CRISPR/Cas9 screen for genome-wide interrogation of essential MYC-bound E-boxes in cancer cells

Abstract

The transcription factor MYC is a proto-oncogene with a well-documented essential role in the pathogenesis and maintenance of several types of cancer. MYC binds to specific E-box sequences in the genome to regulate gene expression in a cell-type- and developmental-stage-specific manner. To date, a combined analysis of essential MYC-bound E-boxes and their downstream target genes important for growth of different types of cancer is missing. In this study, we designed a CRISPR/Cas9 library to destroy E-box sequences in a genome-wide fashion. In parallel, we used the Brunello library to knock out protein-coding genes. We performed high-throughput screens with these libraries in four MYC-dependent cancer cell lines-K562, ST486, HepG2, and MCF7-which revealed several essential E-boxes and genes. Among them, we pinpointed crucial common and cell-type-specific MYC-regulated genes involved in pathways associated with cancer development. Extensive validation of our approach confirmed that E-box disruption affects MYC binding, target-gene expression, and cell proliferation in vitro as well as tumor growth in vivo. Our unique, well-validated tool opens new possibilities to gain novel insights into MYC-dependent vulnerabilities in cancer cells.

Keywords: CRISPR/Cas9; MYC; MYC target genes; high-throughput screen; transcription factor.

Fig. 1

Design and generation of the MYC‐EBOX‐CRISPR library for genome‐wide disruption of MYC binding sites. (A) Library was designed based on publicly available MYC‐ChIP‐seq data in MYC‐dependent K562, MCF7, HepG2, and Burkitt lymphoma (BL) cell lines. After excluding E‐boxes in coding exons, all possible sgRNAs targeting remaining E‐boxes were designed, based on the presence of PAM sequence. sgRNAs with predicted off‐target binding were filtered out. The final library contains 43 350 sgRNAs targeting E‐boxes, 1000 non‐targeting (NT) sgRNAs as a negative control, and four sgRNAs targeting MYC as a positive control. (B) Number of sgRNA constructs per E‐box. (C) Genomic location of E‐boxes targeted by the library. (D) Overlap of targeted E‐boxes for selected cancer cell lines. (E) Percentage of MYC binding sites targeted by the MYC‐EBOX‐CRISPR library in various cell lines, based on available MYC‐ChIP‐Seq data. In red, are cell lines for which the library was designed. (F) Distribution of sgRNA constructs in the MYC‐EBOX‐CRISPR plasmid library determined by NGS. All sgRNAs were present in the library.

Fig. 2

High‐throughput screen with MYC‐EBOX‐CRISPR and Brunello libraries. (A) Experimental approach: (1) MYC‐EBOX‐CRISPR library to destroy E‐box sequences, disrupt MYC binding, and its effect on target‐gene expression (2) Brunello library for genome‐wide gene knockout. (B) Scheme of the high‐throughput screen in cancer cells with MYC‐EBOX‐CRISPR and Brunello libraries. (C) DESeq2 analysis revealed essential genes in Brunello library and (D) essential E‐boxes in MYC‐EBOX‐CRISPR (depleted genes and E‐boxes in blue, enriched genes and E‐boxes in red). (E) Circos plots showing log2 fold change (FC) values for genes in Brunello screen (outer circle) and E‐boxes in MYC‐EBOX‐CRISPR screen (inner circle) across the chromosomes. Blue dots indicate genes and E‐boxes significantly (P _adj < 0.001) depleted or enriched, black lines denote log2FC = 0. (A and B) were created using BioRender.

Fig. 3

Essential MYC‐regulated processes and pathways. (A) Top five Gene Ontology (GO) terms for essential genes from Brunello library. (B) Top five GO terms for genes localized up to 50 kb from essential E‐boxes. (C) Overlap of essential E‐boxes from MYC‐EBOX‐CRISPR library. (D) Overlap of essential genes from Brunello library.

Fig. 4

Validation of selected E‐boxes and target genes. (A) Efficiency of disruption of selected E‐boxes and the spectrum of mutations introduced by individual sgRNAs, demonstrated by TIDE analysis. Size distribution of introduced indels ranged from ≤ −30 to +2 bp. Colors indicate percentage of sequences with a given indel size. (B) qRT‐PCR analysis of genes adjacent to selected E‐boxes upon CRISPR/Cas9 disruption of E‐box sequences. Known MYC‐regulated genes are underlined; genes essential or at least fourfold depleted in Brunello screen are in red. Noncoding genes are in green. Despite the name ‘antisense’, PRKCQ‐AS1 does not overlap with PRKCQ. NT—average of two non‐targeting (negative control) sgRNAs. Mean and SD of two independent experiments, each performed in triplicate, are shown. **, P < 0.01; ***, P < 0.001, Student's t‐test. (C) Cell viability upon disruption of selected E‐boxes and knockout of adjacent genes was measured using CellTiter‐Glo assay at three time points: 0, 48, and 96 h. Shown are mean values and SD from three independent experiments, each performed in triplicate. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, Student's t‐test. (D) Luciferase reporter assay for selected E‐boxes upon MYC knockdown with shRNA. Decreased luminescence signal was observed for all E‐boxes in MYC‐shRNA samples vs. NT control. **, P < 0.01; ****, P < 0.0001, Student's t‐test. Mean and SD of three independent experiments, each performed in triplicate, are shown. EV, empty vector, negative control; MYC‐RE, MYC responsive element, positive control. (E) MYC‐ChIP‐qPCR analysis of MYC binding upon E‐box disruption. Cells were infected with sgRNAs targeting selected E‐boxes. MYC binding was decreased for chr17_BS377 and for chr10_BS212 but not for chr11_BS79. Mean and SD from three replicates are shown.

Fig. 5

E‐box disruption inhibits tumor growth in vivo. HepG2 cells transduced with a non‐targeting (NT) control sgRNA or sgRNAs targeting E‐boxes on chr1, chr11, and chr18 were subcutaneously injected into NOD/SCID mice (NT n = 6 tumors; chr1 n = 8; chr11 n = 4; chr18 n = 8). (A) Confirmation of the decreased growth of HepG2 cells in vitro upon targeting selected E‐boxes. Shown are mean values and SD from three independent experiments, each performed in triplicate. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, Student's t‐test. (B) Luciferase‐based bioluminescence imaging of tumors over 5 weeks, mean and SEM. *, P < 0.05; ***, P < 0.001, 2‐way ANOVA. (C) Representative images of luciferase‐based bioluminescence imaging on Week 5. (D) Volume of tumors excised from mice, median with interquartile range. *, P < 0.05, Kruskal–Wallis test with Dunn's post‐test. (E) Representative images of excised tumors [cm].

Fig. 6

Grammar of the E‐box sequence context. (A) Expression of two genes adjacent to an E‐box on chromosome 10 (chr10_BS212) was examined in monoclonal cell lines derived from K562 cells transduced with an sgRNA targeting this E‐box. A spectrum of clones with various modifications of the E‐box was obtained, including wild‐type (WT) homozygotes (n = 7; K562 cells are triploid); homozygotes with the +G insertion after E‐box (n = 3); clones with two WT alleles and one mutated allele (n = 2); clones with one WT allele and two mutated alleles (n = 3); and clones with various indels on all three alleles (n = 9). Median with interquartile range is shown; *, P < 0.05, Kruskal–Wallis test with Dunn's post‐test. (B) Sequence logo (created using Seq2Logo) of the E‐boxes and 20 nt flanking sequences for non‐essential E‐boxes (left, n = 24 705) and E‐boxes essential in at least one cell line (right, n = 276). (C) Frequency of up to 10 nt upstream/downstream flanking essential (n = 276) and non‐essential (n = 24 705) E‐boxes. Since E‐boxes are (quasi)palindromic and can be read on either strand, G at +1 equals C at −1, etc. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; chi‐squared goodness‐of‐fit test.