Interrogating immune cells and cancer with CRISPR-Cas9

Abstract

CRISPR-Cas9 technologies have transformed the study of genetic pathways governing cellular differentiation and function. Recent advances have adapted these methods to immune cells, which has accelerated the pace of functional genomics in immunology and enabled new avenues for the design of cellular immunotherapies for cancer. In this review, we summarize recent developments in CRISPR-Cas9 technology and discuss how they have been leveraged to discover and manipulate novel genetic regulators of the immune system. We envision that these results will provide a valuable resource to aid in the design, implementation, and interpretation of CRISPR-Cas9-based screens in immunology and immuno-oncology.

CRISPR-Cas9 screens and immune regulation

Cas9 is an RNA-guided endonuclease, capable of cleaving DNA sequences in the mammalian genome. Cas9 is guided to target DNA sequences by clusters of regularly interspaced short palindromic repeats (CRISPR) RNAs, which pair with trans-activating CRISPR RNAs to facilitate ribonucleoprotein complex formation and DNA cleavage [1]. Cas9 genome-editing applications, as discussed in this review, generally use engineered single guide RNAs (sgRNAs) (see Glossary), which fuse CRISPR RNAs and trans-activating CRISPR RNAs into a single RNA molecule. Genome targeting is then specified by a ~20 nucleotide sequence contained in the sgRNA that binds a DNA region, which must neighbor a protospacer adjacent motif (PAM) in the genome.

The earliest adoption of CRISPR-Cas9 to edit mammalian genomes was achieved in the early 2010s, building on decades of previous characterization of the CRISPR immune system in bacteria [1–4]. After target recognition, the Cas9 nuclease creates a double-strand DNA break (DSB) that is repaired by error-prone endogenous DNA-repair pathways, typically non-homologous end-joining (NHEJ) [5,6,93]. NHEJ most frequently results in imperfect repair, leading to small insertions or deletions (indels) at the DSB site, in effect mutating the sgRNA target site in the genome and causing gene inactivation. The CRISPR-Cas9 system can also be modified to enable sequence insertion (knock-in) by directing the DNA damage response pathway toward homology-directed repair [7]. This is accomplished by providing a DNA repair template with homology to the DSB site, usually as DNA oligos or viral donor templates, which are engaged by endogenous homologous recombination machinery. This pathway can be exploited to incorporate novel sequences into specific regions of the genome targeted by CRISPR [2,3,8–10].

The power of CRISPR-Cas9 as a tool for genome editing lies in several strengths of this platform: (i) it is a two-component system that can be easy adapted to target specific genomic sites simply by changing the sgRNA sequence, (ii) it is highly efficient and specific, and (iii) Cas9 itself can be modified for diverse gene regulatory function, such as gene activation or repression, or specific editing of individual nucleotides (Box 1). In this review, we discuss CRISPR-Cas9-based screens (CRISPR screens) that have led to the discovery of novel mechanisms of immune cell regulation, either cell intrinsic regulation or cell-extrinsic regulation by cancer cells.

Box 1. The extended CRISPR-Cas9 toolbox.

The fundamental capabilities of CRISPR-Cas9 have been extended to an increasingly diverse toolbox of engineered Cas9 proteins with multiple functions. For example, catalytically deficient Cas9 (dCas9) retains its genome-targeting ability without creating double-strand breaks (DSB) upon binding [64]. The fusion of dCas9 to transcriptional effector domains has thus enabled targeted transcriptional inhibition (CRISPRi) or activation (CRISPRa). CRISPRi directly inhibits genes by sterically hindering the access of RNA polymerase; dCas9 has inhibitory capability alone, but recent studies have enhanced this capability by fusing dCas9 to a Krüppel-associated box (KRAB) domain to further inhibit the transcription of target genes [64,65]. Conversely, CRISPRa uses dCas9 fused to a transcriptional activator, which can recruit transcriptional machinery and RNA polymerase to target genes to induce transcription [65–70]. Furthermore, linking catalytically impaired Cas9 to a deaminase enzyme enables single base pair editing in the genome [71,72]. These Cas9 base editors (BE) shuttle the deaminase enzyme to the target site and initiate a repair response that leads to the conversion of a single base within a specified window. The current generation of BEs is capable of mediating all four possible mutations and have improved efficacy by incorporating a uracil excision repair inhibitor [71,73,74]. The uses of BEs range from correcting single nucleotide polymorphism disease variants to creating premature stop codons and splice-site disruptions to disrupt genes without creating DSB. Recent methods have leveraged a nickase Cas9/reverse transcriptase fusion protein bound to a chimeric sgRNA encoding the desired edit (dubbed ‘prime editing’), providing a strategy for homologous DNA repair that does not require DSB [75]. Furthermore, recently discovered Cas9 orthologs, such as Cas12a, also extend the range of possible edits [92].

Immune cell CRISPR screens

Precise developmental and functional regulation of immune cells is crucial to an organism’s ability to generate robust immunity to pathogens and cancer. The initial discovery of CRISPR-Cas9 was rapidly followed by its optimization for human and murine systems, including immune cells, both in vitro and in vivo [11–16]. Furthermore, computational optimization of sgRNA design has improved the efficiency and specificity of genome targeting in these systems [17–20]. Over the past few years, these advances have been coupled with traditional immune cell assays, drastically improving the scope and throughput of functional genetic screening in immunology (Figure 1).

Figure 1. General approaches for clusters of regularly interspaced short palindromic repeats (CRISPR)-Cas9 mediated discovery in mammalian immune systems.

Shown are distinct phenotypic readouts that can be used to determine immune mechanisms with CRISPR screens. (Left) Protein expression screens can identify genes responsible for positively or negatively regulating the protein of interest. (Center) Cell expansion screens can reveal regulators of cell proliferation or immune effector/target cell sensitivity. (Right) Single cell RNA-seq (scRNA-seq) screens can assess transcriptional states of CRISPR-modified cells, which enable the interrogation of cell state regulators, as well as coregulated gene programs. Shown on the bottom panels are example findings from each approach. USP22 deubiquitinates the Foxp3 locus to promote and maintain FOXP3 expression in murine regulatory T (Treg) cells [31]. PTPN2 inhibits the IFN response in murine melanoma cells and promotes tumor cell resistance to T cell killing [46]. Transcription factors IRF8 and CEBPB drive distinct cell fates in bone marrow cells in response to lipopolysaccharide (LPS) stimulation, with IRF8 promoting a dendritic cell fate and CEBPB promoting a monocyte/macrophage cell fate [63]. Abbreviations: NK, natural killer; WT, wild type.

Murine CRISPR screens in myeloid cells

The development of the Cas9-expressing transgenic mouse enabled pooled CRISPR screens in Cas9⁺ immune cells [21,22]. In an early study, the authors performed a CRISPR screen in mouse bone marrow-derived dendritic cells (BMDCs) stimulated with the bacterial component and toll-like receptor (TLR) 4 agonist, lipopolysaccharide (LPS), to validate the use of CRISPR screens in a well-defined signaling pathway and to discover novel regulators of BMDC pathogen response [16]. Here, Cas9⁺ BMDCs transduced with a pooled genome-wide lentiviral sgRNA library were stimulated with LPS in vitro, then sorted for expression of tumor necrosis factor alpha (TNFα; an inflammatory cytokine produced by BMDCs in response to LPS stimulation). The top ten positive TNFα regulators were consistent with canonical members of the TLR signaling pathway, validating the power of CRISPR screens in recapitulating key components of entire signaling pathways in a single assay. Further experimental analysis of these genes suggested that regulators of the BMDC response to LPS fell into three distinct modules; these modules were defined by their relative ability to regulate TNFα, IL-6, CD14, MIP-1α, and CD11c protein expression, as well as the distinct transcriptional programs driven by targeting these genes. Two of the identified modules contained members of protein complexes that were not previously known to affect the LPS response: (i) the oligosaccharyltransferase complex (OSTc), an endoplasmic reticulum (ER)-resident complex that modifies nascently translated proteins with oligosaccharide chains important for protein folding and ER transport; and (ii) the PAF complex, which regulates transcription elongation and 3′ mRNA processing. Follow-up experiments confirmed that OSTc components, including Dad1, mediated the expression of early inflammation-related genes, whereas PAF complex members, Paf1 and Rtf1, mediated the expression of antiviral-related genes and a distinct set of sustained inflammation-related genes, highlighting these complexes as novel and synergistic regulators of the innate immune response to LPS. More broadly, this study provided an elegant proof-of-concept that CRISPR screens were capable of identifying novel and biologically relevant pathways in immune cells [16].

A second early immune cell screen sought to identify regulators of antigen crosspresentation in type 1 conventional dendritic cells (cDC1s) [23]. cDC1 progenitors isolated from Cas9⁺ mice were transduced with retroviral sgRNA constructs targeting 94 genes highly expressed in cDC1s, then cultured under cDC1 differentiation conditions. To assess the effect of these knockouts on cross-presentation, the authors quantified OT-1 CD8⁺ T cell (OT-1 cell) proliferation in response to edited cDC1s cultured with cell-associated SIINFEKL antigen in vitro. In this screen, two independent sgRNAs targeting Wdfy4 (a transmembrane BEACH domain-containing protein family member) completely ablated the ability of cDC1s to induce proliferation of OT-1 cells, suggesting an essential role for WDFY4 in the cross-presentation of cell-associated antigens. Wdfy4^−/− mice had specific defects (relative to Wdfy^+/− mice) in the ability to prime transferred OT-1 cells in response to vaccination, as well as endogenous antigen-specific CD8⁺ T cells in response to cowpox or West Nile virus infection [23]. Furthermore, Wdfy4^−/− mice were unable to reject subcutaneously implanted fibrosarcoma cells, emphasizing an essential role for WDFY4 in mediating crosspresentation of tumor antigens. Together, these studies highlight the ability to incorporate CRISPR screens into complex functional assays to identify novel cell-intrinsic response pathways in immune cells.

Murine CRISPR screens in T cells

The T cell response has been the focus of several recent CRISPR screens, with a particular emphasis on the CD8⁺ T cell response to cancer. In vivo models are particularly important in studying the relationship between T cells and cancer, due to the complex modes of suppression that tumors can exert on T cells though cell–cell interactions, secreted factors, and metabolic changes. One study attempted to identify metabolic regulators of the T cell response to cancer using a pooled sgRNA library targeting 3017 metabolic enzymes in murine CD8⁺ T cells [24]. CRISPR-modified CD8⁺ OT-1 cells were adoptively transferred into B16-Ova melanoma-bearing mice and then isolated from the tumor 7 days later to measure sgRNA abundance. Regnase-1-targeted T cells were the most highly enriched genotype in the tumor. Since the enrichment of an sgRNA in the tumor suggests that deletion of the targeted gene improved intratumoral T cell infiltration, this result suggested that REGNASE-1, a ribonuclease and RNA-binding protein, negatively regulated T cell tumor infiltration. Indeed, adoptively transferred Regnase-1-targeted CD8⁺ T cells were more effective at controlling tumor growth and promoting survival (relative to control sgRNA-targeted cells) in C57BL/6 mice harboring melanoma or B cell leukemias, suggesting that the role of REGNASE-1 was conserved across tumor types. Furthermore, Regnase-1-targeted T cells expressed higher amounts of naïve and memory genes (by RNA-seq), including Tcf7, Lef1, and Bach2, compared with control T cells, suggesting that REGNASE-1 deletion reprogrammed the CD8⁺ T cell state. The mechanism of REGNASE-1 signaling was investigated using a secondary genome-wide screen in Regnase-1-targeted T cells. Batf-targeted, Regnase-1-targeted OT-1 cells were depleted in tumors, relative to other Regnase-1-targeted OT-1 cells, suggesting that Regnase-1-targeted T cell expansion and tumor control were mediated by the transcription factor (TF) BATF [24].

The mechanisms of the CD8⁺ T cell antitumor response were further investigated in a study that used two parallel in vitro and in vivo genome-wide screens. Here, edited OT-1 cells were assayed for their ability to infiltrate orthotopically implanted E0771-Ova murine triple-negative breast cancer (TNBC) in vivo in mice, as well as their ability to degranulate (measured as CD107a expression) in response to these cells in vitro [25]. Negative regulators of the T cell response were identified as hits in both screens (enriched in the tumor infiltration screen and increased degranulation in vitro); only three genes (Dhx37, Lyn, and Odc1) were shared as hits. In validation experiments, Dhx37-targeted T cells more effectively rejected TNBC tumors in vivo, compared with Odc1-targeted and control T cells, and therefore DHX37, an RNA helicase with no known role in T cells, was selected for further characterization. In response to IL-2 and IL-7 stimulation, Dhx37-targeted T cells had increased expression of T cell activation markers, such as CD69, granzyme B, and several checkpoint markers, relative to nontargeted T cells. Immunoprecipitation (IP) of DHX37 in primary human CD8⁺ T cell lysates demonstrated that it interacted with the NF-kB binding protein, PDCD11, and the NF-kB TF p65, suggesting that the effect of DHX37 may be mediated by NK-kB. Finally, the authors demonstrated that luminal-A and TNBC breast cancer patients with high DHX37 expression had a lower survival rate compared with those with low DHX37 expression, suggesting that DHX37 also played a role in the CD8⁺ T cell response to human cancers [25].

The same group performed an independent in vivo screen to identify potential strategies to enhance the CD8⁺ T cell response to glioblastoma (GBM) [26]. Antigen-specific CD8⁺ T cells were edited using an AAV-Sleeping Beauty sgRNA library targeting 1658 cell surface protein-encoding genes and transferred into C57BL/6J mice harboring orthotopically implanted GL261 GBM tumors. sgRNA enrichment was quantified in the brain relative to the preinjection T cell pool and several hits were identified across replicates, including Lag3, a known T cell checkpoint, and Pdia3, Mgat5, and Emp1, whose roles in the T cell response are less clear. In single-gene targeting adoptive transfer studies, Pdia3, Mgat5, and Emp1-targeted T cells significantly improved GBM-engrafted mouse survival, when compared with nontargeted T cells. Of note, Pdia3-deficient T cells improved survival in several different tumor types, highlighting a conserved function for Pdia3 in restraining the effector T cell response across tumor types [26].

To interrogate the effector versus memory T cell lineage choice, a recent study performed a CRISPR screen in CD8⁺ T cells responding to lymphocytic choriomeningitis virus (LCMV) infection [27]. A custom sgRNA pool targeting 120 TFs was designed using a DNA-binding domain-targeted sgRNA design approach [28]. CD8⁺ T cells edited with this library were subsequently injected into LCMV-infected C57BL/6 mice and harvested from several anatomical sites at 1 and 2 weeks postinfection to quantify sgRNA enrichment [27]. Known positive regulators of effector T cell function and differentiation, such as Irf4, Batf, and Myc, were strongly depleted in mice infected with acute (Armstrong) or chronic (Clone 13) LCMV strains, confirming the validity of the approach. Additionally, several novel genes without known functions in effector T cell differentiation were among the most enriched, including Atf6, Irf2, and Fli1, suggesting that they repressed effector CD8⁺ T cell differentiation. Fli1-targeted CD8⁺ T cells were the most positively selected, indicating that this ETS TF played a role in restraining effector T cell differentiation. FLI1-bound, regulated chromatin was characterized using the assay for transposase accessible chromatin with sequencing (ATAC-seq) and CUT&RUN, which suggested that FLI1 could bind and repress the accessibility of RUNX family TF chromatin binding sites to repress effector cell differentiation. Confirming these observations, targeting Fli1 in CD8⁺ T cells increased effector cell number without reducing memory cell formation, relative to nontargeted controls [27]. Together, these data identified FLI1 as a key regulator of effector T cell differentiation and suggested that FLI1 deletion might improve T cell-based immune responses without compromising long-term immune memory.

Several studies have also used CRISPR screens to identify regulators of CD4⁺ T cell differentiation. A genome-wide retroviral CRISPR screening approach was used to identify genes that controlled CD4⁺ T helper 2 (Th2) cell differentiation [29]. Here, using the initial screen results, as well as RNA-seq and ATAC-seq profiling results, the authors selected 40 genes for further characterization. The results of targeting each individual candidate gene revealed that perturbing genes implicated in Th2 differentiation often altered Th2 differentiation and T cell activation, suggesting that these processes might be tightly linked in CD4⁺ T cells. Further characterization of these genes highlighted the known role of Stat6 as a mediator of Th2 differentiation, as well several novel genes in Th2 differentiation, including Lrrc40 and Ccdc134 [29], a cytokine-like gene previously implicated in CD8⁺ T cell effector function [30].

Finally, two independent groups performed CRISPR screens in regulatory T cells (Tregs) to identify regulators of FOXP3, the key Treg lineage-defining TF [31,32]. In one study, the nuclear regulators of FOXP3 expression in primary mouse Treg cells were studied using an sgRNA library targeting 486 genes encoding nuclear factors [31]. This approach identified known regulators, including Cbfb, Runx1, and Satb1, as well as novel regulators, including the deubiquitinase, Usp22, and the E3 ubiquitin ligase, Rnf20. Screen hits were confirmed using individual ribonucleoprotein sgRNA-Cas9 complexes (RNPs), which showed that targeting Usp22 in both mouse and human Tregs (USP22) decreased FOXP3 expression, again suggesting a role in positively regulating FOXP3 expression. USP22 is required for the SAGA-mediated deubiquitination of histones, which can regulate the transcriptional activity of genes [33]. Chromatin precipitation followed by genome-wide sequencing (ChIP-seq) confirmed that Usp22-deficient Tregs had aberrant ubiquitination at the Foxp3 locus, compared with wild type (WT) mice [31]. In vivo experiments further confirmed the regulation of FOXP3 by USP22: (i) Usp22^fl/fl Foxp3^YFP-cre (Usp22^TregKO) mice harbored fewer Tregs than WT mice; (ii) Usp22^TregKO mice challenged with MOG peptide-induced experimental autoimmune encephalomyelitis had more severe disease, compared with WT mice; and (iii) Tregs transferred from Usp22^TregKO mice to WT mice were not able to protect against effector CD4⁺ T cell-driven colitis in Rag1^−/− mice, whereas WT Tregs effectively prevented colitis. Finally, disrupting Rnf20, a ubiquitin ligase identified as a negative regulator of FOXP3 in this screen, restored FOXP3 expression Tregs from Usp22^TregKO mice to WT levels, suggesting that FOXP3 expression might be reciprocally regulated by USP22 and RNF20 and that Rnf20 deficiency might compensate for Usp22 loss [31].

In the second study, the authors used a similar strategy to perform a genome-wide screen for regulators of FOXP3 expression [32]. Here, Cas9⁺ Treg cells were isolated and transduced with a retroviral library, then sorted for both high and low FOXP3 expression. Sequencing these populations revealed 254 positive and 490 negative FOXP3 regulators. Among these hits were several functional subunits of the SAGA complex, including the previously identified Usp22 [31] and SWI/SNF complex factors, again suggesting that these complexes might play a role in regulating FOXP3 expression in Tregs [32]. The functions of SWI/SNF complex members were further tested by targeting specific members of the BAF, ncBAF, and PBAF complexes and measuring FOXP3 protein expression. These functional studies revealed that the ncBAF complex, specifically Brd9 and Smarcd1, positively regulated FOXP3 expression, while the PBAF complex negatively regulated FOXP3 expression. ChIP-seq confirmed that BRD9 occupied the FOXP3 locus, as well as a subset of FOXP3 target genes. Functionally, Brd9-targeted Treg cells were less able to protect Rag1^−/− mice from CD4⁺ T cell-induced colitis and less effective in inhibiting effector CD4⁺ and CD8⁺ T cell responses in Rag1^−/– mice challenged with MC38 tumor cells, compared with WT Tregs [32]. Altogether, these two studies in primary murine Tregs revealed novel mediators of FOXP3 expression and the role of chromatin regulation in maintaining the suppressive function of Tregs.

Several challenges remain in immune cell screening, particularly in in vivo screens, which remain difficult due to the large number of cells needed to perform genome-wide screens, and the immune rejection of cells presenting Cas9-associated antigen in immunocompetent mice [34,35]. Potential solutions to these challenges include the use of targeted libraries (to decrease cell number requirements), and transiently providing immunogenic editing components or screening directly in Cas9⁺ mice (to prevent immune rejection) [36]. Nevertheless, these studies demonstrate that CRISPR screens may be feasible in a wide variety of immune cell types, and future studies may continue to advance these strategies to interrogate a breadth of immune-related questions using in vivo models and high-dimensional readouts (Box 2).

Box 2. Single cell CRISPR screening.

Screens based on proliferation, cell survival, or the expression of single gene markers are limited in their ability to assay complex cell phenotypes. For example, it can be difficult to parse T cell or cDC1 differentiation states using single cell surface markers and, similarly, cellular activation states often comprise multiple gene modules whose activities may not be accurately reflected by one cytokine. To address this, several groups have combined CRISPR-Cas9 screens with single cell RNA-seq (scRNA-seq [76,77]) to read-out the effect of CRISPR perturbations on genome-wide transcriptional networks and cell states. Perturb-seq and CRISP-seq were the earliest implementations of this strategy and relied on an sgRNA vector with expressible barcodes that could be captured using existing droplet-based scRNA-seq methods [63,78,79]. Perturb-seq was first used to capture the transcriptional effect of knocking out 24 transcription factors (TFs) in LPS-stimulated BMDCs [78]. The authors identified ‘modules’ of TFs with similar regulatory effects and defined groups of genes that were similarly affected by TF perturbation (gene programs). This study identified four TF modules that directed five gene programs, each characterized by distinct biological processes in response to LPS stimulation. They included known regulators of the BMDC response to LPS (e.g., Stat1 and Stat2 in the antiviral gene program) and identified novel regulators (e.g., Irf8 regulation of antiparasitic guanylate-binding proteins) [78].

Single-cell CRISPR screening can also be achieved through the use of expressible protein barcodes in the form of antigen tags [80]. This method utilizes high-dimensional mass cytometry (CyTOF) to detect combinations of antigen tags expressed from the individual lentiviral sgRNA backbone: one antigenic tag aligns with one sgRNA. The advantage of this system is that the user can directly measure targeted protein expression to determine cell phenotype changes in response to specific gene knockouts. Demonstrating its strength, one study simultaneously measured PD-L1 expression changes in response to Socs1 knockout in a pooled screen in 4T1 mouse mammary gland tumor cells challenged with antigen-specific CD8⁺ T cells, implicating SOCS1 as a novel regulator of PD-L1.

Although powerful, these applications of single-cell CRISPR screening have been limited by low throughput, complex analytical methods, specialized vector systems, and vector recombination during processing, leading to sgRNA:barcode swapping [63,78,79,81,82]. Direct-capture Perturb-seq, a new iteration of Perturb-seq, solves the issue of potential recombination of barcoded vectors by directly sequencing the sgRNA using an sgRNA-specific primer during reverse transcription [59]. This study provided a method for reducing sequencing costs by enriching selected transcript targets through a hybridization approach, effectively depleting noninformative transcripts accounting for a large portion of sequencing reads. Continued innovations in single-cell sequencing technologies can continue to increase the utility of single-cell CRISPR screening in immune cells.

Human CRISPR screens in T cells

The earliest CRISPR screens in human immune cells used CRISPRa/i techniques to interrogate the identity and function of non-protein coding DNA in immune gene regulation. A CRISPRa screen was performed in a dCas9-VP64⁺ Jurkat T leukemia cell line to identify gene enhancers in two autoimmune risk loci, CD69 and IL2RA [37]. sgRNA libraries were used to ‘tile’ the region encompassing each gene locus and CD69⁺ or IL-2Rα⁺ cells were sorted and sequenced to identify enriched sgRNAs. dCas9-VP64 activation of six regions in ~178 kilobases (kb) flanking the IL2RA gene resulted in increased IL-2Rα expression and activation of three regions in ~135 kb flanking the CD69 gene resulted in increased CD69 expression (relative to untransduced cells). These enhancer regions overlapped sites marked by H3K27 acetylation that also contacted the IL2RA promoter in chromosome conformation assays, further supporting their gene enhancer function. Furthermore, one of the IL2RA enhancer elements contained a causal single nucleotide polymorphism associated with type 1 diabetes and Crohn’s disease, suggesting that perturbation of this enhancer might contribute to disease pathogenesis in patients. In support of this hypothesis, CD4⁺ effector T cells from autoimmune-prone nonobese diabetic mice lacking this enhancer had reduced IL-2Rα expression in response to in vivo T cell receptor (TCR) stimulation with anti-CD3 antibodies. These findings demonstrated the ability of CRISPR screens to identify functional noncoding regulatory elements in the genome and suggested that stimulation-responsive enhancers may play a role in inflammatory conditions [37].

The utility of CRISPR screens in interpreting autoimmune disease risk variants has also been performed using gene level analysis. A pooled, genome-wide CRISPR screen was used to identify positive and negative regulators of TCR signaling in Jurkat T cells, using CD69 expression as a surrogate for TCR signaling strength [38]. In this screen, the top-ranked positive regulators were highly enriched for known members of the TCR signaling pathway, confirming the validity of the readout. Among novel putative negative regulators, FAM49B was chosen for further validation because of its high expression in lymphoid organs and reported genetic associations with multiple sclerosis [39]. IP combined with mass spectrometry confirmed that FAM49B directly bound Rac1, implicating a function in Rac1-driven actin polymerization during cytoskeletal remodeling. These results suggested that disease-associated T cells might be deficient in this negative regulator of TCR signaling, leading to overactivation to self-antigens, though further studies in primary T cells are needed to validate this concept.

These screens were performed in Cas9-expressing cell lines due to limitations in delivering Cas9 to primary immune cells. However, recent studies have overcome this challenge by developing techniques to transiently deliver Cas9 protein to primary T cells via electroporation [40,41]. This approach was used to study T cell signaling at genome-scale directly in primary CD8⁺ T cells [41]. Here, primary CD8⁺ T cells were transduced with a genome-wide lentiviral sgRNA library, electroporated with Cas9, then stimulated with anti-CD3/CD28 to induce proliferation. Positive and negative regulators of TCR signaling were identified by measuring sgRNA enrichment in the most and least proliferative T cells (selected using carboxyfluorescein succinimidyl ester staining). Putative positive regulators of TCR signaling were highly enriched for known TCR signaling pathway members, including those identified in the previous study [38] and hits previously found in an shRNA screen in mouse T cells [42]. Several negative regulators were further interrogated for their ability to enhance CD8⁺ T cell function; RASA2, CBLB, SOCS1, and TCEB-targeted CD8⁺ T cells exhibited increased expression of CD69 and CD40L and enhanced killing of target A375 melanoma tumor cells in cytotoxicity assays, compared with nontargeted T cells [41]. T cell intrinsic-mediators of immunosuppressive adenosine signaling were studied in a follow-up screen; here, T cell proliferation was assessed in the presence of an adenosine receptor agonist as a proxy for adenosine-mediated suppression. This screen identified ADORA2A, the expected receptor specifically targeted by the adenosine agonist used in the assay, and FAM105A, which had no previously validated role in primary T cells [41]. In summary, these methods for CRISPR screening in human T cells present exciting opportunities to understand gene regulation in primary cells. Recent reports of CRISPR editing in primary human B cells and monocyte derived-dendritic cells may further extend the possibilities of human immune cell CRISPR screens [43–45]. Furthermore, future screens may increasingly focus on complex immune phenotypes beyond those of proliferation or stimulation; high-dimensional readouts will be particularly useful in studying complex cell states (Box 2).

CRISPR screens in cancers

Several CRISPR screens have focused on understanding cancer cell-intrinsic mechanisms of resistance to CD8⁺ T cell-mediated killing. In these studies, pooled CRISPR-Cas9-modified cancer cell lines were cocultured with effector immune cells, such as CD8⁺ T cells, to identify gene knockouts that rendered the tumor cells resistant or sensitive to immune cell killing (Figure 1). In one such screen, murine B16 melanoma cells were transduced with an sgRNA library targeting 2368 genes that encode specific functional protein classes suspected to be important in mediating tumor/immune cell interactions, such as kinases, cell surface proteins, and mediators of antigen processing and presentation [46]. Modified B16 cells were implanted in WT mice, which were subsequently treated with an irradiated tumor cell vaccine modified to secrete the cytokine granulocyte-macrophage colony stimulating factor (GVAX), with or without anti-programmed death-1 (PD-1) checkpoint antibody immunotherapy, to elicit an antitumor immune response causing T cell proliferation and tumor shrinkage. In 12–14 days after transplant, tumor sgRNA depletion was quantified, relative to tumors implanted into Tcra^−/− mice (which lack T cells), to identify knockouts that sensitized cancer cells to T cell killing. The most-depleted sgRNAs in immunotherapy-treated mice were enriched in four biological pathways: TNF signaling/NF-kB activation, antigen processing and presentation, inhibitory kinase signaling, and the ubiquitin-proteasome pathways. Representative members of each class (Ptpn2, H2-T23, Ripk1, and Stub1, respectively) were validated in in vivo competition assays, where red fluorescent protein expressing (RFP⁺) single gene-targeted tumor cells were co-injected with green fluorescent protein expressing (GFP⁺) control cells. Targeting each of the four genes decreased RFP⁺ cell survival in immunotherapy-treated mice, relative to GFP⁺ control cells. Further, Ptpn2-targeted B16-Ova presented more SIINFEKL antigen after in vitro IFNγ treatment, which led to increased granzyme-B⁺ CD8⁺ T cell infiltration in Ptpn2-targeted B16-Ova in vivo [46]. These data were consistent with previous studies showing that PTPN2 negatively regulated IFNγ signaling by dephosphorylating JAK1 and STAT1 [47,48]. The same screen also identified the RNA-editing enzyme ADAR1 as a novel checkpoint of antitumor immunity in the same model of murine melanoma [49].

A separate in vitro CRISPR screen was used to identify mechanisms of human melanoma resistance to CD8⁺ T cell killing [50]. Mel624 melanoma cells were transduced with a genome-wide sgRNA library and cocultured with antigen-specific CD8⁺ T cells. sgRNA abundance was quantified in surviving Mel624 cells to identify gene knockouts that modified tumor cell sensitivity to T cell killing. sgRNAs targeting genes responsible for antigen processing and presentation, such as TAP1 and B2M, and known mediators of the IFNγ and TNFα pathways, were enriched in the surviving Mel624 population, relative to the starting tumor cell pool, confirming that the screen successfully recovered mediators of sensitivity to CD8⁺ T cell killing. Seventeen novel hits were selected for validation, which revealed that targeting APLNR, the apelin receptor gene, in A375 melanoma cells led to decreased β2M expression after coculture with antigen-specific CD8⁺ T cells. CD8⁺ T cells cocultured with APLNR-targeted A375 produced less IFNγ, relative to CD8⁺ T cells cultured with nontargeted A375 cells, and Aplnr-targeted B16 cells implanted in vivo were less sensitive to killing by antigen-specific CD8⁺ T cells, confirming the importance of Aplnr in vivo [50]. Finally, mutations in APLNR were identified in melanoma and lung cancer patients treated with checkpoint immunotherapy [51] and these mutant forms of APLNR were cloned and transformed into APLNR-knockout A375 cells. Several of these mutations mediated a similar resistance phenotype as CRISPR-mediated disruption of APLNR, suggesting that APLNR loss might also play a role in cancer immunotherapy resistance in humans [50].

The PBAF form of the SWI/SNF chromatin remodeling complex was also identified as a mediator of resistance to CD8⁺ T cell-mediated killing using CRISPR screens [52]. CRISPR-Cas9-edited B16 cells were challenged with antigen-specific CD8⁺ T cells in vitro and sgRNA depletion in the resulting B16 population was quantified to identify mutants preferentially sensitized to T cell killing. Several known immune regulators, including previously identified members of the NF-κB pathway [46], were identified, as well as several genes whose functions in immune regulation are not well characterized. All three components of the PBAF chromatin remodeling complex, Arid2, Pbrm1, and Brd7, were depleted in the screen, suggesting that this complex was important in mediating T cell/tumor interactions. Mice injected with Pbrm1-targeted tumor cells and treated with anti-PD-1/CTLA-4 antibody checkpoint therapy exhibited increased tumor rejection in vivo, relative to nontargeted tumor cells, validating the results of the screen. Consistent with the known role of the SWI/SNF complex in regulating chromatin accessibility, Pbrm1-targeted cells had increased chromatin accessibility at a number of IFNγ-induced genes, relative to control tumor cells, demonstrating that the PBAF complex can suppress inflammatory gene expression in tumor cells in response to T cell recognition [52].

Finally, IFNγ-independent signaling pathways that sensitize tumor cells to CD8⁺ T cell killing were also interrogated using a genome-wide CRISPR screen in a novel IFNγ receptor-deficient (IFNGR1-deficient) tumor cell system [53]. By culturing CRISPR-Cas9-edited IFNGR1-deficient human melanoma cells with MART-1 antigen-specific T cells, the authors identified several sgRNAs that sensitized melanoma cells to T cell killing. The top two hits in this assay were TRAF2 and BIRC2, two genes associated with the TNF-induced cell death pathway, suggesting that the TNF pathway was a second key pathway in T cell-mediated killing and might be particularly important in IFNGR1-deficient tumors. Follow-up experiments confirmed that TRAF2 loss sensitized human and murine tumor cells to TNF-induced T cell killing via caspase-8-mediated cell death, which was supported by in vitro studies showing higher caspase-8 activation in TRAF2-targeted melanoma cells cultured with antigen-specific CD8⁺ T cells [53].

Altogether, these studies have confirmed key roles for antigen processing and presentation, IFNγ signaling, and the TNF response in antitumor immunity and identified novel members of these pathways that are mutated in patient tumors and may represent new therapeutic avenues for activating the immune response to cancer (Box 3 and Figure 2). These screens have also been extended to additional tumor–immune cell interactions, including with natural killer (NK) cells [54–56] and chimeric antigen receptor (CAR) T cells [57]. A significant caveat is that many of these screens have been performed in vitro using two-cell interaction systems, or in immune-deficient in vivo models where exogenous tumor-specific T cells are provided (due to the inherent immunogenicity of CRISPR components in immune-competent mice), which does not account for the complex multicellular interactions that occur during an antitumor immune response. A recent study demonstrated excision of CRISPR components from tumor cells after editing before transplantation; this might provide an experimental platform to perform CRISPR screens in vivo to identify additional putative immune evasion and response mechanisms in immunocompetent mice [58].

Box 3. Preclinical applications of CRISPR-Cas9 editing in immune cells.

Primary human T cells are an attractive target for therapeutic editing since they are: (i) easily cultured and genetically modified in vitro, and (ii) an effective therapeutic modality in diverse cancer types (Figure 2). The development of RNA and protein-based CRISPR-Cas9 reagents has led to higher efficiency editing, relative to plasmid-based reagents [8,83,84]. Many efforts to manipulate primary human T cells with CRISPR have focused on improving CAR T cells. In an early study, the TCR constant regions and β-2 microglobulin (B2M; required for HLA-I expression) were targeted with CRISPR-Cas9 to generate CAR T cells with reduced off-target alloreactivity and graft-versus-host disease (GVHD) in mice [94]. In an in vivo GVHD model in NOD/scid/γc^−/− (NSG) mice, TCR^– and TCR/HLA^– T cell-treated mice did not exhibit the severe GVHD and weight loss experienced by NSG mice treated with wild type (WT) T cells. This study also provided evidence that deleting PD-1 might be therapeutically effective in PD-L1⁺ tumors by targeting PD-1 in prostate stem cell antigen (PSCA)-specific CAR T cells; PD-1-targeted cells better protected NSG mice against PD-L1⁺ PC3 tumors, compared with nonedited PSCA-specific CAR T cells [94]. These results have been confirmed in another publication [85].

The PD-1, TCR, B2M triple knockout approach was also achieved using base editing (BE). BE approaches are attractive, achieving less potential off-target double-strand breaks (DSB) and chromosome translocations than WT Cas9 [86–88]. Several approaches were evaluated using BE variants to create premature stop codons, or disrupt splice acceptors or donor sites [89]. Targeting PDCD1, TRAC, or B2M loci showed that disrupting splice acceptor or donor sites with BE led to targeted protein expression loss comparable with that observed using WT Cas9 [89]. BE T cells had less off-target editing events, quantified by GUIDE-seq, and undetectable chromosomal translocations, quantified by digital droplet PCR and fluorescent in situ hybridization, compared with matched WT Cas9 controls.

CRISPR can also be used to introduce insertions to specific genomic locations through simultaneous DNA donor delivery. This was used to target a CD19-specific CAR cDNA construct to the TRAC locus [9]. By delivering TRAC-targeting RNPs and an adenovirus vector encoding self-cleaving P2A peptide sequence and CD19 CAR cDNA flanked by homology arms, high on-target integration of CD19-CAR into the TRAC locus was achieved. CAR T cells with targeted CD19 integration had reduced constitutive signaling; CAR expression more closely mirrored physiological TCR expression and improved tumor control when compared with retrovirally transduced CAR T cells [9]. Together, these results suggest that targeted genomic placement of CAR constructs may have therapeutic value compared with random viral integration.

Recent work showed that efficient CRISPR-enabled gene insertion can also be achieved with long, single-stranded (ss) or double-stranded (ds) DNA templates [10]. Using nonviral DNA templates and CRISPR-editing, IL2RA mutations were corrected in patient cells and a transgenic TCR sequence was inserted into the TRAC locus, demonstrating the feasibility of this approach, which might complement targeted insertion strategies in cellular therapy. This can also be extended to screening synthetic gene constructs, although possibly currently limited in scale [90]. One caveat is that current CRISPR strategies in immune cells are performed ex vivo, involving several time-consuming and costly steps such as isolating, activating, and expanding cells. Future approaches may focus on developing targeted strategies, including antigen-specific viral delivery methods for in vivo therapeutic gene editing.

Figure 2. Clusters of regularly interspaced short palindromic repeats (CRISPR) approaches for cellular immunotherapy.

Shown are strategies for cellular engineering using CRISPR technologies. (Top) A representation of a gene locus containing three genes that are transcribed and translated. (Top left) Wild type (WT) Cas9 is used to target a gene, or several genes, to promote gene knockout through the error-prone DNA repair pathways, which lead to disrupted transcription or translation. (Top right) CRISPR-mediated gene insertion can be used to insert new sequences into specific loci, including to imbue new specificity of an antigen receptor or alter the intracellular signaling component of a receptor to enable novel function. (Bottom left) CRISPR activation/inhibition can be used to promote or silence gene expression without gene disruption. (Bottom right) Next-generation genome editing will enable multiplexing of many types of edits, including base editing-mediated gene modifications, gene insertions, and gene activation/inhibition.

Concluding remarks

In this review, our goal was to highlight recent studies that have leveraged CRISPR screens to uncover known and novel mechanisms by which immune cells differentiate, function, and interface with cancer and other target cell types. We envision that the next decade of immunological research will draw heavily on the early applications of this technology, described here, to continue unveiling novel mechanisms and pathways across diverse immune cell types and functionally relevant biological systems. Moreover, the coupling of CRISPR screens to single cell transcriptome and epigenetic readouts can expand the utility of these methods to interrogate complex phenotypes where selection is not easily performed [59,60]. In many ways, the challenge in coming years will no longer be to implement CRISPR targeting in immune cells, or to perform pooled genetic perturbations, but rather, to design appropriate screening strategies to access unique cellular phenotypes and behaviors (see Outstanding questions). Finally, early advances in the use of CRISPR in cellular immunotherapy are promising and have rapidly been implemented in first-in-human clinical trials (Box 4) [61,62]. These pioneering studies have focused primarily on achieving acceptable safety profiles and T cell products that are not rejected by a patient’s immune system. With these encouraging studies ongoing or completed, the field can begin to focus on translating the insights gained from genome-wide forward genetic screens in immune cells into safe, viable, and targeted cell products that might enable the treatment of cancers, autoimmunity, and a growing list of human diseases.

Outstanding questions.

To date, the vast majority of immune CRISPR screens have been performed in T cells, or in tumor cells in the context of T cell cytotoxicity. Will the broad adoption of these techniques also lead to similar novel insights in myeloid cell and B cell biology?

To date, the majority of CRISPR studies have focused on coding genes. However, will the editing of noncoding regions (e.g., highly specific enhancers) lead to increased specificity and molecular control of desired immune cell states and behaviors?

Will Perturb-seq and other techniques that couple single-cell transcriptome and epigenome profiling with CRISPR screens lead to new insights into the regulation of complex immune phenotypes that cannot be assayed with simple selection pressures?

Will the continued development of CRISPR-mediated gene insertion and base editing techniques enable precision engineering and regulation of immune gene drives? And will these techniques allow immunologists to design novel or hybrid immune cell phenotypes that are not found in nature?

Will multiplex CRISPR editing of key genes improve immunotherapy efficacy in humans and will this be adopted in the clinic?

Box 4. Therapeutic primary T cell CRISPR editing in the clinic.

Building on years of development of clinical gene editing protocols using zinc-finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs), CRISPR-Cas9 was quickly introduced into the clinic [91]. Recently, the clinical and manufacturing feasibility of CRISPR-modified adoptive cell therapy was demonstrated by a first-in-human Phase I nonrandomized, clinical trial that tested multiplexed CRISPR editing in NY-ESO-1TCR-expressing T cells in three cancer patients (multiple myeloma, melanoma, or synovial sarcoma plus myxoid/round cell liposarcoma), measuring safety, tolerability, and adverse events (NCT03399448)ⁱ [61]. This study edited patient T cells with CRISPR ribonucleoproteins (RNP) targeting TRAC, TRBC, PDCD1, and lentiviral transduction of the NY-ESO-1 transgenic TCR specific for peptides derived from the cancer antigens, NY-ESO-1, and LAGE-1. The authors reasoned that endogenous TCR deletion paired with transgenic expression of NY-ESO-1 TCRs could lead to reduced mispairing of the endogenous and transgenic TCRα and TCRβ sequences and that deletion of PD-1 could lead to increased T cell function. In vitro analysis of edited cells revealed consistent gene disruption across the patient products, albeit at lower levels for TRBC and PDCD1, and robust tumor lysis activity against NY-ESO-1-expressing cells. CRISPR-edited NY-ESO-1⁺ T cells persisted and differentiated into memory T cells in patients, with an average decay half-life in the blood of 83.9 days, compared with roughly 1 week in previous studies using NY-ESO-1 TCR-engineered T cells. Off-target editing events and translocations were interrogated in the engineered cell product and upon reisolation from patients showing: (i) a low prevalence of off-target indels by GUIDE-seq, and (ii) detectable translocations that decreased over time below the limit of detection. The safety and feasibility of CRISPR-edited adoptive T cell therapy was further confirmed in a nonrandomized Phase I, 12-patient clinical trial targeting PD-1 in T cells from patients with metastatic non-small cell lung cancer (primary outcome: adverse events and/or dose-limiting toxicities) (NCT02793856)ⁱⁱ [62]. Here, polyclonal T cells were isolated, electroporated with protein Cas9 and sgRNA plasmids encoding two sgRNAs targeting the PDCD1 gene, and then infused back into patients. PD-1 editing was achieved in all 12 patients, with a median efficiency of 5.81% determined by gene sequencing. Off-target editing events were characterized via targeted and whole genome sequencing, with both methods showing a low, but reproducible level of off-target editing events [62].

Highlights.

Recent methods have enabled primary human and mouse T cell CRISPR-Cas9 screens, identifying novel regulators of T cell differentiation and function.

In vivo T cell screens revealed regulators of T cell differentiation in the context of infection and cancer, including FLI1-mediated restriction of effector CD8⁺ T cell differentiation.

Regulatory T cell (Treg) CRISPR-Cas9 screens unveiled new mechanisms of FOXP3 regulation, including deubiquitination of the Foxp3 gene locus by USP22 in mice.

CRISPR-Cas9 screens in cancer cells identified novel mechanisms of tumor cell-intrinsic immune evasion, such as negative regulation of IFNγ signaling by the SWI/SNF chromatin remodeling complex.

Preclinical studies have shown that CRISPR-Cas9 can be used to enhance CD8⁺ T cell function, leading to the first-in-human clinical trial using CRISPR-Cas9 edited T cells.

Glossary

ATAC-seq

assay for transposase-accessible chromatin using sequencing; assays genome-wide chromatin accessibility by transposition with Tn5 transposase.

Base editing (BE)

Cas9-mediated conversion of single base pairs within a specified window, mediated by fusion of Cas9 to a deaminase enzyme.

Cas9-expressing transgenic mouse

C57BL/6J mouse genetically engineered to constitutively express Cas9; Gt(ROSA) 26Sor^{tm1.1(CAG-cas9*,-EGFP)Fezh}/Gt(ROSA)^{26Sortm1.1(CAG-cas9*,-EGFP)Fezh}.

Catalytically deficient Cas9 (dCas9)

Cas9, engineered to have one or both of its nuclease domains; retains genome homing but loses DSB potential.

Checkpoint antibody immunotherapy

blocking antibodies used to inhibit the action of immune checkpoint receptors such as PD-1 or CTLA-4.

Chimeric antigen receptor (CAR) T cells

T cells that are genetically modified to express a chimeric antigen receptor, typically specific for a tumor-associated antigen and used for adoptive cell transfer immunotherapy.

Chromosome conformation assays

assays for 3D DNA interactions in the nucleus.

Conventional dendritic cells

immune cell type required for antigen presentation and T cell priming.

CRISP-seq/Perturb-seq

methods for CRISPR screening paired with a single-cell RNA sequencing readout.

CRISPRa/i

CRISPR activation or inhibition; leverages dCas9 fusions to activate or inhibit transcription of sgRNA target genes.

Crosspresentation

presentation of cell exogenous antigen on MHC class I by professional antigen presenting cells; MHC class I required for activation of antigen specific CD8⁺ T cells.

CUT&RUN

cleavage under targets and release using nuclease; chromatin profiling method in which antibody-targeted regions of DNA are cleaved by nucleases and sequenced.

Homology arms

homologous DNA strands that flank gene insertion cassette, used to activate cell-intrinsic homology directed repair pathways.

Nickase Cas9

Cas9 with a single nuclease domain inactivation; creates single-stranded nicks on target DNA.

Non-homologous end-joining (NHEJ)

DNA damage response pathway that prioritizes end-joining.

OT-1 CD8⁺ T cells (OT-1 T cells)

CD8⁺ T cells engineered to express the OT-1 TCR, specific for the SIINFEKL antigen on ovalbumin (Ova).

Protospacer adjacent motifs (PAM)

short DNA motif required for Cas9 binding to target site; different Cas9 orthologs have different PAM sequence requirements.

Single guide RNA (sgRNA)

RNA engineered to contain both the crRNA (targeting RNA) and trRNA (Cas9 binding RNA).