High-resolution interrogation of functional elements in the noncoding genome

Abstract

The noncoding genome affects gene regulation and disease, yet we lack tools for rapid identification and manipulation of noncoding elements. We developed a CRISPR screen using ~18,000 single guide RNAs targeting >700 kilobases surrounding the genes NF1, NF2, and CUL3, which are involved in BRAF inhibitor resistance in melanoma. We find that noncoding locations that modulate drug resistance also harbor predictive hallmarks of noncoding function. With a subset of regions at the CUL3 locus, we demonstrate that engineered mutations alter transcription factor occupancy and long-range and local epigenetic environments, implicating these sites in gene regulation and chemotherapeutic resistance. Through our expansion of the potential of pooled CRISPR screens, we provide tools for genomic discovery and for elucidating biologically relevant mechanisms of gene regulation.

Figure 1. CRISPR mutagenesis of noncoding regions flanking three genes involved in BRAF inhibitor resistance

A) Design of sgRNA libraries targeting 100 kb 5` and 100 kb 3` of a gene. The sgRNAs are array synthesized and cloned into a lentiviral vector. BRAF mutant cells are transduced with the pooled lentivirus and treated with vemurafenib (Vemu) or DMSO (Control). A deep sequencing readout identifies sgRNAs enriched after treatment with vemurafenib. Scatterplot of normalized read counts (average of the 2 infection replicates) for B: NF1, C: NF2, D: CUL3 sgRNAs at Day 0 (x axis) and Day 14 (y axis) Read counts from control (gray) and vemurafenib-treated cells (red) are shown relative to 4 standard deviations of the control cell distribution (dotted line) with the percentage of enriched sgRNAs in vemurafenib (>4 s.d.). E) Distribution of log₂ ratio of the normalized read count for each sgRNA in vemurafenib to its normalized read count in control (minimum of the 2 infection replicates). F) All CUL3 sgRNAs plotted by hg19 coordinates and the percent expression of the two most highly expressed CUL3 isoforms (Primary, Alt.). For vemurafenib-enriched sgRNAs, the log₂ enrichment over control (minimum value of 2 replicate screens) is plotted (red); non-enriched sgRNAs are indicated in blue. G) Percent of enriched sgRNAs by genomic category.

Figure 2. Characteristics of functional noncoding elements at the CUL3 locus

A) Plot of 3C interaction frequencies with the CUL3 promoter in A375 cells. Data points represent independent libraries generated with BglII, EcoRI, and HindIII restriction enzymes. The grey curve shows a smoothed estimate of interaction frequency. B) Average enrichment of sgRNAs (log₂ ratio of vemurafenib/DMSO reads) near 3C sites with specified minimum interaction frequency with the CUL3 promoter (43). C) An example of enriched sgRNAs (red) that overlap with a melanoma-specific region of open chromatin. ATAC-seq in A375 melanoma (orange), MCF-7 breast cancer (purple) and U-87 glioblastoma (blue) and Melanoma DNAse I hypersensitivity sequencing (green, ENCODE/Colo-829). Loci investigated relative to CUL3 is shown at top (yellow). Scale bar: 500 bp. D) Fold enrichment of enriched sgRNAs near ATAC-seq open chromatin peaks in melanoma, breast cancer and glioblastoma cell lines. E) Fold enrichment of enriched sgRNAs near DNAse I HS-seq open chromatin peaks in melanoma, breast cancer and glioblastoma cell lines. F) An example of enriched sgRNAs (red) that coincide with regions that show greater primate-specific conservation than placental mammal and vertebrate conservation. Loci investigated relative to CUL3 is shown at top (yellow). Scale bar: 200 bp. G) Fold enrichment of enriched sgRNAs near phastCons peaks in primates, placental mammals and vertebrates.

Figure 3. Noncoding mutations impact CUL3 expression and histone modifications

A) Criteria selecting 25 sgRNAs targeting noncoding regions for validation. B) CUL3 RNA expression (normalized to non-targeting sgRNAs) after transduction with lentivirus expressing non-targeting (triangles), noncoding region-targeting (colored circles) and coding region-targeting (squares) sgRNAs. C) Relationship between CUL3 expression and cell survival after vemurafenib. Linear fit is to noncoding sgRNAs only (r_noncoding = −0.54, p = 0.005) and does not include coding region-targeting or non-targeting sgRNAs. D) Percent change in average H3K4me3 chromatin immunoprecipitation (ChIP) for all validation sgRNAs within 1 kb of the transcription start site (TSS) of CUL3 (left). Percent change in average H3K27ac and average H3K4me2 ChIP for all validation sgRNAs >1 kb from the TSS of CUL3 (right).

Figure 4. Cas9 mutagenesis disrupts predicted transcription factors and DNA binding proteins at target sites of vemurafenib-enriched sgRNAs

A–D) sgRNA target locations in relation to predicted binding sites. E–H) Change in transcription factor/DNA binding protein occupancy around cleavage site and change in CUL3 expression. Both measurements are normalized to cells transduced with non-targeting sgRNAs.