CRISPR-Based Methods for High-Throughput Annotation of Regulatory DNA

Abstract

Developments in CRISPR/Cas9-based technologies provide a new paradigm in functional screening of the genome. Conventional screening methods have focused on high-throughput perturbations of the protein-coding genome with technologies such as RNAi. However, equivalent methods for perturbing the non-coding genome have not existed until recently. CRISPR-based screening of genomic DNA has enabled the study of both genes and non-coding gene regulatory elements. Here we review recent progress in assigning function to the non-coding genome using CRISPR-based genomic and epigenomic screens, and discuss the prospects of these technologies to transforming our understanding of genome structure and regulation.

Introduction

Mammalian genomes are primarily composed of DNA that does not code for gene sequences. This non-coding sequence is thought to integrate cellular signaling into dynamic, cell type-specific patterns of gene expression. In fact, since humans share a large proportion of their protein-coding genome with much simpler organisms, it is thought that the non-coding sequence is critical in specifying diverse and complex phenotypes. Furthermore, the vast majority of genetic variation associated with complex human disease is found within non-coding regions of the genome [1]. Thus, understanding how genomic regulatory elements in the non-coding DNA encode gene expression programs could elucidate mechanisms of disease initiation and genetic contributions to drug responses, providing an avenue to develop next-generation therapies. Large-scale research projects such as the Encyclopedia of DNA Elements (ENCODE) [2] and the Roadmap Epigenomics Project [3] have mapped the epigenetic profiles of many human cell types and revealed common epigenetic signatures of regulatory elements genome-wide. However, these large datasets cannot precisely determine which elements are functional, identify the target gene(s) of a particular element, or specify which gene networks are influenced by dynamic epigenetic changes.

From the >3 billion base pairs in the human genome, millions of putative regulatory elements have been identified through projects such as ENCODE and genome-wide association studies [4]. Given the scale of the human genome and the diversity of genetic variation therein, high-throughput approaches are needed to annotate the function of non-coding regions. Tools for genome engineering based on programmable DNA-binding platforms enable the targeted perturbation of endogenous genomic sites and can be used to assess the functional role of putative regulatory elements in their native context [5]. Although the advent of these technologies initially inspired efforts to use programmable DNA-binding proteins, such as zinc finger domains, for high-throughput functional genomics [6], the considerable protein engineering required to build these systems has limited their scalability. The recently discovered and repurposed clustered regularly interspaced short palindromic repeats (CRISPR) technology for genome editing is much more compatible with high-throughput genetic and epigenetic screening [7]. The RNA-guided nature of CRISPR-based systems permits pooled screening to test thousands of unique hypotheses in parallel. Here, we review recent advances in using CRISPR screening to interrogate the function of the non-coding genome, with particular emphasis placed on epigenetic editing approaches.

CRISPR for genomic and epigenomic editing and application to pooled screening

CRISPR evolved as an adaptive immune system in prokaryotes, and has recently been repurposed for gene editing in mammalian cells [8,9]. In type II CRISPR systems, a CRISPR-associated (Cas) endonuclease is guided to a DNA target via a short guide RNA (gRNA) where it catalyzes a double-strand break [10]. Endogenous cellular pathways can repair the break by either non-homologous end joining (NHEJ) or homology-directed repair (HDR). NHEJ in particular is useful for screening as it is an error-prone process resulting in small insertions or deletions (indels) that can knockout gene function by disrupting the reading frame or alter gene regulation by mutating transcription factor binding sites. CRISPR systems have also been repurposed for programmable gene regulation by deactivating the catalytic activity of the Cas endonuclease [11]. The nuclease-deactivated version retains its ability for DNA targeting via the gRNA, and thus can serve as a scaffold for recruiting transcriptional machinery or epigenetic modifiers. The nuclease-deactivated versions have been engineered as repressors and activators by fusing transcriptional regulatory domains to the termini of the Cas protein [12].

The Cas9 endonuclease from the type II CRISPR system of Streptococcus pyogenes has been the most widely used CRISPR system for implementing high-throughput gRNA screens. The first screens with nuclease-active Cas9 designed pools of gRNAs targeting early constitutive exons of over 18,000 protein coding genes [13,14]. The indels formed in the early exons frequently result in out-of-frame mutations, leading to deactivation of the target gene. Similar screens were implemented with deactivated Cas9 (dCas9) fusion proteins to knockdown or activate genes via targeting with dCas9-based effectors [15,16]. Gene repression is typically achieved by fusing dCas9 to the Krüppel-associated box (KRAB) domain that recruits epigenetic modifiers that catalyze the formation of heterochromatin [17]. Gene activation can be achieved with any one of several distinct dCas9-based effectors, all of which rely on the recruitment of transcriptional machinery via multiple scaffold domains, such as VP64, p65 or HSF1 [15–19]. Contrary to gene knockout with nuclease-active Cas9, dCas9-based transcriptional regulation is achieved by targeting gene promoters and is highly specific in manipulating gene expression, epigenetic marks, and chromatin structure [20,21].

A high-throughput screen with either Cas9 or dCas9-based effectors applies pools of gRNAs to test thousands of hypotheses in parallel. gRNAs are typically synthesized as pools of oligonucleotides and assembled into a lentiviral vector library. The lentiviral library of gRNAs is then transduced into a cell population at a low multiplicity of infection, such that a single gRNA expression cassette integrates into each genome and serves as a genetic barcode for each cell. The cell line can either stably express the Cas9 effector, or the lentiviral library can contain both Cas9 and the gRNA. The cell population is then selected for cells that received a gRNA and cultured for a period of time to ensure adequate expression of the CRISPR components and permit the phenotypic effect resulting from the corresponding genomic perturbation.

In order to screen for genes or genomic regions corresponding to a particular phenotype, there must be selectable criteria by which cells in which that phenotype is perturbed can be identified. Common selection methods include resistance or sensitivity to drug treatment, changes to cellular proliferation rates or viability, or changes in gene expression assessed by fluorescence-activated cell sorting (FACS) or single cell RNA-seq. Following selection, the gRNA abundance in the selected population is quantified via next-generation sequencing to identify enriched or depleted gRNAs relative to the initial gRNA pool. This general approach to CRISPR-based screening was first implemented to identify essential genes in cancer, uncover mechanisms of drug resistance, and dissect cellular signaling networks [13,14,22]. In addition, data collected from such screens have helped to develop algorithms to predict optimal gRNAs for next-generation libraries [23,24].

Although the majority of CRISPR-based screens to date have targeted coding genes, several recent studies have applied a similar screening methodology to probe the function of non-coding regions distal to gene bodies. Non-coding screens have been performed with both Cas9 and dCas9-based effectors [25], each of which harbors its own advantages and disadvantages in elucidating the role of the non-coding genome in coordinating gene expression. With nuclease-active Cas9, saturation mutagenesis screens that densely tile a region with gRNAs can better resolve minimal sites within a putative regulatory element that influence function, such as transcription factor binding sites [26]. In addition, the indels produced via NHEJ can define sequences that impart differing influence on activity of a particular element. Complementing this approach, CRISPR-based screens with dCas9 fusion proteins can aid in elucidating the role of particular genomic regions and epigenetic marks in regulating gene expression. Fusions of epigenetic modifying domains to dCas9 have been shown to deposit chemical modifications to local histones and DNA and influence gene expression [12]. These tools can be applied to high-throughput screens to identify the sites that harbor regulatory function, and also to specify the epigenetic marks that guide that function. Furthermore, in some cases, a single gRNA can modulate the chromatin state of an entire enhancer [20,27], enabling the coverage of a larger region of genomic space with a fixed gRNA library size.

Nuclease Mutagenesis Screens

The largest collection of work performed thus far in screening non-coding sequence has been with CRISPR nuclease-based mutagenesis, exploiting the indel footprint of Cas9 to identify important regulatory sites (Figure 1A). One of the earliest examples probed the erythroid intronic enhancer of BCL11A which regulates fetal hemoglobin levels and is a therapeutic target for several hemoglobin disorders [26]. A gRNA library was designed consisting of ~500 gRNAs targeting the three Dnase I Hypersensitivity Sites (DHSs) in the composite enhancer, ~50 gRNAs targeting exon 2 of BCL11A as a positive control and 120 non-targeting gRNAs. The screen revealed several transcription factor motifs required for BCL11A expression including the erythroid-specific transcription factor GATA1 as well as STAT1, EHF, and EFL1. Another study used a similar approach to interrogate regulatory elements that control PD-1 expression in mouse T cells during exhaustion induced by chronic infection [28]. A later study looked to increase the number of perturbations per region by targeting putative regulatory elements found in a GWAS locus with two variants of the SpCas9 protein that have different PAM requirements [29].

Figure 1.

An overview of library design approaches to screen the non-coding genome. (A) A saturation mutagenesis approach to dissect known or putative enhancers. Targeted regions of DNA are saturated with gRNAs to determine which sequence disruptions, generated by indels via genome editing, will alter gene expression. This method can be used to determine which transcription factor motifs are required for proper enhancer function. (B) An unbiased tiling approach to scan across a genomic region of interest. This method does not rely on any prior knowledge of transcription factor binding or other epigenetic marks, but is limited in the amount of sequence space that can be interrogated. (C) Targeting specific transcription factor binding sites via ChIP-seq signal or transcription factor motifs. This method relies on prior knowledge of important DNA-binding proteins to target and disrupt their binding sites. The focus on particular binding sites facilitates using a reduced number of gRNAs and enables the possibility of genome-wide screens. (D) A DHS-targeted approach that provides intermediate coverage between strategies in (B) and (C). gRNAs are designed to target DNase I hypersensitive sites (DHSs) in order to reduce library size and target a larger genomic region without requiring previous knowledge of important transcription factors. (B) Overlapping paired gRNAs can be used to create deletions that span across a locus and disrupt regulatory DNA. Pairs of gRNAs are delivered together such that in one cell a specific region of DNA is deleted. Signal from overlapping deletions can narrow down which region is important for gene regulation. Approaches A,B,C, and E have been used with Cas9-based nuclease screens while approaches B-D have been used with dCas9-based approaches.

An alternative approach to saturating defined putative regulatory elements is to design an unbiased library of gRNAs surrounding genes of interest (Figure 1B). One such study performed a saturation mutagenesis screen to identify cis-regulatory elements surrounding four genes specifically expressed in mouse embryonic stem cells (mESCs). Each gene had an in-frame knock-in GFP sequence to quantify gene expression via flow cytometry [30]. Additionally, the authors developed a non-viral approach to deliver the gRNA library to cells. A gRNA expression cassette containing a placeholder gRNA was integrated at the ROSA26 locus in mESCs. Subsequently, a pool of oligos encoding gRNA spacer sequences and containing homology arms was transfected into the cell line along with Cas9 and a gRNA targeting the placeholder gRNA sequence to facilitate homology-directed repair and integration of the gRNA library. Cells were sorted based on GFP expression into different bins and gRNA abundance was quantified within each bin. These screens identified promoters of nearby genes as well as unannotated regions containing no chromatin features of canonical regulatory elements (i.e. DNase signal or ChIP-seq signal for known transcription factors) as having an impact on gene expression.

Downstream validation of gRNAs enriched in the initial pooled screens is essential to sufficiently characterize the regulatory function of those defined genomic regions. For example, a recent study used chromosome conformation capture (3C) to validate putative enhancers identified with a gRNA library surrounding three genes found to be involved in vemurafenib resistance in melanoma cells [31]. The greatest enrichment of gRNAs was found in regions surrounding the CUL3 gene. In addition to 3C analysis, the authors found several correlating genomic and epigenomic features at the gRNA-enriched regions, such as increased chromatin accessibility in melanoma cells and increased DNA sequence conservation within primates, indicating evolutionarily important regulatory activity. These types of characteristics could help inform future algorithmic identification of active regulatory elements.

High density saturation mutagenesis surrounding genes of interest is effective at resolving functional regulatory elements proximal to target genes. However, this approach is limited in genomic coverage with a library of constrained size, which is typically restricted by cell number or throughput of the selection method, such as FACS. Rather than saturating regions with all possible gRNAs, a more targeted approach can be used to only focus on previously identified chromatin annotations that attempt to define putative regulatory elements, such as DNase-seq and ChIP-seq (Figure 1C,D). For instance, a recent study looked for regulators of the embryonic stem cell gene POU5F1 (OCT4) by targeting putative enhancers that were previously identified by epigenetic marks [32]. A gRNA library was designed to target 174 DHSs with ~11 gRNAs per DHS resulting in a library of ~1900 gRNAs (Figure 1D). This library was delivered with Cas9 to a POU5F1-GFP reporter stem cell line, and low reporter expressing cells were isolated with FACS. The authors identified known enhancers of POU5F1 as well as previously unknown enhancers that when perturbed, resulted in transient knockdown of reporter expression. Another study designed a gRNA library to target specific transcription factor binding motifs identified by ChIP-seq (Figure 1C). This targeted approach allowed for a reduced library size and increased number of regions targeted. The authors performed screens targeting p53 motifs with a library of 1,116 gRNAs to map enhancers involved in oncogene-induced senescence and ERα motifs with a library of 97 gRNAs to identify sites required for proliferation in breast cancer cell types [33]. These studies demonstrate that prior knowledge of transcription factor binding and chromatin accessibility can significantly reduce the size of the library required for identification of important non-coding regions.

Larger genomic regions can be analyzed with deletion screens, where a pair of gRNAs is co-expressed and mediates a targeted genomic deletion (Figure 1E). This approach was applied in two recent studies. The first delivered a library of paired gRNAs for programmed deletions around the OCT4 locus [34]. Overlap in deletions allowed for a cumulative signal to be recovered after sequencing gRNAs in cells with low GFP expression. The second study used programmed genetic deletions to identify regulatory elements of HPRT1 in HAP1 cells after selection with a chemotherapeutic drug [35]. The average deletion size in this library was ~1–2 kb in length across a 206 kb region of the HPRT1 locus. This study found a lack of distal enhancers essential for HPRT1 expression with the strongest effects of deletion occurring in the exons of the gene.

CRISPRi Screens

Rather than using Cas9 to perturb regulatory elements through DNA sequence mutation, dCas9-based CRISPR interference (CRISPRi) can be used to identify regulatory regions in similar pooled formats. One study used dCas9-KRAB to tile a total of ~1.3 megabases surrounding GATA1 and MYC in K562 cells [36]. These genes are required for proliferation in this cell line, so depletion of gRNAs from the library was measured after 14 population doublings. Two key features of the enriched regions around these genes were DNase I hypersensitivity and H3K27ac signal, supporting the use of these features for designing targeted gRNA libraries. Together with Hi-C enhancer-promoter contact frequencies, these features were used to construct a predictive model of enhancer activity and correctly ranked six of the seven identified regulatory elements as well as other known enhancers.

Another study used dCas9-KRAB to perturb DHSs in two loci, the human β-globin locus and the HER2 locus [37]. DNase I hypersensitivity was used to identify putative regulatory elements and gRNAs were designed to tile each DHS. For the β-globin locus, transcriptional activity was measured via an endogenous HBE1-mCherry reporter while HER2 levels were measured using immunofluorescence. After applying the library of gRNAs and sorting off high and low expressing cells, known enhancers in the β-globin locus control region were enriched as well as promoters of the downstream HBG1/2 genes. For the HER2 screen, a previously characterized HER2 intronic enhancer was identified as well as a downstream promoter of GRB7 and a novel intronic enhancer also located in GRB7.

CRISPRa Screens

Far fewer studies have utilized CRISPR activation (CRISPRa) to probe for enhancer activity, but next-generation dCas9-based activators have made this approach feasible [27]. An early study to use CRISPRa gain-of-function screens for regulatory element discovery utilized the dCas9-p300 fusion for activation [37]. The p300 core domain deposits acetylation marks to histones, including histone H3 lysine 27, which is found at active promoters and enhancers and is thought to open chromatin by reducing electrostatic interactions between DNA and lysine residues. This study used a library of gRNAs targeting DHSs surrounding HER2 in the A431 and HEK293T cell lines. High and low HER2 expressing cells, as measured using cell surface immunostaining, were isolated using FACS. Enriched DHSs were found in both cell lines, some of which were non-overlapping between cell types, possibly attributable to cell type-specific enhancer activity. This demonstrates the importance of choosing appropriate cell lines or systems that best model the biology being studied in order to find the most relevant hits. This study also directly compared CRISPRi and CRISPRa screens at the same locus and found only partially overlapping results, demonstrating the importance of both gain- and loss-of-function screens to map the regulatory landscape of a locus.

In another study, dCas9-VP64 was used to screen for enhancers of CD69 and IL2RA in Jurkat T cells [38]. One of the identified enhancers of IL2RA contained an autoimmunity disease variant identified through GWAS. A mouse model was generated harboring the disease variant and activated T cells showed a delayed expression of Il2ra rather than complete loss of expression. When the same enhancer was deleted, mice produced more pro-inflammatory T helper cells and decreased numbers of regulatory T cells in response to stimulus. These results demonstrate that certain regulatory elements are specific to cell type and stimulus, further supporting the importance of choosing a relevant cell type and screening approach for identifying significant hits.

Single-cell methods

CRISPR-based screens in a pooled format enable testing of thousands of hypotheses in parallel. Such a systematic and scalable approach has revolutionized the annotation of genome function. However, in the case of most CRISPR screens performed to date, the phenotypic readouts are complex cellular functions, such as cell growth and drug resistance. These phenotypes can be influenced by myriad cellular processes, yet a pooled CRISPR screen cannot inform as to the molecular mechanisms deriving from a single or multiplexed genomic or epigenomic perturbation. Such analyses are commonly performed after the initial screen during which individual gRNAs are validated in low throughput.

To address some of the limitations of traditional CRISPR screening approaches, several groups have developed technologies to combine massively parallel single-cell RNA sequencing methods [39–41] with pooled CRISPR genetic and epigenetic perturbations [42–45]. In this way, complete transcriptomic effects can be comprehensively determined for single or combinatorial perturbations. Since this approach is limited in the number of perturbations that can be tested in parallel because of constraints in sequencing depth or cell numbers, the bulk screening approaches described above can aid in narrowing down targets to choose from. For example, Adamson et al. applied their Perturb-seq technology to dissect the regulatory mechanisms of the unfolded protein response (UPR) in mammalian cells. The authors first performed a genome-wide CRISPR knockout screen to identify a set of candidate genes involved in UPR regulation. Subsequently, they implemented Perturb-seq on a subset of 82 genes and assessed the transcriptional effects resulting from the knockdown of these genes. Interestingly, they identified gene sets that modulated the UPR response through distinct signaling branches, and uncovered cellular subpopulations that respond differently to a given perturbation [42].

The initial studies combining single-cell RNA sequencing with pooled CRISPR screens have used gRNA libraries targeting protein-coding genes. However, this approach holds unique value in uncovering functions of non-coding regulatory elements. For example, the unbiased nature of profiling whole transcriptomic effects of CRISPR-based perturbations enables the identification of genes and gene networks downstream of particular regulatory elements. This is of particular importance when the target genes of candidate enhancers are not known a priori, which is often the case in the identification of disease-associated genetic variation from GWAS. In addition, the single-cell approaches facilitate combinatorial analysis of concurrent perturbations, which can help to delineate the contribution of multiple regulatory elements.

Some of these advantages were demonstrated in a recent study which targeted 71 constituent enhancers from 15 super enhancers with dCas9-KRAB and measured the transcriptional profile of 12,444 cells [46]. Cells were transduced such that each cell received an average of ~3.2 gRNAs, allowing for multiple perturbations to be studied in one cell and reducing the overall number of cells needed for the screen. An advantage to this approach is the ability to determine the cell-to-cell variability in the usage of each enhancer in a given region. This was measured by determining the number of cells that had changes to gene expression as well as the level of gene expression change in those cells. Some enhancers were found to be active in a majority of cells and also led to significant knockdown of gene expression. Other enhancers were found to be active only in a subset of cells but also led to significant knockdown of gene expression. Finally, a third class of enhancers were found to be active in the majority of cells but had only mild effects on gene expression. In addition to classification of enhancer usage, combinatorial perturbations can be measured, as was seen with cells harboring gRNAs targeting multiple enhancers within the same super enhancer. Cells containing only one gRNA targeting an enhancer near PIM1 resulted in no detectable knockdown. However, when a cell contained two gRNAs targeting enhancers nearby PIM1, significant knockdown was detected. These types of pooled screen analyses are greatly facilitated through single-cell readouts.

Perspective

High-throughput CRISPR screens have greatly improved our ability to dissect the function of the non-coding genome (Figure 2). The complement of Cas9 and dCas9-based approaches enables the identification of regulatory motifs within putative enhancers and the epigenetic signatures associated with their function. Though most screens thus far have been performed with Cas9 from Streptococcus pyogenes that is limited to target sites with an NGG PAM, recent success implementing Streptococcus aureus Cas9 into pooled screening formats will expand the sequence space for higher resolution perturbations and enable combinatorial and orthogonal screens revealing novel genetic and epigenetic interactions [47].

Figure 2.

Validation and characterization of hits from noncoding CRISPR screens. Pooled screens with dCas9-effectors can potentially scan a larger region of the genome since enhancer activity can be increased or decreased with a single gRNA. The results from these screens can inform target regions for more focused Cas9-based mutagenesis screens. Following a pooled screen, individual gRNAs are assayed independently to validate their function in gene regulation. Genome-wide genomic datasets (RNA-seq, ChIP-seq, ATAC-seq, Hi-C, etc…) can be used to determine the target genes of putative enhancers and the epigenetic marks and genomic topology associated with their function. With Cas9 mutagenesis, secondary screens with individual gRNAs or smaller sub-libraries can be performed to define functional genotypes resulting from Cas9-induced indels. These sequence features can then be compared to datasets of genetic variation associated with human disease, which can direct the generation of in vivo disease models or drug screening platforms. The data acquired from CRISPR-based screens of the noncoding genome can be useful in the development of algorithms to predict enhancer function or define features of highly active gRNAs.

Most CRISPR screens of the non-coding genome to date have identified regulatory elements that harbor epigenetic signatures predictive of enhancers (i.e. DNase I hypersensitivity). This finding reinforces the dogma that epigenetic marks can denote regulatory function and reduces the sequence space needed for future screens. However, one study did identify functional elements lacking any canonical epigenetic signatures [30], although it is unclear how ubiquitous this phenomenon is across the genome. It will be important to perform comparative screens using different Cas9 and dCas9-based effectors to elucidate how the deposition or removal of specific epigenetic marks influences the regulatory function of a genomic site. Data accumulated from such screens could inform computational approaches to better predict enhancer function in homeostasis or in response to perturbation. Lastly, the knowledge gained from CRISPR screens could inform the rational design of de novo enhancers, perhaps with aide from recent technologies for artificially modulating chromosomal topology [48].

Table 1.

A survey of studies using CRISPR-based screens to annotate function of the non-coding genome.

Technology Used	Screen Phenotype Measured	Library Design	Area of Genomic Region Screened	Number of Sites Screened	Number of gRNAs in library
SpCas9	Antibody Staining (FACS)	Saturation Mutagenesis	~4.2 Kb	3 DHSs, 1 Exon	702
SpCas9	Antibody Staining (FACS)	DHS Targeting	~60 Kb	9 DHSs	2,071
SpCas9	Reporter Expression (FACS)	Unbiased Tiling/Predicted Enhancer	40 Kb – 150 Kb	4 genes	3,908
SpCas9	Reporter Expression (FACS)	DHS Targeting	~1 Mb	174 DHSs	1,964
SpCas9	Reporter Expression (FACS)	Paired gRNA Deletions	2 Mb	1 gene	11,570
SpCas9	Proliferation	TF Motif/ChIP-seq Peak Targeting	Genome wide	685 or 73 Regions	1,116 or 97
SpCas9	Resistance to Drug Treatment (Proliferation)	Unbiased Tiling	200 Kb	3 genes	6,682, 6,934, or 4,699
SpCas9	Resistance to Drug Treatment (Proliferation)	Paired gRNA Deletions	206 Kb	1 gene	8,684
SpCas9, SpCas9-VQR	Antibody Staining (FACS)	Saturation Mutagenesis	~340 Kb	98 DHSs	5,542
dCas9-KRAB	Proliferation	Unbiased Tiling	1.29 Mb	2 genes	98,000
dCas9-KRAB	Single cell RNA-seq	DHS Targeting	~2 Mb across 7 TADs	71 constituent enhancers in 15 super enhancers	243
dCas9-KRAB, dCas9-p300	Reporter Expression or Antibody Staining (FACS)	DHS Targeting	~4.5 or ~4 Mb	281 or 433 DHSs	12,472
dCas9-VP64	Antibody Staining (FACS)	Unbiased Tiling	~135 Kb or ~178 Kb	2 genes	10,780 or 20,412

Highlights.

• Non-coding regulatory DNA determines gene expression and complex phenotypes

• Technologies have not been available to annotate function of the non-coding genome

• High-throughput synthesis of gRNAs facilitates CRISPR-based genomic screens

• Screens include Cas9-based mutagenesis and dCas9-based gain- and loss-of-function

• Diverse screening strategies can be employed with varying degrees of genomic coverage