Hypoxia-Induced Epithelial-to-Mesenchymal Transition in Hepatocellular Carcinoma Induces an Immunosuppressive Tumor Microenvironment to Promote Metastasis

Portal vein tumor thrombosis (PVTT) is a significant risk factor for metastasis in hepatocellular carcinoma (HCC) patients and is therefore associated with poor prognosis. The presence of PVTT frequently accompanies substantial hypoxia within the tumor microenvironment, which is suggested to accelerate tumor metastasis, but it is unclear how this occurs. Recent evidence has shown that the hypoxia-inducible factor HIF-1a induces epithelial-to-mesenchymal transition (EMT) in tumor cells to facilitate metastasis. In this study, we investigated whether hypoxia-induced EMT in cancer cells also affects immune cells in the tumor microenvironment to promote immunosuppression. We found that hypoxia-induced EMT increased the expression of the CCL20 cytokine in hepatoma cells. Furthermore, coculture of monocyte-derived macrophages with hypoxic hepatoma cells revealed that the expression of indoleamine 2, 3dioxygenase (IDO) was induced in monocyte-derived macrophages in a CCL20-dependent manner. In turn, these IDOexpressing monocyte-derived macrophages suppressed T-cell proliferation and promoted the expansion of immunosuppressive regulatory T cells. Moreover, high CCL20 expression in HCC specimens was associated with PVTT and poor patient survival. Collectively, our findings suggest that the HIF-1a/ CCL20/IDO axis in hepatocellular carcinoma is important for accelerating tumor metastasis through both the induction of EMT and the establishment of an immunosuppressive tumor microenvironment, warranting further investigation into the therapeutic effects of blocking specific nodes of this signaling network. Cancer Res; 76(4); 818–30. 2016 AACR.


Introduction
Hepatocellular carcinoma is one of the most common malignancies worldwide with a poor prognosis (1)(2)(3).Metastasis is the leading cause of death in the majority of hepatocellular carcinoma patients, of which the overall five-year survival rate is merely 5% to 6% (4)(5)(6).The presence of a portal vein tumor thrombus (PVTT) is considered to be a strong predictor of metastasis and one of the most significant factors for a poor prognosis in hepatocellular carcinoma (7).Venous thrombosis is a major hallmark of metastatic hepatocellular carcinoma with 40% to 90.2% of advanced hepatocellular carcinoma patients reporting PVTT (7)(8)(9)(10).The association of PVTT with hepatocellular carcinoma is mainly due to the highly vascular nature of hepatocellular carcinoma tumors (11,12).
The tumor microenvironment plays an important role in cancer development and metastasis (13).Because of the rapid tumor cell growth, the tumor often outpaces its blood supply, leading to substantial hypoxia in the vicinity of hepatocellular carcinoma tumors (14).Hypoxia triggers overexpression of hypoxia-inducible factor 1a (HIF-1a), which has previously been shown to induce epithelial-mesenchymal transition (EMT) of cancer cells in breast cancer (15).During EMT, epithelial cells are converted into motile, invasive mesenchymal cells.EMT is frequently observed at the invasive front of advanced tumors and significantly correlates with metastasis in tumor progression (16)(17)(18).There is also increasing evidence for a functional relationship between EMT and macrophages, as both cell types are found at the invasive front of tumors (19,20).However, it is unclear whether the EMT of cancer cells contributes to the immunosuppressive microenvironment in hepatocellular carcinoma.
Immune cells within the tumor microenvironment also play an important role in cancer development and metastasis.Monocytes can respond to environmental signals and differentiate into tumor supportive macrophages, which modulate the stroma to prompt tumor invasion and metastasis (21,22).These monocytederived macrophages are the most abundant leukocytes that infiltrate the tumor microenvironment (23,24).A subset of monocyte-derived macrophages has increased expression of indoleamine 2, 3-dioxygenase (IDO), a rate-limiting enzyme required for degradation of tryptophan.These IDO þ monocyte-derived macrophages can assist tumors by suppressing T-cell activation and proliferation through degradation of the essential amino acid tryptophan (25).
While the IDO þ monocyte-derived macrophages play a vital role in establishing and maintaining an immunosuppressive tumor environment, the mechanism through which the microenvironment educates the monocytes and induces a tolerogenic state remains unclear.In this study, we explored that the interaction between hypoxia-induced EMT of cancer cells and monocyte-derived macrophages.We showed that increased levels of the CCL20 cytokine released from cancer cells significantly induced the expression of IDO in monocyte-derived macrophages, helping to form an immunosuppressive microenvironment to accelerate metastasis.

Patients and specimens
Tumor tissue samples were obtained from the Second Affiliated Hospital of Zhejiang University School of Medicine.Ninety patients with hepatocellular carcinoma underwent curative resection between 2007 and 2010, and samples from these patients were used for IHC and quantitative RT-PCR (qRT-PCR).This project was approved by the Ethics Committee of Second Affiliated Hospital of Zhejiang University School of Medicine.All samples were anonymously coded in accordance with local ethical guidelines (as stipulated by the Declaration of Helsinki), and written informed consent was obtained.

Cell culture and treatment
HL-7702 and human hepatoma cell line (Huh-7) were obtained from the Shanghai Institute for Biological Science (Shanghai, China).Hep-G2 cells were obtained from ATCC.They were maintained in DMEM supplemented with 10% FBS (Gibco) and 1% penicillin/streptomycin (Sigma).All cell lines were authenticated by a professional biotechnology company in 2015, and used in our previous study (26).Cells were exposed to hypoxia (1.0% O 2 ) in a hypoxic chamber (Thermal Tech) for the indicated time period.

Chromatin immunoprecipitation assay and luciferase reporter assays
Putative hypoxia response element (HRE)-like sequences were identified in the CCL20 promoter sequence using the regulatory sequence analysis tools (http://rsat.ulb.ac.be/rsat/).After treatment with 100 mmol/L CoCl 2 for 12 hours, Hep-G2 and Huh-7 cells were cross-linked and processed according to the Millipore EZ-ChIP Assay Kit protocol (Millipore).The mouse anti-human HIF-1a ChIP validated antibody (ab1, Abcam), and mouse IgG (negative control) were used.PCR analysis for CCL20 was carried out using the primers (forward: 5 0 -GACCTTTGTATCGCTGTTAAT; reverse: 5 0 -AGTAGCAGCACTGACATCAAA) and the PCR products were analyzed by agarose gel electrophoresis.
The putative (AGAAGGCGTGTTGCCACA) HRE-like sequence synthesized by Hanbio was ligated into the pGL3-basic plasmid (Promega) to construct the CCL20 promoter luciferase reporter plasmid.A VEGF reporter plasmid, containing a À1,005/þ306 fragment from the human VEGF promoter, was constructed as described previously (27).Luciferase reporter assays were carried out in 24-well plates, and 250 ng of CCL20 reporter plasmid or VEGF reporter plasmid was cotransfected with 750 ng HIF-1a plasmid (Vigene Bioscience CH886250) and 25 ng pRL-TK reporter plasmid (Millipore).Reporter activity was evaluated using the dual luciferase reporter system (Promega).

In vitro coculture model
Peripheral blood mononuclear cells (PBMC) were isolated by Ficoll density-gradient centrifugation.Monocytes were purified from PBMCs using anti-CD14 magnetic beads (Miltenyi Biotec), or fluorescence-activated cell sorting (FACS).Monocytes were cultured with cell lines expressing GFP at a ratio of 3:1 for 3 days.Then, the medium was changed to medium with CCL20 (100 ng/mL; PeproTech), medium with IFNg (100 IU/mL; PeproTech), medium with fludarabine (50 mmol/L; Selleckchem), or the cells were directly cultured with hypoxic tumor cells for one day.After one day, fresh medium was added and the cells were cultured for another day.Then, we sorted the GFP-expressing hepatoma cells from the non-GFP-expressing monocytes using FACS (Supplementary Fig. S3C).The supernatants were harvested and kynurenine was detected (28).

In vitro T-cell culture models
Circulating T cells were purified using the Pan T Cell Isolation Kit II (Miltenyi Biotec) or FACS sorting.Autologous or treated monocytes were pretreated with 10 mg/mL mitomycin C (Sigma) for 30 minutes, washed twice, and cocultured for 7 days with polyclonal-stimulated (10 mg/mL anti-CD3 and 10 mg/mL anti-CD28) and prelabeled CFSE T cells at a ratio of 1:10 in the presence or absence of 100 mmol/L (1-MT; Sigma).Functional assays were carried out using the regulatory T cell (T reg ) suppression inspector (Miltenyi Biotec), with five ratios of T eff: T reg cells (0:1, 1:0, 1:1, 4:1 and 8:1).

ELISA
The supernatants were collected and stored at 4 C. Concentrations of CCL20, IL10, IL12, IL6, and VEGF in the conditioned media were detected using ELISA kits (R&D Systems, Inc.).
Paraffin-embedded hepatocellular carcinoma tissue samples were cut into 5-mm sections and processed for IHC (29).The slides were incubated with human anti-CCL20 (1:1,000) or HIF-1a (1:1,000) antibodies.We defined samples without any staining as 1 score.Other samples were defined as low (2 scores), medium (3 scores), or high (4 scores) levels of expression.The scoring system is illustrated using representative samples of each score in Supplementary Fig. S6 and was evaluated by two independent pathologists.

Quantitative reverse transcription PCR
Tumor sample or cells were lysed as previous described.qRT-PCR was performed on the ABI 7900 Prism HT (Applied Biosystems), followed by melting curve analysis.The DC t method was used to assess the gene expression fold change among groups.Three independent experiments were performed.

Statistical analysis
Statistical calculations were performed using Prism 5 software (GraphPad).Data were expressed as mean and SD, unless otherwise indicated.Continuous variables were evaluated using an unpaired Student t test for comparisons between two groups.Categorical variables were compared using a c 2 test (or Fisher exact test).Multivariate analysis was performed using forward stepwise logistic regression analysis.Survival analysis was conducted using the Kaplan-Meier method with the log-rank test.Two-sided tests were performed with a P < 0.05 indicating statistical significance.

HIF-1a is positively associated with PVTT in hepatocellular carcinoma patients
To evaluate the clinical outcomes of PVTT in hepatocellular carcinoma patients, we analyzed the survival of 90 hepatocellular carcinoma patients who received surgery in the Second Affiliated Hospital of Zhejiang University School of Medicine in China.
The results showed that 21 patients within PVTT group had a significantly worse prognosis than the other patients (Fig. 1A).As HIF-1a is one of the most potent proteins linked to liver metastasis and tumor progression, we examined HIF-1a mRNA level in normal livers, tumors without PVTT, and tumors with PVTT (n ¼ 21 in each group; Fig. 1B).HIF-1a mRNA level was significantly higher in the group of tumors with PVTT.Moreover, we found significantly strong nuclear staining for HIF-1a in the tumors with PVTT (Fig. 1C and 1D).

HIF-1a induces EMT in hepatocellular carcinoma cell lines
We examined the potential for hypoxia to induce EMT in hepatocellular carcinoma cell lines, Huh-7 and Hep-G2.After exposure to 1.0% oxygen for 24 hours, the morphology of the Huh-7 and Hep-G2 cells became significantly stretched and elongated (Fig. 1E).The epithelial biomarker E-cadherin was downregulated, while simultaneously the mesenchymal biomarkers vimentin and Snail increased (Fig. 1F).In addition, immunofluorescence showed weaker E-cadherin membrane localization and stronger intracytoplasmic localization of vimentin (Fig. 1G).These results are all typical of events that occur during EMT of tumor cells.

HIF-1a induces the expression of the cytokine CCL20 by mesenchymal cancer cells
To investigate whether there was a change in the cytokine profile associated with hypoxia-induced EMT of cancer cells, the hypoxic medium from Hep-G2 cells was analyzed using the RayBio Human Cytokine Antibody Array (AAH-CHE-1; Fig. 2A).The expression of six cytokines [CCL2, CCL20, CXCL1 (GROa), CXCL8 (IL-8), CXCL9, and CXCL16] increased significantly compared with the condition medium of untreated Hep-G2 cells, and the levels of CCL20 increased most dramatically.We used qRT-PCR to confirm an increase in CCL20 gene expression at 16 and 24 hours in both Huh-7 cells and Hep-G2 cells under hypoxic conditions (Fig. 2B).
To confirm whether the HIF-1a was responsible for this change in cytokine levels, HIF-1a accumulation and CCL20 levels were examined over time in Huh-7 and Hep-G2 cells exposed to 1% oxygen (Fig. 2C).HIF-1a reached a peak at 16 hours, and then gradually declined, which was consistent with the dynamic change in CCL20 gene expression.However, the protein levels of CCL20 did not reach a maximum until 24 hours.Therefore, there may be a time delay between HIF-1a-induced gene expression and secreted protein levels for CCL20.Knockdown of HIF-1a by siRNA (100 nmol/L) in Hep-G2 and Huh-7 cells resulted in a significant decrease in the expression of CCL20 compared with the control (ncRNA at 100 nmol/L; Fig. 2D).After HIF-1a knockdown, there was no upregulation of EMT markers observed either by Western blot or immunofluorescence under hypoxic conditions.Furthermore, the Transwell assay showed that HIF-1a knockdown prevented hypoxic Huh-7 or Hep-G2 cells passing through the Matrigel-coated membranes (Supplementary Fig. S1A-S1C).
To understand whether there was direct association between HIF-1a and CCL20, we overexpressed HIF-1a (using a HIF-1a plasmid) and found that CCL20 dramatically increased both in Hep-G2 and Huh-7 cells (Fig. 2E).To further confirm that the CCL20 gene is regulated directly by hypoxia through a HRE [(A/G)CGTG); refs.[30][31][32], high-scoring HRE-like sequences in the CCL20 gene promoter were identified by bioinformatics analysis.Moreover, a ChIP assay using an anti-HIF-1a showed that the putative HRE in the promoter of the CCL20 gene (AGAAGGCGTGTTGCCACA) was captured (Fig. 2F).To determine whether the putative HRE plays a functional role in hypoxia-dependent transcriptional activation, we cotransfected cells with the HIF-1a plasmid and a CCL20 or VEGF luciferase reporter.Inducibility of the CCL20 reporter by HIF-1a overexpression was compared with that of VEGF reporter, which contains a functional HIF-1a-binding site.We detected a substantial, significant increase of CCL20 promoter transactivation (approximately 5-10 fold relative to control cells) in response to HIF-1a overexpression.This result confirms that the putative HRE (AGAAGGCGTGTTGCCACA) in CCL20 was activated by HIF-1a (Fig. 2G).Taken together, these results suggest that a functional HRE motif exists in the promoter fragment of the CCL20 gene that mediates hypoxiadependent CCL20 expression.
To evaluate the potential relationship of HIF-1a and CCL20 in the tumor microenvironment, we analyzed their expression in adjacent sections of hepatocellular carcinoma tissues by staining for HIF-1a and CCL20 (Supplementary Fig. S2).HIF-1a was expressed at higher levels at the invasive front (Supplementary Fig. S2).We observed the same gradient for CCL20 with higher levels at the tumor edge, where cancer cells displayed more mesenchymal features and expressed high levels of HIF-1a, compared with the center of the tumor (Supplementary Fig. S2).These data indicated that hypoxia-induced mesenchymal cancer cells secreted more CCL20 at the invasive front due to increased levels of HIF-1a.

Tumor-derived CCL20 upregulates IDO expression in monocyte-derived macrophages
Tumor microenvironments recruit monocytes from blood, and educate them, converting them into tumor supportive macrophages.Previous studies used the tumor culture supernatants from hepatoma cells to induce the macrophages (22,33).However, this model may not accurately reflect the effects of cell-to-cell contacts, which are important for cell surface marker function.Therefore, in this study, we cocultured isolated monocytes directly with hepatoma cells (Huh-7 and Hep-G2).After 72 hours, the monocytes cocultured with hepatoma cells, but not with HL-7702, exhibited a high expression of CD206 and CD163 (Supplementary Fig. S3A).Moreover, the levels of HLA-DR were reduced in monocytes exposed to hepatoma cells compared with HL-7702 (Supplementary Fig. S3A).The activation of suppressive monocyte-derived macrophages was also confirmed by IL10, IL12, IL6, and VEGF cytokine production (Supplementary Fig. S3B).
To further identify the effect of the hypoxia-induced cytokines in the tumor microenvironments, we directly cultured monocytederived macrophages with hypoxic hepatoma cells and then sorted the monocyte-derived macrophages from the tumor cells.We found that the expression of IDO increased in the monocytederived macrophages (Fig. 3A).As Lutz and colleagues reported (34), CD14 þ monocytes are heterogeneous, and include different subsets of monocytes.Therefore, we sorted the major subset (the classical CD14 þþ CD16 À subset) from the two minor CD16 þ subsets (the intermediate CD14 þþ CD16 þ and the nonclassical CD14 þ CD16 þþ subsets).Consistent with a previous report (35), the nonclassical CD14 þ CD16 þþ monocytes expressed the lowest levels of CD163 and CD206 compared with the other two subsets, which may be partially due to the proinflammatory properties of this subset (36).When we cocultured these three subsets of monocytes with tumor cells, the expression of CD163 and CD206 increased considerably (Fig. 3B).Furthermore, we sorted the three different subsets of cells and cocultured them separately with hypoxic hepatoma cells.We found that the expression of IDO was almost the same among the three subsets (Fig. 3C).Therefore, all three subsets of monocytes can be affected by the coculture system.
Next, we extensively explored which cytokine was responsible for this effect, and found that CCL20, and not CCL2, CXCL1, CXCL8, CXCL9, or CXCL16, significantly induced the expression of IDO in monocyte-derived macrophages (Fig. 3D).The expression of IDO was confirmed by confocal microscopy results (Supplementary Fig. S4A).In parallel with IDO expression, we observed an increase in its catalyzed product (kynurenine) in the culture supernatant (Supplementary Fig. S4B).To further confirm this, we added the neutralizing anti-CCL20 antibody to the hypoxic medium and found that the expression of IDO was reduced to almost same level as the control (Fig. 3E).
We also evaluated CCL20 and IDO levels in human hepatocellular carcinoma tissue.Immunofluorescence of frozen sections showed that most CCL20 surrounded IDO at the invasive front of the tumor (Fig. 3I).We measured mRNA levels of IDO by qRT-PCR, using the same set of tissue samples shown in Fig. 1A, and found that IDO (Fig. 3F) and CCL20 (Fig. 3G) mRNA significantly increased in the group of tumors with PVTT.Furthermore, the IDO mRNA was positively associated with the CCL20 mRNA level in tumors with PVTT group (Fig. 3H).However, there was no significant association between IDO and CCL20 expression in the group with normal liver tissue (Supplementary Fig. S5A) and the group with tumors without PVTT (Supplementary Fig. S5B).Collectively, these results provided strong support for our hypothesis that the elevation of IDO was likely the result of the elevation of CCL20 as the etiologic factor in the development of tumor thrombus.

Phosphorylation of STAT1 is required for upregulation of IDO expression
Previous studies indicated that JAK-STAT pathway and NF-kB signaling pathways were involved in the regulation of IDO expression in macrophages and dendritic cells (DC).STAT1 activation was accepted as the canonical mechanism for mediating the induction of IDO expression by IFNg.However, the molecular mechanism behind the upregulation of IDO expression in monocyte-derived macrophages after CCL20 treatment is still unclear.Therefore, we studied the expression and phosphorylation of STAT1, STAT3, and P65 (Fig. 4A).The treatment of CCL20 significantly induced the expression and nuclear translocation of phosphorylated-STAT1 (pSTAT1) in monocyte-derived macrophages (Fig. 4B and D).Fludarabine, the specific antagonist of STAT1 phosphorylation, downregulated expression of pSTAT1 and IDO proteins (Fig. 4C), and inhibited STAT1 nuclear translocation (Fig. 4D).These results suggest that CCL20-induced expression of IDO might be a result of activation of phosphorylation and nuclear translocation of STAT1.

IDO þ monocyte-derived macrophages induce T cell dysfunction via an IDO-dependent manner
To define whether CCL20-treated monocyte-derived macrophages exhibited an immunosuppressive function, we cocultured monocyte-derived macrophages with T cells to detect proliferation and cytokine production.Compared with controls, hypoxic tumor-treated monocyte-derived macrophages slightly inhibited the proliferation of T cells.The addition of CCL20-treated IDO þ monocyte-derived macrophages further enhanced the inhibition (Fig. 5A and B).When IDO activity was blocked with the competitive inhibitor 1-MT, the effect of T-cell inhibition was attenuated.On the other hand, CCL20-treated IDO þ monocytederived macrophages significantly decreased secretion of IFNg both in CD4 þ and CD8 þ T cells (Fig. 5C and D).As expected, 1-MT suppressed the inhibitory effect of monocyte-derived macrophages on IFNg secretion.We also examined whether the CCL20-treated IDO þ monocytederived macrophages influence the generation of immunosuppressive T reg -like cells.Autologous-sorted CD4 þ CD25 À effector T cells were cocultured with IDO þ monocyte-derived macrophages.After 7 days, we found high expression of Foxp3 in T cells (Fig. 5E-G).Foxp3 was significantly induced in the (Continued.)HIF-1a mRNA was undetectable in one normal liver tissue and one tumor with PVTT.C, IHC staining showing the accumulation of HIF-1a in normal liver, tumor without PVTT, and tumor with PVTT.Representative images at Â200 magnification from one set of the samples are shown.D, statistical analysis of the pathologic scores of HIF-1a in the various groups.Data are expressed as means AE SD and compared using the unpaired t test ( Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ , P < 0.001).E, bright field microscopy showing the morphologic changes that occur when Huh-7 and Hep-G2 cells are cultured for 24 hours in 1.0% oxygen.F, detection of HIF-1a, E-cadherin, vimentin, and Snail proteins using Western blot analysis.GAPDH was used as loading control.One of three representative Western blots is shown.F, immunofluorescence staining showing that the staining of E-cadherin (green) and vimentin (red) changes under normal and hypoxic conditions in both Huh-7 and Hep-G2 cells.CD4 þ CD25 þ cells other than in CD4 þ CD25 À cells (Fig. 5G).To further determine the function of these T reg -like cells, we sorted the CD4 þ CD25 þ CD127 À cells from the coculture system.We cultured the sorted autologous CD4 þ CD25 À cells with the T reg Suppression Inspector (Miltenyi Biotec) at five different ratios of T eff cells: T reg cells (i.e., 0:1, 1:0, 1:1, 4:1, and 8:1).We found that the proliferation of T cells was significantly inhibited by the T reg -like cells compared with the control (Fig. 5H and I).These findings suggest that IDO þ monocyte-derived macrophages exert an immunosuppressive effect on T cells in an IDO-dependent manner.

High amounts of CCL20 predict poor patient survival
To investigate the clinical relevance of elevated CCL20 levels in hepatocellular carcinoma cancer patients, we performed immunohistochemical staining in the 90 hepatocellular carcinoma samples.We divided patients into "low" and "high" expression groups, based on the median value of CCL20 levels, to assess the association of expression of CCL20 with clinical characteristics (Fig. 6A; Supplementary Fig. S6).We found that the high expression of CCL20 was significantly correlated with the tumor PVTT group (P ¼ 0.030; Supplementary Table.S1).There was a significant inverse correlation of tumor CCL20 expression with both þ monocytes cocultured with Huh-7 and Hep-G2 cells as detected by Western blot analysis.E, the expression of IDO protein after using the neutralizing antibody for CCL20 as detected by Western blot analysis.F and G, relative mRNA levels of CCL20 and IDO, measured by qRT-PCR and normalized b-actin, in tissue specimens from the same set of patients from Fig. 1B.Data are expressed as means AE SD of three independent experiments and were compared using an unpaired t test ( Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ , P < 0.001).CCL20 and IDO mRNA was undetectable in one normal liver tissue.H, the correlation between IDO and CCL20 expression for each individual in the tumor with PVTT group assessed by linear regression.I, physical contact between CCL20 þ cells (green) and IDO þ cells (red) in the invading edge of cancer in hepatocellular carcinoma samples.One of 10 representative images from each set of samples is shown.Magnification at Â200 for the left image and Â400 for the two inset boxes (red and green).One of three representative results is shown in B-E.
We also analyzed the relationship between CCL20 and IDO expression in hepatocellular carcinoma patients with or without vascular invasion from National Center for Biotechnology Information Gene Expression Omnibus database (GSE20238; ref. 37).The database includes a total of 91 hepatocellular carcinoma patients and the associated vascular invasion data for 79 patients.Our analysis showed that the CCL20 levels significantly correlated with IDO in patients with vascular invasion (P ¼ 0.0443; Fig. 6D).In contrast, no significant association was observed in patients without vascular invasion (P ¼ 0.7949; Fig. 6E).These data strongly suggest that increased tumor CCL20 abundance is a significant and independent predictor of poor survival in hepatocellular carcinoma patients.

Discussion
Over the past decade, much research into hepatocellular carcinoma has focused on the cancer cells themselves, without taking into account the unique but complex tumor microenvironment.Hypoxia in the tumor microenvironment plays an important role in driving cell metastasis (38,39).Our study demonstrated that PVTT can cause substantial hypoxia in hepatocellular carcinomas compared with the hepatocellular carcinomas without PVTT.However, whether the reverse is true (i.e., that different levels of HIF-1a directly induce the formation of PVTT) remains unclear.Indeed, some patients with high HIF-1a expression still existed in the PVTTÀ group compared with the PVTTþ group (Fig. 1B).Moreover, based on our findings, we cannot conclude whether HIF-1a has on-off or a more graduated effect on PVTT formation.
In response to hypoxia, HIF-1a is known to accumulate and induce EMT in tumor cells to facilitate metastasis (40).During EMT, tumor cells acquire cell motility, with decreased adhesive ability and a rearranged cytoskeleton (40,41).EMT is also thought to have another function: to create an immunosuppressive microenvironment by educating the infiltrating immune cells (42).We examined whether a change in the secretory cytokine profile during EMT may be the key factor for educating immune cells in the microenvironment.We found hypoxia-induced mesenchymal cancer cells elevated the levels of CCL20 in the microenvironment.Moreover, we identified a functional HRE within the CCL20 promoter for the first time.Therefore, we propose that hypoxia induces HIF-1a, which interacts with this HRE to stimulate the transcription of CCL20.Together, our results suggest that CCL20 expression is under the direct control of HIF-1a via a HRE in the CCL20 gene promoter.Among the cytokines upregulated during hypoxic EMT, CCL20 was required for the induction of IDO expression in monocyte-derived macrophages.Indeed, the analysis of our clinical samples showed that CCL20 was significantly associated with IDO expression in hepatocellular carcinoma with PVTT.Moreover, as monocyte-derived macrophages are typically composed of three subsets, which are characterized by different levels of CD14 and CD16 expression, we assessed whether they showed differential responses to the hypoxia.We showed that all three subsets of monocytes could be affected by our coculture system.Although the subsets showed differential expression of CD163 and CD206, they all showed increased IDO expression.Therefore, CCl20 is required for the induction of IDO expression in monocyte-derived macrophages.
Previous studies showed that the CCL20-CCR6 axis is important for the recruitment of circulating T regs into the tumor microenvironment (43).Here, we demonstrated that CCL20-treated IDO þ monocyte-derived macrophages significantly inhibited the T-cell proliferative activity compared with untreated monocytederived macrophages in an IDO-dependent way.Inhibition of IDO activity by 1-MT significantly attenuated monocyte-derived macrophage-mediated suppression on T-cell proliferation and production of IFNg in CD4 þ and CD8 þ cells.On the other hand, we found the CD25 þ Foxp3 þ T reg -like cells significantly increased in samples cultured with CCL20-treated IDO þ monocyte-derived High amounts of CCL20 are predictive of poor patient survival.A, hepatocellular carcinoma tissue sections (n ¼ 90) were stained with an antibody against CCL20.Representative images (magnification, Â200) from a patient with "low" (weak) CCL20 levels and from a patient with "high" (strong) CCL20 levels are shown.Lower panel represents magnified views (Â400) indicated by boxes in the top panel.B and C, the disease-free survival and cumulative overall survival of hepatocellular carcinoma patients with high or low CCL20 expression levels (based on median CCL20 levels) estimated using the Kaplan-Meier method and compared by the log-rank test in the same set of patients.D and E, the relationship between CCL20 and IDO expression in hepatocellular carcinoma patients without and with vascular invasion from the National Centre for Biotechnology Information Gene Expression Omnibus database (GSE20238) assessed by linear regression.
macrophages compared with those cultured with untreated monocyte-derived macrophages.The potential mechanism for the increase in CD25 þ Foxp3 þ T reg -like cells may be the production of TGFb, IL10, and thrombospondin-1 (TSP1), which have been reported to induce T regs (44,45).Together, these results indicate that the release of CCL20 during hypoxic-EMT plays an important role in creating a tumor immunosuppressive microenvironment.
In macrophages and DCs, STAT1 activation has been shown to induce IDO expression via type I or type II IFN stimulation (46).In addition, posttranslational modification of STAT3 by phosphorylation and/or acetylation has been demonstrated to be essential for the expression of IDO in DCs in response to commensal bacteria or allograft transplant (47)(48)(49).In this study, we examined the expression and phosphorylation of STAT1 and STAT3 and found that the level of pSTAT1, but not pSTAT3, significantly increased in CCL20-treated monocyte-derived macrophages.This suggests that CCL20 induced expression of IDO might be due to phosphorylation of STAT1.Therefore, CCL20 may act in similar way to the classical IFNg signaling pathway that has previously been reported to increase IDO expression.
Macrophages are versatile, plastic cells that can respond to environmental signals and differentiate into a tumor supportive phenotype (24,50).Several studies identified the interaction between cancer cells and macrophages, using condition medium from cancer cell line to induce macrophages (22,33).However, these studies may not reflect the real status in a tumor microenvironment, where they would be in direct cell-to-cell contact.Therefore, it is likely that this induction effect will be reduced if the cancer and immune cells are separately cultured.To mimic a hepatocellular carcinoma cell condition microenvironment in vitro, we cocultured monocytes with hepatoma cells.We found that IDO was weakly expressed in the monocyte-derived macrophages after coculture with hepatoma cells, and there was no IDO expression in macrophages induced by the condition medium alone (33).
In conclusion, our results provide new insights into the action of CCL20 on monocyte-derived macrophages that was exercised to counteract effective T-cell responses.The hypoxia-induced EMT of cancer cells helps to educate newly recruited monocytes by secreting CCL20, which leads to IDO upregulation in the monocytes.This suppressive IDO þ monocyte-derived macrophages, act to suppress T-cell responses and promote to tolerance to tumor antigens.In this way, IDO þ monocyte-derived macrophages create immunosuppressive conditions in the tumor microenvironment and promote metastasis (Fig. 7).As elevated CCL20 levels were associated with a poor prognosis in hepatocellular carcinoma patients, preventing CCL20 elevation may stop the creation of an immunosuppressive tumor microenvironment and slow metastasis.Therefore, targeting CCL20 may be a promising strategy for treating human hepatocellular carcinoma in the future.

Figure 1 .
Figure 1.HIF-1a is positively associated with PVTT development in hepatocellular carcinoma patients.A, the cumulative overall number of hepatocellular carcinoma patients with PVTT compared with those without PVTT over time, estimated using the Kaplan-Meier method.B, relative mRNA levels of HIF-1a measured by qRT-PCR and normalized to b-actin, in normal livers, and in hepatocellular carcinoma tumors without and with PVTT.Data are expressed as means AE SD and compared using the unpaired t test ( Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ , P < 0.001).(Continued on the following page.)

Figure 2 .
Figure 2.Relative levels of CCL20 increase most dramatically among the cytokines under hypoxia.A, cytokine array from the condition medium of Hep-G2 cells and Hep-G2 induced to undergo EMT by culturing for 24 hours in 1.0% oxygen.The relative signal intensity of each of the indicated cytokines is shown.B, relative levels of CCL2, CCL20, CXCL1, CXCL8, CXCL9, and CXCL16 in Huh-7 and Hep-G2 as measured by qRT-PCR (normalized to b-actin) after 0, 16, and 24 hours of culturing in 1.0% oxygen.C, CCL20 concentration (detected by ELISA) and HIF-1a accumulation (detected by Western blot analysis) in Huh-7 and Hep-G2 cells exposed to 1% oxygen from 0 to 48 hours.GAPDH was used as loading control for the Western blot analysis.D, expression of CCL20 after HIF-1a knockdown by siRNA #1 and #2 (100 nmol/L) compared with controls (ncRNA at 100 nmol/L) in Hep-G2 and Huh-7 cells.HIF-1a knockdown was confirmed using Western blot analysis, with GAPDH as a loading control.E, expression of CCL20 (detected by ELISA) and HIF-1a (detected by Western blot analysis) in Huh-7 and Hep-G2 cells after transfection with the HIF-1a plasmid.GAPDH was used as loading control for the Western blot analysis.F, ChIP analysis using anti-HIF-1a to ascertain the existence of the HRE in the promoter of the CCL20 gene.PCR results showing that one fragment containing the putative HRE could be precipitated after treatment of Hep-G2 and Huh-7 cells with CoCl2 100 mmol/L for 12 hours.G, luciferase activities of CCL20 or VEGF cotransfected with HIF-1a plasmid.Results in A-G are expressed as means AE SD of at least three separate experiments and compared using the unpaired t test ( Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ P < 0.001).

Figure 3 .
Figure 3. Tumor-derived CCL20 upregulates IDO expression in monocyte-derived macrophages.A, CD14 þ monocytes cocultured with Huh-7 and Hep-G2 cells in the presence or absence of hypoxic tumor (HT) cells, CCL20, or IFNg.IDO protein was detected by Western blot analysis in each case, with GAPDH used as a loading control.Western blot results are from at least 10 individual experiments.B, three subsets of CD14 þ monocytes were sorted and cocultured with Huh-7 and Hep-G2 cells, and the expression of CD163 and CD206 was analyzed by flow cytometry (solid line) compared with corresponding cells before culture (dash line).C, the expression of IDO was detected in the three different subsets of monocytes after coculture with hypoxic tumor (HT).D, levels of various cytokines in CD14

Figure 4 .
Figure 4. Phosphorylation of STAT1 is required for upregulation of IDO expression.A, protein levels of pSTAT3, STAT3, pSTAT1, STAT1, pP65, and P65 in CD14 þ monocytes cocultured with Huh-7 and Hep-G2 cells in the presence or absence of hypoxic tumor (HT) cells, CCL20, or IFNg as detected by Western blot analysis.GAPDH was used as a loading control.B, statistical analysis of the pSTAT3, STAT3, pSTAT1, STAT1, pP65, and P65 protein levels in the various groups is shown.Protein levels were quantified by densitometry, corrected for the sample load based on GAPDH expression, and expressed as fold-increase or decrease relative to the control lane.Data are expressed as means AE SD and were compared using the unpaired t test ( Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ , P < 0.001).C, protein levels of pSTAT1, STAT1, and IDO in CD14 þ monocytes cocultured with Hep-G2 cells, CCL20, or fludarabine (a STAT1 antagonist at 50 mmol/L) as detected by Western blot analysis.D, immunofluorescent images showing the nuclear location of STAT1 in monocytes after treatment of fludarabine (50 mmol/L) in the presence of CCL20.One of three representative results is shown in A, C, and D.

Figure 6 .
Figure 6.High amounts of CCL20 are predictive of poor patient survival.A, hepatocellular carcinoma tissue sections (n ¼ 90) were stained with an antibody against CCL20.Representative images (magnification, Â200) from a patient with "low" (weak) CCL20 levels and from a patient with "high" (strong) CCL20 levels are shown.Lower panel represents magnified views (Â400) indicated by boxes in the top panel.B and C, the disease-free survival and cumulative overall survival of hepatocellular carcinoma patients with high or low CCL20 expression levels (based on median CCL20 levels) estimated using the Kaplan-Meier method and compared by the log-rank test in the same set of patients.D and E, the relationship between CCL20 and IDO expression in hepatocellular carcinoma patients without and with vascular invasion from the National Centre for Biotechnology Information Gene Expression Omnibus database (GSE20238) assessed by linear regression.