Functional interrogation of DNA damage response variants with base editing screens

Abstract

Mutations in DNA damage response (DDR) genes endanger genome integrity and predispose to cancer and genetic disorders. Here, using CRISPR-dependent cytosine base editing screens, we identify > 2,000 sgRNAs that generate nucleotide variants in 86 DDR genes, resulting in altered cellular fitness upon DNA damage. Among those variants, we discover loss- and gain-of-function mutants in the Tudor domain of the DDR regulator 53BP1 that define a non-canonical surface required for binding the deubiquitinase USP28. Moreover, we characterize variants of the TRAIP ubiquitin ligase that define a domain, whose loss renders cells resistant to topoisomerase I inhibition. Finally, we identify mutations in the ATM kinase with opposing genome stability phenotypes and loss-of-function mutations in the CHK2 kinase previously categorized as variants of uncertain significance for breast cancer. We anticipate that this resource will enable the discovery of additional DDR gene functions and expedite studies of DDR variants in human disease.

Keywords: 53BP1; ATM; BE3 base editor; CHK2; CRISPR-dependent base editing; DNA damage response; TRAIP; clinically relevant variants; variants of uncertain significance.

Figure 1.. Performance of CRISPR-dependent base editing screens targeting DDR genes

A) 27-gene DDR network targeted by the sgRNA sublibrary 1, organized by pathways, and their genetic and physical interactions (grey lines), according to the STRING database. B) Schematic of the protocol utilized for base editing screens in untreated conditions. MCF10A, MCF7 and HAP1 cells expressing BE3 were transduced with a lentiviral sgRNA library targeting the gene network in (A), and then cultured for 18 days after selection. The LFC in sgRNA abundance between day 0 (T0) and day 18 (T18) was then determined following next-generation sequencing (NGS). C) Density plots of LFC values for sgRNAs predicted to generate the indicated mutational outcomes in essential and non-essential genes in MCF10A-BE3 cells (Table S1). LFC density plots are also shown for iSTOP and negative control sgRNAs. Dotted line: LFC for the bottom 5% of negative controls (LFC= −0.46581). D) ROC analyses of MAGeCK ranks for sgRNAs predicted to introduce splice and nonsense mutations (true positives) vs silent and empty-window sgRNAs (false positives) in essential genes. AUC values are indicated in brackets. E) Graphical representation of LFC values and Rule Set 2 on-target scores for iSTOP controls in MCF10A-BE3 cells. F) ROC analyses conducted as in (D) exclusively for sgRNAs with Rule Set 2 on-target scores >0.5. AUC values are indicated in brackets. G) Density plots of LFC values for sgRNAs introducing nonsense and splice variants in essential genes with Rule Set 2 on-target score >0.5 in MCF10A-BE3 cells. Density plots from (C) are depicted in lighter colors. Dotted line: LFC= −0.46581. See also Figure S1 and Tables S1–4.

Figure 2.. Analyses of drug-sgRNA interactions upon treatment with DNA damaging agents

A) 86-gene DDR network targeted by sgRNA sublibraries 1 and 2, represented as in Figure 1A. B) Schematic of the protocol utilized for base editing screens including treatment with genotoxic agents. MCF10A-BE3 cells were transduced with the lentiviral sgRNA sublibraries 1 and 2 targeting the gene network in (A) and cultured in the presence of the indicated genotoxic agents, or left untreated. LFC values in sgRNA abundance between T0 and T18 were then calculated for each individual condition following NGS. C) Heatmap of LFC values for biologically and statistically relevant sgRNAs (LFC beyond top or bottom 1% of negative controls and p-value <0.01, Table S6) hierarchically clustered by Ward’s method. E: enrichment, D: depletion. D) Analyses of the enrichment of sgRNA-targeted genes within the groups defined by the cluster analyses in (C). Dotted line: Fisher’s t-test p-value <0.01. See also Tables S5–6.

Figure 3.. Characterization of LOF and GOF mutations in 53BP1’s tandem Tudor domain that define a previously uncharacterized surface required for USP28 interaction

A) Lollipop plot of 53BP1 sgRNAs and their LFC values in the presence of doxorubicin mapped to the canonical 53BP1 protein isoform. Lollipop transparency reflects sgRNA biological relevance, with non-transparent lollipops indicating sgRNAs with a LFC value beyond the threshold of 1% of negative controls. Symbol size reflects sgRNA statistical significance, with larger symbols indicating sgRNAs with a p-value <0.01. Residues of interest are highlighted next to their corresponding lollipop(s). B) Competitive growth assay in the presence of doxorubicin (2.5 nM) conducted on a BE3-expressing MCF10A cell population expressing the indicated sgRNAs. Data represent the sgRNA_of_interest_GFP/AAVS1_sgRNA_mCherry ratio normalized to the day 1 (T1) time point and to the corresponding GFP/mCherry AAVS1 sgRNA ratio at each experimental time point. Mean ± SD for n=2. C) Time course analysis of editing frequency induced by the indicated 53BP1 sgRNAs at their target loci, as determined by Sanger sequencing and ICE analysis (Hsiau et al., 2019). Mean ± SD for n=2. D) Cartoon of the crystal structure of 53BP1’s TTD bound to a H4K20me2 peptide (orange) (PDB ID: 2IG0, (Botuyan et al., 2006)), colored based on amino acid conservation (V: variable, C: conserved). Representations of overlapping WT (grey) and mutant structures for V1544I and G1560K predicted by Missense3D (red) are shown and AAs structurally affected are highlighted, along with the number of sgRNAs targeting each residue (in brackets). E) Competitive growth assays in the presence of N3 (1 μM) conducted on MCF7-BE3 cells expressing the indicated 53BP1 sgRNAs. GFP/mCherry sgRNA ratios for N3-treated conditions were normalized to the corresponding untreated controls. Mean ± SD for n=3. F) Immunoblot showing USP28 and p53 co-immunoprecipitated by HA-GFP, HA-53BP1 WT and mutants from HEK293T cell lysates using an anti-HA antibody. Images are representative of two independent experiments. G) Dot plot of the number of IR-induced 53BP1 foci per cell in MCF10A-BE3 cells AAVS1-targeted or carrying the indicated 53BP1 mutations (Table S7). Cells were fixed 4 h post-IR treatment (5 Gy) prior to imaging by high-content microscopy. Data belong to two independent experiments, and mean values for each condition are indicated. Statistical analysis was conducted using one-way ANOVA (**** p-value <0.0001). H) Representative images of IR-induced RAD51 foci in control and 53BP1 mutant cells upon treatment with control or BRCA1 siRNA (Table S7). Cells were treated and imaged as in (G). Scale bar: 10 μm. I) Quantification of IR-induced RAD51 foci per cell from (H). Mean ± SD for n=3. Statistical analysis was conducted using unpaired t-test (**** p-value <0.0001). See also Figure S2 and Tables S5–7.

Figure 4.. Identification of variants of the ATM protein with opposing phenotypes

A) Heatmap of LFC values for relevant ATM sgRNAs in MCF10A-BE3 cells hierarchically clustered by the Ward’s method. sgRNA categories are color-coded, and empty-window/non-coding sgRNAs are in white. sgRNAs selected for validation are highlighted. B) Competitive growth assays in untreated conditions or upon doxorubicin (2.5 nM) treatment conducted on MCF10A-BE3 cells expressing the indicated ATM sgRNAs. Data are represented as in Figure 3B. Mean ± SD for n=4 (UNT) and n=3 (DOX). C) Analysis of editing frequency induced by the indicated ATM sgRNAs at their target loci at day 4 post-selection, as determined by Sanger sequencing and ICE analyses. Mean ± SD for n=2. D) Immunoblot showing ATM expression in MCF10A-BE3 cells, either AAVS1-targeted or edited with the indicated ATM sgRNAs. E) Cartoon of the crystal structure of the ATM dimer (PDB ID: 5NP0, (Baretic et al., 2017)) with domains colored as in Figure S3A. ATM mutations verified by Sanger sequencing are highlighted in green (growth advantage) or red (growth disadvantage) in the protein structure. F) Dot plot of the number of doxorubicin-induced γH2AX foci per cell in MCF10A-BE3 cells AAVS1-targeted or carrying the indicated ATM mutations (Table S7). Cells were fixed 8, 24, and 48 h after doxorubicin (20 nM) addition prior to imaging by high-content microscopy. Data belong to two independent experiments and mean values for each condition are indicated. Statistical analysis for each condition relative to its respective AAVS1-targeted time point was conducted using one-way ANOVA (**** p-value <0.0001). G) Analyses of nuclear aberrations (micronuclei, nuclear fragments) in MCF10A cells carrying the indicated ATM mutations upon doxorubicin treatment (20 nM) for 48 h. Mean ± SD for n=3. Statistical analysis for each sample relative to the AAVS1-targeted control was conducted using unpaired t-test (* p-value <0.05, ** p-value <0.01, *** p-value <0.001, **** p-value <0.0001). See also Figure S3 and Tables S5–7.

Figure 5.. Discovery of a functional region in the TRAIP ubiquitin ligase that promotes sensitivity to camptothecin

A) Heatmap of LFC values for relevant TRAIP sgRNAs hierarchically clustered by the Ward’s method, as shown in Figure 4A for ATM sgRNAs. B) Lollipop plot of TRAIP sgRNAs and their LFC values for camptothecin-treated conditions mapped to the canonical TRAIP protein isoform, as shown in Figure 3A for 53BP1. sgRNAs targeting residues selected for validation are highlighted. C) Competitive growth assays upon cisplatin (1 μM) or camptothecin (5 nM) treatment in BE3-expressing MCF10A cell populations carrying the indicated TRAIP sgRNAs. Data are represented as in Figure 3B. Mean ± SD for n=3 (CISP) and n=4 (CPT). D) Time course analysis of editing frequency induced by the indicated TRAIP sgRNAs at their target sites, as determined by Sanger sequencing and ICE analyses. Mean ± SD for n=2. E) RT-PCR analyses of TRAIP transcript levels in MCF10A cells carrying the Q174* and R185* mutations. The identity of the indicated TRAIP splice variants was verified by Sanger sequencing. fl: full-length transcript; Δ^ex7: delta exon 7 transcript; Δ^ex6−8: delta exons 6–8 transcript. F) Immunoblot showing TRAIP expression in MCF10A-BE3 cells targeted with the indicated sgRNAs and subjected to camptothecin (20 nM, 48 h) treatment or left untreated. fl: full-length isoform; s: short isoform. G) Analyses of micronuclei in MCF10A-BE3 cells targeted with the indicated sgRNAs upon camptothecin treatment (10 nM, 48 h). Values are represented as fold change relative to the AAVS1-targeted control. Mean ± SD for n=6. Statistical analysis for each sample relative to the AAVS1-targeted control was performed using unpaired t-test (* p-value <0.05, ** p-value <0.01, *** p-value <0.001, **** p-value <0.0001). H) Analyses of 53BP1 NBs in MCF10A-BE3 cells AAVS1-targeted or carrying the indicated TRAIP mutations, subjected to camptothecin (10 nM, 48 h) or cisplatin (2 μM, 48 h) treatment. The fold change in the number of cells with >2 53BP1 NBs in each condition relative to the AAVS1-targeted control is shown. Mean ± SD for n=3 (CISP) or n=4 (CPT). Statistical analysis was performed as in (G). I) Neutral comet assay on MCF10A-BE3 cells AAVS1-targeted or carrying the indicated TRAIP mutations, subjected to mock or camptothecin (10 nM, 48 h) treatment. Mean tail moment is indicated and data represent three independent replicates. Statistical analysis was conducted by one-way ANOVA (* p-value <0.05, ** p-value <0.01, *** p-value <0.001, **** p-value <0.0001). See also Figure S4 and Tables S5–7.

Figure 6.. Analysis of variants of uncertain significance with pathogenic-like behavior

A) ROC analyses of MAGeCK ranks for sgRNAs predicted to introduce pathogenic and likely-pathogenic mutations (true positives) vs benign and likely-benign sgRNAs (false positives) in essential genes from sublibrary 1. AUC values are indicated in brackets. B) Distribution of low-priority (LFC beyond top or bottom 5% of negative controls and p-value <0.05 in any treatment) or high-priority (LFC beyond top or bottom 1% of negative controls and p-value <0.01 in any treatment, Table S6) sgRNAs predicted to introduce clinically-relevant mutations in MCF10A-BE3 and MCF7-BE3 cells. Statistical analysis relative to the full set of relevant sgRNAs in MCF10A-BE3 or MCF7-BE3 cells was conducted using a chi-squared test (** p-value <0.01, **** p-value <0.0001). C) Venn diagrams of high-priority sgRNAs generating clinically-relevant mutations in MCF10A-BE3 and MCF7-BE3 screens (Table S6). LFC values for each individual sgRNA across the 5 conditions (UNT, CISP, OLAP, DOX, CPT) were averaged and sgRNAs with average LFC>0 were assigned to the enrichment set, while sgRNAs with average LFC<0 were assigned to the depletion set. D) Correlation of LFC values for clinically-relevant sgRNAs targeting the CHK2 gene upon doxorubicin treatment in MCF10A-BE3 vs MCF7-BE3 cells. Pearson correlation value is shown. sgRNAs targeting residues selected for validation are highlighted. E) Competitive growth assays in the presence of doxorubicin (2.5 nM) conducted on MCF10A-BE3 cells expressing the indicated sgRNAs. Data are represented as in Figure 3B. Mean ± SD for n=4. F) Time course analysis of editing frequency induced by the indicated CHK2 sgRNAs at their target loci, as determined by Sanger sequencing and ICE analysis. Mean ± SD for n=2. G) Cartoon of the crystal structure of the CHK2 dimer (PDB ID: 3I6W, (Cai et al., 2009)) with domains colored as in Figure S5B. Mutations verified by Sanger sequencing are highlighted in green. H) Immunoblot showing CHK2 expression and its phosphorylation on T68 in MCF10A-BE3 cells, either AAVS1-targeted or edited by CHK2 sgRNAs, with or without doxorubicin treatment (100 nM, 8 h). I) Correlation of averaged LFC values (CISP, OLAP) for clinically-relevant sgRNAs targeting BARD1, BRCA1 and BRCA2 in MCF10A-BE3 vs MCF7-BE3 cells. Pearson correlation value is shown. sgRNAs that uniquely generate a VUS in the BARD1 gene are indicated. J) Lollipop plot of BARD1 sgRNAs and their average LFC values (CISP, OLAP) in MCF7-BE3 cells mapped to the canonical BARD1 protein isoform. sgRNA categories are distinguished by symbol shape, and larger symbols indicate sgRNAs with a p-value <0.01 in olaparib and/or cisplatin treatments. Colors indicate clinical relevance. sgRNAs that uniquely generate a VUS with pathogenic-like behavior in the BARD1 gene are indicated. See also Figures S5–6 and Tables S5–7.