Mechanistic basis of atypical TERT promoter mutations

Abstract

Non-coding mutations in the TERT promoter (TERTp), typically at one of two bases -124 and -146 bp upstream of the start codon, are among the most prevalent driver mutations in human cancer. Several additional recurrent TERTp mutations have been reported but their functions and origins remain largely unexplained. Here, we show that atypical TERTp mutations arise secondary to canonical TERTp mutations in a two-step process. Canonical TERTp mutations create de novo binding sites for ETS family transcription factors that induce favourable conditions for DNA damage formation by UV light, thus creating a hotspot effect but only after a first mutational hit. In agreement, atypical TERTp mutations co-occur with canonical driver mutations in large cancer cohorts and arise subclonally specifically on the TERTp driver mutant chromosome homolog of melanoma cells treated with UV light in vitro. Our study gives an in-depth view of TERTp mutations in cancer and provides a mechanistic explanation for atypical TERTp mutations.

Fig. 1. Overview of cancer types with TERTp mutations in GENIE.

59 cancer types from GENIE (v11) were considered, all having at least 1% TERTp-mutated samples. For each cancer type, the fraction of samples with ETS-forming TERTp driver mutations and recurrent atypical mutations is indicated, as well as the fraction UV-related single and double base substitutions (mutational signatures SBS7 and DBS1, respectively). All recurrent TERTp mutations (present in ≥10/19,755 samples and being within 300 bp upstream of the TERT translation start site) are indicated by separate colours. Numbers (e.g. −146) refer to base positions relative to the translation start. Cancer types were divided into subcohorts based on the frequency of TERTp driver mutations (above or below 20%) and the fraction UV-related substitutions (above or below 10%). The number of patients (each represented by a single tumour sample) is indicated within parentheses for each cancer type, and OncoTree codes are shown within square brackets. SNV, single nucleotide variant; DNV, double nucleotide variant. Source data are provided as a Source Data file.

Fig. 2. Atypical TERTp mutations arise at pre-existing or de novo-formed UV hotspot positions.

a Mutation frequencies in a 80 bp region of the TERTp (chr5:1295188−1295268, hg19) in highly TERTp-mutated ( ≥ 20% of samples) GENIE cancer types, which were further subdivided into non-UV associated (“High TERTp, no UV”) or UV-associated (“High TERTp, UV”) subcohorts as indicated in Fig. 1. Mutations at a native (preexisting) ETS factor binding site arise only in the UV-exposed set, at positions in the ETS motif known to exhibit a UV damage hotspot effect when occupied (primarily CCTTCCK but also CCTTCCK). The schematic shows GABP binding to mutant TERTp sequences. b Frequencies in the UV-exposed subcohort following further subdivision of tumours by presence (positive axis) or absence (negative axis) of a TERTp driver mutation (−124, −126 or −139/−138 bp). Nearly all native ETS site mutations co-exist with primary driver events in the same patients, suggestive of a two-step model. Inset shows contingency table for driver and native site mutation co-occurrence (P-value from two-sided Fisher’s exact test). Equally, atypical mutations at −149 and −126 bp, predicted secondary hotspot bases at de novo ETS sites from −146 and −124 bp driver SNVs, co-exist with driver events. SNV, single nucleotide variant; DNV, double nucleotide variant; ONV, oligonucleotide variant; CPD, cyclobutane pyrimidine dimer. Source data are provided as a Source Data file.

Fig. 3. Passenger-like distribution of atypical TERTp mutations.

Line plots showing the frequency of key TERTp mutations as a function of UV mutation burden (C > T mutations in dipyrimidine contexts) across the TERTp mutated, UV-exposed, GENIE subcohort (“High TERTp, UV” cancer types in Fig. 1). 1,569 samples were divided into five bins based on burden as indicated on the x-axis (586, 287, 291, 236 and 169 samples per bin). a ETS-forming driver mutations show limited correlation with UV mutation burden. b The frequency of non-ETS-forming atypical mutations is strongly dictated by UV mutation burden, as expected for UV-induced passengers under neutral selection. Error bars indicate 80% confidence intervals based on the standard errors of the proportions. The Pearson correlation coefficient r and associated two-sided P-value, calculated based on the average burden in each bin, is indicated for each category. ns, not significant; *P < 0.05; **P < 0.01. Source data including exact P-values are provided as a Source Data file.

Fig. 4. Atypical TERTp mutations co-occur in cis with canonical drivers in 100k Genomes melanomas.

a Somatic mutations at positions of interest in the TERTp in the Genomics England 100k Genomes melanoma cohort. 335 samples are shown ordered by tumour mutation burden (TMB). The fraction of mutations attributed to the UV signature SBS7 is indicated for each sample (blue = 0%; orange = 100%). b Allele phasing in 12 cases with primary driver mutations (−146 bp, −139/−138 bp, or −124 bp) and atypical mutations. Reads were split by mutation status at the key TERTp driver position, considering only reads covering all the relevant positions. The total number of driver mutant and wild-type reads are indicated by the height of the positive and negative axes, respectively. Mutations in cis with (i.e. in the same read as) the mutant allele or the reference allele are shown on the positive and negative axes, respectively. TMB, tumour mutation burden; SNV, single nucleotide variant; DNV, double nucleotide variant. Source data are provided as a Source Data file.

Fig. 5. Binding of GABP to the TERTp promotes UV damage formation at ETS hotspot sites in vitro.

a DNaseI footprint of wild-type or mutated radiolabelled TERTp dsDNA fragments encompassing the native and two main proto-ETS sites, with and without added recombinant GABPA/B protein complex. ETS sites are indicated and were localized as described in Supplementary Fig. 4. b CPD formation following UV irradiation (5000 J/m2 UVC) of the same TERTp fragments as revealed by T4 endonuclease V digestion, with and without GABP (see Supplementary Fig. 5 for additional gel image data). CPD, cyclobutane pyrimidine dimer. Source data are provided as a Source Data file.

Fig. 6. UV exposure of A375 melanoma cells introduces mutations at predicted TERTp UV hotspot positions specifically on the −146 bp (C250T) mutated chromosome homologue.

a A375 cells, either untreated or UV-exposed (36 J/m2 UVC daily during 6 weeks), were assayed for subclonal TERTp mutations using a modified SiMSen-seq error-corrected amplicon sequencing protocol. A375 cells carry a −146 bp TERTp driver mutation, and mutations were phased in relation to this event by categorizing all error-corrected consensus reads (minimum 3 × oversampling) based on genotype at −146 bp (positive and negative axes). b Per-base mutation counts (same TERTp region as in Fig. 2) in non-exposed cells, which had few recurrent mutations (15,923 consensus reads in total). All individual reads having recurrent mutations outside of the −146 site are shown. c Per-base mutation counts in UV-exposed cells (12,006 consensus reads), showing atypical recurrent mutations at predicted UV hotspot positions at the native ETS site (−101 and −100 bp) and at −149 and −148 bp, which are predicted secondary UV hotspot bases at the −146 bp proto-ETS site. The atypical mutations predominantly occurred in cis with the −146 bp driver mutation. Only mutations detected in > 1 consensus read were considered. CPD, cyclobutane pyrimidine dimer. Source data are provided as a Source Data file.