Dual-Regulated Mechanism of EZH2 and KDM6A on SALL4 Modulates Tumor Progression via Wnt/β-Catenin Pathway in Gastric Cancer

SALL4 has been demonstrated in many cancers and participated in tumorigenesis and tumor progression, however, its expression and function still remain ambiguous in GC, especially its upstream mechanistic modulators. We explored whether the dual mediation of EZH2 and KDM6A could be involved in upstream regulation of SALL4, which promotes GC cell progression via the Wnt/β-catenin pathway. Analysis of discrepant gene expression in GC and normal gastric tissues from The Cancer Genome Atlas (TCGA) dataset. GC cell lines were transfected by siEZH2 and siKDM6A, the transduction molecules of KDM6A/EZH2-SALL4-β-catenin signaling were quantified in the GC cells. Here, we showed that only SALL4 levels of SALL family members were upregulated in nonpaired and paired GC tissues than those in corresponding normal tissues and were associated with its histological types, pathological stages, TNM stages including T stage (local invasion), N stage (lymph node metastasis), M stage (distant metastasis), and overall survival from the TCGA dataset. SALL4 level was elevated in GC cells compared to normal gastric epithelial cell line (GES-1) and was correlated to cancer cell progression and invasion through the Wnt/β-catenin pathway in GC, which levels would be separately upregulated or downregulated by KDM6A or EZH2. We first proposed and demonstrated that SALL4 promoted GC cell progression via the Wnt/β-catenin pathway, which was mediated by the dual regulation of EZH2 and KDM6A on SALL4. This mechanistic pathway in gastric cancer represents a novel targetable pathway.


Introduction
GC is the third most frequent cause of cancer-related death worldwide [1,2] and the fifth most common malignant cancer globally [3,4], especially in Asia [5,6]. The prevalence more dramatically increases in Eastern Asia which is the topmost one in the world [4,7]. Despite the continuous improvement of surgical techniques in China, including complete lymph node dissection (LDN) and R0 tumor resection, the endless novation of perioperative neoadjuvant chemoradiotherapy (NACT) regimens, and the increasing awareness of clinical therapy compliance in GC, the 5-year survival rate of GC still remains lower than 30% [8,9]. The long-term poor prognosis of GC is unsatisfactory due to the privation of specific early diagnostic biomarkers [10,11], as a result, the majority of newly diagnosed GC cases are at the advanced stage when symptoms appear. Oncologists worldwide still focus on the unvarying early-diagnosis and effective-therapy aspects of GC. At present, the novel study model on tumors should be varied from diagnosis and therapy methods to the molecular mechanism due to the unclear tumorigenesis. Contrarily, the explicit molecular mechanism will benefit the exploitation of the target points in clinical therapeutics. Therefore, it is vital to investigate and identify the genes implicated and the genetic alterations, involved in the potential molecules associated with GC phenotype.
The strong positive expression of SALL4 and the lose of trimethylation of Lys-27 in histone 3 (H3K27me3) in both rhabdomyosarcomatous elements and malignant neuroectodermal neoplasm are reported by immunohistochemistry (IHC), which highlight the potential molecular alterations such as CTNNB1 mutations [27]. However, H3K27me3, as a chromatin mark associated with gene transcriptional silencing, is catalyzed by the enhancer of zeste homolog 2 (EZH2), which is the catalytic subunit of the polycomb repressive complex 2 (PRC2) [28]. EZH2 is one of the most important histone methyltransferases (HMTs), which acts as an important epigenetic regulator repressing transcription. The catalytic activation of EZH2 necessitates two other PRC2 components, including embryonic ectoderm development (EED) and suppressor of zeste 12 protein homolog (SUZ12) [29]. H3K27me3 has been commonly identified as a transcriptionally repressive epigenetic event during tissue differentiation and development, which can drive tumorigenesis and tumor progression [30]. The β-catenin (encoded by CTNNB1)-dependent canonical Wnt ligands, including Wnt1, 2, 3, and 3a, have been implicated in cell growth, proliferation, invasion, and migration in numerous solid tumors. In the Wnt/β-catenin pathway, the abnormal regulation of β-catenin accounts for a pivotal component of the Wnt signaling pathway, which can lead to early carcinogenesis [31][32][33]. From the epigenetic regulation of human genome, the trimethylation of H3K27 can be completely reversed by KDM6A (UTX) demethylase. Furthermore, SALL4 expression is not activated or even silenced due to the retention of the repressive H3K27me3 in the case of KDM6A depletion [34]. Nevertheless, the synergistic function of EZH2 and KDM6A in SALL4 via H3K27me3 has not yet been reported in GC.
In this study, we aimed at investigating the upstream regulatory mechanism of SALL4, which included EZH2 and KDM6A hypothesized as the critical regulators in GC. SALL4 was first confirmed to promote GC cell progression via Wnt/β-catenin pathway, which was mediated by the dual regulation of EZH2 and KDM6A on SALL4. This mechanistic pathway in GC represents a novel targetable pathway.

RNA Extraction and Real Time Quantitative PCR (RT-qPCR)
RNA extraction and cDNA reverse amplification were, respectively, performed according to RNeasy plus mini kit (Qiagen), RQ1 RNase-Free DNase kit (Promega) and High-Capacity cDNA Reverse Transcription Kits (Thermo Fisher Scientific) [36], programed the thermal cycler using the conditions below: 25 °C (10 min), 37 °C (120 min), 85 °C (5 min), 4 °C. All real-time PCR reactions were done as triplicatesusing the LightCycler480, GAPDH was performed as the control. The sequences of specific primers are listed in Table 1.

Wound Healing Assay
Cancer cells were seeded into each well of the Culture-Insert 2-Well (ibidi, No: 80241, Culture-Insert 2-Well, Germany) [37], added mitomycin C (10 ug/mL) 2 h before removing wells. And then took images after 24 h by microscope connect AxioCam MRm camera. 10 4 /well of cancer cells were seeded as triplicate into 6-well plates and cultured for 72 h. Cells were counted every 12 h by a hemocytometer under a light microscope. Then, cell growth curves were plotted to assess cell growth. Cell viability assays were evaluated using MTT (Sigma-Aldrich) dye, absorbance at 570 nm was measured (Bio-Rad) every 12 h by spectrophotometry. For the FH535 and XAV939 (the specific inhibitors of β-catenin activity, were, respectively, obtained from Cell Signalling and MedChemExpress) experiments, XAV939 (10 μmol/L) [38] and FH535 (20 μmol/L) [39] were used to pretreat SGC-7901 and MGC-803 for 2 h prior to reseeding and culture.

Boyden-Chamber Assay
Transwell-24 units with 8uM pore-size membrane chamber (Corning, NY) was used for cell migration and invasion. 60ul of Matrigel (BD Biosciences, East Rutherford, NJ) was pre-coated in the upper chamber before seeding cells for 4 h. Seeded 5 × 10 4 cells in FBS-free medium in the upper chamber, after forty-eight hours, images were captured by an Axiovert Observer Z.1 microscope connected an AxioCam MRm camera.

Immunocytochemical Staining
For immunocytochemical staining, cells were seeded on the 13 mm coverslips in 24-well plate, cells were incubated with the primary antibody overnight (rabbit anti-SALL4 1:200, mouse monoclonal anti-E-Cadherin 1:300). Images were taken by Keyence microscope.

Bioinformatics Analysis
GSE79973 dataset was analyzed for expression distribution of genes in human GC and adjacent normal tissues, TCGA RNA-seq data in TPM (transcripts per million reads) format was used for differential expression analysis of the samples. RNA-seq data from TCGA GC project in HTSeq-FPKM format were used for the molecular correlation analysis, and Spearman's correlation analysis were performed using the ggplot2 (version 3.3.3) R package.

Statistics
All data were analyzed by GraphPad Prism 9 Software (GraphPad, San Diego, California, USA) software, and the data are presented as mean ± SD from the triplicate-repeated

Results
To clarify the upstream regulatory mechanism of SALL4, here, we firstly focused on the correlation between SALL4 and the biological characteristics of GC from the TCGA datasets, then evaluated SALL4 level in GC and GES-1 cell lines from gene and protein levels and its prognostic value from TCGA dataset. After confirming its phenotypic functions in GC cells, we identified the hypothesized dual regulation of EZH2 and KDM6A on SALL4 in GC and that SALL4 could promote the tumorigenicity of human GC cells via the Wnt/β-catenin signaling pathway.

SALL Level in GC Tissue and Its Correlation with TNM Stage from TCGA Dataset
To explore the expression level of SALL4 gene in GC, the TCGA dataset (GSE79973) was analyzed to evaluate the distribution of gene expression in GC and its adjacent normal gastric mucosal tissue. The results showed that SALL4 gene expression in GC was up-regulated in comparison with the adjacent non-tumor mucosal tissue (Fig. 1A). On this basis, the expression levels of SALL family members (including SALL1, SALL2, SALL3, and SALL4) in GC were also dissected. Firstly, it was found that only SALL4 among the SALL family members showed increased expression in unpaired and paired GC tissues than corresponding normal tissues in TCGA dataset (Fig. 1B). Simultaneously, compared with the normal tissues, only SALL4 expression elevated not only in various pathological types (Fig. 1C), but also different stages of GC (Fig. 1D). Consistently, with regard to T, N, and M stages, which represented the tumor local extension, spread-to-nearby lymph nodes and long-distance metastasis, respectively, only SALL4 was involved in the different statuses relative to the normal stages (Fig. 1E). These results suggested, unlike SALL2 exerted inhibitive roles in cell proliferation, migration, and invasion of solid cancer [41,42], SALL4 in the SALL family might serve as a biomarker frequently linked with GC to promote metastasis and local invasion, which deserved further investigation,

SALL4 Level in GC Cells and Its Prognostic Value from TCGA Dataset
Encouraged by the above results, to further confirm the SALL4 expression in GC cells, this work analyzed the SALL4 expression in four GC cell lines (HGC-27, AGS, SGC-7901, and MGC803) and normal GES-1 cell line by real-time RT-PCR ( Fig. 2A) and Western-blotting (WB) (Fig. 2B). It was discovered that, SALL4 levels were notably higher in AGS, SGC-7901, and MGC-803 cells than in GES-1 cells, especially in SGC-7901 and MGC-803 cells. Therefore, these two cell lines were selected for subsequent experiments. Thereafter, immunofluorescence staining verified that SALL4 showed a markedly stronger aggregation degree in the nuclei of SGC-7901 and MGC-803 cells than GES-1 cells (Fig. 2C). To facilitate the subsequent study on SALL4 phenotype, this work focused on its prognostic association analysis in TCGA. As a result, GC patients with SALL4-high expression illustrated a worse overall survival (OS), as revealed by Kaplan-Meier survival analysis (Fig. 3D). Similar result was observed in the pathological stages of GC, especially in advanced GC, but not in the early-stage one (Fig. 3E). In the further analysis stratified by TNM stage, OS remained significantly discrepant. For patients with N3 and T4 stage GC, SALL4-high expression predicted the shorter OS than SALL4-low expression ( Fig. 2G and F). These results indicated that the SALL4 level was more correlated with the OS of GC patients with the highest T and N stages, implying that SALL4 was involved in the metastasis and local invasion of GC.

Phenotypic Functions of SALL4 Gene in GC Cells
The high expression of SALL4 in GC cells and its prognostic survival analysis in patients with high SALL4 expression prompted us to focalize its phenotype. Therefore, this study investigated the transfection efficiency of SALL4 siRNA (siSALL4) in SGC-7901 and MGC-803 cells (Fig. 3A). Thereafter, the SALL4 expression was suppressed in SGC-7901 and MGC-803 cells by siSALL4. Compared with control groups, SALL4 silencing conspicuously repressed cancer cell growth (Fig. 3B-1# SGC-7901, 3B-2# MGC-803 cells, red vs. blue dots/lines). In addition, cancer cell migration was also affected after siSALL4 transfection (Fig. 3C-1# SGC-7901, 3C-2# MGC-803 cells). Furthermore, knockdown of SALL4 expression by siSALL4 was sufficient to inhibit the invasiveness of cancer cells, as revealed by a Boyden chamber assay (Fig. 3D). These results further supported the high-SALL4 expression in cancer cells compared with normal gastric mucosal cells. Besides, they bolstered our above results that SALL4 was possibly involved in the metastasis and local invasion of GC patients and was closely related to the tumor malignancy grade and long-term prognosis.

The Dual Regulation of EZH2 and KDM6A on SALL4 in GC Cells
To seek the potential upstream regulatory mechanisms of SALL4, the ARCHS4 dataset (https:// maaya nlab. cloud/ archs4/ gene/ SALL4) principally invested for mining the available RNA-seq Data was analyzed. According to the Z-score rating of human RNA-seq in the ARCHS4 dataset, the top 10 upstream regulatory factors for SALL4 were selected and predicted from the TCGA dataset, and their correlations with SALL4 mRNA expression in GC was also examined (Fig. 4A). According to our results, the FOXM1, KDM6A, SOX2, and CREB1 mRNA expression levels showed positive correlations with SALL4 expression in GC patients from TCGA dataset (Fig. 4B). To further confirm Inspirited by these data, this study focused on investigating the correlation between KDM6A and SALL4 in GC patients. Based on TCGA dataset analysis, our results indicated that KDM6A mRNA expression was up-regulated in both the unpaired (Fig. 4D-1#) and paired (Fig. 4D-2#) GC tissues compared with corresponding normal gastric tissues. Furthermore, the KDM6A mRNA expression was positively related to the SALL4 mRNA expression (Fig. 4E). To clarify KDM6A as the upstream regulator for SALL4, SALL4 expression was down-regulated after transfection with KDM6A siRNA (siKDM6A) in SGC-7901 and MGC-803 cells, as suggested by WB assay (Fig. 4F). The H3K27me3 level is dictated by the balanced regulation of EZH2 and KDM6A [43]. Notably, KDM6A is mainly involved in the demethylation of H3K27me3 [44], whose loss conduces to the activation of SALL4 [45]. Here, it was observed in this study that, SALL4 expression was down-regulated in the SGC-7901 cell line after siKDM6A transfection to silence KDM6A expression, while its expression was up-regulated after plasmid transfection to over-expression KDM6A expression, thus inducing the activation and suppression of H3K27me3 expression, separately (Fig. 4G). EZH2 is the methyltransferase responsible for catalyzing H3K27me3. In this regard, we hypothesized that EZH2 was involved in the SALL4 regulation on GC progression. To validate our hypothesis, it was found that, relative to GC patients with low-EZH2 expression, GC patients with high-EZH2 expression had a conspicuously longer OS in TCGA dataset (Fig. 4H), which was completely different from SALL4. Moreover, EZH2 silencing by EZH2 siRNA (siEZH2) suppressed H3K27me3 expression and increased SALL4 level in SGC-7901 cells (Fig. 4I). These results verified that KDM6A and EZH2 positively and negatively regulated SALL4 expression, respectively, via H3K27me3 in human GC cell lines.

SALL4 Promoted the Tumorigenicity of Human GC Cells via the Wnt/β-Catenin Signaling Pathway
It has been reported that SALL4 modulates the tumorigenicity of BC cells [46], cervical cancer cells [47] and esophageal squamous cell carcinoma (ESCC) cells [48], and induces acute myeloid leukemia in transgenic mice by the activation of Wnt/β-catenin pathway [49]. This study clarified that KDM6A and EZH2, respectively, the demethylase and methyltransferase for H3K27me3, were involved in the regulation of SALL4 in GC.
To further investigate whether these two upstream molecules of SALL4 and its downstream pathway were entangled in the canonical Wnt/β-catenin pathway in GC, here, we first found that the positive correlation between SALL4 and CTNNB1 (encoded β-catenin) was excavated in GC patients from TCGA dataset (Fig. 5A). Moreover, our WB results detected that the β-catenin expression was down-regulated after siSALL4 transfection in SGC-7901 and MGC-803 cells (Fig. 5B), consistent with the correlation analysis in GC patients from TCGA. Besides, β-catenin expression was dominantly decreased by the knockdown of EZH2 and KDM6A in the upstream of SALL4 (Fig. 5C). Hereby, we proposed that SALL4 modulated the evolution of GC cells, which was mediated by the upstream modulators including EZH2 and KDM6A via the activation of Wnt/β-catenin pathway (Fig. 5D). To verify the pertinence between the tumorigenesis and β-catenin in GC, the specific inhibitors of Wnt/β-catenin (FH535 and XAV939) were used, which stably accelerated the inhibition on β-catenin [39,47,50,51]. After validation of the repressed efficacy of FH535 and XAV939 in SGC-7901 (Fig. 5E) and MGC-803 (Fig. 5H) cells, it was found that the cell growth and viability of GC were significantly inhibited by FH535 and XAV939 (Fig. 5F, G, I, and J). These results indicated that KDM6A/EZH2-H3K27me3-SALL4-β-catenin might be the modulatory pathway for regulating GC cell progression.

Discussion
SALL4 expression is identified as the silencing state in completely differentiated cells, while its overexpression is reported in a large proportion of almost all solid tumor. It is widely suggested that SALL4 plays a certain role in carcinogenesis, including cell invasion and migration, cell proliferation and apoptosis, and cell stemness. This unique carcinogenic function demonstrates that SALL4 is involved in the potential correlation between pluripotency and cancer  [52]. More and more studies have reported that SALL4 has evolved into a key biomarker for the early screening of numerous many tumors, such as colorectal cancer [53], lung cancer [54], and breast cancer [55,56]. However, the expression and function of SALL4 in GC remain unclear.
The ultimate clinical prognoses of GC vary greatly depending on patients and the conventional pathological prognostic indicator (namely, TNM stage). Yang et al. [57] reported that SALL4 correlated the clinicopathological features with cancer progression in GC. Our study discovered that only SALL4 from the SALL family was up-regulated in unpaired and paired GC patients compared with the corresponding normal gastric tissues, which was related to the TNM stage and OS of patients from TCGA dataset, especially in T stage and N stage, suggesting that SALL4 was the potentially metastatic indicator in GC [58]. In the present study, our results also revealed that SALL4 was up-regulated in GC cells relative to normal gastric epithelium cell lines, which facilitated GC cell proliferation and cell invasion in vitro.
EZH2 and KDM6A are, respectively, the methyltransferases and demethylases for H3K27me3. The potential correlation of SALL4, H3K27me3, and KDM6A has been reported in oocytes, which presents that SALL4 can modify H3K27me3 by regulating one of the histone demethylase genes KDM6A [59]. Nonetheless, there are no available reports on how to associate the functional role of EZH2/ KDM6A with SALL4 in GC. Demethylation of the CG dinucleotide (CpG) region and up-regulation of SALL4 expression can be observed in myelodysplastic syndrome after treatment with hypomethylating agents, which predicts a worse prognosis [60]. Hypomethylation of the critical CpG region that regulates SALL4 expression can result in SALL4 up-regulation [61]. Heterochromatin binding marker (H3K27me3) is extensively distributed in the hypomethylated CpG region, whereas the polycomb repressive complex (EZH2/SUZ12) recognizing regions are diffusely concentrated in hypermethylated regions [62]. Additionally, H3K27me3 loss is also commonly associated with SALL4 activation [45], and over 60% of the SALL4-binding locus regions in extraembryonic endoderm (XEN) cells are either accompanied by H3K27me3 or by H3K27me3 privation [63,64]. Moreover, the H3K27me3 marks are present in the so-called bivalent domains in the promoter regions of SALL4 [65], and H3K27me3 loss can lead to the activation of bivalent promoters [45]. Our results revealed that KDM6A activated SALL4 expression through the H3K27me3 loss of KDM6A demethylation, which was reversed by KDM6A knockdown, while EZH2 inhibited SALL4 expression by the mobilization of H3K27me3. The H3K27me3-modified gene is especially interrelated with GC susceptibility and synergistically modulated by the two opposite-function enzymes (EZH2 and KDM6A) [66]. We also revealed that KDM6A and EZH2 exhibited the propensity to the inverse regulation of SALL4 via H3K27me3 in GC. SALL4-binding genes are associated with more than 30 diverse signaling pathways in promyelocytic leukemic cell lines, one of the most prominent pathways is WNT/β-catenin [67]. SALL4 level was positively related to tribbles pseudokinase 3 (TRIB3) in GC [57], which activates the Wnt/β-catenin pathway by directly binding with β-catenin in colorectal cancer [68]. However, the interaction domain between β-catenin and SALL4 was forejudged to be the C-terminal portion of SALL4 [63,69,70]. The activation of Wnt/β-catenin pathway by SALL4, which directly binds to CTNNB1 (a transcriptional cofactor in the Wnt/β-catenin signaling pathway), potently promotes cervical cancer development and progression [47]. Studies showed that Wnt/β-catenin pathway activation by SALL4 is related to enhance proliferation, survival, EMT, and metastasis, which are mediated by the increased levels of c-MYC and cyclin D1 [47,49]. SALL4 accelerates cell-cycle progression from G1 to the S-phase via the increment of β-catenin expression in breast cancer [46] and cervical cancer [47]. Moreover, the survival and progression of SALL4 up-regulation in GC are involved in a putative Wnt/β-catenin pathway [57]. Our results uncovered a positive correlation between CTNNB1 and SALL4 in GC cells, which were both positively regulated by KDM6A and negatively regulated by EZH2. Additionally, the growth and viability of GC cells also distinctly decreased after treatment with β-catenin inhibitors (FH535 and XAV939).
In summary, this study is the first to propose and demonstrate that SALL4 enhances the proliferation and growth of GC cells via the Wnt/β-catenin pathway, which is mediated by the dual regulation of EZH2 and KDM6A on SALL4.