Main

In the interplay between mutagenesis and selection, human somatic tissues evolve as mosaics of competing clones driven by mutations, many of which affect cancer genes4,5,6,7,8,9,10,11,12. Although most of these clones do not result in cancer, in some instances they constitute the first step of the evolutionary trajectory towards malignant tumours. Therefore, characterizing the clonal landscape of human tissues can provide a path to understanding the mechanisms of cancer formation and cancer risk. In the case of the bladder, men and people with smoking history have an increased risk of developing cancer, independently of other risk factors1,2 (Supplementary Note 1). The reasons for the sex bias, and whether it is also present in the clonal landscape of the normal urothelium, are not known. The role of smoking—whether purely mutagenic or also as promoter of mutant clones—is not well understood either.

The detection of mutant clones in normal urothelium, as in other normal tissues, is challenging because clones are small and escape detection by conventional bulk sequencing methods. Here we directly probed large mixtures of clones from epithelial brushes of normal bladder using ultradeep DNA duplex sequencing (around 5,000× per sample, amounting to approximately 400,000 haploid genomes in aggregate)13,14. We focused on the detection of clones driven by mutations in 16 genes known to be under positive selection in bladder urothelium10,15, and/or mutated frequently in bladder carcinomas16,17, including the telomerase gene (TERT) promoter18. We then used the magnitude of positive selection for each gene at the sample level to regress out the effect of sex and smoking on their clonal landscape.

We also considered that the capacity to identify the driver mutations of thousands of clones in a human tissue could increase markedly our ability to uncover the functional effect of all mutations in genes relevant in tumourigenesis. Identifying all potential driver mutations in a cancer gene is key to advancing personalized cancer medicine. The selective advantage provided by spontaneous mutations in millions of cells is tested in the interplay between mutagenesis and selection in normal human tissues. We should then be able to probe the results of these natural experiments by sequencing a large enough number of cells in a somatic tissue. We explore this postulate with data from ultradeep sequencing of normal bladder. As a result, we have uncovered a powerful tool with which to quantify positive selection at site resolution.

Mutation landscape of normal urothelium

We collected cells from either one or two regions—the upper (dome) and lower (trigone)—of the bladder urothelium from 45 deceased donors (79 samples) through a brushing of relatively large surface areas (roughly 2 cm2) at autopsy (Fig. 1a and Supplementary Note 2). Clinical data, including exposure to known or suspected bladder cancer risk factors such as tobacco smoking or chemotherapy, was available for most donors (Supplementary Table 1). We used ultradeep DNA duplex sequencing13,14 (average depth ranging between 1,236 and 9,303 across samples, with a median of 5,164; Supplementary Note 2) and a newly developed and carefully calibrated computational pipeline to identify somatic mutations with high sensitivity and specificity in these polyclonal samples (Extended Data Fig. 1a–c and Supplementary Note 3). The accuracy of this technology allows the reliable detection of mutations that are present in a single DNA molecule (Extended Data Fig. 2a–c and Supplementary Note 4).

Fig. 1: Ultradeep DNA sequencing of normal urothelium targeting driver genes shows thousands of mutations.
figure 1

a, Schematic representation of sampling and duplex DNA sequencing of polyclonal epithelial brushes from normal bladders. b, Number of SNVs detected in a panel of selected genes in the normal bladder in this study and comparison with the number of mutations detected in 892 tumours from bladder cancer genomics studies, obtained from intOGen24. c, Number and density of somatic mutations (SNVs, MNVs and indels up to 100 base pairs) identified in 79 samples of normal urothelium obtained from 45 donors. Bladder location (dome or trigone), age, sex and smoking history of donors are shown below the bar plots. d, Trinucleotide substitution profiles of the mutational signatures identified across this cohort of normal urothelium samples through de novo extraction. e, Top, scatter plot representing the relationship between the activity (number of mutations contributed) of SBS-ageing in samples and the age of donors; effect size, regression line and P value of a univariate mixed-effects linear model. Bottom, box plot representing the activity of SBS-chemo in donors exposed or not exposed to chemo/radiotherapy; effect size and P value of a univariate mixed-effects linear model. The boxplots show the quartiles with whiskers extending to the highest and lowest data points within 1.5 times the interquartile range; N = 79 samples for all plots in the panel. Mb, megabase. BioRender was used to create panel a (https://BioRender.com/fgnnet9).

To investigate how bladder cancer risk factors might change the clonal landscape of the urothelium, we focused on 15 genes identified previously to be under positive selection in normal bladder10,15 and/or known to frequently drive bladder tumours16,17 and the promoter of the gene encoding TERT, which is mutated in around 70% of bladder tumours18,19,20,21,22,23. This amounted to 111,876 base pairs of genomic DNA comprising the exonic regions and neighbouring intronic sites of the 15 genes (in full or selected fragments) and a region of the TERT promoter (hereinafter for simplicity, genes; Supplementary Tables 2, 3 and 4). Across samples, we identified a total of 64,278 mutations, between 54 and 1,785 (median 774) mutations per sample, including single nucleotide variants (SNVs), multiple nucleotide variants (MNVs) and indels up to 100 base pairs (Supplementary Table 5). This total number of mutations is approximately 16-fold higher than the number identified in the same genes across 892 bladder tumours (intOGen)24 sequenced over a decade of cancer genomics (Fig. 1b,c).

The overall non-protein-affecting mutation density—mutations per megabase sequenced—was associated, as expected, with age and the exposure to certain mutagenic factors, such as several chemotherapies/radiotherapy (Supplementary Note 5). Using two orthogonal signature extraction methods25,26,27, and supported by the detection of a large number of non-protein-affecting mutations, we discerned the activity of three main mutational signatures (Supplementary Note 5 and Extended Data Fig. 3a–d). These captured the activity of the APOBEC family of cytidine deaminases28 (SBS-APOBEC; Fig. 1d), age-related mutagenesis29 (SBS-ageing; Fig. 1d,e) and an unknown process (harder to reduce to a linear combination of mutational signatures in the COSMIC catalogue30) that seemed to be associated with exposure to chemotherapy (SBS-chemo; Fig. 1d,e and Supplementary Note 5). Overall, the landscape of mutational signatures and the number of mutations per megabase sequenced was similar to that identified in a previous study of normal urothelium10 (Extended Data Fig. 2b,c), with the exception of a signature reported to be associated with tobacco smoking. Mutational signatures were correlated highly between the dome and trigone of the same donor (Supplementary Note 5).

Driver mutations in normal bladder

The 16 genes included in the panel had been demonstrated previously to be under positive selection in normal urothelium or mutated frequently in bladder tumours. Thus, we first examined whether their mutations were positively selected in the samples included in this cohort. To this end, we used four complementary computational methods (Supplementary Note 6) on the basis of comparing features of the pattern of mutations observed in a gene with that expected assuming neutral evolution31,32.

First, we estimated the expected number of protein-affecting mutations with different impact, that is, missense or truncating (including nonsense and essential splice-site mutations) on the basis of the number of observed synonymous mutations in each gene across samples and in each sample using a newly developed method called Omega (Fig. 2a and Supplementary Note 6). This calculation takes into account the number of sites allowing amino acid changes of each type in the protein, the frequency of trinucleotide changes observed across samples and the depth at which each individual genomic site has been sequenced. Then, using a dN/dS approach33, Omega computes the degree of positive selection on missense (dN/dS missense) or truncating (dN/dS truncating) mutations. These values represent the excess of missense or truncating mutations—driver mutations—among those observed in the gene in the sample (Fig. 2b). In addition to the excess of truncating and missense mutations, we quantified three other signals of selection: clustering of mutations in the three-dimensional structure of the protein, functional impact bias and excess of frameshift indels (Supplementary Note 6 and Extended Data Fig. 4a). These metrics indicated that RBM10, KDM6A, STAG2, KMT2D, ARID1A, CDKN1A, TP53, EP300, NOTCH2, CREBBP, FOXQ1, KMT2C and RB1—13 of the genes included in this study—are under positive selection, driving clonal expansion across the 79 normal urothelium samples (Extended Data Fig. 4b,c). We verified that the same is true for PIK3CA through analysis of the mutations in a hotspot affecting the histidine at position 1,047 (Extended Data Fig. 4d).

Fig. 2: Computing positive selection in the normal urothelium.
figure 2

a, Distribution of truncating, missense and synonymous somatic mutations along the coding regions of RBM10 and TP53. Each needle represents the number of samples with mutations with one of the three consequences occurring on an amino acid residue. b, Magnitude of positive selection on truncating and missense mutations in RBM10 and TP53. Left, magnitude of positive selection indicated as the dN/dS ratio on the basis of the number of observed and expected mutations of each type. Right, percentage of driver mutations (observed in excess of neutrality) among truncating and missense in RBM10 and TP53. The number of observed, expected and estimated driver mutations are indicated for each gene. c, Magnitude of selection on truncating and missense mutations in FGFR3. dN/dS below 1 indicates negative selection. d, Magnitude of positive selection of activating mutations in the TERT promoter. Left, distribution of somatic mutations along the sequence of the TERT promoter colour coded according to whether they are activating (seen in tumours). Right, bar representing the magnitude of positive selection on activating TERT promoter mutations. In dN/dS bar plots (bd), the shaded segment in each bar represents the number of truncating or missense (or activating in the case of TERT) mutations expected under neutrality. The numbers above the bar detail the value of dN/dS for each type of mutation. Numbers inside bars represent the mutations in excess over the expectation, that is, the drivers. The P values in bd were calculated using a dN/dS approach (Omega) described in Supplementary Note 6.

This analysis produced two additional findings. First, we discovered that truncating (both SNVs and frameshift indels) mutations of FGFR3 seem to be under negative selection (Fig. 2c and Extended Data Fig. 4b,c). Second, we observed 85 SNVs affecting the TERT promoter (across 42 samples; Fig. 2d). The TERT gene encodes telomerase, and its promoter includes two hotspots that result in the creation of two de novo ETS transcription factor binding sites34 and are mutated frequently in human tumours18,19,22,23. To formally compute the strength of positive selection on the TERT promoter in our cohort, we dichotomized the mutations on the basis of whether or not they have been observed several times across 8,136 tumour whole genomes (Supplementary Note 6; Methods), that is, using their presence in tumours as a surrogate of their capacity to activate the expression of TERT. We could thus calculate a modified dN/dS value for the TERT promoter (dN/dS pTERT; Supplementary Note 6). Although we expected fewer than one activating mutation, we actually observed 56, implying that at least 55 (98.2%) of these mutations detected in the TERT promoter are drivers of clonal expansion, probably by activating the expression of telomerase in normal urothelial cells.

We next computed positive selection separately for each sample in the cohort (Extended Data Fig. 5a). The high depth of sequencing in the study guarantees that these sample-level dN/dS values can be estimated accurately and robustly (Supplementary Notes 7 and 8). The number of clones driven by mutations detected in each of the 13 genes under positive selection varies between zero and several hundreds across the samples analysed from each donor (Fig. 3a). Despite the abundance of driver mutations, the relative size of the clones is very small and thus the fraction of the urothelium covered by mutant clones of these genes is modest (less than 1%) in most cases (Extended Data Fig. 5b). Some activating TERT promoter mutations reach relatively high values of variant allele frequency (VAF), probably due to both clonal expansion and convergent evolution.

Fig. 3: Sex bias of the clonal landscape of normal urothelium.
figure 3

a, Heatmaps with number of driver SNVs (top) and the total number of protein-affecting indels (bottom) in 13 genes across female and male donors. For donors with dome and trigone samples (marked with a dot), the number of driver SNVs in both samples has been added. Donors are sorted by age in ascending order. b, Relationship between age and the density (per Mb) of protein-affecting and non-protein-affecting mutations. c, Association of age with the protein-affecting (top coefficient) and non-protein-affecting (bottom coefficient) mutation density (left), and the magnitude of positive selection on missense (centre) and truncating (right) mutations of 13 genes (gene–mutation consequence combinations for which we could calculate dN/dS at the level of sample for at least 80% of samples were included in the calculation). d, Distribution of dN/dS truncating values of RBM10, CDKN1A, ARID1A and STAG2 and dN/dS missense values of TP53 among male and female donors. Horizontal lines denote the median. e, Association of sex with the protein-affecting (top coefficient) and non-protein-affecting (bottom coefficient) mutation density (left), and the magnitude of positive selection on truncating mutations in RBM10, CDKN1A, ARID1A and STAG2 (right). The plots represent the results of multivariate regressions accounting for age, smoking history, alcohol drinking history, BMI and exposure to chemotherapy. In all regressions (c and e), multivariate linear mixed-effects models (accounting for sex, smoking history, alcohol drinking history, BMI and exposure to chemotherapy) were used to account for the presence of several samples of the same donor, and these models yielded the effect size and P values. Circles represent the point estimate of the effect size of the regressions; horizontal lines represent 95% confidence intervals. Circles with dark outer circumference denote significant associations (false discovery rate (FDR) threshold of 0.2). N = 79 samples in b, c and e. All corrected P values for these regressions appear in Supplementary Table 7. F, female donors; M, male donors.

We also found that the patterns of positive selection of the dome and trigone from the same donor were strikingly similar across the entire cohort (Extended Data Figs. 5c and 6a–c and Supplementary Note 5), indicating that the evolutionary dynamics and selective forces in the two regions are equivalent. Thus, a brushing of cells from approximately 2 cm2 of the urothelium provides a good representation of the clonal structure of the entire tissue.

In summary, applying a dN/dS-based approach to mutations detected across urothelial samples through ultradeep sequencing, we demonstrated that a large percentage of mutations in the profiled genes are drivers, indicating pervasive positive selection; we uncovered negative selection for truncating mutations in FGFR3; we showed positive selection for activating mutations in the TERT promoter and we made robust estimations of positive selection and number of driver mutations per gene in each sample, thus enabling the study of inter-individual differences in clonal selection (Fig. 3a, Extended Data Fig. 7a,b and Supplementary Table 6).

Sex bias in the bladder clonal landscape

To identify a potential sex bias of the urothelial clonal landscape, we resorted to regressions that incorporated meaningful covariates, in particular age and history of smoking1,35,36,37 (Supplementary Note 1). Indeed, part of this heterogeneity across donors is explained by age. The density of protein-affecting mutations increases at a higher pace than that of non-protein-affecting mutations, indicating that clones driven by mutations of the 13 genes under positive selection expand with age, as reported for other normal tissues7,8,9,11,12,38,39 (Fig. 3b,c and Supplementary Note 9). In agreement with this, the dN/dS values of missense and truncating mutations of these genes are also associated significantly with age (Fig. 3c and Supplementary Table 7).

Using multivariate regressions, we determined that the values of dN/dS truncating of RBM10, CDKN1A and ARID1A are increased significantly in the urothelium of men compared with women, independently of age, smoking history, alcohol drinking history, body mass index (BMI) and exposure to chemotherapy (Fig. 3d,e, Extended Data Fig. 8a,b, Supplementary Table 7 and Supplementary Notes 9 and 10). In addition, the dN/dS truncating of STAG2 shows a tendency, although not significant, towards higher values in men. Two of these genes (RBM10 and STAG2) are in the X chromosome, whereas CDKN1A and ARID1A are in autosomes. KDM6A is another X chromosome gene that has been described previously with a higher number of mutations in female bladder tumours40. A closer inspection of X chromosome genes confirms the expectation of higher coverage across women for all three, although only RBM10 (and STAG2 to a lesser extent) shows a sex bias of dN/dS truncating (Supplementary Note 9). This implies that the growth of clones driven by mutations under positive selection in these genes is different in the urothelium of men and women (Supplementary Notes 9 and 10). The number of bladder tumours with truncating mutations in RBM10 and CDKN1A is also significantly higher across men than women (Extended Data Fig. 8c,d and Supplementary Note 11). In summary, clonal selection in normal bladder differs between men and women, mirrored by an increased number of mutations in RBM10 and CDKN1A in male bladder tumours.

TERT clones associate with age and smoking

We discovered that activating mutations in the promoter of TERT (that is, those observed at least twice in tumours18,19,20,21,22,23,34), were much more recurrent among older donors in the cohort. Specifically, all TERT promoter-activating mutations in the cohort, except one, occurred across donors older than 55 years (Fig. 4a,b). Furthermore, in this group, those with a history of smoking exhibited a significantly higher rate of TERT promoter mutations (Fig. 4a,b). Some of these mutations showed relatively large VAF, probably caused by several expanding clones with these mutations (Extended Data Fig. 8e,f and Supplementary Note 10). We found that the association between smoking history in older donors and the rate of TERT promoter-activating mutations is independent of sex, BMI and exposure to chemotherapy (Fig. 4c), as well as the depth of sequencing. The significant increase of activating TERT promoter mutations in smokers contrasts with the absence of association between smoking history and overall mutation density across samples (Supplementary Note 5 and Supplementary Table 7). This result indicates that tobacco carcinogens have a role in promoting the growth of clones with TERT promoter mutations. In summary, the rate of TERT promoter-activating mutations in the normal urothelium is associated significantly with the interaction of age and a history of smoking. More than 70% of bladder tumours have TERT promoter mutations, with a comparable frequency in smokers and non-smokers. In lung tumours, where the frequency of TERT promoter mutations is much lower, we find a significantly higher frequency in smokers (2.1%) compared with non-smokers (0.9%) (Supplementary Note 11). Collectively, these results indicate that exposure of the normal urothelium to tobacco carcinogens may increase the risk of bladder cancer through the promotion of clones bearing mutations of the TERT promoter.

Fig. 4: Association of activating mutations in the TERT promoter with smoking.
figure 4

a, Number of activating (observed in tumours) and other (not observed in tumours) mutations in the TERT promoter across donors younger than 55 years (top), older than 55 years with no history of smoking (middle) or older than 55 years with a history of smoking (bottom). b, Density (mutations per Mb) of activating TERT promoter mutations observed in the three groups of donors. The horizontal line denotes the median of the distribution of the frequency of mutations in the group of donors older than 55 years with a history of smoking. The sample size of the three groups was 25 (samples from donors younger than 55), 14 (samples from donors older than 55 with no smoking history) and 38 (samples from donors older than 55 with smoking history), respectively. The two samples from a donor with unknown smoking history were excluded. c, Association between the frequency of activating mutations in the TERT promoter and the interaction between age and smoking history. A linear mixed-effects model was used to compute the P value. The circle represents the point estimate of the effect size of a multivariate linear mixed-effects regression, and the horizontal line represents 95% confidence intervals. The dark outer circumference denotes a significant association (FDR threshold of 0.2). The corrected P value appears in Supplementary Table 7. Number of samples for this analysis is the same as in a and b.

Natural human saturation mutagenesis

Experimental saturation mutagenesis (that is, the insertion of all possible mutations in a genomic element in experimental systems) has been used to identify sites or regions in proteins that are key for their structure, function or regulation, to understand their role in disease and improve the design of drugs to target them41,42,43,44,45. Recently, we introduced the concept of in silico saturation mutagenesis, exploiting more than 20 years of cancer genomics research on the natural experiments that are human tumours (or clonal expansions in normal blood) to build machine learning models that identify all their potential driver mutations46,47. In contrast to tumours, whose evolutionary history entails the expansion of a single clone, with only one or two mutations observed in a cancer driver gene in most cases, in normal polyclonal samples, several mutations affecting a given gene are detected as drivers of the expansion of several clones. Therefore, sequencing normal human tissues, which can also be considered natural experiments of clonal evolution, can provide a pathway towards natural saturation mutagenesis.

To understand how far along the ultradeep sequencing of roughly 400,000 haploid genomes takes us in the path towards natural saturation mutagenesis, we first calculated the number of mutations observed in each amino acid residue of the protein-coding genes (Fig. 5a). For example, we detected mutations (missense or nonsense) on 230 (58%) of the 394 amino acid residues of TP53 protein product covered by duplex sequencing (Fig. 5a). In normal urothelium, the frequency of mutations by gene differed from those in bladder cancers (Extended Data Fig. 9a) and the proportion of mutated residues was larger in all genes tested (Extended Data Fig. 9b), demonstrating different selective pressures and the advantage of studying mutations in normal tissue to reach saturation mutagenesis.

Fig. 5: Natural saturation mutagenesis.
figure 5

a, Percentage of amino acid residues in each gene with zero, one, two or three or more mutations across the 79 samples. b, Theoretical and observed kinetic of natural saturation mutagenesis for TP53, EP300 and FGFR3 as cumulative depth of sequencing increases. Grey line, theoretical kinetic of saturation mutagenesis (assuming no selection). Red circle, saturation achieved across the cohort. Red dashed line, observed kinetic (obtained by downsampling). c, Natural saturation mutagenesis of TP53 in normal bladder urothelium. From top to bottom, distribution of truncating, missense and synonymous mutations along the coding sequence of the gene; site selection computed for each amino acid residue of TP53 protein product; solvent accessibility along the protein sequence; TP53 protein product domains; duplex sequencing depth per amino acid residue. Left top, TP53 protein product three-dimensional structure with significant site selection of residues highlighted in blue. d, Experimental functional impact51 of TP53 mutations not observed, observed, or observed with significant site selection across the 79 samples. Only mutations with experimental functional impact reported in ref. 51 are included. e, dN/dS truncating and dN/dS missense values for each domain of TP53 protein product. The vertical lines represent the 95% confidence intervals of the dN/dS estimate. Solid border represents significant dN/dS values (P value < 0.05) according to Omega (Supplementary Note 6); N = 79 samples. f, Natural saturation mutagenesis of the TERT promoter. From top to bottom, distribution of mutations; site selection computed for observed mutations; experimental functional impact values of mutations in the TERT promoter according to ref. 55; distribution of mutations observed in the TERT promoter across 8,136 tumours (Supplementary Note 6). g, Experimental functional impact55 of mutations not observed, observed or observed in the TERT promoter with significant site selection across the 79 samples. h, Relationship between site selection and experimental functional impact value55 of all mutations observed in the TERT promoter.

Next, we asked how many more genomes would need to be sequenced to observe mutations on all amino acid residues of each gene. To estimate this, we first calculated the probability to observe each mutation in a gene under neutrality, on the basis of the number of synonymous mutations and the trinucleotide mutational probabilities observed in our cohort. Thus, we obtained a theoretical curve describing the fraction of all possible mutations that are expected to be observed at different sequencing depths (theoretical kinetic of natural saturation mutagenesis; Fig. 5b (continuous curve), Extended Data Fig. 9c and Supplementary Note 12). The shape of this curve is determined by the difference in probability across mutations, governed by their trinucleotide contexts, with some mutations much more likely to occur than others. For example, for TP53, with the accumulated depth provided by the entire cohort (approximately 500,000×), only on the basis of neutral mutagenesis, we expected to have observed around 26% of all possible mutations. In theory, being able to detect every possible mutation in the gene would require an accumulated depth of around 107, assuming an aggregated mutation density and mutational signatures as the samples in the study.

Nevertheless, the observation of mutations in normal urothelium is not neutral, but instead is strongly affected by selection. To obtain the actual curve of the fraction of observed mutations depending on the number of sequenced genomes (observed kinetic of natural saturation mutagenesis), we sub-sampled the number of mutations observed in each gene across the cohort, respecting their VAF but reducing the depth of sequencing at each position (dots in Fig. 5b and Extended Data Fig. 9c). The divergence in the shape of the two curves shows the importance of selection in shaping the probability to observe a given fraction of mutations in a gene. For example, as missense and truncating TP53 mutations are under positive selection, we actually observed more mutations than predicted by the theoretical kinetic. For several genes, we observed this faster accumulation of mutations than expected under neutral mutagenesis due to positive selection (Extended Data Fig. 9c). The curve for FGFR3 falls below the theoretical curve constructed under purely neutral mutagenesis (Fig. 5b), which supports the notion that FGFR3 mutations are under negative selection in the normal urothelium.

Approaching natural saturation mutagenesis provides an opportunity to calculate the strength of positive selection on mutations affecting each amino acid residue (Supplementary Note 6), and for other within-gene structures such as exons or protein domains (Fig. 5c, Extended Data Figs. 9d,e, 10a and Supplementary Fig. 1). In the case of TP53, most sites under significant selection are in the p53 DNA-binding domain48 (Fig. 5c), and are observed more frequently across tumours24,49,50 (Extended Data Fig. 10b), more likely to be identified as cancer drivers46 (Extended Data Fig. 10b) and tend to score higher on an experimental saturation mutagenesis assay51 (Fig. 5d), as described previously in duplex sequencing analysis of TP53 in other normal tissues52,53,54. In this and other genes that act as tumour suppressors, the mutations with significant site selection appear in buried areas of the protein (Fig. 5c, Extended Data Fig. 10c,d and Supplementary Fig. 2).

The calculation of dN/dS at the level of domains shows that only the p53 DNA-binding domain exhibits a significant excess of missense mutations (Fig. 5e). By contrast, TP53 dN/dS truncating values present a more homogeneous distribution across exons and domains (Fig. 5e and Supplementary Fig. 2). The same is true for other genes such as EP300 and CREBBP where only the histone acetyltransferase domain (HAT-KAT11) shows significant dN/dS missense values (Extended Data Fig. 9d and Supplementary Figs. 1 and 2). For RBM10, with a much stronger signal of truncating mutations, residues with the highest site selection are distributed more uniformly than in TP53 (Supplementary Fig. 2). Genes affected mostly by truncating driver mutations, such as STAG2 and RBM10, show a higher dN/dS truncating than dN/dS missense values for virtually all exons, except the first and last (Supplementary Fig. 2). Despite the fact that almost all genes sequenced are tumour suppressors, in all cases, a few SNVs exhibit much stronger site selection than the rest (Extended Data Fig. 10a and Supplementary Fig. 2). Whereas FGFR3 mutations showed negative selection overall (Fig. 2c), mutations affecting a single residue (G380) were positively selected in normal urothelium. This residue did not correspond to the three most common FGFR3 hotspots observed in bladder cancer (Supplementary Fig. 3).

TERT promoter mutations with significant site selection in normal urothelium appear more frequently across tumours (Extended Data Fig. 10e), and present overall higher experimental functional impact values55 (Fig. 5f–h). In summary, natural saturation mutagenesis, brought about by the possibility of identifying mutations at extremely low VAF through ultradeep sequencing of normal tissues, opens up the possibility of directly assessing the functional impact of mutations affecting different genes in human tissues.

Discussion

The reasons behind the wide heterogeneity in clonal landscape observed across normal human tissues from different people7,8,9,10,11,12,38,39 and the potential role of known cancer risk factors are not understood. In this study we demonstrated the suitability of ultradeep DNA duplex sequencing13,14 (approximately 400,000 haploid genomes) to accurately measure the magnitude of positive selection on each of 16 genes in 79 individual samples (Supplementary Note 6). The high number of mutations identified proved key to showing associations between clonal expansions and bladder cancer risk factors. Specifically, we discovered that the magnitude of positive selection on truncating mutations in RBM10, ARID1A and CDKN1A is significantly higher among men even when adjusting for age, smoking history, alcohol drinking history, and BMI. Importantly, mutations in RBM10 and CDKN1A are also significantly more abundant in bladder cancers of men than of women (Supplementary Note 11). We can speculate that the observed differences between men and women could be related to differences in exposure to internal factors (for example, sex hormones) or extrinsic exposures that disproportionately affect men. Discerning whether the bias towards the expansion of clones bearing mutations in these and other genes is somehow related with the observed increased risk of bladder cancer across men requires further investigation. We also found a high increase of TERT promoter mutations among older people with a history of smoking, which may explain at least part of the link between smoking and increased bladder cancer risk35,36,37, given the high frequency of TERT promoter mutations across bladder tumours18,19,20,21,22,23,34. In this cohort, we do not find a specific smoking-associated mutational signature. The high increase of activating TERT promoter mutations in smokers, rather than a general increase in mutation density, points to a non-mutagenic—acting rather as promoter56,57,58—role of tobacco smoking in bladder tumourigenesis. Scaling up these analyses to probe the influence of other bladder cancer risk factors requires the possibility of obtaining samples in a minimally invasive manner, which could be conceivably achieved through urine or lavages59.

The proposal of a pathway towards the study of natural saturation mutagenesis through ultradeep error-correcting duplex DNA sequencing of large mixtures of clones from normal tissues is a very compelling one. In this study we have provided a strong argument for the plausibility of this approach, paving the way for future studies. We have shown that it is possible to obtain a fine map of the degree of selection acting on mutations at individual sites or regions of a protein throughout the normal development of tissues of people exposed to different external factors. Saturation mutagenesis—a key development towards personalized cancer medicine60—through ultradeep sequencing in human tissues, thus provides a cost-effective and direct alternative to complement experimental and in silico approaches. In addition, characterization of the patterns of positive (and negative) selection in different genes and human tissues will be critical to understanding how somatic evolution influences cancer risk.

Methods

Sample collection

Two epithelial brushes (2–3 cm2) from the bladder top (dome) and the bladder floor (trigone) were obtained from 53 people without known bladder pathology and no history of bladder cancer upon autopsy (average 4 days post-mortem) at the University of Washington. Next of kin consented to autopsies and research on leftover specimens. The study of de-identified collected specimens and linked clinical history from the deceased donors was reviewed and deemed not human subjects research by the University of Washington Institutional Review Board (STUDY00016707; IRB Federal Wide Assurance number, FWA 00006878). We obtained the following relevant clinical information for all donors: age, sex, BMI, tobacco smoking history, alcohol use, previous cancer and chemotherapy exposure (Supplementary Table 1). Three donors with active or chronic inflammation and four donors with insufficient DNA were discarded. The two samples from another donor were also discarded from the study upon visual inspection of their mutational profile, resulting in a total of 45 deceased donors in this study (Supplementary Notes 2 and 3).

Duplex DNA sequencing

Capture panel design

We designed a panel including ten genes identified in a previous study as being under positive selection in the normal urothelium and with more than ten mutations across the 1,647 microbiopsies analysed10. We added five genes that are mutated frequently in bladder tumours30 and the TERT promoter—also known to be under positive selection in bladder carcinomas22,23. This resulted in a panel containing the entire (or almost entire) coding region of 12 genes (ARID1A, NOTCH2, FOXQ1, CDKN1A, KMT2D, RB1, CREBBP, TP53, EP300, KDM6A, RBM10 and STAG2), three genes for which only selected regions were targeted because of clustering of cancer mutations in those regions and/or difficulties for capturing the full gene (PIK3CA, FGFR3 and KMT2C) and the TERT promoter10,61. The panel was constructed by TwinStrand Biosciences and covered 111,876 base pairs, including 65,086 base pairs in coding regions and 46,790 base pairs in non-coding regions (Supplementary Tables 2 and 3 and Supplementary Note 2).

DNA extraction, duplex library preparation and sequencing

After centrifugation of epithelial brushes, DNA was extracted using the DNeasy Blood and Tissue (Qiagen) kit, following manufacturer’s instructions with some variations (Supplementary Note 2). The DNA integrity number (DIN) was measured using Agilent 4200 TapeStation Genomic tapes. Duplex sequencing libraries were prepared using commercially available kits (TwinStrand Biosciences)13,14,62,63,64 and 250 ng of genomic DNA. The DNA fragmentation step was carried out taking into account the starting DIN of samples and monitored using TapeStation. Fragmented DNA was subject to end-repair, A-tailing, ligation to duplex sequencing adaptors, library conditioning and PCR amplification, according to protocol. The PCR product was captured at 65 °C for 16–20 h. Upon PCR amplification, libraries were pooled for sequencing. Sequencing was performed with a NovaSeq 6000 at the Department of Laboratory Medicine and Pathology at University of Washington or a NovaSeq X Plus at Novogene or the Fred Hutchinson Cancer Center using 2 × 150 base-pair paired-end reads (around 115 million reads per sample; Supplementary Note 2).

Duplex DNA sequencing was successful for 34 donors on both samples. For 11 other donors, we produced duplex DNA sequencing data only for dome or trigone because of insufficient DNA or too fragmented DNA (DIN < 1.4) in the paired sample. Thus, the final number of donors with available data for at least one sample was 45 (Supplementary Table 1) and the final number of samples processed was 79 (Supplementary Table 4).

Somatic mutation calling

We constructed a computational pipeline (deepUMIcaller) in Nextflow65 to call mutations from duplex sequencing data on the basis of an early version of nf-core/fastquorum pipeline66, which implements the fgbio Best Practices FASTQ to Consensus Pipeline (https://github.com/fulcrumgenomics/fgbio/blob/main/docs/best-practice-consensus-pipeline.md) and downstream variant calling by VarDictJava (https://github.com/AstraZeneca-NGS/VarDictJava). A series of filters to discard potential artefacts are included in the pipeline. Code implementing deepUMIcaller is available at (github.com/bbglab/deepumicaller). For a detailed description of the pipeline, see Supplementary Note 3.

To estimate the error rate of TwinStrand DNA duplex sequencing, we compared the density of mutations and mutational profile identified across three cord blood samples (purchased from StemCell) with that expected on the basis of colonies obtained from human haematopoietic stem cells67,68. The estimated error rate resulting from this analysis was around 4 × 10−8, which is two orders of magnitude lower than the mutation density across samples in this study (Extended Data Fig. 2a and Supplementary Notes 4 and 8).

Mutational signatures

De novo signature extraction

We extracted mutational signatures de novo with a Bayesian hierarchical Dirichlet process using HDP_sigExtraction pipeline (https://github.com/McGranahanLab/HDP_sigExtraction) and the R-package hdp developed by N. Roberts (https://github.com/nicolaroberts/hdp)9 and SigProfilerExtractor (https://github.com/AlexandrovLab/SigProfilerExtractor)26. The HDP_sigExtraction pipeline was run with the default parameters with no previous signatures assigned. Five signatures in addition to the null signature were extracted. De novo signatures were extracted using SigProfiler using the nonnegative matrix factorization approach. The same input data were used. The upper bound for the number of signatures was set to ten; however, the most robust solution was three signatures that were similar to the three most active signatures extracted by HDP. These two sets of mutational signatures were decomposed into known COSMIC signatures when possible (Supplementary Note 5).

Assessing biases in trinucleotide composition of the panel

To account for biases due to the trinucleotide composition of the panel, the number of substitutions of each class was re-calculated. The rate of each substitution was calculated as the number of observed substitutions with the consequence in question divided by the number of corresponding sequenced trinucleotide sites (number of trinucleotides with this consequence in the panel weighted by sequencing depth). This mutational probability was multiplied by the number of the corresponding trinucleotide in the genome to get the expected number of substitutions of this type in the whole genome. The mutational probability of each substitution for the whole genome was calculated by dividing the expected number of substitutions with this consequence per genome by the sum of all expected substitutions per genome. Finally, the expected number of substitutions with each consequence was obtained by multiplying the probability by the total number of substitutions observed in the sample to keep the absolute number of observed mutations.

Statistical association between mutational signatures and clinical variables

We tested for possible associations of de novo extracted signatures with age, sex and bladder location. To do this we used linear mixed-effect regression models (Supplementary Note 5). We tested both the number of mutations attributed to the signature (counts) and the relative contribution (proportion) of each signature as a response variable. Age, sex and bladder location were used as fixed effects (separated regression was prepared for each fixed effect variable) and donors were used as random effects to control for non-independence between dome and trigone samples from the same person. P values for the association were obtained using likelihood-ratio tests (ANOVA function) comparing models including and excluding the variable of interest. Associations with all other clinical variables (smoking history, drinking history, chemotherapy history, and so on) were tested in the same way but always including age, sex and bladder location in the regression together with the variable of interest.

Positive selection

We used four methods to compute positive selection on the mutations observed across genes. One method (Omega)—a dN/dS approach to assess the strength of selection on the mutational pattern of genes—was developed de novo for this study and is described at length in Supplementary Note 6 (https://github.com/bbglab/omega). Two others, OncodriveFML and Oncodrive3D, which compute the deviation in the average functional impact and clustering in the three-dimensional structure of proteins, respectively, from those expected under neutrality, had been developed previously, and were adapted here to work on duplex sequencing data. These are also described in Supplementary Note 6. A fourth method, assessing the relative enrichment for frameshift indels observed across genes was also developed de novo for this study and is described thoroughly in Supplementary Note 6. For the TERT promoter, all mutations that were observed at least twice across 8,136 whole-genome sequencing tumour samples sequenced by the Hartwig Medical Foundation and the Pan-Cancer Analysis of Whole Genomes consortium (Supplementary Note 6) were considered activating. The remaining mutations were used to calculate the expected number of activating mutations under neutrality.

Fraction of the urothelium covered by clones with driver mutations

The number of missense and truncating driver mutations of each gene in a sample was obtained from the dN/dS missense and dN/dS truncating values. From these values the number of missense and truncating mutations in excess in each sample were calculated. Specifically, the 95% confidence intervals of both dN/dS values were used to compute two extreme values of driver mutations in each sample, whereas the mean dN/dS value was used to compute an expected number of driver mutations.

This mean number was thus used to select the most likely driver mutations in the sample. To this end, mutations were ranked in descending order according to their VAF, and the top number of mutations corresponding to the expected number of driver mutations were selected. Then, for a gene G, we computed the fraction of genomes bearing driver mutations out of those sequenced at each genomic position covered by the DNA duplex sequencing as:

$$P(G)=\mathop{\sum }\limits_{i=1}^{n}{(-1)}^{i-1}\sum _{| S| =i}\prod _{x\in S}{p}_{x}=\sum _{x}{p}_{x}-\sum _{x\ne y}{p}_{x}{p}_{y}+\cdots +{(-1)}^{n-1}{p}_{{x}_{1}}{p}_{{x}_{2}}\cdots {p}_{{x}_{n}}$$

where px is the all-molecules VAF of a driver mutation x (considering both duplex and non-duplex reads; Supplementary Note 3) and n is the number of driver mutations from the previously selected set in the gene. The formula is the realization of the probabilistic principle of inclusion-exclusion assuming independence across mutations69.

The fraction of genomes with driver mutations is equal to the fraction of cells with driver mutations under the assumption that two mutations, one in each homologous copy of the gene, are required to drive the clonal expansion. This extreme would be true in the case that all studied genes behave as classical tumour suppressors and bear deleterious mutations in both alleles. An exception to this rule are the mutations in genes in the X chromosome in men, whereby one mutation would suffice to drive the clonal expansion. In general, it is possible for some genes to be capable of driving the clonal expansion (or mutations at specific positions in a gene) with only one mutated allele. It is also possible that the other allele is affected by a large deletion or methylation event not seen by DNA duplex sequencing. In the extreme that only one mutation per cell per gene is required, the fraction of cells with driver mutations will be double the fraction of genomes. In the plot in Extended Data Fig. 5b, we use the two-hit assumption.

Association of the urothelium clonal structure with bladder cancer risk factors

We designed a two-step strategy to assess the influence of known bladder cancer risk factors on features of the urothelium clonal structure across samples from donors in the cohort. As dependent variables representing the urothelium clonal structure, we selected the (protein-affecting and non-protein-affecting) mutation density, and the magnitude of positive selection on genic mutations (dN/dS missense and dN/dS truncating). As the dependent variables are continuous, we chose linear regressions, specifically mixed-effects linear models, to account for two samples from the same donor and use all available samples. As independent variables, we included a list of available clinical features: age (in decades), sex, smoking history (binarized as ever or never smoker), alcohol consumption history (also binarized), BMI (re-scaled within the 0–1 interval), and exposure to chemo/radiotherapy (binarized). For the case of TERT promoter mutation density association, we used the interaction of age and smoking history instead of age and smoking history separately. In the first step, we applied univariate mixed-effects linear models to identify any clinical feature with a significant association with any of the dependent variables for any gene or for all genes. In the second step, we applied multivariate mixed-effects linear models to rule out confounding effects for the associations that seemed significant in the first step. Associations with FDR below 0.2 were deemed significant. We also carried out a binomial test to rule out a spurious dependence of the mutation density on group differences in terms of sequencing depth. For details, see Supplementary Notes 9 and 10.

Tumour data

Mutations identified across 622 muscle-invasive and 105 non-muscle-invasive bladder cancer (MIBC and NMBIC, respectively) cohorts were downloaded from cBioPortal49,70,71,72,73 together with the clinical data of the combined study. From the MIBC Beijing Genomics Institute cohort74 only samples labelled as invasive were considered MIBC and added to the MIBC dataset. Mutations identified in a further cohort of 79 NMIBCs were obtained from the literature40 and included as part of the NMIBC dataset. Only non-silent protein-affecting mutations in 14 of the genes included in the panel were analysed. In case of duplicated samples across MIBC and NMBIC cohorts, the mutations from the most recent published study were kept. In total, mutations across 806 bladder tumours were obtained (622 MIBC and 184 NMIBC). The mutation density per gene across the two datasets was calculated by dividing the number of observed mutations per gene by the length of its coding region. To compute a mutation density metric comparable with that used in the normal urothelium that accounts for the number of megabases sequenced, we multiplied the coding length by two times the number of samples in which that gene was sequenced, to take into account the two copies in a diploid genome. Under the assumption that, in tumours each mutation belongs to a single clonal expansion and in normal tissues we have a mixture of clones, we reasoned that the number of different tumour genomes sequenced would be equivalent to the number of genomes sequenced in normal urothelium. The mutation density per megabase was calculated multiplying the mutation density by 106.

Mutations identified across 33,218 tumours (892 bladder tumours, mostly MIBC) in intOGen24 were downloaded from intogen.org. These mutations were used to obtain the total number of mutations observed in each gene, their distribution along the sequence of the genes in the study and the percentage of sites affected by different numbers of mutations. The same data of 109,017 tumours (3,909 bladder tumours) were obtained from the GENIE project50 to calculate the frequency of mutations in each of the genes and the TERT promoter. Mutations in the TERT promoter were also obtained from two cohorts of tumours (included in intOGen) sequenced at the whole-genome level (Hartwig Medical Foundation21: N = 5,582 and Pan-Cancer Analysis of Whole Genomes19: N = 2,554). The classification (and score) of all possible mutations in TP53 into drivers and passengers through in silico saturation mutagenesis was obtained from boostDM (intogen.org/boostDM). Details of analyses involving tumour mutations appear in Supplementary Note 11.

Natural saturation mutagenesis

To compare the distribution of somatic mutations along the sequence of each of the genes in normal urothelium and bladder tumours, we first downloaded somatic mutations identified across 892 bladder tumours from the intOGen (intogen.org) platform. Although this cohort is larger than the 806 muscle-invasive and non-muscle-invasive tumours used to explore differences in mutation density, it is composed mostly of muscle-invasive carcinomas.

For each gene, we started the analysis with all genomic sites in the coding sequences and splicing sites that passed the pipeline’s filtering criteria of sufficient duplex coverage across samples (Supplementary Note 3). We then mapped these DNA sites to protein positions using the Ensembl REST API75, allowing us to determine the total number of protein positions (including amino acid residues and stop codon) covered by the panel. Then, we computed the number of mutations affecting each protein residue of every gene studied here across the 79 normal urothelium samples and the 892 bladder tumours. Next, we obtained the consequence type (missense, synonymous, nonsense or splice-affecting) of these mutations on the MANE transcript76 of each gene from the output of the Variant Effect Predictor v.111 (ref. 77), run within the intOGen pipeline24 (https://github.com/bbglab/intogen-plus). The method to compute the natural saturation mutagenesis kinetics in Fig. 5b is described in Supplementary Note 12. The residue-level sequencing depth shown in Fig. 5c, Extended Data Fig. 9d and Supplementary Figs. 2 and 3 was calculated as the average depth across all sites corresponding to each codon.

Calculation of site selection

We developed a metric to measure selection per site by comparing the observed with the expected number of mutations at each site or residue. The expected number of mutations per site was obtained by distributing the expected number of mutations in a given gene under neutrality along all the possible changes in that same genomic region, with the assumption that only the mutation probability of each trinucleotide and the sequencing depth are responsible for the within-gene differences of mutation probability. We computed this value for each possible mutation in the positively selected genes including the TERT promoter, and for the protein-coding genes we also obtained a value per residue. A more complete explanation on this metric (and the detection of positive selection in sub-genic structures such as exons and domains) can be found in Supplementary Note 6.

Structural representation and features

Structural models for all proteins used to run Oncodrive3D were obtained from the AlphaFold database (AlphaFold 2 v.4)78,79. Three-dimensional protein structures and protein structural features represented across figures were also obtained from the AlphaFold database. Solvent accessibility and secondary structure information were extracted from the AlphaFold-predicted PDB structures using PDB_Tool (https://github.com/realbigws/PDB_Tool). Structural visualizations of proteins were produced using UCSF ChimeraX80.

Comparison with experimental saturation mutagenesis

The results of two saturation mutagenesis experiments estimating the functional impact of mutations in the TP53 DNA-binding domain51 and along the sequence of the TERT promoter55 were obtained directly from supplementary tables from both papers. In the case of TP53, we used the transformed score. In the case of the TERT promoter, values calculated for the glioblastoma SF7996 cells system were used. In these comparisons, the site selection of individual mutations was used.

Comparison of orthogonal studies of normal bladder

We obtained the mutations identified through whole-genome sequencing in clonal or quasi-clonal samples obtained from the normal bladder of donors by laser capture microdissection in a previous study10. We used these samples to construct the mutational profile of normal urothelium, which we compared with that reconstructed in the present study through ultradeep sequencing (Extended Data Fig. 2b). From the same previous study, we obtained the mutations identified through whole-exome sequencing. Mutations obtained from samples sequenced at depth lower than 80× were discarded. From these data, we calculated the number of mutations per megabase observed in each sample, and averaged this value across donors. A linear regression was then constructed between these values and the age of donors, and a trend line calculated. The number of mutations per megabase calculated across the samples in our cohort was overlaid upon the obtained regression. The results of this comparison are presented in Extended Data Fig. 2c.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.