Background
Cancer cells adapt to hypoxic microenvironments using aberrant vasculature to enable tumor growth. Hypoxia-inducible factors (HIFs) 1α and 2α are major transcription factors required for adaptive responses to hypoxia. HIFs in complex with the aryl hydrocarbon receptor nuclear translocator (ARNT) can bind hypoxia response elements (HREs) of target genes, where they act to enhance transcription [
1]. This conventional mechanism regulates genes such as
VEGF and
HO to improve relatively mild hypoxic conditions [
2,
3]. However, many tumor types, especially those of pancreatic and cervical origin, are known to experience more severe hypoxic stresses [
4].
To respond to dynamic changes in hypoxic conditions combined with variable nutrient, growth factor, and hormone environments, tumor cells employ multiple stress response mechanisms. In addition to HIF signaling, endoplasmic reticulum (ER) stress followed by the unfolded protein response (UPR) is another cellular response involved in adaptation to severe hypoxia [
3]. The UPR can stimulate the transcription of additional genes required for tumor survival and growth.
Mammalian target of rapamycin (mTOR) signaling is also important for adaptation of cancer cells to hypoxia. mTOR is a serine/threonine kinase and a regulator of protein expression at the translational level [
5]. mTOR is activated via phosphorylation of its serine residues, and its activation regulates levels of cellular nutrients, energy availability, and growth factors [
6]. In general, mTOR becomes inactivated in hypoxic tumor tissues. However, it is known that mTOR dysregulation promotes the translation of HIF1 [
5] and cooperates in responses to hypoxic conditions, suggesting that the interaction between the HIF- and mTOR-signaling pathways is involved in adaptation to hypoxia.
We previously demonstrated that the
FVII gene, which encodes blood coagulation factor VII (fVII) [
7], can be induced in ovarian clear cell carcinoma (CCC) cells in response to hypoxia [
8,
9]. Ectopic expression of this pro-coagulant promotes phenotypic changes in cancer cells [
8,
10]. We also demonstrated that a physical interaction between the ubiquitously expressed transcription factor Sp1 [
11] and HIF2 is a major cause of
FVII activation in CCC cells under hypoxic conditions [
12]. Furthermore,
FVII activation is synergistically enhanced via a UPR-independent pathway when serum-starved cells are cultured under hypoxic conditions, explaining why thrombosis frequently occurs in CCC patients [
12]. However, little is known of the mechanisms underlying this synergism.
Ovarian cancer is most the malignant of the gynecological neoplasms [
13,
14]. CCC in particular is highly resistant to chemotherapy and has a poor prognosis [
14]. We hypothesized that genes regulated by the above-described Sp1-dependent mechanism are responsible for adaptive responses to hypoxia with a severity and/or duration that is characteristic of CCC tissues. The aim of this study was to test this hypothesis. We first tested whether multiple genes are activated under hypoxia in an Sp1-dependent manner in CCC cells and whether such transcriptional activations occur synergistically during serum starvation. We further investigated the molecular mechanisms of synergistic gene activation, including regulation by associated serum factors. The study was further extended to investigate
ICAM1 gene expression
in vivo and its contribution to CCC progression.
Discussion
Here, we demonstrated that Sp1 mediates the activation of multiple genes in cooperation with HIFs. This transcriptional activation is synergistically enhanced under SSH, in contrast to the activation of authentic HIF-dependent genes. We further showed that an insufficient supply of LCFAs can lead to synergistic gene activation. The ICAM1 protein is dramatically induced by this mechanism and is critical for tumor growth in vivo, thus indicating a novel mechanism for adaptation to LCFA starvation and hypoxia as a means to promote tumor growth.
Evidence presented in this study suggests that direct binding of both HIFs and NFκB to the
ICAM1 promoter region is essential for synergistic
ICAM1 activation. Notably, HIF2 binding appears to be mediated via interaction with Sp1. The synergistic activation of
KLF6 and
JUN occurs independently of NFκB. This suggests that transcriptional cooperation between Sp1, HIFs, and NFκB promotes rigorous
ICAM1 upregulation under lipid starvation and hypoxia through a distinct mechanism. In addition, the question arises as to how NFκB may be activated under SSH, a condition where exogenous stimuli such as cytokines are absent. Our study provides 2 answers to this question. First, the activation is associated with mTOR activity. This may correlate with earlier observations that, under non-hypoxic conditions, mTOR can cause NFκB activation, leading to
ICAM1 induction in cancerous and normal cells [
25,
28]. Second, NFκB can be activated by autonomously produced TNFα in an autocrine manner.
Recent evidence has shown that cancer cells utilize aberrant mechanisms for energy metabolism [
37]. Some cancer cell types predominantly consume fatty acids rather than glucose, relying on β-oxidation for energy production [
34,
35]. In addition to their normal role as an energy source, recent studies have revealed that LCFAs may be involved in the tolerance of cancer cells to various cellular stresses [
35,
38]. LCFAs exist in the bloodstream in complex with albumin. It is possible that the cellular uptake of LCFAs may be facilitated only in association with albumin, resulting in their conversion to neutral lipids (acyl-CoA derivatives) [
39] in CCC cells. In the present study, we showed that the removal of LCFAs synergistically enhances Sp1-dependent activation of multiple genes under hypoxia. This effect appears to be greater for unsaturated LCFAs than for saturated LCFAs. This difference may be attributed to their roles in transcriptional regulation in addition to altered metabolism pathways among LCFAs, as saturated and unsaturated LCFAs could be differentially utilized as ligands for some nuclear receptors [
40]. It is necessary to uncover the detailed mechanism(s) by which CCC cells sense LCFA-starvation and hypoxia to trigger cellular signaling and induce unusually high gene expression.
We found that ICAM1 enhances cell survival under SSH at least in part through the inhibition of apoptosis. This is distinct from the survival mechanism whereby unsaturated LCFAs protect against UPR-dependent cell death following ischemic conditions mediated by mTOR activity [
38]. Furthermore, ICAM1 promotes CCC tumor growth in mice. Previous reports have found that ICAM1 induction during inflammation enhances eosinophil survival by activating cellular signaling mechanisms, resulting in eosinophilia in asthma [
41]. Thus, CCC cells might similarly activate signaling cascades via ICAM1, enabling cells to survive under severely hypoxic conditions. In additional, ICAM1 is known to be overexpressed in cancer cells where it recruits immune cells to promote tumor progression [
42]. Thus, immune cell-mediated mechanisms may contribute to the acceleration of tumor growth, although it is not clear to what extent immune cells contributed in our immunodeficient mouse model. Furthermore, our data contradict earlier reports showing that ICAM1 suppresses ovarian cancer progression [
30,
36]. These discrepancies may be due to different histological types or experimental settings tested.
The severity of hypoxia within tumor tissues varies depending on the sparsity and aberrant nature of local vasculatures, thus creating microenvironments with limited gradients of molecules supplied from the blood. Diffusion-limited hypoxia occurs at ~100-μm distances from blood vessels, and hypoxic severity in tissues increases with increasing distances [
43]. Given that the sizes of O
2, glucose, and amino acids are smaller than that of LCFA (in complex with albumin), it is likely that areas with lipid starvation, hypoxia, expression of HIFs, and a nutrient supply exists in tumor tissues (Figure
9h). In this case, tissue lipid levels may be used to define a more severe hypoxia
via Sp1 (Figure
9h). It was recently reported that immunotherapy targeting ICAM1 is effective in treating multiple myeloma [
44]. Thus, further clinical investigations will likely support the validity of lipid starvation and hypoxia as a target of cancer therapy.
Methods
Reagents
The following reagents were obtained from Sigma (St. Louis, MO, USA): tunicamycin (T7765), HP-albumin from human serum, fatty acid-free (A3782), LP-albumin from human serum (A1653), palmitic acid (P0500), oleic acid (O1008), linoleic acid (L1376), and stearic acid (S4751). Rapamycin (553210) was purchased from (Millipore, Temecula, CA, USA).
Cell culture
Human cancer cell lines used in this study were cultured as previously described [
8].
Transfection with expression vectors and siRNAs
Expression vectors and siRNAs targeting Sp1 and HIFs were described previously [
12]. We employed 2 individual siRNA sequences targeting each target protein. See Additional file
4 for siRNAs.
cDNA microarray analysis
Total RNA was isolated as previously described [
8]. cDNA microarray analysis of genes activated in response to CoCl
2 stimulation in an Sp1-dependent manner was performed by Takara (Shiga, Japan), using a human whole genome array (Agilent Technologies, Santa Clara, CA, USA). The effects of serum deprivation on the expression of genes under normoxic and hypoxic conditions were analyzed by cDNA microarray analysis using a NimbleGen Catalog Eukaryotic Gene Expression 385 K Array (#5543886, Roche, Madison, WI, USA). Arrays were scanned using a GenePix 4100A Microarray Scanner (Axon Instruments, Union City, CA, USA), and data were analyzed by Subio, Inc. (Kagoshima, Japan).
Real-time RT-PCR analysis
mRNA levels were determined as previously described [
8,
12]. See Additional file
4 for primer and probe sequences. The mRNA expression levels of 18S ribosomal RNA were measured using SYBR green-based detection [
45].
Protein quantification
Protein levels were quantified using the Micro BCA Protein Assay Kit (23235; Thermo Scientific, Rockford, IL, USA).
Western blotting
Western blotting was performed using Novex NuPAGE 4–12% Bis-Tris gels (Life Technologies, Carlsbad, CA, USA). Detection was performed using ECL-Prime or ECL-Advance (GE Healthcare, Buckinghamshire, UK). See Additional file
4 for a description of the antibodies used.
Reporter gene assays
Reporter gene assays were performed as previously described [
12]. Plasmid constructs encoding the
ICAM1,
KLF6, and
JUN 5′promoter regions used in ChIP assays were prepared as previously described [
12]. See Additional file
4 for more details.
ChIP assays
ChIP analyses were performed primarily as previously described [
8,
12]. See Additional file
4 for more details.
Preparation of cytoplasmic and nuclear fractions
Cytoplasmic and nuclear fractions were prepared using an NE-PER Nuclear and Cytoplasmic Extraction Reagents Kit (Thermo Scientific).
Preparation of reduced-albumin serum
Albumin was removed from human serum (Takara Bio, Shiga, Japan) using a Proteoprep Immunoaffinity Albumin & IgG Depletion Kit (Sigma).
Preparation of cells stably transfected with small hairpin RNAs
Cells were stably transfected with short hairpin RNAs (shRNAs) using pcDNA6.2-GW/miR, pLenti6/V5-DEST, and the ViraPower Lentiviral Expression System (Life Technologies). Several independent clones were isolated by selecting cells in 10 μg/mL blasticidin. See Additional file
4 for the shRNA sequences used.
Cell viability and caspase assays
Cell viabilities were estimated in MTS assays using the CellTiter96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, IL, USA). Caspase activities were determined by performing Caspase-Glo assays (Promega).
Xenograft tumor experiments
The institutional review board at the Kanagawa Cancer Center Research Institute approved this study. OVISE (1.5 × 107 cells) cell clones were subcutaneously injected into 9 (for Scr-shRNA or ICAM1 shRNA #1) and 10 (for Scr-shRNA or ICAM1 shRNA sh#2) NOD-SCID mice (Charles River Laboratories Japan Inc., Yokohama, Japan). Tumor growths were monitored by calculating tumor volumes as the long diameter × (short diameter)2 × 1/2. A Hypoxyprobe-1 Kit (Natural Pharmacia International, Burlington, MA, USA) was used to detect tissue hypoxia. Prior to tumor isolation, mice were injected intraperitoneally with pimonidazole-HCl solution (60 mg/kg). At 1 h post-injection, mice were sacrificed under general anesthesia with isoflurane, and tumors were isolated for further experiments.
Immunohistochemistry and immunofluorescence
Routinely processed formalin-fixed, paraffin-embedded specimens were sectioned (4 μm thickness) and stained with antibodies. Immunoreactivity was visualized by the peroxidase-labeled amino acid polymer method, using Histofine Simple Stain MAX-PO (Nichirei Co., Tokyo, Japan) and the avidin-biotin-peroxidase complex method (LSAB+; DakoCytomation Co., Tokyo, Japan), according to the manufacturer’s instructions. Sections were counterstained with H&E. In the case of ICAM1 immunofluorescence, a secondary antibody (Alexa Fluor 488 anti-mouse IgG, Life Technologies) was used. Sections were counterstained with DAPI using Prolong Gold Antifade Reagent with DAPI (Life Technologies). Antibodies against the following targets were used for IHC at the indicated concentrations: ICAM1 (sc-8439, 4 μg/mL; Santa Cruz Biotechnology), pimo-adduct (Hypoxyprobe-1 Kit, 2.3 μg/mL; Hydroxyprobe), CD31 (6.7 μg/mL; Dianova GmbH, Hamburg, Germany), HIF-1α (2.5 μg/mL; BD Biosciences, San Jose CA, USA), HIF-2α (10 μg/ml; NB100-122, Novus Biologicals, Littleton, CO), and Ki67 (1:25 dilution; BD Biosciences).
Microscopic detection of neutral lipid in cultured cells and tumor tissues
Neutral lipid in cultured cells and tumor tissues was detected by fluorescence microscopy using BODIPY lipid probes (Life Technologies). Cultured OVSAYO cells were washed twice with serum-free culture media and treated with 25 μM of BODIPY lipid probes in serum-free media for 5 min in a 37°C incubator. Cells were then washed twice with serum-free media and used for fluorescence detection. Images were acquired on an OLYMPUS CKX41 microscope. Fluorescence areas were quantified using ImageJ software (
http://rsb.info.nih.gov/ij/). For tumor tissues, frozen sections (8-μm thickness) of xenograft tumors were treated with 100 μM of BODIPY lipid probes (Life Technologies) in 50 mM Tris-HCl (pH 7.6) for 30 min at ambient temperature. Tissues were then washed twice with 50 mM Tris-HCl (pH 7.6), and fluorescence detection was performed. Images were acquired on a BZ-9000 fluorescence microscope (Keyence, Osaka, Japan). Quantitative analysis of images was performed using equipped BZ-Analyzer software (Keyence).
Antibody labeling
Labeling of an anti-pimo antibody with Alexa Fluor dye was performed using Zenon Mouse IgG Labeling Kits (Life Technologies), according to manufacturer’s recommended procedure.
In situ hybridization
In situ hybridization was performed using an RNAscope 2.0 FFPE Assay Kit (Advanced Cell Diagnostics, Hayward, CA, USA) and an ICAM1 probe (402951, Advanced Cell Diagnostics). Routinely processed formalin-fixed paraffin-embedded specimens were sectioned (4-μm thickness) and used in this study.
Clinical samples
FFPE tissues from surgically removed specimens were prepared from patients of the Yokohama City University Hospital under written agreements in the study, which was approved by our institutional review board.
Statistics
Statistical significances were evaluated using 2-sided Student’s t-tests for 2 data sets. P values < 0.05 were considered statistically significant.
Accession numbers
The GEO database accession number for the microarray dataset is GSE55565.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SK designed research, performed majority of experiments, analyzed data, wrote the paper. SI and RY performed animal experiments. YN and MY performed immunohistochemistry of xenograft tumor tissues. MF performed immunohistochemistry of clinical samples. EM, FH, and YT analyzed data. YM prepared cells stably transfected with shRNAs, contributed to draft the manuscript, and supervised the project. All authors read and approved the final manuscript.