14-3-3ζ promotes hepatocellular carcinoma venous metastasis by modulating hypoxia-inducible factor-1α

Portal vein tumor thrombus (PVTT) is a type of intrahepatic metastasis arising from hepatocellular carcinoma (HCC) and is highly correlated with a poor prognosis. Hypoxia is common in solider tumors, including HCC, where it alters the behavior of HCC cells. We asked whether and how hypoxia contributes to PVTT formation. We demonstrated that increased intratumoral hypoxia is strongly associated with PVTT formation in HCC. We also showed that 14-3-3ζ is induced by hypoxia in HCC cells and correlates with PVTT formation in clinical HCC samples. In addition, 14-3-3ζ up-regulates HIF-1α expression by recruiting HDAC4, which prevents HIF-1α acetylation, thereby stabilizing the protein. Under hypoxic conditions in vitro, 14-3-3ζ knockdown inhibits hypoxia-induced HCC invasion by the HIF-1α/EMT pathway. Blockade of 14-3-3ζ in HCC cells reduces PVTT formation and distant lung metastasis in vivo. Moreover, a combination of 14-3-3ζ and HIF-1α expression is more prognostic for HCC patients than either protein alone. These results suggest that the hypoxia/14-3-3ζ/HIF-1α pathway plays an important role in PVTT formation and HCC metastasis.


INTRODUCTION
Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide and the third leading cause of death from cancer worldwide [1,2]. Both intra-and extrahepatic metastases result in a poor prognosis for HCC patients [3,4]. Portal vein tumor thrombus (PVTT), arising from the invasion of HCC cells into the portal vein (trunk or branch), is a special type of intrahepatic metastasis of HCC [5]. Approximately 20%-70% of HCC is accompanied by PVTT as determined by autopsy and/or radiographic examination, while up to 90% of HCC patients have PVTT determined by microscopic examination [6]. HCC patients with PVTT exhibited a poorer clinical outcome [5,7]. Clarification of the mechanisms underlying the formation and development of PVTT is crucial for developing novel therapeutic strategies for HCC patients.
Hypoxia is common in tumors including HCC. Hypoxia alters HCC cell proliferation, apoptosis, metastasis, chemoresistance, and radioresistance [8]. In HCC, hypoxia results from a shortage of blood circulation caused by liver cirrhosis and the rapid growth of tumor cells. Interestingly, liver cirrhosis and tumor size (> 8 cm) are both independent predictors of PVTT in HCC [9]. In addition, expression of protein disulfide-isomerase A6 (PDIA6), apolipoprotein A-I (APO A-I) and CXC chemokine receptor 4 (CXCR4), correlate with PVTT, and can also be induced by hypoxia [10,11]. Thus, PVTT formation may be associated with the hypoxic microenvironment of HCC. However, there is no direct evidence showing that hypoxia contributes to the formation of PVTT.
In the present study, we uncovered a causative link between intratumoral hypoxia and PVTT formation. Hypoxia increased the expression of 14-3-3ζ, which www.impactjournals.com/oncotarget induced hypoxia-induced factor-1α (HIF-1α) expression by stabilizing HIF-1α protein. This resulted in an enhanced EMT response of HCC cells, promoting the formation of PVTT and HCC metastasis. These results reveal that the hypoxia/14-3-3ζ/HIF-1α pathway underlies PVTT formation and may contribute to the development of new therapeutics for HCC patients.

Both HIF-1α levels and PVTT formation in HCC are strongly correlated with 14-3-3ζ expression
Since HIF-1α plays a crucial role in hypoxia [12], we sought to identify proteins interacting with HIF-1α under the assumption that these proteins will be potential drug targets to inhibit HIF-1α. We first screened proteins associated with HIF-1α using Scansite software (http://www.motifscan. mit.edu). 35 proteins were found to potentially interact with HIF-1α (Figure 2A, left panel and Supplementary  Table 1). We next performed cDNA microarray analysis using paired HCC/normal tissues (n = 5) to profile the genes up-regulated in HCC. 1012 genes were found increased in tumor tissues compared with the matched nontumorous tissues (Figure 2A, left panel and Supplementary Table 2). Combined analysis of both results from bioinformatics and microarray analysis showed that only YWHAZ (also known as 14-3-3ζ) and PRKDC were potential HIF-1α binding proteins that were also increased in HCC (Figure 2A, left panel and Supplementary Figure 1). 14-3-3ζ was selected for further investigation due to its correlation with tumor metastasis and functions in regulating proteins stability [13]. Gene set enrichment analysis (GSEA) confirmed the positive correlation of 14-3-3ζ with cell migration and tumor invasiveness ( Figure 2A, right panel). The previous IHC analysis using TMAs ( Figure 1) showed a positive linear correlation between the expression of 14-3-3ζ and HIF-1α in primary tumor tissues, especially in PVTT(+) tumor tissues ( Figure 2B), suggesting a potential relationship between the two proteins.
We then investigated 14-3-3ζ expression in HCC tissues using real-time PCR and western blots. Both 14-3-3ζ mRNA and protein were higher in tumor tissues (T) than that in nontumorous tissues (N) (Supplementary Figure 2A-2B). A similar result was observed in published data from Liver hepatocellular carcinoma (LIHC) in The Cancer Genome Atlas (TCGA) database (Supplementary Figure 2C). Then, the expression of 14-3-3ζ in HCC tumors from patients with (PVTT+, n = 10) or without PVTT (PVTT−, n = 10) was assessed by western blot to determine the relationship between 14-3-3ζ expression and PVTT status in HCC. 14-3-3ζ protein level was significantly higher in PVTT(+) tumors than in PVTT(−) tumors ( Figure 2C). Similar results were also observed in IHC analysis ( Figure 2D), further confirming the potential correlation between 14-3-3ζ overexpression and the PVTT formation.

14-3-3ζ regulates the aggressive behavior of tumor cells by the HIF-1α /EMT signaling pathway under both hypoxic and normoxic conditions
Having documented that hypoxia induced 14-3-3ζ expression regulated HIF-1α expression, we sought to elucidate the role of 14-3-3ζ in tumor cell behavior. Transwell invasion assays were used to assess the potential invasion capacity of tumor cells with 14-3-3ζ modulation. 14-3-3ζ down-regulation reduced the invasion capacity of CSQT-2 and HCC-LM3 cells under both of normoxic and hypoxic conditions ( Figure 5A and Supplementary Figure 5), which could be rescued by HIF-1α ( Figure 5B). These results suggested that hypoxia-induced invasion of HCC cells is, at least in part, a result of activation of 14-3-3ζ/HIF-1α axis.
To verify the involvement of 14-3-3ζ in tumor metastasis in vivo, we performed orthotopic transplantation assays in nude mice. CSQT-2 cells were used because they are derived from PVTT and prone to extend to portal veins forming PVTT, and sh14-3-3ζ was used for knock-down as previously described [16]. The incidence of PVTT was lower in mice transplanted with CSQT-2/ sh14-3-3ζ cells than in those transplanted with CSQT-2/shcon cells ( Figure 5C). We also established a mouse model of PVTT by injecting HCC-CSQT-2/sh14-3-3ζ cells or HCC-CSQT-2/shcon cells into mice through the tail vein. Histopathological examination revealed fewer micrometastatic lesions in the lungs of mice injected with HCC-CSQT-2/sh14-3-3ζ cells than in those of control mice injected with HCC-CSQT-2/shcon cells ( Figure 5D, left panel). Similar results were obtained using LM3 cells ( Figure 5D, right panel). These results suggest that 14-3-3ζ down-regulation in HCC cells could inhibit metastases to lung and the formation of PVTT in vivo.

SMMC-7721 cells (Supplementary
In addition, we found that the up-regulation of E-cadherin and down-regulation of Vimentin and Snail induced by sh14-3-3ζ can be rescued by HIF-1α (Supplementary Figure 6F). Moreover, down-regulation of 14-3-3ζ inhibited hypoxia induced down-regulation of E-cadherin and up-regulation of Vimentin and Snail (Supplementary Figure 6G). The expressions of these proteins in 14-3-3ζ down-regulated tumor cells was rescued by ectopic HIF-1α expression under hypoxic conditions (Supplementary Figure 6G). These results suggest that 14-3-3ζ promotes HCC cells metastasis via the HIF-1α/EMT signaling pathway under hypoxic conditions.
For patients whose tumors expressed above-average levels of HIF-1α (HIF-1α high), adverse outcomes were exacerbated (Supplementary Figure 8). moreover, the HCC patients with above-average levels of both HIF-1α and 14-3-3ζ (HIF-1α high/14-3-3ζ high) displayed even worse prognoses, indicating that a combination of these two parameters increases the prognostic value ( Figure 6B). Thus, evaluation of both HIF-1α and 14-3-3ζ expression is a powerful predictor of poor prognosis in HCC, leading further support to a model of 14-3-3ζ increasing HIF-1α expression through enhancement of protein stability, resulting in an enhanced EMT followed by metastases of HCC cells and the formation of PVTT ( Figure 6C).

DISCUSSION
As an important microenvironmental factor of solid tumors, hypoxia plays an critical role in tumor metastasis including HCC. Although a positive relationship between intratumoral hypoxia (reflected by HIF-1α) and PVTT formation in HCC patients has been found in several clinical studies [18,19], the mechanisms associated with hypoxia-induced PVTT formation are still obscure. In the current study we found that increased intratumoral hypoxia is strongly correlated with PVTT formation in HCC patients and the hypoxia/14-3-3ζ/HIF-1α signaling axis contributes to the PVTT development.
HIF-1α is a major factor regulating the cellular response to hypoxia. Aberrant expression of HIF-1α has been previously observed in almost every solid tumor, and HIF-1α has been found to promote growth and metastasis of tumors [8,20]. Considering the central role of the interaction between HIF-1α and other proteins under hypoxic conditions [21], we investigated the novel target proteins of HIF-1α and investigated whether these target proteins contribute to hypoxia-induced PVTT formation. We showed that 14-3-3ζ, prevents acetylation of HIF-1α by recruiting HDAC4 protein, resulting in the increased stabilization of HIF-1α protein. Importantly, knockingdown of 14-3-3ζ attenuates hypoxia-induced invasion in vitro and inhibits PVTT formation as well as distant lung metastasis in vivo, indicating that 14-3-3ζ plays an important role in hypoxia-mediated metastasis and PVTT formation. Therefore, our results reveal for the first time that 14-3-3ζ functions as a positive regulator of HIF-1α, and the induction of HIF-1α by 14-3-3ζ augments the metastatic potential of tumor cells under hypoxic conditions. 14-3-3ζ was previously found to participate in the formation of breast cancer, non-small cell lung cancer (NSCLC), and head and neck squamous cell carcinomas (HNSCCs) [22][23][24]. Elevated 14-3-3ζ expression in these cancers promoted proliferation, inhibited apoptosis, and enhanced chemotherapy resistance in cancer cells. 14-3-3ζ was also increased in HCC, while 14-3-3ζ inhibition suppressed tumor cell proliferation and enhanced the anti-cancer effects of cis-diamminedichloridoplatium (CDDP) [25]. Additionally, 14-3-3ζ up-regulation was involved in tumor invasion and metastasis. By forming a complex with ErbB2 or αB-crystallin, 14-3-3ζ promoted tumor metastasis by inducing EMT of tumor cells. [26][27][28]. In accordance with these studies, in the present study we demonstrated elevated 14-3-3ζ expression in HCC (Supplementary Figure 2) and its role in HCC metastasis via EMT ( Figure 5). Moreover, we found that 14-3-3ζ stabilized HIF-1α protein through preventing its acetylation via recruiting HDAC4 and then induced EMT of HCC cells, suggesting a novel pathway involving 14-3-3ζ. Aberrant expression of 14-3-3ζ, previously observed in several cancers, is largely a result of genomic alterations and post-transcriptional modulation of stability [13,29]. Our results also showed that both 14-3-3ζ mRNA and protein are overexpressed in HCC tissues ( Figure 2) and hypoxia increases 14-3-3ζ protein by inhibiting the proteasome-mediated degradation, while 14-3-3ζ mRNA was not affected by hypoxia (Figure 3). These results indicated that multiple mechanisms contribute to the aberrant expression of 14-3-3ζ. We speculated that genomic alterations increased 14-3-3ζ mRNA levels and that the hypoxic microenvironment further stabilizes 14-3-3ζ proteins, resulting in 14-3-3ζ overexpression in tumor tissues.
Although we have demonstrated that the hypoxia/14-3-3ζ/ HIF-1α pathway plays an important role in HCC metastasis and PVTT formation, our study also raises several critical questions. Previous studies showed inflammatory cytokines, including TGF-β and IL-6, contribute to the PVTT formation [7,30]. Hypoxia may also be involved in increasing TGF-β and IL-6 in tumor cells [31,32]. In addition, several inflammatory cytokines, including TNF-α, IL-1, IL-6, TGF-β and IL-18, have been documented to increase HIF-1α expression or transcriptional activation in tumor cells [33][34][35][36]. Therefore, 14-3-3ζ may be involved in hypoxia-induced inflammatory cytokine expression, which also promotes PVTT formation in HCC patients. Additionally, our previous research found that non-coding RNAs such as miRNAs also involved in PVTT development [37]. Whether or not these non-coding RNAs also interact with the hypoxia/14-3-3ζ/ HIF-1α pathway to promote PVTT development is also unclear. A comprehensive investigation for the synergistic actions of a hypoxic microenvironment, inflammatory responses, and non-coding RNAs will be helpful for revealing the full mechanism underlying the PVTT formation [36]. Moreover, in addition to 14-3-3ζ, several other isoforms of 14-3-3 such as 14-3-3β and 14-3-3ε have also been shown to contribute to cell migration and EMT of HCC [38,39]. These other 14-3-3 isoforms are also overexpressed and associated with high metastatic risk and poorer survival rates of HCC [38][39][40][41][42]. However, whether or not these proteins are involved in PVTT formation and are regulated by the same mechanism in HCC remains unclear.
In conclusion, we found that hypoxia up-regulates 14-3-3ζ expression in HCC by increasing its protein stability through a ubiquitin pathway. 14-3-3ζ promotes hypoxia-induced HCC invasion in vitro and PVTT formation in vivo by stabilizing HIF-1α protein. These findings reveal a novel mechanism underlying PVTT formation by the hypoxia/14-3-3ζ/HIF-1α pathway, which may contribute to the development of new therapeutics for HCC patients.

Patients and samples
This study was approved by the Ethics Committee of Eastern Hepatobiliary Surgery Hospital and all participants provided written consent. The criteria for inclusion and exclusion was in accordance with previous reports [16]. According to these criteria, 144 HCC patients (collected from June 2009 to May 2010, Supplementary Table 6), who underwent curative resection in the Eastern Hepatobiliary Surgery Hospital (Shanghai, China) were randomly retrieved to construct a tissue microarray (TMA) as previously described [43]. Each patient was followed until September 2012 with the longest follow-up up to 39 months. The time of the surgery was used to calculate the time to a given event. Overall survival (OS) and time to recurrence (TTR) were defined as previously described [44]. In

Cell lines and cell culture
The liver cancer cell line SMMC-7721 was purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The MHCC-LM3 liver cancer cell lines was provided by Professor Wei-zhong Wu (Zhong Shan Hospital, Fu Dan University, Shanghai, China). The PVTT-originated cell line CSQT-2 was generated in our laboratory [16]. Cells were maintained at 37°C in a humidified incubator containing 5% CO 2 in Dulbecco's Modified Eagle's Medium (Gibco-BRL, CO.LTD,USA), supplemented with 10% heat-inactivated fetal bovine serum (Gibco-BRL, CO.LTD, USA) and passaged every 3 days to maintain logarithmic growth. To mimic hypoxic conditions, the cells were incubated in the presence of 200 μM cobalt chloride (CoCl 2 , St. Louis, MO USA).

Microarray analysis
Total RNA was isolated from 5 paired HCC tissues (tumor tissues and matched nontumorous tissues using Trizol reagent (Invitrogen) and was sent to Gminix Co., Ltd (Shanghai) for cDNA microarray analysis as previously described [45].

Bioinformatics analysis
The liver hepatocellular carcinoma (LIHC) dataset consisting of 371 HCC tumor tissues and 50 non-tumorous tissues was downloaded from TCGA website (https://tcgadata.nci.nih.gov/tcga/) following approval of this project by the consortium. To probe the YWHAZ (14-3-3ζ) gene-associated pathways in an unbiased basis way, we performed gene set enrichment analysis (GSEA) using GSEA version 2.0 from the Broad Institute at MIT. Expression of the gene was used as a phenotype label, and "Metric for ranking genes" was set to Pearson Correlation.

Reverse transcriptase-polymerase chain reaction (RT-PCR) and quantitative RT-PCR
Total RNA was isolated from the cell lines and clinical samples using TRIzol reagent according to the manufacturer's recommended protocol (Invitrogen, Carlsbad, CA). RT-PCR performed as described previously with specific primers [16]. Quantitative RT-PCR was performed using SYBR Premix Ex Taq TM (TaKaRa BIOTECHNOLOGY CO.,LTD, Dalian, China) and the StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA). Gene expression was calculated relative to the expression of β-actin in tumor cell lines or clinical samples using the 2 -ΔΔCt method. The primers used were provided in the Supplementary Table 7.

Western blot analyses
Western blot analyses were performed as previously described [46]. The total soluble proteins (40 μg) were used and β-actin was used as an internal control. The antibodies used are listed in the Supplementary Table 8.

Immunohistochemical analyses
Paraffin-embedded tissue sections or tissue microarray slides were analyzed using immunohistochemistry (IHC) as previously described [47]. The antibodies that were used for IHC are listed in Supplementary Table 8. Slides were scanned with an Aperio ScanScope GL. Aperio ImageScope software (Aperio Technologies, Vista, CA) was then used to assess the staining of the scanned images based on the percentage of positively stained cells and the staining intensity. Scores were then generated by software.

In vitro cell behavior assays
CSQT-2/14-3-3ζ, MHCC/sh14-3-3ζ, SMMC-7721/ 14-3-3ζ or control cells were subjected to invasionassays using Millicell inserts (Millipore, Billerica, M A, USA), which had been coated with Matrigel (BD Biosciences, Sparks, MD, USA). Cells (5.0 × 10 4 ) were plated onto the upper chambers. After 24 hours, the non-invasive cells on the upper chambers were removed using a cotton swab and the cells that had invaded into the underside of the membrane were stained using crystal violet staining. Cell numbers were plotted as the average number of invaded cells from 5 random microscopic fields.

In vivo metastasis assays
CSQT-2/sh14-3-3ζ, MHCC-LM3/sh14-3-3ζ or control cells were subjected to metastasis assays in vivo. Twenty 6-week-old male nude mice were used in metastasis assays. 2 × 10 6 cells were used to inject into tail vein of each nude mice. The mice were sacrificed at 8 weeks post-injection and examined microscopically using H and E staining for the development of lung metastatic foci. All of the mice were maintained in a pathogen-free facility and used in accordance with the institutional guidelines for animal care.

Immunoprecipitation
Immunoprecipitation was performed on total protein extracts from cells by following a routine protocol [15]. The antibodies that were used are listed in Supplementary  Table 8.

Statistical analyses
Statistical analyses were performed using SAS 9.1.3 software (SAS Institute Inc, Cary, NC). Student's t-tests were used to compare two groups, unless otherwise indicated (χ 2 test). Quantitative variables were analyzed using student's t-test or Pearson's correlation test. Kaplan-Meier and log-rank analyses were used to assess patient survival between subgroups. The data were presented as the means ± SEM, unless otherwise indicated. P < 0.05 was considered to be statistically significant.