The origins and genetic interactions of KRAS mutations are allele- and tissue-specific

Abstract

Mutational activation of KRAS promotes the initiation and progression of cancers, especially in the colorectum, pancreas, lung, and blood plasma, with varying prevalence of specific activating missense mutations. Although epidemiological studies connect specific alleles to clinical outcomes, the mechanisms underlying the distinct clinical characteristics of mutant KRAS alleles are unclear. Here, we analyze 13,492 samples from these four tumor types to examine allele- and tissue-specific genetic properties associated with oncogenic KRAS mutations. The prevalence of known mutagenic mechanisms partially explains the observed spectrum of KRAS activating mutations. However, there are substantial differences between the observed and predicted frequencies for many alleles, suggesting that biological selection underlies the tissue-specific frequencies of mutant alleles. Consistent with experimental studies that have identified distinct signaling properties associated with each mutant form of KRAS, our genetic analysis reveals that each KRAS allele is associated with a distinct tissue-specific comutation network. Moreover, we identify tissue-specific genetic dependencies associated with specific mutant KRAS alleles. Overall, this analysis demonstrates that the genetic interactions of oncogenic KRAS mutations are allele- and tissue-specific, underscoring the complexity that drives their clinical consequences.

Fig. 1. The contribution of mutational processes to KRAS mutagenesis.

a The frequency of KRAS mutations in each cancer. b The distribution of KRAS allele frequencies at the four hotspots, codons 12 (left), 13 (middle-left), 61 (middle-right), and 146 (right) in each cancer. The size of the circle reflects the percent of KRAS mutations that are the indicated allele in each cancer. Each cancer is assigned a different color. The number of tumor samples whose sequencing data was collected for this study is indicated along the y-axis. c The average composition of mutational signatures in tumor samples grouped by KRAS allele. Each color represents a different mutational signature. Mutational signatures of know etiology are annotated. d The average probability of each mutational signature to have caused the KRAS mutation in a tumor sample. This value accounts for the level of each mutational signature in the tumor sample, and the ability of the mutational signature to cause the indicated KRAS allele. In c and d, only KRAS alleles found in at least 15 tumor samples of the cancer type are included. Source data are provided in the Source data file.

Fig. 2. The predicted frequencies of cancer-specific KRAS alleles.

a The predicted versus observed frequency of KRAS alleles for the common alleles of each cancer. Triangles indicate rejection of the null hypothesis that the observed and predicted frequencies are the same (χ-squared test, p values were adjusted using the Benjamini–Hochberg FDR correction method, hereon referred to as FDR-adjusted p values; FDR-adjusted p value < 0.05); circles indicate the failure to reject the null hypothesis (χ-squared test, FDR-adjusted p value $\ge$ 0.05). Error bars indicate bootstrapped 95% confidence intervals of the predicted values. b The average probability of the indicated KRAS allele in tumors samples with the KRAS allele (closed circle), tumors samples with a different KRAS mutation (open circle), and tumor samples with WT KRAS (upside-down triangle). The errors bars indicate bootstrapped 95% confidence intervals of the mean. For each allele, differences in the probabilities between tumor samples with the allele and those with another allele, and between tumor samples with the allele and those with WT KRAS were tested via a Wilcoxon rank-sum test (FDR-adjusted p values < 0.05 are indicated). Source data are provided in the Source data file.

Fig. 3. The comutation networks of oncogenic KRAS alleles.

a The comutation network of the KRAS alleles in COAD with each edge representing a significant comutation interaction between an allele and another gene (p value < 0.01). The color of the edge indicates whether the interaction was an increase (blue) or decrease (green) in the frequency of comutation. Genes with multiple interactions are represented by a gray dot to disambiguate them from where edges intersect. b A subset of the network shown in a of genes that encode proteins known to physically interact with KRAS, are in one of its canonical upstream or downstream pathways, or are validated oncogenes or tumor suppressors. The width of the edge indicates the strength of the association. c Cellular functions enriched in the comutation networks of the KRAS alleles in COAD (left), LUAD (center), and PAAD (right). The size of the dot indicates the number of genes in both the function and the comutation network, and the transparency indicates the FDR-adjusted p value of the enrichment. d, e A visualization of the increased (d) or decreased (e) comutation of select genes with KRAS G12D in COAD. Rows of the central plot represent genes. Each column of the central plot is a different tumor sample. A filled space denotes a mutation of the gene in the sample, the color describing the type of variant. The bar plots above and to the right indicate the marginal values of the central plot. f A comparison of the comutation frequencies in PAAD of the genes producing proteins in the PPIN of SMAD1-3. Each column is a gene with a comutation interaction with a KRAS allele and in at least one of the gene sets. The black tiles on top indicate that the gene was in the PPIN of the indicated SMAD protein. The bar plot shows the distribution of the comutation events of each gene across tumor samples with the various KRAS mutations. n = 4145 COAD, 5051 LUAD, 1262 MM, and 2314 PAAD biologically independent tumor samples for the increased comutation analysis, and n = 1536 COAD, 891 LUAD, 1395 PAAD biologically independent tumor samples for the reduced comutation analysis. Source data are provided in the Source data file.

Fig. 4. Allele-specific genetic dependencies in COAD cell lines.

a Gene sets with significant enrichment for increased (lower dependency score; purple) or reduced (higher dependency score; orange) genetic dependency in COAD cell lines. The size of the dot relates the FDR-adjusted p value of the association and the color indicates the strength of the enrichment (“normalized enrichment score”). b, c Heatmaps ranking the cell lines by dependency (“dep.”) score of the genes at the leading edge of enrichment for two gene sets. Each row represents a gene and each cell represents a cell line colored by its KRAS allele. The cell lines are arranged in ranking order by their dependency score for the gene. Thus, each column indicates a rank. The line plots above the heatmaps indicate the representation (density) of each KRAS allele at each rank across the genes. d Hierarchically clustered heatmaps of the genes that demonstrated differential genetic dependency amongst cell lines of different KRAS alleles. Each column is a cell line labeled by its DepMap identifier and each row is a gene. e Examples of genes that demonstrated differential genetic dependency amongst cell lines of different KRAS alleles (t tests; FDR-adjusted p values). For the box plots, the box demarcations represent the 25th, 50th, and 75th percentiles, and the whiskers extend from the box to the largest and smallest data points at most 1.5 times the interquartile range away from the median. n = 23 biologically independent COAD cell lines. Source data are provided in the Source data file.

Fig. 5. Some dependency interactions can be explained by comutation events.

a The nonzero coefficients for the model of STARD9 dependency in COAD cell lines regressed on KRAS G12D (versus all other KRAS alleles) and its comutation interactors (top), and the actual dependency scores for KRAS G12D mutant and TP53-mutant cell lines (bottom). Cell lines without either mutation or with both are not shown. KRAS G12D has reduced comutation with TP53 in COAD. b–d The nonzero coefficients for the models of b EEF1E1, c ABL1, and d MYBL2 dependency in PAAD cell lines regressed on KRAS G12D (versus all other KRAS alleles) and its comutation interactors (top), and the actual dependency scores for KRAS G12D mutant and SMAD4 mutant cell lines (bottom). Cell lines without either mutation or with both are not shown. KRAS G12D has reduced comutation with SMAD4 in PAAD. For the box plots, the box demarcations represent the 25th, 50th, and 75th percentiles, and the whiskers extend from the box to the largest and smallest data points at most 1.5 times the interquartile range away from the median. n = 23 biologically independent COAD cell lines, and n = 23 biologically independent PAAD cell lines. Source data are provided in the Source data file.