Genetic perturbation provides biological insight. The development of genetic tools such as RNA interference (RNAi) have opened the door for genome-wide targeted perturbation enabling understanding of cancer cell and immune cell dependencies (1, 2). However, RNAi screening can be limited by incomplete protein knockdown and off-target effects (3–6). Discovery of CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats)-associated Cas9 nuclease has revolutionized perturbation approaches by enabling knock out at the DNA level (7–10). Furthermore, the CRISPR-Cas9 system has enabled pooled genetic screening to be performed in an efficient manner compared to arrayed screens (11, 12). While arrayed screens use multiwell plates to separately target genes (13), pooled screens perturb a pool of genes in a single population of cells for increased throughput. Critical modules in performing a CRISPR screen include selection of an endonuclease system and guide library, screen phenotype selection, and hit evaluation. CRISPR-Cas9 screens have substantially advanced our understanding of innate and adaptive immune responses in the setting of infection and cancer (14–18). Furthermore, significant advances are being made in high-content phenotypic readouts at the transcriptomic, epigenetic, proteomic, and optical levels to increase resolution and overcome screening bottlenecks (19–22). This review highlights the critical modules necessary to perform a pooled CRISPR-Cas9 screen as well as discusses the recent development of high-content CRISPR screening systems. Furthermore, we detail how use of this technology is enabling discovery of novel therapeutic targets in the field of tumor immunology.

The Cas9 protein, derived from Streptococcus pyogenes, is the most commonly used Cas protein for CRISPR genome editing (23) due to its ability to introduce blunt double-stranded breaks (DSBs) in a cell’s genome at designated target sites. Error-prone non-homologous end joining (NHEJ) repair of these DSBs introduces a coding region frameshift through nucleotide insertion or deletion (indel) mutations which result in permanent gene silencing (7, 8). Cas9 is RNA-guided and can be programmed by guide RNAs (gRNA) to target specific genomic regions of interest, thus permitting robust loss-of-function studies (11–13). However, a limitation to the use of Cas proteins includes the inherent immunogenicity of this bacterial protein, which may reduce its efficacy when introduced in vivo (24).

More recently, the CRISPR system has been adapted for CRISPR interference (CRISPRi) and CRISPR activation (CRISPRa) studies (25–30). These methods utilize a catalytically inactivated Cas9 protein (dCas9) (7) that binds to genomic regions but is unable to cleave its target sequence for generation of DSBs. Fusion of dCas9 to activator or repressor domains results in targeted transcriptional activation or suppression at endogenous promotors, respectively. VP64 is a common transcriptional activator utilized for this system. Meanwhile, the Kruppel-associated box (KRAB) repressor domain can be coupled to Cas9 for transcriptional suppression (31). Thus, multiple CRISPR systems exist for genetic knockout, activation, or suppression studies and have become an appreciable tool for biological discovery.

Conventional pooled CRISPR screens in numerous studies have uncovered genes that regulate the cancer immunity cycle (82) and may improve the antitumor response (83) ( Figure 1 ). Readouts of these screens include cell survival, cell expansion, and phenotypic selection with fluorescence-activated cell sorting (FACS) based on fluorescent reporter or marker expression (84–86), followed by sequencing to identify enriched and depleted gRNAs. CRISPR screening of antigen-presenting cells (APCs) may detect genes regulating APC stimulation and antigen presentation efficiency to T cells ( Figure 1A ). For instance, a pooled screen in mouse primary bone marrow-derived dendritic cells (DCs) has identified factors controlling Tumor Necrosis Factor (TNF)α production in response to lipopolysaccharide (LPS) stimulation (86). These factors include the previously uncharacterized oligosaccharyltransferase (OST) complex and PAF transcription complex (86). Likewise, screens in immortalized and primary murine macrophages have similarly uncovered regulators of LPS-induced pathways (52, 87). For instance, Mettl3-mediated m⁶A RNA modification has been found to positively regulate LPS-induced macrophage activation to control tumor growth (87). Furthermore, autocrine TNF signaling has been surprisingly identified as a negative regulator of NF-κB (52). Overall, these screens used readouts of TNFα and NF-κB reporter expression (52, 86, 87). This screening approach may be applied in the future to investigate DC stimulation in response to tumor antigens. Finally, a pooled CRISPRi screen in the THP-1 human monocyte cell line using a cyclic dinucleotide (CDN)-inducible reporter system has revealed a CDN transporter, SLC19A1, as required for activation of the stimulator of interferon genes (STING) pathway (88). Importantly, induction of the STING pathway in antigen-presenting cells enhances antitumor responses (89).

Regulators of T cell activation have also been discovered by pooled CRISPR screens (16, 90, 91) ( Figure 1B ). Formation of the immunological synapse between T cells and APCs requires cytoskeletal remodeling to provide a proper signaling scaffold for T cell activation (92). A CRISPR screen in the immortalized Jurkat T cell line has uncovered FAM49B as a previously uncharacterized negative regulator of this cytoskeleton reorganization in response to TCR stimulation (90). Furthermore, known regulators of TCR signaling and proliferation have been confirmed in primary human CD8⁺ T cells (16), including the negative regulators SOCS1, CBLB, CD5, and RASA2, and positive regulator LCP2. CRISPRa screening has additionally been used to identify a novel stimulation-dependent IL-2Rα enhancer, Cns2, which influences T cell polarization. Deletion of this enhancer impaired regulatory T (Treg) development and favored differentiation into a pro-inflammatory T helper (Th) 17 state (91). Expression of common activation and proliferation features such as CD69, IL2Rα, and CSFE (16, 90, 91), have been used as a screen readout by these studies, followed by sorting and gRNA sequencing.

CD4⁺ T cells differentiate into distinct phenotypes in response to extracellular signals such as cytokines. Th1 cells are pro-inflammatory and enhance antitumor immunity. On the other hand, anti-inflammatory Th2 cells constrain Th1 response and Treg cells suppress immune functions to promote cancer progression (93). Pooled CRISPR screens in these T cell populations have revealed factors controlling their differentiation (94–98). An in vivo proliferation screen has identified Socs1 as a negative regulator of CD4⁺ T cell proliferation, survival, and effector function indicative of a functional Th1 response. Notably, depletion of Socs1 in CAR T cells enhanced antitumor immunity (94). Meanwhile, transcription factors Pparg and Bhlhe40 identified by an in vitro screen have been revealed to mediate Th2 differentiation (95). Regulators of Foxp3 expression in Treg cells have also been pinpointed by in vitro screens. Positive regulators of Foxp3 include ubiquitin-specific peptidase 22 (Usp22) (a member of the SAGA complex) and the Brd9-containing ncBAF complex, while Rnf20, an E3 ubiquitin ligase, has been identified as a negative regulator (96, 97). Finally, mTORC1 signaling in Treg cells has been found to be promoted by Sec31a and the SWI/SNF complex and inhibited by the Ccdc101-containing SAGA complex (98. Inactivation of Usp22, Brd9, and Ccdc101 enhanced antitumor immunity but also led to spontaneous inflammation in some cases (96–98). Thus, regulators of CD4⁺ T cell differentiation may be targeted to skew the balance towards a beneficial Th1 effector phenotype.

In vivo CRISPR screens have revealed mechanisms controlling tumor infiltration by CD8⁺ T cells in models of triple-negative breast cancer, glioblastoma, and melanoma (85, 99, 100) ( Figure 1C ). To do this, library-transduced CD8⁺ T cells were adoptively transferred into tumor-bearing mice and tumors were then sequenced. Depletion of factors identified by these screens have been shown to control CD8⁺ T cell effector function ( Figure 1D ) including ribonuclease Regnase-1, RNA helicase Dhx37, and ER-associated protein Pdia3 (85, 99, 100), as well as the Fli1 transcription factor identified by an in vivo LCMV Clone 13 infection screen (101). Modulation of effector function includes increased degranulation and tumor cell killing during in vitro co-culture, production of effector cytokines, expression of effector markers, as well as expression of inhibitory receptors (85, 99–101). Furthermore, adoptive transfer of CD8⁺ T cells deficient in these factors has been shown to reduce tumor growth and increase survival in murine tumor models. Notably, one of these studies has engineered regulator-depleted CAR T cells possessing Pdia3 knockout to demonstrate potential clinical application (100).

Many T cell CRISPR screens involve ex vivo stimulation and transduction of isolated T cells; however, this process may disturb their homeostasis and differentiation. To address this issue, Sharpe et al. has developed an in vivo CRISPR-Cas9 delivery system, termed CHIME (CHimeric IMmune Editing) (102). CHIME consists of transducing bone marrow stem cells with a gRNA library ex vivo, followed by transfer into irradiated mice. In this manner, both innate and adaptive immune cell lineages can be rapidly screened in vivo without disrupting differentiation. This screening system has been applied in the context of LCMV Clone 13 infection (102) and may be implemented in future immunology screens to identify regulators of antitumor immunity.

Finally, multiple screens have elucidated mechanisms mediating resistance of cancer cells to CD8⁺ T cell killing and immune checkpoint blockade (ICB) (84, 103–106) ( Figure 1E ). To identify genes regulating resistance, cancer cells were first transduced with a gRNA library followed by either co-culture with CD8⁺ T cells in vitro or transplantation in vivo with or without ICB. Surviving cancer cells were then sequenced to identify genes that mediate resistance or susceptibility to the screening conditions. Screens in human melanoma cells have revealed mechanisms mediating susceptibility to in vitro CD8⁺ T cell killing, including an apelin receptor (APLNR) required for interferon gamma (IFNγ) susceptibility and components of the TNF pathway including TRAF2 (84, 105). Alternatively, in vitro and in vivo screens have identified the SWI/SNF complex and a protein tyrosine phosphatase Ptpn2 as mediators of resistance against the IFNγ response pathway, with inactivation of these factors enhancing the efficacy of ICB (103, 106). Moreover, an in vitro screen of CD19⁺ acute lymphoblastic leukemia cells co-cultured with CD19-targeted CAR T cells has found that deficiency in the death receptor apoptotic signaling pathway mediates cancer cell resistance and induces CAR T cell dysfunction (104). On the other hand, factors controlling macrophage antibody-dependent cellular phagocytosis (ADCP) have additionally been discovered by CRISPR-KO and CRISPRa screening of a B lymphocyte cell line (107). These factors include known ADCP regulators CD47 and CD20, as well as a previously undescribed enzyme adipocyte plasma membrane-associated protein (APMAP). In this study, screened immortalized B lymphocytes were co-cultured with stimulated macrophages and anti-CD20/CD47 antibodies to induce ADCP, then surviving cancer cells were sequenced. Loss of APMAP in cancer cells was shown to synergize with monoclonal antibody therapy and sensitize multiple tumor models to macrophage phagocytosis. Furthermore, a complementary genome-wide screen in macrophages revealed intercellular regulators of APMAP-deficient cancer cell uptake by macrophages, specifically the G-protein coupled receptor GPR84 (107).

Conventional CRISPR screens have also been performed in an arrayed format to investigate regulators of the tumor-immunity cycle. For instance, arrayed CRISPR screening has been used to discover factors necessary for cross-presentation of tumor-antigens by conventional DC1s (cDC1s), most notably Wdfy4 (15) ( Figure 1A ). In this study, primary murine CD8⁺ T cells were co-cultured in vitro with screened cDC1s, and T cell proliferation was measured as a readout of cross-presentation efficiency. Importantly, mice lacking Wdfy4 were unable to suppress tumor growth (15). Furthermore, an arrayed screen in human monocyte-derived DCs (moDCs) has identified known regulators or Toll-like receptor (TLR) and MyD88 signaling in response to LPS from the human microbiome (67). Future screens may similarly investigate regulators of moDC signaling in response to tumor antigens. Finally, arrayed screening has revealed p38 kinase as a regulator of CD8⁺ T cell activation ( Figure 1B ). Inhibition of p38 in CD8⁺ T cells resulted in increased cell expansion and expression of the memory marker CD62L, as well as reduced ROS and DNA damage. Furthermore, CAR T cells depleted in p38 kinase using a small molecule p38 inhibitor exhibited enhanced effector function, shown by increased IFNγ production and cytolytic activity ex vivo (64) ( Figure 1D ). Thus, pooled and arrayed CRISPR screening is a tremendous tool used to uncover key regulators of the cancer-immunity cycle that may potentially be targeted in a clinical setting to improve therapeutic response.

Multiple techniques have been developed that couple CRISPR screens to epigenetic analyses ( Table 1 ). Although these systems originally investigated contexts not pertaining to tumor immunology, future studies could use these methods to probe for epigenetic regulators of anti-tumor immunity in primary immune cells. For instance, CRISPRi and CRISPR-KO screens performed in an arrayed format have been linked with epigenetic data using Perturb-ATAC (115). Perturb-ATAC builds upon ATAC-seq (Assay for Transposase-Accessible Chromatin with sequencing), a method that implements the hyperactive Tn5 transposase with DNA sequencing to assess epigenetic profiling (124). To perform Perturb-ATAC, individual cells are captured on an integrated fluidic circuit and ATAC-seq is performed following pooled screening. The addition of pooled primers that target regions flanking the barcode or gRNA allows for simultaneous amplification of both ATAC-seq fragments and the barcode or gRNA linking the epigenetic data with the CRISPR-induced genetic perturbation ( Figure 2C ). Arrayed CRISPRi screening with Perturb-ATAC has been implicated in human immortalized B lymphoblasts to identify factors that induce nucleosome structure alterations, such as the chromatin regulator DNMT3A (115).

However, a limitation to Perturb-ATAC is that it is performed in an arrayed format and requires a fluidics platform which can only capture ninety-six cells. More recently, development of CRISPR-sciATAC, which uses the CROP-seq vector, has eliminated the need for a fluidics platform and instead implements 96 barcoded transposases for combinatorial barcoding of ATAC and gRNA fragments in 96-well plates (116) ( Figures 2A-C ). Pooled CRISPR-sciATAC screening has been used to screen chromatin-related genes in human myelogenous leukemia K562 cells. Linkage of specific gene perturbations to ATAC-seq data reveals changes in chromatin accessibility. For instance, loss of a SWI/SNF complex component, ARID1A, results in reduced accessibility at JUN and FOS transcription factor binding sites (116).

Furthermore, Pierce et al. has also increased throughput capacity by developing Spear-ATAC (Single-cell perturbations with an accessibility read-out using scATAC-seq) (20), derived from scATAC-seq (125). Since scATAC-seq is droplet-based, it allows for higher throughput and compatibility with pooled screens. In Spear-ATAC-seq, to detect each perturbation, adapters used in ATAC-seq are also inserted around the gRNA sequence ( Figure 2A ). Therefore, both ATAC-seq fragments and the gRNA are directly amplified from the genomic DNA, as compared to typical gRNA detection from RNA transcripts. Consequently, each gRNA perturbation can be easily paired to epigenetic data without the use of a barcoded vector system. Spear-ATAC coupled with pooled CRISPRi screening has been used to phenotype human K562 leukemia and MCF7 breast cancer cell lines, as well as GM12878 B lymphoblasts (20). By perturbing key transcription factors, Spear-ATAC has identified epigenetic alterations to transcription factor binding sites in K562 cells with increased throughput and reduced cost as compared to Perturb-ATAC (115). For instance, GATA1 deletion resulted in increased accessibility of hematopoietic transcription factor binging regions such as RUNX, SPI1, and IRF1 (20). Hence, these systems may be applied in future high-content pooled CRISPR screens to reveal potential epigenetic regulation relevant to antitumor immunity in various immune cell types.

Proteomic data can be coupled to pooled CRISPR screens ( Table 1 ), and this technology has exhibited promise in the field of tumor immunology. ECCITE-seq (Expanded CRISPR-compatible cellular indexing of transcriptomes and epitopes by sequencing) has been adapted from CITE-seq for the purpose of pooled CRISPR screening (117). CITE-seq functions to pair single-cell transcriptome data with protein epitope and TCR clonotype data using barcoded oligonucleotide tags conjugated to protein-specific antibodies that are simultaneously detected during scRNA-seq (126). ECCITE-seq builds off this concept by introducing an additional gRNA scaffold-specific primer to the scRNA-seq reverse transcription reaction ( Figure 2C ). Therefore, the gRNA sequence is directly captured and allows for linkage of transcriptomic, proteomic, and TCR clonotype data to gRNA-specific perturbations introduced during pooled CRISPR screens (117). ECCITE-seq was initially shown to be compatible with small-scale CRISPR screening in K562 cells using a small sub-pool library targeting known cell surface markers and intracellular signaling molecules (117). Recently, ECCITE-seq has identified both transcriptional and post-transcriptional regulators of PD-L1 expression in human monocyte leukemia THP-1 cells. Results from this study illustrate that PD-L1 upregulation following interferon IFNγ signaling is mediated by NRF2 and KEAP1 (118). Additionally, ECCITE-seq has been implemented in human cutaneous T-cell lymphoma samples to reveal transcriptional profiles indicative of malignancy in different T cell clonotypes, as well as transcriptional and proteomic differences in the skin microenvironment of CTCL patients compared to the blood (119).

More recently, Frangieh et al. has developed Perturb-CITE-seq for large-scale pooled screens by combining a modified CROP-seq vector with CITE-seq, as well as a computational framework for analysis of the perturbation data (21) ( Figure 2A ). Notably, Perturb-CITE-seq has been used to screen patient-derived melanoma cells with a library consisting of genes previously associated with ICI resistance. Melanoma cells are cultured with autologous TILs to identify known mechanisms of ICI resistance ( Figure 1E ), such as impaired IFNγ-JAK/STAT and antigen presentation pathways, as well as novel mechanisms such as CD58 downregulation. Importantly, simultaneous scRNA-seq and CITE-seq phenotyping employed by Perturb-CITE-seq has revealed perturbation-induced effects at both the RNA and protein level in this model system.

Protein-level phenotyping of pooled CRISPR screens has also been achieved through high-throughput CyTOF mass cytometry (120). To do this, protein-encoding barcodes (termed Pro-Codes) are generated consisting of triplet combinations of linear epitopes that are detectable by known antibodies. These Pro-Code sequences are then paired with gRNA sequences within the lentiviral vector ( Figure 2A ). Following transduction, metal-conjugated antibodies specific to the linear epitopes detect each Pro-Code reporter by CyTOF mass cytometry to report the corresponding genetic perturbation of each cell. Since CyTOF allows for detection of up to 45 distinct conjugated antibodies, numerous phenotypic targets can be matched to these Pro-Codes. This system has been implemented in breast cancer 4T1 cells to screen for positive and negative regulators of breast cancer immune evasion. This study has revealed that knockout of two interferon-stimulated genes, Psmb8 and Rtp4, contribute to cancer cell resistance to T cell killing and that Socs1 negatively regulates PD-L1 expression (120) ( Figure 1E ). Overall, ECCITE-seq, Perturb-CITE-seq, and CyTOF proteomic screening methods have demonstrated their potential in cancer cell line screens to reveal tumor immune-related mechanisms. However, additional studies are needed to apply these techniques for screening of immune cell populations.

Pooled CRISPR screens can additionally be linked to single-cell optical phenotypes in fixed and live cells ( Table 1 ). Initial work done by Feldman et al. (121, 127) details the use of barcoded single gRNA libraries as well as non-barcoded CROP-seq libraries for this purpose ( Figure 2A ). Following a pooled CRISPR screen, immunostaining and microscopy is employed to delineate single cells in a region of interest prior to in situ amplification and sequencing. This method can even detect multiple perturbations per cell for potential genetic interaction studies. Optical pooled screening has been applied to multiple human cancer cell lines and successfully screened human cervical cancer (HeLa) cells for NF-κB regulators in response to IL-1β or TNFα stimulation. To do this, p65 nuclear localization was used as an optical readout. In this study, MED12 and MED24 were found to be previously unknown negative regulators of p65 translocation and NF-κB activation (121). Therefore, this approach can uncover regulators controlling tumor cell responses to immune stimuli and may be applied to additional cell types in the future.

Pooled optical screens show promise for in vivo settings as well. Induction of CRISPR perturbations in vivo followed by imaging and in situ gRNA sequencing of extracted tissues allows for characterization of native spatial phenotypes, such as the tumor microenvironment (TME). Moreover, additional implementation of spatial transcriptomics can link genetic alterations with both optical and transcriptomic data. This approach has been recently exemplified by the Perturb-map system (22) ( Figure 2A ) that specifically uses Pro-Codes (120) to detect each perturbation and has been used to screen lung adenocarcinoma cells in vivo with a pooled library of known cytokine signaling regulators. Perturb-map has enabled visualization of immune cell infiltration and identification of molecular signatures in extracted lung tumors in response to specific perturbations (22). Notably, loss of TGFβ receptor (TGFβR2) in tumor lesions led to T cell exclusion and enrichment of a TGFβ-activated lung fibroblast signature, indicating stroma remodeling and a more immunosuppressive TME (22). Thus, Perturb-map is an exciting new avenue to uncover spatial regulators of the native TME.

When conducting a conventional pooled CRISPR screen, quality control evaluations, including determination of the gRNA coverage and correlation between biological replicates are essential (128, 129). Inclusion and evaluation of negative non-target control guides can assist in limiting false positive discovery. To determine which genes are most relevant to the screened phenotype, numerous analysis tools have been developed (130, 131) including MAGeCK (132), HiTSelect (133), and CasTLE (134). Broadly, these tools evaluate the enrichment or significance of each sgRNA to deduce gene level analyses and rankings.

The primary readout of a high-content pooled CRISPR screen is sequencing data containing barcode or gRNA counts, which can then be used to detect and match each perturbation to a corresponding cell. Furthermore, transcriptome data is generated for scRNA-seq-based screens, microscopy data for optical screens, or cytometry data for CYTOF proteomic screens. Many bioinformatic methods have been developed to analyze scRNA-seq-coupled CRISPR screen data, such as FBA, LRICA, MELD, MIMOSCA, MILO, Mixscape, MUSIC, Normalisr, SCEPTRE, and scMAGeCK (118, 135–141).

The initial step to analyzing a pooled scRNA-seq CRISPR screen is processing the raw sequencing data. This can be done by using standard scRNA-seq analysis tools, or tools tailored for scRNA-seq-CRISPR screens such as FBA and scMAGeCK (135, 141) utilizing the barcode or gRNA sequences to demultiplex cells. Multiple analysis pipelines have been developed including aggregating cells based on gRNA expression followed by differential gene analysis of transcriptome profiles (109), or assignment of transcriptome signatures to each gRNA perturbation using principal component analysis (19). Low-rank independent component analysis (LRICA) can also distinguish components with distinct genetic perturbations (51). Conversely, unsupervised clustering can instead first group cells based on transcriptome profile, followed by analysis of enriched gRNA within each cluster (16, 108, 113). The effect of a given perturbation can then be calculated by comparing the average transcriptome profile of perturbed to control cells (108). A more specific analysis tool, MIMOSCA (Multi-Input-Multi-Output-Single-Cell-Analysis), was additionally created for analysis of Perturb-seq data (54). Furthermore, perturbation modules and affected gene programs can be visualized using heat maps (54).

More recently, defined packages have been created to analyze scCRISPR screen data including FBA, MELD, MILO, Mixscape, MUSIC, Normalisr, SCEPTRE, and scMAGeCK (118, 135–141). These tools include additional CRISPR-specific quality control steps, such as removing unperturbed cells or cells with invalid edits (138–140), filtering a minimum number of cells per perturbation (138), or detecting off-target effects (139). A feature barcoding analysis (FBA) package can be used to process and cluster cells based on their perturbations (135). Manifold learning algorithms and dimension reduction algorithms that estimate and visualize the relative likelihood ratios of perturbations influencing phenotypes have also been developed (136, 142). Furthermore, the MELD algorithm uses manifold learning to generate relative likelihood scores that may then be used for vertex frequency clustering (VFC) of perturbed cells (136). Finally, differential expression analysis can be performed on clusters or neighborhoods to obtain perturbation signatures.

Packages that have been validated with CRISPRi, such as SCEPTRE, Normalisr, and scMAGeCK, can additionally detect gene enhancer links and regulatory networks (139–141). scMAGeCK’s linear regression-based method, scMAGeCK-LR, is capable of simultaneously analyzing the expression of all genes in a cell to deduce networks, including cells possessing multiple perturbations. Meanwhile, the Robust Rank Aggregation (RRA) component, scMAGeCK-RRA, identifies regulatory relationships by ranking cells based on expression of a certain gene of interest, which is then linked to the cell’s corresponding perturbation (141). Finally, specific perturbations may be ranked according to their effect on a cell’s phenotype using the MUSIC algorithm (138). This ranking may be according to an overall perturbation effect, a functional topic-specific effect, or a relationship between different perturbations. These emerging algorithms are promising tools to ease analysis of high-throughput single-cell CRISPR screens evaluating tumor immunology in the future.

Elucidation of gene function is key to the discovery of prospective therapeutic targets in many disease contexts. The CRISPR-Cas9 system has become a central tool in this search by enabling genetic screens to be easily performed at the genome scale. Recent studies have effectively coupled CRISPR screens to content-rich phenotypes using transcriptomic, epigenetic, proteomic, and optical readouts. These systems can deconvolute genotypic phenotypic relationships and have successfully identified regulators of numerous contexts including T cell stimulation and exhaustion, ICI resistance, and cancer immune evasion.

In recent years, CRISPR technology has been extended to the clinic (143–145) ( Figure 2D ). Initial proof-of-concept clinical trials in advanced esophageal (NCT03081715), metastatic non-small cell lung (146), and advanced, refractory cancer (NCT03399448) (147) have demonstrated CRISPR-Cas9 editing of autologous CAR T cells prior to adoptive transfer to be safe and feasible. Ongoing Phase I clinical trials have used CRISPR editing to remove disadvantageous genes from autologous CAR T cells, such as HPK1 (NCT04037566) or TGFβR (NCT04976218), to improve the T cell antitumor immune responses. CRISPR-Cas9 editing has also been used in allogeneic CAR T cells to diminish immunogenicity through the deletion of β2 microglobulin as well as enable precise insertion of the CAR construct into the endogenous TRAC locus (NCT04035434, NCT04557436, NCT04637763, NCT04244656, NCT04502446). The PoKI-seq and ModPoKI-seq studies highlight the potential of high-content CRISPR screens to enhance CAR T function (32, 110). In the future, high-content CRISPR screens in immune cells can be used to identify genes controlling antitumor activity, such as activation or effector functions ( Figure 1 ). These scientific advances can be rapidly translated into immune cell therapeutics with the potential to improve cancer care ( Figure 2D ).

Despite its influential application in gene discovery, there remain limitations to CRISPR screening. While in vitro genome wide screens can be performed in immune cells and have discovered novel regulators of T cell activation and proliferation (16, 90, 91), these readouts only partially approximate T cell biology. Pooled in vivo T cell screens can offer a more holistic evaluation of T cell function, but the low number of tumor-infiltrating lymphocytes can contribute to dropouts and false negatives (99). Additionally, many analysis tools for high-content CRISPR screens have yet to be extensively implemented, warranting future validation in the context of tumor immunology. Furthermore, although pooled CRISPR screens have been used to elucidate intercellular regulators governing cell-to-cell interactions (107), screens that focus on intercellular regulators remain largely unexplored. In the future, additional pooled CRISPR screens which focus on intercellular regulators are needed to better dissect the complexities of interactions within the tumor microenvironment.

In conclusion, high-content screens are likely to become a core approach for the future of CRISPR. The application of this technology to the tumor immunology field has the potential to reveal genes that regulate antitumor function. These genes may be therapeutically targeted in a clinical setting to improve patient outcome.

EH and MG conceptualized this review and drafted the original manuscript. AP, KJ, AT, JJ, LJ, and AH contributed to the writing and critical revision. All authors have approved the final submission.

Funding was provided to EH by the Immunology Training Program. Funding was provided to KJ by the Pharmacological Sciences Training Program (PSTP) T32 Training Grant (GM007767), the Rackham Merit Fellowship, and the Rackham Graduate School Research Grant. Funding was provided to AP by NIAID Training Grant T32 (AI007413). Funding was provided to MG by the Lung Precision Oncology Program (VA 150CU000182), LUNGevity, Veterans Affairs (I01 BX005267), Melanoma Research Alliance (MRA689853), NCI (CA252010), and the Breast Cancer Research Foundation.

Figures were created with BioRender.com.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.