Luo X, Morgan A, Wang J, Hoang MP, Lo J, et al. An ultraviolet-radiation-independent pathway to melanoma carcinogenesis in the red hair/fair skin background. Nature. 2012;491:449–53.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Chang Y-M, Barrett JH, Bishop DT, Armstrong BK, Bataille V, Bergman W, et al. Sun exposure and melanoma risk at different latitudes: a pooled analysis of 5700 cases and 7216 controls. Int J Epidemiol. 2009;38:814–30.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Veierød MB, Adami H-O, Lund E, Armstrong BK, Weiderpass E. Sun and solarium exposure and melanoma risk: effects of age, pigmentary characteristics, and nevi. Cancer Epidemiol Biomarkers Prev. 2010;19:111–20.

    Article  PubMed  CAS  Google Scholar 

  10. Landi MT, Bishop DT, MacGregor S, Machiela MJ, Stratigos AJ, Ghiorzo P, et al. Genome-wide association meta-analyses combining multiple risk phenotypes provide insights into the genetic architecture of cutaneous melanoma susceptibility. Nat Genet. 2020;52:494–504.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Cherobin ACFP, Wainstein AJA, Colosimo EA, Goulart EMA, Bittencourt FV. Prognostic factors for metastasis in cutaneous melanoma. An Bras Dermatol. 2018;93:19–26.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Kibrité A, Milot H, Douville P, Gagné ÉJ, Labonté S, Friede J, et al. Predictive factors for sentinel lymph nodes and non-sentinel lymph nodes metastatic involvement: a database study of 1,041 melanoma patients. Am J Surg. 2016;211:89–94.

    Article  PubMed  Google Scholar 

  13. Brandt A, Sundquist J, Hemminki K. Risk of incident and fatal melanoma in individuals with a family history of incident or fatal melanoma or any cancer. Br J Dermatol. 2011;165:342–8.

    Article  CAS  PubMed  Google Scholar 

  14. Ribero S, Davies JR, Requena C, Carrera C, Glass D, Rull R, et al. High nevus counts confer a favorable prognosis in melanoma patients. Int J Cancer. 2015;137:1691–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Zhu G, Duffy DL, Eldridge A, Grace M, Mayne C, O’Gorman L, et al. A major quantitative-trait locus for mole density is linked to the familial melanoma gene CDKN2A: a maximum-likelihood combined linkage and association analysis in twins and their sibs. Am J Hum Genet. 1999;65:483–92.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Wachsmuth RC, Gaut RM, Barrett JH, Saunders CL, Randerson-Moor JA, Eldridge A, et al. Heritability and gene-environment interactions for melanocytic nevus density examined in a U.K. Adolescent twin study. J Invest Dermatol. 2001. https://doi.org/10.1046/j.0022-202X.2001.01415.x.

    Article  PubMed  Google Scholar 

  17. Olsen CM, Carroll HJ, Whiteman DC. Estimating the attributable fraction for cancer: a meta-analysis of nevi and melanoma. Cancer Prev Res. 2010;3:233–45.

    Article  Google Scholar 

  18. Burke LS, Hyland PL, Pfeiffer RM, Prescott J, Wheeler W, Mirabello L, et al. Telomere length and the risk of cutaneous malignant melanoma in melanoma-prone families with and without CDKN2A mutations. PLoS ONE. 2013;8: e71121.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Rachakonda S, Srinivas N, Mahmoudpour SH, Garcia-Casado Z, Requena C, Traves V, et al. Telomere length and survival in primary cutaneous melanoma patients. Sci Rep. 2018;8:10947.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  20. Meisner A, Kundu P, Zhang YD, Lan LV, Kim S, Ghandwani D, et al. Combined utility of 25 disease and risk factor polygenic risk scores for stratifying risk of all-cause mortality. Am J Hum Genet. 2020;107:418–31.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Liyanage U, MacGregor S, Bishop DT, Shi J, An J, Ong JS, et al. Multi-trait genetic analysis identifies auto-immune loci associated with cutaneous melanoma. J Invest Dermatol. 2021. https://doi.org/10.1016/j.jid.2021.08.449.

    Article  PubMed  Google Scholar 

  22. Loh P-R, Danecek P, Palamara PF, Fuchsberger C, A Reshef Y, K Finucane H, et al. Reference-based phasing using the Haplotype Reference Consortium panel. Nat Genet. 2016;48:1443–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, 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 Med. 2015;12: e1001779.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562:203–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Newton-Bishop JA, Davies JR, Latheef F, Randerson-Moor J, Chan M, Gascoyne J, et al. 25-Hydroxyvitamin D2 /D3 levels and factors associated with systemic inflammation and melanoma survival in the Leeds Melanoma Cohort. Int J Cancer. 2015;136:2890–9.

    Article  CAS  PubMed  Google Scholar 

  26. Newton-Bishop JA, Chang Y-M, Iles MM, Taylor JC, Bakker B, Chan M, et al. Melanocytic nevi, nevus genes, and melanoma risk in a large case-control study in the United Kingdom. Cancer Epidemiol Biomark Prevent. 2010. https://doi.org/10.1158/1055-9965.epi-10-0233.

    Article  Google Scholar 

  27. Bishop DT, Demenais F, Iles MM, Harland M, Taylor JC, Corda E, et al. Genome-wide association study identifies three loci associated with melanoma risk. Nat Genet. 2009;41:920–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4:7.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  29. Therneau T. A package for survival analysis in r. r package version 3.2–3 [Internet]. 2020 [cited 2020 Dec 1]. https://CRAN.R-project.org/package=survival

  30. Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics. 2010;26:2190–1.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Watanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional mapping and annotation of genetic associations with FUMA. Nat Commun. 2017;8:1826.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  32. Lambert SA, Abraham G, Inouye M. Towards clinical utility of polygenic risk scores. Hum Mol Genet. 2019;28:R133–42.

    Article  CAS  PubMed  Google Scholar 

  33. Olsen CM, Green AC, Neale RE, Webb PM, Cicero RA, Jackman LM, et al. Cohort profile: the QSkin Sun and Health Study. Int J Epidemiol. 2012;41:929–929i.

    Article  PubMed  Google Scholar 

  34. Vilhjálmsson BJ, Yang J, Finucane HK, Gusev A, Lindström S, Ripke S, et al. Modeling linkage disequilibrium increases accuracy of polygenic risk scores. Am J Hum Genet. 2015;97:576–92.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  35. Nagelkerke NJD. A note on a general definition of the coefficient of determination. Biometrika. 1991;78:691–2.

    Article  Google Scholar 

  36. Kundu S, Aulchenko YS, van Duijn CM, A Cecile J. PredictABEL: an R package for the assessment of risk prediction models. Eur J Epidemiol. 2011;26:261–4.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Balduzzi S, Rücker G, Schwarzer G. How to perform a meta-analysis with R: a practical tutorial. Evid Based Mental Health. 2019. https://doi.org/10.1136/ebmental-2019-300117.

    Article  Google Scholar 

  38. Duffy DL, Zhu G, Li X, Sanna M, Iles MM, Jacobs LC, et al. Novel pleiotropic risk loci for melanoma and nevus density implicate multiple biological pathways. Nat Commun. 2018;9:4774.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  39. Dudbridge F, Allen RJ, Sheehan NA, Schmidt AF, Lee JC, Jenkins RG, Wain LV, Hingorani AD, Patel RS (2019) Adjustment for index event bias in genome-wide association studies of subsequent events. Nat Commun 10(1):1561. https://doi.org/10.1038/s41467-019-09381-w

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Howe LJ, Dudbridge F, Schmidt AF, Finan C, Denaxas S, Asselbergs FW, et al. Polygenic risk scores for coronary artery disease and subsequent event risk amongst established cases. Hum Mol Genet. 2020;29:1388–95.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Broadbent JR, Foley CN, Grant AJ, Mason AM, Staley JR, Burgess S (2020) MendelianRandomization v0.5.0: updates to an R package for performing Mendelian randomization analyses using summarized data. Wellcome Open Res 5:252. https://doi.org/10.12688/wellcomeopenres.16374.2

    Article  PubMed  PubMed Central  Google Scholar 

  42. Iiizumi M, Arakawa H, Mori T, Ando A, Nakamura Y. Isolation of a novel gene, CABC1, encoding a mitochondrial protein that is highly homologous to yeast activity of bc1 complex. Cancer Res. 2002;62:1246–50.

    CAS  PubMed  Google Scholar 

  43. Sherman EA, Strauss KA, Tortorelli S, Bennett MJ, Knerr I, Morton DH, et al. Genetic mapping of glutaric aciduria, type 3, to chromosome 7 and identification of mutations in c7orf10. Am J Hum Genet. 2008;83:604–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Cust AE, Drummond M, Kanetsky PA, Australian Melanoma Family Study Investigators, Leeds Case-Control Study Investigators, Goldstein AM, et al. Assessing the Incremental Contribution of Common Genomic Variants to Melanoma Risk Prediction in Two Population-Based Studies. J Invest Dermatol. 2018;138:2617–24.

  45. Welch HG, Mazer BL, Adamson AS. The rapid rise in cutaneous melanoma diagnoses. N Engl J Med. 2021;384:72–9.

    Article  PubMed  Google Scholar 

  46. Glasziou PP, Bell KJ, Barratt AL. Estimating the magnitude of cancer overdiagnosis in Australia. Med J Aust. 2020;213(4):189.e1.

    Article  Google Scholar 

  47. Watts CG, McLoughlin K, Goumas C, van Kemenade CH, Aitken JF, Soyer HP, et al. Association between melanoma detected during routine skin checks and mortality. JAMA Dermatol. 2021. https://doi.org/10.1001/jamadermatol.2021.3884.

    Article  PubMed  PubMed Central  Google Scholar 

  48. Berwick M, Armstrong BK, Ben-Porat L, Fine J, Kricker A, Eberle C, et al. Sun exposure and mortality from melanoma. J Natl Cancer Inst. 2005;97:195–9.

    Article  PubMed  Google Scholar 

  49. Rosso S, Sera F, Segnan N, Zanetti R. Sun exposure prior to diagnosis is associated with improved survival in melanoma patients: results from a long-term follow-up study of Italian patients. Eur J Cancer. 2008;44:1275–81.

    Article  PubMed  Google Scholar 

  50. Holm J, Li J, Darabi H, Eklund M, Eriksson M, Humphreys K, et al. Associations of breast cancer risk prediction tools with tumor characteristics and metastasis. J Clin Oncol. 2016;34:251–8.

    Article  CAS  PubMed  Google Scholar 

  51. Li J, Ugalde-Morales E, Wen WX, Decker B, Eriksson M, Torstensson A, et al. Differential burden of rare and common variants on tumor characteristics, survival, and mode of detection in breast cancer. Cancer Res. 2018;78:6329–38.

    Article  CAS  PubMed  Google Scholar 

  52. Zhong C, Chao CR, Song JY, Weisenburger DD, Luo J, Ding YC, et al. Follicular lymphoma polygenic risk score is associated with increased disease risk but improved overall survival among women in a population based case-control in Los Angeles County California. Cancer Epidemiol. 2020. https://doi.org/10.1016/j.canep.2020.101688.

    Article  PubMed  PubMed Central  Google Scholar 

  53. Baretta Z, Mocellin S, Goldin E, Olopade OI, Huo D. Effect of BRCA germline mutations on breast cancer prognosis: a systematic review and meta-analysis. Medicine. 2016;95: e4975.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Adami H-O, Csermely P, Veres DV, Emilsson L, Løberg M, Bretthauer M, et al. Are rapidly growing cancers more lethal? Eur J Cancer. 2017;72:210–4.

    Article  PubMed  Google Scholar 

  55. Halpern AC, Marghoob AA. Thin melanoma: still “excellent prognosis” disease? J Clin Oncol. 2004;22(18):3651–3.

    Article  CAS  PubMed  Google Scholar 

  56. Maurichi A, Miceli R, Camerini T, Mariani L, Patuzzo R, Ruggeri R, et al. Prediction of survival in patients with thin melanoma: results from a multi-institution study. J Clin Oncol. 2014;32:2479–85.

    Article  PubMed  Google Scholar 

  57. Autier P, Funck-Brentano E, Aegerter P, Boniol M, Saiag P. Re: High nevus counts confer a favorable prognosis in melanoma patients by S ribero and co-workers, published in the International Journal of Cancer, 2015 (online 21 march 2015). Int J Cancer. 2015;137(12):3006–7.

    Article  CAS  PubMed  Google Scholar 

  58. Marston NA, Kamanu FK, Nordio F, Gurmu Y, Roselli C, Sever PS, et al. Predicting benefit from evolocumab therapy in patients with atherosclerotic disease using a genetic risk score: results from the FOURIER trial. Circulation. 2020;141:616–23.

    Article  PubMed  Google Scholar 

  59. Damask A, Steg PG, Schwartz GG, Szarek M, Hagström E, Badimon L, et al. Patients with high genome-wide polygenic risk scores for coronary artery disease may receive greater clinical benefit from alirocumab treatment in the ODYSSEY OUTCOMES trial. Circulation. 2020;141:624–36.

    Article  PubMed  Google Scholar 

  60. Trotter SC, Sroa N, Winkelmann RR, Olencki T, Bechtel M. A global review of melanoma follow-up guidelines. J Clin Aesthet Dermatol. 2013;6:18–26.

    PubMed  PubMed Central  Google Scholar 

  61. Farma JM, Abdulla A. Follow-up guidelines for resected melanoma. Melanoma. 2015. https://doi.org/10.1093/med/9780199971015.003.0008.

    Article  Google Scholar 

Download references

Acknowledgements

This study was conducted using data from UK Biobank (application number 25331), MIA (Australia), QSkin (Australia), Leeds Melanoma Cohort (UK), 23andMe Research (USA) and GWAS summary data from the melanoma meta-analysis consortium. We gratefully acknowledge Simone Cross, as well as Susan List Armitage and the Sample Processing Facility, at QIMR Berghofer Medical Research Institute for their assistance in genotyping MIA samples, and Maria Teresa Landi at the National Cancer Institute for the genotyping them. We would like to thank the research participants and employees of 23andMe for making this work possible. The authors thank and acknowledge the valuable contributions of Hazel Burke and Valerie Jakrot, and their colleagues, in the clinical, data management, and biospecimen banking teams at Melanoma Institute Australia.

Funding

QIMR Berghofer Medical Research Institute: The study was supported by a program grant (APP1073898) and a project grant (APP1063061) from the Australian National Health and Medical Research Council (NHMRC). SM and DCW are supported by Research Fellowships from the NHMRC. MS was supported by the Australian Government Research Training Program (RTP) and the Faculty of Health Scholarship at Queensland University of Technology, Australia. Melanoma Institute Australia: This study was supported by funding from Melanoma Institute Australia, the Australian National Health and Medical Research Council (NHMRC) through a program grant to GJM, RAS, JFT & GVL (APP1093017) and from Cancer Institute New South Wales and infrastructure grants from Macquarie University and the Australian Cancer Research Foundation. R.A.S. and G.V.L. are supported by NHMRC Fellowships (APP1141295 for R.A.S), and G.V.L. is supported by the University of Sydney Medical Foundation. Leeds Melanoma Cohort: The Leeds Melanoma Cohort was funded by the Cancer Research UK (under project grant C8216/A6129 and programme award C588/A19167), and by the National Institutes of Health (NIH) (R01 CA83115) and EU FP6 Network of Excellence award to GenoMEL. Participant recruitment was also supported by the UK National Cancer Research Network. DTB and MMI were supported in part by the Cancer Research UK awards.

Author information

Authors and Affiliations

  1. Statistical Genetics Lab, QIMR Berghofer Medical Research Institute, 300 Herston Road, Herston, QLD, 4006, Australia

    Mathias Seviiri, Stuart MacGregor & Matthew H. Law

  2. School of Biomedical Sciences, Faculty of Health, Queensland University of Technology, Brisbane, QLD, Australia

    Mathias Seviiri, Stuart MacGregor & Matthew H. Law

  3. Center for Genomics and Personalised Health, Queensland University of Technology, Brisbane, QLD, Australia

    Mathias Seviiri

  4. Melanoma Institute Australia, The University of Sydney, Sydney, NSW, Australia

    Richard A. Scolyer, Serigne N. Lo, Johnathan R. Stretch, Robyn P. M. Saw, Omgo E. Nieweg, Kerwin F. Shannon, Andrew J. Spillane, John F. Thompson & Georgina V. Long

  5. Faculty of Medicine and Health, The University of Sydney, Sydney, NSW, Australia

    Richard A. Scolyer, Serigne N. Lo, Johnathan R. Stretch, Robyn P. M. Saw, Omgo E. Nieweg, Andrew J. Spillane, John F. Thompson & Georgina V. Long

  6. Department of Tissue Oncology and Diagnostic Pathology, Royal Prince Alfred Hospital, Sydney, NSW, Australia

    Richard A. Scolyer

  7. NSW Health Pathology, Sydney, NSW, Australia

    Richard A. Scolyer

  8. Division of Haematology and Immunology, Leeds Institute of Medical Research at St James’, University of Leeds, Leeds, UK

    D. Timothy Bishop & Julia A. Newton-Bishop

  9. St James’s Institute of Medical Research, University of Leeds, Leeds, UK

    Mark M. Iles

  10. Leeds Institute of Data Analytics, University of Leeds, Leeds, UK

    Mark M. Iles

  11. Department of Melanoma and Surgical Oncology, Royal Prince Alfred Hospital, Camperdown, NSW, Australia

    Johnathan R. Stretch, Robyn P. M. Saw, Omgo E. Nieweg, Kerwin F. Shannon & John F. Thompson

  12. Sydney Head & Neck Cancer Institute, Chris O’Brien Lifehouse Cancer Center, Sydney, NSW, Australia

    Kerwin F. Shannon

  13. Department of Breast and Melanoma Surgery, Royal North Shore Hospital, Sydney, NSW, Australia

    Andrew J. Spillane

  14. Genetic Epidemiology Lab, QIMR Berghofer Medical Research Institute, Brisbane, QLD, Australia

    Scott D. Gordon

  15. Cancer Control Group, QIMR Berghofer Medical Research Institute, Brisbane, QLD, Australia

    Catherine M. Olsen & David C. Whiteman

  16. Faculty of Medicine, University of Queensland, Brisbane, QLD, Australia

    Catherine M. Olsen

  17. Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA

    Maria Teresa Landi

  18. Department of Medical Oncology, Mater Hospital, North Sydney, NSW, Australia

    Georgina V. Long

  19. Department of Medical Oncology, Royal North Shore Hospital, St Leonards, NSW, Australia

    Georgina V. Long

Authors
  1. Mathias Seviiri
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Richard A. Scolyer
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. D. Timothy Bishop
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Julia A. Newton-Bishop
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Mark M. Iles
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Serigne N. Lo
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Johnathan R. Stretch
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Robyn P. M. Saw
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Omgo E. Nieweg
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Kerwin F. Shannon
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Andrew J. Spillane
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Scott D. Gordon
    View author publications

    You can also search for this author in PubMed Google Scholar

  13. Catherine M. Olsen
    View author publications

    You can also search for this author in PubMed Google Scholar

  14. David C. Whiteman
    View author publications

    You can also search for this author in PubMed Google Scholar

  15. Maria Teresa Landi
    View author publications

    You can also search for this author in PubMed Google Scholar

  16. John F. Thompson
    View author publications

    You can also search for this author in PubMed Google Scholar

  17. Georgina V. Long
    View author publications

    You can also search for this author in PubMed Google Scholar

  18. Stuart MacGregor
    View author publications

    You can also search for this author in PubMed Google Scholar

  19. Matthew H. Law
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Conceptualization: MS, GVL, RAS, MHL, SM; Data collection, generation, and curation: MS, MHL, SDG, MTL, DCW, CMO, SM, SNL, JRS, RPMS, OEN, KFS, AJS, JFT, DTB, JAN-B; Genotyping and imputation: MTL, SM, SDG, MMI, DTB, MHL. Formal analysis: MS, MHL, DTB; Funding acquisition: SM, MHL, DCW, GVL, RAS, DTB, MTL; Investigation: MS, MHL, GVL, RAS, DCW, CMO, SM; Methodology: MS, RAS, GVL, MHL, SM; Project administration: MS, MHL, SM; Resources: MHL, GVL, RAS, MTL, DCW, SM; Software: MS; Supervision: SM, MHL; Visualisation: MS; Writing—original draft preparation: MS. Writing—subsequent drafts preparation: MS, SM, MHL, DCW, CMO, SDG, GVL, RAS, MTL, DCW, MMI, DTB, JAN-B, SNL, JRS, RPMS, OEN, KFS, AJS, JFT. All authors contributed to the final version of the manuscript.

Corresponding authors

Correspondence to Mathias Seviiri or Matthew H. Law.

Ethics declarations

Ethics approval and consent to participate

The Melanoma Institute Australia study protocols (protocol HREC/10/RPAH/530 and HREC/11/RPAH/444) were approved by the Sydney Local Health District Ethics Review Committee, Royal Prince Alfred Hospital, Camperdown, Australia. The UK Biobank study was approved by the United Kingdom’s National North West Multi-Centre Research Ethics Committee (ref. 11/NW/0382). Leeds Melanoma Cohort approved by Northern and Yorkshire Research Ethics Committee in the United Kingdom. The Human Research Ethics Committee of QIMR Berghofer Medical Research Institute, Brisbane, Australia approved the QSkin study protocol (P1309), and this study (P2121). In all the respective studies, all participants provided written informed consent to participate.

Consent for publication

Not applicable.

Competing interests

JFT has received honoraria for advisory board participation from BMS Australia, MSD Australia, GSK and Provectus Inc, and travel and conference support from GSK, Provectus Inc and Novartis. RAS has received fees for professional services from F. Hoffmann-La Roche Ltd, Evaxion, Provectus Biopharmaceuticals Australia, Qbiotics, Novartis, Merck Sharp & Dohme, NeraCare, AMGEN Inc., Bristol-Myers Squibb, Myriad Genetics, GlaxoSmithKline. The rest of the authors declare no conflict of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Figure 1

. Manhattan plot for the MSS GWAS meta-analysis between MIA and UKB cohorts. Figure 2. Skin colour polygenic risk score model performance in the validation cohort (QSkin). The x-axis represents the different melanoma polygenic risk score (PRS) models of varying fractions of causal SNPs (i.e. 1 (F0), 0.1 (F1), 0.01 (F2),0.001(F3),0.0001 (F4) and 0.00002 (F5)) at the different radii of the linkage disequilibrium (LD) (i.e. 2 megabase (mb) and 5 mb). The y-axis represents Nagelkerke’s R2 (%) for each of the 12 PRS models. The horizontal dashed black line highlights the optimal model (F1-2mb) (i.e. with the highest Nagelkerke’s R2). Figure 3. The association between polygenic risk scores and the risk of melanoma in QSkin. Figure 4. The association by quartile of polygenic risk for melanoma susceptibility and melanoma specific survival in the MIA cohort. Figure 5. Association of standalone skin colour and naevus PRSs and melanoma specific survival in MIA and UKB.

Additional file 2: Table 1

. Cohorts used for the cutaneous melanoma susceptibility discovery meta-analysis GWAS. Table 2. Per study melanoma specific survival  GWAS results. Table 3. Pigmentation loci excluded from the cutaneous melanoma susceptibility polygenic risk score. Table 4. Naevi loci excluded from the cutaneous melanoma susceptibility polygenic risk score. Table 5. Naevous count SNPs included in the naevus polygenic risk score.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Seviiri, M., Scolyer, R.A., Bishop, D.T. et al. Higher polygenic risk for melanoma is associated with improved survival in a high ultraviolet radiation setting. J Transl Med 20, 403 (2022). https://doi.org/10.1186/s12967-022-03613-2

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12967-022-03613-2

Keywords

Journal of Translational Medicine

ISSN: 1479-5876

Contact us