Transposable elements drive widespread expression of oncogenes in human cancers

Abstract

Transposable elements (TEs) are an abundant and rich genetic resource of regulatory sequences^1-3. Cryptic regulatory elements within TEs can be epigenetically reactivated in cancer to influence oncogenesis in a process termed onco-exaptation⁴. However, the prevalence and impact of TE onco-exaptation events across cancer types are poorly characterized. Here, we analyzed 7,769 tumors and 625 normal datasets from 15 cancer types, identifying 129 TE cryptic promoter-activation events involving 106 oncogenes across 3,864 tumors. Furthermore, we interrogated the AluJb-LIN28B candidate: the genetic deletion of the TE eliminated oncogene expression, while dynamic DNA methylation modulated promoter activity, illustrating the necessity and sufficiency of a TE for oncogene activation. Collectively, our results characterize the global profile of TE onco-exaptation and highlight this prevalent phenomenon as an important mechanism for promiscuous oncogene activation and ultimately tumorigenesis.

Fig. 1:. The TE onco-exaptation landscape across cancer types.

a, Frequency of onco-exaptation events per tumor across cancer types. Donut plot reports the percent of tumor samples with at least one event. b, Enrichment of TE class in onco-exapted TEs across cancer types. c, A series of boxplots that highlight the distribution of the total number of tumor samples per candidate that is present in a certain number of cancer types. We have zoomed in on 1–11 so that the distribution can be more clearly seen. Each box represents the median and interquartile range, and the whiskers are 1.5× the IQR. Below each boxplot, we have labeled the number of candidates. We have also labeled all the outlier candidates. d, The top 10 most prevalent onco-exaptation candidates are presented. The left-most panel gives the TE-oncogene candidate label as well as a diagram of the transcript structure of the candidate. The next two panels display the number of tumor samples each candidate is present in as well as the distribution of the candidate across cancer types. The “Total Expression” panel displays the expression of the oncogene across all the tumor samples as grey dots, and the samples with the onco-exaptation candidate are highlighted in red. The “Fraction Expression” panel displays a boxplot of the percent of total expression of the oncogene contributed by the onco-exaptation candidate across the samples in which the candidate is present. Each box represents the median and interquartile range, and the whiskers are 1.5× the IQR.

Fig. 2:. TEs provide bona fide promoters for oncogenes in lung cancer cell lines.

a, CAGE-seq profile of H727 across onco-exaptation candidates (ARID3A & SYT1) visualized on WashU Epigenome Browser. Signals in CAGE-seq represent TSS locations. b, CAGE-seq and epigenetic profiles of the AluJb TE in the H1299 and H838. Signal in ATAC-seq represent open chromatin regions. Grey bars in the BS-seq track represent CpG locations while the height of blue bars indicate methylation %. c, Luciferase assays for transcriptional activity of various TE arrangements in H1299 (left) and H838 (right) (n = 3 independent experiments). d, Luciferase assays for promoter activity in H1299 (left) and H838 (right) with mutagenized transcription factor motifs in AluJb-P (n = 3 independent experiments). c, d, P values were derived from two-tailed Welch t test. All data are represented as means ± standard error (SE).

Fig. 3.. AluJb drives LIN28B expression and contributes to oncogenesis in lung cancer cell lines.

a, Schematic describing sgRNA locations and sequence targets within AluJb-P and LIN28BP. b, Cropped Western blot for LIN28B protein in H1299 (top) and H838 (bottom) CRISPR clones. This experiment was repeated twice with similar results. c, Relative let-7a, let-7b, and let-7g miRNA levels compared to WT in CRISPR-knockout clones of H1299 (n = 4 independent experiments) and H838 (n = 3 independent experiments) as measured by qPCR. d, The effect of AluJb-P or LIN28BP deletion on cell growth rate as determined by CCK-8 assay in H1299 and H838 cells (n = 3 independent experiments). e, The effect of AluJb-P or LIN28BP deletion on cell migration in H1299 (top) and H838 (bottom) as measured by scratch migration assay (n = 3 independent experiments). f, Tumor growth of H1299 WT and H1299 CRISPR-knockout clones injected in nude mouse. Resected tumors of WT and LIN28BP #1 xenografts. g, Cropped Western blot (repeated twice with similar results) of re-expression of human FLAG-LIN28B or AluJb-LIN28B in AluJb KO clones and its effect on relative let-7 miRNA levels (number of independent experiments indicated in figure as n) and growth rate (n = 3 independent experiments). d,e,g, P values from CCK-8 growth assays and scratch migration assays were derived from comparing to WT with two-tailed Welch t test. All data are represented as means ± SE.

Fig. 4.. Targeted DNA methylation dynamics uncover epigenetic control of AluJb promoter activity.

a Schematics illustrating CRISPR-SunTag models for targeted de/methylation of AluJb. DNMT3A was recruited to AluJb loci in H1299 to increase methylation. b, TET1CD was recruited to AluJb in K562 to remove DNA methylation from the TE. c, Methylation levels of AluJb in WT and CRISPR-SunTag-DNMT3A clones of H1299 measured by BSPCR-seq. d, Relative abundance of LIN28B in H1299 CRISPR-SunTag-DNMT3A Clone #1 (left) and Clone #2 (right) compared to WT as measured by qPCR (n = 3 independent experiments) and cropped Western blot (repeated twice with similar results). P values were derived from two-tailed Welch t test. All data are represented as means ± SE. e, Methylation levels of AluJb in WT and CRISPR-SunTag-TET1CD clones of K562. f, Cropped Western blot (repeated twice with similar results) illustrating the presence of larger LIN28B protein, similar size as AluJb-LIN28B in H1299 and H838, in K562 CRISPR-SunTag-TET1CD clones.