The human colorectal cancer cell line LoVo, which was purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA), was cultured in F12K medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), penicillin (100 U/ml) and streptomycin (100 g/ml), in a humidified 5% CO₂ incubator at 37°C.

The ChIP was performed following the instructions of an Enzymatic Chromatin IP kit (cat no. 9003; Cell Signaling Technology, Inc., Danvers, MA, USA). Briefly, ~4×10⁷ cells were prepared for each experiment. Formaldehyde (1% concentration) was added to crosslink proteins to DNA for 10 min. Following cell lysis and nuclei collection, the chromatin was fragmented by partial digestion with micrococcal nuclease to obtain chromatin fragments of 1–5 nucleosomes in size (150–900 bp). The crosslinked chromatin preparation was then immunoprecipitated with 5 µg polyclonal ZEB1 antibody (cat no. sc-25388 X; Santa Cruz Biotechnology, Inc., Dallas, TX, USA) or negative control normal rabbit IgG (cat no. 2729; Cell Signaling Technology, Inc.) at 4°C overnight. Elution of chromatin from the crosslinked complex using Protein G magnetic beads was performed with KingFisher Flex Magnetic Particle Processors. After reversing the crosslinks, DNA was purified using the reagents and spin columns provided in the Enzymatic Chromatin IP kit. In addition, a positive control histone H3 rabbit monoclonal antibody (cat no. 4620; Cell Signaling Technology, Inc.) and a primer for its known binding gene, ribosomal protein L30, were also included in the experiment to analyze IP efficiency. Three biological replicates were performed and successful enrichment was validated in each experiment.

ChIP DNA was amplified to 7.5 µg, using the high-performance liquid chromatography-purified primers: A, 5′-GTTTCCCAGTCACGG TC(N)₉-3′; and B, 5′-GTTTCCCAGTCACGGTC-3′. The dUTP-containing post-amplified samples was then fragmented by uracil-DNA glycosylase and labeled by terminal deoxynucleotidyl transferase (cat no. 72033) with the biotin-like labeling reagent (cat no. 79015) (both from USB Corp.; Affymetrix, Inc., Santa Clara, CA, USA). The labeled DNA was hybridized to six Affymetrix GeneChip Human Promoter 1.0R arrays (three biological replicates using ZEB1 antibody and three using normal rabbit IgG) at 45°C for 16 h. The arrays were washed and stained by R-phycoerythrin streptavidin (cat no. S-866; Molecular Probes; Thermo Fisher Scientific, Inc., Waltham, MA, USA) using Fluidics Station 450 and then scanned using GeneChip Scanner 3000 7G, which was controlled by Affymetrix GeneChip Command Console software (Affymetrix, Inc.).

Affymetrix Tiling Analysis software (Affymetrix, Inc.) and Integrated Genome Browser were used to select the positive intervals that satisfied the two conditions: that the signal intensity of the ZEB1-binding DNA and the intensity ratio of the ZEB1-binding DNA signal/normal rabbit IgG-binding DNA signal were within the top 1% among all the intervals. The filter criteria ruled out the false-positive regions with a high intensity ratio that resulted from the division of two infinitely small numbers. The positive intervals of each chromosome in bed files were uploaded to the GALAXYP online platform (usegalaxyp.org) to produce a tail-to-head concatenation and then submitted to the Cis-regulatory Element Annotation System (CEAS) to produce nearby gene mapping and motif identification. For each positive ChIP region, CEAS identified the nearest RefSeq gene names and predicted locations within 300 kb upstream or downstream of the gene. Regions within 1 kb upstream from the gene 5′ start site were assumed to be proximal promoters; those within 1 kb downstream from the gene 3′-end were reported to be immediate downstream regulatory elements; while those distributed >1 kb from the gene were assumed to be enhancers (20).

The LoVo cell line was selected as a suitable host for transfection. To avoid off-target effects, two different sequences were used to knock down ZEB1. Using Endofectin™ Plus, cells were transfected with plasmids (both from GeneCopoeia, Inc., Rockville, MD, USA) containing short hairpin RNA targeting ZEB1 HSH017963-4-LVRH1GP (OS241004) or HSH017963-5-LVRH1GP (OS397209). Parallel transfection was performed using plasmids with non-targeting shRNA CSHCTR001-1-LVRH1GP (OSNEG20) to generate scrambled control clones. The cells were then selected with puromycin and after 4 weeks, single colonies were analyzed for ZEB1 expression by western blotting assay. The results revealed that the transfections had successfully induced a decrease in the protein level of ZEB1. The shRNA targeted sequences were as follows: ZEB1-knockdown-1, 5′-GGTCAACTATCACTAGTGT-3′; ZEB1-knockdown-2, 5′-TGATCAGCCTCAATCTGCA-3′; and scrambled control, 5′-GCTTCGCGCCGTAGTCTTA-3′. Knockdown is referred to as KD hereinafter.

Total RNA from the cells was extracted and reverse transcribed into cDNA according to the protocols of the Total RNA kit I (Omega Bio-Tek, Inc., Norcross, GA, USA) and First Strand cDNA Synthesis kit (GeneCopoeia, Inc.). The cDNA template (100 ng) was amplified using 0.2 µM primers and the 20 µl SYBR-Green-based qPCR reaction system (GeneCopoeia, Inc.). A Bio-Rad IQ 5 instrument (Bio-Rad Laboratories, Inc., Hercules, CA, USA) was used to perform the reaction and detect the fluorescent signals. Standard curves were created to confirm the amplification efficiency of each gene, while melting curves ensured the specificity of the amplification. Normalization to the reference gene (β-actin) was performed for each sample. Fold changes between the mRNA level of the ZEB1-KD groups and the scrambled control group were calculated according to the 2^−ΔΔCq method (21). The sequences of primers were as follows (5′-3′): USP17 forward, AGGTGAGTGGCAGTTCAACC and reverse, GGAAGCTTCTTCCTGGGAGC; CHD1L forward, ACTAGCATTCCTGTATTCTGGGG and reverse, CACGCTCATAGCTGTAGCCTC; DUX4 forward, CGATGGCCCTCCCGACA and reverse, GGCGTGACCTCTCATTCTGA; and β-actin forward, CCTAGAAGCATTTGCGGTGG and reverse, GAGCTACGAGCTGCCTGACG.

Cells were seeded into 96-well plates at 5,000 cells/200 µl/well, treated with increasing concentrations of the DNA-damaging agents cisplatin (cat no. P4394) and etoposide (cat no. E1383) (both from Sigma-Aldrich; Merck Millipore, Darmstadt, Germany), and then cultured for 24 h. Cell proliferation was assessed using Cell Counting Kit-8 (Boster Biological Technology, Ltd., Wuhan, China), and IC₅₀ values were calculated.

Cells were seeded in 6-well plates at 5×10⁵ cells/2.5 ml/well and treated with DNA-damaging agents at the indicated concentrations for 24 h. For cell cycle analysis, cell suspensions were harvested, washed with PBS, and resuspended in propidium iodide (PI) stain buffer [0.1% (v/v) Triton X-100, 10 µg/ml PI and 100 µg/ml DNase-free RNaseA; Sigma-Aldrich; Merck Millipore] for 30 min in the dark. The DNA contents of the cells were analyzed using a BD FACSCalibur (BD Biosciences, Franklin Lakes, NJ, USA). For apoptosis analysis, the cells were stained with PI and allophycocyanin-conjugated Annexin V and the FACSCalibur was used to detect early-stage apoptotic cells.

Data are presented as the mean ± standard deviation from at least three independent experiments. When data followed a normal distribution, the statistical significance between experimental values was assessed by unpaired Student's t-test using SPSS software, and P<0.05 was considered to indicate a statistically significant difference.

The Human promoter 1.0R array is tiled with >25,500 human promoter regions. Each region covers ~7.5 kb upstream to 2.45 kb downstream of the 5′ transcription start sites of >1,300 cancer-associated genes. The DNA fractions that co-immunoprecipitated with ZEB1 were hybridized to the array, in order to identify putative downstream targets of ZEB1 (normal rabbit IgG-binding DNA was used as the background).

Using the Homo sapiens hg18 genome (Refgene; NCBI Build 36.1; March 2006), continuous positive regions enriched along the USP17 (Fig. 1A), CHD1L (Fig. 1B) and DUX4 (Fig. 1C) gene sequences were identified. CEAS predicted ZEB1 binding sites located at the enhancer, promoter and immediate downstream regions of the USP17 gene; at the enhancer, intron and exon regions of the DUX4 gene; and at the enhancer regions of the CHD1L gene.

As ZEB1 is a well-known E-box binding transcriptional repressor, the sequences of each positive region from the three putative target genes (in FASTA format) were screened for the presence of E-box elements (5′-CANNTG-3′). As shown in Fig. 2A-a, CACCTG, CACGTG, CAGCTG, CAGGTG were the four most canonical E-box sequences. Duplication of the sequences, particularly in an AGGTG-N_n-CACCT structure, have been reported to possess strong ZEB1-binding affinity (22,23). In 24 positive intervals along the USP17 gene, 10 CACCTG sequences, 1 CAGGTG sequence and 12 CAGCTG sequences, and, notably, an AGGTG-N_n-CACCT duplication were identified (Fig. 2A). In 12 upstream positive regions of the CHD1L gene, two CACCTG sequences were detected. The 7 positive regions along the DUX4 gene were identified to contain CAGCTG sequences.

In addition to E-boxes, CEAS identified ~40 predicted ZEB1-binding motifs using the default algorithm, arranged in ascending order of p-value. The top four ranked sequences predicted to be ZEB1-interacting DNA motifs following CEAS analysis are presented in Fig. 2B, namely E2f, Pax-2, Elk-1 and STAT3 motifs. The FASTA files were also searched for the de novo predicted ZEB1-binding elements, and identified those consensus binding sites within the positive regions along the three genes.

To confirm the ChIP-on-chip data, the mRNA level of the three genes was examined for their responses to ZEB1 knockdown by RT-qPCR. As demonstrated in Fig. 3, the three genes exhibited significantly enhanced expression following ZEB1 depletion compared with the scrambled control group. The relative expression levels of USP17, CHD1L and DUX4 reached 1.48-, 11.21- and 2.16-fold in ZEB1-KD-1 cells, and 2.66-, 10.59- and 1.97-fold in ZEB1-KD-2 cells, respectively.

Overexpression of USP17 and CHD1L was previously reported to cause alterations to the DDR, resulting in increased sensitivity to DNA-damaging agents (24,25). Therefore, it was investigated whether silencing of ZEB1 may induce similar effects. Data demonstrated that when ZEB1 was knocked down by two specific shRNA constructs, cell proliferation was significantly impaired and cells were more susceptible to cisplatin and etoposide-induced DNA damage. As demonstrated in Fig. 4, ZEB1 knockdown enhanced the inhibitory effect of cisplatin and etoposide on cell proliferation. Compared with the scrambled control group, the two ZEB1-knockdown LoVo cell lines exhibited decreased IC₅₀ values for cisplatin and etoposide as shown in Table I.

A disabled G1/S checkpoint, together with its mediated anti-apoptotic effect, is another characteristic of an unbalanced DDR network (26). It has been previously reported that overexpression of USP17 or DUX4 may lead to G1/S arrest and a higher apoptotic rate (27,28). Therefore, the present study evaluated whether ZEB1 silencing impacted the cell cycle and apoptosis in a similar way. Cells were synchronized in G1 by serum starvation, then treated with cisplatin at 0.5 µg/ml or etoposide at 3 µg/ml for 24 h and analyzed using flow cytometry. As expected, >70% of ZEB1-depleted cells failed to exit G1 and progress into the S phase following drug treatment. By contrast, only 57 and 42% of scrambled shRNA-transfected cells accumulated in the G1 phase following cisplatin or etoposide treatment, respectively (Fig. 5A). The results of three independent experiments are summarized in Fig. 5B. With regard to the apoptosis assay, the proportion of cells undergoing early apoptosis was also markedly upregulated in response to ZEB1 depletion. Cisplatin and etoposide treatment induced early apoptosis in only 0.53 and 0.48% of scrambled control cells, respectively; however, the same treatment induced early apoptosis in 1.02 and 0.69% of ZEB1-KD-1 cells and 1.99 and 3.13% ZEB1-KD-2 cells (Fig. 5C). The results of three independent experiments are summarized in Fig. 5D.