Towards a CRISPeR understanding of homologous recombination with high-throughput functional genomics

Abstract

CRISPR-dependent genome editing enables the study of genes and mutations on a large scale. Here we review CRISPR-based functional genomics technologies that generate gene knockouts and single nucleotide variants (SNVs) and discuss how their use has provided new important insights into the function of homologous recombination (HR) genes. In particular, we highlight discoveries from CRISPR screens that have contributed to define the response to PARP inhibition in cells deficient for the HR genes BRCA1 and BRCA2, uncover genes whose loss causes synthetic lethality in combination with BRCA1/2 deficiency, and characterize the function of BRCA1/2 SNVs of uncertain clinical significance. Further use of these approaches, combined with next-generation CRISPR-based technologies, will aid to dissect the genetic network of the HR pathway, define the impact of HR mutations on cancer etiology and treatment, and develop novel targeted therapies for HR-deficient tumors.

Introduction

Homologous recombination (HR) is a genome maintenance pathway required for the accurate repair of DNA double-strand breaks (DSBs). HR competes with non-homologous end joining (NHEJ) and microhomology-mediated end joining (MMEJ) for the repair of DSBs, and the choice between these pathways is dictated by the resection of DSB ends [1]. DSBs are protected from resection by the 53BP1-RIF1 complex, which promotes the direct ligation of DNA ends by NHEJ [2]. The end protection activity of 53BP1 is counteracted by the BRCA1-BARD1 complex, which stimulates end resection to promote HR [3–5]. Initial end resection by the MRE11-RAD50-NBS1 complex promotes the removal of DNA end-blocking lesions and generates short 3’-single-strand DNA (ssDNA) tails [6]. If microhomology-containing sequences are revealed by resection, the 3’ tails of the resected ends can undergo annealing, followed by trimming of non-homologous DNA flaps and end joining by MMEJ [7]. Alternatively, more extensive resection by the EXO1 and DNA2 nucleases generates long 3’-ssDNA tails onto which RAD51 nucleoprotein filaments are assembled by BRCA2, with the assistance of PALB2, BRCA1-BARD1 and the RAD51 paralogs [3, 8]. The RAD51 nucleoprotein filament then catalyzes DNA strand invasion and pairing with a homologous sequence to initiate DNA synthesis and copy the missing genetic information [9]. In addition to these activities, RAD51, BRCA1/2 and other HR factors protect nascent DNA from nucleolytic degradation in response to replication stress [8], suppress the formation of ssDNA gaps during DNA replication and promote their repair [10, 11], outlining the complexity of the functions exhibited by HR factors.

HR proteins act as tumor suppressors, as indicated by the observation that individuals carrying mutations in BRCA1, BRCA2 or other HR genes (e.g., PALB2, RAD51 paralogs) display a higher risk of developing breast, ovarian, prostate and pancreatic cancer [12, 13]. Furthermore, HR deficiency is often observed in the most aggressive subtypes of sporadic breast (i.e., triple-negative) and ovarian (i.e., high-grade serous) cancer [12]. Defining the impact of HR mutations on cancer etiology and uncovering vulnerabilities of HR-deficient tumors is therefore of utmost importance for developing targeted therapies against those tumors. In this review, we outline CRISPR-based functional genomics technologies that have enabled the discovery of novel vulnerabilities of HR-deficient tumors and the characterization of HR gene mutations.

CRISPR-based functional genomics technologies

CRISPR-mediated genome editing relies on the targeting of Cas effector proteins to genomic sequences of interest. The most commonly used Cas effector, Cas9, is an endonuclease that generates DSBs at its target locus. Cas9, like all Cas proteins, is targeted to genomic loci using guide RNAs (gRNAs) consisting of constant structural scaffolds and short programmable sequences known as spacers [14]. Paring of the 20 nucleotide gRNA spacer with its genomic complement results in Cas9-induced DSBs in the presence of specific protospacer-adjacent motifs (PAMs) flanking the target sequence [15, 16]. NHEJ- and MMEJ-mediated repair of Cas9-induced DSBs can lead to the formation of indels, which can disrupt genes by causing loss-of-function (LOF) mutations (Figure 1a) [17, 18]. Given its high efficiency, Cas9 has been utilized to conduct gene LOF studies on a high-throughput scale using screening technologies for functional genomics.

Figure 1. CRISPR-Cas effectors and applications for functional genomics screening.

(a) CRISPR-Cas enzymes and their respective strengths and weaknesses for functional genomics screening applications. (b) Examples of CRISPR-based functional genomics screening applications: 1) Genetic interaction profiling in isogenic cell backgrounds with distinct genetic perturbations or treatments. 2) Generation of epistasis maps from multi-dimensional screening approaches, such as combinatorial genetic interaction screens, multi-background screens (e.g., DepMap), multi-treatment screens and multi-variate phenotypic screens (e.g., Repair-seq, Perturb-seq, CROP-seq). Epistasis maps can be derived from direct gene-gene interaction data or the correlation of multiple phenotypic readouts resulting from distinct genetic perturbations. 3) Nucleotide variant classification by mutagenesis screens using CRISPR-dependent HDR, base editing or prime editing.

CRISPR screens allow for the simultaneous testing of thousands of genetic perturbations within a single experiment. At their simplest, CRISPR screens involve infection of a Cas9-expressing cell population with a pooled lentiviral gRNA library targeting a set of genes of interest. The starting and final gRNA library representations are determined through targeted sequencing of the genome-integrated gRNAs, enabling the quantification of cellular fitness for each perturbation induced by a unique gRNA [19]. Cellular phenotypes that can be studied using CRISPR screens include, among others, alteration of cell growth, expression of cell markers or fluorescence reporters, and changes in mRNA transcription. CRISPR screening can also be used to interrogate a wide range of cell perturbations beyond gene knockout using distinct CRISPR-Cas effectors. For example, transcriptional repressors or activators fused to catalytically dead Cas9 (dCas9) can induce CRISPR interference (CRISPRi) or activation (CRISPRa) of gene expression [20–23], while cytosine and adenine base editors (CBE and ABE) and prime editors (PEs) can generate site-specific base substitutions (Figure 1a) [24–26]. In addition, Cas12a enables simultaneous targeting of multiple genomic loci [27, 28], whereas Cas13 is capable of cleaving RNA molecules upon RNA targeting (Figure 1a) [29, 30]. Below we outline recent CRISPR screening approaches that have been employed to define genetic interactions and characterize the function of nucleotide variants, with a focus on the HR genes BRCA1 and BRCA2.

Exploring genetic interactions using CRISPR-based knockout screens

Genetic interactions define functional relationships between genes. One such relationship is epistasis, a phenomenon where a genetic mutation behaves differently depending on interactions with secondary mutations [31]. Epistatic interactions can have a positive or negative effect, depending on whether two mutations produce a better or worse than expected outcome, respectively [31]. Synthetic lethality describes a class of negative epistasis where two mutations are individually viable but when combined lead to cell death [32]. Synthetic lethality provides a unique strategy for targeted treatments of HR-deficient tumors, given that it enables the identification of genetic perturbations that impair the growth of HR-deficient cancer cells, while displaying neutral effects in HR-proficient non-cancer cells [33]. Interrogation of synthetic lethal genetic interactions can be conducted on a genome-wide scale using CRISPR-based knockout screens [34–37]. The simplest CRISPR screens for identifying synthetic lethal interactions utilize a common gRNA library for parallel screens in paired isogenic cell lines, one with and one without a LOF mutation in a gene of interest. Genes that cause lethality in the mutant, but not wildtype, background are deemed synthetic lethal candidates (Figure 1b). The need to generate isogenic cell lines limits the number of genetic interaction studies that can be conducted using this screening approach. This limitation has been circumvented by recent screening methodologies that remove the need for single cell cloning, enabling the testing of a larger number of gene pair interactions [38]. These approaches identify genetic dependencies that can be leveraged for the treatment of cancer cells harboring mutations in one of the identified synthetic lethal gene pairs.

To define dependencies across cancer types, whole-genome CRISPR-based knockout screens have been performed in hundreds of different cancer cell lines, as part of the Cancer Dependency Map (DepMap) initiative (https://depmap.org/portal/). DepMap is not particularly well suited to identify specific synthetic lethal interactions between gene LOF mutations given the non-isogenic nature of the cell lines utilized. However, by comparing the effects of genetic perturbations across hundreds of different cell backgrounds, DepMap enables the identification of genes whose loss causes highly correlated phenotypes, indicating that they are likely to participate in the same cellular processes [39, 40]. Using a related strategy that examined the gene-level correlation between CRISPR screens performed in the same cell line under a variety of genotoxic drug treatments, recent studies have generated a genetic map of the DNA damage response, including the HR pathway [41]. These large projects that carry out identical screens across many cell lines or perturbations provide useful resources for uncovering genes that might display positive epistatic interactions (Figure 1b).

Genetic interactions can be examined on a high-throughput scale using combinatorial CRISPR-based knockout screens [42–45]. Combinatorial screens use libraries of gRNA pairs that interrogate all possible single- and double-gene knockouts within a gene set of interest. This screening approach allows for the interrogation of a larger genetic interaction space compared to CRISPR screens conducted in isogenic knockout backgrounds. Furthermore, combinatorial screens can directly identify interacting genes (Figure 1b), thus providing an advantage over correlation-based datasets, like DepMap. Combinatorial knockout screening approaches utilize multi-guide constructs and Cas9 [42–45] or Cas12a [46–48] enzymes. Cas9-based combinatorial screens can be technically challenging, as dual gRNA constructs need multiple promoters and trans-activating CRISPR RNA (tracrRNA) sequences, making them susceptible to recombination during library preparation and sequencing [49]. For this reason, orthogonal Cas9 enzymes (e.g., Streptococcus pyogenes and Staphylococcus aureus Cas9) with distinct promoters and tracrRNAs have been utilized to limit recombination and increase the efficiency of combinatorial screens [43]. The use of Cas12a is also particularly useful for combinatorial screens, given that Cas12a can process multiple gRNAs from a single transcript (Figure 1a) [28, 46, 50], thus requiring a single promoter and simplified sequencing and cloning strategies. One common limitation of knockout-based interaction screens is the potential impact on cellular fitness of inducing multiple DSBs [51]. To circumvent this caveat, DSB-independent CRISPR technologies, such as CRISPRi, have been employed for generating genetic interaction maps in human cells [52]. CRISPRi also provides other advantages over knockout-based genetic interaction screens by enabling the study of genetic interactions exhibited by essential genes and avoiding heterogeneous repair outcomes that can confound downstream analysis (Figure 1a). Together, these approaches hold great promise for dissecting the genetic network of the HR pathway and developing novel targeted therapies for HR-deficient tumors.

Functional characterization of nucleotide variants using large-scale precision genome editing screens

Functional characterization of single nucleotide variants (SNVs) in HR genes is integral to understanding and treating HR-deficient tumors. Advances in genome sequencing have resulted in a large increase in the number of observed SNVs in HR genes, the majority of which remain of unknown relevance. In the case of BRCA1 and BRCA2, >18,000 SNVs are listed in the ClinVar database of clinically relevant variants, >8,000 of which are classified as variants of uncertain significance (VUSs). Given the vast number of uncharacterized variants, cellular assays have been developed to examine the phenotype of gene variants at scale. These assays can be conducted by ectopically expressing individual cDNAs carrying a defined set of mutations [53, 54] or by performing deep mutational scanning (DMS) screens, which enable the interrogation of a large number of gene variants in a pooled format [55]. Classically, DMS experiments utilize viral- or plasmid-based libraries of mutant open reading frames, thus not allowing the assessment of gene variant functions within their endogenous genomic context [56]. More recently, the development of CRISPR-based precision genome editing has enabled high-throughput functional studies of nucleotide variants introduced at endogenous loci.

Precision genome editing technologies include CRISPR-dependent homology-directed repair (HDR), base editing and prime editing. CRISPR-dependent HDR relies on the delivery of ssDNA or dsDNA donor molecules containing a sequence of interest flanked by regions of homology to the Cas9 cut site (Figure 1a) [57]. After Cas9 cutting, these DNA donors are used by the HR machinery as templates for DSB repair, resulting in the insertion of the sequence of interest at the Cas9-induced DSB [57]. This approach can generate any nucleotide variant of interest, but it is a low efficiency editing strategy due to competition between HR and NHEJ, and can cause cell cycle arrest or cell death because of DSB formation [51, 58, 59]. Alternatively, nucleotide substitutions can be generated by CRISPR-dependent base editing in a manner independent from DSBs and DNA donors, resulting in reduced toxicity and greater efficiency (Figure 1a). Base editing relies on a nickase mutant Cas9 (nCas9) tethered to adenine or cytosine deaminases that generate A>G or C>T transitions, respectively, within a window (~6 bases) of base editing activity [24, 25]. In addition to adenine and cytosine base editors (ABE and CBE), base editors that generate C>G transversions (CGBE) have recently been developed [60–62]. Unlike base editors, prime editors are capable of generating all 12 base-to-base conversions, along with templated insertions and deletions (Figure 1a) [26]. This technology employs a prime editing guide RNA (pegRNA) for nCas9 localization and reverse transcriptase-mediated templated insertion of a sequence of interest from the extended pegRNA. The efficiency of prime editing is highly variable between cell types [26, 63], and is stimulated by nicking the non-edited strand using a second gRNA [26].

Recent studies have shown that the above precision genome editing technologies can be employed for functional screening of nucleotide variants (Figure 1b). HDR screens are highly accurate in their ability to insert variants, but they have only been carried out in NHEJ-deficient (LIG4 knockout) haploid cells, given the low editing efficiency of HDR, and are restricted in the number of sites that can be edited, since separate DNA donor libraries of variants need to be utilized for each Cas9-induced DSB [64]. For these reasons, HDR screens have been performed on a single-gene basis. Distinct from HDR screens, base editing screens are compatible with the study of variants in any number of genes simultaneously and have been successfully conducted in non-haploid cell lines [65, 66]. Both HDR and base editing screens have been reported to enable the functional classification of VUSs and the identification of LOF variants [64–66]. Base editing screens have also been shown to allow high-throughput gene disruption studies through the generation of nonsense and splice variants and enable multi-gene analyses of gain- and separation-of-function variants, novel protein domains, interaction surfaces and small molecule binding sites [65–69]. It is important to note that current screening technologies that utilize base editors with strict PAM requirements (e.g., NGG) may display a significant number of false negatives [65, 66], given that, in the vast majority of cases, nucleotide variants rely on the efficiency of a single gRNA for being generated. Furthermore, when multiple editable bases are present within the activity window, base editors can lead to multiple editing outcomes [65, 66]. These limitations can be circumvented by using base editors with more relaxed PAM specificities (e.g., NG, SpG, SpRY) [70–72] and more restricted activity window (<6 bp) [73], thus increasing the number of overlapping gRNAs that could generate the same specific mutation, while reducing the number of outcomes per gRNA. Prime editing screens are less constrained by PAM sequences than base editing screens, given the broader editing window of pegRNAs. Similar to HDR screens, prime editing screens are highly accurate, but have only been applied for the editing of a limited number of haploid loci [74]. Despite their current limitations, these technologies provide unprecedented opportunities to dissect the multiple functions of HR factors at the nucleotide level and determine the clinical relevance of HR gene mutations for cancer etiology and therapy. Below we highlight discoveries on the functions of BRCA1/2 obtained using CRISPR screens, which illustrate the potential of these technologies.

Elucidating the functions of BRCA1 and BRCA2 using CRISPR screens

CRISPR screens have been instrumental in defining genetic interactions exhibited by BRCA1 and BRCA2 and uncovering their relevance for the treatment of BRCA1/2-mutant tumors with poly (ADP-ribose) polymerase (PARP) inhibitors. PARP inhibition (PARPi) induces trapping of PARP1/2 at DNA single-strand breaks (SSBs), which during DNA replication can cause the collapse of replication forks in BRCA1/2-deficient cells [75]. PARPi can also lead to the accumulation of toxic ssDNA gaps in BRCA1/2-deficient cancer cells [76]. While PARPi is initially efficacious, resistance to PARPi arises over time in BRCA1/2-mutant tumors [77].

CRISPR-based knockout screens conducted in BRCA1/2-deficient cells have provided new insights into the mechanisms that promote resistance to PARPi and identified novel factors whose loss enhances PARPi sensitivity [78–85]. In addition, CRISPR-based knockout screens have led to the identification of novel BRCA1/2 synthetic lethal partners that could be targeted for developing new therapies for BRCA1/2-mutant tumors [86–88]. More recently, precision genome editing screens have been employed to identify BRCA1/2 VUSs of likely pathogenicity and determine their phenotype in response to PARPi and other DNA damaging treatments [64–66, 74, 89], as discussed below.

Defining the mechanisms that promote resistance to PARP inhibition in BRCA1/2-deficient cells

CRISPR-based knockout screens have contributed to defining the mechanisms that regulate the response to PARPi in BRCA1-deficient cells. In particular, these screens have revealed that loss of members of the shieldin complex results in PARPi resistance in BRCA1-deficient cells [84, 85]. These studies, together with additional reports, showed that the shieldin complex acts as a 53BP1-RIF1 mediator that suppresses HR-dependent DSB repair (Figure 2) [84, 85, 90–95]. Shieldin has been proposed to suppress HR by multiple mechanisms, including the inhibition of DSB resection by the EXO1 and DNA2 nucleases, the refilling of resected ends mediated by the DNA polymerase Polα in complex with CTC1-STN1-TEN1 (CST), and the inhibition of BRCA2/PALB2-dependent loading of RAD51 on resected ends (Figure 2) [84, 85, 90–96]. Notably, loss of the CST complex was shown to promote PARPi resistance in CRISPR screens conducted in BRCA1-deficient cells [82]. CRISPR screens also uncovered that loss of DYNLL1, a MRE11-interactor that promotes efficient 53BP1 oligomerization at DSBs and suppresses MRE11-dependent end resection [83, 97], or disruption of its transcriptional regulator ATMIN causes PARPi resistance in BRCA1-mutant cells (Figure 2) [83]. Together, these studies have defined the mechanisms by which DSB end protection controls the response to PARPi in BRCA1-deficient cells.

Figure 2. Functions of the shieldin-CST-Polα complex, DYNLL1 and FEN1 in the context of DNA double-strand break repair.

Following DSB formation, DSBs can be directly religated with minimal end processing by NHEJ (1). Alternatively, during the S and G2 phases of the cell cycle, DSBs can undergo end resection. DSB resection is initiated by the MRE11-RAD50-NBS1 (MRN) complex together with CtIP and is promoted by the BRCA1-BARD1 complex (2). DYNLL1 suppresses end resection by inhibiting MRE11 and facilitating the assembly of 53BP1 at DSBs, which competes with BRCA1-BARD1 for binding to histone H2A ubiquitinated on lysine 15 at DSBs (2). Following short-range end resection, 53BP1 recruits RIF1, which localizes the shieldin complex to DSBs (3). The shieldin complex binds the resected ends and protects them from further resection by the DNA2 and EXO1 nucleases, while also inhibiting the assembly of HR factors (4). The shieldin complex also recruits Polα through the binding of the CST complex, leading to the filling of the resected ends, which can then be religated by NHEJ (5). If microhomology-containing regions are exposed by short-range end resection, they can be repaired by MMEJ through Polθ-dependent annealing and filling of the resected ends (6). Polθ-mediated DNA synthesis can lead to 5’ flaps that can be removed by FEN1 (6). Alternatively, long-range end resection by EXO1 and DNA2 generates long 3’-ssDNA tails, which are protected by RPA (7). BRCA2 in complex with PALB2, BRCA1 and BARD1 then catalyzes the assembly of RAD51 nucleoprotein filaments on the 3’-ssDNA ends, which then initiate strand invasion and homology search to copy the missing genetic information and repair DSBs by HR (8).

Distinct from the above findings, disruption of the ubiquitin ligase HUWE1 and the acetyl-transferase KAT5 was shown through CRISPR screening to cause PARPi resistance in BRCA2-deficient cells [81]. Loss of HUWE1 was proposed to cause PARPi resistance by partially restoring HR in BRCA2-deficient cells, while KAT5 deficiency was suggested to increase the binding of 53BP1 to DSBs [81]. Further work will be required to investigate whether elevated 53BP1 activity might promote PARPi resistance in BRCA2-deficient cells. These studies have confirmed that resistance to PARPi arises through different mechanisms in BRCA2- vs BRCA1-deficient cells.

Defining endogenous DNA lesions that enhance PARP inhibition sensitivity

Multiple CRISPR-based knockout screens interrogating the mechanisms that drive PARPi sensitivity have identified aberrant genomic nucleotides as endogenous sources of DNA damage that cause cell lethality upon PARPi [78, 80, 98]. Genome-wide CRISPR screens in a BRCA1-mutant and two BRCA1-wildtype cell lines discovered that loss of RNase H2, an endonuclease that removes genome-embedded ribonucleotides, causes sensitivity to PARPi [80]. RNase H2 loss was reported to trigger ribonucleotide incision by the topoisomerase TOP1 and subsequent formation of SSBs with 3’-ends blocked by a 2’−3’-cyclic phosphate, which can engage PARP and lead to PARP trapping and cell death following PARPi (Figure 3) [80]. These studies have uncovered a critical role for ribonucleotide processing in the response to PARPi.

Figure 3. Roles of the RNase H2 complex, APE2, DNPH1 and ALC1 in the repair of DNA base damage.

Misincorporated genomic ribonucleotides undergo processing by the RNase H2 complex, followed by restoration of the original DNA sequence through the action of other enzymes involved in ribonucleotide excision repair (RER) (1–3). Alternatively, genome-embedded ribonucleotides can be acted upon by TOP1, which forms SSBs with 3’ ends blocked by 2’−3’-cyclic phosphate (2’−3’-P) adducts that prevent DNA synthesis and ligation (4). APE2 can process 2’−3’-P blocked ends, generating repairable SSBs (5). If not processed by APE2, persistent SSBs with 3’-blocked ends can lead to replication fork collapse, which is exacerbated by PARP trapping in the presence of PARP inhibitors (6–7). Collapsed forks are then repaired by HR (8). Misincorporation of modified deoxynucleotides and DNA base alkylation also induce DNA repair events. DNPH1 limits the misincorporation of hmdU into the genome by hydrolyzing its precursor hmdUMP (9–10). If misincorporated, hmdU is processed by SMUG1, resulting in the generation of apurinic/apyrimidinic (AP) sites (11–12). AP sites are also generated by MPG following the excision of alkylated bases (13). Nucleosome occupancy at the AP sites prevents access of downstream BER factors (14). PARP-mediated PARylation events stimulate ALC1-dependent nucleosome remodeling, allowing BER factors, such as APE1, access to the AP sites (14–15). After chromatin remodeling, APE1 can process the AP sites, generating SSBs (16). PARP trapping at SSBs in the presence of PARP inhibitors can lead to replication fork collapse (17). If base damage is not repaired by ALC1, unrepaired AP sites can impair DNA replication and lead to the accumulation of ssDNA gaps, which can trap PARP in the presence of PARP inhibitors (18). Repair of ssDNA gaps is promoted by HR (19).

In addition to the processing of genomic ribonucleotides, processing of modified deoxynucleotides has been found to enhance PARPi sensitivity. In particular, CRISPR screens conducted following PARPi in cells knocked out for MUS81, an endonuclease involved in Holiday junction resolution during HR, uncovered that the loss of DNPH1, a sanitizer of the cellular nucleotide pool, enhances PARPi sensitivity in HR-deficient cells, such as BRCA1/2-deficient cells [78]. DNPH1 loss was shown to cause the genomic misincorporation of 5-hydroxymethyl-deoxyuridine (hmdU), thus triggering hmdU processing by the DNA glycosylase SMUG1 and the consequent formation of base excision repair (BER) intermediates leading to PARP trapping and replication fork collapse in HR-deficient cells subjected to PARPi (Figure 3) [78]. Release of trapped PARP has been reported to be facilitated by ALC1, a poly(ADP-ribose)-dependent chromatin remodeler that enhances chromatin accessibility to promote the repair of DNA base damage in HR-deficient cells (Figure 3) [79, 98–100]. ALC1 loss was shown by CRISPR screening to be a key determinant of PARPi sensitivity in BRCA1/2-mutant cells [79]. ALC1 has been proposed to operate downstream of the DNA glycosylases SMUG1 and MPG for the repair of misincorporated uracil and alkylated bases, respectively (Figure 3) [98]. Inability to repair base lesions in the absence of ALC1 leads to the accumulation of replication-associated ssDNA gaps and DSBs that cause toxicity in HR-deficient cells (Figure 3) [79, 98]. Altogether, these studies have revealed a novel interplay between chromatin remodeling, BER and HR in the response to PARPi.

Identifying vulnerabilities of BRCA1/2-deficient cells

CRISPR-based knockout screens have also been performed to identify synthetic lethal interactions in unperturbed BRCA1/2-deficient cells. Screens conducted in BRCA2-deficient cancer cell lines revealed that loss of the 5’-flap endonuclease FEN1 caused increased lethality in BRCA2-deficient relative to BRCA2-proficient cells [88]. Similar findings were also obtained in BRCA1-deficient cells and confirmed using FEN1 inhibitors [88, 101]. The FEN1-BRCA1/2 synthetic lethal interaction was ascribed to FEN1’s function in removing 5’-flaps that could originate following Polθ-mediated DNA synthesis during MMEJ (Figure 2) [88]. Similar to FEN1 deficiency, Polθ loss or inhibition causes synthetic lethality in combination with BRCA1/2 deficiency [102–106]. The ability of FEN1 to remove 5’-flaps during Okazaki fragment maturation has also been suggested to contribute to maintain the viability of BRCA1/2-deficient cells [76, 101], given that defective Okazaki fragment ligation could result in the accumulation of ssDNA gaps and DSBs, which would require HR for repair.

The above CRISPR screens also identified a synthetic lethal interaction between APEX2 and BRCA1/2 [88]. APEX2 encodes APE2, a paralog of the apurinic/apyrimidinic (AP) endonuclease APE1. Unlike APE1, APE2 has poor AP endonuclease activity, while exhibiting strong 3’−5’ exonuclease and 3’-phosphodiesterase activities [107]. Recent CRISPR screens performed in APEX2 knockout cells identified genes encoding the subunits of the RNase H2 complex as synthetic lethal pairs with APEX2 [87]. These genetic screens, together with additional biochemical studies, suggested that APE2 is required for the removal of 3’-blocked DNA ends generated by TOP1-mediated processing of genomic ribonucleotides, and that loss of this activity causes DSBs or other DNA lesions during DNA replication that require HR for repair (Figure 3) [87].

Finally, a set of CRISPR screens was conducted in isogenic pairs of BRCA1/2-mutated and BRCA1/2-wildtype cell lines to identify common BRCA1/2 synthetic lethal interactions. These screens identified CIP2A as a novel synthetic lethal partner for both BRCA1 and BRCA2 [86]. CIP2A data analysis on DepMap hinted at a positive epistatic interaction between CIP2A and TOPBP1, which was confirmed by showing that CIP2A physically associates with TOPBP1 and together form filamentous structures on mitotic chromosomes in response to replication stress. These findings raise the possibility that the CIP2A-TOPBP1 complex may operate by holding together broken chromosomes resulting from the processing of incompletely replicated DNA [86], an activity that would be particularly important for the viability of HR-deficient cells, given their accumulation of mitotic aberrations resulting from under-replicated DNA [108]. Altogether, the above studies discovered novel vulnerabilities of BRCA1/2-deficient cells, which could be exploited to develop novel targeted therapies for BRCA1/2-mutant tumors.

Identifying BRCA1/2 SNVs of likely pathogenicity

CRISPR-dependent precision genome editing technologies have recently been employed to characterize the large number of BRCA1/2 VUSs. In an effort to define functional variants that predispose to breast and ovarian cancer, HDR-dependent CRISPR screens have been conducted in the haploid HAP1 cells to survey BRCA1 SNVs that cause BRCA1 disruption and consequent loss of viability [64]. This functional screening approach utilized dsDNA donor plasmids to introduce 3,893 SNVs at Cas9-induced DSBs generated within the 13 exons encoding the BRCA1 RING and BRCT domains. This screen, which accounted for 96.5% of all possible SNVs across these domains, including ~400 VUSs listed in the ClinVar database, led to the identification of ~400 non-functional BRCA1 SNVs that impaired HAP1 cell viability.

More recently, CRISPR-dependent base editing screens have been employed to study the functional impact of BRCA1/2 nucleotide variants [65, 66, 89]. These screens utilized the cytosine base editor BE3 in combination with >475 and >550 gRNAs to generate C>T transitions across the gene body of BRCA1 and BRCA2, and measured the effects induced by the generated variants on cell viability in both haploid and non-haploid cell lines with and without treatment with PARP inhibitors, topoisomerase inhibitors or cisplatin [65, 66, 89, 109]. Base editing screens enabled the identification of non-functional and likely pathogenic BRCA1/2 missense mutations, which largely localized to known functional domains, such as the RING and BRCT motifs of BRCA1 and the DNA binding domain of BRCA2 [65, 66]. Recent studies with prime editing screens have investigated the impact of 465 BRCA2 variants generated in 10 BRCA2 loci on the growth of cells engineered to be haploid for BRCA2 [74]. In line with the above base editing screens, this work uncovered non-functional BRCA2 missense mutations primarily localizing within the BRCA2 DNA binding domain [74]. Further work will be warranted to characterize the non-functional mutants identified in these studies and determine which activities (e.g., DSB repair vs fork protection) and/or physical interactions of BRCA1/2 they might disrupt. Together, the above studies have provided novel approaches to investigate the function of BRCA1/2 SNVs at scale.

Conclusions and future directions

The CRISPR screens that have been performed to study BRCA1/2 and other HR factors have utilized cell proliferation in response to DNA damage as their main readout. Future CRISPR screens that couple gene knockout, CRISPRi/a and/or base editing with fluorescence-activated cell sorting or fluorescence microscopy plus in situ gRNA sequencing [110] will enable the use of DNA damage markers as readouts to uncover novel genetic interactions exhibited by HR factors and further define the phenotype of HR mutations. The functions of HR genes and their mutations could also be further elucidated using CRISPR screening technologies coupled to profiling of DSB repair events (e.g., Repair-seq) [111]. In addition, the use of single-cell RNA-seq (e.g., Perturb-seq and CROP-seq) [112–115] and single-cell ATAC-seq (e.g., Spear-ATAC, Perturb-ATAC) [116, 117] technologies combined with CRISPR screening will allow the investigation of the impact of HR gene mutations and interactions on gene expression and chromatin accessibility. Finally, the development of combinatorial CRISPR screens using Cas9 or Cas12a will generate a more refined map of the epistatic interactions exhibited by HR genes and their mutations. Taken together, these and other new genome editing technologies that will be developed in the coming years, will provide unparalleled insights into the interplay between HR and other cellular pathways and uncover the functional impact of HR mutations on cancer etiology and therapy, thus leading to the development of novel targeted treatments for HR-deficient tumors.