A CRISPR Competition Assay to Identify Cancer Genetic Dependencies

[Abstract] The CRISPR/Cas9 system is a powerful tool for genome editing, wherein the RNA-guided nuclease Cas9 can be directed to introduce double-stranded breaks (DSBs) at a targeted locus. In mammalian cells, these DSBs are typically repaired through error-prone processes, resulting in insertions or deletions (indels) at the targeted locus. Researchers can use these Cas9-mediated lesions to probe the consequences of loss-of-function perturbations in genes of interest. Here, we describe an optimized protocol to identify specific genes required for cancer cell fitness through a CRISPR-mediated cellular competition assay. Identifying these genetic dependencies is of utmost importance, as they provide potential targets for anti-cancer drug development. This protocol provides researchers with a robust and scalable approach to investigate gene dependencies in a variety of cell lines and cancer types and to validate the results of high-throughput or whole-genome screens.

2017; Giuliano et al., 2018;Thomenius et al., 2018). In contrast, head-to-head comparisons have demonstrated that CRISPR/Cas9 constructs exhibit significantly fewer off-target effects compared to RNAi-based approaches, underscoring the significant potential that this technology harbors to identify true cancer dependencies (Morgens et al., 2016;Smith et al., 2017).
As with RNAi, the highly programable nature of CRISPR facilitates targeted screening approaches to uncover genetic addictions in cancer cell lines (Shalem et al., 2014;Wang et al., 2014). Through rational guide design, researchers can disrupt protein function and interrogate potential drug targets. In this protocol, we describe an optimized screening methodology first described by Shi et al. (2015) in which cells that constitutively express the Cas9 nuclease are transduced with a vector that co-expressed a sgRNA and GFP. The resulting population is comprised of both transduced (GFP+) and untransduced cells (GFP-), and the relative abundance of each population is tracked via flow cytometry over the course of several passages ( Figure 1). Guides that target the functional protein domains of genes essential to cellular fitness are consistently outcompeted by untransduced cancer cells, which is easily detectable as a decreasing abundance of GFP+ cells over time. Here we describe a straightforward protocol to determine whether a gene of interest is required for the viability or fitness of a particular cancer cell line. This protocol can be easily scaled-up and offers robust results in as little as three weeks. We have extensively validated this approach through the use of control gRNAs targeting known essential genes and non-essential loci. Researchers can investigate potential dependencies through this GFP competition assay by utilizing guides targeting their gene of 3 www.bio-protocol.org/e3682 interest in parallel with appropriate positive and negative controls. We have found that the results of these GFP competition assays are consistent with other in vitro assays measuring cellular fitness, including 2-D proliferation assays and soft-agar assays to measure anchorage-independent growth (Lin et al., 2017 and. This protocol also represents a straightforward approach to validate individual hits recovered in high-throughput or whole-genome CRISPR screens.  14. gRNA plasmid backbone (see Table 1 for suggested plasmids and their respective Addgene numbers)  i. To optimize the concentration of puromycin for selection, we suggest plating wild-type cells at approximately 50% confluence on 6 or 12-well plates.

Materials and Reagents
ii. Replace media on the cells 24 h later with media containing puromycin at a range of 1-4 µg/ml. We generally test in increments of 0.5 μg/ml, with each well containing a different concentration of puromycin. 6 www.bio-protocol.org/e3682 c. Maintain selection until Cas9 is stably expressed. We find that for most cell lines, only one round of transduction, followed by 3 to 5 days of puromycin selection, is sufficient for generating stable Cas9 expression. However, some cell lines may require more than one round of transduction and/or a longer selection period. iii. Transduction with guides targeting essential genes. While this may not be suitable for the first GFP competition assay you conduct, our lab routinely performs control competition assays with several positive and negative control guides (as described later in this protocol) to verify Cas9 functionality in a cell line of interest.
b. It is of note to mention that qPCR and western blotting for Cas9 can confirm Cas9 expression but not functionality. Transduction with guides with known biological activity can serve to confirm Cas9 functionality. For these reasons, we recommend using a combination of these approaches to verify Cas9 expression and functionality in any Cas9-expressing cell lines you generate.
Note: In this section, we will design guides targeting the gene of interest, and clone these guides, alongside appropriate positive and negative control guides, into guide plasmids for sgRNA and fluorescent protein expression. We discuss considerations for ideal guide design and provide control guides extensively validated by our lab and our methods for plasmid cloning. 7 www.bio-protocol.org/e3682 Bio-protocol 10 (14) i. For this protocol, we use single guides with 20 base pairs of homology to the target sequence, as this is best-suited for CRISPR-knockout.
ii. The PAM sequence should be NGG on the 3′ side of the guide sequence, as the Cas9 strain we use is SpCas9. The PAM sequence is necessary for Cas9-mediated DNA cleavage and is found 3-4 base pairs downstream of the cut site.
b. Guides targeting functional protein domains produce the most-efficient knockout of the target gene. For this reason, we highly recommend cross-referencing exons with protein domains to identify which guides are best-suited to disrupt protein function, as guides targeting the first exon alone can lead to false negatives ( Figure 2A) (Shi et al., 2015).
c. Additionally, we recommend designing 3 or more guides per gene, ideally targeting different exons.
d. Order oligonucleotides through IDT or your preferred commercial source for DNA synthesis. The fold dropout over the course of five passages is shown for each guide. It can be seen that the negative control guides targeting Rosa26 exhibit minimal depletion and the positive control guides targeting RPA3 display over ten-fold dropout at the final timepoint. Consistent with prior knowledge, these results verify that AURKB is a cancer cell dependency.
2. Identify suitable positive and negative control guides a. Negative control guides should target noncoding loci or genes dispensable for cellular fitness such as Rosa26 or AAVS1. These loci are considered canonical safe-harbor sites, and thus introducing indels at these loci will have minimal effects on cancer cell fitness.
b. Positive control guides should target essential genes, such as PCNA or RPA3. These genes are critical for DNA replication, and thus introducing indels at these loci will be detrimental to cancer cell fitness.
c. We have provided our control guides in Table 2.   To ensure experimental consistency, we suggest saving a program with your laser voltages and analysis templates to use for successive time-points.
Note: In this section, we will continue the competition assay through its completion.
E. Successive timepoints of competition assay 1. Harvest cells 2. Seed cells at appropriate confluency for next passage.
At the fifth timepoint, the assay is complete, and it is not necessary to re-plate the cells for the next passage. 2. Therefore, the fold change at passage 1 should be 1. An example of these calculations can be seen in Figure 2B.

B. Prepare grouped graphs
We suggest using Prism, but any software that can graph numerical data (i.e., Excel) will suffice ( Figure 2C).
C. Set threshold for dropout 1. The threshold should be greater than the fold change of the negative controls, as these guides target loci dispensable for cancer cell fitness.
2. Our lab uses a threshold of 2.5-fold change for determining dependencies, as we have found that this threshold is always above the level of depletion observed with negative control guides.
3. This can be adjusted based on the cell line as needed, as the dynamic range for fold change is dependent on a variety of factors including Cas9 expression, cell type differences, and GFP % at passage 1.
4. If multiple guides targeting a gene exhibit minimal dropout above background levels, this gene is unlikely to be a genetic dependency in the cell line tested.
5. If multiple guides targeting a gene exhibit consistent dropout above background levels, this is evidence for the gene being a dependency in that cell line. to probe for genetic dependencies in other contexts. For instance, this assay can be modified to probe for synthetically lethal relationships between two genes (i.e., gene A and gene B). If guides targeting each gene individually exhibit minimal dropout, but the researcher has evidence to suggest these genes have some degree of functional redundancy, then a dualcompetition assay can be conducted instead. In this experiment, researchers can co-transduce target cells with two guides that express different fluorescent protein markers (Table 1) double-positive population exhibits consistent depletion, but the single-positive populations do not, this is evidence for a potential synthetically lethal relationship between genes A and B.
2. The development of additional CRISPR tools for transcriptional modulation provides researchers with other possible approaches for probing genetic dependencies. The CRISPRinterference (CRISPRi) system, wherein catalytically-inactive Cas9 (dCas9) is fused to a Krüppel-associated box (KRAB) domain, provides researchers the ability to suppress target gene expression without double-strand break formation (Gilbert et al., 2013). The resulting partial loss-of-function phenotype provides yet another context for researchers to investigate potential drug targets. The use of such an orthogonal approach complements the total loss-offunction phenotype generated by standard CRISPR/Cas9. Partial loss-of-function approaches, such as CRISPRi provide models of genetic perturbation more closely resembling the effects of targeted therapies, as pharmacological inhibition of any potential genetic dependency is unlikely to be absolute. This competition assay can be readily tailored to a CRISPRi approach through the use of a dCas9-KRAB expression vector in place of a Cas9 expression vector (Addgene #85969), and through the use of guides targeting promoters rather than functional exons. Our lab regularly conducts such assays, and we have found our results from this modified approach to be largely consistent with traditional Cas9 screening (Lin et al., 2019).
3. While the CRISPR/Cas9 system offers numerous advantages over RNAi for investigating genetic dependencies, there are some limitations to consider. Generally, the Cas9 nuclease is utilized to produce an assortment of indels at the target site, and it is thought that the resulting premature stop codon leads to nonsense-mediated decay (NMD) of the mutant mRNA. However, the production of indels does not necessarily result in the ablation of the target gene product.  follows that these limitations could potentially be minimized through use of alternative systems for perturbation that do not cleave DNA, such as CRISPR-interference. 5. Beyond the limitations of the CRISPR system itself, it is important to note that the results of this GFP competition assay are by no means dispositive of a gene's status as being necessary or dispensable for cellular fitness in other contexts. While the consistent lack of dropout for guides targeting a gene across multiple trials and cell lines indicates that a targeted gene is dispensable for cell-autonomous growth in vitro, it remains possible that the gene is a dependency in other cellular contexts, like anchorage-independent growth, low oxygen conditions, in the presence of chemotherapeutics, in vivo, etc. Thus, a gene that does not score as a genetic dependency through this assay can still provide a valuable drug target for anticancer therapies. On the other hand, if researchers do see consistent dropout for guides targeting a gene, this is strong evidence for the gene being a genetic dependency. Such results should be followed up with further investigation through orthogonal approaches. 6. We believe this assay will provide researchers with a valuable approach to investigate putative cancer dependencies. Results from our lab show this assay to be robust in its findings, and its relative simplicity allows researchers to simultaneously conduct this assay in many cell types.
This assay is best utilized as a first-pass approach, and complemented with thorough, in-depth investigation of any putative cancer dependencies.