Most cancers carry a substantial deleterious load due to Hill-Robertson interference
- Department of Biology, Stanford University, United States
- Department of Genetics and Genome Sciences, Case Western Reserve University, United States
- Department of Medicine, Division of Oncology, Stanford University School of Medicine, United States
- Department of Genetics, Stanford University School of Medicine, United States
- Stanford Cancer Institute, Stanford University School of Medicine, United States
Figures
Figure 1
Two Hill-Robertson interference processes that accumulate deleterious mutations at high mutation rates.
(A) Genetic hitchhiking. Each number identifies a different segment of a clone genome within a tumor. De novo beneficial driver mutations that arise in a clone can drive other mutations (passengers) …
Figure 2 with 16 supplements
Attenuation of selection and increased protein folding stress in high mutation load tumors.
(A) dN/dS of passenger (red) and driver (green) gene sets within 10,288 tumors in TCGA stratified by total number of substitutions present in the tumor (dN(observed)+dS(observed)). dN/dS is …
Figure 2—figure supplement 1
Schematic of our permuted dN and dS calculation.
Permuted synonymous and nonsynonymous counts are used to account for mutational biases in dN/dS calculations. Observed mutations and their tri-nucleotide context is shown in a solid gray bar. …
Figure 2—figure supplement 2
Permutation-based null model of mutagenesis corrects for mutational biases in dN/dS calculations.
(A) Simulations (N=100) of negative selection under extreme mutational bias scenarios where all mutations are generated from a single mutational signature (e.g. APOBEC or smoking, COSMIC signatures …
Figure 2—figure supplement 3
Attenuation of selection with increasing mutational burden in both oncogenes and tumor suppressors.
dN/dS of passenger and driver gene sets (Bailey et al., 2018) within tumors in TCGA stratified by the total number of substitutions present in the tumor (dN+dS). dN/dS is calculated with error bars …
Figure 2—figure supplement 4
No common germline polymorphisms observed in low mutation rate cancers.
(A) Fraction of mutations that overlap all germline polymorphisms in the 1000 Genomes Project within tumors stratified by the total number of substitutions. (B–D) Fraction of mutations that overlap …
Figure 2—figure supplement 5
Weaker signals of positive selection within cancer-specific drivers.
dN/dS values of passenger and different driver gene sets within tumors in TCGA stratified by the total number of substitutions present in the tumor (dN+dS). dN/dS is calculated with error bars using …
Figure 2—figure supplement 6
Patterns of attenuated selection persist across tumor purity thresholds.
(A) Correlation between tumor purity (calculated by GDC using the ABSOLUTE [Johnson, 1999] algorithm, Materials and methods) and the total number of substitutions in all TCGA samples (R2 = −0.01). …
Figure 2—figure supplement 7
Comparison of dN/dS to results in Martincorena et al., 2017, for tumors stratified by mutational burden.
(A) dN/dS in driver (green), passenger (red), and all gene sets (gray) of tumors in TCGA stratified by the total number of substitutions using nine bins of equal width (log-scale TMB), as depicted …
Figure 2—figure supplement 8
Random permutations of the positions of observed copy number alterations (CNAs) exhibit neutral values of dE/dI.
The stop and start location of each observed CNA was randomly permuted, while preserving its length. dE/dI was calculated for CNAs (with and without non-focal amplifications) using both metrics: …
Figure 2—figure supplement 9
Fractional overlap of copy number alterations (CNAs) within exomic regions (dE) relative to intergenic regions (dI) exhibits similar patterns of selection as fractional overlap.
Calculations of fractional overlap (Zack et al., 2013) of exomic regions (dE) to intergenic (dI) regions within passenger and GISTIC (Mermel et al., 2011) driver gene sets in tumors stratified by …
Figure 2—figure supplement 10
Signal of negative selection in subclonal mutations are robust to variant allele frequency (VAF) threshold.
dN/dS calculations within clonal and subclonal passenger and driver gene sets within tumors in TCGA stratified by the total number of substitutions using a permutation-based null model of …
Figure 2—figure supplement 11
Attenuation of negative selection within different functional gene sets.
dN/dS of passengers within different functional gene sets in the highest (black) and lowest (gray) mutational burden bin across all tumors using a permutation-based null model of mutagenesis. Dotted …
Figure 2—figure supplement 12
Attenuation of selection in somatic nucleotide variants (SNVs) persists across cancer subtypes and broad cancer group categories.
(A) dN/dS in passenger and driver gene sets within tumors stratified by the total number of substitutions in broad tumor sub-categories. Error bars are 95% confidence intervals determined by …
Figure 2—figure supplement 13
Attenuation of selection in copy number alterations (CNAs) in cancer subtypes and broad cancer group categories.
dE/dI in driver (green) and passenger (red) gene sets in tumors stratified by the total number of CNAs for the six most commonly sequenced cancer subtypes (presented in Figure 2) calculated using (A)…
Figure 2—figure supplement 14
Upregulation of heat shock protein pathways in tumors with elevated mutational burdens.
(A) Z-scores of median gene expression of (i) all genes, (ii) HSP90, (iii) Chaperonins, and (iv) the Proteasome averaged across tumors stratified by the total number of copy number alterations …
Figure 2—figure supplement 15
The power to detect signals of selection is dependent on the quality of mutation calls.
dN/dS values in passenger and driver gene sets within tumors in TCGA stratified by the total number of substitutions when using low-quality mutations (bottom panels, ‘Mutect2 SNP Calls’) and …
Figure 2—figure supplement 16
Quantity of mutations within each mutational burden bin for data depicted in Figure 2.
(A–D) all report the total number of samples used in their respective figure pane within Figure 2. (A) Counts of mutations in passenger (red) and driver (green) gene sets within tumors stratified by …
Figure 3 with 6 supplements
Approximate Bayesian computation (ABC) procedure estimates the strength of selection in passengers and drivers.
(A) Schematic overview of the ABC procedure used. A model of tumor evolution with genome-wide linkage contains two parameters – sdrivers (mean fitness benefit of drivers) and spassengers (mean …
Figure 3—figure supplement 1
dN/dS rates of drivers and passengers in simulated cancers with various fitness coefficients.
Ten-thousand simulated tumors were generated for various combinations of mean driver fitness benefits (sdrivers) and mean passenger fitness costs (spassengers, Materials and methods). For some …
Figure 3—figure supplement 2
Probability of cancer by age and mutational burdens in simulated cancers at various fitness coefficients.
Clinical summary statistics of simulated tumors at various combinations of mean driver fitness benefits (sdrivers) and mean passenger fitness costs (sp, Materials and methods). (A) Initial …
Figure 3—figure supplement 3
Implementation and use of approximate Bayesian computation (ABC) for model selection and parameter estimation.
(A) Leave-one-out cross validation (CV) on the simulated data was used to select an optimal rejection tolerance and optimal rejection method. Observed data can be compared to simulated data using …
Figure 3—figure supplement 4
Evidence of positive selection on synonymous mutations within driver genes at low mutational burdens.
(A) The quantity of synonymous mutations within driver genes was compared to the quantity of synonymous mutations within passenger genes and both were normalized by their expected frequencies using d…
Figure 3—figure supplement 5
Distribution of mutation rates of simulated tumors.
(A) Mutation rates of all simulated tumors were randomly sampled from a uniform distribution (in log-space) from 10–12 to 10–7 nucleotide–1 · generation–1. (B) In simulations that best agreed with …
Figure 3—figure supplement 6
Relative contribution of genetic hitchhiking and Muller’s ratchet to fix deleterious passengers.
Using analytical theory developed in Neher and Shraiman, 2012; McFarland et al., 2014; Bachtrog and Gordo, 2004, we can estimate the relative rates of genetic hitchhiking and Muller’s ratchet in our …
Author response image 1
Additional files
-
MDAR checklist
- https://cdn.elifesciences.org/articles/67790/elife-67790-mdarchecklist1-v2.docx
-
Supplementary file 1
Model assumptions of tumor evolution and its anticipated effects.
- https://cdn.elifesciences.org/articles/67790/elife-67790-supp1-v2.docx
-
Supplementary file 2
Broad (meta-categories) of cancer groupings used in Figure 2 and Figure 2—figure supplement 12-13.
- https://cdn.elifesciences.org/articles/67790/elife-67790-supp2-v2.docx
