Novel HDGF/HIF-1α/VEGF axis in oral cancer impacts disease prognosis

Background Hepatoma-derived growth factor (HDGF) participates in angiogenesis and represents a negative prognostic factor in oral cancer. The current study was designed to elucidate the regulatory mechanism between HDGF and vascular endothelial growth factor (VEGF) and the clinical impact of oral cancer. Methods TCGA data and surgical samples from oral cancer patients were used for the clinicopathological parameter and survival analysis. Human oral cancer SCC4 and SAS cells were treated with recombinant HDGF protein. VEGF gene expression and protein level were analyzed by RT-PCR, Western blotting, and enzyme-linked immunosorbent assay. The signaling pathways for regulating VEGF expression were investigated. The nucleolin neutralizing antibody and HIF-1α inhibitor were applied to SCC4 cells to investigate their effects on the HDGF-stimulated VEGF pathways. Results TCGA and immunohistochemical analysis revealed a positive correlation between HDGF and VEGF expression in oral cancer tissues. Recombinant HDGF significantly increased VEGF gene and protein expression in oral cancer SCC4 cells in a dose-dependent manner. HDGF enhanced the phosphorylation levels of AKT and IkB and the protein level of HIF-1α and NF-κB. The nucleolin-neutralizing antibody abolished HDGF-stimulated HIF-1α, NF-κB and VEGF protein expression in SCC4 cells. The HIF-1α inhibitor antagonized the HDGF-induced VEGF gene expression. High VEGF expression was strongly correlated with HDGF expression, advanced disease, and poor survival. Conclusion This study postulated a new pathway in which HDGF activated HIF-1α and then induced VEGF expression through binding to membrane nucleolin under normoxic conditions, leading to poor disease control. The HDGF/HIF-1α/VEGF axis is important for developing future therapeutic strategies.


Background
Oral cancer is characterized by its aggressive behavior. Even after radical surgery followed by adjuvant radiotherapy and chemotherapy, the survival rate of oral cancer patients remains poor due to relentless recurrence or metastasis [1,2].
Angiogenesis is required for tumor growth [3] and facilitates tumor recurrence and metastasis [4,5] through perturbing the balance of proangiogenic and antiangiogenic factors. Among the proangiogenic factors, vascular endothelial growth factor (VEGF) is the most important one [6]. Angiogenesis plays a critical role in disease progression and mediates treatment resistance [7]. Therefore, understanding angiogenesis, particularly the VEGF pathway, is urgently needed for the risk stratification of oral cancer patients and the development of novel therapeutic targets.
We have previously demonstrated that HDGF overexpression contributes to oncogenic processes and constitutes a novel negative prognostic factor for oral cancer [20]. HDGF expression has been hypothesized to play an important role in tumorigenesis and angiogenesis in oral cancer, which may be associated with the induction of angiogenic factors, leading to a more aggressive pattern of growth and poor prognosis [21]. However, the possible regulatory mechanism between HDGF and VEGF has not been explored.
Thus, the current study was designed to elucidate the possible interaction or regulatory mechanism between HDGF and VEGF and the possible clinical impact in oral cancer.

Western blotting
Whole cell extracts were prepared and quantified by the Coomassie Plus Assay as previously described [22]. The PVDF membrane was blocked with 5% skim milk in TBS-T for 1 h and then incubated with the indicated primary antibodies and secondary antibodies conjugated with HRP (1:5000; Santa Cruz Inc.; Santa Cruz, CA, USA) for 1 h each. The signals on the membrane were detected using HRP chemiluminescent substrate (Millipore Corporation; Billerica, MA, USA) and exposed to X-ray film for signal detection.

Enzyme-linked immunosorbent assay (ELISA)
The VEGF-A concentrations in the supernatants were determined by ELISA using a commercially available kit (Boster Biological Technology, Valley Ave, Pleasanton, CA). Briefly, after supernatant collection, the total cellular proteins were extracted and then measured by the bicinchoninic acid assay to assess the cell number in each group. The secreted VEGFA concentration was normalized to the total cellular protein level and is shown as the mean ± SD.

Immunohistochemical staining and assessment
The surgically resected specimens from 102 oral cancer patients were included with approval from the institutional review board. Immunohistochemical analysis using the tissue microarray (TMA) consisting of surgically resected samples from oral cancer patients was performed as previously described [20] to delineate the correlation between HDGF expression, VEGF expression, and clinic-pathological parameters. Briefly, the slides were incubated with primary HDGF antibody (1: 200 dilution) and VEGF antibodies (1:250; Santa Cruz; Santa Cruz, CA, USA) for 30 min and visualized using a peroxidase-conjugated secondary antibody, a polymer detection system (Zymed Laboratories, San Francisco, CA, USA) and 3,3-diaminobenzidine tetrahydrochloride (Sigma, St. Louis, MO). The sections were then counterstained with hematoxylin and eosin.
The percentage of tumor cells with definite moderate to intense nuclear or cytoplasmic immunoreactivity was scored, and the median of scores from multiple cores in the same patient was adopted as the labeling index (LI) for each marker, as previously described [12,20,24]. A total of 95 patient samples containing at least two preserved-tissue cores were scored and analyzed. Seven patients were excluded due to insufficient TMA samples. The cutoffs of the LIs to define high expression of HDGF were determined as follows: (1) high expression of nuclear HDGF (HDGF-N) if ≥40% of tumor nuclei were stained, (2) high expression of cytoplasmic HDGF (HDGF-C) if ≥40% of tumor cytoplasm was stained and (3) VEGF high expression if ≥50% of tumor cytoplasm was stained.

Immunofluorescent staining of paraffin-embedded tissues
Immunofluorescence staining was performed on surgically resected specimens of oral cancer patients as described previously [25]. To investigate the expression of HDGF and VEGF, tissue sections were incubated with primary HDGF antibody (1:200 dilution) and VEGF antibodies (1:250; Santa Cruz; Santa Cruz, CA, USA). After wash step, tissue sections were incubated with appropriate fluorescent-labeled secondary antibodies then nuclei were stained with DAPI (Sigma-Aldrich, St. Louis, MO, USA). Finally, tissues were mounted with coverslips in fluorescence mounting medium (Dako corporation; Glostrup, Denmark). The fluorescent color of HDGF was Green (AlexaFluor488); VEGF was red (Alexa-Fluor546); nuclei were stained with blue color (DAPI). The microscope images were captured using Zeiss LSM 510 confocal imaging (200x magnification) and processed with ZEN 2 microscope image analysis software (Carl Zeiss; Jena, Germany).

Computational biology analysis
HDGF and VEGF mRNA expression data were obtained from The Cancer Genome Atlas (TCGA). All the software and graphics for transcriptomics analysis were developed using in-house code implemented in MATLAB (MathWorks, Natick, MA, USA). HDGF and VEGF expression in the TCGA are reported as the fold changes between 1) oral cancer and healthy tissues and 2) head and neck cancer and healthy tissue. The correlation of HDGF and VEGFA mRNA expression in the TCGA dataset was analyzed by UCSC Xena (http://xena.ucsc. edu/).

Statistical analysis
For the Western blotting, RT-PCR, and ELISA data, comparisons were performed by using one-way ANOVA followed by the Newman-Keuls post hoc test or t-test (for multiple comparisons) using Prism 5 (GraphPad Software, Inc., La Jolla, CA, USA). All the in vitro experiments in this study were triplicated. A probability value < 0.05 is considered to be statistically significant.
The associations among clinicopathologic factors, HDGF expression, and VEGF expression were evaluated using the X 2 test, t-test, and ANOVA as appropriate. Estimates of disease-specific survival (DSS), metastasis-free survival (MFS), and local recurrence-free survival (LRFS) were calculated using the Kaplan-Meier method with log-rank test. The multivariable analyses of DSS, MFS, and LRFS were performed by using the Cox proportional hazards model with a stepwise approach. All tests were two-tailed, with a probability value < 0.05 considered to be statistically significant. Clinical statistical analyses were performed using SPSS 14 software (SPSS, Chicago, IL, USA).

Correlation of HDGF and VEGF expression and clinicopathological parameters in oral cancer tissues
Because HDGF overexpression is correlated with angiogenesis and tumorigenesis, including in oral cancer [26], we investigated whether there was a relationship between HDGF and VEGF expression in oral cancer and head and neck cancer. According to TCGA data analysis (n = 522, oral cancer; n = 566, head and neck cancer [TCGA, Provisional cohort]), the HDGF and VEGF mRNA expression profile exhibited a strong positive correlation (P = 0.0107; R 2 = 0.01247, oral cancer; P = 0.0001; R 2 = 0.02643, head and neck cancer) ( Fig. 1A and B). Therefore, these results suggested that HDGF expression was positively correlated with VEGF expression in human head and neck cancer and oral cancer. An immunohistochemistry assay revealed a positive correlation (P = 0.006) between HDGF-N expression and VEGF expression (Table 1). In addition, high expression of VEGF and HDGF-N were closely linked to advanced status of oral cancer, more advanced primary T stage and poorly differentiated histological grade. Higher VEGF expression also correlated with more advanced nodal status (P = 0.021). Immunohistochemical staining of HDGF and VEGF and immunofluorescence staining of oral cancer patients were shown in Fig. 1c and d.

Recombinant HDGF induced VEGF expression and release in oral cancer cells
To investigate whether HDGF regulated VEGF expression in oral cancer cells, SCC4 cells and SAS cells were treated with different concentrations of recombinant HDGF protein and then harvested for subsequent analysis. RT-PCR showed that exogenous HDGF protein significantly increased VEGF gene expression by approximately 1.5-fold compared with the control group in SCC4 cells (Fig. 2a, rHDGF 100 ng/ml, P < 0.01). Western blotting assays showed the protein levels of VEGF were also increased by HDGF stimuli in a dosedependent manner ( Fig. 2b and Additional file 1: Figure  S1, rHDGF 100 ng/ml, P < 0.05). We next analyzed the secreted levels of VEGF by Western blotting and ELISA. As expected, more VEGF protein was secreted into the culture medium under HDGF stimulation than in the control group (Fig. 2c, rHDGF 100 ng/ml, P < 0.05). ELISA analysis revealed that HDGF enhanced a small but significant levels of VEGF secreted by SCC4 cells in a dose-dependent manner (Fig. 2d). Approximately additional 50 pg/ml VEGF was secreted in 100 ng/ml-rHDGF-treated group, comparing to the control group ( Fig. 2d, P < 0.01). Therefore, these results supported that additional HDGF induced VEGF upregulation and expression in human oral cancer cells. SAS cells were treated with recombinant HDGF protein for 24 h before harvest. Western blotting showed the protein levels of VEGF was upregulated by HDGF stimulation in a dosedependent manner (Additional file 1: Figure S2A-B).
HDGF stimulates AKT/HIF-1α/NF-κB signaling in oral cancer cells Given the well-known signaling pathways for regulating VEGF expression [27,28], we then focused on the activation of specific transcription factors, including HIF-1α, NF-κB, and STAT3. SCC4 cells were treated with recombinant HDGF, and the levels of HIF-1α, NF-κB, and Correlation between HDGF and VEGF mRNA levels in oral cancer and head and neck cancer patients obtained by analysis of data from TCGA. HDGF expression is positively correlated with VEGFA expression in human head and neck squamous cell carcinoma tissues, including oral cancer. c Tissue microarray analysis of the correlation between HDGF and VEGF expression in oral cancer patients. The photographs were from two representative oral cancer patients. Case 1 (pT2N0M0, stage II) exhibited lowexpression HDGF and VEGF immunostaining, whereas Case 2 (pT3N2M0, stage III) showed high-expression staining of both HDGF and VEGF. Scale bars, 20 μm. d Immunofluorescence staining of oral cancer patients. The fluorescent color of HDGF was Green (AlexaFluor 488); VEGF was red (AlexaFluor 546); nuclei were stained with blue color (DAPI). Case 3 (pT1N0M0, stage I) exhibited both high-intensity staining of HDGF and VEGF, whereas Case 4 (pT2N0M0, stage II) showed intermediate-intensity HDGF and VEGF immunofluorescence staining, and Case 5 (pT1N0M0, stage II) showed low-intensity staining of HDGF and VEGF. Scale bars, 20 μm STAT3 were measured and quantified by Western blotting ( Fig. 3a-d and Additional file 1: Figure S3A-D). HDGF enhanced the phosphorylation levels of AKT and IκB in the HDGF-treated group compared with the control group in SCC4 cells (Fig. 3a-b and Additional file 1: Figure S3A-B, rHDGF 10 ng/ml, P < 0.01). In addition, the protein levels of the transcriptional factors HIF-1α and NF-κB p65 were also upregulated under HDGF stimulation (HIF-1α, Fig. 3c and Additional file 1: Figure  S3C, rHDGF 1 ng/ml, P < 0.01; NF-κB p65, Fig. 3d and Additional file 1: Figure S3D, rHDGF 10 ng/ml, P < 0.05), indicating that HDGF triggered the AKT/HIF-1α/ NF-κB signaling pathway. HIF-1α was upregulated under HDGF stimulation in SAS cells (Additional file 1: Figure  S2C, rHDGF 1 ng/ml, P < 0.01). However, HDGF treatment (even at a high dose of 100 ng/ml) did not affect the phosphorylation of STAT3, suggesting that HDGF did not elicit STAT3 activation in SCC4 cells (Fig. 3e and Additional file 1: Figure S3E). Together, these results implied that HDGF stimulated AKT/HIF-1α/NF-κB signaling, thereby modulating VEGF expression in oral cancer cells.
Antibody neutralization of surface nucleolin abolished HDGF-stimulated AKT/HIF1α/NF-κB/VEGF signaling in oral cancer cells Because the surface nucleolin/AKT axis has been found to participate in transmitting the oncogenic signaling of HDGF [22], we investigated whether blockage of the HDGF/nucleolin axis by antibody neutralization affected HDGF-stimulated HIF-1α, NF-κB and VEGF expression in SCC4 cells. Western blotting analysis demonstrated that the additional recombinant HDGF was unable to enhance phosphorylation levels of AKT and HIF-1α protein under the co-treatment with the neutralizing antibodies against nucleolin in SCC4 cells (Fig. 4a-b and Additional file 1: Figure S4A-B). Moreover, blocking the HDGF/nucleolin axis not only diminished the HDGFstimulated phosphorylation of IκB and NF-κB p65 but also significantly reduced VEGF protein expression (Fig.  ; HDGF-C, Cytoplasmic expression of hepatoma-derived growth factor; VEGF, Vascular endothelial growth factor; *, Statistical significance; All values are presented as mean ± SD. The correlation analyses were evaluated using the X 2 test, t-test, and ANOVA as appropriate 4c-d and Additional file 1: Figure S4C-E, P < 0.05). These results suggested that the nucleolin-mediated signaling pathway is important for HDGF-modulated VEGF expression.

Application of the HIF-1α inhibitor chetomin antagonized HDGF-induced VEGF upregulation in oral cancer cells
To further investigate which of the transcription factors HIF-1α and NF-κB were dominant in HDGF-induced VEGF gene expression, we employed the HIF-1α inhibitor chetomin and the NF-κB inhibitor Bay 11-7082. RT-PCR analysis showed no significant difference in VEGF mRNA levels with or without additional HDGF in the chetomin group (Fig. 5a, P, not statistically significant). The application of chetomin potently suppressed the HDGF-induced VEGF gene expression. On the other hand, HDGF treatment was able to induce VEGF upregulation even in the presence of Bay 11-7082 (Fig. 5a). Western blot assays demonstrated that chetomin suppressed the VEGF protein expression induced by HDGF (Fig. 5b). Although Bay 11-7082 could inhibit the basal level of VEGF, VEGF was still enhanced in HDGFtreated cells (Fig. 5b). Moreover, ELISA also revealed that chetomin eliminated the increased secretion of VEGF protein induced by HDGF (Fig. 5c, P, not statistically significant). Thus, HIF-1α signaling plays a critical role in HDGF-induced VEGF gene regulation.

Univariate log-rank analyses of survival
According to univariate survival analysis, postoperative concurrent chemoradiotherapy (Post-OP CCRT), histological grade, and high expression of HDGF-N and VEGF were statistically significant prognostic predictors for DSS, MFS, and LRFS. The univariate survival analysis is summarized in Additional file 1: Table S1. High VEGF expression predicted a higher rate of local and distant recurrence and shorter DSS in the Kaplan-Meier survival analysis (Fig. 6).

Multivariate analyses of survival
In the multivariate comparison (Table 2)

Discussion
Angiogenesis is essential for cancer progression, metastasis and treatment resistance. The regulation of angiogenesis involves a number of critical growth factors, cytokines, signaling cascades and cellular processes that are triggered in response to either a hypoxic or an inflammatory stimulus [29]. Hypoxia-and inflammationdriven angiogenesis are regulated via distinctly different and yet overlapping pathways [30]. Through correlation analysis of the immunohistochemistry assay and TCGA data, these results provided support for the interaction between HDGF and VEGF expression in oral cancer. In this study, for the first time, we demonstrated that HDGF enhanced VEGF  The mechanism through which HDGF induces or regulates VEGF expression in tumor cells remains unclear. HDGF has been reported to stimulate the proliferation and invasion of hepatocellular carcinoma cells via PI3K/ AKT signaling [22,31]. Indeed, activation of the PI3K/ AKT pathway in both tumor and endothelial cells increases VEGF secretion by both HIF-1α-dependent and HIF-1α-independent mechanisms [32][33][34].
In hypoxia-driven angiogenesis, hypoxia activates the PI3K/AKT pathway to prevent the posttranslational hydroxylation and the subsequent degradation of HIF-1α, allowing it to accumulate and then translocate to the nucleus, where it upregulates VEGF production pathways [29,[35][36][37]. The inflammatory stimulus activates the PI3K/AKT pathway, leading to the phosphorylation of IκBα. IκBα is degraded, allowing NF-κB subunits p50 and p65 to translocate into the nucleus and activate VEGF production [29,38].
HIF-1 has been shown to essentially control the cellular response to hypoxia. Evidence has emerged that HIF-1α is also responsive to stimuli under normoxic conditions [39]. One important mechanism underlying these normoxic conditions is the transcriptional regulation of HIF-1α by NF-κB [40], which is the key promoter in the inflammatory angiogenic pathway [29,39]. Recently, HIF-1α has been reported that directly bound to the HDGF promoter region, which was highly correlated with pancreatic cancer-associated fibrosis under normoxic condition [41].
Our data have shown that exogenous HDGF protein not only stimulated the phosphorylation levels of AKT and IκB but also increased the protein levels of the transcriptional factors HIF-1α and NF-κB p65 in oral cancer cells. The Western blotting results (Fig. 3) showed HDGF in a dose of 10 ng/ml was able to enhance more than two folds of phosphorylation levels of AKT and IκB; additional HDGF in the low-dose of 1 ng/ml could induce two folds of the protein levels of HIF-1α. The upregulation of phosphorylated IκB implied a loss of NF-κB blockage by IκB, leading to NF-κB activation and subsequently to modulated HIF-1α expression or VEGF production. This finding offered a rationale for how HDGF simultaneously triggered the AKT/HIF-1α and NF-κB signaling pathways in oral cancer cells.
HDGF has been demonstrated to bind directly to surface nucleolin (NCL) and activate the NCL/PI3K/AKT axis in hepatoma cells during liver carcinogenesis [22]. Here, we applied a neutralizing antibody against nucleolin that was able to abolish the HDGF-stimulated phosphorylation levels of AKT, IκB and NF-κB p65 and the HDGF-stimulated protein levels of HIF-1α VEGF. These results suggest that surface nucleolin plays a pivotal role in mediating the HDGF-induced AKT/HIF-1α signaling and NF-κB signaling pathways, ultimately modulating VEGF expression in oral cancer cells.
Studies have shown that the binding of both STAT3 and HIF-1α to the VEGF promoter is essential for the maximum transcription of VEGF mRNA under hypoxia [42]. STAT3 signaling is required for VEGF-and PI3K/ AKT-mediated HIF-1α expression. Blocking STAT3 abolished both HIF-1 and VEGF expression [43]. However, whether STAT3 contributes to HIF-1 expression/ activity independently of AKT remains to be determined.
Here, HDGF did not modulate the phosphorylation levels of the transcription factor STAT3 even at a high dose (100 ng/ml), suggesting that STAT3 activation was not modulated by recombinant HDGF in SCC4 cells. This result implied that HDGF-stimulated VEGF expression might act through alternative AKT/HIF-1α and NF-κB signaling pathways but not the STAT3 pathway in oral cancer cells.
To confirm the signaling pathway between HDGF and VEGF, a HIF-1α inhibitor (chetomin) and an NF-κB inhibitor (Bay 11-7082) were used. Mild upregulation in VEGF mRNA level, protein level, and secreted protein level were noted in the chetomin alone group. In the cotreatment group of HDGF and chetomin, the VEGF levels were reduced without further enhancement that suggested chetomin was able to eliminate HDGFinduced VEGF expression pathway in SCC4 cells. In the other hand, Bay 11-7082 has some suppressive effects on the VEGF mRNA level, protein level, and secreted protein level. The HDGF cotreatment with Bay 11-7082 was able to upregulate the VEGF mRNA and protein levels even under the possible suppression caused by Bay 11-7802. Therefore, the current study revealed the pivotal role of HIF-1α signaling in the HDGF-mediated upregulation of VEGF (Fig. 5d). There are some limitations to this study. We only analyzed three common signal transduction pathways for the possible regulatory VEGF pathways by Western blotting. Western blotting can only identify a single protein-protein interaction but not for weak or transient interactions and evaluate numerous pathways will be time-consuming. However, the two cell lines in the current study showed consistent results. Also, we validated the correlation between HDGF and VEGF in the clinical data.
In the current study, the cohort of patients with oral cancer who received radical treatment was selected to assess the prognostic value of VEGF immunohistochemical staining. High expression of VEGF was strongly correlated with HDGF-N expression, primary T stage, nodal status and histological grade. In a previous study, high expression of HDGF seemed limited to locally aggressive behavior only [20]. Here, high expression of VEGF was associated with a greater likelihood of both local and distant recurrence. VEGF is capable of increasing vascular permeability in both blood and lymphatic vessels and helps cancer cells enter lymphatic or blood vessels and become established in both local lymph nodes and at distant sites [4,44]. In multivariate analysis, high expression of VEGF was the most significant predictor for all survival endpoints (LRFS, DMS, and DSS).
Antiangiogenic agents can potentially modulate the tumor microenvironment and induce radiosensitivity and chemosensitivity. Using antiangiogenic agents alone or in combination with conventional therapies in oral cancer is a promising new approach [45]. In the current study, HDGF activated HIF-1α and then induced VEGF expression, leading to poor disease control. The combination of antiangiogenic agents and HIF-1 inhibitors might be efficacious, as antiangiogenic agents would cut off the tumor's blood supply, and HIF-1α inhibitors could potentiate the effect of antiangiogenic agents and reduce the potential for the development of drug resistance [46]. Therefore, the HDGF/nucleolin/HIF-1α/VEGF axis is a very attractive target for oral cancer treatment.

Conclusions
In summary, this study is the first to report the association between HDGF and VEGF and prognosis in oral cancer. Our study postulated a new pathway in which HDGF activated HIF-1α and the NF-κB signaling pathway and then increased VEGF expression through binding to membrane NCL under normoxic conditions. The HDGF/HIF-1α/VEGF axis is important for developing future therapeutic strategies.
Additional file 1: Table S1. Univariate log-rank analyses of HDGF and VEGF. Figure S1. Effect of HDGF on VEGF expression in oral cancer cells. Figure S2. Effect of HDGF on VEGF expression in oral cancer cells. Figure  S3. HDGF triggered AKT/HIF-1α/NF-κB signaling in SCC4 oral cancer cells. Figure S4. The Neutralizing antibody against nucleolin eliminates HDGFstimulated AKT/HIF-1α/NF-κB/VEGF signaling in SCC4 oral cancer cells. Funding This work was supported in parts by grants from Kaohsiung Veterans General Hospital, Taiwan (VGHKS 106-112 & VGHNSU108-001) for the design of the study and collection, analysis, and interpretation of data; the Ministry of Science and Technology, Taiwan (MOST 107-2320-B-110-003; 108-2314-B-110-003-MY2) for collection, analysis, and interpretation of data; Kaohsiung Armed Forces General Hospital, Taiwan (102-5) for analysis and interpretation of data and in writing.

Availability of data and materials
All data analyzed during this study are included in this published article.

Ethics approval and consent to participate
The Institutional Review Board (IRB) at Chi Mei Medical Center approved this study (IRB09703-001). The IRB waived the need for written informed consent from the participants because this was a retrospective review study. No ethics approval is required for these purchased cell lines (SCC4 and SAS) from the bioresource centers.