Thioredoxin 1 mediates TGF-β-induced epithelial-mesenchymal transition in salivary adenoid cystic carcinoma.

Epithelial-mesenchymal transition (EMT) plays an important role in the invasion and metastasis of salivary adenoid cystic carcinoma (SACC) which is characterized by wide local infiltration, perineural spread, a propensity to local recurrence and late distant metastasis. Our recent studies have disclosed that TGF-β is a crucial factor for EMT in metastatic SACC. In this study, we further uncovered small redox protein thioredoxin 1 (TXN) as a critical mediator of TGF-β induced EMT. Immunohistochemistry analysis revealed significantly higher expressions of TXN, thioredoxin reductase 1 (TXNRD1) and N-cadherin, and lower expression of E-cadherin in human metastatic SACC compared to non-metastatic SACC tissues. Consistently, cultured SACC cells with stable TXN overexpression had decreased E-cadherin and increased N-cadherin as well as Snail and Slug expressions. The enhanced migration and invasion potential of these cells was abrogated by Akt or TXNRD1 inhibitors. Expression of N-cadherin and Akt p-Akt decreased, whereas E-cadherin expression increased in a BBSKE (TXNRD1 inhibitor)-dose-dependent manner. In a xenograft mouse model, TXN overexpression facilitated the metastatic potential of SACC-83 cells to the lung. Our results indicate that TXN plays a key role in SACC invasion and metastasis through the modulation of TGF-β-Akt/GSK-3β on EMT. TXN could be a potential therapeutic target for SACC.


INTRODUCTION
Salivary adenoid cystic carcinoma (SACC) is one of the most common malignant tumors of salivary glands that arise from ductal, myoepithelial, and basal cells [1,2]. Because of persistent indolent growth, high rates of recurrence, and distant metastasis, the overall survival rate is 71% at 5 years, but 37% at 15 years. Survival rates after diagnosis of tumor progression at 5 and 15 years are 35% and 0%, respectively [3]. Surgical dissection and postoperative radiotherapy, the standard treatment, has provided reasonable local control rates but limited response for distant metastasis [4]. A better understanding of the biological mechanisms underlying SACC metastasis is needed to improve this situation.
Epithelial-mesenchymal transition (EMT) is the molecular basis of cancer metastasis [5], and involves loss of cell-cell adhesion, cell polarity, acquisition of migratory and invasive capabilities. In EMT, there are decreased expression of cell adhesion molecules such as E-cadherin and increased expression of mesenchymal markers such as vimentin and N-cadherin [6]. TGF-β, the most potent factor of EMT, has been implicated in various types of cancer cells [7]. Our previous studies have shown that human metastatic SACC tissue samples and SACC-LM cell line had higher expression of TGF-β. In addition, TGF-β-induced EMT promoted the migration and invasion of SACC-83 cells [8,9]. However, the mechanisms underlying TGF-β-induced EMT in SACC are still unclear, which needs to be investigated. www.impactjournals.com/oncotarget The thioredoxin system including thioredoxin, thioredoxin reductase and nicotinamide adenine dinucleotide phosphate (NADPH) is involved in aspects of tumor physiology such as proliferation, apoptosis, and metastasis [10]. Thioredoxin 1 (TXN), a 12-kDa redox protein, is important in regulation of cellular redox homeostasis and anti-apoptotic functions. TXN stimulates cell proliferation and cell cycle progression, induces hypoxia-inducible factor-1α (HIF-1α) and angiogenesis, and alters the balance between the matrix metalloproteinases and their tissue inhibitors [11,12]. Recent studies have shown that TXN had high expression in many human primary tumor tissues such as the liver, colon, pancreas, and the uterine cervix [13][14][15][16]. Increased TXN expression may be associated with oncogenesis, a poorer survival rate, and higher reactive oxygen species (ROS) generation [17]. Recent findings have demonstrated that ROS-evoked signals are associated with TGF-β signaling-mediated EMT in breast and renal tubular epithelial cells and in cardiac and pulmonary fibrosis [18][19][20][21]. As a negative regulator of TXN, thioredoxin binding protein-2 (TBP-2) deficiency can upregulate the expression of Snail or Slug in TGF-β-driven EMT [22]. However, little is known about the role of TXN in tumor mobility and invasion during the process of TGF-β-induced EMT. We hypothesize that TXN may be involved in the TGF-β-mediated EMT-induced tumor mobility and invasion in SACC.
In this study, we compared the expression of TXN, TXNRD1, E-cadherin and N-cadherin in SACC patients with or without lung metastasis. We also manipulated TXN expression in two SACC cell lines, SACC-83(low lungmetastatic cell line) and SACC-LM (high lung-metastatic cell line) [9] to compare cell-invasive behavior. Silence of TXN Figure 1: Immunohistochemical staining reveals differential expressions of thioredoxin 1 (TXN), thioredoxin reductase 1 (TXNRD1) and epithelial-mesenchymal transition signs in salivary adenoid cystic carcinoma (SACC) samples from patients with/without lung metastasis. Higher expression of TXN (AA'), TXNRD1 (CC') and N-cadherin (GG') in SACC tissues with metastasis compared to those without metastasis (BB', DD', HH'), respectively. Lower expression of E-cadherin in SACC tissues with metastasis (EE') compared to those without metastasis (FF'). Scale bar = 20 μm, A'-H' show enlarged fields of the inset squares in panels A-H. www.impactjournals.com/oncotarget decreased SACC-LM cell mobility and invasion, whereas overexpression of TXN increased SACC-83 cell mobility and invasion which was dependent on the Akt/GSK-3β pathway. Our in vivo study further found that SACC-83 cells overexpressing TXN had significantly increased potential of lung metastasis. In addition, EMT measured by E-cadherin and N-cadherin and cell invasion was promoted by manipulating TXN expression in SACC-83 cells. These findings suggest that the EMT mediated by TXN and TXNRD1 plays an important role in SACC metastasis. Therefore, TXN and TXNRD1 could be novel targets for SACC treatment in future.

Endogenous TXN and TXNRD1 expressions are correlated with the potential of metastasis and poor survival in SACC patients
Expressions of TXN, TXNRD1, E-cadherin, and N-cadherin were detected by immunohistochemical analysis in SACC tissues with (n = 25) or without (n = 22) metastasis. SACC tissues with metastasis had high expression of TXN and TXNRD1, which were correlated with high expression of the mesenchymal marker N-cadherin and low expression of the epithelial marker E-cadherin ( Figure 1A-1H). Occasional nuclear staining of TXN was also found in the stained tissue sample ( Figure 1A'). Kaplan-Meier survival analysis for above 47 SACC patients demonstrated that TXN and TXNRD1 expressions were correlated with poor survival rate (P = 0.0072 and P = 0.0224, respectively) ( Figure 2A and 2B). TXN was expressed in 21 out of 25 SACC samples with metastasis and in 2 out of 22 SACC samples without metastasis (Table 1). Correlations between TXN expression and clinicopathological features of SACC were summarized in Table 1. As shown in Table 1, high TXN expression in SACC was significantly correlated with clinical stage (P = 0.012) and distant metastasis (P < 0.001). TXN expression in SACC was also positively associated with TXNRD1 (P < 0.001), N-cadherin (P = 0.018), and negatively associated with E-cadherin (P = 0.01) expression (Table 2).

TXN expression in SACC cell lines potentially impacts on EMT, migration and invasion
Our previously established SACC-83 and SACC-LM cell lines have identical STR profiling, and express epithelial markers such as pan-cytokeratin and cytokeratin AE1, and the luminal markers such as CK8/18 and S100P, indicating that both cell lines were originated in oral adenoepithelial cells and not contaminated by other cancer cell lines [8]. In this study, we further identified that SACC-83 and SACC-LM cell lines expressed the intrinsic SACC biomarkers including c-myb, FABP7 and NTF3F ( Figure S1, see Table  S1 for primer sequences). Using these two cell lines, we explored the role of TXN expression in EMT, migration and invasion. Our data showed that SACC-LM cell line had higher expressions of TXN and N-cadherin, but lower expression of E-cadherin compared to SACC-83 cell line ( Figure 3A). The overexpression of TXN by transfection of GV230 vector in SACC-83 (SACC-TXN) cells showed a fibroblast-like morphology ( Figure 3B). Wound closure and transwell assays revealed that overexpression of TXN increased SACC-83 cell migration and invasion whereas knockdown of TXN decreased SACC-LM cell migration and invasion ( Figure 3C and 3D). In addition, TXN overexpression upregulated N-cadherin and downregulated E-cadherin expression in SACC-TXN cells ( Figure 3E  downregulated N-cadherin expression ( Figure 3E). These data suggested that TXN may play a potential role in promoting motility, invasion and EMT of SACC cells.

TXN is a critical mediator in TGF-β-induced EMT in SACC cells
SACC cells respond to TGF-β and acquire a mesenchymal phenotype through EMT, which is accompanied with a distinctive pattern of gene expression [9]. TGF-β stimulates EMT, migration, invasion, and metastasis of adenoid cystic carcinoma cells. To investigate how TGF-β affects TXNdependent EMT in SACC cell lines, we knocked down TGF-β in SACC-LM by siRNA, and found that TXN protein expression and N-cadherin were downregulated, while E-cadherin were upregulated ( Figure 4A). Treatment of SACC-83 with recombinant TGF-β induced TXN and N-cadherin expressions in a dose-dependent manner ( Figure  4B), whereas TXN knockdown by siRNA suppressed effects of TGF-β on EMT ( Figure 4C). These data suggest that TXN involves in the TGF-β -induced EMT in SACC cells.

TXN regulates EMT through Akt/GSK-3β/Snail signaling pathway
Previous studies have demonstrated that TGF-β regulates Snail expression through PI3K/Akt / GSK-3β signaling [23,24]. To further elucidate the mechanism of TXN induced cell motility, we examined above signaling components which are known to regulate cell migration and invasion. We found that overexpression of TXN increased the levels of Snail, Slug and Akt/GSK-3β phosphorylation, whereas knockdown of TXN reduced the levels of Snail, Slug, Akt/GSK-3β phosphorylation ( Figure 5A and 5B). LY49002, PI3K inhibitor, suppressed motility and invasion ( Figure 5C and 5D), and increased Snail, Slug, Akt/GSK-3β phosphorylation, TXN, TXNRD1 and N-cadherin while increased E-cadherin of SACC-TXN cells in a dose-dependent manner ( Figure  5E and 5F). These findings suggest that TXN and PI3K/ Akt/GSK-3β signaling coordinately regulate EMT of SACC cells.

TXNRD1 targets Akt/GSK-3β to regulate EMT of SACC cells
TXNRD1 plays an important role in catalyzing the NADPH dependent regulation of TXN [25]. TXNRD1 knockdown markedly suppresses tumor progression and metastasis and decreases transcription levels of cancerrelated proteins [26]. To examine the role of TXNRD1 in EMT of SACC-LM cells, we used TXNRD1 inhibitor, 1, 2-[bis (1, 2-Benzisoselenazolone-3 (2H) -ketone)] ethane (BBSKE) to treat SACC-LM cells and found that activity of TXNRD1 was decreased in a dose-dependent manner ( Figure 6A). In addition, N-cadherin expression and Akt/GSK-3β phosphorylation were significantly decreased while E-cadherin expression was increased after BBSKE treatment in SACC-LM cells ( Figure 6B). These findings imply that TXN and TXNRD1 could target PI3K/Akt/GSK3β which may impact EMT of SACC cells (Figure 7).

TXN overexpression in SACC cells increases lung metastasis in vivo
To investigate the role of TXN in invasiveness and metastatic potential of SACC cells in vivo, 3 × 10 6 SACC-TXN or Mock cells were injected into the tail vein of NOD/ SCID mice. Eight weeks later, 83% mice injected with SACC-TXN cells developed lung metastases, compared to 17% mice injected with Mock cells ( Figure 8A). Lung tissues of mice in SACC-TXN group had more tumor cells compared with those of mice in Mock group ( Figure 8B). These results suggest that TXN might enhance SACC tumor cell metastasis in vivo.

DISCUSSION
Previous studies suggest that both TXN and TXNRD1 are overexpressed in aggressive tumors with a high proliferation capacity, a low apoptosis rate, and an elevated metastatic potential [27]. In this study, we found higher expressions of TXN and TXNRD1 in the SACC tissue samples with distant metastasis. We also found that overexpression of TXN could increase lung metastasis of the SACC cells in a xenograft mouse model. Overexpression of TXN promoted an EMTlike phenotype with increased N-cadherin expression, enhanced cell mobility and invasion, and induced Snail/ Slug expression through PI3K/Akt/GSK3β signaling pathway. These findings suggest that TXN and its related signaling pathway could be novel targets for therapeutic treatment of SACC.
TXN increases proliferation and resistance to cell death, and promotes metastatic progression of tumor cells [28]. It is predominantly located in cytoplasm, but  also in the nucleus [29]. Nuclear translocation of TXN activates transcriptional factors involved in cellular redox regulation [30] and impacts high infiltration and/ or metastatic capability of cancer cells [31]. Our present study found the occasional nuclear staining of TXN and TXNRD1 in SACC tumors with distant metastasis. In addition, increased expressions of TXN and TXNRD1 were correlated with clinical stage and distant metastasis in SACC patients, which supports previous report that TXN overexpression is an independent prognostic factor in metastatic cancers [30]. SACC-LM cells with high potential of metastasis had a higher expression of TXN compared with SACC-83 cells which have low potential of metastasis. In addition, when the SACC-83 cells were overexpressed TXN, they got an increase in EMT, cell migration and invasion. Combining with the role of EMT in chemoresistance and cancer cell invasion/metastasis [32], our data suggest that TXN may play a crucial role in regulation of SACC metastasis through EMT.
Our previous studies have shown that TGF-β induces EMT progression in SACC. In this study, we have further identified the role of TXN in TGF-β-induced EMT. Either knocking down TGF-β with siRNA in SACC-LM cells, or treating with TGF-β at different doses in SACC-83 cells, we found that TXN, TXNRD1, E-cadherin, and N-cadherin were regulated by TGF-β accordingly. In addition, knockdown of TXN followed by treatment with TGF-β could attenuate the effects of TGF-β on EMT. These data indicate that TXN is involved in the TGF-βinduced EMT in SACC.
Snail, a member of the zinc-finger transcription factor family, is a master regulator that promotes EMT and mediates invasiveness and metastasis in many different types of malignant tumors [33][34][35]. Snail and Immunoblot analysis of expression of TXN, TXNRD1, E-cadherin, and N-cadherin in SACC-83 cells treated with or without TGF-β (2, 5 10 ng/ml) for 24 hours. C. After pre-transfection of TXN or Con siRNAs for 24 hours, siTXN or siCon cells were treated with or without TGF-β (10 ng/ml) for another 24 hours. Then expressions of TXN, TXNRD1, E-cadherin, and N-cadherin were analyzed. β-actin was used as loading control. Data represent means ± SD of three independent experiments (*P < 0.05). www.impactjournals.com/oncotarget Slug are recognized as essential factors in EMT of SACC cells [36,37]. In particular, Ras-MAPK activates Snail and Slug, which transcriptionally inhibit E-cadherin and promote EMT [38,39]. Previous studies found that Snail was aberrantly expressed and essential for TGFβ-induced EMT in SACC [36,40]. In this study, both Snail and Slug were upregulated in SACC-TXN cells. In addition, knockdown of TXN in SACC-LM cells inhibited Snail and Slug expressions. However, it is still unclear how Snail and Slug affect EMT in TXN overexpression SACC cells. Multiple signalings including Ras/MAPK, p38MAPK, Rho kinase, PI3-kinase, and Smad are related to the TGF-β-induced EMT [41]. The Akt/GSK-3β signaling pathway can modulate stability of Snail [24,42]. PI3K/Akt pathway is also involved in cell migration, invasion and cell survival [43]. Our previous study have found that PI3K inhibitor LY294002 suppressed the EMT as well as decreased the migration  ability of SACC cells [44]. TXN binds to and inhibits proapoptotic proteins, including apoptosis signal regulating kinase-1 (Ask-1) [45] and tumor suppressor PTEN [46], a protein that attenuates the PI3K/Akt cell survival pathway when it is present in cancer cells. Our data suggested that TXN might contribute to EMT modulated by PI3K/Akt / GSK-3β. It is supported by our findings that LY294002 suppressed expression of TXN and TXNRD1, while overexpression of TXN in SACC-83 cells impacts on Akt/ GSK-3β phosphorylation. Previous studies have showed that GSK-3β is inhibited by TGF-β through increased serine 9 phosphorylation and increased stabilization of Snail via the PI3K/AKT signaling network [47].
TXNRD1 associated with aggressive tumor growth and poor survival rate plays an important role in regulation of TXN [25]. We further used BBSKE, a TXNRD1 activity inhibitor and demonstrated that BBSKE decreased TXNRD1 activity of SACC-LM in a dose dependent manner and downregulated N-cadherin expression, and increased E-cadherin expression via PI3K/Akt/GSK3β signal. It suggests that BBSKE may be a potential targeting agent in SACC metastasis through Akt/ GSK-3β signaling pathway.
In summary, our present study demonstrated that TXN mediates TGF-β-induced EMT and promotes SACC metastasis through stabilization of the transcriptional factors, Snail and Slug, and cooperation with the PI3K/ Akt/GSK-3β signaling. Our results suggest that TXN may serve as a prognostic marker for predicting the risk of developing distant metastases in SACC. In addition, TXN and its related signaling pathway could be potential therapeutic targets for SACC.

In vitro wound closure assay
After overnight serum deprivation, confluent monolayers of cells were scratched with a 200 μl pipette tip to create wounded areas with width of 400-600 μm. Wounded monolayers were photographed at 0 and 9/12 hours after scratching. Wound closure was determined by the reduction in the width of wounded areas. Average rates of wound closure were calculated as reduced width/ incubated time.

Transwell invasion assay
Transwell invasion assays were performed in the transwell chambers with a polycarbonate membrane (Millipore, Bedford, MA, USA) coated with 20 μg extracellular matrix (ECM) gel (Sigma). Cells were serum starved overnight, and then seeded at 2 × 10 5 cells/ well in medium without serum in the upper chamber, but with 15% fetal bovine serum (FBS) in the lower chamber. 24 hours after incubation, cells that had invaded through the matrix gel into the lower chamber were fixed with 95% ethanol and stained with 1% crystal violet (Sigma). Cells on the upper surface of the membrane were wiped off. Membranes were photographed in 4 random fields and the number of cells counted by light microscopy. Every experiment was repeated independently at least three times.

Experimental metastasis assay
Five-week-old NOD/SCID female mice were purchased from Vital River Laboratories (Beijing, China). Mice were randomly grouped (6 mice per group). Mice were tail-vein injected with 3 × 10 6 SACC-TXN or Mock cells in 100 μl PBS. After 8 weeks, mice were sacrificed. Lungs from mice were removed, fixed in 10% buffered formalin for 24 hours, examined with a microscope and scored for visible surface tumors. In addition, formalin fixed and paraffin embedded sections (4 μm) were used for Hematoxylin and Eosin staining. All animal care and procedures were approved by Peking University School of Stomatology Institutional Review Board for Animal Experiments.

Immunohistochemistry
A cohort of 47 patients (20 male and 27 female with a median age of 46.5 years) diagnosed with SACC and treated at the Department of Oral and Maxillofacial Surgery, Peking University School of Stomatology during 1996-2006 were followed up. The mean follow-up period was 89.8 months. X-rays were performed to determine presence of pulmonary metastasis. Clinicopathological data were summarized in Table 1. Informed written consent was obtained from all subjects, and the study was approved by Peking University School of Stomatology Institutional Review Board.

Statistical analysis
Data were presented as means ± SD. Statistical analysis was performed using SPSS Version 19.0 software. Data were grouped according to treatments and analyzed by one-way ANOVA. χ 2 test was used to determine the correlation between TXN expression and clinicopathologic characteristics as well as TXNRD1, E-cadherin, and N-cadherin expressions. Overall survival rate was estimated using Kaplan-Meier method. P < 0.05 was considered statistically significant.