Depletion of NK6 Homeobox 3 (NKX6.3) causes gastric carcinogenesis through copy number alterations by inducing impairment of DNA replication and repair regulation

Genomic stability maintenance requires correct DNA replication, chromosome segregation, and DNA repair, while defects of these processes result in tumor development or cell death. Although abnormalities in DNA replication and repair regulation are proposed as underlying causes for genomic instability, the detailed mechanism remains unclear. Here, we investigated whether NKX6.3 plays a role in the maintenance of genomic stability in gastric epithelial cells. NKX6.3 functioned as a transcription factor for CDT1 and RPA1, and its depletion increased replication fork rate, and fork asymmetry. Notably, we showed that abnormal DNA replication by the depletion of NKX6.3 caused DNA damage and induced homologous recombination inhibition. Depletion of NKX6.3 also caused copy number alterations of various genes in the vast chromosomal region. Hence, our findings underscore NKX6.3 might be a crucial factor of DNA replication and repair regulation from genomic instability in gastric epithelial cells.


INTRODUCTION
Genome stability is necessary for the integrity of human cells that are exposed continuously to the genotoxic agent. DNA repair is an important defense mechanism, maintaining genome integrity. Defective DNA repair can lead to cancer development by inducing mutations and chromosomal aberrations [1]. p53 is activated by DNA-damaged checkpoint signals and its activation coordinates various DNA damage responses to maintain genomic stability to prevent cancer development [2]. In particular, DNA double-strand breaks (DSBs), one of the most deleterious forms of DNA damage, are mostly repaired by two main mechanisms: (1) nonhomologous end-joining (NHEJ), which is active in the G1 cell cycle; and (2) homologous recombination (HR), which preferentially repairs DSB in the late S/G2 phase of the cell cycle [3,4]. Although p53 can regulate DSB repair genes, including RAD51 [5], the molecular mechanisms that underlie the formation of DNA copy number alterations (CNAs) in gastric carcinogenesis remain poorly understood.
Members of the NKX family are involved in various developmental processes, such as determining cell fate in the central nervous system, gastrointestinal tract, and pancreas [6]. In the stomach, NK6 Homeobox 3 (NKX6.3) is expressed in the epithelium of the most distal region, which eventually segregates to the lower/base region of the gastric unit [6,7]. Previously, we reported that NKX6.3, which showed frequently lost or reduced expression in gastric cancers, acted as a gastric tumor suppressor by inhibiting cell proliferation and inducing apoptosis [8,9]. It also protects gastric epithelial cells against the expression of the AICDA/APOBEC family, NFκB, and CBFβ gene expression, known to generate genome-wide somatic mutations and the accumulation of genetic alterations in gastric epithelial cells [10].
In the present study, we investigate subsequent effects caused by the loss of NKX6.3 on DNA replication and the DNA damage response in gastric mucosal epithelial cells to understand the molecular mechanisms underlying DNA CNAs during gastric carcinogenesis. We identify a novel role for NKX6.3 in genome maintenance, which provides new insights into its role in gastric tumorigenesis.

Depletion of NKX6.3 aberrantly regulates DNA Replication
To test our hypothesis that NKX6.3 depletion might be associated with abnormal DNA replication, we performed DNA fiber analysis, a method of measuring DNA synthesis and replication stress (RS) levels [11]. The cells were pulse-labeled with 5-chloro-2′-deoxyuridine (CIdU; red) and 5-iododeoxyuriduine (IdU; green). Ongoing replication can be detected by the simultaneous staining of the red and green DNA strands, but the singly-labeled red or green tracts indicate stalled forks and newly-initiated forks (new origin), respectively (Fig. 1A). Compared to HFE-145 shCtrl cells, NKX6.3depleted HFE-145 cells increased the frequencies of stalled forks 6-and 8-fold, and newly-initiated forks by approximately 5.5-and 8.5-fold, respectively (Fig. 1A). Moreover, the depletion of NKX6.3 increased the velocity of fork progression and fork asymmetry in the HFE-145 shNKX6.3 cells (Fig. 1B). In an analysis of the replication program by 5-bromo-2′-deoxyuridine (BrdU)/5-ethynyl-2′-deoxyuridine (EdU) dual-labeling, markedly reduced early-to-early S phase cells but increased mid-to-late and early-to-late S phase cells were observed in the HFE-145 shNKX6.3 cells (Fig. 1C). Furthermore, the NKX6.3 depletion increased the expression levels of CDT1, Orc1, Cdc6, MCM2, 3, and 7, while RPA1 expression was reduced (Fig. 1D). Because NKX6.3 was found to be a transcription factor for various genes [10], we asked whether NKX6.3 play a role as a transcription factor for DNA replicationrelated genes. ChIP analysis confirmed that the depletion of NKX6.3 increased the NKX6.3 binding activity to the promoter regions of the CDT1 and RPA1 genes. Then, NKX6.3 depletion increased the expression of CDT1, whereas it decreased RPA1 expression at the mRNA and protein levels ( Fig. 1E-G), suggesting that NKX6.3 could act as a transcription factor for CDT1 and RPA1. When we examined the effect of CDT1 overexpression and RPA1 depletion on DNA replication in the HFE-145 shCtrl cells, the ectopic expression of CDT1 and RPA1 deficit increased the fork progression speed ( Supplementary Fig. S1A). Besides, the knockdown of CDT1 and ectopic expression of RPA1 decreased the frequencies of stalled forks and newly-initiated forks, respectively, compared to those in the HFE-145 shNKX6.3#1 and HFE-145 shNKX6.3#2 cells ( Supplementary Fig. S1B, C). Moreover, the knockdown of CDT1 and ectopic expression of RPA1 in the NKX6.3-depleted HFE-145 cells reduced the velocity of fork progression and fork symmetry induced by NKX6.3 depletion (Supplementary Fig. S1D). These results thus suggest that NKX6.3 may modulate DNA replication via the transcriptional regulation of CDT1 and RPA1.
Previously, we found over 150 target genes of NKX6.3 through ChIP cloning and sequencing [8]. The target genes were analyzed in the TFBS site (http://tfbsdb.systemsbiology.net), and CDT1 and RPA1 genes involved in DNA replication were selected for further studies. Expectedly, the ectopic expression of CDT1 and depletion of RPA1 increased the expression levels of γH2AX in the HFE-145 shCtrl cells ( Supplementary Fig. S3A). The ectopic expression of CDT1 resulted in an increase in HR and the total GFP-positive population, whereas the depletion of RPA1 repressed HR and decreased the total GFP-positive population in the HFE-145 shCtrl cells ( Supplementary Fig. S3A). However, the ectopic expression of CDT1 and depletion of RPA1 did not increase NHEJ in the HFE-145 shCtrl cells ( Supplementary Fig. S3A). In the NKX6.3-depleted HFE-145 cells, the knockdown of CDT1 and ectopic expression of RPA1 significantly reduced the tail moment ( Supplementary Fig.  S3B). The expression of γH2AX and 53BP1 was significantly reduced, whereas the expression of Rad51 and RPA32 was markedly increased ( Supplementary Fig. S3C, D). Additionally, the knockdown of CDT1 and ectopic expression of RPA1 restored the effects on HR and NHEJ by NKX6.3-depleted HFE-145 cells ( Supplementary Fig. S3E). Taken together, these results suggest that NKX6.3 contribute to the repair of DSBs through the regulation of HR and NHEJ by regulating the expression of CDT1 and RPA1.
When we analyzed the CNAs of 34 paired samples from The Cancer Genome Atlas Stomach Adenocarcinoma (TCGA_STAD) dataset, gained or amplified copy numbers were found in chromosomes 6, 14, and 20, and loss or deleted copy numbers were found in chromosomes 4 and 18 ( Supplementary Fig. S5A). This analysis led us to focus on the CNAs in chromosomes 4, and 14. As expected, the DNA copy number of 438 genes, including SORBS2 and SPACL1 in chromosome 4, showed loss or deletion ( Fig. 4D and Supplementary Table S4), whereas the DNA copy number of 315 genes, including DLGAP5 and CDKN3 in chromosome 14 showed gain or amplification in the HFE-145 shNKX6.3#1 and HFE-145 shNKX6.3#2 cells (Fig. 4E, F and Supplementary Table  S4). Since GSEA analysis showed that genes with CNA due to NKX6.3 depletion were most closely associated with mitotic spindle formation (Fig. 4C), we verified the CNAs of genes involved in mitotic spindle formation, such as SORBS2, SPARCL1, DLGAP5, and CDKN3, using real-time PCR. The results showed that the depletion of NKX6.3 reduced the DNA copy number and mRNA expression levels of SORBS2 and SPARCL1 but increased the DNA copy number and mRNA expression levels of DLGAP5 and CDKN3 in the HFE-145 shNKX6.3#1 and HFE-145 shNKX6.3#2 cells (Fig. 4F, G). In Western blot analysis, the depletion of NKX6.3 decreased the expression levels of SORBS2 and SPARCL1 proteins but increased the expression levels of DLGAP5 and CDKN3 proteins in the HFE-145 cells (Fig. 4H). We further confirmed that altered DNA copy number and expression of mRNA and protein of these genes in the tumors derived from mice implanted with HFE-145 shNKX6.3#1 and HFE-145 shNKX6.3#2 cells ( Supplementary Fig. S5B-D). Thus, we concluded that the depletion of NKX6.3 in gastric epithelial cells could induce CNAs, which subsequently affected the expression levels of various cancer-related proteins.  CDKN3, we examined NKX6.3 expression and these genes' CNAs and expression in gastric cancer tissues. The expression levels of CDKN3 and DLGAP5 were markedly increased, whereas those of SPARCL1 and SORBS2 were decreased in 60 human gastric cancers (Fig. 5A). In addition, the mRNA expression levels of CDKN3, DLGAP5, SPARCL1, and SORBS2 were positively correlated with CNAs in our gastric cancer samples (Fig. 5B). Furthermore, NKX6.3 protein expression was positively correlated with the expression of SORBS2 and SPACL1 but inversely correlated with the expression of CDKN3, and DLGAP5 proteins and CNAs (Fig. 5C, D). In addition, the mRNA expression levels of CDKN3, DLGAP5, SPARCL1, and SORBS2, were positively correlated with CNAs in the TCGA STAD data set (Fig. 5E). These results indicate that NKX6.3 depletion in normal gastric epithelial cells may activate oncogenes, such as CDKN3 and DLGAP5, and inactivate tumor suppressor genes, such as SPARCL1 and SORBS2, which subsequently lead to cancer development.

DISCUSSION
DNA RS, which indicates a change in the rate of replication fork progression and an increase in stalled and collapsed forks, is known to be an important mechanism causing genomic instability that can cause various diseases, including cancer [16][17][18]. DNA RS is an important mechanism for genomic instability [14]. It can accelerate cell growth by inactivating tumor suppressors and activating oncogenes [19]. Here, we provide evidences showing that NKX6.3 depletion could lead to DNA RS, which results in genomic instability. DNA replication is initiated by DNA binding of the origin recognition complex and the sequential binding of other replication initiator proteins [20,21]. CDT1 and RPA1 is a key protein in the DNA replication, recombination and repair, and the activation of damage checkpoints [22][23][24][25][26][27]. Previous reports have shown that the overexpression of CDT1 and depletion of RPA1 could increase the activation of origin, induce DNA damage, stalled fork replication, defects in DNA damage repair, and promote genomic instability during tumor development [28][29][30]. Interestingly, we showed that NKX6.3 regulated CDT1 and RPA1 expression by acting as a transcription factor and that restoring the expression of these two proteins relieved the DNA RS caused by NKX6.3 depletion. These results demonstrate that NKX6.3 was a key factor regulating DNA replication by regulating CDT1 and RPA1 expression.
RS can lead to aberrant DNA replication fork speed and induce DNA damage [13]. DNA DSBs are a major type of DNA damage that evokes a checkpoint response and early-acting regulators, such as the MRN complex and master kinase ataxia-telangiectasia mutation (ATM), which collaborate in initiating this response cascade and repair pathway [31]. Importantly, the MRN complex can recognize DSBs and locate to the DSBs during all stages of the cell cycle [32], suggesting that the MRN complex plays an important role in checkpoint signaling from stalled replication forks. In addition, DNA DSB repair involves NHEJ or HR [33]. HR is the major error-free mechanism involved in DSB repair associated with replication forks [3]. It is active in the late S/G2 phase and involves strand invasion using the sister chromatid as a template [34]. Single-stranded DNA generated by DNA end-resection can activate ATR kinase [35], which promotes the contributions of ATR and its effector Chk1 kinase in the S and G2 phases [35]. DNA damage was shown to reduce CDT1 expression, and the overexpression of CDT1 induced re-replication, leading to DNA damage and inhibiting DNA repair [36][37][38]. In addition, the RPA protein is known to play an important role in DNA damage and the loss of expression and mutation of RPA1 interfere with the DNA strand exchange activity of RAD51 and lead to DSB repair failure [29,[39][40][41]. Notably, when RAD51, the rate-limiting factor in HR repair [42][43][44], becomes exhausted due to DNA damage, it leads to RAD52-dependent error-prone DNA repair pathways such as BIR and SSA [44][45][46]. In this study, the depletion of NKX6.3 induced the repression of HR and increased DNA damage, BIR, SSA, and NHEJ. Therefore, we hypothesized that the abnormal expression of CDT1 and RPA1 would play an important role in DNA damage and repair by NKX6.3 depletion. Expectedly, the restoration of CDT1 and RPA1 expression, which is abnormally regulated by NKX6.3 depletion, induced the recovery of DNA damage and repair induced by NKX6.3 depletion. Additionally, the expression of NKX6.3 shows a strong inverse correlation with CDT1 and a strong positive correlation with RPA1 in tumors derived from mice implanted with HFE-145 shNKX6.3#1 and HFE-145 shNKX6.3#2 cells and human gastric cancer tissues. All of these findings suggest that NKX6.3 acts as an important factor in regulating DNA DSB repair by modulating DNA replication and the HR system by regulating CDT1 and RPA1. Future studies are needed to determine whether DSB repair defects caused by NKX6.3 depletion could contribute to the progression of gastric cancer.
Previous studies have suggested that RS can lead to CNAs [47,48]. CNAs occur due to errors in the repair process. For example, the HR and NHEJ of DSBs are induced by a number of factors, such as DNA RS [15]. CNAs are major genomic alterations that contribute to carcinogenesis and tumor progression caused by oncogene activation and tumor suppressor gene inactivation. Here, NKX6.3 depletion could lead to frequent gain or amplification in chromosomes 6 and 14, with loss or deletion in chromosomes 4 and 18, suggesting that NKX6.3 depletion might lead to the aberrant expression of various cancer-related proteins by inducing CNAs. The results suggest that CNAs caused by NKX6.3 depletion in normal gastric epithelial cells might lead to cancer development through activating oncogenes and inactivating tumor suppressor genes.
Herein, for the first time, we demonstrate that NKX6.3 plays a critical role in maintaining genomic stability by regulating DNA replication and DNA damage repair. NKX6.3 regulates cell-cycle progression, DNA replication, and DNA damage repair through transcriptional regulation of the novel target genes, CDT1 and RPA1. We demonstrated that depletion of NKX6.3-induced CNA, which subsequently leads to abnormal expression of cancerrelated proteins, consequently promoting the development of gastric cancer. Our findings represent a NKX6.3 plays an important role in gastric carcinogenesis by controlling CNAs through regulating DNA replication and DNA repair. These findings not only partially uncovered the role of NKX6.3 in the DNA repair pathway, but also provide an invaluable basis for understanding the development of gastric cancer with DNA damage repair defects.

MATERIALS AND METHODS Samples and ethics approval
The Chonnam National University Hwasun Hospital Biobank, a member of the Korea Biobank Network, provided a total of 60 frozen human gastric cancer tissues. The study was approved by the Institutional Review Board of The Catholic University of Korea, College of Medicine (MC15SISI0015). Written informed consent was obtained from all subjects in accordance with the Declaration of Helsinki. There was no evidence of familial cancer in any of the patients. Animal experiments were performed in mice maintained under pathogen-free conditions after approval by the Animal Experiment Ethics Committee of the Catholic University of Korea College of Medicine (CUMC-2017-0018-01).

In vivo xenograft mouse experiment
For the in vivo xenograft assay, 5 × 10 6 HFE-145 shCtrl , HFE-145 shNKX6.3#1 , and HFE-145 shNKX6.3#2 cells were mixed with 0.2 ml of phosphate-buffered saline (PBS, pH7.4) and 30% (v/v) Matrigel (BD Biosciences, San Jose, CA, USA). BALB/c-nude mice between four and five weeks old were purchased from ORIENT (Seongnam, Korea). The cells were inoculated subcutaneously into the side of the flank of the nude mice. The mice were examined twice per week for tumor formation at the injection site. After three weeks, the tumor volumes were measured every three days, and volume calculated as length × width 2 × 0.5. After seven weeks of tumor formation, the tumors were carefully removed, photographed, and weighed. Each experimental group consisted of five mice and tumor growth was quantified by measuring the tumor sizes using calipers.

Whole-genome sequencing and copy number data analysis
CNAs were analyzed using whole-genome sequencing (WGS), as described previously [10]. Briefly, 0.5-3 μg of DNA from the HFE-145 shCtrl , HFE-145 shNKX6.3#1 , and HFE-145 shNKX6.3#2 cells were used to prepare the sequencing library by shearing the DNA followed by ligating the sequencing adapters. The captured DNA was sequenced using the Illumina HiSeq X Ten platform and paired-end sequencing reads for the HFE-145 shCtrl , HFE-145 shNKX6.3#1 , and HFE-145 shNKX6.3#2 cells were generated. We used FastQC 0.1 software to verify the quality of the sequence reads. The paired-end library of~500-bp inserts was sequenced with 150-bp pairedend reads. The raw sequencing data were processed with an ultrafast Isaac DNA sequence aligner designed to align the next-generation sequencing data with a low-error rate using the human genome hg19, GRCh37.
The WGS data were analyzed using Nexus Copy Number version 9.0 software (Biodiscovery Inc., El Segundo, CA, USA). The Nexus calling algorithm, SNPrank segmentation is based on the Circular Binary Segmentation model [49]. Abnormal CNVs and allele ratios were defined as the Log2 ratio values for gains and losses at 0.2 and −0.2, respectively. The thresholds for high copy number gains and homozygous deletions were set at 0.65 and −0.65, respectively.
After quantification of the genomic DNA extracted from the HFE-145 shCtrl , HFE-145 shNKX6.3#1 , and HFE-145 shNKX6.3#2 cells, a real-time SYBR Green quantitative polymerase chain reaction (qPCR) was performed on a Bio-Rad IQ5 real-time PCR platform (Bio-Rad Laboratories, Hercules, CA, USA). Specific primers for the analysis of the DLGAP5, CDKN3, SORBS2, and SPARCL1 DNA copy numbers were designed according to GenBank's genomic sequence. All samples were PCR amplified and normalized using oligonucleotide primers specific for constitutively expressed genes and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primers for SYBR Green analysis were designed based on gene-specific nonhomologous DNA sequences. The primer sequences are described in Supplementary Table S5.

DNA fiber assay
The cells were pulse-labeled with 25 μM CIdU (Sigma-Aldrich, St. Louis, MO, USA) for 30 min, then gently washed with fresh pre-warmed medium and exposed to hydroxyurea (2 mM) or untreated media, and a second pulse of 250 μM IdU (Millipore-Sigma) for 30 min. The DNA fibers were spread as described [52] before the standard detection of the CIdU and IdU tracts. For all experimental conditions, three slides were stretched and stained for each condition. CIdU was detected by a rat anti-BrdU antibody (Abcam, ab6326) and IdU by a mouse anti-BrdU antibody (Abcam, ab8152). DyLight 550 anti-rat (ThermoFisher Scientist, Waltham, MA, USA) and Alexa Fluor 488 anti-mouse (Invitrogen) secondary antibodies were used. The fibers were imaged using a Carl Zeiss LSM800 w/Airyscan confocal microscope and analyzed using ImageJ software. Statistical analyses were conducted using Prism software.

Replication program
For double labeling, cells grown on coverslips were pulse-labeled with 10 μM BrdU for 30 min and washed and incubated in a fresh medium for 4 h. After the chase, the cells were incubated with 10 μM EdU for 30 min. The cells were then fixed in methanol. DNA was denatured in 2 M HCl for 30 min and BrdU was detected using a mouse anti-BrdU antibody and the appropriate secondary antibody. EdU was detected using the Click-iT EdU Alexa Fluor 568 Imaging Assay Kit (Life Technology, Carlsbad, CA, USA) according to the manufacturer's protocol. High content Carl Zeiss LSM800 w/Airyscan microscopy was used for S phase analysis. The replication and S phase patterns were clustered into early, mid or late groups in the scatter plots of BrdU and EdU versus DNA content.

Comet assay
The cells were collected and resuspended in PBS at a density of 5,000 per μl. Samples of 10 μl were mixed with 100 μl of 37°C low melting point agarose in PBS and spotted on the slides pre-coated with normal melting point agarose in ddH 2 O. Slides were covered with coverslips and kept flat for 10 min at 4°C. The coverslips were gently removed and the slides were immersed in cold alkaline electrophoresis buffer at 4°C for 20 min and electrophoresis was performed at 4°C for 25 min. The slides were then held for 10 min at 4°C, and washed with cold PBS and cold ddH 2 O. Next, the slides were dehydrated by continuous 5 min washes in a cold gradedethanol series (70%, 90%, and 100%) at 4°C, air-dried, and stored at room temperature. The following day, the slides were rehydrated in ddH 2 O and stained with SybrGold for 5 min at room temperature. After washing with PBS, the slides were mounted with Vectashield mounting reagent. The images were acquired using a Carl Zeiss LSM800 w/Airyscan microscope.

Co-immunoprecipitation (Co-IP)
Co-immunoprecipitations were performed as follows. Cells in one nearly confluent 10 cm dish were harvested in 1 ml of protein lysis buffer as described above. Subsequently, BSA-blocked protein G beads (Fastflow; Millipore-Sigma) were added and incubated at 4°C for 30 min to preclear the lysates. One percent of the total dose was separated and rotated for 1 h at 4°C in the presence of 2 μg of anti-p53, anti-PI3K p85α, anti-Myc-taq, and BSA-blocking protein G beads to allow immunoprecipitation. After washing the beads with lysis buffer, the proteins were eluted using Laemmli loading buffer and immunoblots were performed as described above.

Statistical analyses
The MTT, colony formation assays, FACS, and western assays were performed at least twice, and the means ± SD are displayed in the bar graphs. One-or two-way ANOVA was conducted to assess the differences J.H. Yoon et al. between the means. GraphPad Prism version 7.00 for Windows (GraphPad Software, La Jolla, CA, USA) was used for the statistical analyses. All other data were analyzed using the student's t-test. Statistical significance is indicated as asterisks in figures: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.