Downstream-of-gene (DoG) transcripts contribute to an imbalance in the cancer cell transcriptome

Abstract

Downstream-of-gene (DoG) transcripts are an emerging class of noncoding RNAs. However, it remains largely unknown how DoG RNA production is regulated and whether alterations in DoG RNA signatures exist in major cancers. Here, through transcriptomic analyses of matched tumors and nonneoplastic tissues and cancer cell lines, we reveal a comprehensive catalog of DoG RNA signatures. Through separate lines of evidence, we support the biological importance of DoG RNAs in carcinogenesis. First, we show tissue-specific and stage-specific differential expression of DoG RNAs in tumors versus paired normal tissues with their respective host genes involved in tumor-promoting versus tumor-suppressor pathways. Second, we identify that differential DoG RNA expression is associated with poor patient survival. Third, we identify that DoG RNA induction is a consequence of treating colon cancer cells with the topoisomerase I (TOP1) poison camptothecin and following TOP1 depletion. Our results underlie the significance of DoG RNAs and TOP1-dependent regulation of DoG RNAs in diversifying and modulating the cancer transcriptome.

Fig. 1.. Readthrough transcription is prevalent in major cancers.

(A) Number and extension strength of the DoG RNAs identified by DoGFinder (33) in breast, colon, and liver nonneoplastic (n = 22) and tumor (n = 22) samples from TCGA. The extension strength is shown in log₁₀ scale. Box plots enclose values between first and third quartiles, midlines show medians, and whiskers extend to data points within 1.5 the interquartile range from the box; outliers are shown. Statistical significance was determined by one-way Wilcoxon rank sum test (alternative = “less”). P values include 0.99, 5.1 × 10⁻⁷, and 3 × 10⁻⁵ for NTs versus breast, colon, and liver tumors, respectively. (B) Volcano plots of the differentially expressed DoG RNAs (log₂ FC > 0.58 or log₂ FC < −0.58, q < 0.05) in breast, colon, and liver tumors compared with paired NTs. The significant down-regulated (blue) and up-regulated (pink) DoG RNAs are denoted, and the qPCR-validated DoG RNAs in (C) are denoted as darker shades of blue and pink. (C) qRT-PCR analysis of the denoted mRNAs and their respective DoG RNAs normalized to 18S ribosomal RNA (rRNA) in breast (left), colon (middle), and liver (right) tumors. Expression levels in tumors are relative to the levels in paired NTs. (D) Kaplan-Meier plots showing high versus low DoG RNA expression levels associated with overall patient survival in patients with breast (left), colon (middle), and liver (right) tumors. Statistical significance was determined using a log-rank test.

Fig. 2.. DoG RNAs in colorectal carcinoma tissues are differentially expressed and associated with tumorigenic pathways.

(A) Number and extension strength of DoG RNAs identified by DoGFinder (33) in three nonneoplastic (NT, blue) and three COAD (red) tumors. The extension strength is shown in log₁₀ scale. Box plots enclose values between first and third quartiles, midlines show medians, and whiskers extend to data points within 1.5 the interquartile range from the box; outliers are shown. Statistical significance was determined by one-way Wilcoxon rank sum test (alternative = less). P values include 2 × 10⁻⁷, 2 × 10⁻⁹, and 0.7 for patients 1 to 3, respectively. (B) Volcano plot showing differentially expressed DoG RNAs (log₂ FC > 0.58 or log₂ FC < −0.58, q < 0.05) in the same three NT and three COAD tissues shown in (A). The significantly down-regulated (blue) and up-regulated (pink) DoG RNAs are shown. (C) Integrative Genomics Viewer (IGV) tracks of total RNA-seq signal in log reads per kilobase of transcript per million mapped reads (RPKM) at ASCL2 and PCK1 loci in three paired NT and COAD samples shown in (A) and (B). NT and COAD tissues are represented in blue and red, respectively. The horizontal bars define the DoG region determined by DoGFinder (33). (D) Scatter plot showing log₂ FC expression for the DoG-producing genes on the x axis and the log₂ FC DoG RNA expression on the y axis. Down-regulated and up-regulated DoG RNAs (log₂ FC > 0.58 or log₂ FC < −0.58, q < 0.05) and their host genes’ expression (silenced: log₂ FC < −0.58 and q < 0.05; not changed: −0.58 ≤ log₂ FC ≤ 0.58 and q ≥ 0.05; activated: log₂ FC > 0.58 and q < 0.05) in COAD samples relative to NT are represented in blue and red, respectively. DEDoGs, differentially expressed DoGs.

Fig. 3.. DoG RNA production is prevalent in colorectal carcinoma cell lines.

(A) DoG number and extension strength identified in FHC, HCT116, and SW480 cells by DoGFinder (33). Box plots enclose values between first and third quartiles, midlines show medians, and whiskers extend to data points within 1.5 the interquartile range from the box; outliers are shown. Statistical significance determined by one-way Wilcoxon rank sum test (alternative = less). P < 0.05. Extension strength is shown in log₁₀ scale. (B) Volcano plot of differentially expressed DoG RNAs (log₂ FC > 0.58 or log₂ FC < −0.58, q <0.05) in SW480 versus FHC and HCT116 versus FHC cells. (C) qRT-PCR analysis of DoG RNAs denoted in (B) and normalized with GAPDH. Expression levels in SW480 or HCT116 cells are relative to the levels in FHC cells. Data represent the mean and SEM that are representative of three independent experiments. P values are shown. (D) Top five (P < 0.05) Molecular Signatures Database (MSigDB) pathways for host genes showing up-regulated and down-regulated DoG RNAs in SW480 versus FHC (left) and HCT116 versus FHC (right). TNF-α, tumor necrosis factor–α; NF-κB, nuclear factor κB. (E) Scatter plots showing log₂ FC for the DoG-producing gene on the x axis and log₂ FC for the DoG RNA expression on the y axis. Down-regulated and up-regulated DoG RNAs (log₂ FC > 0.58 or log₂ FC < −0.58, q < 0.05) and their host genes’ expression (silenced: log₂ FC < −0.58 and q < 0.05; not changed: −0.58 ≤ log₂ FC ≤ 0.58 and q ≥ 0.05; activated: log₂ FC > 0.58 and q < 0.05) in SW480 versus FHC (left) and HCT116 versus FHC (right) cells. (F) Immunoblot of GAPDH and heterogeneous nuclear ribonucleoprotein (hnRNP) A2/B1 from SW480 cytoplasmic and nuclear extracts. Image is representative of three independent experiments. (G) qRT-PCR analysis of the denoted mRNAs and their respective DoG RNAs normalized with GAPDH using RNA from SW480 cytoplasmic and nuclear fractions. Data represent the mean and SEM that are representative of three independent experiments. P values are shown.

Fig. 4.. TOP1 regulates DoG RNA production in colorectal carcinoma.

(A) Box with jitter plot for TOP1 RNA-seq levels in 349 NTs and 275 COADs determined with GTEx and TCGA data from GEPIA server (32, 35, 66). Statistical significance was determined by one-way analysis of variance (ANOVA) test (*P < 0.05). (B) Immunoblot analysis of TOP1 and β-actin from paired NT and COADs from three patients. (C) Left: Dose response (0, 0.1, 1, 2.5, and 5 μM) for CPT treatment in SW480 cells. SW480 cell number was examined relative to CPT treatment for 24 hours. Right: Immunoblot analysis of TOP1 and β-actin in SW480 treated with 1.15 μM CPT or equal volume DMSO for 3 hours. P values are shown. n.s., nonsignificant. (D) Heatmap of RNA-seq distribution spanning 3-kb upstream of TSS to 3-kb downstream of TES of the CPT-specific DoG RNAs (n = 532). The log₂ ratio of RPKM is represented in SW480 cells treated with CPT versus DMSO. (E) qRT-PCR (left) and immunoblot analysis (right) of SW480 cells stably expressing control (Ctrl) or TOP1 shRNA. Data represent the mean and SEM representative of three independent experiments. P = 9.5 × 10⁻⁶. (F) Heatmap of RNA-seq distribution spanning 3-kb upstream of the TSS to 3-kb downstream of the TES of the TDR genes (n = 555). The log₂ ratio of RPKM is represented in TOP1 knockdown versus shCtrl SW480 cells. IGV tracks of RNA-seq signal (RPKM) at the EXO5 and MARVELD1 loci in SW480 cells treated with (G) DMSO versus CPT or (H) expressing Ctrl versus TOP1 shRNA. The horizontal bar defines the DoG region determined by DoGFinder (33). qRT-PCR analysis of DoG RNA induction in SW480 cells treated with (I) CPT versus DMSO or (J) shTOP1 versus shCtrl. Data represent the mean and SEM that are representative of three independent experiments. P values are shown.

Fig. 5.. RNAPII and TOP1 accumulate at the promoter- and TES-proximal regions of paused genes.

(A) Box plot showing the counts of TDRs and non-DoG–producing genes with low, medium, and high transcriptions in SW480 cells. The counts were grouped using the bottom and top quartiles into three groups, low (quartile 1), medium (quartile 2), and high (quartile 3) expressions according to the total RNA-seq reads. Box plots enclose values between first and third quartiles, midlines show medians, and whiskers extend to data points within 1.5 the interquartile range from the box; outliers are shown. (B) Schematic of TOP1-seq, a method to identify TOP1cc (22). (C) Metaplots of TOP1, TOP1-seq, and RNAPII ChIP-seq signal at TDRs (n = 555) and non-DoG–producing genes with low, medium, and high transcriptions in SW480 cells. TOP1-seq and ChIP-seq signal is represented as log₂-transformed FC of bins per million over input and spans 2-kb upstream of the TSS to 5-kb downstream of the TES. (D) Illustration depicting the pause index determination for TDR genes. The TSS pause index is defined as the ratio of PRO-seq reads at paused site 1 at the TSS (P1, spanning from −50 to +300 bp) over the PRO-seq reads at gene body (P2, spanning from +300 to end of the gene). The TES pause index is defined as the ratio of PRO-seq reads at paused site 2 on TES (P3, spanning from +500 to +1500 bp) over the PRO-seq reads at DoG region identified by DoGFinder (33) (P4, spanning from +1500 to end of the DoG defined by DoGFinder) (33). (E) Box plot showing the TSS and TES pause indices in highly expressed non-DoG–producing genes (green), SW480 DoG-producing genes (blue), and TDR genes (red) in SW480 cells. Statistical significance was determined by two-way Wilcoxon rank sum test.

Fig. 6.. TOP1 associates with RNAPII to regulate TDR genes.

(A) Coimmunoprecipitation (IP) with TOP1 and immunoglobulin G (IgG) antibodies from SW480 nuclear extracts and immunoblot analysis of RNAPII and TOP1. An image is shown that is representative of three independent experiments. (B) Left: qRT-PCR analysis of TOP1 mRNA. Data represent the mean and SEM of three independent qRT-PCR experiments. P = 1.5 × 10⁻⁶. Right: Immunoblot analysis of TOP1, RNAPII, and β-actin in SW480 cells expressing Ctrl and TOP1 shRNA. A representative image is shown that is representative of three independent experiments. (C) Heatmap of RNAPII ChIP-seq distribution spanning 3-kb upstream of the TSS to 10-kb downstream of the TES of the TDR genes (n = 555). The RNAPII ChIP-seq represents the log₂ ratio of ChIP-seq signal in TOP1 knockdown (KD) over shCtrl in SW480 cells. Box plot showing the pause index at TDR genes at two regions, TES to 3 kb and 3 to 8 kb past the TES in SW480 cells expressing shTOP1 over shCtrl. Statistical significance was determined by two-way Wilcoxon rank sum test. P = 0.032. (D) IGV tracks of RNAPII ChIP-seq signal at the IGFBP4 and ADAMTS15 loci in SW480 cells expressing Ctrl and TOP1 shRNA. (E) Heatmaps of RNAPII ChIP-seq and RNA-seq distribution spanning 3-kb upstream of the TSS to 10-kb downstream of the TES of the highly transcribed non-DoG genes. The log₂ ratio of RNA-seq signal (RPKM) is represented in shTOP1 versus shCtrl SW480 cells. The RNAPII ChIP-seq signal is represented as the log₂ ratio of ChIP-seq signal in shTOP1 over shCtrl SW480 cells. (F) Box plot showing the TSS and TES pause indices (calculated from RNAPII ChIP-seq) on TDRs genes in SW480 cells expressing Ctrl and TOP1 shRNA. Statistical significance was determined by two-way Wilcoxon rank sum test. P = 0.043 and P = 0.00035.

Fig. 7.. Working model for how DoG RNAs reshape normal and cancer transcriptomes.

Comparative analyses of DoG RNA signatures in normal versus cancer tissues provide previously unknown insights into this emerging class of differentially expressed tissue- and stage-specific ncRNAs. Dysregulated expression of DoG RNAs in breast, liver, and colon tumors is significantly correlated with poor patient survival. Up-regulated DoG RNAs are associated with DoG-producing host genes that exhibit tumor-promoting functions, and down-regulated DoG RNAs are linked to host genes involved in normal developmental and tumor-suppressor pathways. Treatment of colon cancer cells with the TOP1 poison, CPT, leads to an induction of DoG RNA production, which is consistent with DoG RNAs exhibiting a potential therapeutic benefit in patients with colon cancer. Mechanistically, we confirm that TOP1 depletion promotes DoG RNA induction by lowering the RNAPII PI at the TES and promoting RNAPII release well beyond the ends of TDR host genes.