Epigenetic silencing of DACH1 induces the invasion and metastasis of gastric cancer by activating TGF-β signalling

Gastric cancer (GC) is the fourth most common malignancy in males and the fifth most common malignancy in females worldwide. DACH1 is frequently methylated in hepatic and colorectal cancer. To further understand the regulation and mechanism of DACH1 in GC, eight GC cell lines, eight cases of normal gastric mucosa, 98 cases of primary GC and 50 cases of adjacent non-tumour tissues were examined. Methylation-specific PCR, western blot, transwell assay and xenograft mice were used in this study. Loss of DACH1 expression correlated with promoter region methylation in GC cells, and re-expression was induced by 5-Aza-2′-deoxyazacytidine. DACH1 is methylated in 63.3% (62/98) of primary GC and 38% (19/50) of adjacent non-tumour tissues, while no methylation was found in normal gastric mucosa. Methylation of DACH1 correlated with reduced expression of DACH1 (P < 0.01), late tumour stage (stage III/IV) (P < 0.01) and lymph node metastasis (P < 0.05). DACH1 expression inhibited epithelial–mesenchymal transition and metastasis by inhibiting transforming growth factor (TGF)-β signalling and suppressed GC cell proliferation through inducing G2/M phase arrest. The tumour size is smaller in DACH1-expressed BGC823 cell xenograft mice than in unexpressed group (P < 0.01). Restoration of DACH1 expression also sensitized GC cells to docetaxel. These studies suggest that DACH1 is frequently methylated in human GC and expression of DACH1 was controlled by promoter region methylation. DACH1 suppresses GC proliferation, invasion and metastasis by inhibiting TGF-β signalling pathways both in vitro and in vivo. Epigenetic silencing DACH1 may induce GC cells' resistance to docetaxel.


Introduction
Gastric cancer (GC) is one of the most common malignancies, ranking fourth in males and fifth in females, respectively, and is a leading cause of cancer-related death worldwide [1]. Lacking early detection markers and effective therapeutic strategies, GC is usually diagnosed in late stages. Most of GC patients die of recurrence and metastasis, with a poor 5-year survival [2,3]. Dysregulation of normal signalling pathways is part of the transformation process in malignancy. Transforming growth factor (TGF)-b signalling functions in several biological processes, including cell proliferation, differentiation, migration and apoptosis. It plays important but paradoxical roles in carcinogenesis and cancer progression. TGF-b signalling promotes the invasion and metastasis by induction of epithelial-mesenchymal transition (EMT) in the later stages of cancer [4][5][6][7][8]. The EMT phenotype is characterized by loss of epithelial marker (E-cadherin) and increasing of mesenchymal factor (vimentin) expression, during which epithelial cells acquire mesenchymal properties while losing cell-cell interactions and apicobasal polarity [9,10]. EMT plays an important role in tumour metastasis and invasion and TGF-b signalling was reported to promote GC invasion and metastasis [11][12][13][14].
Dachshund homolog 1 (DACH1), a homologue of Drosophila dac in human, is located in chromosome 13q22. Dac protein contains two conserved domains: dachbox-N and dachbox-C, both of which are highly conserved from drosophila to humans. DACH1 is expressed extensively in normal tissues and loss of DACH1 expression was reported related to poor prognosis in breast, endometrial and prostate cancers [15][16][17]. DACH1 has been shown to inhibit TGF-b signalling in breast and colorectal cancer [18,19].
In this study, we examined whether epigenetic changes in DACH1 occurred in GCs and explored the role of DACH1 in tumour growth, invasion, metastasis and chemosensitivity in human GC.

Primary human GC samples and cell lines
Ninety-eight cases of primary GC and eight cases of normal gastric mucosa were collected as fresh frozen tissue from Chinese PLA General Hospital. Gastric cancer was classified by TNM stage, including stage I (N = 4), II (N = 8), III (N = 26) and IV (N = 60). Among 98 cancer samples, 32 cases of paraffin blocks are available with matched adjacent tissue. Eight cases of normal gastric mucosa were collected by biopsy under endoscopy from non-cancer patients. All samples were collected under the approved guidelines of the Chinese PLA General Hospital's institutional review board.
Bisulphite modification, methylation specific PCR (MSP) and bisulfite sequencing (BSSQ) Genomic DNA from GC cell lines and GC tissue samples were prepared by proteinase-K method. MSP and BSSQ were performed as described previously [22,23]. MSP primers and BSSQ primers was designed according to genomic sequences around transcription start site in the CpG island of DACH1 gene (NM_080759.4) promoter region and synthesized (BGI, Beijing, china) to detect unmethylated (U) and methylated (M) alleles [19].

Plasmid construction
The expression vectors for DACH1 wild-type or mutant type (DS and DDS) were gift from Dr. Cvekl. Reporter constructs, including SBE-4 Luc, which contains Smad-binding elements, were described as previously [18]. DACH1 was subcloned into plenti6-GFP lentivirus expression vector, and DACH1 expression lentiviral or empty vectors were packaged by using ViraPowerTM lentiviral expression system (Invitrogen) to infect BGC823 and AGS cell lines to establish stable expression cells. Lipofectamine 2000 (Invitrogen) was used for plasmid transfection. All constructs were confirmed by sequencing.

Dual-Luciferase reporter assay
BGC823 and AGS cells were seeded at 1 9 10 4 /well in 96-well plates and incubated for 24 hrs. The transfection was described previously [19]. Relative luciferase activities were measured with the Dual Luciferase Reporter Assay system (Promega, Shanghai, China) according to the manufacturer's protocol. For each experiment, the luciferase reporter assay was repeated three times.

Transwell migration assay
The effect of DACH1 on cell migration was detected by using the COSTAR transwell (Corning Incorporated, Beijing, China). DACH1- unexpressed and -re-expressed BGC823 and AGS cells were harvested and suspended in the serum-free medium. Cell suspensions were then placed into the upper well at a concentration of 2 9 10 4 cells/200 ll, while the complete medium with 10% FBS was placed into the lower well (500 ll). The chamber was incubated for 16 hrs. The cells still on the upper surface were scraped gently and washed out with PBS three times. The cells that migrated to the lower surface of the membrane were stained with crystal violet and counted in three independent highpower fields (9200).
Cell invasion assay 2 9 10 4 DACH1 -unexpressed and -re-expressed BGC823 and AGS cells were suspended in 200 ll of serum-free medium and loaded onto the upper compartment of an invasion chamber containing a polycarbonate membrane with an 8 lm pore size, which was coated with a layer of extracellular matrix (ECM; Matrigel TM , BD, Beijing, China). After 24 hrs of incubation, the invasive cells that migrated through the ECM layer to the complete medium in the lower compartment were stained with crystal violet and the number of invaded cells was counted in three independent high-power fields (9200).

Colony formation and cell viability detection
DACH1 unexpressed and stably expressed cells (1000 cells/well) were plated in 2 ml complete growth medium. The medium and reagents were changed every 48 hrs. Two weeks later, cells were fixed with 75% ethanol for 30 min. and stained with 0.2% crystal violet (Beyotime, Nanjing, China) for 20 min. and counted. The experiment was repeated three times. DACH1 unexpressed and stably expressed cells were seeded onto 96-well plates (3 9 10 3 cells/well), cell viability was measured daily by the Cell Counting Kit-8 (CCK8 kit; Dojindo Laboratories, Gaithersburg, MD, USA) for 4 days following the instruction of manufacturer.

Flow cytometry analysis
DACH1 -unexpressed and -re-expressed BGC823 and AGS cells were starved 12 hrs for synchronization, the cells were re-stimulated with 10% FBS for 24 hrs. The cells were treated by Cell Cycle Detection Kit (KeyGen Biotech, Nanjing, China) following the instruction of manufac- In vivo tumourigenicity DACH1 unexpressed and stably expressed BGC823 cells (3 9 10 6 cells in 0.2 ml phosphate-buffered saline) were inoculated subcutaneously into the dorsal right flank of 4-week-old male Balb/c nude mice (n = 7) respectively. Tumour volume was assessed every 2 days for 3 weeks since 5 days after implantation. Tumour volume was calculated according to the following formula: V = L 9 W 2 /2 where V, volume (mm 3 ); L, biggest diameter (mm); W, smallest diameter (mm). All procedures were approved by the Animal Ethics Committee of the Chinese PLA General Hospital.
Statistical Analysis SPSS 15.0 software was employed and all data are presented as means AE SD of at least three independent experiments. Chi-squared test or Fisher's exact test was used to analyse the association of clinical factors and the methylation status and the association of DACH1 expression and methylation status. Student's t-test was employed to compare two means. Statistical differences are presented at probability levels of P < 0.05 and P < 0.01.

DACH1 was silenced by promoter region hypermethylation in GC
DACH1 was found frequently methylated in hepatocellular carcinoma and colorectal cancer in our previous study [19,24]. To determine whether this change occurred in GC, we first analysed epigenetic changes and DACH1 expression in seven human GC cell lines, one immortalized normal gastric epithelia cell (GES-1) and one case of normal gastric mucosa. Loss of DACH1 expression and complete methylation was found in BGC823, AGS, SGC7901 and MGC803 cells. Partial methylation was found in NCI-N87, MKN45 and NUGC3 cells, and reduced expression was found in NCI-N87 cells. The normal gastric epithelial cell line GES-1 and the normal gastric mucosa were found to be completely unmethylated and to express DACH1 ( Fig. 1A and B). To further validate the specificity of MSP primers and the methylation density in promoter region, BSSQ was performed (Fig. 1C). Dense methylation was found in BGC823 cells, partial methylation was detected in NCI-N87 and a completely unmethylated promoter was found in normal gastric mucosa, confirming the findings obtained by using MSP. Loss of DACH1 expression correlated with promoter region hypermethylation. To confirm that DACH1 expression was directly regulated by promoter region methylation, MKN45, AGS, NUGC3, N87 and BGC823 cells were treated with the demethylating agent 5-AZA. Re-expression of DACH1 was found in BGC823 and AGS cell lines (originally without expression) and increased expression was found in NCI-N87 and MKN45 cell lines (partially methylated), while no expression change was found in NUGC3, which already expresses DACH1 (Fig. 1D). These results suggest that DACH1 expression was repressed by promoter region hypermethylation in most of the GC cell lines.
DACH1 was frequently methylated in human primary GC and promoter region hypermethylation was associated with reduction of DACH1 expression To examine DNA methylation of DACH1 in GC, 98 cases of primary GC samples, eight cases of normal gastric mucosa and 50 cases of adjacent non-tumour tissues were examined by MSP ( Fig. 2A). Methylation was found in 63.3% (62/98) GC and 38% (19/50) adjacent non-tumour tissues. This methylation frequency was significantly higher in GC samples than in adjacent nontumour tissues (P < 0.01), and no methylation was found in nor- mal gastric mucosa from patients without cancer. The presence of methylation in some cases of adjacent normal appearing tissue suggests that DACH1 methylation is an early event of GC. As shown in Table 1, DACH1 methylation was significantly associated with late tumour stage (stage III/IV; P < 0.01) and lymph node metastasis (P < 0.05), but no association was found between DACH1 methylation and age, gender, differentiation, intravascular cancerous emboli or tumour size.
The association of promoter region hypermethylation and DACH1 expression was analysed in 32 cases with available matched GC and adjacent tissue samples. DACH1 staining was observed in both the nucleus and cytoplasm. The expression of DACH1 was reduced significantly in cancer tissue compared with the adjacent tissue (P < 0.01, Fig. 2B and C). Among 20 cancer cases with lost/reduced DACH1 expression, 18 cases were methylated (90%). In contrast, among 12 cases of DACH1-expressed cancer tissue samples, only five cases were methylated (41.67%). Loss/reduction of DACH1 expression was significantly associated with promoter region hypermethylation in the tumours (P < 0.01, Fig. 2D). Despite the presence of methylation in some adjacent normal tissues, DACH1 expression was observed in the majority of normal appearing cells, suggesting that DACH1 methylation, detected with the sensitive MSP approach, was present in the minority of normal appearing cells and clonal progression, with methylation present in all cells within transformed tissues, was not yet present in these normal areas. Our results comparing DNA methylation with protein expression suggest that DACH1 expression is repressed by promoter region hypermethylation in most primary GCs.

Restoration of DACH1 expression inhibits TGF-b signalling in GC
DACH1 protein contains two domains: DachBox-N and DachBox-C. Human DachBox-N shares approximately 28% amino acid identity with the SKI/SNO proteins, inhibitors of TGF-b signalling. Thus, the DachBox-N domain is also known as SKI/SNO (DS) domain [25,26]. TGF-b signalling was inhibited by DACH1 in breast and colorectal cancer by DS domain [18,19]. The function of the full-length DACH1, DS (DACH1 fragment, which encodes Smad-binding proteins' domain) and DDS (fragment of DACH1 deleted DS) was analysed by Dual-Luciferase reporter assay. SBE4 luciferase activity was inhibited by full-length DACH1 and DS significantly in both BGC823 and AGS cells (P < 0.05, Fig. 3A), but the activity of SBE4 has no changes before and after DDS transfection (P > 0.05). The results demonstrate that DS domain is critical for DACH1 function in TGF-b signalling in GC.
To further characterize the role of DACH1 in TGF-b signalling, phosphorylated SMAD2 and phosphorylated SMAD3 were examined by western blot in DACH1 -unexpressed and -re-expressed BGC823 and AGS cells. The TGF-b induced phosphorylation of SMAD2 was reduced in DACH1 expressing cells, while no change was found in phosphorylated SMAD3 after restoration of DACH1 expression (Fig. 3B). The inhibition function in TGF-b signalling was further confirmed by knocking down DACH1 in GES-1 cells (Fig. 3C) where DACH1 knock-down was associated with an increase in TGF-b-induced phosphorylation of SMAD2. These results suggest that DACH1 inhibits TGF-b signalling in GC through inhibition of SMAD2 phosphorylation.

DACH1 suppresses cell invasion and migration by inhibiting EMT caused by TGF-b signalling in GC
Epithelial-mesenchymal transition is related to cancer invasion and metastasis [27,28]. TGF-b1 may induce EMT by down-regulating Ecadherin and up-regulating vimentin in GC cells [29]. To see the role of DACH1 in cell migration and invasion, a transwell assay was carried out with BGC823 and AGS cells. The number of migrated BGC823 cells under each field of microscope was 48.67 AE 4.04, 77.67 AE 3.51 and 34.33 AE 5.86 in empty vector group, empty vector adding TGF-b1 group and re-expression of DACH1 adding TGF-b1 group respectively. The number of migrated AGS cells under each microscope was 134 AE 7.55, 208.67 AE 8.14 and 87.67 AE 7.02 in empty vector group, empty vector adding TGF-b1 group and re-expression of DACH1 adding TGF-b1 group respectively. Thus, an increase in migrating cells was found with TGF-b1 treatment compared with empty vector (P < 0.01), but less migration cells were found when DACH1 expression was restored, even when cells also received TGF-b1 treatment (P < 0.01, Fig. 4A). The results suggest that cell migration was promoted by TGF-b1 and was suppressed by DACH1 in BGC823 and AGS cells. As shown in Figure 4B, similar results were obtained by using transwell matrigel experiment, which mimics invasion. TGF-b1 promotes cell invasion and DACH1 suppresses cell invasion in BGC823 and AGS cells (P < 0.01, Fig. 4B). To see the effect of DACH1 on EMT, E-cadherin and vimentin, EMT-related markers, were examined by western blot in DACH1 -unexpressed and -re-expressed BGC823 and AGS cells. Expression of E-cadherin was up-regulated and vimentin was down-regulated in DACH1-expressed BGC823 and AGS cells (Fig. 4C) compared with the non-expressing parental cell lines. As a family of zinc-containing endopeptidases, matrix metalloproteinases (MMPs) are the major enzymes involved in ECM degradation [30,31]. Two representative members, MMP-2 and MMP-9, were reported highly expressed in invasive tumours [32,33]. The expression of MMP-2 and MMP-9 was reduced after re-expression of DACH1 in GC cells, and this relationship was confirmed by knocking down DACH1 in GES-1 cells (Fig. 4D), which show an increase in MMP2 and MMP9 with reduction of DACH1 expression.

Restoration of DACH1 expression inhibits GC cell proliferation and induces G2/M arrest
To evaluate the effect of DACH1 on GC cell proliferation, colony formation and cell viability were examined. As shown in Figure 5A  2.08 AE 0.19 versus 1.32 AE 0.17 in AGS cells (P < 0.05) before and after re-expression of DACH1 (Fig. 5B), with a reduction of viability by 37-50% after re-expression of DACH1 in GC cells. These results suggest that DACH1 inhibits GC cell proliferation. To examine the way in which this occurred through cell cycle alterations, flow cytometry technique was employed. The cell phase distribution of DACH1 -unexpressed and -re-expressed BGC823 cells was as follows: 60.22 AE 1.41% versus 48.85 AE 1.37% for G0/1 phase (P < 0.01), 28.17 AE 2.2% versus 23.33 AE 1.2% for S phase (P < 0.05) and 11.6 AE 0.8% versus 27.83 AE 2.05% for G2/M phase (P < 0.01). The cell phase distribution of DACH1 -unexpressed and -re-expressed AGS cells was as follows: 59.03 AE 1.65% versus 47.11 AE 1.12% for G0/1 phase (P < 0.01), 25.66 AE 0.68% versus 23.32 AE 0.33% for S phase (P < 0.05) and 15.31 AE 1.12% versus 29.53 AE 0.85% for G2/M phase (P < 0.01). For both cell lines, G0/1 and S phase were reduced and G2/M phase was increased significantly in BGC823 and AGS cells (Fig. 5C), suggesting induction of a G2/M arrest. To further validate the effect of DACH1 on cell cycle, the expression of cdc2 and cyclinB1, the mitosis initiators, was examined by western blot in BGC823 and AGS cells (Fig. 5D). Both cdc2 and cyclinB1 were decreased after reexpression of DACH1. These results further demonstrated the G2/M phase arrest induced by DACH1 in GC.

DACH1 sensitizes GC cells to docetaxel
G2/M arrest was induced by DACH1 in BGC823 and AGS cells. Docetaxel is a cytotoxic reagent that causes microtubule poly-  merization-stabilization by binding to the tubulin. This, then suppresses dynamic properties at the mitotic spindle and induces cancer cell mitotic arrest (G2/M arrest) and apoptosis. To determine whether there was an interaction between DACH1 expression and docetaxel, cell viability was analysed by CCK-8 kit in DACH1 unexpressed and expressed GC cells. As shown in Figure 6, the IC50 of docetaxel in DACH1 -unexpressed and -re-expressed BGC823 cells was 4.0343 AE 0.6241 versus 7.1243 AE 1.0073 (P < 0.05, Fig. 6A). The IC50 of docetaxel in DACH1 -unexpressed and -re-expressed AGS cells was 0.2933 AE 0.0645 versus 3.4377 AE 0.7735 (P < 0.01, Fig. 6B). The results suggest that restored expression of DACH1 sensitized GC cells to docetaxel.

DACH1 suppresses GC growth in xenograft mice
To further explore the effect of DACH1 on GC growth in vivo, BGC823 cells with and without DACH1 expression were grown as xenograft in immune-compromised mice. As shown in Figure 7, the tumour size is smaller in DACH1 expressing BGC823 cell xenografts than in the parental non-expressing BGC823 cell xenografts (150.174 AE 191.064 mm 3 versus 912.674 AE 373.444 mm 3 ; P < 0.01, Fig. 7A and B), confirmed by measuring tumour weight, which is less in DACH1 expressing BGC823 cell xenografts than in parental BGC823 cell xenografts (54.29 AE 79.76 mg versus 450 AE 210 mg, P < 0.01, Fig. 7C).
The expression of DACH1, E-cadherin, vimentin, MMP-2 and MMP-9 in xenografts was confirmed by IHC staining. Increased expression of E-cadherin and reduced expression of vimentin, MMP-2 and MMP-9 were revealed in DACH1-stably expressed BGC823 cell xenograft (Fig. 7D). These results suggest that DACH1 suppresses GC cell growth in xenograft mice, confirming the effects seen in vitro.

Discussion
The expression of DACH1 has been reported to be reduced in different types of cancers, but increased in ovarian cancer [16,18,19,24,34,35]. We demonstrate here that DACH1 was frequently methylated in human GC and the expression of DACH1 was suppressed by promoter region methylation. DACH1 methylation is associated with higher tumour stage and lymph node metastasis. These results suggest that methylation of DACH1 may serve as a potential GC detection and prognostic marker. DACH1 suppresses GC growth both in vitro and in vivo, and further study found that DACH1 suppresses GC growth by inhibiting TGF-b signalling. It has been reported that TGF-b signalling suppresses tumour growth in the early stage and promotes the invasion and metastasis of cancer by induction of EMT in the later tumour stage [4][5][6][7][8]. Our study provides insight into this important but poorly understood switch. Aberrant activation of TGF-b signalling facilitates degradation of ECM, and degradation of the ECM is a necessity for invasion and dispersion of cancer cells and has been thoroughly studied during the last decade [36][37][38]. MMP-2 and MMP-9 are two major components of MMPs, which have been identified as key enzymes in this process. MMP-2 and MMP-9 can promote malignant cell progression and may facilitate the tumour growth, invasion and metastasis because of their ability to degrade type IV collagen of ECM and basement membrane [39][40][41][42]. MMP-2 promotes cleavage of ECM proteins and is intensively expressed by tumour and stromal components of cancer, while MMP-9 modulates permeability of vascular endothelium [42,43]. In our study, methylation of DACH1 was associated with late tumour stage and lymph node metastasis. Cell migration and invasion were suppressed by DACH1 through inhibiting TGF-b signalling in BGC823 and AGS cells. The expression of E-cadherin was up-regulated and vimentin was down-regulated by DACH1 in BGC823 and AGS cells, and the expression of MMP-2 and MMP-9 was reduced by DACH1 through inhibiting TGF-b signalling. These results were validated by xenograft. Our results suggest that cell migration and invasion are suppressed by DACH1 through inhibiting EMT, which was induced by TGF-b signalling. As DACH1 suppresses GC growth both in vitro and in vivo, it suggests that DACH1 is a tumour suppressor in human GC. Loss of DACH1 with progression of GC allows the growth-promoting activities of TGF-b to persist, while the invasive and growth inhibitory effects are lost. Cell cycle checkpoint dysfunction is often associated with sensitivity to chemotherapeutic agents. Microtubule-targeting drugs function in suppressing spindle microtubule dynamics, thus inhibiting the metaphase-anaphase, blocking mitosis and inducing apoptosis. Both docetaxel and paclitaxel are microtubule inhibitors.
Paclitaxel suppresses spindle microtubule dynamics by allowing microtubule attachment, but altering the tension across the kinetochore in mitosis [44,45]. Docetaxel disrupted centrosome organization by affecting the late S phase, G2/M phase, which results in incomplete mitosis, accumulation of cells in the G2/M phase and cell death [46,47]. However, in GC, the response rate to docetaxel is 24%, with only partial responses and a median survival of 7.5 months [48]. As docetaxel is poorly tolerated clinically, it is important to find sensitive markers for personalized treatment. DACH1 was silenced by promoter region hypermethylation and G2/ M phase arrest was induced by DACH1 in BGC823 and AGS cells.
To explore the possible connection of DACH1 expression and docetaxel sensitivity, we detected the IC50 of docetaxel in DACH1 unexpressed and re-expressed BGC823 and AGS cells. Our results suggest that re-expression of DACH1 sensitized BGC823 and AGS cells to docetaxel. It suggests that DACH1 may enhance docetaxel to exert its cytotoxic activity through stabilizing microtubule assembly, leading to cell cycle arrest at G2/M phase. Another report demonstrated that enhancing acetylation of P53 could have potential implication for increasing the sensitivity of cancer cells to Taxol [49]. DACH1 was identified as a novel P53 binding partner [50,51]. Further study is needed to understand if DACH1 sensitizes GC cells to docetaxel through P53. Our results suggest that DACH1 methylation may serve as a docetaxel-resistant marker in GC and epigenetic therapy that reverses DACH1 silencing might result in sensitization to doecetaxel.
In conclusion, DACH1 is frequently methylated in human GC and the expression of DACH1 was silenced by promoter region hypermethylation. DACH1 methylation is associated with late tumour stage and lymph node metastasis. DACH1 suppresses GC proliferation, invasion and metastasis by inhibiting TGF-b signalling pathways both in vitro and in vivo. DACH1 re-expression sensitized GC cells to docetaxel.