A hypoxia-dependent upregulation of hypoxia-inducible factor-1 by nuclear factor-κB promotes gastric tumour growth and angiogenesis

Background: The underlying mechanisms involved in the activation of hypoxia-inducible factor-1 (HIF-1) in gastric cancer remain unclear. As nuclear factor-κB (NF-κB) as well as HIF-1 have been implicated in angiogenesis of various cancers, we investigated their relationship in gastric cancer. Methods: Nuclear expressions of HIF-1α and NF-κB/RelA were assessed in 251 human gastric carcinoma specimens by immunohistochemical tissue array analysis. Stable human gastric cancer cells, infected with a retroviral vector containing super-suppressive mutant form of IκBα (IκBαM), were used for animal studies as well as cell culture experiments. Xenografted tumours were measured and IκBαM effects on angiogenesis and HIF-1α activation were assessed by immunohistochemistry, western blotting, luciferase reporter assay, and semiquantitative reverse transcription–polymerase chain reaction. In addition, NF-κB effects on the HIF-1α degradation and synthesis were examined. Results: Hypoxia-inducible factor-1α activation positively correlated with RelA activation in clinical gastric cancer samples (P<0.001). The IκBαM overexpression suppressed tumour growth, microvessel density, and HIF-1α activation in xenografted tumours. Cell culture experiments showed that hypoxia-induced HIF-1α expression was reduced by NF-κB inhibition under hypoxic conditions at the translational level. Conclusion: The hypoxia-dependent activation of the NF-κB/HIF-1α/VEGF pathway contributes, at least in part, to gastric cancer promotion via enhancement of angiogenesis.

Both HIF-1 and NF-kB are involved in cancer progression (Rayet and Gélinas, 1999;Hare et al, 2008) and have been implicated in the tumour responses to hypoxia (Royds et al, 1998;Karashima et al, 2003;Yokoi and Fidler, 2004). Previously, a positive correlation between HIF-1a and NF-kB was shown in surgical colorectal cancer specimens (Kwon et al, 2010). In addition, in vivo animal and in vitro cell culture studies showed that HIF-1a activation was induced by NF-kB in various cancer cells, including lung cancer cells, colon cancer cells, and osteosarcoma cells (Jung et al, 2003;Van Uden et al, 2008). However, these experiments were done mainly under normoxic conditions and little is known about this relationship in cancer cells under hypoxic conditions.
In gastric cancer, both HIF-1a (Zhong et al, 1999;Urano et al, 2006) and NF-kB (Lee et al, 2005) have been reported to be overexpressed in surgical samples. Although HIF-1a was shown to increase angiogenesis and tumour growth of gastric cancer (Stoeltzing et al, 2004), the regulatory mechanism of HIF-1a activation in gastric cancer remains unclear. The present study investigated the correlation between HIF-1a and NF-kB RelA in 251 surgically excised human gastric carcinoma tissues. In addition, we performed cell culture and animal studies after establishment of a stable gastric cancer cell line overexpressing supersuppressive mutant form of IkBa (IkBaM).

Patients and tissue array methods
A total of 251 surgically resected human gastric cancer specimens were obtained from the Department of Pathology, Seoul National University College of Medicine from 1 January to 30 June, 1995, and six paraffin array blocks were prepared by Superbiochips Laboratories (Seoul, Korea), as previously described . Briefly, core tissue biopsies (2 mm in diameter) were taken from individual paraffin-embedded gastric tumours (donor blocks) and arranged in a new recipient paraffin block (tissue array block) using a trephine apparatus. As we have reported previously , the staining results of the different intratumoural areas of gastric carcinomas in these tissue array blocks showed an excellent agreement. A core was chosen from each case for the analysis. We defined an adequate case as a tumour occupying 410% of the core area. This protocol was reviewed and approved by the institutional review board of the Seoul National University (approval no. C-0603-162-170).

Cell culture
A human gastric cancer cell line SNU-668 was obtained from the Korean Cell Line Bank (Seoul, Korea), cultured in RPMI-1640 medium (Life Technologies, Grand Island, NY, USA) containing 10% fetal bovine serum (Life Technologies), and maintained in a 37 1C humidified incubator containing 95% air and 5% CO 2 .

Infection with retroviral vectors expressing IjBa supersuppressor
The control retroviral vector MFG.EGFP.IRES.puro has been previously described (Oh et al, 2001). MFG.EGFP.IRES.puro and the retroviral vector MFG.IkBaM.IRES.puro, which encodes an IkBaM, were generated and infected into three gastric cancer cell lines, as described previously (Lee et al, 2005;Cho et al, 2008).

In vivo tumourigenesis assay
Eight-week-old male athymic nude mice (BALB/cSlc-nu) were purchased from SLC Inc. (Hamamatsu, Shizuoka, Japan). All animal procedures were performed in accord with the procedures described in Workman et al (2010). Tumours were established by injecting 1 Â 10 7 gastric cancer cells (SNU-668 Vector or SNU-668 IkBaM cells) in 100 ml of Matrigel (provided by Professor Hynda K Kleinman, George Washington University, Washington DC, USA) subcutaneously (s.c.) into both flanks of each of ten mice. Tumours were measured on alternate days using a caliper, and the tumour volumes were calculated using the following formula: (length Â width Â height) Â (p/6). After killing, tumour xenografts were removed and prepared for immunohistochemistry or immunoblotting.

Tumour histology and immunohistochemistry
Tissue specimens were fixed with 10% neutral-buffered formalin, and 4 mm paraffin sections were then prepared. One section was stained with haematoxylin and eosin (H&E) for histological assessment, and the other sections were immunostained using a streptavidin peroxidase procedure after microwave antigen retrieval. The primary antibodies were anti-NF-kB p65 (1 : 50; Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-proliferating cell nuclear antigen (PCNA) (1 : 100; DAKO, Carpinteria, CA, USA), anti-cleaved caspase-3 (1 : 100; Cell Signaling Technology, Beverly, MA, USA), anti-CD31 (1 : 100; Santa Cruz Biotechnology), and anti-HIF-1a (1 : 50, provided by Dr Jong-Wan Park in Seoul National University, Seoul). Specimens were incubated with the biotinylated secondary antibody against the corresponding primary antibody and then with avidin -biotin -peroxidase complex (Vectastain Elite ABC kit, Vector Laboratories, Burlingame, CA, USA). Visualisation was performed using diaminobenzidine (DAB). All immunostained sections were then lightly counterstained with Mayer's haematoxylin. Throughout the above analysis, negative controls were prepared by omitting the primary antibody. For PCNA staining, we evaluated 500 cells and counted the cells with nuclear staining for each specimen. The proliferation index was defined as follows: proliferation index (%) ¼ 100 Â PCNA-positive cells/total cells.

Quantification of microvessel density in xenograft tumours
Microvessel densities (MVDs) were determined by light microscopy/optical image analysis after immunostaining xenograft tumour sections with anti-CD31 antibody as described previously (Stoeltzing et al, 2004). The three most highly vascularised areas in areas of tumours near the tumour -normal tissue interface were selected. Photographs of CD31-immunopositive vessels in tumour sections were taken under a light microscope, and the crosssectional areas of CD31-immunopositive structures (i.e., vessel areas) were quantified by capturing images, converting them to greyscale, and analyzing CD31-stained areas using NIH Image Analysis software (version 1.62; National Institute of Health, Bethesda, MD, USA) after setting one consistent intensity threshold for all slides. Then, CD31-positive areas were expressed as pixels squared per high-power field and were measured for all tumours.

Transient transfection and luciferase reporter assay
The NF-kB-luciferase reporter plasmid (pNF-kB-luciferase) (Stratagene, La Jolla, CA, USA) contains a 5 Â NF-kB response element fused to luciferase. To determine whether NF-kB activity is controlled by hypoxia, SNU-668 cells were transiently co-transfected with 0.4 mg of pNF-kB-luciferase and 0.4 mg of b-galactosidase vector, an internal control, using LipofectAMINE Plus (Life Technologies). At 24 h after transfection, SNU-668 cells were incubated under either normoxic or hypoxic conditions. Luciferase activity was measured on an AutoLumat LB 9505c luminometer (Berthold Analytical Instruments, Nashua, Germany) and normalised by b-galactosidase activity. Luciferase activities are presented as relative values vs the normoxic level of the empty vector control.
Lentivirus-mediated short hairpin RNA (shRNA) silencing of HIF-1a Hypoxia-inducible factor-1a shRNA lentiviral particles and control shRNA lentiviral particles were purchased from Santa Cruz Biotechnology. The HIF-1a shRNA lentiviral particles is a pool of concentrated, transduction-ready viral particles containing three target-specific constructs that encode 19 -25 nt (plus hairpin) shRNA designed to knock down gene expression. Construct in the control shRNA lentiviral particle encodes a scrambled shRNA sequence. The viral infection was performed by incubating SNU-668, SNU-216, and SNU-484 gastric cancer cells in the   (22) 160 (78) Abbreviations: HIF-1 ¼ hypoxia-inducible factor-1; NF-kB ¼ nuclear factor-kB. a Considered to be statistically significant.
HIF-1a and NF-jB in gastric cancer SY Nam et al culture medium containing lentiviral particles for 12 h in the presence of 5 mg ml À 1 Polybrene (Santa Cruz Biotechnology). Pooled puromycin (2 mg ml À 1 )-resistant cells were harvested and stored for further analysis.

Protein isolation and western blotting
Cells were lysed with 1 Â Laemmli lysis buffer (2.4 M glycerol, 0.14 M Tris (pH 6.8), 0.21 M SDS, and 0.3 mM bromophenol blue) and then boiled for 10 min. Protein contents were measured using BCA Protein Assay Reagent (Pierce, Rockford, IL, USA). Samples were diluted with 1 Â lysis buffer containing 1.28 M b-mercaptoethanol, and equal amounts of protein were loaded onto 8 -12% SDS -polyacrylamide gels. Proteins were electrophoretically transferred to PVDF or nitrocellulose membranes, and membranes were then blocked with 5% nonfat dry milk in PBS/Tween-20 (0.1%, vol/vol) at 4 1C overnight. They were then incubated with a primary antibody against IkBa (1 :

Assessment of cell viability
SNU-668, SNU-484, and SNU-216 gastric cancer cells (2.5 Â 10 4 cells) were seeded into each well of 24-well plates and were allowed to grow under hypoxic conditions for 0-72 h. Cell numbers were then measured indirectly using the method reported by Kim et al (Hur et al, 2003). Briefly, cells were stained with 0.2% crystal violet aqueous solution in 20% methanol for 10 min, dissolved in 10% SDS, transferred into 96-well plates, and the absorbance was measured at 570 nm using an ELISA reader (Bio-Rad, Hercules, CA, USA).

Statistical analysis
For tissue array analysis, statistical analyses were conducted using SPSS Version 11.0 statistical software program (SPSS, Chicago, IL, USA), and the w 2 -test was used to determine the correlation between the nuclear expressions of NF-kB and HIF-1a. For the animal and cell experiments, data were analyzed using SAS software (version 8.1; SAS Institute Inc., Cary, NC, USA), and the two-tailed Student's t-test was used to determine the significances of the results. The P-values of o0.05 were considered significant for all statistical analyses.

RESULTS
HIF-1a activation in relation to NF-jB activation in clinical gastric cancer samples To confirm the correlation between HIF-1a and NF-kB in human gastric cancer tissues, we performed immunohistochemistry on tissue array slides containing 251 human gastric cancer specimens. Figure 1 shows that both HIF-1a ( Figure 1A) and NF-kB ( Figure 1B) are expressed in both the nucleus and the cytoplasm of tumour cells. Cells showing distinct nuclear staining, regardless of the presence of cytoplasmic staining, were considered to express activated HIF-1a or NF-kB. Immunostaining results were considered to be positive when X5% (for HIF-1a) and X10% (for NF-kB) of tumour cell nuclei were stained ( Figure 1A and B, respectively). Positive immunoreactivity for nuclear HIF-1a was found in 69 of 251 (27%) gastric cancer specimens. In addition, nuclear NF-kB was found in 46 of 251 (18%) of gastric cancer specimens. Data concerning the correlation between the activations of HIF-1a and NF-kB are summarised in Table 1. Nuclear HIF-1a expression was found to be significantly and positively correlated with nuclear NF-kB expression (Po0.001).

Effect of NF-jB inhibition on tumour growth and MVD in the nude mouse xenograft model
In order to confirm the correlation between HIF-1a and NF-kB observed in the clinical gastric cancer samples, we produced stable SNU-668 cell lines overexpressing either empty vector (SNU-668 Vector ) or IkBaM (SNU-668 IkBaM ). Western blotting showed that IkBaM overexpression inhibited the nuclear translocation of NF-kB p65 in SNU-668 cells (Figure 2A). We then established gastric carcinoma xenografts derived from these cells and examined the effects of NF-kB p65 inhibition on tumour growth. We found that 7 out of 10 mice showed tumour formation ( Figure 2B). On the final day of measurement (day 54), SNU-668 IkBaM tumours (mean ¼ 170.0 mm 3 ) were significantly smaller than SNU-668 Vector tumours (mean ¼ 1490.6 mm 3 ) (P ¼ 0.035; Figure 2C). These results indicated that NF-kB inhibition suppresses the gastric tumour growth.
To investigate the mechanism by which NF-kB promotes gastric tumour growth, we performed immunohistochemistry on xenograft tissue sections. First, we identified that nuclear NF-kB p65 expression was less frequent in SNU-668 IkBaM tumours than in SNU-668 Vector tumours ( Figure 3C and D), indicating that NF-kB activation was inhibited by IkBaM overexpression in tumours. Next, the protein expressions of PCNA, a proliferation-related marker, and cleaved caspase-3, an apoptosis-related marker, were analyzed. The immunoreactivity of PCNA was found less frequently in the nuclei of SNU-668 IkBaM tumour cells than in those of SNU-668 Vector tumour cells (P ¼ 0.012;Figures 3E and F and 4A). In addition, the immunoreactivity of cleaved caspase-3 was enhanced in SNU-668 IkBaM tumour cells compared with in SNU-668 Vector tumour cells ( Figure 3G and H). Thus, the NF-kB activity in the gastric cancer xenograft appears to be correlated positively with cell proliferation, and negatively with apoptosis.
We then examined whether NF-kB activity was associated with tumour angiogenesis. Immunohistochemical staining of an endothelial cell marker CD31 was performed on tissue sections of xenografted tumours ( Figure 3I and J). Optical image analyses showed that the CD31-positive vessel area was noticeably smaller in SNU-668 IkBaM tumours (mean ¼ 2.6 Â 10 6 pixels 2 ) than in SNU-668 Vector tumours (mean ¼ 10.4 Â 10 6 pixels 2 ) (Po0.001; Figure 4C). Thus, the NF-kB p65 activity is likely to be positively correlated with angiogenesis in gastric tumours.

Effect of NF-jB inhibition on HIF-1a and VEGF expression
Hypoxia-inducible factor-1a was previously found to increase angiogenesis and tumour growth in gastric cancer (Stoeltzing et al, 2004). As the present study showed that IkBaM overexpression suppressed MVD in xenografted tumours derived from SNU-668 cells, we assessed whether HIF-1a mediates the effect of NF-kB activation in gastric cancer. Immunohistochemistry showed that the frequency of HIF-1a-positive cells was significantly lower in SNU-668 IkBaM tumours (mean ¼ 16.6%) than in SNU-668 Vector tumours (mean ¼ 26.2%; P ¼ 0.001; Figures 3K, L and 4B). Thus, these results indicate that HIF-1a is a downstream molecule of NF-kB.
Furthermore, to exclude the possibility that this difference in HIF-1a immunoreactivity is correlated with the size of xenograft tumours rather than NF-kB activity, we compared HIF-1a expression in three sets of similarly sized xenograft tumours: the sizes of SNU-668 IkBaM tumours were 251.2, 255.2, and 255.8 mm 3 , and the sizes of corresponding vector control tumours were 255.4, 256.3, and 259.5 mm 3 . Western blot analysis showed that IkBaM overexpression decreased the expressions of HIF-1a and VEGF in similar sized tumours ( Figure 4D).
As hypoxia is an important environment for tumour growth and progression, we performed in vitro cell culture experiments and investigated the correlation between hypoxia and the NF-kB activation in gastric cancer cells. Luciferase reporter assay showed that hypoxia induced NF-kB activation in SNU-668 Vector cells after 2 h of hypoxia (P ¼ 0.009 vs normoxic control), and that basal NF-kB activity was essential for hypoxia-induced NF-kB activation in gastric cancer cells ( Figure 5A). In addition, IkBaM overexpression reduced HIF-1a protein expression under hypoxic conditions ( Figure 5B, top). However, SQ RT -PCR showed that IkBaM overexpression did not alter the expression of HIF-1a mRNA, although it decreased the expression of hypoxia-induced VEGF mRNA in SNU-668 IkBaM cells ( Figure 5B, bottom). Consistent results were shown in other gastric cell lines SNU-484 ( Figure 5C) and SNU-216 ( Figure 5D).

Effects of IjBaM overexpression on the degradation and de novo synthesis of HIF-1a protein
To obtain a better understanding of the mechanism involved in NF-kB-induced HIF-1a activation in gastric cancer cells, we investigated whether IkBaM stimulates HIF-1a degradation or inhibits its synthesis. Using three gastric cancer cell lines, we first analyzed the effect of NF-kB on the oxygen-dependent degradation of HIF-1a protein. Hypoxia-inducible factor-1a was first stabilised by exposure to hypoxia, and then destabilised by re-oxygenation. We found that IkBaM overexpression did not affect the oxygendependent degradation rate of HIF-1a protein ( Figure 5E). Next, we checked whether IkBaM overexpression inhibits the synthesis of HIF-1a protein. Cells were pretreated with cycloheximide to remove the remaining HIF-1a, and further degradation of HIF-1a protein was blocked using a proteasome inhibitor MG132. We then examined the time course of HIF-1a accumulation. Figure 5F shows that HIF-1a protein was substantially synthesised after 4 h and that the de novo synthesis of HIF-1a was impaired by IkBaM overexpression. These results indicate that IkBaM overexpression inhibits synthesis, not stabilisation, of HIF-1a protein. Thus, the hypoxic activation of NF-kB seems to contribute to the expression of HIF-1a protein at the translational level.

Effect of HIF-1a shRNA expression on gastric cancer cell viability in vitro under hypoxic conditions
To investigate the role of HIF-1a in gastric tumour growth derived from SNU gastric cancer cells, we blocked the HIF-1a pathway in three gastric cancer cell lines using shRNA expression. Our results showed that shRNA-mediated downregulation of HIF-1a expression reduced the cell viability of SNU-668, SNU-484, and SNU-216 gastric cancer cells in vitro under hypoxic conditions ( Figure 6). These data are consistent with a previous in vivo findings reported by Stoeltzing et al (2004), who showed that xenografted tumours derived from human gastric cancer TMK-1 cells overexpressing dominant-negative-HIF-1a had smaller volumes than those derived from vector control cells.

DISCUSSION
Gastric cancer is one of the most common malignancies worldwide and the leading cause of cancer-related motility in Asian countries (Parkin et al, 1993). Although angiogenesis is an important aspect of tumour growth and progression and is considered the most important predictor of overall survival in gastric cancer (Gong et al, 2005), little is known about the molecular events critical to the gastric cancer angiogenesis. Previously, it was found that HIF-1a activation was significantly correlated with VEGF protein expression, and this overexpression was a prognostic factor in patients with gastric cancer (Mizokami et al, 2006;Urano et al, 2006). Although it was shown that HIF-1a inhibition reduced angiogenesis and tumour growth in xenografted gastric tumours (Stoeltzing et al, 2004), it has become evident that the mechanism underlying the activation of HIF-1 in various cancer cells depends on cancer type (Li et al, 2005). In the present study, NF-kB inhibition suppressed angiogenesis, as was manifested by decreased MVD, and tumour growth of gastric cancer xenografts. Consistently, HIF-1a shRNA expression decreased the cell growth of gastric cancer cell lines. In addition, immunohistochemical tissue array analysis showed that HIF-1a was constitutively expressed in 27% of the 251 surgical samples of gastric carcinomas, and that this was positively correlated with NF-kB activation (Po0.001). Thus, we investigated whether HIF-1a mediates the role of NF-kB in gastric cancer angiogenesis. Cells were cultured to reach 60% confluence, and were preincubated under hypoxic conditions for 8 h. Then, at the indicated time of re-oxygenation, HIF-1a levels were analyzed by western blotting. (F) HIF-1a protein synthesis. After pretreatment with cycloheximide (100 mM) for 1 h, cells were incubated with MG132 (10 mM) to prevent the degradation of newly synthesised HIF-1a. The HIF-1a protein levels were determined at the indicated time by western blotting.
In xenografted gastric tumours, we used stable gastric cancer cells infected with a retrovirus overexpressing IkBaM, which could prove to be a powerful tool to obtain further insight into the role of NF-kB in the regulation of HIF-1 activity. We found that IkBaM overexpression decreased HIF-1a protein expression as well as MVD in xenograft tumours. In addition, IkBaM overexpression decreased VEGF mRNA expression in cell culture experiments.
Thus, these results suggest that HIF-1a is a downstream molecule of NF-kB in the angiogenesis pathway in gastric cancer.
Hypoxia-inducible factor-1 activation was shown to be induced by NF-kB in lung cancer cells and colon cancer cells under normoxic conditions (Jung et al, 2003;Van Uden et al, 2008). In contrast, HIF-1 activation by NF-kB was hypoxia dependent in smooth muscle cells and embryonic kidney cells (Belaiba et al, 2007;Van Uden et al, 2008). In the present study, through a cell culture system, we found that hypoxia was involved in the NF-kB effect on the expressions of HIF-1a protein and VEGF mRNA in gastric cancer cells. Thus, the NF-kB/HIF-1/VEGF pathway in gastric cancer cells might be activated under hypoxic conditions.
In the present study, cell culture experiments showed that NF-kB inhibition suppressed hypoxia-induced HIF-1a protein expression, but not HIF-1a mRNA expression. Furthermore, we found that IkBaM overexpression inhibited synthesis, not degradation, of HIF-1a protein. Thus, the hypoxic activation of NF-kB seems to contribute to the accumulation of HIF-1a protein at the translational level, but not at the transcriptional or post-translational level. These results are not in accord with earlier findings shown in normal cells such as smooth muscle cells (Belaiba et al, 2007) and embryonic kidney cells (Van Uden et al, 2008), which suggested that NF-kB targets and transactivates the HIF-1a gene. This discrepancy may come from the different cell types (cancer cell vs non-cancer cells), as the mechanism underlying the activation of HIF-1a is specifically tailored according to cell type (Li et al, 2005).
In conclusion, the results obtained from human gastric cancer specimens, gastric tumour xenografts, and cell culture experiments indicate that HIF-1a mediates NF-kB-induced angiogenesis by increasing VEGF expression and MVD, and that this occurs under hypoxic conditions. Thus, HIF-1a and NF-kB may be candidate molecular targets for gastric cancer therapy, and in particular, blocking the NF-kB/HIF-1a pathway with appropriate inhibitors might be useful therapeutically for treating gastric carcinoma.  Figure 6 Effect of HIF-1a suppression induced by lentivirus-delivered shRNA on the cell viability of gastric cancer cell lines SNU-668 (top), SNU-484 (middle), and SNU-216 (bottom). Cells were infected with a lentivirus containing a construct, which encodes either HIF-1a shRNA (denoted as shHIF-1a) or scrambled shRNA (denoted as LVcontrol). Expression levels of HIF-1a and b-actin proteins were determined by western blotting after cells were exposed to normoxia (N) or hypoxia (H) for 8 h (A). Cell viability was measured at the indicated time of hypoxia exposure and was represented as the percentage of cell number expressing HIF-1a shRNA vs vector control cells (B). Values represent the means ± s.d. (n ¼ 4). *Po0.05.