Somatic mutations in facial skin from countries of contrasting skin cancer risk

The incidence of keratinocyte cancer (basal cell and squamous cell carcinomas of the skin) is 17-fold lower in Singapore than the UK1–3, despite Singapore receiving 2–3 times more ultraviolet (UV) radiation4,5. Aging skin contains somatic mutant clones from which such cancers develop6,7. We hypothesized that differences in keratinocyte cancer incidence may be reflected in the normal skin mutational landscape. Here we show that, compared to Singapore, aging facial skin from populations in the UK has a fourfold greater mutational burden, a predominant UV mutational signature, increased copy number aberrations and increased mutant TP53 selection. These features are shared by keratinocyte cancers from high-incidence and low-incidence populations8–13. In Singaporean skin, most mutations result from cell-intrinsic processes; mutant NOTCH1 and NOTCH2 are more strongly selected than in the UK. Aging skin in a high-incidence country has multiple features convergent with cancer that are not found in a low-risk country. These differences may reflect germline variation in UV-protective genes.

1. Abstract line one, and final sentence page 2. References 1-3, used to report the rela*ve incidences of SCC across Singapore and the UK are from 2005 and earlier -are there no more recent surveys of skin cancer incidence in these countries?Figure 1a shows incidence rates up un*l 2015 sugges*ng different sources have been used then were cited (ref 1) in figure 1#s footnote; Suggested ac*on: check/update references 2. The final sentence of the abstract offers no conclusion as to why skin from a low-risk country does not show mul*ple features convergent with cancer; the likely answer from the data is the degree of skin pigmenta*on and propensity to tan.This should be addressed.Also, see points 4 and 11.
Suggested ac*on: Update abstract/text to address 3. Introduc*on, paragraph 1, page 2, sentence 5: Refs 18-21 aren#t cita*ons to a large twin study (I assume the authors mean Mucci 2016 PMID: 26746459) nor are they all cita*ons to GWAS of KC cancer.
18 is a PRS study, 20 isn#t a GWAS, and there are other published KC GWAS other than refs 19 and 21 that should be cited.
Suggested ac*on: check/update references 4. Final sentence of introduc*on paragraph 1: In my opinion saying that some GWAS SNPs for KC are linked to skin pigmenta*on somewhat undersells this fact.For example, in ref 21 the majority of the most strongly associated loci (and thus represent the largest combina*on of effect sizes and minor allele frequency, and thus the propor*on of trait variance explained) are all pigmenta*on SNPs.While other pathways are important in KC risk, pigmenta*on is by and far the strongest gene*c risk for KC, which is why the rates are so elevated in the UK and other countries with majority fair-skinned popula*ons.Related -in the introduc*on paragraph 3 while there may be addi*onal genes under selec*on in East Asian popula*ons that are relevant to skin cancer risk, a likely major reason for the reduced (gene*c) risk for KC is the absence or greatly reduced frequency of low pigmenta*on risk alleles enriched in European ancestry popula*ons e.g.• rs12203592, a pigmenta*on SNP influencing IRF4 func*on and a KC risk SNP has a low pigmenta*on T allele freq of ~14% in EUR pops, and < 0.01% in East Asian popula*ons.• rs1805007, one of the strongest red hair alleles and a strong risk allele for KC, has a risk T allele freq of ~7% in EUR but only 0.07% in East Asian popula*ons (source: GNOMAD).
Suggested ac*on: The importance of pigmenta*on SNPs/genes should be more fully and clearly discussed since it has bearing on their later results.5. Introduc*on page 3 -sentence 1. Expand briefly on the behaviour differences with respect to UV exposure and if relevant discuss how it may impact the differences in muta*onal burden observed.Suggested ac*on: Update abstract/text to address 6. Introduc*on page 3 -sentence 2. OCA2 isn#t a precursor to melanin; it is a channel transporter protein most likely involved in maintaining the pH of melanosomes, allowing Tyrsoinase#s produc*on of melanin -see PMID 32333855 Suggested ac*on: Update abstract/text to address 7. Table 1 a.It would be very helpful to merge the key muta*onal load columns from Sup table 12A into this table b.Why are the age listed as ranges e.g SG1 is 68-71?c.While the mean ages are roughly similar between the UK and SG samples, SG2 is young (28-31) which is relevant given the premise here is to report on ageing skin.The similarity of this sample to the others in terms of muta*on load etc should be discussed -e.g. are they an outlier?d.The most striking difference in this table is that all the SG samples have higher Fitzpatrick skin types.This makes it difficult to interpret what differences are due to country (and UV exposure) vs host factors (a more photoprotec*ve skin type) as they are confounded.The authors should discuss this and report on the regression of Fitzpatrick score onto muta*onal load -how much of the varia*on in muta*onal load is explicable by Fitzpatrick score?Suggested ac*on: update tables to address (a), and clarify table or footnote to address (b).(c) and (d) please update analysis/manuscript as required 8. Methods -mul*ple tes*ng.There are a large number of tests performed -mul*ple tes*ng correc*on should be applied and the impact in terms of the significance of results reported.
Suggested ac*on: List and report total number of tests and adjust the manuscript appropriately.9. Page 8 final paragraph; the jus*fica*on for removing the large TP53 clone from SG1 is not clearly presented.Why is it valid to exclude this sample?This should be made clear in the methods/results Suggested ac*on: Discuss/address in manuscript 10.Page 10, second last sentence.!-we would s*ll expect to observe at least 3 muta*ons at codons R248/R282 across all Singaporean samples" isn#t clear.If I understand it correctly the authors are sugges*ng 0 observa*ons out of 191 samples is unusual given they expected 3 out of 191 -is that especially unlikely given the events are rare?If that is the correct interpreta*on is this difference significant using an appropriate non-parametric test?If not clarify this result.
Suggested ac*on: Clarify in manuscript and include addi*onal results if relevant 11. Figure 4 indicates that the SG popula*on essen*ally lacks any of the major pigmenta*on risk alleles.This is interes*ng (but perhaps not surprising).In general the implica*ons of this aren#t really addressed in the manuscript, and relates to point 4 -that the likely main reason the SG samples have lower muta*on burdens than UK samples is they have greater skin pigmenta*on and that greater protec*on against UV damage.Specific comments about the SNP selec*on and figure 4: a) Please include chr and posi*on and present SNPs in their chr/bp order order b) Please include an appropriate reference popula*on allele freqs for EUR and EA (e.g. from GNOMAD ) as a column to help contextualise the results.This is relevant because many SNPs are rare (e.g.rs192481803 has a MAF ~ 0 in all pops which is likely why everyone is missing the risk allele) and/or many of the risk alleles related to pigmenta*on are essen*ally absent in East and South Asian popula*ons e.g.see point 4. c).Please check if the listed SNPs are in LD with each other at least in EUR pops (as their BCC/SCC associtons were iden*fied n EUR ancestry pops) e.g.rs4268748 is an intronic variant near MC1R, and is in LD with the rs1805007 func*onal missense allele (LD r2 0.23 in EUR) which is likely why it has previously been reported for BCC/SCC.As rs1805007 is wildtype in SG samples I am not sure it is relevant this SNP is heterozygous in some SG samples.The authors should consider removing SNPs in LD with other func*onal SNPs if they exist.d).Some SNP IDs are out of date e.g. the ID rs74664507 has been merged with rs11445081 e).Some lised SNPs have known func*ons/associa*ons assigned since the publica*on of the cited reviews (41 and 42) e.g.rs74664507 / rs11445081 is strongly associated with skin colour.The authors should survey more recent literature, or databases like Open targets gene*cs (e.g.hDps:// gene*cs.opentargets.org/variant/9_16913838_T_A)to update the figure to beDer reflect current informa*on.
Suggested ac*on: update/check SNP data entered into figure 4 and related to points 2 and 4 address implica*ons in the manuscript.
12. Methods -page 17.Please provide more details on the reference panel used to call muta*ons.Specifically, what was the ancestry of the reference popula*on used, and did it match each sequenced popula*on e.g. using an off-ancestry reference panel could be expected to increase the muta*on detec*on rate due to real SNP differences between panel and target.
Suggested ac*on: update methods and discuss if relevant.
13. Methods page 18 paragraph 1 final sentence + sup figure 3. Please report the significance value for the correla*on R = 0.33 between coverage and muta*ons detected.
Suggested ac*on: Report significance, and if this is significant the statement star*ng !no evidence" will need to be adjusted and this limita*on addressed in the discussion 14.Methods video model + model images.Are these simula*ons derived in any way from the data other than star*ng the muta*on burden 4 fold higher in video 2? E.g. were the model#s se{ngs for fitness derived from the reported dataset?If so please state so clearly.If not, please clarify what these videos are showing us about the data presented, or how they are otherwise providing nuance to the manuscript.
Suggested ac*on: Clarify methods for these videos and their relevance to the paper Reviewer #2: Remarks to the Author: This paper compares the accumulated muta*ons in human epidermis of individuals of two different ethnic backgrounds, UK and Singapore.The two ethnici*es have a different life*me risk of developing cSCC, independently of UV exposure.The comparison shows individuals from the high-risk ethnicity (UK) have a dis*nct paDern of muta*ons and muta*onal processes that overlap with cancer-driving cSCC genes, and individuals from the cSCC low risk group (Sing) in contrast have fewer muta*ons, fewer muta*ons in known SCC driver genes, and importantly, have an increase in the muta*ons that are linked to skin ageing but not skin cancer.
This study is important, novel and opens many new ques*ons that will impact scien*sts and clinicians working in the life sciences, not only cancer biology.The poten*al impact to medicine is clear.
The main limita*on is there are no valid hypotheses put forward to explain the difference in muta*onal landscape between UV exposed skin of different ethnic backgrounds, no experimental work to explain why different ethnic backgrounds develop/select different muta*ons.
Ques*ons a.Previous work shows sbs1 sig reflects the number of cell divisions, as it increases at a constant rate per cell division.The rate of cell division in human skin depends on sun exposure and age.Can the authors infer whether the rate of sbs1 muta*ons over *me is equal in both skin types, taking into account sbs7?It is interes*ng to test whether cell cycle regula*on underpins ethnic differences.
b.If there are differences in the cell division rate inferred from the sbs1 rate, adjusted by age and sbs7, for example faster cycling in UK skin, could these differences link to less *me for DNA repair, i.e more muta*ons?Have they compared DNA repair?c.You propose that notch1 can prevail and expand in Singapore skin due to less compe**on from other clones.You propose this is a possible explana*on for the difference in clone selec*on.This possibility could be confirmed in skin of young UK people, with fewer muta*ons.Does younger UK skin have high notch1, and will ageing reduce these in favour of the more UK-prevalent p53 clones?My hypothesis is that this will not be the case, and young skin will select R348W and R282W TP53.There are no alterna*ve hypotheses in the paper to explain why the accumula*on of specific muta*ons is not stochas*c across both skins.d.Is there any evidence ethnic loci risk scc variants promote selec*on of specific DNA muta*ons?Can this be tested?e.Many of the high risk loci are linked to pigmenta*on.Is the ra*onale that Singapore skin will make !more" pigment to protect beDer from UV?In reality, Singapore skin has a similar level of pigmenta*on compared to Uk skin, at least the difference is not large enough to explain the drama*c difference in incidence.Moreover, many Mediterranean countries have extremely high incidence of scc, compared to Singapore, and these differences cannot be explained by pigmenta*on level alone.Do they have any mechanis*c insight to what any of the variants do?Reviewer #3: Remarks to the Author: We know that skin cancer risk varies enormously across popula*ons around the world.But the ques*on is -why?King and colleagues present an analysis of the soma*c muta*onal landscapes of eyelid skin from subjects from Singapore (~17X lower skin cancer risk compared to UK, despite much higher UV exposure), comparing results from a previously reported UK cohort.The use of eyelid samples should control for differences in skin coverage with clothing, with the caveat that sunglasses can block UV exposure.Their studies reveal striking differences in muta*onal burden and signatures, selec*on, and muta*onal hotspots in expansions across the two cohorts, with the UK skin exhibi*ng greater features found in skin cancers.Results are very clearly presented and well wriDen.Data are suppor*ve of claims, which are not overstated.Methods are well described.One caveat of the study is that they are comparing two cohorts that differ both by gene*cs and by sun/UV exposure, making it more difficult to ascertain rela*ve contribu*ons to the muta*onal paDerns observed.S*ll, it is interes*ng that muta*onal burden and !pro-cancer" selec*ve paDerns are greater for the UK subjects despite substan*ally less overall UV exposure, perhaps rela*ng to greater gene*c risk (resul*ng in less protec*on by skin pigmenta*on).
In all, this study makes an important contribu*on to the field of soma*c muta*ons in normal *ssues and how they could contribute to cancer risks, including risks that vary across popula*ons.Major findings include: 1. Mean muta*onal burden is 4X higher in UK subjects than Singapore subjects, and the UK subjects had substan*ally greater UV muta*onal signatures and 8X as many double base subs*tu*ons characteris*c of UV exposure.CNAs were evident in 13% of UK eyelid specimen compared to just 1.0% of Singaporean.Thus, muta*onal burden correlates with skin cancer risk.
2. Perhaps one of the most interes*ng findings from this study is that fitness landscapes appear to differ for the two popula*ons, with NOTCH1 and NOTCH2 clonal expansions occupying less and TP53 expansions occupying more of the epidermis for the UK subjects than expected by the muta*onal burden.For the TP53 expansions, this was only true if an outlier clone was removed from one Singaporean subject (a reasonable exercise).In all, while about half of the epidermis in the Singapore group is occupied by clones with disrup*ve muta*ons in at least one of the 74 examined genes, this number is about 95% for the UK cohort.The distribu*on of clone size is smaller for the UK cohort, perhaps indica*ng clonal compe**on, which may also explain why NOTCH1/2 clones are significantly smaller (suppressed) in the UK skin samples (in contrast, TP53 clone size was similar across the cohorts).This compe**on is illustrated in 2 videos of a computa*onal model.3.They observed differences in the muta*onal spectra for TP53, FGFR3 and RAS genes, with the muta*ons most commonly observed in skin cancers (or cancers in general) almost exclusively restricted to the UK cohort.4.Not surprisingly, gene*c (germline) variants associated with pigmenta*on and SCC risk were differen*ally found in the UK cohort.Other risk alleles showed similar distribu*ons across the two cohorts.
5. Analyses of cSCC from South Korea (where risk of SCC is similar to that of Singapore) revealed that the UV muta*onal signature s*ll dominates.If we can extrapolate to SCCs from Singapore (not ideal, but not totally unreasonable), this suggests that despite lower muta*onal burden in the skin and a reduced UV signature in Southeast Asians, that the cancers are s*ll driven by UV mutagenesis.
In all, their data support their conclusion that UK skin shares more features with skin cancers (in terms of muta*onal burden and signatures, CNAs, genes under selec*on, and hotspot muta*ons) compared to Singaporean skin.
Ques*ons to address: 1. Do Brits tend to holiday in warmer places with more sun exposure, and perhaps experience sun burns?Periodic high exposure may be worse than regular moderate exposure, as the skin can adapt to the laDer.Could increased vaca*oning of Brits closer to the equator have contributed to the 2-3X rise in skin cancer incidence over the last few decades? 2. Were the UK donors of North European descent (Anglo-Saxon) and the Singapore donors of South Asian descent?It#s important to make this clear, as there has clearly been a lot of migra*on of humans.
3. From Supplementary table 3 it can be observed that the number of non-UV associated SBS1 and SBS5 muta*ons are more than twice in the UK cohort compared to the Singapore cohort -this should be pointed out in the manuscript.One interpreta*on of this observa*on is that the skin of UK subjects is more prone to muta*on-driven clonal expansions independent of UV mediated mutagenesis, such as UV induced selec*on; alterna*vely, these muta*ons could be due to indirect UV mutagenesis or other exposures.S2a, there seems to be a substan*al difference in dN/dS for missense and nonsense+splice TP53 muta*ons comparing Singapore and UK subjects, both of which are much higher in the Singapore subjects.Could this just reflect the overall lower muta*onal burden in the Singapore cohort, and thus when a TP53 muta*on is detected it is more likely to alter the protein?Some specula*on might be warranted, unless we are just missing something.5. Also, with regard to TP53, while it#s reasonable to assume that the large TP53 clone that is observed in donor SG1 is an outlier, it s*ll begs the ques*on if it might have some biological significance.The muta*on in ques*on appears to be P278S, which is not one of the most common in cancers.In addi*on, supplementary table 4 shows that two CNAs were iden*fied in donor SG4 (and none in the UK cohort).With regard to TP53 muta*on, missense muta*ons can be more deleterious than copy-number loss due to gain-of-func*on/dominant nega*ve effect.Combining the observa*ons above with their findings that aggressive muta*ons, such as those in codon R248, were not observed in Singapore donors, might suggest that more aggressive muta*ons are nega*vely selected in the skin of people from Singapore while clones with damaging but less aggressive muta*ons are not counteracted as efficiently and expand at levels similar to those seen in people for the UK.Is there evidence that this might be the case?AdmiDedly, ge{ng into this in the manuscript may be overly specula*ve.

For Fig
6.The authors state that !four of the five Singaporean donors and zero UK donors have introgression of the HYAL2 region at chromosome 3p21."They should briefly refer to the significance of this (a Neanderthal introgression that is associated with UV responses).
7. Is there no sequencing of SCCs from Singaporean subjects available?Such a comparison would have been more ideal.
8. One minor point, on page 3 the authors say !OCA2, which encodes a precursor to the UV-absorbing pigment melanin".From our understanding OCA2 encodes a protein that is necessary for the synthesis of melanin but is not a precursor of melanin.

Response to Reviewers: Somatic Mutations in Facial Skin from Countries of Contrasting Skin Cancer Risk
We would like to thank referees for the time taken to review our manuscript and are most grateful for their constructive comments which have significantly improved the revised text.Our responses to each point are in blue.

Referee expertise:
Referee #1: skin cancer risk, genetics Referee #2: skin cancer and ageing Referee #3: genomics, cell competition (signed report) Reviewer #1: The authors have systematically compared the mutational landscape of a large number of biopsies (191) taken from a sun-exposed site from 5 individuals from Singapore.They have contrasted these results with a similar dataset from 6 people from the UK.The samples from individuals from Singapore showed much lower rates of mutations and smaller/fewer mutant clones.
The overall mutational burden was also very low compared to the samples from the UK, and proportionally lower was the signal from UV-induced mutations even though Singapore experience 2-3 fold higher rates of exposure.WES of cSCC lesions from 19 Korean individuals (chosen as were similar to the Singapore population) was very similar to the mutation profiles of SCCs from elsewhere in the world suggesting that while the profiles of healthy skin may differ by population and region, SCC tumours follow a similar oncogenic process across populations.The discussion of the reasons for these differences is cursory.The clearest difference, and most likely cause, is that the 6 individuals from Singapore have a greater Fitzpatrick sore (3-4) vs the UK samples (2), and essentially lack any low pigmentation risk alleles (Figure 4) indicating they likely have a greater degree of skin pigmentation and tanning ability than UK samples.This is of interest both to better understand the causes of differences in ageing skin, but also to highlight that better adherence to UV protective behaviours in UK populations (and other fair-skinned populations) may reduce the mutation burden in the skin.
This study aims to compare the mutational landscape of mutant clones in normal facial skin from two populations of contrasting skin cancer risk.Our aim is to give insight into which features of ageing skin correlate with differences in cancer risk and which (for example, mutations in NOTCH1) do not.To do this, we stratify individuals by country.An individual's keratinocyte skin cancer risk is difficult to accurately quantify but population data on keratinocyte skin cancer incidence is readily available for both the UK and Singapore.KC incidence reflects a combination of behavioural, environmental and genetic factors.
We agree with the reviewer that differences in pigmentation and the propensity to tan play a major role in KC risk and we now emphasise this in the revised text.We have increased our SNP panel to 189 genomic sites associated with skin cancer risk, pigmentation and tan response to better explore the genetic differences between the UK and Singapore and gain more insight into cancer protective mechanisms.Specifically, we selected SNPs from the NHGRI-EBI GWAS and Open Targets databases and combined loci associated with keratinocyte carcinoma (entry EFO_0010176 -keratinocyte carcinoma), tan response (EFO_0004279 -suntan) and 'Ease of skin tanning' (UK Biobank: 1727).We added these to 16 positively selected loci in Singaporean genomes (Wu et al., 2019, Supplementary Table 8).
We have updated Figure 4 to highlight genotypes which differ most between donors of each country: We find at least 11 SNPs associated with pigmentation, particularly near to HERC2, MC1R and OCA2, that differ by donor country.OCA2 encodes a melanosomal anion channel that is essential for melanin synthesis 1 and is under positive selection in Singaporean genomes 2 .A regulatory element within an intron of HERC2 is known to inhibit the OCA2 promoter and both these genes have associations with hair, eye and skin pigmentation 3 .SNPs near IRF4 show associations with skin tanning, pigmentation and keratinocyte cancer risk (https://genetics.opentargets.org/gene/ENSG00000137265).rs12203592-T, which has an allele frequency of 0.18 in the UK but is absent in East Asians, lowers IRF4 levels and hence reduces expression of the pigmentation enzyme encoded by TYR 4 .As the reviewer points out, these genetic differences are reflected in the pigmentation and tanning phenotypes of UK and Singaporean skin, assessed by the Fitzpatrick score.However, differences in pigmentation may not be the only explanation for variation in KC risk between the two countries.In the SNPs that differ by country, at least ten seem to affect KC risk through mechanisms unrelated to pigmentation (Figure 4).For example, rs2111485 is associated with inflammatory and autoimmune diseases such as psoriasis, vitiligo, asthma and inflammatory bowel disease, whilst rs12129500 and rs2243289 affect the expression of IL6R and IL4, respectively.
Other SNPs that differ by country include: rs663743, shown to alter the expression of PPP1R14B, a gene highly expressed in multiple tumour tissues with functions related to cell growth, cell cycle regulation and apoptosis (PMID: 34858479); rs7335046, shown to alter UBAC2 expression, a gene involved in regulating the Wnt signalling pathway; and rs4409785, which alters SESN3 expression, a gene involved in cellular response to stress, e.g.reactive oxygen species.Space precludes a discussion of these genes and SNPs in the manuscript.
Finally, we find four of our Singaporean donors have introgression of a large LD block at chromosome 3p21.31.This region is under positive natural selection in East Asians and contains multiple genes associated with tumour suppression (HYAL1, RASSF1, TUSC2), cell motility (NAA80) and cellular response to UVB (HYAL2) (Ding et al., 2014).We do not propose that the differences observed in the somatic mutational landscape here are due to genes in this region as it is not present across all Singaporean donors in this study, however, we highlight this as an example of a mechanism unrelated to pigmentation which may act to suppress keratinocyte carcinogenesis.
In the revised text we now discuss these genetic differences at two points.
1.In the introduction, emphasising pigmentation differences, as suggested in comment 4 below -p.2: !The incidence of many cancers varies substantially worldwide, reflecting genetic differences between populations and their environmental exposures.This is well-illustrated by keratinocyte skin cancers (KC), where incidence varies 140-fold globally 5 .KC risk increases with an individual#s cumulative UV exposure which depends on age, outdoor work, sunbathing, use of tanning beds [6][7][8][9][10][11] and phenotypes such as freckles, low levels of skin pigmentation and poor tan response 12 .The most strongly associated KC risk loci are found near to pigmentation genes such as HERC2, OCA2, MC1R and IRF4, which have much greater allele frequencies in high-risk populations such as in the UK than in low-risk populations in East Asia 13 ."!To gain insight into differences in genetic background that may exist between donors we genotyped individuals for SNPs associated with KC risk (Methods, Fig. 4, Table S8).At 36 risk loci we found differences in donor genotype by country.Multiple SNPs were associated with genes linked to pigmentation, such as SLC45A2, IRF4, BNC2, OCA2 and HERC2 (https:// genetics.opentargets.org).For example, rs12203592 alters IRF4 levels and expression of the pigmentation enzyme encoded by TYR 4 .rs4778210, positively selected in Singaporeans 2 , is adjacent to OCA2, encoding a melanosomal anion channel that is essential for melanin synthesis 1 .These findings are consistent with the marked difference in UV mutational burden between countries.However, we also observed differences in non-pigmentation or tanning related SNPs.KC risk is strongly linked to immunosuppression and is also associated with inflammatory diseases 12 .rs2111485 is associated with inflammatory and autoimmune diseases such as psoriasis and inflammatory bowel disease, whilst rs12129500 and rs2243289 affect the expression of IL6R and IL4, respectively (https://genetics.opentargets.org).We found four of our Singaporean donors have a large introgression at chromosome 3p21.31 2 .This region is under positive natural selection in East Asians and contains the skin tumour suppressor gene RASSF1 14 and the UV induced genes HYAL1 and HYAL2 15 .Other KC risk SNPs were linked to genes with diverse functions."Comments 1. Abstract line one, and final sentence page 2. References 1-3, used to report the relative incidences of SCC across Singapore and the UK are from 2005 and earlier -are there no more recent surveys of skin cancer incidence in these countries?Figure 1a shows incidence rates up until 2015 suggesting different sources have been used then were cited (ref 1) in figure 1's footnote; Suggested action: check/update references Thank you for highlighting this point.We have now replaced the out-of-date references with those used to generate Figure 1a.We have also added a key reference underpinning the UK data 16 .2. The final sentence of the abstract offers no conclusion as to why skin from a low-risk country does not show multiple features convergent with cancer; the likely answer from the data is the degree of skin pigmentation and propensity to tan.This should be addressed.Also, see points 4 and 11.Suggested action: Update abstract/text to address Thank you for this suggestion.We have revised the abstract as below to reflect the likely significance of germline variation in tanning and pigmentation genes as suggested.However, without data on behaviour, we cannot make a conclusive statement on this point.The final two sentences of the abstract now read: !Ageing skin in a high incidence country has multiple features convergent with cancer, not found in a low-risk country.These differences may reflect germline variation in UV protective genes."Thank you.We have rewritten this section of the text, see above, and amended the references.4. Final sentence of introduction paragraph 1: In my opinion saying that some GWAS SNPs for KC are linked to skin pigmentation somewhat undersells this fact.For example, in ref 21 the majority of the most strongly associated loci (and thus represent the largest combination of effect sizes and minor allele frequency, and thus the proportion of trait variance explained) are all pigmentation SNPs.While other pathways are important in KC risk, pigmentation is by and far the strongest genetic risk for KC, which is why the rates are so elevated in the UK and other countries with majority fair-skinned populations.Related -in the introduction paragraph 3 while there may be additional genes under selection in East Asian populations that are relevant to skin cancer risk, a likely major reason for the reduced (genetic) risk for KC is the absence or greatly reduced frequency of low pigmentation risk alleles enriched in European ancestry populations e.g.• rs12203592, a pigmentation SNP influencing IRF4 function and a KC risk SNP has a low pigmentation T allele freq of ~14% in EUR pops, and < 0.01% in East Asian populations.• rs1805007, one of the strongest red hair alleles and a strong risk allele for KC, has a risk T allele freq of ~7% in EUR but only 0.07% in East Asian populations (source: GNOMAD).Suggested action: The importance of pigmentation SNPs/genes should be more fully and clearly discussed since it has bearing on their later results.
We have updated the text to emphasise the importance of pigmentation on KC incidence.The first paragraph of the introduction has been rewritten, see above.We also highlight the importance of tanning/ pigmentation alleles in Figure 4 and the associated discussion, see above.
5. Introduction page 3 -sentence 1. Expand briefly on the behaviour differences with respect to UV exposure and if relevant discuss how it may impact the differences in mutational burden observed.Suggested action: Update abstract/text to address Thank you, we have added detail on risk behaviours and now cite a Singaporean study on sun exposure.We have amended the text as follows: !KC risk increases with an individual#s cumulative UV exposure which depends on age, outdoor work, sunbathing, use of tanning beds [6][7][8][9][10][11] and phenotypes such as freckles, low levels of skin pigmentation and poor tan response 12 ."6. Introduction page 3 -sentence 2. OCA2 isn't a precursor to melanin; it is a channel transporter protein most likely involved in maintaining the pH of melanosomes, allowing Tyrsoinase's production of melaninsee PMID 32333855 Suggested action: Update abstract/text to address Thank-you.We have amended our description to: !rs4778210, positively selected in Singaporeans 2 , is adjacent to OCA2, encoding a melanosomal anion channel that is essential for melanin synthesis 1 ."7. Table 1 a.It would be very helpful to merge the key mutational load columns from Sup table 12A into this table b.Why are the age listed as ranges e.g SG1 is 68-71?c.While the mean ages are roughly similar between the UK and SG samples, SG2 is young (28-31)  which is relevant given the premise here is to report on ageing skin.The similarity of this sample to the others in terms of mutation load etc should be discussed -e.g. are they an outlier?d.The most striking difference in this table is that all the SG samples have higher Fitzpatrick skin types.This makes it difficult to interpret what differences are due to country (and UV exposure) vs host factors (a more photoprotective skin type) as they are confounded.The authors should discuss this and report on the regression of Fitzpatrick score onto mutational load -how much of the variation in mutational load is explicable by Fitzpatrick score?Suggested action: update tables to address (a), and clarify table or footnote to address (b).(c) and (d) please update analysis/manuscript as required a. Thank-you, genome-wide burden estimates have been added to Table 1.b. Donor ages are listed as ranges to help maintain donor anonymity.c.SG2 is ~20 years younger than the youngest UK donor, however, we have not excluded SG2 from the study because donor age is comparable by country (mean: SG = 62 years, UK = 68 years; median: SG = 70 years, UK = 70 years) and there is no evidence to suggest SG2 is an outlier.Although SG2 has the lowest mutation burden of all donors, it is comparable to that of SG3 (72-75 years) and not an outlier (Fig. 2a, Fig. S1a).
d.The Fitzpatrick score is an easily obtained subjective measure of skin UV response and has been very widely adopted, but its application to the global range of skin phenotypes is subject to debate 17 .While a regression of Fitzpatrick score by mutational burden is significant (Kruskal-Wallis test: p = 0.02), after post hoc testing with multiple testing correction, we find the main contribution to donor mutation burden is country.Furthermore, since Fitzpatrick score correlates with country, it cannot be considered an independent factor.We note that the effect of Fitzpatrick score on burden within Singapore is also non-significant (p = 0.4).

Response Figure 1: Fitzpatrick score and mutational burden
We agree that differences in UV responsiveness are likely to be of major importance in shaping the mutational landscape but would submit that including the additional genotypic information in Figure 4 is more informative than Fitzpatrick score, which we continue to provide in Table 1 for interested readers.8. Methods -multiple testing.There are a large number of tests performed -multiple testing correction should be applied and the impact in terms of the significance of results reported.
Suggested action: List and report total number of tests and adjust the manuscript appropriately.
As requested, we now list all tests with their reported significance.Duplicate mentions of tests in both text and figures have been removed from the main text, tests are now only reported in the figure legends with p-values and test names.The tests performed are listed below: Fig. 2a.Estimates of genome-wide mutation burden per donor by country (Student's t-test: p = 8.8x10 -3 ).
Here we performed a single t-test and therefore do not require multiple testing correction.Fig. 2c.We used unsupervised clustering to test if UV signature distributions were different between countries (p < 1x10 -5 ).We do not require multiple testing correction.

Fig S1b.
Propor*on of SBS1 muta*ons/mm 2 per donor by country, t-test: p = 0.01.Here we performed a single t-test and therefore do not require multiple testing correction.

Fig S1c.
Propor*on of SBS5 muta*ons/mm 2 per donor by country, t-test: p = 2.7 x10 -4 .Here we performed a single t-test and therefore do not require multiple testing correction.

Fig S1d.
Propor*on of SBS assigned to each signature, split by muta*ons above and below median variant allele frequency (VAF) for each country, Pearson's chi-square UK: p < 2.2x10 -16 ; SG: p = 1.3x10 -14 .No mul*ple test correc*on required for the pairwise comparisons within each country.
FigS1f.Count of insertions and deletions per mm 2 in each donor by country (UK mean = 0.72 indels/mm 2 , SG mean = 0.37 indels/mm 2 , Welch t-test: p = 0.07).We use Welch's t-test (for unequal variance), therefore do not require multiple testing correction.
Supplementary Figure 2: a Ratio of observed/expected non-synonymous mutations for positively selected genes by country (q < 0.01).No synonymous mutations were detected in Singaporean skin for TP53 and AJUBA, leading to high dN/dS ratios.Line drawn at y = 1.b Sizes of all clones with protein-altering mutations in the top four positively selected genes, by country (samples with known CNA removed), t-test adjusted by Bonferroni multiple test correction: p = 6x10 -4 (NOTCH1), 1.5x10 -2 (NOTCH2), 0.68 (FAT1) and 1.0 (TP53).c Distributions of mutations across codons in TP53.The most frequently mutated codon in cancer, R248 (shown red), is the most common codon change in UK skin but is absent in Singaporean skin.
Fig. S2b Distribution of clone size for those carrying protein-altering mutations in the top four positively selected genes, by country (samples with known CNA removed).We have added a Bonferroni multiple test correction and reported adjusted p-values, including non-significant values.p.9 Difference in 'hotspot' TP53 codon distribution (bootstrap test) by country.We tested how likely it was to find a non-zero number of hotspot TP53 mutations in Singaporean skin, p < 0.001.

Methods, p.16.
'There was no evidence to suggest a difference in mean mutation burden between the eyebrows and eyelids of donors (Welch's t-test: p = 0.14).' Welch's t-test, two groups (mutation burden in eyelids and brows per donor), does not require multiple testing correction.9. Page 8 final paragraph; the justification for removing the large TP53 clone from SG1 is not clearly presented.Why is it valid to exclude this sample?This should be made clear in the methods/results of zero (green = bootstrapped estimation from each sub-sampling, red = Singaporean observation) or the 95% upper confidence interval limit (0.0156).
We describe this test in the Methods as follows: !We applied non-parametric bootstrapping to estimate the significance of observing 0 mutations at R248 and R282 TP53 codons in Singaporean skin, given the decreased burden, decreased TP53 selection and decreased UV signature compared to UK skin.The non-parametric bootstrap statistical method is appropriate for this data because it does not make assumptions on distribution 18 .The method samples from a given distribution (here, UK skin), maintaining variability and other observable parameters.For a bootstrap test we simulated expected values for mutation distribution in Singaporean skin given: (i) UK mutation distribution per donor and per sample of all, C>T/CC>TT, TP53 and R248W/R282W TP53 mutations and (ii) proportions of all, C>T/CC>TT and TP53 mutations between UK and SG.If the null hypothesis that the similarity of UK (adjusted) and SG distributions is true, we would expect that the simulated "UKadjusted#$values would correspond to the SG observations, on average.We applied the "rule of three#$approach, often used in clinical trials, 19 to estimate the robustness of zero outcomes in Singaporean skin.This approach suggests the upper limit of a 95% confidence interval is 3/(n+1), where n is the number of SG samples.Here, this value is 3/ (191+1)= 0.0156.We ran 1000 simulations to obtain estimation for an event of zero R248/R282 TP53 codon mutation counts in SG samples.Of 1,000 simulations, none produced a value lower than 0.0156.The bootstrap simulation shows that (i) an expected mutation frequency at R248/ R282 TP53 codons is around 3 mutations across all SG samples and (ii) the UK and SG distributions are significantly different (p < 0.001)." We now state in the manuscript p.9: !Even after adjusting for C>T/CC>TT burden across TP53 by country, we would still expect to observe approximately 3 mutations at codons R248/R282 across all Singaporean samples (Methods, Bootstrap test, p<0.001)".
11. Figure 4 indicates that the SG population essentially lacks any of the major pigmentation risk alleles.This is interesting (but perhaps not surprising).In general the implications of this aren't really addressed in the manuscript, and relates to point 4 -that the likely main reason the SG samples have lower mutation burdens than UK samples is they have greater skin pigmentation and that greater protection against UV damage.Specific comments about the SNP selection and figure 4 4 and related to points 2 and 4 address implications in the manuscript.
Thank-you, all revised -please see Figure 4 and the SNP analysis above.We present SNPs in chr/bp order with East Asian and GBR allele frequencies (1000 Genomes Project Phase 3).We note that allele frequencies for SouthEast Asia or Singapore are unavailable.East Asia allele frequencies are calculated from 504 individuals (~300 Chinese, ~100 Japanese and ~100 Vietnamese).GBR allele frequencies are taken from 91 individuals across England and Scotland.We have kept all SNPs that we genotyped across the HYAL region in Figure 4 but have highlighted in the legend that they form the same LD block.As the reviewer recommended, we have updated associated genes and phenotypes using the Open Targets Genetics database.
12. Methods -page 17.Please provide more details on the reference panel used to call mutations.Specifically, what was the ancestry of the reference population used, and did it match each sequenced population e.g. using an off-ancestry reference panel could be expected to increase the mutation detection rate due to real SNP differences between panel and target.Suggested action: update methods and discuss if relevant.
The reference panel for deepSNV is used to construct a base by base model of the background error rate of the sequencing platform [20][21][22] .This allows the identification of loci that have a statistical excess of mismatched base calls, enabling the reliable detection of low VAF mutations.To generate the large reference panel, 51 deeply sequenced samples of muscle or fat were used all from donors of the UK.Variants are called if the VAF exceeds the error rate at the base concerned.
The reviewer raises the issue of germline variants.These would be present in all samples (n = 34-45) in a given donor at a high VAF, allowing us to exclude any such variants as germline.We have now explained these points in more detail in the Methods.13.Methods page 18 paragraph 1 final sentence + sup figure 3. Please report the significance value for the correlation R = 0.33 between coverage and mutations detected.Suggested action: Report significance, and if this is significant the statement starting "no evidence" will need to be adjusted and this limitation addressed in the discussion Thank you.The significance of the correlation for the combined SG and UK data shown in the original Fig. S3 is high (p is very small).We note, a statistically significant correlation coefficient does not mean there is a strong association.It tests the null hypothesis that there is no relationship but provides no information about the strength of the relationship or its importance.The relationship between two variables is generally considered strong when r is greater than 0.7 23 .
We have, however, incorrectly calculated the correlation in the original Fig. S3 due to two statistical pitfalls in combining two different populations (UK and Singapore) 24 , these are: (i) when data has two subgroups, within each of which there is no correlation (ii) when variability in values on the x-axis changes with values on the y-axis (and we note that variance grows with coverage here) We do indeed have both (i) and (ii) in our original approach.Plotting within each group (see amended Fig. S3 below), we show there is no significant correlation: Supplementary Figure 3: Correlation between mean quality sequencing coverage per sample and the number of mutations detected, by country.Mean coverage was calculated after removing off-target reads, duplicates and those with mapping quality of 25 or less and base quality of 30 or less.Samples of neither country show a correlation between coverage and mutation counts (SG: R = 0.087, p = 0.089; UK: R = 0.092, p = 0.058).
We are most grateful to the reviewer for leading us to correct this error.
The aim of Fig. S3 is to show that even at the ranges where the coverage is equal (700 to 1000x), more mutations are called in UK samples.The Shearwater algorithm (deepSNV package) does not require checking of any correlation beyond ensuring a minimum coverage of more than 500x.Above this level, there is enough statistical power for an unbiased variant call, so our correlation is perhaps unnecessary.
However, for completeness, we have replaced the misleading original Fig. S3 with separate panels for each data set.This demonstrates that when samples are grouped appropriately, there is no correlation between coverage and mutation calls and, most importantly, that coverage exceeds 500x in both groups.
We also now state in the Methods p.15: !We filtered sequencing reads for quality by mapping and base quality.Mean quality sequencing depth of coverage exceeded 500x per sample in both Singapore and UK samples and the number of mutations detected did not correlate with sequencing depth in either group (Fig. S3)." 14. Methods video model + model images.Are these simulations derived in any way from the data other than starting the mutation burden 4 fold higher in video 2? E.g. were the model's settings for fitness derived from the reported dataset?If so please state so clearly.If not, please clarify what these videos are showing us about the data presented, or how they are otherwise providing nuance to the manuscript.Suggested action: Clarify methods for these videos and their relevance to the paper The videos aim to illustrate the effect of mutation burden on clone size distributions.We show that, with a 4-fold higher mutation burden, clones are more restricted in their growth, leading to a smaller clone size distribution (as observed in the UK skin samples).It may be counter-intuitive for some readers that clone sizes are larger in Singaporean skin compared to the UK, so the aim of the video is to aid understanding by illustrating our hypothesis: high competition in UK skin leads to restriction of clone growth.
We have amended the Methods to now read: "We illustrated mutant clone growth in a sparsely (Video 1) and densely (Video 2) mutated environment to simulate clonal competition in skin from Singapore and the UK, respectively.In Video 1, mutant clones of two arbitrary levels of fitness divide against a wild-type background of lower fitness until approximately 50% of the space is mutant.The starting mutation burden of Video 2 is 4-fold higher than for Video 1, with cells dividing for the same number of divisions.
Final mutant clone sizes are larger in a sparsely mutated environment (Video 1) compared to in a densely mutated environment (Video 2), reflecting the different clone size distributions we observe by country (Fig. 3f)."Reviewer #2: This paper compares the accumulated mutations in human epidermis of individuals of two different ethnic backgrounds, UK and Singapore.The two ethnicities have a different lifetime risk of developing cSCC, independently of UV exposure.The comparison shows individuals from the high-risk ethnicity (UK) have a distinct pattern of mutations and mutational processes that overlap with cancer-driving cSCC genes, and individuals from the cSCC low risk group (Sing) in contrast have fewer mutations, fewer mutations in known SCC driver genes, and importantly, have an increase in the mutations that are linked to skin ageing but not skin cancer.This study is important, novel and opens many new questions that will impact scientists and clinicians working in the life sciences, not only cancer biology.The potential impact to medicine is clear.The main limitation is there are no valid hypotheses put forward to explain the difference in mutational landscape between UV exposed skin of different ethnic backgrounds, no experimental work to explain why different ethnic backgrounds develop/select different mutations.Questions a.Previous work shows sbs1 sig reflects the number of cell divisions, as it increases at a constant rate per cell division.The rate of cell division in human skin depends on sun exposure and age.Can the authors infer whether the rate of sbs1 mutations over time is equal in both skin types, taking into account sbs7?It is interesting to test whether cell cycle regulation underpins ethnic differences.
SBS1 has indeed been argued to reflect the number of cell divisions in earlier work, though the progressive increase of SBS1 mutations with age in post-mitotic cells such as neurons and smooth muscle cells challenges this assumption 25,26 .
The number of SBS1 mutations/mm 2 of skin is significantly higher in the UK than in Singapore (Response Figure 4).Dividing by donor age gives mean rates of SBS1 mutation of 0.0006/year/mm 2 and 0.010/year/mm 2 in Singapore and the UK respectively.
One way to explore the impact of cell division on signatures is to consider clone size.We examined whether signatures were different in clones smaller or larger than the median VAF, finding this was the case in both countries (Pearson's chi-square UK: p < 2.2x10 -16 ; SG: p = 1.3x10 -14 ).The proportion of SBS1 mutations is higher in larger clones than in smaller clones in both countries.
Response Figure 5: Proportion of SBS assigned to each signature, split by mutations above and below median VAF for each country.
In terms of adjusting for SBS7 we find that, across both countries, SBS1 (and SBS5) are positively correlated with SBS7 (Pearson's R = 0.90 and 0.91, respectively).This might reflect clonal expansion driven by UV light.
We thank the reviewer for suggesting this line of enquiry.The increased proportion of SBS1 in larger clones suggests SBS1 may increase with increased cell division in larger clones, though these may also be older in chronological time.The uncertainty over the link between cell division and SBS1 perhaps argues for a cautious interpretation of this data.We present this result in the text p.6 as follows: "The number of both SBS1 and SBS5 mutations/mm 2 was increased in UK skin compared with Singaporean skin (Fig. S1b-c).In epithelial cells, the proportion of SBS1 mutation has previously been correlated with the cumulative number of cell divisions in a tissue.Here, in both countries, we find increased SBS1 in larger clones (Fig. S1d).These differences may reflect the effects of UV, which increases the rate of proliferation and proportion of dividing cells in the epidermis and can drive mutant clone expansion 27 " b.If there are differences in the cell division rate inferred from the sbs1 rate, adjusted by age and sbs7, for example faster cycling in UK skin, could these differences link to less time for DNA repair, i.e more mutations?Have they compared DNA repair?
The burden of SBS1 and SBS5 mutations correlate with SBS7 burden (Pearson's R = 0.90 and 0.91, respectively).As discussed above, UV driven clonal expansion may contribute to the increased number of SBS1 mutations in UK skin.
We are limited in our ability to compare DNA repair beyond analysis of mutational signatures.It is hypothesised that SBS7a and SBS7b are each the consequence of the two major known UV photoproducts: cyclobutane pyrimidine dimers and 6-4 photoproducts, whilst SBS7c and SBS7d may be the consequence of translesion DNA synthesis by error-prone polymerases inserting T or G respectively, rather than A, opposite UV induced photodimers.SBS7d, is increased in UK skin as expected with increased UV exposure.
C>T mutations caused by UV damage are more frequently observed on the untranscribed strand of genes, due to the repair of lesions on the transcribed strand by transcription-coupled nucleotide excision repair 28 .However, C>T mutations caused by SBS1 do not exhibit strong transcriptional strand bias 29 .There is a stronger transcriptional strand bias of C>T mutations in skin from donors of the UK compared to Singapore, consistent with increased UV damage in the UK (transcribed/untranscribed = 0.829 and 0.896, respectively).This is consistent with less repair of UV damage in UK skin, compared to Singapore.c.You propose that notch1 can prevail and expand in Singapore skin due to less competition from other clones.You propose this is a possible explanation for the difference in clone selection.This possibility could be confirmed in skin of young UK people, with fewer mutations.Does younger UK skin have high notch1, and will ageing reduce these in favour of the more UK-prevalent p53 clones?My hypothesis is that this will not be the case, and young skin will select R348W and R282W TP53.There are no alternative hypotheses in the paper to explain why the accumulation of specific mutations is not stochastic across both skins.
Unfortunately, as blepharoplasty surgery is performed on ageing skin, it is not possible for us to sample facial skin from young UK donors.However, we have attempted to address the reviewer's question through analysis of previously published samples of skin from UK donors that have been taken from less sun-exposed sites 21 .
The Singaporean samples carry 1-28 mutations per sample.We identified 7 previously published UK donors with less than 30 mutations per sample in other body sites.
Response Figure 6: The number of mutations detected through previously published targeted sequencing (74 genes) across skin samples from 7 UK donors and three body sites (Leg, Trunk and Ab = Abdomen).
In total, 1,658 clones are detected across these donors (ages 26-68) in Abdominal, Trunk and Leg skin which we can compare with the 1,839 clones detected across the Singaporean donor facial skin.
Response Figure 7: a Mutational signature decomposition for three UK donors with >200 single-base substitutions using SigProfiler.b Ratio of observed/expected non-synonymous mutations for positively selected genes by country (q < 0.01), using dNdScv.
Only three of the UK donors had enough mutations (>200) for reliable signature calling.Despite similar mutation burdens to SG donors, the majority of mutations in the additional UK samples were caused by UV damage.This likely reflects germline differences in SNPs related to UV response, discussed below in point d, between the UK and Singaporean populations.
Four genes were identified as positively selected in the low burden UK skin (NOTCH1, NOTCH2, FAT1 and TP53).
Response Figure 8: a The number of mutations of each consequence for positively selected genes (dNdScv: q < 0.01) between SG facial skin and low-burden UK body skin.b Plot of non-synonymous mutations per gene in SG facial vs. UK body samples.Positively selected genes (purple) are labelled.Red indicates positively selected genes with a significant (p < 0.001) difference in dN/dS ratio by country, after accounting for global differences.
Mutant NOTCH1 is more strongly selected in Singaporean skin compared to UK skin of similar burden, while mutant FAT1 is more strongly selected in UK skin (however, we note that mutant FAT1 was reported as depleted in facial skin in pan-body analysis) 21 .

29
Response Figure 9: Sizes of all clones with protein-altering mutations in positively selected genes, by country (SG facial skin vs. UK body skin).
Amongst the selected mutant genes, clone sizes in the UK low burden skin were consistently smaller than in the UK.There are multiple potential explanations for this, including body site (Singaporean skin is exposed to UV regularly whereas UK skin is in intermittently sun exposed sites) and donor age (UK donors being younger than Singaporean).
Across the 225 UK low-burden samples, we find two TP53 R248W mutations (36 year old abdominal skin and 30 year old trunk) and one occurrence of TP53 R282W (68 year old abdominal skin).
Unfortunately, variation in body site, age and genetic background confound comparison between UK and Singapore and we would submit that discussion of this analysis in the revised text is overly speculative for these reasons.d.Is there any evidence ethnic loci risk scc variants promote selection of specific DNA mutations?Can this be tested?e.Many of the high risk loci are linked to pigmentation.Is the rationale that Singapore skin will make "more" pigment to protect better from UV?In reality, Singapore skin has a similar level of pigmentation compared to Uk skin, at least the difference is not large enough to explain the dramatic difference in incidence.Moreover, many Mediterranean countries have extremely high incidence of scc, compared to Singapore, and these differences cannot be explained by pigmentation level alone.Do they have any mechanistic insight to what any of the variants do?Points d and e are both interesting and important issues related to donor genotype.We selected SNPs from the NHGRI-EBI GWAS and Open Targets databases and combined loci associated with keratinocyte carcinoma (entry EFO_0010176 -keratinocyte carcinoma), tan response (EFO_0004279 -suntan) and 'Ease of skin tanning' (UK Biobank: 1727).We added these to 16 positively selected loci in Singaporean genomes (Wu et al., 2019, Supplementary Table 8).
We have updated Figure 4 to highlight genotypes which differ most between donors of each country: As the reviewer points out, many of the SNPs that differ by country are linked to pigmentation or tanning.We find at least 11 SNPs associated with pigmentation, particularly near to HERC2, MC1R and OCA2, that differ by donor country.In terms of mechanistic insights, OCA2 encodes a melanosomal anion channel that is essential for melanin synthesis 1 and is under positive selection in Singaporean genomes 2 .SNPs near IRF4 show associations with skin tanning, pigmentation and keratinocyte cancer risk (https:// genetics.opentargets.org/gene/ENSG00000137265).rs12203592-T, which has an allele frequency of 0.18 in the UK but is absent in East Asians, lowers IRF4 levels and hence reduces expression of the pigmentation enzyme encoded by TYR 4 .
The reviewer also highlights that the differences in cancer incidence 'cannot be explained by pigmentation level alone'.At least 10 of the risk SNPs seem to have mechanisms independent of pigmentation.For example, rs2111485 is associated with inflammatory and autoimmune diseases, whilst rs12129500 and rs2243289 affect the expression of IL6R and IL4, respectively.rs663743, alters the expression of PPP1R14B, a gene highly expressed in multiple tumour tissues with functions related to cell growth, cell cycle regulation and apoptosis.rs7335046 alters UBAC2 expression, a gene involved in regulating the Wnt signalling pathway, and rs4409785, which alters SESN3 expression, a gene involved in cellular response to stress.Space precludes a discussion of these genes and SNPs however.
Finally, we find four of our Singaporean donors have introgression of a large LD block at chromosome 3p21.31.This region is under positive natural selection in East Asians and contains multiple genes associated with tumour suppression (HYAL1, RASSF1, TUSC2), cell motility (NAA80) and cellular response to UVB (HYAL2) (Ding et al., 2014).
Regarding point d, we do not have the statistical power to determine if these mutations alter the selection of specific mutant genes.
Addressing point e, we have added the following to the text along with the revised Figure 4: !To gain insight into differences in genetic background that may exist between donors we genotyped individuals for SNPs associated with altered risk for KC (Methods, Fig. 4, Table S8).
At 36 risk loci we found differences in donor genotype by country.Multiple SNPs were associated with genes linked to pigmentation, such as SLC45A2, IRF4, BNC2, OCA2 and HERC2 (https://genetics.opentargets.org).For example, rs12203592 alters IRF4 levels and expression of the pigmentation enzyme encoded by TYR 4 .rs4778210, positively selected in Singaporeans 2 , is adjacent to OCA2, encoding a melanosomal anion channel that is essential for melanin synthesis 1 .These findings are consistent with the marked difference in UV mutational burden between countries.However, we also observed differences in non-pigmentation or tanning related SNPs.KC risk is strongly linked to immunosuppression and is also associated with inflammatory diseases 12 .rs2111485 is associated with inflammatory and autoimmune diseases such as psoriasis and inflammatory bowel disease, whilst rs12129500 and rs2243289 affect the expression of IL6R and IL4, respectively (https://genetics.opentargets.org).We found four of our Singaporean donors have a large introgression at chromosome 3p21.31 2 .This region is under positive natural selection in East Asians and contains the skin tumour suppressor gene RASSF1 14 and the UV induced genes HYAL1 and HYAL2) 15 .Other KC risk SNPs were linked to genes with diverse functions."

Reviewer #3:
We know that skin cancer risk varies enormously across populations around the world.But the question is -why?King and colleagues present an analysis of the somatic mutational landscapes of eyelid skin from subjects from Singapore (~17X lower skin cancer risk compared to UK, despite much higher UV exposure), comparing results from a previously reported UK cohort.The use of eyelid samples should control for differences in skin coverage with clothing, with the caveat that sunglasses can block UV exposure.Their studies reveal striking differences in mutational burden and signatures, selection, and mutational hotspots in expansions across the two cohorts, with the UK skin exhibiting greater features found in skin cancers.Results are very clearly presented and well written.Data are supportive of claims, which are not overstated.Methods are well described.One caveat of the study is that they are comparing two cohorts that differ both by genetics and by sun/UV exposure, making it more difficult to ascertain relative contributions to the mutational patterns observed.Still, it is interesting that mutational burden and "pro-cancer" selective patterns are greater for the UK subjects despite substantially less overall UV exposure, perhaps relating to greater genetic risk (resulting in less protection by skin pigmentation).
In all, this study makes an important contribution to the field of somatic mutations in normal tissues and how they could contribute to cancer risks, including risks that vary across populations.Major findings include: 1. Mean mutational burden is 4X higher in UK subjects than Singapore subjects, and the UK subjects had substantially greater UV mutational signatures and 8X as many double base substitutions characteristic of UV exposure.CNAs were evident in 13% of UK eyelid specimen compared to just 1.0% of Singaporean.Thus, mutational burden correlates with skin cancer risk.2. Perhaps one of the most interesting findings from this study is that fitness landscapes appear to differ for the two populations, with NOTCH1 and NOTCH2 clonal expansions occupying less and TP53 expansions occupying more of the epidermis for the UK subjects than expected by the mutational burden.For the TP53 expansions, this was only true if an outlier clone was removed from one Singaporean subject (a reasonable exercise).In all, while about half of the epidermis in the Singapore group is occupied by clones with disruptive mutations in at least one of the 74 examined genes, this number is about 95% for the UK cohort.The distribution of clone size is smaller for the UK cohort, perhaps indicating clonal competition, which may also explain why NOTCH1/2 clones are significantly smaller (suppressed) in the UK skin samples (in contrast, TP53 clone size was similar across the cohorts).This competition is illustrated in 2 videos of a computational model.3.They observed differences in the mutational spectra for TP53, FGFR3 and RAS genes, with the mutations most commonly observed in skin cancers (or cancers in general) almost exclusively restricted to the UK cohort.4.Not surprisingly, genetic (germline) variants associated with pigmentation and SCC risk were differentially found in the UK cohort.Other risk alleles showed similar distributions across the two cohorts.5. Analyses of cSCC from South Korea (where risk of SCC is similar to that of Singapore) revealed that the UV mutational signature still dominates.If we can extrapolate to SCCs from Singapore (not ideal, but not totally unreasonable), this suggests that despite lower mutational burden in the skin and a reduced UV signature in Southeast Asians, that the cancers are still driven by UV mutagenesis.In all, their data support their conclusion that UK skin shares more features with skin cancers (in terms of mutational burden and signatures, CNAs, genes under selection, and hotspot mutations) compared to Singaporean skin.Questions to address: 1. Do Brits tend to holiday in warmer places with more sun exposure, and perhaps experience sun burns?Periodic high exposure may be worse than regular moderate exposure, as the skin can adapt to the latter.Could increased vacationing of Brits closer to the equator have contributed to the 2-3X rise in skin cancer incidence over the last few decades?
The rise in overseas holidays to the Mediterranean region is indeed one of the behavioural factors that is suspected to be responsible for the increasing incidence of keratinocyte cancer, but multiple other behaviours are also likely to contribute, such as the use of indoor tanning facilities and lack of awareness and/or adoption of 'safe sun' practices in previous decades [30][31][32] .Keratinocyte cancer risk reflects cumulative UV exposure over a lifetime as well as episodes of sunburn 8,9 .The accurate measurement of UV exposure over such long periods makes research in this area very challenging and so we have chosen not to speculate on the multiple possible behavioural differences that may lie behind the different mutational landscapes in the UK and Singapore.2. Were the UK donors of North European descent (Anglo-Saxon) and the Singapore donors of South Asian descent?It's important to make this clear, as there has clearly been a lot of migration of humans.
All Singaporean donors self-reported their ethnicity as either Chinese or Malay.We report, where available, the Fitzpatrick score of each donor in Table 1 as a phenotypic measure of skin type and taking this, and our analysis of donor genotype, we believe our donors are typical of the ancestries of a small sample of each country's population.3. From Supplementary table 3 it can be observed that the number of non-UV associated SBS1 and SBS5 mutations are more than twice in the UK cohort compared to the Singapore cohort -this should be pointed out in the manuscript.One interpretation of this observation is that the skin of UK subjects is more prone to mutation-driven clonal expansions independent of UV mediated mutagenesis, such as UV induced selection; alternatively, these mutations could be due to indirect UV mutagenesis or other exposures.
Both SBS1 and SBS5 are associated with tissue ageing and are ubiquitous amongst normal tissues and cancers in humans.SBS1 is caused by the endogenous deamination of 5-methyl-cytosine to thymine and rates of SBS1 acquisition in different cell types correlate with estimated rates of stem cell division 29 .After adjusting for the area sampled in each donor, there are indeed significantly more SBS1 and SBS5 mutations in UK skin than in Singaporean skin.
Response Figure 10: Proportion of SBS1 and SBS5 mutations/mm 2 per donor by country.This is an interesting finding.The reviewer's hypothesis that SBS1 and SBS5 mutations are more likely to be expanded in UK skin, through mutation and/or UV-driven clonal expansion is plausible.In both the UK and Singapore, we find proportions of SBS signatures vary with clone size, see figure below.In both countries, SBS7a was over-represented in mutations below the median VAF and SBS1 was over-represented in mutations above the median VAF.It is possible that larger clones have undergone more cell division, leading to an increase of SBS1 mutations.
We have added the plots above to Supplementary Figure 1 and update our discussion p.6 in the manuscript to now read: "The number of both SBS1 and SBS5 mutations/mm 2 was increased in UK skin compared with Singaporean skin (Fig. S1b-c).In epithelial cells, the proportion of SBS1 mutation has previously been correlated with the cumulative number of cell divisions in a tissue.Here, in both countries, we find increased SBS1 in larger clones (Fig. S1d).These differences may reflect the effects of UV, which increases the rate of proliferation and proportion of dividing cells in the epidermis and can drive mutant clone expansion 27 " 4. For Fig S2a, there seems to be a substantial difference in dN/dS for missense and nonsense+splice TP53 mutations comparing Singapore and UK subjects, both of which are much higher in the Singapore subjects.Could this just reflect the overall lower mutational burden in the Singapore cohort, and thus when a TP53 mutation is detected it is more likely to alter the protein?Some speculation might be warranted, unless we are just missing something.
No synonymous mutations were detected in TP53 in Singaporean skin, leading to the high dN/dS ratios (Supplementary Table 5).This reflects a relative depletion of synonymous mutations across all genes in Singaporean skin compared with the UK (Wilcoxon: p = 0.0043).We may speculate that synonymous clones in the skin are likely to follow neutral drift and are far less likely than positively selected mutants to reach the lower limit of detection 18 , most synonymous clones that are detected are passengers in mutant clones under selection.If this is the case it would explain why there are more synonymous mutations in "We found four of our Singaporean donors have a large introgression at chromosome 3p21.31 2 .This region is under positive natural selection in East Asians and contains the skin tumour suppressor gene RASSF1 14 and the UV induced genes HYAL1 and HYAL2 15 ."7. Is there no sequencing of SCCs from Singaporean subjects available?Such a comparison would have been more ideal.
We agree, but unfortunately DNA sequencing of skin cancers in non-European ancestry populations is sparse and, at time of writing, no keratinocyte skin cancer sequencing is available for patients from Singapore.8.One minor point, on page 3 the authors say "OCA2, which encodes a precursor to the UV-absorbing pigment melanin".From our understanding OCA2 encodes a protein that is necessary for the synthesis of melanin but is not a precursor of melanin.
Thank-you for correcting us on this point.We have changed our description to: !rs4778210, positively selected in Singaporeans 2 , is adjacent to OCA2, encoding a melanosomal anion channel that is essential for melanin synthesis 1 ."Signed: James DeGregori and Marco De Dominici For all corresponding authors listed on the manuscript, please follow the instruc*ons in the link below to link your ORCID to your account on our MTS before submi{ng the final version of the manuscript.If you do not yet have an ORCID you will be able to create one in minutes.hDps://www.springernature.com/gp/researchers/orcid/orcid-for-nature-researchIMPORTANT: All authors iden*fied as "corresponding author#$on the manuscript must follow these instruc*ons.Non-corresponding authors do not have to link their ORCIDs but are encouraged to do so.Please note that it will not be possible to add/modify ORCIDs at proof.Thus, if they wish to have their ORCID added to the paper they must also follow the above procedure prior to acceptance.
To support ORCID's aims, we only allow a single ORCID iden*fier to be aDached to one account.If you have any issues aDaching an ORCID iden*fier to your MTS account, please contact the <a href="hDp:// pla€ormsupport.nature.com/">Pla€ormSupport Helpdesk</a>.I am delighted to say that your manuscript "Soma*c muta*ons in facial skin from countries of contras*ng skin cancer risk" has been accepted for publica*on in an upcoming issue of Nature Gene*cs.
Over the next few weeks, your paper will be copyedited to ensure that it conforms to Nature Gene*cs style.Once your paper is typeset, you will receive an email with a link to choose the appropriate publishing op*ons for your paper and our Author Services team will be in touch regarding any addi*onal informa*on that may be required.
A‚er the grant of rights is completed, you will receive a link to your electronic proof via email with a request to make any correc*ons within 48 hours.If, when you receive your proof, you cannot meet this deadline, please inform us at rjsproduc*on@springernature.com immediately.
You will not receive your proofs un*l the publishing agreement has been received through our system.
Due to the importance of these deadlines, we ask that you please let us know now whether you will be difficult to contact over the next month.If this is the case, we ask you provide us with the contact informa*on (email, phone and fax) of someone who will be able to check the proofs on your behalf, and who will be available to address any last-minute problems.Your paper will be published online a‚er we receive your correc*ons and will appear in print in the next available issue.You can find out your date of online publica*on by contac*ng the Nature Press Office

Final Decision Le*er:
As soon as your ar*cle is published, you will receive an automated email with your shareable link.
You can now use a single sign-on for all your accounts, view the status of all your manuscript submissions and reviews, access usage sta*s*cs for your published ar*cles and download a record of your refereeing ac*vity for the Nature journals.
An online order form for reprints of your paper is available at <a href="hDps://www.nature.com/reprints/author-reprints.html">hDps://www.nature.com/reprints/author-reprints.html</a>.Please let your coauthors and your ins*tu*ons' public affairs office know that they are also welcome to order reprints by this method.
If you have not already done so, we invite you to upload the step-by-step protocols used in this manuscript to the Protocols Exchange, part of our on-line web resource, natureprotocols.com.If you complete the upload by the *me you receive your manuscript proofs, we can insert links in your ar*cle that lead directly to the protocol details.Your protocol will be made freely available upon publica*on of your paper.By par*cipa*ng in natureprotocols.com,you are enabling researchers to more readily reproduce or adapt the methodology you use.Natureprotocols.com is fully searchable, providing your protocols and paper with increased u*lity and visibility.Please submit your protocol to hDps:// protocolexchange.researchsquare.com/.A‚er entering your nature.comusername and password you will need to enter your manuscript number (NG-LE61339R1).Further informa*on can be found at hDps://www.nature.com/nature-por€olio/editorial-policies/repor*ng-standards#protocolsSincerely, Safia Danovi Editor Nature Gene*cs

Figure 4 :
Figure 4: 36 SNP loci where donor genotype differs by country.SNPs are either associated with keratinocyte cancer (KC), tan response (Tan) or previously found to be under selection in Singaporean genomes (SG); -for genotype indicates a call could not be made (Methods).Associated genes and mechanisms are suggested for each SNP using PheWAS and eQTL data from the Open Targets Genetics database.ECM, extra-cellular matrix, -indicates mechanism is unknown.Population allele frequencies are reported for East Asia and Great Britain using data from 1000 Genomes Project Phase 3 (Methods).Note all SNPs in the HYAL region form part of the same LD block on chr 3p.21.

Figure 3 :
Figure 3: Clonal selection and competition differs by country.a The number of mutations of each consequence for positively selected genes (dNdScv: q < 0.01) by country.ARID2 is not significantly positively selected in SG samples.b Plot of non-synonymous mutations per gene in SG vs. UK samples.Gradient of line = total number of non-synonymous mutations in UK/SG = 6846/1432.Positively selected genes (purple) are labelled.Red indicates positively selected genes with a significant (p < 0.001) difference in dN/dS ratio by country, after accounting for global differences.c A representation of protein-altering mutations in 1 cm 2 of skin from donors of Singapore and the UK.Samples were randomly selected and mutations displayed as circles, randomly distributed in the space.Sequencing data, including copy number, was used to infer the size and number of clones and, where possible, the nesting of sub-clones.Otherwise, sub-clones are nested randomly.d Estimated percentage of cells with at least one non-synonymous mutation per positively selected gene, by country (samples with known CNA removed).Wilcoxon: p = 0.03 (NOTCH1), 0.7 (NOTCH2), 9 x10 -3 (FAT1) and 0.05 (TP53), p-adjusted by Holm multiple comparison.e Estimated percentage of cells with at least one non-synonymous mutation across 74 genes by country (Methods, Wilcoxon: p = 4.3 x10 -3 ).f Violin plot comparing clone size distributions (summed VAF) per mutation by country.Mutations from UK donors had a lower mean summed VAF (Welch#s t-test: p = 4.27 x10 -14 ).

Fig. 3d .
Fig. 3d.Percentage of cells with at least one non-synonymous mutation per positively selected gene, by country, by gene (samples with known CNA removed).We now report the multiple testing correction in the figure legend.

Fig 3e .
Fig 3e.Percentage of cells with at least one non-synonymous mutation across 74 genes per country.Wilcoxon test, two groups, this does not require multiple testing correction.Fig 3f.Clone size distributions (summed VAF) per mutation by country.Wilcoxon test, two groups, does not require multiple testing correction.

Figure 4 :
Figure 4: 36 SNP loci where donor genotype differs by country.SNPs are either associated with keratinocyte cancer (KC), tan response (Tan) or previously found to be under selection in Singaporean genomes (SG); -for genotype indicates SNP could not be called (Methods).Associated genes and mechanisms are suggested for each SNP using PheWAS and eQTL data from the Open Targets Genetics database.ECM, extra-cellular matrix, -indicates mechanism is unknown.Population allele frequencies are reported for East Asia and Great Britain using data from 1000 Genomes Project Phase 3 (Methods).
are you?I hope you're well.
3. Introduction, paragraph 1, page 2, sentence 5: Refs 18-21 aren't citations to a large twin study (I assume the authors mean Mucci 2016 PMID: 26746459) nor are they all citations to GWAS of KC cancer.18 is a PRS study, 20 isn't a GWAS, and there are other published KC GWAS other than refs 19 and 21 that should be cited.Suggested action: check/update references : a) Please include chr and position and present SNPs in their chr/bp order.b)Please include an appropriate reference population allele freqs for EUR and EA (e.g. from GNOMAD ) as a column to help contextualise the results.This is relevant because many SNPs are rare (e.g.rs192481803 has a MAF ~ 0 in all pops which is likely why everyone is missing the risk allele) and/or many of the risk alleles related to pigmentation are essentially absent in East and South Asian populations e.g.see point 4.
c) Please check if the listed SNPs are in LD with each other at least in EUR pops (as their BCC/ SCC associations were identified n EUR ancestry pops) e.g.rs4268748 is an intronic variant near MC1R, and is in LD with the rs1805007 functional missense allele (LD r2 0.23 in EUR) which is likely why it has previously been reported for BCC/SCC.As rs1805007 is wildtype in SG samples I am not sure it is relevant this SNP is heterozygous in some SG samples.The authors should consider removing SNPs in LD with other functional SNPs if they exist.d) Some SNP IDs are out of date e.g. the ID rs74664507 has been merged with rs11445081 e) Some listed SNPs have known functions/associations assigned since the publication of the cited reviews (41 and 42) e.g.rs74664507 / rs11445081 is strongly associated with skin colour.The authors should survey more recent literature, or databases like Open targets genetics (e.g.https://genetics.opentargets.org/variant/9_16913838_T_A[genetics.opentargets.org])to update the figure to better reflect current information.Suggested action: update/check SNP data entered into figure