Enhancer RNA Transcription Is Essential for a Novel CSF1 Enhancer in Triple-Negative Breast Cancer

Abstract

Enhancers are critical regulatory elements in the genome that help orchestrate spatiotemporal patterns of gene expression during development and normal physiology. In cancer, enhancers are often rewired by various genetic and epigenetic mechanisms for the activation of oncogenes that lead to initiation and progression. A key feature of active enhancers is the production of non-coding RNA molecules called enhancer RNAs, whose functions remain unknown but can be used to specify active enhancers de novo. Using a combination of eRNA transcription and chromatin modifications, we have identified a novel enhancer located 30 kb upstream of Colony Stimulating Factor 1 (CSF1). Notably, CSF1 is implicated in the progression of breast cancer, is overexpressed in triple-negative breast cancer (TNBC) cell lines, and its enhancer is primarily active in TNBC patient tumors. Genomic deletion of the enhancer (via CRISPR/Cas9) enabled us to validate this regulatory element as a bona fide enhancer of CSF1 and subsequent cell-based assays revealed profound effects on cancer cell proliferation, colony formation, and migration. Epigenetic silencing of the enhancer via CRISPR-interference assays (dCas9-KRAB) coupled to RNA-sequencing, enabled unbiased identification of additional target genes, such as RSAD2, that are predictive of clinical outcome. Additionally, we repurposed the RNA-guided RNA-targeting CRISPR-Cas13 machinery to specifically degrade the eRNAs transcripts produced at this enhancer to determine the consequences on CSF1 mRNA expression, suggesting a post-transcriptional role for these non-coding transcripts. Finally, we test our eRNA-dependent model of CSF1 enhancer function and demonstrate that our results are extensible to other forms of cancer. Collectively, this work describes a novel enhancer that is active in the TNBC subtype, which is associated with cellular growth, and requires eRNA transcripts for proper enhancer function. These results demonstrate the significant impact of enhancers in cancer biology and highlight their potential as tractable targets for therapeutic intervention.

Keywords: CRISPR-Cas9; Cas13; breast cancer; dCas9-KRAB; eRNA; enhancer; gene expression; ovarian cancer.

Figure 1

eRNA transcription and its associated epigenomic modifications identify a novel eRNA producing enhancer upstream of CSF1 in triple-negative breast cancer. (A). GRO-seq (red peaks indicate transcription in the sense strand and blue peaks indicate transcription in the anti-sense strand) and H3K27ac ChIP-seq browser tracks showing eRNA production and H3K27ac enrichment at the CSF1 enhancer (highlighted in blue) across multiple breast cancer cell lines. (B). RT-PCR of human tumors collected from TNBC and non-TNBC patients to confirm the expression and detection of eRNA (top) and mRNA (middle) transcripts. GAPDH (bottom) expression is shown as a loading control. The positive control (MDA-MB-436 cDNA), reverse transcriptase negative control, and no template control are shown on the right. C and D. qRT-PCR showing changes in CSF1 mRNA expression (C) and CSF1 eRNA expression (D) across cell lines grouped by normal mammary epithelial cells (gray), non-TNBC cancer cell lines (yellow), and TNBC specific cell lines (green). Each point represents the fold change (relative to the normal group) for an individual replicate and the bar shows the mean value for each group. Significance determined by a two-sided t-test comparing each KO line to WT (* p < 0.05, ** p < 0.01). (E). Boxplot comparison of CSF1 mRNA expression across the METABRIC cohort of breast cancer patients, grouped by subtype. Each point represents an individual patient sample. Significance determined by a two-sided t-test comparing each KO line to WT (*** p < 0.001). (F). Boxplot of mean ATAC-seq scores demonstrating increased chromatin accessibility of the novel CSF1 enhancer in basal/claudin-low (TNBC) subtype breast cancer patients as compared to other subtypes. Each point represents the mean ATAC-seq score for a single library across the 5 bins spanning the CSF1 enhancer region. Asterisks indicate significant difference when comparing each group to TNBC (Wilcoxon rank-sum test, *** p < 0.001). (G). Boxplot of mean ATAC-seq scores overlapping the CSF1 transcription start site (TSS) demonstrating consistent accessibility across breast cancer subtypes within the TCGA cohort. ** p < 0.01, The designation of NS represents statistically not significant.

Figure 2

Knockout of the CSF1 enhancer confirms cis-regulatory activity and leads to reduced cell proliferation, migration, and colony formation. (A). Diagram showing the use of Cas9 and flanking sgRNAs to target and cut regions flanking the CSF1 enhancer, resulting in a ~2100 bp deletion. (B). Genotyping was performed by PCR amplification both within the WT allele and across the deletion site to create PCR products representative of genomic deletion (top). Genotyping PCR of CSF1e-KO clones confirms multiple heterozygous knockouts (bottom). (C). qRT-PCR analysis showing that CSF1 enhancer RNAs (both positive and negative strand) and the CSF1 mRNA are significantly reduced in heterozygous KO clones. Bars show mean fold change (relative to WT cell line) in CSF1 eRNAs (negative strand left, positive strand middle) or mRNA (right); points represent fold change values for individual replicates. Error bars are standard error of the mean. n = 4. Statistical significance determined by a two-sided t-test comparing each KO line to WT (** p < 0.01, *** p < 0.001) (D). Proliferation assay shows a significant reduction in cell growth following deletion of the CSF1 enhancer. Each point shows the mean absorbance reading for each cell line across all four replicates. Error bars standard error of the mean. Significance determined by a two-sided t-test comparing each KO line to WT (** p < 0.01, *** p < 0.001). (E). Migration assays of KO clones demonstrate a reduced potential for migration when the CSF1 enhancer has been deleted. Brightfield images (left) show the presence of stained cells on migration columns. Bars represent the mean number of cells counted (relative to WT cell line) across all replicates per cell line (right). Individual points show number of cells counted per field of view captured. Error bars are standard error of the mean. n = 4 with 10 fields counted per replicate (*** p < 0.001). (F). Colony formation assays show a reduction in ability to form colonies from a single cell following deletion of the CSF1 enhancer. Representative well images show a decrease in colony formation for CSF1e-KO cell lines (left). Bar plots show results of mean colony counts for each cell line across all four replicates (right). Individual points are shown for each replicate. Significance determined by a two-sided t-test comparing each KO line to WT (* p < 0.05).

Figure 3

Epigenetic silencing of the CSF1 enhancer reduces CSF1 mRNA expression and identifies target genes significant to patient outcome. (A). Diagram showing the approach for CRISPRi epigenetic silencing of the CSF1 enhancer. A construct containing four guide RNAs targeting distinct areas of the CSF1 enhancer was transduced into dCas9-KRAB expressing MDA-MB-436 cells, facilitating targeting of the enhancer for epigenetic repression. (B). Western blot showing expression of dCas9-KRAB protein in cell lines used for this work. β-tubulin is shown below as a loading control. (C). Genotyping PCRs demonstrating insertion of the guide RNA expression vector within TNBC cell lines. (D). qRT-PCR showing changes in CSF1 eRNA or mRNA expression resulting from dCas9-KRAB targeting of the CSF1 enhancer. CSF1-A and CSF1-B are independent cell lines (biological replicates) expressing dCas9-KRAB and the same sgRNA expression vector. Bars represent mean fold change (relative to scrambled sgRNA expressing cell line) and points show individual fold change values for each replicate. Significance determined by a two-sided t-test comparing each KO line to WT (** p < 0.01, *** p < 0.001). n = 4. (E). RNA-seq heatmap displaying the changes in gene expression resulting from CRISPRi perturbation of the CSF1 transcription start site (top) or enhancer (bottom). CSF1 TSS or enhancer targeting replicates shown on the left, n = 4. Negative control scrambled replicates shown on the right, n = 6. (F). Volcano plots showing differential gene expression (p-value versus fold change difference) results by CRISPRi perturbation of the CSF1 transcription start site (left) or enhancer (right). Genes shown in red are significantly downregulated. (G). Forest plot showing hazard ratios for the genes downregulated in the CSF1 TSS targeting samples. (H). Kaplan–Meier plot showing patient overall survival probability separated by high or low expression (median centered) of RSAD2 using the METABRIC patient cohort. (I) Kaplan–Meier plot showing patient overall survival probability separated by high or low expression of CSF1 and RSAD2.

Figure 4

Cas13-targeted disruption of the CSF1 eRNA transcript causes reduction in CSF1 mRNA expression. (A). Vector map displaying modifications to pC0046-EF1a-PspCas13b-NES-HIV vector to allow for eRNA targeting within the nucleus. (B). Western blot showing expression of the HA tagged PspCas13b protein in MDA-MB-436 cells. (C). Diagram showing approach for Cas13b-mediated RNA degradation. Guide RNAs were pooled and transiently transfected to enable targeting of the either the CSF1 eRNAs or the CSF1 mRNA. (D). Predicted structure of the positive strand eRNA demonstrating potential secondary structures. Sites selected for sgRNA targeting labeled. (E). qRT-PCR displaying the relative fold change in CSF1 mRNA expression when targeting either the CSF1 mRNA or CSF1 eRNAs with PspCas13b. Each bar represents the mean fold change (relative to a non-targeting guide RNA) and each point shows the individual fold change per replicate. Error bars show standard error of the mean. Significance determined by a two-sided t-test comparing each KO line to WT (F). qRT-PCR displaying the relative fold change in CSF1 mRNA expression when targeting the CSF1 eRNA (positive strand) with RfxCas13d in two independent cell lines (biological replicates). Each bar represents the mean fold change (relative to a non-targeting guide RNA) and each point shows the individual fold change per replicate. Error bars show standard error of the mean. Significance determined by a two-sided t-test comparing each KO line to WT (G). qRT-PCR showing effects of siRNA-mediated eRNA knockdown on CSF1 mRNA expression. Each bar represents the mean fold change (relative to a scrambled siRNA) and each point shows the individual fold change per replicate. Error bars show standard error of the mean. (* p < 0.05, ** p < 0.01, *** p < 0.001).

Figure 5

The CSF1 enhancer is active in ovarian cancer patients and requires eRNA transcripts for its activity. (A). A Boxplot showing the expression of CSF1 (RSEM) across all cancer types of the TCGA dataset. Cancer types are arranged by median in descending order. Inset: Kaplan–Meier plot showing patient overall survival probability separated by high or low expression of CSF1 in ovarian cancer. (B). Browser tracks displaying H3K27ac (green) and BRD4 (blue) ChIP-seq overlapping the CSF1 enhancer and TSS in OVCAR3 High-grade serous ovarian cancer cells. The CSF1 enhancer is highlighted in light blue and BRD4-bound H3K27ac peaks are indicated below. (C). Top left: Western blot showing expression of dCas9-KRAB protein in OVCAR3 cells. β-tubulin is shown below as a loading control. Top right: Diagram showing the targeting of dCas9-KRAB to either the enhancer (purple) or the promoter (orange). Bottom: qRT-PCR displaying the relative fold change in CSF1 eRNA (positive and negative strand transcripts) or mRNA expression when targeting the CSF1 enhancer or promoter with dCas9-KRAB. Each bar represents the mean fold change (relative to a scrambled guide RNA), and each point shows the individual fold change per replicate. Error bars show standard error of the mean. Significance determined by a two-sided t-test comparing each promoter or enhancer sample to scrambled control (* p < 0.05, ** p < 0.01), NS represents statistically not significant. (D). Top left: Western blot showing expression of the HA tagged PspCas13b protein in OVCAR3 cells. Top right: Diagram showing the targeting of Cas13b to either the enhancer RNA (purple) or the mRNA (orange). Bottom: qRT-PCR displaying the relative fold change in CSF1 mRNA expression when targeting the CSF1 eRNA or mRNA with Cas13b. Each bar represents the mean fold change (relative to a scrambled guide RNA), and each point shows the individual fold change per replicate. Error bars show standard error of the mean. Significance determined by a two-sided t-test comparing each promoter or enhancer sample to scrambled control (** p < 0.01, *** p < 0.001).