in normal urothelium are all known cancer
genes. Somatic mutations in certain genes
could also lead to cellular death or differ-
entiation, which would lead to a depletion of
protein-altering mutations in surviving clones.
Although this dataset is not powered to detect
negative selection at the level of individual
genes, exome-wide dN/dS ratios excluding
known cancer genes were close to, and not
significantly lower than, 1 (Fig. 2F). This is
consistent with the vast majority of somatic
coding point mutations being tolerated by
normal cells and accumulating passively, in
line with observations in cancer genomes ( 19 ).
Similar nonsignificant results were obtained
when focusing on putative antigenic regions ofthe exome, thereby providing no clear evidence
of immune editing against these mutant clones
(fig. S4) ( 22 ).Extreme variation in driver preference
across individuals
Having identified many independent mu-
tant clones per donor, we were able to studySCIENCEsciencemag.org 2 OCTOBER 2020•VOL 370 ISSUE 6512 77
Fig. 2. Positive selection of
bladder cancer genes in normal
urothelium from organ donors.
In (A) to (D), analyses are shown for
17 genes under positive selection
in normal urothelium and for
four other genes frequently mutated
in bladder cancer. (A) Number and
consequence of mutations detected
in histologically normal urothelium.
(B) Observed-to-expected ratios
for missense substitutions, truncat-
ing (nonsense and essential splice
site) substitutions, and indels.
(C) Estimated percentage of
urothelial cells bearing a mutation
for donors aged≥50 years from
samples with median on-target
coverage≥50×. (D) Percentage of
urothelial carcinomas in TCGA with a
nonsynonymous substitution or
indel. Error bars depict 95%
binomial confidence intervals.
(E) Scatterplot comparing mutation
frequency in bladder cancer (D)
and the number of nonsynonymous
mutations in normal urothelium
(A) for driver genes (colored
by biological function) identified
in this study and in previous
studies ( 18 , 19 ). (F) Comparison of
dN/dS values for the 321 cancer
genes in the targeted panel to
19,107 cancer passenger genes
[defined in ( 19 )]. Dashed line indi-
cates a dN/dS value of 1, which
indicates neutral expectation.
(G) Heatmap showing the number of
unique nonsynonymous mutations
in abundant (≥10 mutations) driver
genes across transplant organ
donors. Sample numbers refer to
samples with at least one mutation.
Blue boxes indicate statistically
significant combinations of gene and
donor ( 22 ). (HandI) Histology
images annotated with driver
mutations and their cellular fractions
in sequenced microbiopsies from
two transplant organ donors
exhibiting enrichment of drivers in
KDM6AandARID1A, respectively.
RESEARCH | RESEARCH ARTICLES