Genomic hallmarks and therapeutic targets of ribosome biogenesis in cancer

Abstract

Hyperactive ribosome biogenesis (RiboSis) fuels unrestricted cell proliferation, whereas genomic hallmarks and therapeutic targets of RiboSis in cancers remain elusive, and efficient approaches to quantify RiboSis activity are still limited. Here, we have established an in silico approach to conveniently score RiboSis activity based on individual transcriptome data. By employing this novel approach and RNA-seq data of 14 645 samples from TCGA/GTEx dataset and 917 294 single-cell expression profiles across 13 cancer types, we observed the elevated activity of RiboSis in malignant cells of various human cancers, and high risk of severe outcomes in patients with high RiboSis activity. Our mining of pan-cancer multi-omics data characterized numerous molecular alterations of RiboSis, and unveiled the predominant somatic alteration in RiboSis genes was copy number variation. A total of 128 RiboSis genes, including EXOSC4, BOP1, RPLP0P6 and UTP23, were identified as potential therapeutic targets. Interestingly, we observed that the activity of RiboSis was associated with TP53 mutations, and hyperactive RiboSis was associated with poor outcomes in lung cancer patients without TP53 mutations, highlighting the importance of considering TP53 mutations during therapy by impairing RiboSis. Moreover, we predicted 23 compounds, including methotrexate and CX-5461, associated with the expression signature of RiboSis genes. The current study generates a comprehensive blueprint of molecular alterations in RiboSis genes across cancers, which provides a valuable resource for RiboSis-based anti-tumor therapy.

Keywords: drug response; impaired ribosome biogenesis checkpoint; pan-cancer multi-omics; ribosome biogenesis; therapeutic target.

Figure 1

Development and validation of the approach to quantify ribosome biogenesis (RiboSis) activity. (A) Schematic of RiboSis in human cells adapted from [27]. rDNA: ribosomal DNA; rRNA: ribosomal RNA; RNA pol Ι/II/III: RNA polymerase Ι/II/III; snoRNA: small nucleolar RNA; RP: ribosome protein; RPS: ribosomal protein small subunits; RPL: ribosomal protein large subunits; 40S: small 40S ribosomal subunits; 60S: large 60S ribosomal subunits; 80S: mature 80S ribosome. (B) Summary of cancer dependencies of RiboSis genes. Left, the proportion of all 331 RiboSis genes that were essential genes, non-essential genes and undefined genes; Right, the proportion of RiboSis genes involved in each substep and the number of genes involved is marked on the right. (C) Workflow for evaluating RiboSis activity. The input includes the gene expression matrix and the RiboSis gene set, and the output is the RiboSis activity score of each sample. g: a specific gene; N: the total number of genes in the gene expression matrix; S: a specific sample; n: the total number of samples in the gene expression matrix; ES: enrichment score; G: the RiboSis gene set; r: the rank of a specific gene; N_G: the total number of RiboSis genes. A detailed description of this workflow is available in the supplementary methods. (D) The scatter diagram shows the correlation between RiboSis activity (calculated by the above workflow) and the protein abundance of RRP1 (obtained from CPTAC) in each patient with breast invasive carcinoma. (E) Boxplot showing the difference in the protein abundance of FBL between breast invasive carcinoma samples with high and low RiboSis activity. (F) tSNE representation (Left) and the difference (Right) in colorectal cancer cells’ RiboSis activity after treatment with different doses of 5-fluorouracil. ^*P < 0.05 and ^****P < 0.0001.

Figure 2

Hyperactive ribosome biogenesis (RiboSis) in human cancers. (A) Violin and boxplot showing RiboSis activity across 33 human cancer types. (B) The paired point plot shows the difference in the average activity of RiboSis between tumor and normal samples across 26 cancer types with a sufficient sample size. ^***P < 0.001. (C) Heatmap showing RiboSis activity among malignant cells, immune cells and stromal cells across various human cancer types. (D) tSNE representation of RiboSis activity of different cell types in ESCC single-cell RNA-seq data. (E) Heatmap showing RiboSis activity among different cell subtypes across various human cancer types.

Figure 3

Upregulated ribosome biogenesis (RiboSis) genes in human cancers. (A) ROC curve showing the performance of RiboSis activity in distinguishing primary cancer samples from normal samples in READ, LUSC, COAD and LUAD. ROC: receiver operating characteristic; AUC: area under the ROC curve. (B) Progression-free interval (PFI) between primary cancer patients with high and low RiboSis activity. The number of patients is enclosed in brackets. (C) Hazard ratio between patients with high and low RiboSis activity across different cancer types. A Cox proportional hazards model was used to calculate the hazard ratio. The number of patients and 95% confidence interval (CI) of the hazard ratio are enclosed in brackets. (D) Bar diagram showing the proportion of genome-wide differentially expressed genes (up) and differentially expressed RiboSis genes (down) across 26 cancer types. (E) Heatmap of differentially expressed RiboSis genes across 26 cancer types. Each column represents a different RiboSis gene and each row represents a different cancer type. The shade of color represents the degree to which the expression has changed.

Figure 4

Characterization of somatic genetic alterations in ribosome biogenesis (RiboSis) genes. (A) The proportion of patients with genetic alterations in RiboSis genes across 33 cancer types. (B) Boxplot showing the ratio of patients with genetic alterations in RiboSis genes among SNVs, deletions and amplifications. (C) Area diagram showing somatic amplification and deletion on 22 autosomes at the pan-cancer level. The shade of color represents the number of cancer types, and the sites prone to alteration in more cancers are darker. The top 10 (5) RiboSis genes more likely to be amplified (deleted) are labeled. (D) Density plot showing the G-score of significantly altered peaks in RiboSis genes. Boxplot showing the difference in the number of RiboSis genes with high level (G-score > 1.0) genetic alterations between amplification and deletion. (E) Bubble diagram showing the RiboSis genes with the G-score of the top 40 across various cancer types. (F) Histogram showing the enrichment ratio of RiboSis genes that reside in the amplification peaks (identified by GISTIC2, q < 0.25). The enrichment ratio: fractions of RiboSis genes compared to non-RiboSis genes that reside in the amplification peaks. Significant amplification enrichments are detected with P < 0.05 (Fisher’s exact test). ^****P < 0.0001.

Figure 5

Characterization of ribosome biogenesis (RiboSis) gene-based therapeutic targets. (A) Sankey diagram showing cancer genetic dependencies of RiboSis genes based on target development level and Pubtator score. The width of the bar is proportional to the number of RiboSis genes at each corresponding level. (B) Bar diagram showing the number of RiboSis genes defined as potential targets in 7–11, 12–16 or > 16 cancer types at upregulated, downregulated, amplification, deletion or SNV levels. RiboSis genes defined as potential targets in more than 16 cancer types are labeled. (C) Boxplot showing the difference in RiboSis activity between samples with and without TP53 mutation at the pan-cancer level. MUT: with TP53 mutation (n = 3248); WT: without TP53 mutation (n = 5212). ^****P < 0.0001. (D) Progression-free interval (PFI) among patients with high and low RiboSis activity and with and without TP53 mutation. (E) Network diagram showing the top 10 ranked RiboSis genes within each cancer type. The size of the node is scaled according to the degree of its connection.

Figure 6

Putative drugs against ribosome biogenesis (RiboSis). (A) The workflow for identifying significant RiboSis gene-drug pairs. The drug sensitivity of patients is predicted by machine learning (oncoPredict). Combined the expression of RiboSis genes with the predicted drug sensitivity of each patient to identify significant RiboSis gene-drug pairs (|r| > 0.8, P < 0.05). (B) Point diagram showing the number of significant RiboSis gene-drug pairs across different cancer types. (C) Sankey diagram showing the enrichment result of significantly negatively correlated RiboSis genes-drug pairs. The width of the bar is proportional to the number of significantly negatively correlated RiboSis gene-drug pairs. Left: RiboSis gene enriched substeps of ribosome biogenesis; Right: drug target pathway. (D) Volcano plot showing the results of differential drug response analysis between cell lines with high and low RiboSis activity. Each point represents a drug. Drugs reported to be able to inhibit RiboSis are labeled. (E) Venn diagram showing drugs with differential drug responses in RiboSis and its five substeps. Twenty-three drugs that were identified as significantly differential drugs in all differential drug response analyses are labeled.