Hypoxia-inducible factor-1α/interleukin-1β signaling enhances hepatoma epithelial–mesenchymal transition through macrophages in a hypoxic-inflammatory microenvironment

CELL CULTURE AND REAGENTS

Cell culture methods are described in the Supporting Information. For macrophage differentiation, THP-1 cells were stimulated overnight with 200 nM phorbol 12-myristate 13-acetate (Sigma-Aldrich, St. Louis, MO). Human monocyte-derived macrophages (MDMs), mouse bone marrow–derived macrophages, and mouse peritoneal macrophages were obtained as described in the Supporting Information. For Toll-like receptor 4 (TLR4) signaling inhibition, Pepinh-TIR domain–containing adapter-inducing interferon-β (TRIF), Pepinh-MYD, CLI-095, and antihuman TLR4 antibody were purchased from InvivoGen (San Diego, CA).

GENERATION OF NECROTIC DEBRIS AND CONDITIONED MEDIUM

To generate cancer cell necrotic debris, Huh-7 cells were cultured in hypoxia for 48 hours after reaching 100% confluence. For apoptotic cell debris, cells were treated with 6 mM cisplatin (Sigma-Aldrich) for 24 hours. Cells and supernatants were collected and centrifuged (3000 g, 15 minutes). Cell debris was then washed three times with phosphate-buffered saline to completely remove soluble components. The debris was stored at −30°C and resuspended in Roswell Park Memorial Institute 1640 when used. The induction efficiency of necrotic and apoptotic debris was confirmed by flow cytometry with fluorescein isothiocyanate–annexin V/propidium iodide staining (BD Biosciences). The conditioned medium was harvested after cells were cultured for 48 hours, followed by centrifugation (3000 g, 10 minutes) and filtered through 0.2-μm pore filters to remove debris. Before use, the conditioned medium was mixed 1:1 with complete medium. All samples were tested for the presence of bacterial endotoxin using the Limulus amebocyte lysate test (Genscript, Nanjing, China) for qualification (<0.02 EU/mL).

TRANSWELL ASSAYS

Migration and invasion capacities of cells were detected by transwell assays as described in the Supporting Information.

TRANSFECTION, RNA INTERFERENCE, AND LUCIFERASE REPORTER ASSAYS

Small interfering RNA (siRNA) transfection was performed as reported.6 For the luciferase reporter assays, cells were seeded into a 96-well plate and transfected with hypoxia-response element–luciferase reporter plasmids (a gift from Navdeep Chandel14; Addgene plasmid 26731). Luciferase activities were measured using the Luciferase Reporter Assay System (Promega, Madison, WI). Nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) reporter plasmids were a gift from Prof. Qingqing Wang at Zhejiang University School of Medicine.

IMMUNOBLOTTING, IMMUNOFLUORESCENCE, AND IMMUNOHISTOCHEMISTRY

Immunoblotting and immunofluorescence were performed as described.6 Paraffin-embedded HCC tissue samples were cut into 5-μm sections and processed for immunohistochemistry. Detailed procedures are shown in the Supporting Information.

ENZYME-LINKED IMMUNOSORBENT ASSAY

Supernatants were collected and stored at −80°C after removal of cells and debris by centrifugation. Concentrations of IL-1β, IL-6, and tumor necrosis factor-α (TNF-α) were detected using specific enzyme-linked immunosorbent assay (ELISA) kits (Biolegend, San Diego, CA, and R&D Systems, Minneapolis, MN). For IL-1β detection, 5 mM adenosine 5′-triphosphate (Sigma-Aldrich) was added for 30 minutes before sample collection. Experiments were performed according to the manufacturer's instructions.

QUANTITATIVE REAL-TIME PCR

The relative expression of specific genes was measured with the Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) using the comparative threshold cycle method. Detailed procedures are described in the Supporting Information. All primers were synthesized by Sangon Biotech (Shanghai, China; Supporting Table S1).

FLOW CYTOMETRY

Single-cell suspensions were generated, and cells were stained with specific antibodies according to the manufacturer's instructions. Antibodies including human leukocyte antigen (HLA) A, HLA-B, HLA-C (W6/32), cluster of differentiation 68 (CD68; FA-11), CD45 (2D1), CD206 (19.2), CD163 (GHI/61), CD11b (ICRF44), and IL-1β (JK1B-1) were purchased from eBioscience (San Diego, CA) or Biolegend. Isotype controls were used for all staining. Cell sorting was performed using a FACSAria II cell sorter (BD Biosciences). Flow-cytometric analysis was performed on a BD FACS Canto-II Flow Cytometer (BD Biosciences).

MOUSE MODELS

Male BALB/c nude mice, 5-6 weeks old, were purchased from Shanghai Experimental Animal Center (Shanghai, China). Spontaneous HCC mice (Mst1^–/–;Mst2^–/flox;Alb-Cre) were a gift from Professor Bin Zhao (Life Science Institute Zhejiang University). Detailed procedures are shown in the Supporting Information.

HUMAN TISSUE ACQUIREMENT

Human tumor samples were obtained from patients with HCC who underwent curative resection between 2013 and 2017 at the Second Affiliated Hospital, Zhejiang University School of Medicine. Informed consent of patients was obtained. The protocol was approved by the Ethics Committee of the Second Affiliated Hospital, Zhejiang University School of Medicine. The clinicopathological information of all samples used is shown in Supporting Tables S2-S6. HCC tissue microarrays were purchased from Shanghai Outdo Biotech (Shanghai, China).

STATISTICAL ANALYSIS

The data were analyzed using GraphPad Prism 6 (San Diego, CA) and are presented as means ± standard deviation or standard error of the mean (SEM). Continuous variables were analyzed using an unpaired Student t test or Mann-Whitney test as indicated to compare between two groups. Categorical variables were compared using a chi-squared test or Fisher exact test. Survival analysis was conducted using the Kaplan-Meier method with a log-rank test. P < 0.05 was considered statistically significant.

MODERATE HYPOXIA FACILITATES IL-1β SECRETION BY MONOCYTES AND MACROPHAGES THROUGH HIF-1α

To investigate the impact of hypoxia on proinflammatory cytokine secretion of the monocyte–macrophage system, we first examined the levels of IL-1β and IL-6 secreted by THP-1, a widely used human monocyte cell line. Compared with unstimulated THP-1 cells cultured under normoxia, IL-1β secretion increased mildly but significantly with desferrioxamine treatment, a chemical mimicry of hypoxia (Fig. 1A). Similarly, moderate hypoxia mildly increased the level of IL-1β rather than IL-6 in both THP-1 cells and THP-1-derived macrophages (Fig. 1B). When activated by lipopolysaccharide (LPS), THP-1-derived macrophages and human MDMs were much more sensitive to hypoxia-enhanced IL-1β secretion (Fig. 1C). Similar effects were observed in mouse bone marrow–derived macrophages and peritoneal macrophages in the presence of LPS (Fig. 1D). Because LPS induces IL-1β synthesis by up-regulating HIF-1α,15 we used HIF-1α siRNA to confirm the role of HIF-1α in LPS-induced IL-1β production in THP-1-derived macrophages. As expected, inhibition of HIF-1α eliminated the LPS-induced pro-IL-1β expression in THP-1-derived macrophages (Fig. 1E). These results suggested that hypoxia in HCC could create a proinflammatory microenvironment.

SEVERE HYPOXIA-ASSOCIATED NECROTIC DEBRIS OF CANCER CELLS STIMULATES IL-1β PRODUCTION IN MACROPHAGES

Clinically, in some cases, hypoxia or even anoxia is so severe and persistent that HCC cells are forced to undergo necrosis. Macrophages are natural scavengers and are recruited to these regions to clear the necrotic debris.16 In patients with large HCCs exhibiting necrotic centers, we observed substantial CD68⁺ TAM infiltration in the tumor margin, tumor body, and necrotic center (Supporting Fig. S1). To mimic the local microenvironment in the necrotic center of HCC, we investigated whether Huh-7 conditioned medium or persistent hypoxia-associated necrotic debris of Huh-7 cells could stimulate IL-1β secretion in THP-1-derived macrophages in vitro. Persistent hypoxia-associated necrotic debris was confirmed by flow cytometry to ensure that the necrotic components were similar to those exhibited by human HCC samples (Supporting Fig. S2). Stimulation with Huh-7 necrotic debris significantly enhanced IL-1β production in macrophages (Fig. 2A). Moreover, expression of pro-IL-1β was enhanced in a time-dependent manner following the stimulation (Fig. 2B). Although the mRNA levels of proinflammatory cytokines, including IL-1β, IL-6, and TNF-α, were dramatically up-regulated following treatment with Huh-7 necrotic cell debris (Fig. 2C; Supporting Fig. S3), the secreted protein levels of IL-6 and TNF-α did not significantly increase (Fig. 2A).

To further confirm the effect of necrotic debris of cancer cells, we acquired MDMs from healthy donors and HCC patients. MDMs from healthy donors showed consistent alterations upon stimulation of Huh-7 necrotic debris with THP-1-derived macrophages (Fig. 2D). The proinflammatory effect of cancer cell necrotic debris was further enhanced in hypoxia (Fig. 2D). Necrotic debris isolated from HCC samples was then used to stimulate MDMs from the same patients, and IL-1β secretion by macrophages was significantly up-regulated (Fig. 2E). Altogether, we proved that HCC necrotic debris could prime macrophages, enhancing IL-1β production.

O-GLYCAN STRUCTURES CONTRIBUTE TO NECROTIC DEBRIS–INDUCED IL-1β PRODUCTION

We noted that Huh-7 conditioned medium failed to induce IL-1β secretion of macrophages (Supporting Fig. S4A), though it facilitated the differentiation of THP-1-derived macrophages (Supporting Fig. S4B). We further excluded any residual soluble components using debris eluent, and no IL-1β inductive effect was detected (Supporting Fig. S4C). Thus, direct cell-debris contact but not secretory cytokines or soluble damage-associated molecular patterns played a pivotal role in cell debris–induced macrophage activation. Furthermore, living cells and apoptotic cell debris were incapable of inducing IL-1β secretion in macrophages (Supporting Fig. S5). However, proteinase K and O-glycosidase, rather than peptide-N-glycosidase F, treatment could partially or completely inhibit the stimulatory effect of necrotic debris (Fig. 3A,B). For further validation, we used debris of benzyl-N-acetyl-D-galactosamine (an O-glycosylation inhibitor)–pretreated Huh-7 cells to stimulate human MDMs and observed no IL-1β inductive effect (Fig. 3C). More importantly, O-glycosidase treatment of Huh-7 necrotic debris simultaneously inhibited cytokine overexpression and reversed the expression of polarization-related genes including NOS2, CD163, ARG1, and MRC1 (Fig. 3D,E). We then investigated the protein change by gel electrophoresis (Supporting Fig. S6) and, using mass spectrometry, identified a group of membrane-located proteins that potentially contributed to IL-1β induction (Supporting Table S7). These results indicated that O-glycan structures in certain glycoproteins were required for the HCC necrotic debris–induced proinflammatory effect in macrophages.

NECROTIC DEBRIS INDUCES PROINFLAMMATORY EFFECTS AND M2 POLARIZATION IN MACROPHAGES

HCC cells can “educate” TAMs to promote cancer progression by inducing TAM M2 polarization toward an anti-inflammatory phenotype.17 Unexpectedly, we detected M2 polarization of macrophages upon stimulation with cancer cell necrotic debris, which paradoxically induced proinflammatory effects. In the MDMs of HCC patients stimulated with the same patient-derived HCC necrotic debris, we detected increased CD206 expression and reduced inducible nitric oxide synthase (iNOS) expression (Fig. 4A,B). We also found an accumulation of CD206⁺ TAMs in both the tumor margin and necrotic center of HCC (Fig. 4C). Notably, the necrotic center had even more CD206⁺ TAMs than the tumor margin (Fig. 4D). Intriguingly, TAMs in the necrotic center of HCC also had an increased capacity of IL-1β production compared with those at the tumor margin (Fig. 4E,F). Using RNA sequencing, we found a general activation of proinflammatory response, and particularly, NF-κB signaling was activated (Supporting Fig. S7). In addition, Huh-7 necrotic debris and LPS exhibited the same metabolic changes (increased glycolysis and decreased oxidative phosphorylation), which are critical for M1 polarization of macrophages,18 in both human and mouse macrophages of distinct origins (Supporting Figs. S7 and S8). Therefore, our findings demonstrated the proinflammatory effects of M2 TAMs in response to stimulation with HCC necrotic debris.

TLR4/TRIF/NF-κB SIGNALING MEDIATES CELL NECROTIC DEBRIS–INDUCED IL-1β PRODUCTION

Next, we explored the potential mechanisms underlying the enhanced IL-1β production in necrotic debris–primed macrophages. We initially assumed that macrophages could nonselectively engulf necrotic debris and subsequently trigger a proinflammatory response; however, we did not observe a significant engulfment of necrotic debris by macrophages (Supporting Fig. S9A). Additionally, the enhanced secretion of IL-1β was not impaired when phagocytosis was suppressed by cytochalasin D (Supporting Fig. S9B), suggesting a phagocytosis-independent manner.

IL-1β secretion employs a two-step activation mechanism, and TLR signaling is crucial for up-regulating the transcription of IL1B. Because some glycoproteins harbor LPS-like domains and can activate TLR4 to induce a proinflammatory response,19 we assumed that O-linked glycoproteins activated TLR4 in a manner similar to LPS. Indeed, when we used the small molecular inhibitor CLI-095 or an antihuman TLR4 antibody to block TLR4 signaling, the IL-1β inductive role of necrotic debris was ablated (Fig. 5A). Consistently, the transcriptional activity of NF-κB, a downstream effector of TLR4 signaling, was induced by Huh-7 necrotic debris treatment and inhibited by CLI-095 and antihuman TLR4 antibody treatments (Fig. 5B; Supporting Fig. S10A,B). We further analyzed the expression of NF-κB and phosphorylated NF-κB (p-NF-κB) following Huh-7 necrotic debris stimulation. p-NF-κB was up-regulated in a time-dependent manner and reached its peak at 3 hours, while the total NF-κB remained stable (Fig. 5C). In addition, the HIF-1α level was elevated in parallel with p-NF-κB, while the expression of pro-IL-1β increased continuously. Using a TLR signaling pathway quantitative PCR array, we further confirmed the activation of NF-κB instead of the Jun N-terminal kinase/p38 mitogen-activated protein kinase, Janus kinase/signal transducer and activator of transcription, and interferon regulatory factor pathways (Supporting Fig. S11).

Myd88 and TRIF are two TLR4 adapters and mediate overlapping but distinct signaling pathways in response to distinct TLR4 ligands.20 Inhibition of Myd88 using Pepinh-MYD did not affect the induced IL-1β production in both protein and mRNA levels following Huh-7 necrotic debris stimulation (Supporting Fig. S12). Interestingly, Pepinh-TRIF abrogated the enhanced IL-1β secretion induced by Huh-7 necrotic debris (Fig. 5D). Moreover, TRIF inhibition was associated with a consistent decrease of IL1B mRNA and reduced p-NF-κB expression and nuclear location (Fig. 5E; Supporting Fig. S13). The involvement of TLR4/TRIF/NF-κB signaling in IL-1β production and similar changes of M2 phenotypic markers were further verified using Huh-7 necrotic debris–stimulated human MDMs (Supporting Fig. S14). In addition, Huh-7 necrotic debris enhanced the recruitment of TLR4 into the cell membrane (Supporting Fig. S15), possibly due to the presence of saturated fatty acids in the cell membrane,21 further enhancing TLR4 signaling in macrophages. Collectively, these results indicated that TLR4/TRIF/NF-κB signaling mediated cancer cell necrotic debris–induced IL-1β production in macrophages.

The maturation and secretion of IL-1β require caspase-1, which itself needs to be activated.22 We found that Huh-7 necrotic debris induced caspase-1 activation (Supporting Fig. S16), indicating the possible existence of a second signal in our setting and the sufficiency of cancer cell necrotic debris to induce IL-1β production by macrophages.

MACROPHAGE-DERIVED IL-1β INDUCES EMT IN HCC CELLS

To confirm the inflammatory environment in HCC patients, we quantified the serum level of IL-1β in HCC patients, along with patients with cirrhosis and healthy donors. The IL-1β level in HCC patients was significantly elevated compared with that in the healthy controls (Fig. 6A), suggesting a possible association between inflammation and HCC. Notably, HCC patients with necrotic tumors had markedly higher IL-1β levels than patients without evidence of tumor necrosis (81.88 versus 30.48 pg/mL), indicating that tumor necrosis partially contributed to the increased level of IL-1β. To investigate the level of inflammation within HCC, we performed successive cross sections of human HCC tissues and found that the IL-1β level was closely related to the presence of CD68⁺ TAMs (Fig. 6B; Supporting Fig. S17). Using an HCC tissue microarray, we confirmed that a high level of IL-1β in the tumor margins was significantly associated with a poor prognosis in HCC patients (P = 0.0011; Supporting Table S8 and Fig. S18). Univariate and multivariate analyses showed that tumor–node–metastasis stage and elevated IL-1β expression in tumor margins were two independent factors for overall survival (P < 0.01; Supporting Table S9). These findings indicated that TAM-derived IL-1β may promote HCC progression.

We observed that in human HCC tissues the cancer cells in the areas with abundant TAMs exhibited a higher expression of IL-1 receptor and HIF-1α, as well as more significant mesenchymal characteristics (up-regulation of vimentin and down-regulation of E-cadherin) compared with those located in areas with poor TAM infiltration (Fig. 6B,C; Supporting Fig. S17). Given that HIF-1α mediates hypoxia-induced EMT in HCC cells,6 we hypothesized that IL-1β derived from TAMs promoted EMT in HCC cells by inducing a local proinflammatory microenvironment. To test this hypothesis, the conditioned medium from THP-1-derived macrophages was used to culture Huh-7 cells. The conditioned medium elevated vimentin expression and reduced E-cadherin expression in Huh-7 cells; however, this effect was reversed by the IL-1β neutralizing antibody (Fig. 6D), suggesting that IL-1β mediated macrophage-induced EMT in HCC cells. In addition, immunofluorescence assays revealed that there were morphological changes associated with the increased vimentin and decreased E-cadherin expression in Huh-7 cells treated with IL-1β or THP-1-conditioned medium but not LPS stimulation (Fig. 6E). Transwell assays further confirmed that the THP-1-conditioned medium significantly enhanced the migratory and invasive capacities of both Huh-7 and HepG2 cells (Supporting Fig. S19). This effect was retarded by IL-1β neutralizing antibodies, indicating an IL-1β-dependent effect (Supporting Fig. S19A,B). Furthermore, we detected enhanced expression of vimentin, Snail, Twist, and zinc finger E-box binding homeobox 1, as well as a decreased level of E-cadherin in HCC cells treated with IL-1β (Fig. 6F). In summary, these data indicated that macrophage-derived IL-1β facilitated EMT in HCC cell lines.

HIF-1α IS UP-REGULATED BY IL-1β AND MEDIATES IL-1β-INDUCED EMT IN HCC CELLS

Previous studies have demonstrated that IL-1β regulates HIF-1α expression in a cell type–dependent manner. For instance, IL-1β led to HIF-1α overexpression in lung cancer cells but promoted HIF-1α degradation in glioblastoma cells.23, 24 We showed that in both Huh-7 and HepG2 cells HIF-1α rather than HIF-2α was up-regulated by IL-1β at a dose as low as 250 pg/mL in normoxia (Fig. 7A). Consistently, HIF-1α mRNA was selectively up-regulated only in HepG2 and Huh-7 cells (both are epithelial phenotype) but not in SNU-449 cells (mesenchymal phenotype; Fig. 7B). In line with this, the transcriptional activity of HIF-1α was dramatically increased in Huh-7 cells but not in SNU-449 cells (Supporting Fig. S20A). Moreover, IL-1β was unable to affect the expression of EMT-related proteins, including E-cadherin, zonula occludens-1, vimentin, and Snail, in SNU449 cells (Supporting Fig. S20B).

It has been demonstrated that IL-1β activates HIF-1α through the NF-κB/cyclooxygenase-2 (COX-2) pathway.25 To test this theory in HCC cells, a COX-2 selective inhibitor, celecoxib, was added prior to IL-1β treatment. Interestingly, the inhibition of COX-2 activity by celecoxib abrogated IL-1β-enhanced HIF-1α expression and transcriptional activity (Fig. 7C-E), as well as the IL-1β-induced EMT process in HCC cells (Fig. 7E; Supporting Fig. S21). Additionally, we treated a spontaneous HCC mouse model with celecoxib to confirm the role of COX-2 in vivo. As expected, celecoxib significantly reduced the expression of HIF-1α and vimentin (Fig. 7F). We then analyzed the necrotic center of surgically resected HCC from patients who received preoperative transarterial chemoembolization. The necrotic center also revealed increased expression of HIF-1α, IL-1β, and vimentin as well as decreased expression of E-cadherin (Supporting Fig. S22). Altogether, these results indicated that COX-2/HIF-1α mediated IL-1β-induced EMT in HCC cells.

THE HIF-1α/IL-1β LOOP ENHANCES EMT AND METASTASIS OF HCC CELLS IN VIVO

To confirm IL-1β-induced HCC EMT through the up-regulation of HIF-1α in vivo, we first used an intratumoral injection of LPS to enhance the IL-1β secretion of TAMs in a Huh-7 xenograft mouse model. Xenografts in the LPS group showed clearly increased TAM accumulation and overexpression of HIF-1α (Supporting Fig. S23). The up-regulation of vimentin and down-regulation of E-cadherin were also quite apparent in mice locally treated with LPS. In xenografts of mice with decreased TAMs depleted by clodronate,26 expression of IL-1β, HIF-1α, and vimentin was decreased, while E-cadherin expression was recovered (Supporting Fig. S24). Moreover, by neutralizing TAM-secreted IL-1β in the mouse model, we observed reduced Hif-1α expression accompanied by up-regulated E-cadherin and down-regulated vimentin in HCC cells (Supporting Fig. S25). We next established a metastatic mouse model of a red fluorescence protein (RFP)-LM3 xenograft to further verify the role of IL-1β in vivo. We injected recombinant human IL-1β into the xenograft weekly for a total of 5 weeks to create an inflammatory microenvironment within the tumor (Fig. 8A). Two weeks of IL-1β treatment already significantly increased the number of RFP⁺ cells in the circulation of mice compared to those treated with phosphate-buffered saline (Fig. 8B). HLA staining confirmed the increased number of circulating tumor cells in the peripheral blood of mice that received an IL-1β injection (Fig. 8C), indicating the potential role of IL-1β in facilitating HCC cell detachment and metastasis through the bloodstream. Three months later, we observed lung macrometastatic nodules in two of six mice treated with IL-1β and zero of eight in the control group. Hematoxylin and eosin staining confirmed a higher incidence and an increased number of micrometastases in IL-1β-treated mice (P < 0.05; Fig. 8D; Supporting Table S10). We also observed increased recruitment of myeloid-derived suppressor cells in the lungs of IL-1β-treated mice (Supporting Fig. S26A), indicative of an immunosuppressive effect. Immunohistochemistry of the original xenografts revealed overexpression of HIF-1α and COX-2 in mice treated with IL-1β, in conjunction with elevated vimentin expression and reduced E-cadherin expression (Fig. 8E); however, the tumor burden of the xenografts was not significantly affected by the IL-1β injection at the time of analysis (Supporting Fig. S26B,C). Because IL-1β was injected intratumorally and the serum levels of IL-1β, IL-6, and TNF-α were comparable between the experimental and control groups (Supporting Fig. S26D), we believed that metastases formation was primarily mediated by IL-1β-induced HCC EMT within the xenografts.