Expression patterns of seven key genes, including β-catenin, Notch1, GATA6, CDX2, miR-34a, miR-181a and miR-93 in gastric cancer

Gastric cancer (GC) is one of the most prevalent cancers and a major cause of cancer related mortality worldwide. Incidence of GC is affected by various factors, including genetic and environmental factors. Despite extensive research has been done for molecular characterization of GC, it remains largely unknown. Therefore, further studies specially conducted among various ethnicities in different geographic locations, are required to know the precise molecular mechanisms leading to tumorigenesis and progression of GC. The expression patterns of seven candidate genes, including β-catenin, Notch1, GATA6, CDX2, miR-34a, miR-181a, and miR-93 were determined in 24 paired GC tissues and corresponding non-cancerous tissues by quantitative Real-Time PCR. The association between the expression of these genes and clinicopathologic factors were also investigated. Our results demonstrated that overall mRNA levels of GATA6 were significantly decreased in the tumor samples in comparison with the non-cancerous tissues (median fold change (FC) = 0.3143; P = 0.0003). Overall miR-93 levels were significantly increased in the tumor samples relative to the non-cancerous gastric tissues (FC = 2.441; P = 0.0002). β-catenin mRNA expression showed a strong positive correlation with miR-34a (r = 0.5784; P = 0.0031), and miR-181a (r = 0.5652; P = 0.004) expression. miR-34a and miR-181a expression showed a significant positive correlation (r = 0.4862; P = 0.016). Moreover, lower expression of Notch1 was related to distant metastasis in GC patients with a borderline statistical significance (p = 0.0549). These data may advance our understanding of the molecular biology that drives GC as well as provide potential targets for defining novel therapeutic strategies for GC treatment.


Results
Patients and tumor samples. Gastric carcinoma tissues and normal adjacent gastric tissues were obtained from 24 patients undergoing endoscopy for diagnostic purposes. Clinico-pathological data was shown in Table 1. Briefly, at the time of diagnosis, the age of patients (6 female/18 male) ranged from 48 to 89 years (mean 70.25 years). The gastric carcinomas were classified as intestinal (n = 15) and diffuse histological types (n = 9) according to Lauren system 57 . Intestinal type tumors were graded into poorly, moderately or well differentiated. Well-and moderately differentiated tumors were grouped together for purposes of statistical analysis. All diffuse cancers were classed as poorly differentiated. In the other words, 10 poorly, and 14 differentiated tumor tissues were included in our study. Given that the classification of tumor samples based on grading was rather similar to that of histological types (intestinal and diffuse types), our presented data related to altered gene expression in histological types reflects the changes also for two groups of cancer grades.

Expression patterns of candidate genes in gastric tumor and non-tumor tissues.
Individual samples of RNA were evaluated for the transcript levels of seven genes, including β-catenin, Notch1, GATA6, CDX2, miR-34a, miR-181a, and miR-93; and fold change of gene expression was calculated. Differential expression in tumor versus corresponding non-tumor tissues in each patient was shown as Figs. 1, 2, 3, 4, 5, 6 and 7 (left). Overall RNA expression levels in tumor samples compared to non-tumor samples were also presented in Figs. 1, 2, 3, 4, 5, 6 and 7 (right). The qRT-PCR analysis showed that overall mRNA levels of GATA6 were significantly decreased in the tumor samples relative to the adjacent non-cancerous tissues (median FC = 0.3143; P = 0.0003). Overall miR-93 levels were significantly increased in the tumor samples relative to the non-cancerous gastric tissues (median FC = 2.441; P = 0.0002). Overall RNA levels of β-catenin and miR-34a were decreased in the tumor samples in comparison with non-cancerous tissues (median FC = 0.535 and 0.915, respectively), but these differences were not statistically significant (P = 0.2068, P = 0.6714, respectively). In addition, overall RNA levels of Notch1, CDX2 and miR-181a were increased in tumor samples relative to non-cancerous tissues (median FC = 2.742, 3.47 and 1.5, respectively). However, these differences were not statistically significant (P = 0.3382, P = 0.3261, P = 0.1531, respectively) (Figs. 1, 2, 3, 4, 5, 6, 7 right).

Expression levels of candidate genes between intestinal-and diffuse-type GCs. The RNA
(mRNA or miRNA) expression of the candidate genes was compared between 15 intestinal-type and 9 diffusetype cancer tissues. As mentioned above, expression levels of GATA6 mRNA were significantly decreased in tumor samples relative to the corresponding non-tumor samples. Further analysis showed the frequent reduc- Table 1. Patients and histopathological characteristics of cancerous tissues. M male, F female, WD welldifferentiated, PD poorly differentiated, MD moderately differentiated, ND not determined. a Additional information was not available because of some reasons such as elderly patients rarely underwent surgery and some patients went to another city for medical care.  Intestinal  MD  IIIb  -T4  62/M  Diffuse  PD  II  ND   T5  63/M  Intestinal  MD  IV  +   T6  66/F  Diffuse  PD  ND a  -T8  79/F  Intestinal  MD  III  -T9  82/M  Intestinal  WD  ND  ND   T10  80/M  Intestinal  MD  IV  +   T11  www.nature.com/scientificreports/ tion of GATA6 expression in intestinal-type tumors than the diffuse-type tumors when compared to their corresponding non-tumor tissues (p = 0.0095 and p = 0.0142, respectively). However, significant statistical difference was not detected between these two cancer types in GATA6 expression (p = 0.715) (Fig. 8). A significant increased level of miR-93 expression was also observed in intestinal-and diffuse-type cancers relative to the corresponding non-tumor samples (p = 0.0.221 and p = 0.0.098, respectively) but found no statistically significant difference in miRNA level between the two histological types (p = 0.6005). Furthermore, expression levels of miR-34a were decreased specifically in diffuse-type cancers without significant statistical differences compared to the intestinal-type (p = 0.3105) and also to non-tumor samples (p = 0.5455). Additionally, expression levels of β-catenin, Notch1, CDX2, and miR-181a showed no significant difference between intestinal-and diffuse-type cancers (p = 0.8153, 0.3105, 0.3246, 0.6399, respectively), although the expression of Notch1 and CDX2 were somewhat up-regulated in diffuse-type cancer (Fig. 8).  www.nature.com/scientificreports/ Correlation between the expressions of candidate genes in GC. With regards to the concomitant expression of the seven RNAs in total tumor samples, β-catenin mRNA expression showed a strong positive correlation with miR-34a (r = 0.5784; P = 0.0031), and miR-181a (r = 0.5652; P = 0.004) expression ( Fig. 9). In addition, miR-34a and miR-181a expression showed a significant positive correlation (r = 0.4862; P = 0.016) (Fig. 9). Also, a positive correlation was detected between β-catenin and GATA6 expression. However, the statistical significance of this finding was borderline (r = 0.389; p = 0.0603) (Fig. 9). When we assessed the expression of the genes with significant relationships between two histological types, we found some minor differences in the correlation patterns. A significant positive correlation between β-catenin with miR-34a and miR-181a were detected only in diffuse-type cases (r = 0.8667; p = 0.0045, and r = 0.7167; p = 0.0369, respectively) ( Supplementary Fig. 1). Although β-catenin expression was also positively associated with miR-34a and miR-181a in intestinal-type cases, but these correlations were not statistically significant (r = 0.4629; p = 0.0838, and r = 0.4821; p = 0.0711) ( Supplementary Fig. 1). Furthermore, a remarkably positive correlation was detected between miR-34a and    Fig. 1), which was not statistically significant.

Association of the genes under investigation with distant metastasis.
We further investigated whether expression levels of these genes are associated with distant metastasis in GC. We compared the RNA expression between 9 patients with distant metastasis and 7 patients without metastasis. According to the qRT-PCR analysis, lower expression of Notch1 was detected in patients with distant metastasis. The difference of Notch1 expression between the two groups of patients (with and without metastasis) was borderline significant (p = 0.0549), thus, the lower expression level of Notch1 was relatively associated with distant metastasis in GC. The expression levels of other genes were not obviously associated with distant metastasis (Fig. 10).

Discussion
In this study, the expression profiles of seven genes, including β-catenin, Notch1, GATA6, CDX2, miR-34a, miR-181a, and miR-93 were determined in GC tissues and corresponding noncancerous tissues in order to find characteristic changes associated with pathogenesis and progression of GC. We compared the ratios of β-catenin RNA expression fold change in tumor tissue to that in corresponding non tumor tissue in each patient, and found that 29.1% of patients showed a decreased level (FC T/N < 0.5) of β-catenin expression, whereas, 12.5% of them showed an elevated level (FC T/N > 2) of expression (Fig. 1). An overall analysis showed lower level of β-catenin expression in tumor tissues relative to non-tumor tissues that was statistically non-significant (Fig. 1). With reference to previous studies, we found that transcription levels of β-catenin gene in GC tissues has not been convincingly investigated by RT-PCR. However, studies on abnormalities of β-catenin protein expression, specially using immunohistochemistry technique, are noticeable. In a study by Jawhari et al. 6 , loss of membranous β-catenin protein expression was reported in 58% of diffuse-type GCs and in 38% of intestinal-type GCs. In Ramesh et al. study, ectopic intracellular expression of β-catenin protein was relatively rare. Nonetheless, reduced or loss of membranous β-catenin protein expression was found in 83.4% of diffuse and in 28.6% of intestinal type cancers 58 . In another study, the quantitative analysis of β-catenin protein levels in GC samples versus their matched normal gastric mucosa by western blot did not reveal any significant difference 5 . However, in contrast with our findings, the study of Ebert et al. revealed that overall mRNA levels of β-catenin were significantly increased in GC samples. In addition, increased β-catenin mRNA levels were reported more frequent in intestinal-type GCs than diffuse-type GCs 4 . These inconsistence results may be due to different mechanisms that disturb β-catenin expression such as mutations of the β-catenin gene-although β-catenin gene mutations seems to be infrequent in GCs 4,59-61 -, mutations of APC gene 4,5 , or other components involved in Wnt pathway 59 , hypermethylation of the β-catenin promoter 5 , and hypermethylation of the APC promoter 62 .
We detected an increased level (FC T/N > 2) of Notch1 expression in 50% of patients and a decreased level (FC T/N < 0.5) of that in 29.1% patients (Fig. 2). However, we found no significant alteration in overall Notch1 expression between GC tissues and non-cancerous tissues (Fig. 2). Little is known about the dysregulated expression of Notch1 in GC tissues, as well as, its correlation with other genes and clinical features. Yeh et al. study showed that 63.3% of GC patients expressed Notch1 protein in cancer tissues. Furthermore, the activation of Notch1 signaling promoted tumor progression of stomach adenocarcinoma SC-M1 cells through induction of COX-2 (cyclooxygenase-2 (expression 11 . Another study reported that the activated form of Notch1 (N1IC) elevated the progression of several human GC cell lines through STAT3 and Twist expression 63 . N1IC also enhanced GC progression through miR-151-5p 64 .
Noteworthy, the results of our study suggested that down-regulation of Notch1 may be associated with the potential for distant metastases in GC. The statistical significance of this finding was borderline (p = 0.0549), possibly due to the small size of sample or extensive variations in expression levels of Notch1 between individual patients.
In our study, 58.3% of patients showed a decreased level (FC T/N < 0.5) of GATA6 expression, whereas, 4.1% of them showed an elevated expression level (FC T/N > 2) of that (Fig. 3). We found that overall mRNA levels of GATA6 were significantly decreased in the tumor samples in comparison with non-cancerous gastric tissue (P = 0.0003) (Fig. 3).  www.nature.com/scientificreports/ In contrast with our findings, Sulahian et al. reported that GATA6 gene is amplified or overexpressed in gastric, esophageal and pancreatic adenocarcinomas. They further found that depletion of GATA6 impairs gastrointestinal cancer cell growth and induces cell cycle arrest in G2/ M phase 66 . Thus, the exact role of GATA6 in gastric carcinogenesis has remained controversial due to the limited and inconsistent reports.
We determined an elevated level (FC T/N > 2) of CDX2 expression in 52.1% of GC patients and a decreased level (FC T/N < 0.5) of that in 26% of patients (Fig. 4). Furthermore, significant differences were no observed in increased level of CDX2 mRNA between total GC tissues and non-cancerous tissues (Fig. 4). The potential roles of CDX2 in the gastric carcinogenesis and progression are also complex and remain unclear. In spite of some conflicting studies 29,67,68 , an overall experimental insight provided from previous studies is that CDX2 acts as tumor suppressor in GC 31,32 . Kim et al. showed that CDX2 mRNA expression was considerably higher in gastric tumor tissues than non-tumor tissues due to DNA hypomethylation. Consistent with our result, they also found no statistically significant difference in CDX2 mRNA levels between two GC types 69 . As well as, in Almedia et al. 28 study, CDX2 protein expression was not detected in the nuclei of normal mucosa cells adjacent to gastric carcinomas, while that was observed in 54% of cases, regardless of GC histological types. However, inappropriate activation of CDX2, as an intestine-specific gene, in gastric may be involved in intestinal differentiation, a pathway towards the gastric carcinogenesis 28,67,68 . Thus, due to inconsistent results, the precise role of CDX2 overexpression in gastric carcinogenesis and malignancy remains to be further clarified.
In our study, a decreased level (FC T/N < 0.5) of miR-34a expression was detected in 25% of GC patients and an increased level (FC T/N > 2) of that was observed in 8.3% of patients (Fig. 5). Although decreased level of miR-34a was not statistically significant between two GC types, this reduction was more frequent in diffusetype cases than intestinal-type cases (median miR-34a expression FCs were 0.237 and 1.064, respectively for diffuse-and intestinal-type) (Fig. 8). According to qRT-PCR analysis, numerous studies reported that the levels of miR-34a expression were significantly decreased in the GC patients 35,70 . Meanwhile, miR-34a expression levels were detected lower in patients with metastasis than in patients without metastasis 35 . However, there are controversial studies that reported a significant up-regulation of miR-34a level in GC tissues compared to normal gastric tissues 71,72 . It is noteworthy that these studies used the microarray analysis to determine the expression profile of miRNA in GC and normal tissues without further validation of miR-34a expression by qRT-PCR.
We detected an increased level (FC T/N > 2) of miR-181a expression in 33.3% of patients and a decreased level (FC T/N < 0.5) of that in 8.3% patients (Fig. 6). The overall difference in miR-181a expression was not statistically significant between GC tissues and non-cancerous tissues (Fig. 6). Indeed, miR-181a-5p was found as an onco-miRNA promoting cell proliferation, metastasis, invasion and EMT in GC cell lines 47,[73][74][75] . Although there are numerous studies reporting the up-regulated expression of miR-181a-5p in GC tissues 48,75,76 , in controversial study, Lin et al. reported that the expression of miR-181a in GC tissues was significantly lower than in adjacent tissues. Their results suggested that miR-181a acts as a tumor suppressor and its down-regulation may involve in the progression and metastasis of GC 77 . Therefore, the molecular mechanisms by which miR-181a mediate the pathogenesis of GC still need to be further elucidated.
In the present study, an elevated level (FC T/N > 2) of miR-93 expression was detected in 63.6% of patients (Fig. 7). Overall miR-93 levels were strongly increased in the tumor samples relative to the non-cancerous gastric tissues (P = 0.0002) (Fig. 7). Several studies reported the higher expression of miR-93 in GC tissues compared with the noncancerous tissues, suggesting that miR-93 functions as a promoter for tumor progression in GC patients. In vitro and in vivo studies confirmed that miR-93 plays an oncogenic role in GC [78][79][80] . However, Stanitz et al. found no significant difference in the expression levels of miR-93 between GC tumor and normal tissues in the populations that they studied 81 .
The evaluation of correlation between different genes is important to explore the mechanisms leading to GC. We explored a strong positive correlation between β-catenin mRNA and miR-34a expression (Fig. 9). miRNAs regulate the expression of genes at the post-transcriptional and translational levels frequently by binding to complementary sequences in the 3′-UTRs of target mRNAs 79 . β-catenin, encoded by CTNNB1, is a predicted target for miR-34a in the databases such as miRTarbase (https ://mirta rbase .mbc.nctu.edu.tw), mirDIP (https ://ophid .utoro nto.ca/mirDI P/), and TargetMinter (https ://www.isica l.ac.in/~bioin fo_miu/targe tmine r). The potential binding sites for miR-34a were located in the 3′-UTR of β-catenin mRNA ( Supplementary Fig. 2) suggesting that β-catenin is a putative target of miRNA-34a. Furthermore, several studies experimentally demonstrated that β-catenin is a direct target of miR-34a which inversely affects the β-catenin mRNA and protein levels, as verified through luciferase reporter assay, immunoblot analysis and qRT-PCR [82][83][84] . Thus, our finding related to the positive correlation between miR-34a and β-catenin mRNA expression in GC tissues was in contrast to our primary assumption on the basis of previous studies. With further literature review, we found a study reporting that miR-34a expression was directly regulated by β-catenin and significantly induced by the overactivation of β-catenin signaling in mouse tumors and hepatocellular carcinoma patients 37 . However, a positive association between miR-34a and β-catenin, due to well-known oncogenic function of β-catenin as a co-transcriptional factor in Wnt/β-catenin signaling pathway, and common tumor suppressor function of miR-34a in majority of cancers, seems to be related to a little known mechanism, probably independent of p53 pathway. Nonetheless, we speculate that oncogenic function of miR-34a remains of matter of debate and needs to be clarified through further investigation.
Furthermore, we observed that the expression of β-catenin mRNA was positively correlated with the expression of miR-181a in GC tissues (Fig. 9). Consistent with this observation, we found a study by Ji et al. that reported a positive correlation between β-catenin expression and miR-181 family members in HCC (hepatocellular carcinoma) cell lines. In addition, they found that forced expression of β-catenin or Tcf4-a co-transcriptional activator of β-catenin-induced miR-181 expression 85  www.nature.com/scientificreports/ Moreover, we detected a concordant expression of miR-181a and miR-34a in GC tissues (Fig. 9), this may be due to the fact that both miR-34a and miR-181-a are transcriptional target of β-catenin. However, the confirmation of this finding needs to further research.
In previous reports, Notch1 was confirmed as direct target of miR-34a 90,91 . In this research, the inverse correlation between the RNA expression of miR-34a and Notch1 was not considerable (r = −0.1666; p = 0.4365). Furthermore, it was reported that miR-181a directly targets GATA6 and CDX2 92 . In our study, the expression of miR-181a was not noticeably correlated with that of GATA6 (r = 0.0591; p = 0.7836) and CDX2 (r = −0.2638; p = 0.2238). To elucidate the aforementioned findings, we note two important points in the following. Firstly, miRNAs as well as their targets could be cancer-specific. In support of this notion, Shi et al. 36 suggested that the targets of the same miRNA in several cell lines of glioma may be different because each type of cell is likely to have a specific miRNA milieu for regulation of gene expression. Secondly, microRNAs directly suppress their target gene expression via mRNA degradation or translational repression 93 . Thus, to understand the precise roles of microRNAs in different malignancies, the evaluation of alterations at target protein level is beneficial in addition to that at mRNA expression level.
In the present study, there were no age-dependent differences in the expression patterns of seven candidate genes (data not shown). The majority of patients (22 of 24) included in the current study, were at TNM stage of III and IV. This prevented us from further investigation of possible differences in gene expression levels between early-and advanced stages. Moreover, no significant association was identified between the gene expression and clinical features of GC, including distant metastasis or between two intestinal-and diffuse-GC types (Figs. 8, 10). One explanation would be due to the extensive variations in expression levels between each individual patient (see Figs. 1, 2, 3, 4, 5, 6, 7 left).

conclusions
Despite the abundant research conducted to identify the molecular pathways implicated in GC pathogenesis, it remains largely unknown. The majority of GC patients are diagnosed in advanced stages due to poor prognosis of the disease. Thus, limited treatment options can be helpful for cure or improving survival of them. Therefore, there is an urgent need for robust treatment method. In this regard, the conduction of studies at molecular level would be more beneficial. However, it seems that complex nature of cancer gene-expression as well as the heterogeneity existing among the cancer subtypes or even between each individual patient are major obstacles affecting to achieve a better health outcome from the commonly used treatment procedures or methods that target a specific gene transcript/protein. On the other hand, the development of personalized medicine or targeting of the genes with prevalent dysregulation may lead to improve the health outcomes for individual patients after treatment. Nonetheless, further studies are required to identify the most prevalent molecular targets in GC patients.
Our study revealed that GATA6 expression was frequently decreased in GC patients. In addition, miR-93 expression was frequently increased in GC patients. We did not detect any correlation between the expression level of GATA6 and miR-93 (data not shown) implying these genes may affect on GC pathogenesis through two distinct signaling pathways. Notably, we determined a considerable positive correlation between β-catenin mRNA and miR-34a expression in GC tissues. Despite β-catenin was proved as a direct target of miR-34a, this finding may be attributed to the fact that miR-34a involves in a little known pathway that its induction occurs by β-catenin activity.
Overall, although the data presented in our study still needs to be proved by further studies, it may provide potential targets for the exploration of novel therapeutic strategies for GC treatment.

Methods
Collection of tissue samples. Twenty-four pairs of GC tissues and corresponding adjacent noncancerous gastric tissues (48 samples) were obtained from untreated patients who underwent routine endoscopy for diagnostic purposes at the following institutions placed in Sari, Iran: Tuba Clinic, Maziar Clinic, Imam Hospital, and Shafa Hospital. Biopsies were immediately placed in Fix RNA reagent (EURx, E0280, Gdańsk, Poland) and stored at 4 °C until RNA extraction that it was performed in less than one week. The clinical and histopathological parameters, such as gender, age, histological type, grade, and pathological stage were determined according to the medical reports of the patients and are summarized in Table 1. The study was approved by the Ethics Committee of the Mazandaran University of Medical Sciences and performed in accordance with the relevant guidelines. Informed consent was also obtained from patients prior to the study.

RNA extraction and reverse transcription (RT).
Total RNA was extracted from tissue samples using Accuzol reagent (Bioneer, K-3090, Republic of Korea) according to the protocol. The concentration of total RNA was measured by UV spectrophotometry using a PicoDrop instrument. The extracted RNA was treated with RNase-free DNase I (Thermo Scientific, #EN0521) as described in the product manual to remove the possible contamination with genomic DNA.
Complementary DNA (cDNA) was synthesized from ~ 1 µg of DNase I-treated total RNA for each RT reaction using RevertAid First Strand cDNA synthesis kit (Thermo Scientific, #K1622) following the manufacturer's instructions. Stem-loop RT primers were designed on the basis of Chen et al. study 94 and used for the specific cDNA synthesis of miRNAs. Reverse transcription of U6 small non coding RNA-as internal control for the normalization of miRNA expression-was also performed using a specific primer taken from a published study 95