Fig. 5: CTNNB1 mutation effect scores predict signaling activation and immune exclusion in HCC.
From: Mutational scanning reveals oncogenic CTNNB1 mutations have diverse effects on signaling
a, Proportion of HCC samples from TCGA cohort with CTNNB1 missense or AXIN1 coding mutations (left). The CTNNB1 missense mutations are then further divided into three categories (right): strong and weak mutations within the exon 3 hotspot (Supplementary Fig. 6b) and mutations that occurred elsewhere in the gene (other). A total of 18 further samples fell into more than one category or had deletions within CTNNB1; therefore, they were excluded. TCGA sample IDs and classifications are listed in Supplementary Table 9. b, Stacked histograms show the frequency and proportion of CTNNB1 exon 3 mutations from TCGA HCC cohort classified as weak or strong that lie in the docking site for β-TRCP (positions 32–37) versus elsewhere in the mutation hotspot. P value shows Fisher’s exact test (two-sided). c, Expression of n = 10 β-catenin target genes in HCC stratified by β-catenin pathway mutation status. For each gene, the median expression value was calculated across all samples in the indicated patient group, then expressed as a percentage of the median value for the same gene in the strong group (indicated by the dashed horizontal line at 100%). FPKM, fragments per kilobase per million. Solid horizontal lines show the median percentage value across all ten genes, boxes show the upper and lower quartiles and whiskers show the range. *P < 4 × 10−3, **P < 1 × 10−4 from negative binomial generalized linear model (two-sided), with Tukey’s adjustment for a family of five estimates. Gene-level data are shown in Supplementary Fig. 7a. d, Heatmap representation of β-catenin pathway activation for TCGA HCC samples with exon 3 hotspot mutations separated into weak and strong categories. LGR5 and GLUL are individual HCC targets. ‘Hallmark’ indicates a multi-gene score calculated across 42 genes known to be activated by the accumulation of β-catenin (Hallmark wnt_b-Catenin gene set, MSigDB). P values show one-tailed t-tests. e, Boxplots show Hallmark β-catenin target gene set activation in TCGA tumors with strong or weak mutations in the docking motif, compared to those with mutations at T41 or S45. Horizontal lines show the median value, boxes show the second and third quartiles and whiskers show the range. Samples with copy number gain spanning lower-effect mutations (n = 6) were excluded. Differences between groups were not significant in a one-way ANOVA (P = 0.098). f, Enrichment of Gene Ontology terms in transcripts ranked among the most upregulated in pairwise comparisons between strong and weak CTNNB1-mutant HCC samples from TCGA. Terms were selected from the full list shown in Supplementary Table 6. Normalized enrichment scores (NES) show enrichment scores normalized to the size of the gene set. P value estimation is based on an adaptive multi-level split Monte Carlo scheme, adjusted for multiple testing using Benjamini–Hochberg correction. GSEA, gene set enrichment analysis. g, H&E tumor sections from TCGA were scored using an ordinal scale according to the level of inflammatory infiltrate from zero (no visible immune cells) to three (diffuse or nodular aggregates of immune cells), then scores were compared across patients based on β-catenin pathway mutation status. P value shows Fisher’s exact test (two-way) for differences between weak and strong groups in the fraction of patients with an immune score of zero versus one or greater. Statistical analysis was not performed on other groups.
