Atypical UV Photoproducts Induce Non-canonical Mutation Classes Associated with Driver Mutations in Melanoma

Abstract

Somatic mutations in skin cancers and other ultraviolet (UV)-exposed cells are typified by C>T and CC>TT substitutions at dipyrimidine sequences; however, many oncogenic "driver" mutations in melanoma do not fit this UV signature. Here, we use genome sequencing to characterize mutations in yeast repeatedly irradiated with UV light. Analysis of ~50,000 UV-induced mutations reveals abundant non-canonical mutations, including T>C, T>A, and AC>TT substitutions. These mutations display transcriptional asymmetry that is modulated by nucleotide excision repair (NER), indicating that they are caused by UV photoproducts. Using a sequencing method called UV DNA endonuclease sequencing (UVDE-seq), we confirm the existence of an atypical thymine-adenine photoproduct likely responsible for UV-induced T>A substitutions. Similar non-canonical mutations are present in skin cancers, which also display transcriptional asymmetry and dependence on NER. These include multiple driver mutations, most prominently the recurrent BRAF V600E and V600K substitutions, suggesting that mutations arising from rare, atypical UV photoproducts may play a role in melanomagenesis.

Keywords: BRAF; NRAS; UV damage; UV mutagenesis; skin cancer; thymine-adenine photoproduct; transcriptional asymmetry.

Figure 1.. Genome Sequencing of UV-Induced Mutations in Yeast Reveals Non-canonical Mutation Signatures

(A) Experimental procedure for genomic sequencing of UV-induced mutations accrued following 15 exposures to UVC light (25 J/m²) in independent yeast isolates. (B) Percentage of surviving cells following exposure of wild-type (WT) yeast to a single dose of UVC light. Arrow indicates the dose used for genome sequencing experiments (25 J/m²). (C) Number of mutations per isolate of WT accrued following 9× or 15× exposures to UVC light (25 J/m²). Mutations were identified by genome sequencing of each independent yeast isolate. (D) Mutation profile of single-nucleotide substitutions in UV-exposed yeast (9× and 15× doses). The mutation count for each substitution type (e.g., C>A, C>G, etc.) and trinucleotide context is depicted. The middle base of each trinucleotide context is mutated.

Figure 2.. UV-Induced Mutations in Yeast Are Suppressed by NER and Show Transcriptional Asymmetry

(A) Number of mutations in each genome-sequenced WT or NER-deficient isolate is plotted. Inset: schematic showing that repair of the TS by the transcription-coupled-nucleotide excision repair (TC-NER) pathway is primarily dependent upon Rad26 in yeast, while the global genomic-nucleotide excision repair (GG-NER) pathway requires Rad16. (B) Deletion of RAD16 significantly increases the frequency of both canonical (i.e., C>T) and non-canonical (e.g., T>A or T>C) UV-induced mutations. Mutations are classified as being either in the 5′ or 3′ position of a dipyrimidine (“Dipyr”) or not in a dipyrimidine (“No”; see Figures S1C and S1D). Mean ± SEM is depicted for WT or rad16Δ mutant isolates. Significant differences were determined using a t test with the Holm-Sidak correction for multiple hypothesis testing. **p < 0.001; *p < 0.01. (C) Transcriptional asymmetry (i.e., normalized ratio of mutations on NTS relative to TS across all yeast genes) is plotted relative to total number of mutations for each trinucleotide context and each C>N mutation class in WT, rad16Δ, and rad26Δ mutants. Only mutation classes with at least 30 mutations are plotted. The color of the circle indicates the mutation class (e.g., C>T); mutation classes in a dipyrimidine are plotted as a circle with a black outline. (D) Same as (C), except for T>N mutation classes. (E) Fraction of T>C mutations in genes occurring on the non-transcribed strand (NTS) relative to the transcribed strand (TS) for the indicated trinucleotide contexts in rad16Δ mutant cells. Statistical significance was determined using the chi-square test and Bonferroni correction for multiple hypothesis testing. **p < 0.001; *p < 0.05. (F) T>A mutations are significantly enriched on the TS relative to the NTS at NTA sequences (e.g., ATA, CTA, GTA, TTA) in rad16Δ mutant cells. Statistical significance was determined using the chi-square test and Bonferroni correction for multiple hypothesis testing. **p < 0.001; *p < 0.05. (G) Schematic showing that elevated NTA mutations on the TS indicate that causative lesion is located on the other DNA strand (i.e., NTS) at TAN sequences (e.g., TAA, TAC, TAG, TAT).

Figure 3.. Mapping Atypical TA Photoproducts in UV-Irradiated DNA and Cells

(A) Design of DNA oligonucleotide containing a stretch of five thymine-adenine (TA) sequences (underlined), but no dipyrimidine sequences. (B) Analysis of UV lesions by denaturing gel electrophoresis. Double-stranded DNA oligos were irradiated with increasing doses of UVC light (0.86–25.7 kJ/m²) and cleaved by UV DNA endonuclease (UVDE). The locations of the different TA photoproducts in the DNA sequence are indicated (TA 1–5), based on size standards in the leftmost lane. (C) Quantification of UVDE-cleaved TA photoproducts induced by different doses of UVC light. (D) Chemical structure of TA photoproduct. (E) UVDE-seq method for mapping non-CPD UV photoproducts. CPD lesions are removed by photoreactivation with purified CPD photolyase, and the remaining UV photoproducts are cleaved with UVDE. (F) UVDE-seq reads are enriched at dipyrimidine sequences and TA dinucleotides immediately following irradiation of a rad16Δ mutant with 600 J/m² UVC light, consistent with UV-induced formation of 6–4PPs and TA photoproducts. “No UV” sample is included as a control.

Figure 4.. UV Light Induces Novel Tandem Mutations in Yeast

(A) Spectrum of tandem mutations derived from genome sequencing of WT yeast following repeated UV exposure (9 or 15 doses). (B) UV-induced tandem substitutions are elevated in repair-deficient rad16Δ mutant cells. Mutations per sequenced isolate for 15 dose experiments are plotted. Significant differences in the number of mutations in each mutation class per isolate in WT relative to the rad16Δ mutant strain was determined using a t test with the Holm-Sidak correction for multiple hypothesis testing. **p < 0.001; *p < 0.01. (C and D) UV-induced tandem substitutions are elevated on the NTS of yeast genes in WT and rad16Δ mutant cells. Statistical significance was determined using the chi-square test and Bonferroni correction for multiple hypothesis testing. **p < 0.001; *p < 0.05. (E) Sequence logo representation of DNA flanking all AC>NN tandem substitutions (e.g., AC>TT, AC>CT, etc.) in yeast. Logo was generated using weblogo (Crooks et al., 2004).

Figure 5.. Non-canonical Mutation Classes in Skin Cancers Are Associated with UV Exposure and Show Transcriptional Asymmetry

(A) Density of mutations for 140 sequenced cutaneous melanomas (Hayward et al., 2017) associated with the 5′ position of dipyrimidine (Dipyr.), 3′ position of dipyrimidine, or not associated with a dipyrimidine (No), as defined in Figure S1C. Inset shows data just for mutations in coding exons of melanoma driver genes. (B) Ratio of mutation frequency per tumor for 140 cutaneous melanomas relative to 35 acral melanomas (Hayward et al., 2017). Dashed line indicates the median ratio across all mutation classes. (C and D) Fraction of C>T mutations in genes occurring on the NTS relative to the TS for the indicated trinucleotide contexts in cutaneous squamous cell carcinomas (cSCCs) derived from XPC^−/− patients is plotted. Transcriptional asymmetry is plotted for genes (C) highly expressed (top quartile) and (D) lowly expressed (bottom quartile) in keratinocytes. Statistical significance was determined using the chi-square test and Bonferroni correction for multiple hypothesis testing. **p < 0.001; *p < 0.05. (E and F) Same as (C) and (D), except for T>C mutations. (G and H) T>A mutations are significantly enriched on the TS relative to the NTS at NTA sequences (e.g., ATA, GTA, TTA) in XPC^−/− cSCCs in genes highly expressed (top quartile; G), but not in low-expressed genes (bottom quartile; H).

Figure 6.. Non-canonical Tandem Mutations in Melanoma Are Associated with UV Exposure and Show Transcriptional Asymmetry

(A) Spectrum of tandem mutations in 140 sequenced cutaneous melanomas (Hayward et al., 2017). (B) Spectrum of tandem mutations in the coding exons of melanoma driver genes. (C) Ratio of tandem mutation frequency per tumor for 140 cutaneous melanomas relative to 35 acral melanomas (Hayward et al., 2017). Only tandem mutation classes with at least 150 mutations in the cutaneous melanoma dataset and 10 mutations in the acral melanoma dataset are plotted. (D) Transcriptional asymmetry (normalized ratio of NTS relative to TS) of tandem mutations in high-expressed melanocyte genes (top quartile) plotted versus total number of each type of tandem mutation. (E) Same as (D), except transcriptional asymmetry is plotted for tandem mutations in low-expressed melanocyte genes (bottom quartile).

Figure 7.. UV Light Induces Oncogenic Tandem Substitutions Found in BRAF

(A) Odds ratio of recurrent substitution mutations (circles) in the BRAF gene in skin cancers relative to non-skin cancers from the COSMIC database (Tate et al., 2019). Each mutation is positioned along the x axis in accordance with its position within the BRAF cDNA. Nucleotide (nt) and amino acid (aa) positions are indicated below the schematic of the BRAF protein domains. Specific substitution types are color coded, and the number of times each recurrent mutation occurs in the dataset is indicated by the size of the circle. (B) UV light induces AC-to-TT tandem substitutions in yeast. The yeast ura3 K93V reporter is inactive due to an AA-to-GT (red text) substitution at codon 93, resulting in the change of the catalytic lysine (Miller et al., 2001) to valine. Reversion of the ura3 K93V to WT via an AC>TT mutation therefore mimics the BRAF V600K substitution. Median URA⁺ reversion frequencies were determined from six independent measurements for yeast treated with 0, 50 J/m² UVC light, or 300 J/m² UVB light. Error bars indicate ranges. For yeast treated with 0 J/m² UVC, no URA⁺ revertants were recovered in any of the six replicates, so a maximum estimated frequency was calculated as if each replicate contained a single URA⁺ colony. p = 0.0022 by Mann-Whitney ranked sum test. (C) T>A mutations in a GTG context are significantly enriched on the TS relative to the NTS in highly expressed genes in melanocytes (top quartile), but not in lowly expressed genes (bottom quartile). **p < 0.001. (D) UV light induces T>A and TG>AA substitutions in a GTG context. Same as (B), except the yeast trp5 E50V reversion reporter was used to mimic the BRAF V600E mutation. Median TRP⁺ reversion frequencies were determined from six independent measurements for yeast treated with 0, 25 J/m² UVC light, or 300 J/m² UVB light. Error bars indicate ranges. p = 0.0022 by Mann-Whitney ranked sum test. Right panel: estimated frequency of T>A (i.e., GTG-to-GAG) and TG>AA (i.e., GTG-to-GAA) mutations in the trp5 reversion assay, based on sequencing of TRP⁺ revertants.