ARTICLE

https://doi.org/10.1038/s41467-020-20432-5 OPEN Ultraviolet radiation drives in a subset of mucosal

Piyushkumar A. Mundra 1,11, Nathalie Dhomen1,11, Manuel Rodrigues 2,3, Lauge Hjorth Mikkelsen 4, Nathalie Cassoux5, Kelly Brooks1,6, Sara Valpione1,7, Jorge S. Reis-Filho 8, Steffen Heegaard4, ✉ Marc-Henri Stern 2,9, Sergio Roman-Roman10 & Richard Marais 1

Although identified as the key environmental driver of common cutaneous , the

1234567890():,; role of ultraviolet radiation (UVR)-induced DNA damage in mucosal melanoma is poorly defined. We analyze 10 mucosal melanomas of conjunctival origin by whole genome sequencing and our data shows a predominance of UVR-associated single base substitution signature 7 (SBS7) in the majority of the samples. Our data shows mucosal melanomas with SBS7 dominance have similar genomic patterns to cutaneous melanomas and therefore this subset should not be excluded from treatments currently used for common cutaneous melanoma.

1 Molecular Group, Research UK Manchester Institute, The , Alderley Park SK10 4TG, UK. 2 Institut Curie, PSL Research University, INSERM U830, DNA Repair and Uveal Melanoma (D.R.U.M.), Equipe labellisée par la Ligue Nationale contre le Cancer, 75248 Paris, France. 3 Institut Curie, PSL Research University, Department of Medical Oncology, 75248 Paris, France. 4 Department of Pathology/Eye Pathology Section, University of Copenhagen, Rigshospitalet 2100 Copenhagen, Denmark. 5 Institut Curie, PSL Research University, Department of Ocular Oncology, 75248 Paris, France. 6 QIMR Berghofer Medical Research Institute, Brisbane, Queensland 4006, Australia. 7 The Christie NHS Foundation Trust, Manchester M20 4GJ, UK. 8 Experimental Pathology Service, Department of Pathology, Memorial Sloan Kettering Cancer Center, New York City, NY, USA. 9 Institut Curie, PSL Research University, Department of , 75248 Paris, France. 10 Institut Curie, PSL Research University, Translational Research Department, ✉ 75248 Paris, France. 11These authors contributed equally: Piyushkumar A. Mundra, Nathalie Dhomen. email: [email protected]

NATURE COMMUNICATIONS | (2021) 12:259 | https://doi.org/10.1038/s41467-020-20432-5 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20432-5

elanomas are a heterogeneous group of tumors with variants (SNVs) (median 100,098 [range 42,649–274,061] vs. Mdistinct genomic features that may be broadly classed as 10,391 [range 8426–20,538]; P < 0.0001; Fig. 1e, f). Similarly, epithelium-associated melanomas (includes cutaneous, compared to their non-SBS7v2 counterparts, the SBS7v2 cuta- acral, and mucosal melanomas) or non-epithelium associated neous melanomas presented higher proportions of C-to-T tran- melanomas (includes uveal and leptomeningeal melanoma)1. sitions at dipyrimidines (median 82.34% versus 36.09%; P < Non-epithelium associated melanomas have distinct clinical and 0.0001; Fig. 1c, d) and higher numbers of SNV (median 119,058 genomic features1,2, but even among epithelium-associated mel- [range 20,021–938,462] vs. 11716 [range 9838–12,607]; P < anomas, the relative frequency and combinations of genomic 0.0001; Fig. 1e, f). Notably, the SBS7v2 mucosal and common alterations vary between subtypes. For example, KIT and SF3B1 cutaneous melanomas presented similar proportions of SNV and mutations are more common in mucosal melanomas, whereas C-to-T transitions at dipyrimidines (Fig. 1b, d, f). Equally, the BRAF and NRAS mutations are more common in common non-SBS7v2 mucosal and common cutaneous melanomas pre- cutaneous melanomas1. sented similar proportions of SNVs, and similar proportions of Ultraviolet radiation (UVR)-induced DNA damage is a clini- C-to-T transitions at dipyrimidines (Fig. 1b, d, f) and other cally relevant factor that distinguishes the different melanoma nucleotide transitions/transversions (Supplementary Fig. 1a–e). subtypes3. It is clearly linked to the development of common Thus, nine conjunctival mucosal melanomas exhibited features of cutaneous melanomas, but its contribution to the rarer subtypes UVR exposure, whereas one conjunctival and the other eight is largely assumed to be negligible, because they tend to arise in mucosal melanomas did not present these features. sun-protected tissues. Mucosal melanoma (1.4% of melanomas) is We validated our findings in an independent cohort of 65 an example of such a rare melanoma subtype, which arises in the published mucosal melanoma whole genomes9. Here we mucosa of the eyes, mouth, nose, and gastrointestinal and geni- identified eight samples (12%) with SBS7v2 predominance tourinary tracts. It presents distinct biological and clinical fea- (Supplementary Fig. 2a), which also presented higher SNV loads tures, responds poorly to treatment, and is characterized by (Supplementary Fig. 2b), higher proportions of C-to-T transitions distinct genomic traits, with high numbers of chromosomal at dipyrimidines, and lower proportions of other transitions/ structural changes, low burden, and specific patterns of transversions than their non-SBS7v2 counterparts (Supplemen- driver oncogenes4–8. This is thought to be because mucosal tary Fig. 2c–h). melanomas arise in distinct microenvironments and are not driven by UVR1. However, two recent studies7,9 reported that 9% Structural variants distinguish mucosal melanomas from (6/67, 6/65 respectively) of mucosal melanomas present >50% of cutaneous melanomas and are independent of UVR mutation COSMIC single base substitution signature 7 (SBS7), a mutation signature status. Previous studies have reported increased signature associated with UVR10. It was also recently reported numbers of structural variants and indels in mucosal melanomas that uveal melanoma, another rare melanoma subtype not gen- compared to common cutaneous melanomas6. We investigated erally associated with UVR-exposure, can present SBS7- whether UVR-induced DNA damage influenced the extent or predominance if it arises on the iris11. We hypothesized that pattern of structural variation in mucosal melanomas. Consistent UVR drives melanomagenesis independent of tissue micro- with previous studies6,7, the mucosal melanomas presented more environment, so to test this we performed whole-genome structural variants and indels than common cutaneous melano- sequencing (WGS) on mucosal melanomas from the con- mas (Fig. 2a). Critically, we observed no significant differences junctiva, because this tissue is largely UVR exposed12. between the SBS7v2-dominant and the non-SBS7v2 cohorts In this study, we present a comparison of the genomic land- (Fig. 2a and Supplementary Fig. 3a), and similarly no significant scape of these UVR-exposed mucosal melanomas to other pri- differences in copy number variations (Fig. 2b). This was reca- mary mucosal melanomas and to primary cutaneous melanomas, pitulated in the validation cohort, where we again observed no to provide better understanding of the and mutational significant difference in chromosomal structural variants or processes that drive this particular melanoma subtype. number of indels between the SBS7v2-dominant and non-SBS7v2 mucosal melanomas (Supplementary Fig. 3b, c). Results UVR-driven DNA damage is predominant in mucosal mela- SBS7 dominance in mucosal melanoma is a better indicator of nomas of conjunctival origin. We performed WGS on 10 fresh UVR-exposure than tumor site. Large areas of the conjunctiva frozen primary conjunctival melanomas (median patient age 66 are highly sun-exposed, and this is reflected in the SBS7v2 years, range 38–84; six females, four males; Supplementary dominance in nine of ten genomes from our primary conjunctival Table 1) and compared our results to published WGS from eight melanomas. MuM12, MuM13, MuM16, and MuM17 were loca- mucosal melanomas originating on sun-protected sites (nasal, lized to the limbus and MuM14 to the upper part of the bulbar genitourinary, rectal; median patient age 63 years, range 46–84; conjunctiva, areas that are frequently sun-exposed (Supplemen- six females, two males; Supplementary Table 2)6. Mutational tary Table 1). Note, however, that SBS7v2 also dominated the signature analysis on our WGS data revealed a predominance of genomes of MuM10, MuM11, and MuM18, which were from the COSMIC SBS7v2 in nine of the conjunctival melanomas, whereas tarsal conjunctiva which is considered to be more sun-protected, one conjunctival melanoma and the other eight mucosal mela- and SBS7v2 dominated the genome of MuM15, which was a large nomas were dominated by the SBS1v2 (age-related), SBS5v2 lesion spanning the sun-protected fornix and the sun-exposed (ubiquitous), and SBS3v2 (BRCA1, BRCA2, and PALB2-asso- caruncle (Supplementary Table 1). Note also that MuM1, which ciated) (Fig. 1a, b). To facilitate direct comparison with common presented the lowest mutation burden and was the only primary cutaneous melanoma, we used our pipeline to analyze published conjunctival melanoma that did not exhibit SBS7v2 dominance, WGS from 54 primary common cutaneous melanomas6 and arose from the fornix, considered to be the most sun-protected observed SBS7v2 predominance in 51 of these samples (Fig. 1a, part of the conjunctiva. Similarly, the conjunctival melanoma b). Compared to their non-SBS7v2 counterparts, the SBS7v2 with the next lowest mutation burden and SBS7v2 contribution mucosal melanomas presented higher proportions of C-to-T was MuM10, another forniceal tumor. transitions at dipyrimidines (mean 84.7% versus 29.0%; P < In our validation cohort, the SBS7v2-dominant mucosal 0.0001; Fig. 1c, d) and higher numbers of single nucleotide melanomas were from potentially sun-exposed sites, including

2 NATURE COMMUNICATIONS | (2021) 12:259 | https://doi.org/10.1038/s41467-020-20432-5 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20432-5 ARTICLE

ns Mucosal melanomas Common cutaneous melanomas p<0.0001 p<0.0001 a Non- b 1.0 Non-SBS7v2 SBS7v2SBS7v2 SBS7v2 1.0 Unassigned

%) 0.8

Signature 18 s( ht

0.8 g 0.6

Signature 16 i e Signature 11 0.4 0.6 Signature 9 7v2 w S Signature 8 0.2 SB 0.4 Signature 7 Signature 5 0.0 t t 0.2 Signature 3 M M u u u C M Mu 2 2C Signature 2 2 v v

Mutational Signature Weights v 7 7 7 7v2 S S 0.0 Signature 1 B B BS BS S S 4 7 3 3 9 6 1 0 1 2 7 4 8 9 2 6 4 6 - t t5 t t9 S S 51 2 21 30 M M2 M3 M8 - n t35 t15 t t t3 u u u u M1 M1 M1 n o u u Signature 12 ut4 ut50 u ut2 ut2 ut20 u u u Cut2 Cut7 Cut8 C C C Cut1 Cut3 C u u uM1 o N Cut11 Cut17 Cut40 Cut41 Cut42 Cut45 Cut49 C Cut52 Cut54 Cut24 Cut26 Cut27 Cut28 Cut31 Cut33 Cut34 C C Cut3 C C Cu Cut16 Cut18 C C Cut10 Cut12 Cut13 Cut14 C Cut4 C Cut Cut46 Cut4 Cut48 Cut53 Cut25 Cut Cut32 Cut37 Cut38 Cut19 Mu M Mu MuM4 MuM5 MuM MuM7 M MuM9 N Mu MuM1 M MuM13 M M MuM1 MuM15 MuM16

c ns 100 d p<0.0001 p<0.0001 C>A 100 C>G 80 C>T 50 T>A 60 T>C

T>G C>T (%) 40

0 20 Fractions of Mutations (%) 0 4 6 8 9 7 2 9 3 6 5 6 9 3 7 4 1 6 7 0 5 7 7 5 t3 t t t 3 3 1 1 17 1 4 1 M2 M M1 M M t52 t3 t t t22 t u M M12 M M14 M1 uM u u u u uM3 u ut31 u ut5 ut50 ut5 u ut35 u ut28 u ut1 ut14 ut4 ut24 Cut2 Cut8 Cut1 Cu Cut4 Cut Cu Cu C u u u uM u Cut11 Cut53 C C C Cut42 Cut44 Cut33 Cut34 C Cut36 C C Cut30 C C Cut Cut2 C C Cut18 Cu Cut23 C C C Cut47 Cut48 Cut49 Cut41 Cut43 Cut45 C Cut29 Cut38 Cut25 Cut26 Cut10 Cut12 Cut19 Cut20 Cut21 C Cut1 Cut16 M M M M MuM5 M Mu MuM M MuM11 M M MuM18 M Mu MuM M M 0 t M u uM u M M 2Cut 2C 2 2 v v v v 7 S S7 S7 B B BS7 -S S e 1×106 SB S n o on- N N Mucosal Melanoma Cutaneous Melanoma ns 8×105 f p<0.0001 p<0.0001 1000000 6×105

100000 tal SNV

o 4×105 T

Total SNV 10000

2×105

1000 t t M u 0 u uM Cu C M M 2 2 6 1 0 8 4 4 2 0 9 7 2 5 4 2 3 0 7 4 2 6 9 1 8 5 t4 t1 t7 2 v v 1 4 5 4 1 2 4 4 3 25 10 1 1 20 22 4 5 1 M8 M3 M4 M M M t29 t t t t t12 t t t43 t38 t48 t4 t t19 t23 ut9 u v M M17 M1 M M ut11 u u u u ut32 u u u u u ut53 ut27 u u u Cut2 Cu Cut5 Cut6 Cu Cut8 C C Cut3 7 S7 S7 u uM1 u u C Cut45 Cut Cut Cut33 Cut35 Cut37 Cu C Cut28 Cu Cut Cut4 Cut Cu Cut51 Cut26 Cut1 Cut34 Cut36 Cut16 Cut21 C C Cut31 C C Cut39 Cut50 Cu Cut C C C Cut1 Cut14 Cut Cut17 C Cut C Cu C Cu C MuM5 Mu M MuM1 MuM Mu Mu M M S B MuM Mu MuM13 M M MuM16 M M Mu BS7v2 B -S SB -S S n o N Non

Fig. 1 Mutation spectra in mucosal and common cutaneous melanomas. a Mutation signatures weighted by relative contribution to spectrum of mutations in individual tumor genomes. Indices above indicate mucosal (black, n = 18) or common cutaneous (red, n = 54) melanomas with subdivision into non-SBS7v2 (blue, n = 9, n = 3, respectively) and SBS7v2 (magenta, n = 9, n = 51 respectively) genomes. Columns represent individual tumors. b Relative contribution of SBS7v2 to mutation spectra (SBS7v2 Weights) for individual mucosal (MuM) or cutaneous (Cut) melanomas with SBS7v2 (magenta n = 9, gray n = 51) or non-SBSv2 (maroon n = 9, blue n = 3) genomes. c Proportion of six nucleotide transitions/transversions for individual tumor genomes (individual columns). d Proportions of C > T nucleotide transitions in individual mucosal (MuM) or cutaneous (Cut) melanomas with SBS7v2-dominant (magenta n = 9, gray n = 51) or non-SBSv2 (maroon n = 9, blue n = 3) genomes. e Total SNVs in individual mucosal (MuM, magenta, n = 18) or cutaneous (Cut, gray, n = 54) melanomas. Columns represent individual tumors. f Total SNVs in mucosal (MuM) and cutaneous (Cut) melanomas with SBS7v2-predominant (magenta n = 9, gray n = 54) or non-SBSv2 (maroon n = 9, blue n = 3) genomes. Panels b, d, f show median and 95% confidence intervals, dots denote individual tumors, P-values determined by two-tailed Mann–Whitney U,ns= not significant: 0.4140 in b, 0.8387 in d, 0.6838 in f, respectively. the lips (3/5), gingiva (2/28), nasal cavity (1/2), multi-sites (lip We observed that six of the nine UVR mucosal melanomas car- and gingiva) (1/2), and oropharynx (1/1)9. Thus, in these samples ried BRAF mutations, whereas only one of the nine non-UVR also, SBS7v2-dominant mucosal melanomas were largely from mucosal melanomas had a BRAF mutation (Fig. 2c). Notably, potentially sun-exposed sites, but it should be noted that the eight of the nine UVR-exposed mucosal melanomas and all 51 SBS7v2-domimant mucosal melanomas were still in the minority. UVR-exposed cutaneous melanomas carried mutations in one to Thus, although the precise location of these melanomas is not eleven known melanoma genes, with the remaining mucosal known, our analysis suggests that the SBS7v2 dominance is a melanoma (MuM10) carrying a frame-shift mutation in the more specific marker of UVR-driven processes than tumor site melanocyte gene TYRP1 (Fig. 2c and Supplementary Data 1). and henceforth we refer to these tumors as UVR-exposed Thus, FGFR2/4, ERBB4, NF1, CDKN2A, NFKBIE, SALL4, TERT, mucosal melanomas. GRIN2A, and TP53 mutations were restricted to UVR-exposed melanomas (Fig. 2c). Conversely, the non-UVR-exposed mucosal and cutaneous melanomas carried mutations in only two UVR-exposed mucosal and cutaneous melanomas present (1 sample), one (5 samples), or none (6 samples) of the melanoma similar driver mutations. We next investigated genes (Fig. 2c). Additionally, 31 (52%) UVR-exposed melanomas mutations in common melanoma-associated oncogenes. BRAF had TERT and/or TERT promoter mutations, but only one (8%) mutations are generally associated with common cutaneous non-UVR-exposed melanoma had a mutation in this gene melanoma, but only weakly associated with mucosal melanoma1. (Fig. 2c, d). These data show remarkable enrichment for the same

NATURE COMMUNICATIONS | (2021) 12:259 | https://doi.org/10.1038/s41467-020-20432-5 | www.nature.com/naturecommunications 3 niae ee o niiultmr clms.Dte ie eoesgeainit o-B72adSSv oiatgenomes. dominant SBS7v2 and non-SBS7v2 into segregation denote lines Dotted (columns). tumors individual for genes indicated rmsit(le uain nkonmlnm noee ( oncogenes melanoma known in mutations (blue) frameshift mut fcrmsmlgisadlse fato fgnm)i niiulgnms oun ersn niiultumors. individual represent Columns genomes. individual in genome) of (fraction losses and gains chromosomal of amounts (blue; 4 melanomas. (black, cutaneous mucosal common indicate and above mucosal Indices in mutations genomes. gene melanoma and alterations Structural 2 Fig. ARTICLE TERT TERT n = (5:1295250) (5:1295228) e d c b a 9, CDKN2A PCDHAC1 MAP3K1 PKHD1L1 GRIN2A NFKBIE ERBB4 FGFR4 FGFR2 TYRP1 DNAH10 SALL4 PCDH15 n NRAS BRAF

DNAH9 Numbers of Structural Variants TERT USH2A LRP1B GPR98 TP53 XIRP2 =

NF1 Fraction of Genome KIT TP53 0.0 0.5 1.0 100 200 300 400 ,rsetvl)adSSv-oiat(magenta; SBS7v2-dominant and respectively) 3, AUECOMMUNICATIONS NATURE 0

MuM1 MuM1 MuM1 MuM1 MuM1 Mucosal melanomas MuM2 MuM2 MuM2 MuM2 MuM2 Non-SBS7v2 MuM3 MuM3 MuM3 MuM3 MuM4 MuM3 MuM4 MuM4 MuM4 MuM4 MuM5 MuM5 MuM5 MuM6 MuM5 MuM6 MuM5 MuM6 MuM7 MuM6 MuM7 MuM7 MuM6 MuM8 MuM8 MuM7 MuM8 MuM7 MuM9 MuM9 MuM9 MuM8 MuM10 MuM8 MuM10 MuM10 MuM9 MuM11 MuM9 MuM11 MuM11 MuM12 MuM12 MuM10 MuM13 MuM10 MuM12 MuM11 MuM14 MuM13 MuM13 MuM15 MuM11 MuM14 MuM14 MuM12 MuM16 SBS7v2 SBS7v2 MuM12 MuM15 MuM15 MuM13 MuM17 MuM13 MuM16 MuM16 MuM14 MuM18 MuM17 Cut1 (01 229 tp:/o.r/013/447002425|www.nature.com/naturecommunications | https://doi.org/10.1038/s41467-020-20432-5 | 12:259 (2021) | MuM17 MuM15 MuM14 MuM18 Cut2 MuM18 MuM16 Cut3 MuM15 Cut1 Cut1 MuM17 Cut4 SBS7v2

Cut2 Non- MuM16 Cut2 MuM18 Cut5 MuM17 Cut3 Cut3 Cut6 Cut4 Cut4 Cut1 Cut7 Cut8 MuM18 Cut5 Cut5 Cut2 Cut6 Cut9 Cut1 Cut6 Cut3 Cut10 Cut2 Cut7 Cut7 Cut4 Cut11 Cut8 Cut8 Cut5 Cut12 Cut3 Cut9 Cut13 Cut9 Cut6 n c Cut4 Cut10 Cut10 Cut14

,or ), Cut15 = Cut7 Cut5 Cut11 Cut11 Cut12 Cut8 Cut16

8 n omnctnos(red, cutaneous common and 18) Cut6 Cut12 Cut17 Cut13 Cut13 Cut9 Cut18

n Cut7 TERT Cut14 Cut14 Cut10 Cut19 = Cut8 Cut15 Cut15 Cut11 Cut20 Cut16 Cut21 Cut9 Cut16 Common cutaneous melanomas 9, Cut17 Cut12 Cut22 Cut17 Cut23 Cut10 Cut13 https://doi.org/10.1038/s41467-020-20432-5 | COMMUNICATIONS NATURE rmtrrgos( regions promoter Cut18 n Cut18 Cut14 Cut24 Cut11 Cut19 Cut25

= Cut19 Cut12 Cut20 Cut20 Cut15 Cut26

1 epciey eoe.Clmsrpeetidvda tumors. individual represent Columns genomes. respectively) 51, Cut21 Cut13 Cut21 Cut16 Cut27 Cut22 Cut17 Cut28 Cut14 Cut22 Cut29 Cut23 Cut18 Cut23 Cut24 Cut15 Cut24 Cut19 Cut16 Cut25 Cut25 Cut26 Cut20 Cut17 Cut26 Cut27 Cut21 Cut27 Cut28 Cut18 Cut28 Cut22 Cut19 Cut29 Cut23 Cut29 Cut30 6 Cut20 Cut30 Cut24 d C Cut27 Cut31 Cut25 fidvda tumors. individual of ) Cut21 Cut28 Cut32 Cut2 CuCutt2229 Cut33 Cut29 Cut30 Cut30 CutCut2330 Cut31 Cut24 ut31 Cut31 Cut32 n Cut31 ubr n ye fsrcua ainsi individual in variants structural of types and Numbers Cut32 Cut32 Cut33

= Cut25 Cut32 Cut33 Cut33 Cut34 Cut26 Cut34 Cut34 Cut35

4 eaoa ihsbiiinit non-SBS7v2 into subdivision with melanomas 54) Cut33 Cut35 Cut34 Cut36 Cut27Cut34 Cut35 Cut36 Cut36 Cut35 Cut37 Cut28Cut35 Cut36 Cut38 Cut37 Cut37 Cut39 Cut36 Cut38 Cut38 Cut37 Cut40 Cut37 Cut39 Cut39 Cut38 Cut41 Cut40 Cut40 Cut42 Cut38 Cut39 Cut43 Cut41 Cut41 Cut40 Cut39 Cut42 Cut44 Cut42 Cut41 Cut45 e Cut43 Cut43 Cut40 Cut42 Cut46 isnemttos(rne nthe in (orange) mutations Missense Cut41 Cut44 Cut44 Cut47 Cut45 Cut45 Cut43 Cut48 Cut42 Cut46 Cut46 Cut44 Cut49 Cut43 Cut47 Cut47 Cut45 Cut50 Cut48 Cut51 Cut44 Cut48 Cut46 Cut52 Cut49 Cut49 Cut45 Cut47 Cut53 Cut50 Cut50 Cut54 c Cut46 Cut51 Cut51 Cut48 , Cut52 Cut52 Cut49 d Cut47 Cut53 Cut53 Cut50

isne(rne or (orange) Missense Cut48 Cut54 Cut54 Cut51 Cut49 Cut52 Cut50 Cut53 Translocation Deletion Cut51 Cut54 Inversion Duplication Cut52 Cut53 Cut54 Loss Gain b Relative NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20432-5 ARTICLE driver oncogene mutations in UVR-exposed cutaneous mela- homogeneous group of diseases and suggest that tumors arising noma and UVR-exposed mucosal melanoma. from mucosal melanocytes are subject to a common tumorigenic process resulting in the accumulation of structural genome var- iations, to which are added UVR-driven processes in a subset of Ten-gene panel as surrogate to UVR signature in mucosal cases. This aligns with recent reports that SF3B1 R625 mutations melanomas. We previously reported that mutations in a ten-gene are recurrently present in vulvo-vaginal and anorectal melanomas panel (LRP1B, GPR98, XIRP2, PKHD1L1, USH2A, DNAH9, but not in other mucosal locations25,26. Our study therefore PCDH15, DNAH10, TP53, PCDHAC1) were a surrogate of UVR contributes to the definition of biologically and clinically relevant exposure3. We therefore investigated if this panel could segregate subsets of mucosal melanomas, providing better insight into their UVR-exposed from non-UVR exposed mucosal melanomas. and opening avenues for precision medicine. Remarkably, this panel correctly segregated 71/72 (97%) of the The UVR-driven mucosal melanomas tend to arise on sun- UVR-exposed mucosal and cutaneous melanomas (Fig. 2e), exposed sites such as the conjunctiva and lips, but sun exposure including all nine UVR mucosal melanomas, eight of which per se does not identify this subset of mucosal melanomas, as carried mutations in two or more of these genes (Fig. 2e). This highlighted by the presence of non-SBS7v2 tumors at potentially panel therefore provides a targeted approach that allows rapid sun-exposed sites and SBS7v2 tumors at more sun-protected sites. screening for UVR-exposed mucosal melanomas. Some of our SBS7v2 cases came from gingiva, oropharynx, or nasal cavity. Whilst imprecision in site reporting may play a role Discussion in this, one possibility is that mucosal cells in sun-exposed sites Although our cohort was small due to the challenges inherent in may accumulate UVR-induced mutations and expand into large obtaining samples, we present WGS for ten conjunctival mela- clones of mutant cells, extending into sun-protected areas where nomas, and our results extend previous studies by showing that such cells could further develop into a melanoma. Clonal conjunctival mucosal melanomas have similar genomes to com- expansions of this nature have been reported in the skin27 and in mon cutaneous melanoma12–14. We also show that like common Barrett’s esophagus28. Conversely, the presence of non-SBS7v2 cutaneous melanoma3, mucosal melanomas present two broad tumors at potentially sun-exposed sites is consistent with our groups, one with SBS7v2 predominance that appears to be UVR- previously report that in a UVR/BRAFV600E-driven mouse mel- driven, and one that is not UVR-driven, but curiously, both anoma model, a small number of tumors developed without groups present the large structural alterations more commonly evidence of UVR-associated DNA damage3. As outlined above, observed in mucosal melanoma. Our analysis revealed particu- distinguishing UVR and non-UVR melanomas in the mucosal larly striking similarities between UVR-exposed mucosal mela- and other settings could have important implications for clinical nomas and UVR-driven common cutaneous melanomas, as both care. Our data shows that this cannot be determined by the site of presented high mutation burdens and abundant mutations that the primary tumor, but we propose that our 10-gene panel could activate the BRAF–ERK pathway. provide rapid testing for the UVR signature without the cost Notably, BRAF mutations are rare in mucosal melanoma and complexity of WGS. Together with the knowledge that the compared to common cutaneous melanoma15,16,soBRAF UVR-associated mucosal melanomas could benefit from treat- mutation testing is not recommended or reimbursed in some ments currently used in common cutaneous melanoma, our jurisdictions. However, we show that UVR-driven mucosal mel- findings highlight an approach to delivering precision medicine to anomas harbor high frequency BRAF mutations, so could benefit this patient group for whom current treatment options are from BRAF and MEK targeted therapies. We note that the first limited. published case of a patient with a BRAF-mutated metastatic conjunctival melanoma treated with vemurafenib was reported 7 Methods years after the first clinical results of this drug17, and 5 years after Sample collection and ethics. Conjunctival melanoma samples MuM1 and FDA approval for cutaneous melanoma patients18. Moreover, MuM10–18 comprise two cohorts, one from Institut Curie, Paris (MuM10, – although CKIT mutations are present in about 15% of mucosal MuM12, MuM15 18) and one from the department of Pathology/Eye Pathology 1 Section, University of Copenhagen, Rigshospitalet (MuM1, MuM11, MuM13, melanomas response rates to KIT inhibitors range from only 5 to MuM14). The studies were approved by the Internal Review Board of Institut Curie 26%, and no KIT drugs are approved for use in these patients. (2014) and the Danish National Ethics Committee (j. no. 1700673), respectively. Mucosal melanoma patients are unfortunately also generally The samples were collected during surgery and frozen immediately. Paired blood excluded from immunotherapy trials19–21 and because of this samples were also collected and stored for sequencing. Patients provided written informed consent to perform germline and somatic genetic analyses for WGS on exclusion, immunotherapies are not approved for mucosal mel- archived frozen samples. anoma in the adjuvant setting. Note, however, that response rates to immune checkpoint inhibitors in advanced mucosal melano- DNA extraction and WGS. For Institut Curie cohort, DNA was extracted by the mas are at least half of that seen in non-glabrous skin melano- Centre de Ressources Biologiques (Institut Curie tumor biobank) from frozen mas22,23, and it was recently reported that four of five samples using phenol (Invitrogen), then subsequently purified on Zymo-Spin IC conjunctival melanoma patients responded to immunotherapy24. (Zymo Research). DNA was extracted from paired whole blood samples using the QuickGene DNA whole-blood kit with QuickGene-610L equipment (Fujifilm, The clear similarities between UVR-driven cutaneous melanoma Japan). DNA concentrations were quantified by Qubit (Thermo Fisher Scientific). and UVR-associated mucosal melanoma suggest that mucosal Integrities were assessed by a BioAnalyzer 2100 device (Agilent Technologies, Santa melanoma patients with SBS7v2 predominance may benefit from Clara, CA, USA). For Rigshospitalet cohort, DNA/RNA was extracted from frozen BRAF/MEK inhibitor combinations and from immunotherapies tumor samples using Norgen (Biotek Corp.) kit and from blood samples using QIAamp DNA Blood and Tissue kit (Qiagen, Manchester, UK) as per manu- in both the advanced and adjuvant settings. facturer’s instructions. Concentrations were quantified by Qubit (Thermo Fisher Despite the similarities in mutation burden and oncogene Scientific). pathway activation, we note that chromosomal structural varia- tions did not distinguish UVR from non-UVR mucosal mela- Sequencing and processing. WGS of the tumor-normal pairs from patients nomas, but did distinguish mucosal from common cutaneous included in the Institut Curie cohort (MuM10, MuM12, MuM15-18) was per- melanoma. This suggests that UVR imposes additional processes formed in the New York Genome Center (NYGC). DNA libraries were prepared fi using TruSeq PCR-free approach following the protocols implemented in the over the microenvironment-speci c mutational processes that NYGC. WGS of the tumor-normal pairs of the Rigshospitalet cohort (MuM1, otherwise drive the different melanoma subtypes. Together, these MuM11, MuM13, MuM14) was performed by Edinburgh Genomics (The Roslin data show us that mucosal melanomas do not belong to a Institute, University of Edinburgh) using TruSeq Nano library preparation method.

NATURE COMMUNICATIONS | (2021) 12:259 | https://doi.org/10.1038/s41467-020-20432-5 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20432-5

Sequencing was performed on Illumina HiSeqX machine for both the cohorts with 11. Johansson, P. A. et al. Whole genome landscapes of uveal melanoma show an aimed coverage of 30× and 60× for normal (blood) and tumor samples, ultraviolet radiation signature in iris tumours. Nat. Commun. 11, 2408 (2020). respectively. 12. Rivolta, C. et al. UV light signature in conjunctival melanoma; not only skin WGS BAM files of other primary mucosal melanomas (MuM2–MuM9) and should be protected from solar radiation. J. Hum. Genet. 61, 361–362 (2016). primary cutaneous melanomas6 (Cut1-Cut54) were downloaded from EGA using 13. Griewank, K. G. et al. Conjunctival melanomas harbor BRAF and NRAS accession ID EGAS00001001552 using ASPERA (v3.5.4). mutations and copy number changes similar to cutaneous and mucosal melanomas. Clin. Cancer Res. 19, 3143–3152 (2013). Bioinformatics analysis. FASTQ files were extracted for BAM files for 14. Swaminathan, S. S. et al. Molecular characteristics of conjunctival melanoma MuM2–MuM9 and cutaneous melanoma samples using samtools29 (v1.3.1). The using whole-exome sequencing. JAMA Ophthalmol. 135, 1434–1437 (2017). 2 × 150 read pairs were mapped to the reference genome GRCh37 (v75) using 15. Cosgarea, I. et al. Targeted next generation sequencing of mucosal melanomas BWA-mem30 (v0.7.7) tool. This was followed by duplicate removal using PICARD identifies frequent NF1 and RAS mutations. Oncotarget 8, 40683–40692 (2017). (v1.96) and INDELs realignment and recalibration of base qualities using GATK31 16. Flaherty, K. T., Hodi, F. S. & Bastian, B. C. Mutation-driven drug development (v3.6). in melanoma. Curr. Opin. Oncol. 22, 178–183 (2010). The final BAM files from both tumor and corresponding blood sample were 17. Flaherty, K. et al. Phase I study of PLX4032: proof of concept for V600E BRAF used for subsequent somatic variant calling using MUTECT32 (v1.1.7) with default mutation as a therapeutic target in human cancer. J. Clin. Oncol. 27, parameters. Small insertions and deletions were determined using Strelka33 9000–9000 (2009). (v1.0.4). Only “Passed” calls were considered. Variant effect predictor34 (Ensembl 18. Maleka, A., Astrom, G., Bystrom, P. & Ullenhag, G. J. A case report of a version 73) was used to annotate the mutations. Known variants present in dbSNP patient with metastatic ocular melanoma who experienced a response to were excluded. Structural variants were determined using DELLY35 (v0.8.1) with treatment with the BRAF inhibitor vemurafenib. BMC Cancer 16, 634 (2016). default parameters. 19. Weber, J. et al. Adjuvant nivolumab versus ipilimumab in resected stage III or Mutational signatures were determined by fitting somatic SNVs with tri- IV melanoma. N. Engl. J. Med. 377, 1824–1835 (2017). nucleotide context to the 30 COSMIC mutational signatures using deconstructSigs 20. Eggermont, A. M. M. et al. Adjuvant pembrolizumab versus placebo in (v1.8.0)36 package using default parameters. Signatures with contribution weights resected stage III melanoma. N. Engl. J. Med. 378, 1789–1801 (2018). less than 6% were excluded. 21. Robert, C. et al. Pembrolizumab versus ipilimumab in advanced melanoma. N. 37 Copy number alterations were determined using Sequenza (v2.1.9999b0) Engl. J. Med. 372, 2521–2532 (2015). = = = package with parameters (mufreq.treshold 0.05, min.reads 10, min.fw.freq 22. Mignard, C. et al. Efficacy of immunotherapy in patients with metastatic − 0.1). One cutaneous sample with missing gender information in clinical data was mucosal or uveal melanoma. J. Oncol. 2018, 1908065 (2018). excluded from the analysis. Fraction of genomic alteration for each sample was 23. Nathan, P. et al. Safety and efficacy of nivolumab in patients with rare calculated using an in-house script. melanoma subtypes who progressed on or after ipilimumab treatment: a single-arm, open-label, phase II study (CheckMate 172). Eur. J. Cancer 119, Reporting summary. Further information on research design is available in the Nature 168–178 (2019). Research Reporting Summary linked to this article. 24. Sagiv, O. et al. Immunotherapy With programmed death 1 inhibitors for 5 patients with conjunctival melanoma. JAMA Ophthalmol. 136, 1236–1241 Data availability (2018). 25. Quek, C. et al. Recurrent hotspot SF3B1 mutations at codon 625 in The whole-genome sequencing data generated in this study from conjunctival melanoma vulvovaginal mucosal melanoma identified in a study of 27 Australian samples have been deposited in the EGA database under accession code mucosal melanomas. Oncotarget 10, 930–941 (2019). EGAS00001004697. The data is available under restricted access, which can be obtained 26. Yang, H. M. et al. Identification of recurrent mutational events in anorectal by contacting Prof. Richard Marais. The whole-genome sequencing data accessed for use melanoma. Mod. Pathol. 30, 286–296 (2017). in this study corresponding to cutaneous melanoma and other mucosal melanoma 27. Klein, A. M., Brash, D. E., Jones, P. H. & Simons, B. D. Stochastic fate of p53- samples are available from the EGA database under accession code EGAS00001001552. mutant epidermal progenitor cells is tilted toward proliferation by UV B The remaining data are available within the Article, Supplementary Information, or during preneoplasia. Proc. Natl Acad. Sci. USA 107, 270–275 (2010). available from the author upon request. 28. Maley, C. C. et al. Selectively advantageous mutations and hitchhikers in neoplasms: p16 lesions are selected in Barrett’s esophagus. Cancer Res. 64, Code availability 3414–3427 (2004). In-house codes used to compute fraction of genome altered are available at: https:// 29. Li, H. et al. The Sequence Alignment/Map format and SAMtools. github.com/mpiyush21/MucosalNatureComms. Bioinformatics 25, 2078–2079 (2009). 30. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows- Wheeler transform. Bioinformatics 25, 1754–1760 (2009). Received: 10 July 2020; Accepted: 25 November 2020; 31. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010). 32. Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat. Biotechnol. 31, 213–219 (2013). 33. Kim, S. et al. Strelka2: fast and accurate calling of germline and somatic References variants. Nat. Methods 15, 591–594 (2018). 1. Bastian, B. C. The molecular pathology of melanoma: an integrated taxonomy 34. McLaren, W. et al. The ensembl variant effect predictor. Genome Biol. 17, 122 of melanocytic neoplasia. Annu. Rev. Pathol. 9, 239–271 (2014). (2016). 2. Robertson, A. G. et al. Integrative analysis identifies four molecular and 35. Rausch, T. et al. DELLY: structural variant discovery by integrated paired-end clinical subsets in uveal melanoma. Cancer Cell 32, 204–220 e215 (2017). and split-read analysis. Bioinformatics 28, i333–i339 (2012). 3. Trucco, L. D. et al. Ultraviolet radiation-induced DNA damage is prognostic 36. Rosenthal, R., McGranahan, N., Herrero, J., Taylor, B. S. & Swanton, C. for outcome in melanoma. Nat. Med. 25, 221–224 (2019). DeconstructSigs: delineating mutational processes in single tumors 4. Tyrrell, H. & Payne, M. Combatting mucosal melanoma: recent advances and distinguishes DNA repair deficiencies and patterns of carcinoma evolution. future perspectives. Melanoma Manag. 5, MMT11 (2018). Genome Biol. 17, 31 (2016). 5. Furney, S. J. et al. Genome sequencing of mucosal melanomas reveals that they 37. Favero, F. et al. Sequenza: allele-specific copy number and mutation profiles are driven by distinct mechanisms from cutaneous melanoma. J. Pathol. 230, from tumor sequencing data. Ann. Oncol. 26,64–70 (2015). 261–269 (2013). 6. Hayward, N. K. et al. Whole-genome landscapes of major melanoma subtypes. Nature 545, 175–180 (2017). 7. Newell, F. et al. Whole-genome landscape of mucosal melanoma reveals Acknowledgements diverse drivers and therapeutic targets. Nat. Commun. 10, 3163 (2019). R.M. is supported by Wellcome Trust (100282/Z/12/Z), European Research Council 8. Mikkelsen, L. H. et al. Mucosal malignant melanoma - a clinical, oncological, (ERC-ADG-2014 671262), Cancer Research UK (A27412 and A22902). M.R. and pathological and genetic survey. APMIS 124, 475–486 (2016). M.-H.S. are funded by Institut Curie, the Ligue Nationale Contre le Cancer (Labellisa- ’ 9. Zhou, R. et al. Analysis of mucosal melanoma whole-genome landscapes tion), and Site de Recherche Intégrée sur le Cancer (SiRIC) de l Institut Curie. M.-H.S. is reveals clinically relevant genomic aberrations. Clin. Cancer Res. 25, also funded by Institut National de la Santé et de la Recherche Médicale. R.M., M.-H.S., 3548–3560 (2019). and S.R. are supported by the European Commission under the Horizon 2020 program 10. Alexandrov, L. B. et al. Signatures of mutational processes in human cancer. (U.M. Cure; project number: 667787). L.H.M. was funded by a non-restricted Candys Nature 500, 415–421 (2013). Foundation grant. J.S.R.-F. is funded in part the Breast Cancer Research Foundation. S.V.

6 NATURE COMMUNICATIONS | (2021) 12:259 | https://doi.org/10.1038/s41467-020-20432-5 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20432-5 ARTICLE

is supported by a Harry J Lloyd Charitable Trust Career Development Award. WGS costs Peer review information Nature Communications thanks the anonymous reviewers for were partly supported by the Manchester NIHR Biomedical Research Centre. their contribution to the peer review of this work. Peer reviewer reports are available.

Reprints and permission information is available at http://www.nature.com/reprints Author contributions Conceptualization, P.A.M., M.R., L.H.M., N.D., and R.M.; Methodology, K.B., S.V., N.C., Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in J.S.R.-F.; Formal analysis, P.A.M., M.R., L.H.M., and N.D.; Resources, M.R., L.H.M., N.C., published maps and institutional affiliations. S.H., M.-H.S., S.R., J.S.R.-F.; Writing, P.A.M., N.D., and R.M.; Supervision, S.H., M.-H.S., S.R., and R.M. Open Access This article is licensed under a Creative Commons Competing interests Attribution 4.0 International License, which permits use, sharing, R.M. consultants for Pfizer, and as a former Institute of Cancer Research (London) adaptation, distribution and reproduction in any medium or format, as long as you give employee, he may benefit from commercialized programs. J.S.R.F. consultants for appropriate credit to the original author(s) and the source, provide a link to the Creative Goldman Sachs, REPARE Therapeutics and Paige.AI, and serves on the scientific advi- Commons license, and indicate if changes were made. The images or other third party sory boards of Volition Rx and Paige.AI, and ad hoc on the scientific advisory boards of material in this article are included in the article’s Creative Commons license, unless Ventana Medical Systems, Roche Tissue Diagnostics, Genentech, Novartis, and InVicro. indicated otherwise in a credit line to the material. If material is not included in the All other authors declare no competing interests. article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from Additional information the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Supplementary information is available for this paper at https://doi.org/10.1038/s41467- licenses/by/4.0/. 020-20432-5. © The Author(s) 2021 Correspondence and requests for materials should be addressed to R.M.

NATURE COMMUNICATIONS | (2021) 12:259 | https://doi.org/10.1038/s41467-020-20432-5 | www.nature.com/naturecommunications 7