Multiple genetic mutations caused by NKX6.3 depletion contribute to gastric tumorigenesis

NKX family members are involved in a variety of developmental processes such as cell fate determination in the central nervous system, gastrointestinal tract, and pancreas. However, whether NKX6.3 contributes to gastric carcinogenesis remains unclear. The objective of this study was to examine roles of NKX6.3 depletion in mutagenesis and gastric carcinogenesis, focusing on its effects on genetic alterations and expression of genes. Our results revealed that NKX6.3 depletion induced multiple genetic mutations in coding regions, including high frequency of point mutations such as cytosine-to-thymine and guanine-to-adenine transitions caused by aberrant expression of AICDA/APOBEC family in human gastric epithelial cells. Interestingly, NKX6.3 downregulated AICDA/APOBEC family, NFκB, and CBFβ genes by acting as a transcription factor while inhibiting deaminase activity in gastric epithelial cells. Functional relevance of NKX6.3 was validated in xenograft mice injected with NKX6.3 depleting cells. NKX6.3 depletion resulted in tumor formation and mutations of tumor-associated genes, including p53 and E-cadherin. Moreover, expression levels of NKX6.3 and its target genes were analyzed in tumors derived from mice implanted with NKX6.3 depleting cells and tissue samples of gastric cancer patients. Our results indicate that NKX6.3 depletion in gastric epithelial cells activates AICDA/APOBEC family, leading to accumulation of genetic mutations and eventually driving the development of gastric cancers.

Depletion of NKX6.3 leads to genetic mutations in coding regions. Next Table S2). Among these mutations, a high frequency (93%) of point mutations such as missense, silent, and nonsense mutations rather than frameshift, deletion, and insertion were found. The percentage of missense mutations in point mutations was higher than that of silent and nonsense mutations (Fig. 1C). In addition, substitutions from C to T and from G to A showed high percentages ( Supplementary Fig. S2B). In detail, missense and nonsense mutations were found in coding regions of 750 genes in HFE-145 shNKX6.3#1 cells and 767 genes in HFE-145 shNKX6.3#2 cells. Of them, 633 genes were overlapped. Mutations in these overlapped genes were found to be mainly C to T and G to A transitions (54.7%). In addition, frame shift, insertion, and deletion mutations were observed in coding regions of 95 genes in HFE-145 shNKX6.3#1 and 104 genes in HFE-145 shNKX6.3#2 cells. Of them, 69 genes were overlapped (Fig. 1D). Next, we analyzed the associated biological pathway using FunRich software. Results of analysis showed that NKX6.3 depletion-induced mutant genes including TP53, RhoA, EP300, and PIK3CA were involved in biological pathways such as intracellular junction assembly and/or maintenance, cellular proliferation, and apoptosis (Fig. 1E, Supplementary Fig. S2C and Supplementary Table S3). Mutation sites of these genes are shown in Fig. 1F. These results strongly indicate that depletion of NKX6.3 can lead to genetic mutations in several genes involved in cell growth, cell proliferation, and apoptosis.
To probe the biological significance of NKX6.3 as a transcriptional regulator of APOBEC gene family, effects of NKX6.3 on deaminase activity in HFE-145, AGS, and MKN1 cells were determined using fluorescence resonance energy transfer (FRET)-based assay. Results showed that deaminase activities in HFE-145 shNKX6.3 , AGS Mock , and MKN1 Mock cells that expressed AICDA/APOBEC gene family were increased. However, stable NKX6.3 expression in these cell lines resulted in significant decreases in deaminase activity ( Fig. 2C and Supplementary  Fig. S3F). Thus, we can conclude that NKX6.3 transcriptionally downregulates expression of AICDA/APOBEC gene family, Ung, and Apex1 in gastric epithelial cells. of CBFβ and NFκB were inversely associated with levels of NKX6.3 expression but positively associated with mutation rates in TCGA datasets ( Supplementary Fig. S4A). Importantly, NKX6.3 binding activities to promoter regions of CBFβ and NFκB p65 genes were observed in HFE-145 shCtrl , AGS NKX6.3 , and MKN1 NKX6.3 cells ( Fig. 2D and Supplementary Fig. S4B). Both mRNA and protein expression levels of CBFβ and NFκB p65 were significantly increased in HFE-145 shNKX6.3 cells but decreased in NKX6.3-expressing cells based on luciferase activity assay, real-time RT-PCR, and immunoblot analysis ( Fig. 2D and Supplementary Fig. S4C-F). In addition, knockdown of CBFβ or NFκB p65 significantly decreased mRNA and protein expression of AICDA/APOBEC gene family and deaminase activity in HFE-145 shNKX6.3 cells ( Fig. 2E and Supplementary Fig. S5A,B). Next, we determined 3D-PCR error rate using the first PCR amplification corresponding to 139 bp to confirm the effect of CBFβ or NFκB p65 on NKX6.3 depletion-induced somatic mutations in TP53 coding region in HFE-145 shNKX6.3 cells. By decreasing the denaturation temperature (Td), it was found that 87.9 °C was the minimal temperature allowing amplification of TP53 in HFE-145 shCtrl cells. By contrast, 3D-PCR could amplify TP53 gene in HFE-145 shNKX6.3 cells at temperatures as low as 85.6 °C. As expected, knockdown of CBFβ or NFκB p65 in HFE-145 shNKX6.3 was able to recover DNA at 86.6 °C, which was a higher temperature than that of NKX6.3-depleted cells ( Supplementary  Fig. S5C). In addition, 3D-PCR products from amplification at Td of 87.9, 85.6, and 86.6 °C were cloned and up to 11,9,13, and 11 clones were sequenced per sample. Knockdown of CBFβ or NFκB p65 reduced NKX6.  Table S4). In addition, a strong AICDA mutation pattern with the highest enrichment in wrC motif and a strong preference for C to T and C to A changes in wrC motif of AICDA/APOBEC mutation in NKX6.3-depleted cells were found ( Fig. 3A and Supplementary Table S4). Furthermore, APOBEC-and AICDA-induced mutations were mainly detected in coding regions and 3′-UTRs ( Supplementary Fig. S6A,B). Of mutations associated with NKX6.3 depletion, APOBEC and AICDA-induced mutations were found in 78 and 150 genes, respectively (Fig. 3B,C). According to related biological pathways, we analyzed gene set enrichment of APOBEC-and AICDA-induced mutant genes. Frequent mutations in genes involved in Nef and signal transduction, adrenoceptors and VEGFR3 signaling in lymphatic endothelium (including TP53, PIK3CA, ELMO1, and SMAD2), and genes involved in proline catabolism, DNA replication initiation, and NAD phosphorylation and dephosphorylation (including YWHAB, POLA1, PRIM2, and FFAR1) were found (Fig. 3B,C, Supplementary Fig. S6C,D, and Supplementary Table S5). Knockdown of APOBEC3B and AICDA partially recovered NKX6.3 depletion-induced mutations of PIK3CA and POLA1 genes in HFE-145 shNKX6.3 cells (Fig. 3D,E). These results suggest that NKX6.3 depletion can lead to a variety of mutations, including APOBEC-and AICDA-mutation patterns.

NKX6.3 regulates CBFβ and
Effects of NKX6.3 depletion on apoptosis and invasion. Since NKX6.3 was associated with mutations in CDH1, TP53, RhoA, PIK3CA, and EP300 genes (Fig. 1), we examined the expression of mutant genes and effects of NKX6.3 depletion on apoptosis and invasion. When we examined the expression and activity of E-cadherin, TP53, RhoA, PIK3CA, and EP300 in HFE-145 shNKX6.3 cells, NKX6.3 depletion reduced E-cadherin and p21 expression, increased RhoA expression and Akt phosphorylation, and induced decreased and fragmented expression of EP300 possibly due to p53, RhoA, or PIK3CA mutations (Fig. 4A). In addition, increased expression of cleaved PARP and increased caspase-3/7 activity after anti-Fas treatment were significantly inhibited in NKX6.3-depleted cells possibly due to caspase-8 mutation (Fig. 4B). Furthermore, mutations of these genes in HFE-145 shNKX6.3 cells were confirmed by sequencing (Fig. 4C). Next, to determine the effect of NKX6.3 depletion on cell migration and invasion, we performed transwell chemotaxis and Matrigel invasion assays. Cell migration and invasion activities were significantly increased in HFE-145 shNKX6.3 cells (Fig. 4D,E). In tumor sphere assay in vitro, depletion of NKX6.3 dramatically increased sphere number and size in HFE-145 cells (Fig. 4F).

NKX6.3 expression is inversely correlated with AICDA/APOBEC gene family and DNA repair genes in both gastric cancers and non-cancerous gastric mucosae.
To further confirm that NKX6.3 expression could regulate the expression of AICDA, NFκB p65, CBFβ, APOBEC3A and APOBEC3B, mRNA expression levels of these genes were compared with NKX6.3 expression in 55 non-neoplastic gastric mucosae. NKX6.3 expression was inversely correlated with expression of these genes ( Supplementary Fig. S7A) while expression levels of AICDA, APOBEC3A, and APOBEC3B were positively correlated with expression of NFκB p65 and CBFβ (Supplementary Fig. S7B).
In gastric cancers, mRNA expression levels of AICDA, NFκB p65, CBFβ, APOBEC3A, and APOBEC3B were markedly increased, similar to their expression profiles in gastric mucosae with H, pylori infection. However, NKX6.3 expression was inversely correlated with expression of these genes (Fig. 6C). In addition, NKX6.3 mRNA expression was significantly lower in gastric cancers with higher TNM stage ( Supplementary Fig. S7C), consistent with results of our previous study 15 . As expected, expression levels of AICDA, NFκB p65, CBFβ, APOBEC3A, and APOBEC3B were increased in gastric cancers with higher TNM stage ( Supplementary Fig. S7C). We also found that mRNA expression levels of CDH1, CDKN1A, and EP300 were significantly reduced while those of RhoA, ROCK1, ROCK2, PIK3CA, and CCND1 were significantly increased in gastric cancers with reduced or loss of NKX6.3 expression compared to those in NKX6.3 positive cases (Fig. 6C). We also examined the presence of CagA gene in gastric cancer tissues by western blot analysis, as described previously 14 . Results showed that expression levels of NKX6.3, CDH1, CDKN1A, and EP300 were decreased while expression levels of NFκB p65, AICDA, CBFβ, APOBEC3A, APOBEC3B, RhoA, ROCK1, ROCK2, PIK3CA, and CCND1 were increased in CagA positive gastric cancers compared to those in CagA negative cases (Fig. 6C).
Next, we determined the relationship of NKX6.3 expression with p53, AICDA, and EP300 protein levels based on immunohistochemistry data of 151 human gastric cancer tissue specimens 15 of 151 specimens, respectively ( Fig. 6D and Table 1). There was a significant relationship between altered expression levels of NKX6.3, p53, AICDA, and EP300 proteins and clinicopathologic parameters, including depth of invasion (Chi-Square test: P = 0.0026, P = 0.2302, P = 0.039, and P = 0.0553, respectively), lymph node metastasis (Chi-Square test: P < 0.0001, P = 0.148, P = 0.4184, and P = 0.0889, respectively), and TNM stage (Chi-Square test: P < 0.0001, P = 0.1373, P = 0.0539, and P = 0.029, respectively). Importantly, NKX6.3 protein expression was positively correlated with expression of EP300 and but inversely correlated with expression of AICDA protein (Spearman's test: P < 0.0001). EP300 protein expression was inversely correlated with that of AICDA protein (Spearman's test: P = 0.0058) ( Table 1). Kaplan-Meier analysis showed that there was no significant correlation between patient survival and p53 expression, EP300 expression, or AICDA expression alone (data not shown). However, combined analysis revealed that patients with negative NKX6.3 expression and/or negative p53, negative EP300, and positive AICDA expression had shorter overall survival time compared to those with positive NKX6.3 expression and/or positive p53, positive EP300, negative AICDA expression (P < 0.0001) (Fig. 6E). In 32 gastric cancers from TCGA datasets, cases with APOBEC3B and/or AICDA expression without NKX6.3 expression showed significantly higher number of mutations than those with NKX6.3 expression ( Supplementary  Fig. S7D). Taken together, these results indicate that NKX6.3 can regulate cell proliferation and apoptosis as well as cell motility and invasion processes by preventing accumulation of genetic alterations through downregulation  of AICDA, NFκB, CBFβ, and APOBEC3B expression in gastric epithelial cells, thereby contributing to the development and progression of gastric cancer.

Discussion
Recently, several studies have revealed genetic alterations underlying gastric carcinogenesis by multi-omics profiling [3][4][5]18 . A mutational signature highlights specific carcinogenic and mutation processes in gastric cancer, including microsatellite instability, CpG associated deamination, and activation of cytidine deaminases such as AICDA and APOBEC3B 3,4,7,18 . Here, we observed that NKX6.3 depletion could lead to somatic mutations, mainly G or C to A or T base pair transitions in coding regions in gastric epithelial cells, consistent with results of TCGA-based analysis ( Fig. 1A and Supplementary Fig. 1). For example, NKX6.3 depletion induced some driver mutations in CDH1, TP53, RhoA, PIK3CA, and EP300 genes, consistent with previous findings in gastric cancer [3][4][5] . CDH1 acts as a tumor suppressor. Decreased expression of CDH1 has been observed in gastric cancer with somatic mutation and methylation in the gene 19,20 . TP53 is another important tumor suppressor that negatively regulates cell cycle. Aberrant activity of TP53 caused by mutations is required for tumor formation 21 . Loss of p53 function can lead to impaired DNA replication, malignant transformation, genetic instability, and increased survival of cells with increased mutational load 22 . RhoA is a key member of the Rho family of small GTP-binding proteins that act as mediators between cell surface receptors and different intracellular signaling proteins 23 . RhoA overexpression has been observed in various cancers and RhoA activity has been implicated in tumorigenesis and tumor cell invasion 24 . In addition, highly recurrent mutations of RhoA have been detected in diffuse-type of gastric cancers 5,25 .
Mutations in PIK3CA which encodes p110 catalytic subunit of PI3K have been observed in various human cancers, including gastric cancers. These mutations can lead to activation of PI3K/AKT and its downstream signaling pathways, thereby contributing to carcinogenesis [26][27][28] . EP300 functions as a histone acetyltransferase. It is an important modulator in cell proliferation and differentiation through transcriptional regulation via chromatin remodeling 29,30 . Loss of EP300 by inactivating mutations may contribute to tumorigenesis in human cancers, including gastric cancers 31 . Caspase-8 is activated by its recruitment to Fas-mediated apoptosis 32 . Somatic mutations of caspase-8 can lead to attenuation of its proapoptotic function and contribute to gastric carcinogenesis 33 .
Here, we found that depletion of NKX6.3 could drive tumorigenesis in xenograft mouse model while tumor associated genes including p53, PI3K, E-cadherin, and EP300 were aberrantly expressed in these tumors. Consistent with results of NKX6.3 depleted cells, xenograft tumor also showed mutations in these genes (Figs 1, 4 and 5). Thus, NKX6.3 depletion might lead to multiple genetic mutations known to be closely associated with gastric tumorigenesis. APOBECs and AICDA can mutate a host's DNA. Significant numbers of APOBEC and AICDA-induced mutations have been observed in many types of human cancers [34][35][36][37][38] . Gastric cancer cells are highly enriched in mutation signatures [TC(A|T) → (T|G)] and [(A|T)(A|G)C → (T|G)] that are characteristic of a subclass of APOBEC and AICDA cytidine deaminases [39][40][41] . Two members of the APOBEC family, APOBEC3A and APOBEC3B, contribute substantially to mutations in cancers by deaminating cytosines in the TpCpW context [34][35][36][37]40,41 . It has been found that TpCpW mutations occur more frequently in gastric cancers with APOBEC expression 38 . APOBEC3B could activate genome-wide mutagenesis in various cancers [34][35][36][37]40 . In addition, aberrant expression of AICDA via NFκB activation causes accumulation of mutation in gastric epithelium with H. pylori infection 7 . AICDA-induced mutagenesis can lead to genome-wide alterations in the known preferred AICDA target sequence such as WRC motifs 39,42 . Moreover, CBFβ is required for APOBEC3 gene expression while NFκB is required for AICDA gene expression 7,13 . In the present study, we found that expression of CBFβ and NFκB was significantly increased in gastric cancer tissues while knockdown of CBFβ or NFκB dramatically inhibited the expression of APOBEC family and AICDA genes, diminished deaminase activity, and reduced the frequency of NKX6.3 depletion-induced mutations in TP53, PIK3CA, and POLA1 genes. Notably, binding of NKX6.3 to the promoter region of CBFβ and NFκB downregulated their expression while depletion of NKX6.3 dramatically increased expression of APOBEC gene family and AICDA genes at mRNA and protein levels, suggesting that NKX6.3 might function as a transcriptional repressor of these genes and downregulate deaminase activity in gastric epithelial cells (Fig. 2). Thus, NKX6.3 might inhibit the APOBEC-and AICDA-induced mutations by acting as a transcriptional repressor of APOBEC family and AICDA genes, thereby protecting gastric epithelial cells against genome-wide genetic mutations.
In non-neoplastic gastric mucosae and gastric cancer tissues, expression levels of CBFβ, NFκB p65, APOBEC family, and AICDA were increased in the cases with reduced or loss of NKX6.3 expression while NKX6.3 expression was inversely correlated with expression of these genes. In addition, expression levels of CDH1, CDKN1A, and EP300 were reduced while expression levels of RhoA, ROCK1, ROCK2, PIK3CA, and CCND1 were increased in gastric cancer tissues with reduced or loss of NKX6.3 expression (Fig. 6). Previously, we have reported that NKX6.3 expression is involved in gastric cancer progression and patients' survival 15 . Concordant with these results, gastric cancer patients with reduced or loss of NKX6.3 expression and/or high mutation rates had shorter overall survival time ( Supplementary Fig. S8E). In immunohistochemistry analysis, NKX6.3 expression was inversely associated with AICDA expression but positively associated with EP300 expression in 151 gastric cancer tissues (Fig. 6D). Thus, patients with reduced or loss of NKX6.3 expression and/or negative p53, negative (n = 151) by immunohistochemistry. (E) Kaplan-Meier curves for overall survival of patients (n = 151) with gastric cancer with expression levels of NKX6.3, p53, EP300, and AICDA proteins. EP300, and positive AICDA expression had shorter overall survival time (Fig. 6E), providing strong support for our hypothesis.
In conclusion, our study provides a new paradigm for the role of NKX6.3 in the pathogenesis of gastric cancer. We demonstrated that NKX6.3 depletion could lead to aberrant expression of AICDA and APOBEC family genes in gastric epithelial cells, thereby generating widespread genetic mutations and eventually driving the development and progression of gastric cancer (Fig. 6F). These findings unambiguously indicate that depletion of NKX6.3 in gastric epithelial cells can prompt accumulation of genetic mutations, providing important genetic evidence for the molecular mechanism involved in the development of gastric cancers.

Materials and Methods
Human Gastric Samples. A total of 65 frozen gastric cancers were obtained from Chonnam National University Hwasun Hospital supported by the Ministry of Health, Welfare and Family Affairs. In addition, a total of 55 patients with sporadic gastric cancer who underwent gastrectomy at Seoul St. Mary's Hospital were included. Fifty-five non-neoplastic gastric mucosae remote from the tumor (>5 cm) at each corresponding region were enrolled in this study. Informed consent was provided according to the Declaration of Helsinki. Written informed consent was obtained from all subjects. This study was approved by the Institutional Review Board of The Catholic University of Korea, College of Medicine (approval numer: MC16SISI0130). There was no evidence of familial cancer in any of these patients. For immunohistochemical analysis, tissue microarray recipient blocks were constructed, containing 151 gastric cancer tissues from formalin-fixed paraffin embedded specimens at Seoul St. Mary's Hospital. Three tissue cores from each cancer (2 mm in diameter) were taken and placed in a new recipient paraffin block using a commercially available microarray instrument (Beecher Instruments, Micro-Array Technologies, Silver Spring, MD, USA) according to established methods 15 . One cylinder of normal gastric mucosa adjacent to each tumor was also transferred to the recipient block. Sections were cut in thickness of 2 μm the day before use. They were stained according to standard protocols.
Other details regarding cell culture, transfection, whole genome sequencing, gene set enrichment analysis, detecting AICDA/APOBEC mutation pattern, in vitro deaminase assay, 3D-PCR, cloning, sequencing, validation detection, caspase 3/7 activity assay, motility and invasion assays, spheroid cell culture, in vivo xenograft mouse experiment, RT-qPCR and ChIP-qPCR, immunoblotting and immunofluorescence (IF), immunohistochemistry (IHC), construction of plasmids, transient reporter assay, bacterial strain and animal infection, datasets, and statistical analyses are available in Supplementary Materials and Methods.