High-Throughput Approaches to Pinpoint Function within the Noncoding Genome

Abstract

The clustered regularly interspaced short palindromic repeats (CRISPR)-Cas nuclease system is a powerful tool for genome editing, and its simple programmability has enabled high-throughput genetic and epigenetic studies. These high-throughput approaches offer investigators a toolkit for functional interrogation of not only protein-coding genes but also noncoding DNA. Historically, noncoding DNA has lacked the detailed characterization that has been applied to protein-coding genes in large part because there has not been a robust set of methodologies for perturbing these regions. Although the majority of high-throughput CRISPR screens have focused on the coding genome to date, an increasing number of CRISPR screens targeting noncoding genomic regions continue to emerge. Here, we review high-throughput CRISPR-based approaches to uncover and understand functional elements within the noncoding genome and discuss practical aspects of noncoding library design and screen analysis.

Keywords: CRISPR; Cas9; conservation; enhancers; functional genomics; gene editing; gene expression; mutagenesis; noncoding genome; pooled screens.

Figure 1. Function in the Noncoding Genome

(A) Distribution of coding and noncoding sequences in the human genome. Noncoding sequences represent the vast majority of the human genome (Dunham et al., 2012). (B) Estimated biochemically functional portion of the human genome according to ENCODE project mapping (Dunham et al., 2012). This estimate is derived from different biochemical signatures: transcribed regions (protein-coding and noncoding such as enhancer RNAs), regions bound by proteins (e.g., transcription factors), histone modifications associated with function (e.g., histone 3 lysine 27 acetylation, H3K27Ac), DNA methylated regions (methylation of cytosine at CpG dinucleotides), three-dimensional chromosome architecture (gene-distal regions that physically interact and regulate gene expression), and DNase I hypersensitivity sites (regions of open chromatin). (C) Distribution of GWAS SNPs in the human genome. Most disease-associated SNPs are present in noncoding sequences (Maurano et al., 2012).

Figure 2. Different Pooled CRISPR Screening Modalities

(A)CRISPR nucleases such as Cas9 can be used to disrupt coding and noncoding regions by making use of NHEJ to introduce insertion or deletions (indels) in the sequence of interest (CRISPRn) (Cong et al., 2013; Mali et al., 2013a). (B) Catalytically inactive Cas9 (dCas9) can be fused to cytidine deaminase (e.g., rat APOBEC1) to introduce single-nucleotide C-to-T transitions in the target sequence (Komor et al., 2016). (C) dCas9 can be applied to achieve transcriptional repression (CRISPRi). dCas9 can physically repress transcription through steric hindrance (Gilbert et al., 2013; Qi et al., 2013) or, alternatively, can be fused to repressor domains, e.g., the Krüuppel-associated box (KRAB) domain (Gilbert et al., 2013). (D) dCas9 can be used to activate transcription through different strategies (CRISPRa). Examples are dCas9 fused to the Herpex simplex virus protein 64 (VP64) transcriptional activator alone (Mali et al., 2013b) or combined with other activators, e.g., p65 and Epstein-Barr virus R transactivator (Rsa) (Chavez et al., 2015), modified gRNA-MS2 that recruits MCP fused to additional activators such as p65, and heat shock factor 1 (HSF1) (Konermann et al., 2015). (E–G) The CRISPR system can also be used to introduce epigenetic modifications. (E) dCas9 fused to p300 (Hilton et al., 2015; Klann et al., 2017) or histone deacetylase 3 (HDAC3) (Kwon et al., 2017) can perform targeted acetylation or de-acetylation to activate or inhibit transcription, respectively. Purple circles indicate the acetyl group. (F) dCas9 fused to DNA methylases (e.g., DNMT3A) (Lei et al., 2017; Liu et al., 2016; Vojta et al., 2016) or demethylases (TET1) (Liu et al., 2016) introduces or removes methyl groups at CpG dinucleotides (silencing of gene expression), respectively. Orange hexagons in (F) and (G) indicate methyl groups. (G) dCas9 fused to lysine-specific histone demethylase 1 (LSD1) (Kearns et al., 2015) catalyzes the demethylation of histone 3 lysine 4 and 9 (H3K4 and H3K9), resulting in transcriptional repression.

Figure 3. PAM Distribution across the Human Genome

(A) Schematic of the distribution of GC-rich and AT-rich protospacer-adjacent motifs (PAMs) in a single gene. Distribution of target sites/PAMs is computed as the inverse of the median distance between PAMs for SpCas9 (NGG) and LbCpf1 (TTTN). The PAM distribution of a particular nuclease is an important parameter to consider when targeting specific noncoding elements. (B) Median distance (in bp) between PAMs of SpCas9 (NGG) and LbCpf1 (TTTN). Although SpCas9 and LbCpf1 PAMs display an even distribution across the human genome, the SpCas9 PAM is more frequent in GC-rich regions such as promoters, whereas the LbCpf1 PAM (TTTN) is more represented in AT-rich regions such as introns (Canver et al., 2017a).

Figure 4. Pooled CRISPR Screen Workflow and Phenotypic Selection

In pooled CRISPR screens, gRNAs are synthesized, cloned, and constructed as a pool. Typically, viral transduction is performed at a low multiplicity of infection so that each cell receives one viral particle. Viral integration into the genome enables amplification of the gRNA cassette through PCR and readout through next-generation sequencing. During readout, the abundance of the different gRNA cassettes is quantified and a differential analysis of gRNA abundance between samples is performed. For example, gRNAs targeting coding or noncoding regulatory regions of genes that are essential for cell survival will drop out in a pooled screen (Shalem et al., 2014; Wang et al., 2014). In screens targeting noncoding regions that regulate transcription of the gene of interest, gRNAs that alter transcription factor binding motifs will display altered abundance when sorting for gene expression (Canver et al., 2015). By contrast, gRNAs that confer resistance to drugs or an aggressive in vivo metastatic phenotype will be more abundant (Sanjana et al., 2016; Chen et al., 2015; Shalem et al., 2014; Wang et al., 2014). Pooled screens have also been paired with single-cell RNA-seq to dissect genetic networks (Adamson et al., 2016; Dixit et al., 2016; Jaitin et al., 2016).

Figure 5. Consistent Enrichment/Depletion in Pooled CRISPR Screens

The analysis of pooled CRISPR screens is based on the abundance of gRNAs from next-generation sequencing of the genomically integrated gRNA cassette across samples. Enrichment and/or depletion scores are calculated by comparing the abundance after phenotypic selection with an earlier, pre-selection time point. Red arrows: gRNAs displaying enrichment across samples in the screen; blue arrows: gRNAs not showing changes in abundance across samples. (A) In pooled screens targeting coding genes, the enrichment/dropout of multiple gRNAs targeting the same gene (consistent effects) enables identification of potential candidate genes. Genes where gRNAs are not consistently enriched/depleted are not chosen for further validation and analysis. (B) In noncoding screens, candidate functional sequences are identified by enrichment or dropout of distinct gRNAs that target within a defined window.