Hypoxia upregulates CD147 through a combined effect of HIF-1α and Sp1 to promote glycolysis and tumor progression in epithelial solid tumors

Abstract

Hypoxia is one of the most pervasive physiological stresses within tumors. Hypoxia signaling contributes to the aggressive tumor behaviors through promoting tumor cells to undergo the fundamental metabolism adaptation. A series of evidence indicates that this process is mainly mediated by hypoxia-inducible factor (HIF). However, key molecules involved in tumor hypoxia adaptation remain to be characterized. In this study, we investigated the functional role of CD147, a transmembrane glycoprotein highly overexpressed on the surface of tumor cells, in hypoxic microenvironment using in vitro and in vivo assays. Immunohistochemical staining showed that CD147 expression was upregulated in hypoxic region of epithelial solid tumor tissues. In addition, our data indicated that hypoxia induced the upregulation of CD147 expression at both mRNA and protein levels in epithelial carcinoma cells in a time- and dose-dependent manner. Moreover, we demonstrated that hypoxia-induced CD147 upregulation was mainly mediated by a combined effect of transcription factors HIF-1 and specificity protein 1 (Sp1) on the activation of CD147 promoter. We also explored the metabolic functions of hypoxia-induced CD147 and found that upregulated CD147 promoted glycolysis in both tumor cell lines and nude mice tumor xenograft model, partially through the functional cooperation with MCT-1 and MCT-4. Finally, we observed that CD147 promoted tumor growth, inhibited tumor cell apoptosis and enhanced their invasion ability under hypoxia. In conclusion, our findings reveal a novel mechanism of hypoxia adaptation mediated by CD147 in epithelial solid tumors and suggest that CD147 may be a promising therapeutic target in cancer treatment.

Introduction

Hypoxic microenvironment in tumor tissues is developed when solid tumor growth outpaces the delivery ability of existing vasculature. Although hypoxia is toxic to both cancer and normal cells, cancer cells undergo genetic and adaptive changes that allow them to survive and even proliferate in a hypoxic environment ( 1 ). Thus, tumor hypoxia has become a central issue in the studies of tumor physiology and cancer treatment.

Hypoxia, as one of the most pervasive physiological stresses within tumors ( 2 ), selects tumor cells to undergo glycolysis and strong evidence indicates that the glycolytic phenotype of cancer cells is a crucial component of malignancy that confers a significant growth advantage ( 3 , 4) . In recent years, the regulatory mechanism of glycolytic switch by hypoxia has been extensively studied. A group of transcription factors has been reported to be implicated in regulating a wide spectrum of genes responsible for the metabolic changes under hypoxia ( 5 , 6) . A pivotal component of this complex regulatory system is the HIF, a heterodimeric protein composed of a constitutively expressed HIF-1β subunit and an oxygen sensitive HIF-1α subunit. The HIFα–HIFβ dimmer binds to a conserved DNA consensus on the promoters of its target genes known as hypoxia-responsive element ( 7 , 8) . HIF induces a vast array of gene products controlling essential cellular processes, such as energy metabolism, neovascularization, survival, pH and cell migration, all of which are crucial features for hypoxic adaptation ( 9 ). In fundamental metabolic alterations, HIF-1 drives the overexpression and increased the activity of several glycolytic proteins, including transporters (Glut-1, MCT-4) and enzymes (HK1, HKII, LDH-A, PGK1, PYK-M2), all of which play a key role in controlling the changes in glycolytic flux inside cancer cells ( 10 ). Although the biology of hypoxia signaling has been progressively elucidated and many of the HIF-induced gene products have been characterized, the mechanism underlying hypoxia adaptation in tumor cell and the key molecules involved in this process remain not very clear.

CD147, a member of the immunoglobulin superfamily, has been characterized as an inducer of matrix metalloproteinase synthesis ( 11 ). As a transmembrane glycoprotein, CD147 often exhibits different molecular weight in different cell types dependent on the degree of glycosylation. The low-molecular-weight protein of 28kDa commonly represents the unglycosylated CD147. The overexpression of CD147 is observed on the surface of numerous types of malignant cancer cells and associated with the malignant potential and poor prognosis in these malignancies ( 12 ). In particular, CD147 plays a pivotal role as an ancillary protein required for the expression and function of monocarboxylate transporter MCT1 and MCT4 ( 13 , 14) , all of which are fundamental to the glycolytic phenotype that characterizes most cancers ( 15 ). Further study has demonstrated that silencing of CD147 dramatically decreases the glycolytic rate and lactate efflux in carcinoma cell line, indicating that CD147 is involved in tumor glycolysis ( 16 , 17) . Previous studies have reported that CD147 expression is upregulated under ischemic conditions in neuronal and cardiac cells ( 18–20 ). Most recently, CD147 has been reported to be induced by hypoxia in a colon carcinoma cell line LS174 ( 21 ), and suspected as a putative HIF-1 target gene through a genome-wide chromatin immunoprecipitation (ChIP)-on-chip assay ( 22 ). However, thus far, the systematic investigation on hypoxia and CD147 expression is far more insufficient in solid tumors. Further evidence from human tumor tissue and more tumor cell types are urgently warranted.

Considering that biological functions of CD147 are closely associated with hypoxia adaptation, we hypothesized that CD147 may serve as a new hypoxia-responsive molecule essential for the glycolytic switch under hypoxia. To test this hypothesis, we investigated the CD147 expression in hypoxic regions of epithelial solid tumor tissues and examined the CD147 mRNA and protein levels in three epithelial carcinoma cell lines under different hypoxic conditions. Further, we investigated the combined effect of HIF-1α and Sp1 on the activation of CD147 promoter and explored the metabolic functions of hypoxia-induced CD147 and the effect of upregulated CD147 under hypoxia on cellular functions, such as growth, apoptosis and invasion.

Materials and methods

Tissue collection and immunohistochemical staining

Human tumor tissue samples were obtained from the Department of Pathology at Xijing Hospital affiliated with Fourth Military Medical University, Xi’an, China, with signed informed consents. All tissues were assessed by H&E staining to select suitable regions for immunohistochemical examination. Immunohistochemical staining was performed as previously described ( 23 ). The following primary antibodies were used, including anti-HAb18G/CD147 prepared in our lab ( 24 ), anti-carbonic anhydrase (CA) IX (1:50; Santa Cruz), anti-Glut-1 (1:200; Thermo Fisher Scientific), anti-MCT-1 (1:100; Chemicon) and anti-MCT-4 (1:50; Chemicon). Antigen retrieval was carried out in 0.01mol/l citrate buffer (pH 6.0) using an autoclave oven at 110°C for 2min.

Cell culture and hypoxia treatment

Human lung adenocarcinoma cell line A549, mammary carcinoma cell lines MCF-7 and MDA-MB-468, hepatocellular carcinoma cell lines HCC-9204 and HepG2, and embryonic kidney cell line 293 (HEK293) were purchased from Shanghai Institute for Biological Sciences (Shanghai, China) and cultured in DMEM media (Hyclone Laboratories, Logan, UT) supplemented with 10% fetal bovine serum (Gibco, Rockville, MD). For hypoxic exposure, cells were placed in a hypoxia incubator (HERAcell 240) flushed with 1% O ₂ /5% CO ₂ /94% N ₂ . In addition, hypoxic effect was reversed in cultured cells through the treatment of a chemical HIF-1α inhibitor 3-(5ʹ-hydroxy-methyl-2ʹ-furyl)-1-benzylindazole (YC-1; Sigma), and mithramycin A (MMA; Sigma) was used for specific inhibition of Sp1 activity.

Western blot analysis and quantitative RT–PCR

Cells were lysed with RIPA buffer (Pierce Biotechnology) and western blot assay was performed as previously described ( 25 ). The following primary antibodies were used, including anti-HAb18G/CD147 ( 24 ), anti-MCT1 (1:400; Chemicon), anti-MCT4 (1:200; Chemicon), anti-HIF-1α (1:100; BD Biosciences), anti-Sp1 (1:200; Santa Cruz) and anti-α-tubulin (1:200; NeoMarkers, Freemont, CA) antibodies. Quantitative RT–PCR was carried out as earlier descried ( 26 ). The hypoxia-induced change in CD147 mRNA expression was calculated based on the threshold cycle (Ct) as relative to β-actin using a 2 ^{−△(△Ct)} where △Ct = △Ct _CD147 −△Ct _β-actin and △(△Ct) = △Ct _1% −△Ct _21% . The sequences of PCR primers are listed in Table S1, available at Carcinogenesis Online.

siRNA silencing and cell transfection

The siRNA sequences corresponding to the cDNA sequences of HIF-1α (NM_001530) and CD147 (NM_001728) were designed as previously reported ( 26 , 27) . A549 cells were transfected with HIF-1α siRNA (100nM), and cells transfected with scramble negative control (SNC) or mock serve as control. MCF-7 cells were transfected with HIF-1α expression vector (pCEP4/HIF-1α), HIF-1α dominant negative vector (pCEP4/HIF-1α/DN) or empty vector (pCEP4). Cells were allowed to recover overnight after transfection using Lipofectamine 2000 Reagent (Invitrogen, Carlsbad, CA) and then exposed to normoxia or hypoxia (1% O ₂ ). After 24h stimulation, the effect of HIF-1α on CD147 expression was determined through immunoblot assay. The sequences of specific siRNA or SNC fragments were listed in Table S1, available at Carcinogenesis Online.

Construction of reporter plasmids and site-directed mutagenesis

Luciferase reporter vectors containing the full-length CD147 promoter region (−1761 to +37, relative to the transcription start site of CD147 gene) or a series of truncated CD147 promoter sequences have been developed in our previous study or newly constructed in this study as previously described ( 28 ). The QuickChange Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA) was used to generate the site-directed mutant of HIF-1-binding element at CD147 promoter region from −133 to −130 according to the manufacturer’s instructions. The sequences of PCR products were confirmed by sequencing in Shanghai Sangon Co. (Shanghai, China). All PCR primers are listed in Table S1, available at Carcinogenesis Online.

Luciferase reporter assay

HEK-293 cells were cotransfected with a combination of different luciferase reporter vectors (pGL3-CD147 constructs or pGL3-basic) and HIF-1α expression vectors (or control vector pCEP4) using Lipofectamine 2000 Reagent. The HIF-1α expression vector pCEP4/HIF-1α/AdCA5 encodes a mutant version of HIF-1α that is resistant to degradation ( 29 ). (Plasmids of HIF-1α expression vectors including pCEP4/HIF-1α, pCEP4/HIF-1α/AdCA5, pCEP4/HIF-1α/DN were generous gifts from Dr Yan Zheng in our university.) Cells were harvested 48h after transfection and the luciferase activity was determined using a dual-luciferase reporter assay system kit (Promega) with a luminometer (Tecan, Switzerland).

ChIP assay

ChIP assay was carried out using the EZ-ChIP™ kit (17-371, Upstate, Millipore Corporation, Billerica, MA). Immunoprecipitation was carried out with anti-HIF-1α antibody (1:25; Novus) or mouse isotype control IgG (1 ug/ml; Millipore). PCR reactions generate two 215-bp productions from the promoter region of CD147 gene and promoter-specific primers are listed in Table S1, available at Carcinogenesis Online.

Measurement of in vitro glycolytic capacity

Cells were seeded onto a six-well plate and cultured in DMEM containing 4500mg/l glucose supplemented with 10% FBS under 1% O ₂ hypoxia or normoxia. After 24h, the culture media was collected. Extracellular pH was measured with a pH meter (PB-11 Basic Meter; Sartorius). Measurements were made within 2min of sample collection. Glucose uptake was determined using the Amplex Re Glucose/Glucose Oxidase Assay kit (Molecular Probes, Carlsbad, CA). Lactate production in the medium was detected by using the Lactate Assay Kit (BioVision, Mountain View, CA). Measurement of LDH activity was performed using Lactate Dehydrogenase Activity Assay Kit (BioVision). Absorbance was measured using GENios Plus Plate Reader (Tecan). All results were normalized on the basis of the total protein amounts of cells that were determined using micro BCA Assay (Pierce, Rockford, IL).

Nude mice xenograft model and 18F-fluorodeoxyglucose uptake assay

The hairpin oligonucleotides corresponding to the cDNA sequence of CD147 (NM_001728) were synthesized and cloned into Ambion’s pSilencer 2 vector. A549 cell clones stably expressing the CD147 shRNA (defined as A549–CD147 shRNA cell) were established as previously described ( 30 ). CD147 expression was examined by flow cytometry as previously described ( 25 ). Xenografts were initiated by subcutaneous injection of 5×10 ⁶ A549 cells into nude mice on the right side and 5×10 ⁶ A549–CD147 shRNA cells on the left side ( n = 6). A time–volume curve was plotted to investigate the xenograft growth. Six weeks later, mice were injected intraperitoneally with ¹⁸ F-fluorodeoxyglucose ( ¹⁸ F-FDG) at a concentration of 10 µCi/g and sacrificed exactly 60min after injection. Tumor tissues were weighed and assayed for radioactivity using a Gamma Scintillation Counter (Gamma 8000; Beckman Instrument, CA). Tumor ¹⁸ F-FDG uptake was calculated as the radioactivity per gram of tumor tissue ( 31 ). Relative uptake level was compared between left and right side tumor tissues within the same mouse. CD147 expression was evaluated in xenograft tumor tissues using immunohistochemical staining. In addition, the relationship between ¹⁸ F-FDG uptake (µCi/g tissue) and CD147 expression level [staining score calculated as previously described ( 28 )] in xenograft tumor tissues was investigated using Pearson correlation coefficient analysis. All experiments were performed in accordance with institutional guidelines for the care and use of experimental animals.

Apoptosis assay and in vitro transwell invasion assay

Cells transfected with CD147 siRNA or SNC were cultured in DMEM with 0.5% FBS under hypoxia (1% O ₂ ) for 48h. Then, necrotic or late-stage apoptotic cells were determined by double-staining with annexin V-FITC and propidium iodide (PI) using an Apoptosis Detection Kit (Calbiochem, San Diego, CA) with a FACSCalibur flow cytometer. TUNEL assay was performed on xenograft sample sections using a FragEL™ DNA Fragmentation Detection Kit (Calbiochem) according to the protocol instructions. In addition, the invasion ability of cultured cells was measured with the in vitro transwell migration assay using modified Boyden chambers with polycarbonate filters as earlier reported ( 26 ).

Statistical analysis

All statistical analyses were performed using the SPSS 16.0 statistical software package (SPSS, Chicago, IL). Statistical differences were analyzed by Student’s t -test. All the statistical tests were two-sided and P -value of <0.05 was considered to be significant.

Results

CD147 expression is upregulated in hypoxic regions of epithelial solid tumor tissues

To investigate the relationship between CD147 expression and hypoxia, immunohistochemical staining was performed in tissue samples of three common epithelial solid tumors using anti-CD147 monoclonal antibody and anti-hypoxia markers polyclonal antibodies. Our results indicated weak to moderate staining intensity of Glut-1 and Glut-3 in these tissues (data not shown). Therefore, CA IX was used as an endogenous hypoxia marker due to the significant correlation between its expression level and hypoxia condition in solid tumors ( 32 ). As shown in Figure 1 , CD147 was remarkably upregulated in hypoxic regions indicated by positive CA IX staining among tissue samples from lung cancer, hepatocellular carcinoma and breast cancer. Hypoxic regions were mainly located in the center area of cancer cell nests. These results suggest that hypoxic microenvironment in epithelial solid tumor tissues may induce the upregulation of CD147 expression.

Fig. 1.

Immunohistochemical staining of ( A ) CD147 and ( B ) CA IX (an endogenous hypoxic marker) in hypoxic regions of lung cancer, hepatocellular carcinoma and breast cancer tissues. Cell nuclei were counterstained with hematoxylin. Scale bars, 100 µm.

Hypoxia induces the upregulation of CD147 expression in epithelial carcinoma cells in a time- and dose-dependent manner

To further determine the effect of hypoxia on CD147 expression, we examined the CD147 mRNA and protein levels in three epithelial carcinoma cell lines under different hypoxic conditions using quantitative RT–PCR and western blot, respectively. Our results showed that CD147 expression was significantly upregulated in a time-dependent manner at both mRNA and protein levels in all three cell lines (A549, MCF-7 and HCC-9204) under a hypoxia condition of 1% O ₂ ( Figure 2A and 2B ). In addition, a dose–response relationship was observed between CD147 protein expression and hypoxic level in all three cell lines ( Figure 2C ), indicating that hypoxic conditions of both 1% O ₂ and 5% O ₂ significantly increased the CD147 protein expression in a time-dependent manner, which was not observed under a normoxic condition (21% O ₂ ). A condition of 1% O ₂ exhibited a stronger induction of CD147 protein expression than 5% O ₂ at both 12 and 24h in all three cell lines. These data demonstrate that CD147 serves as a hypoxia-responsive molecule that is induced by hypoxia in a time- and dose-dependent way.

Fig. 2.

CD147 and HIF-1α are significantly upregulated under hypoxia in three carcinoma cell lines (A549, MCF-7 and HCC-9204). (A) Western blot analyses for HIF-1α and CD147 protein levels under hypoxia (1% O ₂ ) condition over the time as indicated. (B) qRT–PCR analyses for CD147 mRNA levels under hypoxia (1% O ₂ ) over the time as indicated. β-Actin that shows no response to hypoxia was used as an internal control for qRT–PCR. Data were represented as mean ± SD from three independent experiments. * P < 0.05 (C) Western blot analyses for CD147 expression at different exposure time (0, 12, 24h) in different O ₂ concentrations (21, 5, 1%). (D) Western blot analyses for Sp1 and CD147 protein levels under hypoxia condition (1% O ₂ ) over the time as indicated. Tubulin was used as a protein-loading control in western blot assay.

Hypoxia-induced CD147 upregulation is significantly associated with increased expression of HIF-1α and Sp1

Considering that hypoxia-induced CD147 upregulation is mainly through transcriptional activation, we further evaluated the expression level of HIF-1α, a key transcriptional factor involved in hypoxic response, using western blot and analyzed the association of HIF-1α expression with CD147 upregulation. Our data showed that HIF-1α expression was evidently increased in a time-dependent manner in all three cell lines under hypoxic condition of 1% O ₂ ( Figure 2A ). The increase of HIF-1α expression occurred earlier than CD147 expression upregulation in A549 and MCF-7 cell lines, suggesting that HIF-1α might act as a transcriptional activator to upregulate CD147 expression under hypoxic condition. However, in HCC-9204 cell line, the increase of HIF-1α expression occurred much later than CD147 upregulation, indicating that other transcriptional factors might contribute to the upregulation of CD147 expression under hypoxic condition in this hepatocellular carcinoma cell line. Furthermore, the expression level of Sp1, a transcription factor with putative capability of binding to CD147 promoter, was measured under hypoxic condition of 1% O ₂ ( Figure 2D ). Western blot analysis demonstrated that Sp1 expression had a similar change in all three cell lines after exposing to hypoxic condition, indicating that the protein level of Sp1 was first increased at a certain time point after hypoxic exposure and then decreased to the normal levels. We reexamined the time-dependent expression level of CD147 in HCC-9204 cells treated by MMA (100nM), which can effectively inhibit the transcriptional activity of Sp1 under hypoxic condition (1% O ₂ ). As indicated in Supplementary Figure 1 , available at Carcinogenesis Online, our data demonstrated that MMA-treated HCC-9204 cells exhibited a significant lower CD147 expression level and late increase of CD147 expression under hypoxia (1% O ₂ ) when compared with MMA-untreated HCC-9204 cells. These results support an early role of Sp1 in hypoxia-induced CD147 upregulation, although we cannot completely rule out the possibility of decreased CD147 protein degradation under hypoxia. We further investigated the effect of p53 mutation status on the CD147/HIF-1α/Sp1 time-course expression under hypoxic condition (1% O ₂ ). As illustrated in Supplementary Figure 2 , available at Carcinogenesis Online, our data indicated that HepG2 (p53wt) cells exhibited a very similar result of CD147/HIF-1α/Sp1 time-course expression with HCC-9204 (p53 mutant). It is also the case for MDA-MB-468 (p53 mutant), which showed a similar time-course expression pattern of CD147/HIF-1α/Sp1 with MCF-7 (p53wt). All these results suggest that HIF-1α and Sp1 may exhibit a combined effect on hypoxia-induced CD147 upregulation under hypoxic condition, which is independent of p53 mutation status.

HIF-1α significantly upregulates CD147 expression

To investigate the regulatory role of HIF-1α in hypoxia-induced CD147 expression, we examined the CD147 protein level in the three cell lines under the treatment of YC-1, a chemical HIF-1α inhibitor, which significantly inhibits HIF-1α accumulation under hypoxia at post-translational level via the factor-inhibiting HIF-1α-dependent C-terminal transactivation domain inactivation ( 33 , 34) . As expected, western blot analysis indicated that all three cell lines exhibited a notable decrease of hypoxia-induced HIF-1α protein expression when treated with 100 µM of YC-1 for 24h. Accordingly, CD147 protein expression was significantly decreased in the YC-1-treated cell lines ( Figure 3A ). Furthermore, we assessed the CD147 expression in A549 cells after knockdown of HIF-1α by siRNA under hypoxia. Our data indicated that cells transfected with siRNA targeting HIF-1α exhibited a significant reduction of hypoxia-induced HIF-1α and CD147 expression, compared with cells transfected with SNC siRNA or mock control ( Figure 3B ). Consistently, ectopic expression of HIF-1α in MCF-7 cells under hypoxia induced a significant increase of HIF-1α and CD147 protein levels in cells transfected with HIF-1α expression vector (pCEP4/HIF-1α), compared with cells transfected with dominant-negative form of HIF-1α (pCEP4/HIF-1α/DN) or mock control ( Figure 3C ). All these results support that functional HIF-1α protein can significantly upregulate CD147 expression.

Fig. 3.

HIF-1α significantly upregulates CD147 expression. ( A ) Western blot analysis for HIF-1α and CD147 protein levels in A549, MCF-7 and HCC-9204 cells treated with 100 µM YC-1, an inhibitor of HIF-1α at treatment time of 24h under hypoxia (1% O ₂ ) respectively. ( B ) Left: western blot analysis of HIF-1α and CD147 expression in A549 cells transfected with HIF-1α siRNA or SNC siRNA under normoxia and hypoxia (1% O ₂ ). ( C ) Left: western blot analysis of HIF-1α and CD147 expression in MCF-7 cells transfected with mock vector pCEP4 (pCEP4), dominant negative form of HIF-1α vector pCEP4/HIF-1α/DN or HIF-1α expression vector pCEP4/HIF-1α (HIF-1α) under normoxia or hypoxia (1% O ₂ ). ( B and C ) Right: scanning densitometry analysis for relative expression levels of CD147 and HIF-1α calculated as the ratio divided by tubulin. Data were represented as mean ± SD from three independent experiments. * P < 0.05.

HIF-1α transactivates CD147 expression through directly binding to CD147 promoter

To further explore the mechanism underlying CD147 upregulation by HIF-1α, we evaluated the effect of HIF-1α on CD147 promoter activity in HEK293 cells. Western blot analysis indicated that HEK293 cells transfected with HIF-1α expression vector pCEP4/HIF-1α/AdCA5, a mutant version of HIF-1α that is resistant to degradation under normoxia, had a notable HIF-1α and CD147 protein accumulation, which was not observed in cells transfected with empty vector pCEP4 and untransfected cells ( Figure 4A ). Luciferase reporter assay demonstrated that transcriptional activity of CD147 promoter (−1761/+37) was significantly increased in HEK293 cells transfected with pCEP4/HIF-1α/AdCA5, compared with cells transfected with pCEP4 or pCEP4/HIF-1α/DN. The increase of promoter activity was found to be dependent on the transfection dose of the HIF-1α expression vector ( Figure 4A ). These data strongly suggest that the CD147 promoter region may contain a binding site of HIF-1α. To test this finding, we analysed the DNA sequence of CD147 promoter and identified eight putative HREs with the characteristic motif 5′-RCGTG-3′ ( 35 ) ( Figure 4B ). To further define the binding site of HIF-1α in CD147 promoter, we developed a series of truncated promoter constructs based on the HREs’ distribution and determined their transcriptional activity in HEK293 cells transfected with pCEP4 or pCEP4/HIF-1α/AdCA5. Luciferase reporter assay showed that all CD147 promoter constructs truncated from −1761 to −338 exhibited similar HIF-1α-related transcriptional activities in HEK293 cells and the construct without the promoter region (from −338 to +37) completely abolished the activity of the reporter gene in all transfected cells ( Figure 4B ). These results indicate that the transcriptional activity of CD147 promoter is dependent on the core region from −338 to +37, which is likely to contain the binding site of HIF-1α. We further generated truncated constructs for the core promoter region from −338 to +37 and found that deletion of the region from −145 to −125 totally abrogated the transactivation function of HIF-1α, suggesting that the third HRE within this region may be the binding site for HIF-1α ( Figure 4B ). This finding was further confirmed by site-directed mutation assays, in which the third HRE was mutated from ‘CGTG’ at −133 to −130 to ‘CGAA’. Our results showed that the transactivation function of HIF-1α was completely abolished upon the mutation of the third HRE in CD147 promoter ( Figure 4C ). In addition, ChIP assay provided additional lines of evidence supporting a direct binding of HIF-1α to CD147 promoter within A549 cells, through the identification of PCR products that were obtained under hypoxic condition of 1% O ₂ when specific antibody against HIF-1α, but not non-specific IgG, was used ( Figure 4D ). Taken together, these results demonstrate that HIF-1α directly binds to HRE site in the CD147 promoter region from −145 to −125 to transactivate CD147 expression.

Fig. 4.

HIF-1α directly binds to CD147 promoter region and promotes CD147 transcriptional activity in a combined way with Sp1. (A) Left: western blot analysis of HIF-1α and CD147 in HEK293 cells transfected with pCEP4 or pCEP4/HIF-1α/AdCA5 (CA5) which encodes a mutant version of HIF-1α that is resistant to degradation. Untransfected cells were used as blank control. Right, a dual-luciferase reporter assay was performed in HEK293 cells cotransfected with the pGL3-Basic vector containing the Luc gene under the control of the human CD147 promoter (−1761/+37) fragment (150ng) with different amounts of pCEP4/HIF-1α/AdCA5 (200 or 400ng). The corresponding vectors (pCEP4, 310ng) or (pCEP4/HIF-1α/DN, 400ng) were used as control. (B) Identification of HRE on CD147 promoter region. Left: a scheme depicting the deletion of the 5′-flanking region of CD147 promoter is presented and putative HRE sites defined as 5′-RCGTG-3′ are indicated as numbered 1–8. Right: luciferase reporter assay in HEK293 cells cotransfected with reporter vectors containing a series of nested 5′-deletions in CD147 promoter (−1761/+37) and pCEP4/HIF-1α/AdCA5 or pCEP4. (C) Mutation analysis of critical HRE (−133 to −130) region on luciferase activity. The mutation from ‘CGTG’ to ‘CGAA’ was generated by site-directed mutagenesis as shown by underlined bold letters. (D) ChIP assay with an anti-HIF-1α antibody or mouse isotype control IgG in A549 cells under normoxic or hypoxic (1% O ₂ ) condition. Upper, PCR products from primers specific for HRE 3; lower, PCR products from control primers specific for HRE 8. (E) Functional analysis of HIF-1 and Sp1 in CD147 transactivation. HEK293 cells were transfected with a combination of pCEP4 or pCEP4/HIF-1α/AdCA5 and different reporter constructs with or without treatment of MMA 24h before the luciferase activity was determined. Reporter constructs including pGL3/CD147p-145 wide-type, pGL3/CD147p-145 HRE-mutant and pGL3/CD147p-108 were used for luciferase reporter assay. Cotransfection of pRL-TK (28ng) was served as an internal control. Data are presented as mean ± SEM from three separate experiments, each in triplicates. * P < 0.05.

HIF-1α and Sp1 exhibit a combined effect on the activation of CD147 promoter

We further examined the combined effect of HIF-1α and Sp1 on CD147 transcription using luciferase reporter assay ( Figure 4E ). Our results showed that transcriptional activity of the minimal core region (from −145 to +37) in CD147 promoter, which contains the binding sites of HIF-1α and Sp1, was significantly inhibited by Sp1-specific inhibitor MMA (50nM) in HEK293 cells transfected with expression vector pCEP4/HIF-1α/AdCA5 or pCEP4. The inhibitory effect was also noted when the binding site of HIF-1α was mutated or deleted in the minimal core region of CD147 promoter. In addition, the transcriptional activity significantly increased in HEK293 cells transfected with expression vector pCEP4/HIF-1α/AdCA5 regardless of Sp1 inhibition. However, the increase of transcriptional activity was much more evident in transfected cells without Sp1 inhibition than those with Sp1 inhibition, when compared with corresponding control cells transfected with pCEP4. Furthermore, the increase of transcriptional activity was abolished by the mutation or deletion of the binding site of HIF-1α. These results indicate that HIF-1α and Sp1 cooperatively promote CD147 expression and the transcriptional activity of HIF-1α might partially be dependent on the background Sp1 activity.

Hypoxia-induced CD147 upregulation promotes glycolysis in tumor cells

Given that hypoxic microenvironment is closely related to the upregulation of glycolysis in solid tumors, we further investigated the metabolic functions of hypoxia-induced CD147. In vitro assays showed that lung cancer cell A549 exhibited a significantly higher glucose uptake rate, lactate secretion rate and LDH activity when cultured under hypoxia (1% O ₂ ), compared to those under normoxia (21% O ₂ ) ( Figure 5A ). In comparison, pH value in A549 cell culture medium was obviously lower under hypoxia. However, these hypoxia-enhanced glycolytic phenotypes were significantly inhibited when CD147 expression was knocked down using a specific siRNA targeting CD147 (si-CD147) in A549 cells ( Figure 5B ). Western blot assay demonstrated that knockdown of CD147 expression resulted in a significant decrease of MCT-1 and MCT-4 expression levels in A549 cells ( Figure 5C ). Similar results for glycolytic phenotype analyses and western blot assay were also obtained when MCF-7 and HCC-9204 cell lines were used (data not shown). Furthermore, in vivo assay was performed based on the nude mice model transplanted with wild-type A549 cells or A549 cells stably transfected with shRNA targeting CD147. Our data indicated that tumor tissues developed by CD147 shRNA-transfected A549 cells exhibited a significantly decreased ability of ¹⁸ F-FDG uptake (average 68% of wild-type), compared with that in tumor tissues developed by wild-type A549 cells ( Figure 5D and 5E ). Pearson correlation coefficient analysis indicated a significant trend of higher CD147 expression level with higher ¹⁸ F-FDG uptake in both A549–CD147 shRNA xenografts ( R² = 0.694, P = 0.040) and A549 xenografts ( R² = 0.676, P = 0.041) ( Figure 5F ). Immunohistochemical staining demonstrated that CD147 expression was effectively knocked down in A549–CD147 shRNA xenograft tissue even under hypoxic conditions. In addition, the expression levels of MCT-1 and MCT-4 were significantly decreased in CD147 shRNA-transfected tumor tissues, whereas no obvious expression change was observed for Glut-1 ( Figure 5G ). These results suggest that CD147 may be involved in the metabolic reprogramming of tumor cells under hypoxia, at least partially through affecting the expression of MCT-1 and MCT-4.

Fig. 5.

Knockdown of CD147 expression inhibits glycolysis through inhibition of MCT-1 and MCT-4 both in vitro and in vivo . (A) Measurement of glucose and lactate concentration in culture medium, LDH activity within A549 cells as well as extracellular pH value under normoxia or hypoxia (1% O ₂ ). Data are presented as relative values normalized to normoxia. (B) Measurement of glucose uptake and lactate secretion rate, LDH activity as well as pH in medium in A549 cells transfected with CD147 siRNA or SNC under hypoxia (1% O ₂ ). Data are normalized to SNC-treated cells and represented as mean ± SD from three independent experiments (* P < 0.05). (C) Western blot analyses for CD147, MCT-1, MCT-4 expression in A549 cells transfected with CD147-siRNA or SNC under hypoxia (1% O ₂ ). (D) Flow cytometric analysis of CD147 expression in A549 and A549–CD147 shRNA cells. (E) Measurement of ¹⁸ F-FDG uptake in nude mice xenograft model ( n = 6). Relative uptake level was compared between xenografts developed by A549–CD147 shRNA and A549 cells. (F) Pearson correlation coefficient analysis between CD147 expression level and ¹⁸ F-FDG uptake in xenograft tissues. (G) Representative immunohistochemical staining images of CD147, CA IX, Glut-1, MCT-1 and MCT-4 in serial sections of xenograft tumor tissues developed by A549 or A549–CD147 shRNA cells. Scale bars, 50 µm.

Upregulated CD147 inhibits tumor cell apoptosis and enhances their invasion ability under hypoxia

To explore the biological significance of hypoxia-induced CD147, we examined the effect of upregulated CD147 under hypoxia on cellular functions such as apoptosis and invasion, which have been previously reported to be significantly associated with CD147 expression under normoxia ( 30 ). Our results showed that A549 cells transfected with CD147 siRNA had a significantly lower percentage of viable cells with double-negative staining of annexin V-FITC and PI under the starvation condition (0.5% FBS) than cells transfected with the control siRNA (79.47% versus 86.67%). In addition, the CD147 siRNA-transfected cells exhibited a notable increase of late-stage apoptotic cells with double-positive staining (12.81% versus 6.15%) ( Figure 6A ). TUNEL assay showed that A549–CD147 shRNA xenograft tissues exhibited a significantly higher percentage of TUNEL positive cells than A549 xenograft tissues (16.17% ± 2.94% versus 6.50% ± 1.78%, P = 0.001) ( Figure 6B ). These results suggest that the upregulated CD147 under hypoxia may protect tumor cell from apoptosis. In vitro invasion assay demonstrated that knockdown of CD147 dramatically decreased the invasion ability of A549 cells under hypoxia. Compared to the number of invading cells in the SNC group, the average percentage of invading cells in the A549 cells transfected with CD147 siRNA was 44.67% ± 4.33% ( Figure 6C ). Our data also indicated that A549–CD147 shRNA xenograft tumors exhibited a significant decrease of growth capacity compared with A549 xenograft tumors ( P = 0.012, repeated measures ANOVA) ( Figure 6D ).

Fig. 6.

Upregulated CD147 inhibits tumor cell apoptosis and enhances their invasion ability under hypoxia. (A) Flow cytometry analysis for starvation-induced cell apoptosis and necrosis through double-staining of annexin V-FITC and PI under hypoxia (1% O ₂ ). (B) Left: representative images of TUNEL assay in the tissue sections of A549 and A549–CD147 shRNA xenografts. Scale bar, 50 µm. Right: percentage of positive staining cells is presented as mean ± SD ( n = 6, * P < 0.05). (C) Left: representative images of transwell invasion assay using A549 or A549 CD147-siRNA cells. Scale bar, 50 µm. Right: column graph for three independent experiments. Data were normalized to SNC-treated cells and represented as mean ± SD, * P < 0.05. (D) Time–growth curves of tumor xenografts. Tumor volumes are represented as mean ± SD ( n = 6). Repeated measures ANOVA model was used for statistical analysis.

Discussion

Our previous study has systematically evaluated the CD147 expression profile in human normal and tumor tissues and found that the positivity rate of CD147 was 67.76% in epithelium-derived carcinoma, including lung cancer, hepatocellular carcinoma and breast cancer, whereas only 5.18% in normal epithelial tissues ( 23 ). Recent studies have been focused on the mechanisms underlying the CD147 overexpression and provided more and more evidence to support the transcriptional activation of CD147 in tumor cells. Our previous study has also demonstrated that promoter hypomethylation upregulates CD147 expression in HCC through increasing Sp1 binding ( 36 ). Recently, several cytokines have been reported to be correlated with upregulation of CD147, such as TGF-β1 ( 37 ) and epidermal growth factor ( 38 ). In the present study, we observed that CD147 expression was upregulated in the putative hypoxic regions of epithelial tumor tissues from various cancers. Further investigation using cell model demonstrated that hypoxia significantly upregulated the expression of CD147 in a time- and dose-dependent way. Our results for the first time provide a relatively comprehensive survey on the induction of CD147 expression by hypoxia in solid tumors, which extends the systematic evidence from previously reported ischemic diseases to current general tumor types. These findings clearly indicate that CD147 is a hypoxia-responsive molecule.

We sought to elucidate the molecular mechanism underlying the hypoxia-induced CD147 upregulation. We found that HIF-1 and Sp1 cooperatively played a critical role in the transcriptional activation of CD147 promoter under hypoxia conditions. It has been well established that HIF-1 stabilized and activated by hypoxia is a master transcriptional regulator in several hypoxia-mediated signaling pathways ( 8 , 39) . HIF-1 is known to upregulate over 100 genes with a diverse plethora of functions, including cellular proliferation, survival, angiogenesis and energy metabolism ( 9 , 40) . Using bioinformatic analysis, we identified eight putative HREs located in the DNA sequence of CD147 promoter region. Therefore, it is conceivable that HIF-1 may be involved in hypoxia-induced transactivation of CD147. Our study showed that hypoxia-stabilized HIF-1 upregulated CD147 expression in tumor cells. Moreover, we provided the first evidence that HIF-1 directly binds to a specific HRE located at −133 to −130 on the CD147 promoter region. Our results are consistent with the finding of an earlier study that reported CD147 gene as one of potential direct HIF-1 targets through an integrative analysis through ChIP-on-chip and mRNA expression profiling ( 22 ).

Our study demonstrated that, in addition to HIF-1, the expression of Sp1 was also increased in a time-dependent manner in all three cell lines under a hypoxic condition, which is consistent with a previous report showing that Sp1 is constitutively involved in facilitating promoter activation under hypoxia condition ( 41 ). Moreover, our group recently identified a critical Sp1 binding site located at −87 to −81 and confirmed its essential role in upregulating CD147 promoter activity ( 28 ). These lines of evidence led us to further examine the combined effect of HIF-1α and Sp1 on CD147 transcription. Our data demonstrated that either inhibition of Sp1 or HIF-1α significantly reduced CD147 promoter activity, indicating a synergistic effect conferred by both factors on CD147 production. Taken together, our results validate that the hypoxia-responsive region of the CD147 promoter is composed of the HRE- and GC-rich sequences and provide the first evidence that CD147 is a novel HIF-1 target gene and its transcriptional activation under hypoxia is mediated at least partially through the cooperation of HIF-1 and Sp1. However, we also observed that there was not a strong correlation between HIF-1α and CD147 in MCF-7 and HCC-9204 cells under 1% O ₂ . This result suggests that there may be a complicated cell-specific mechanism of hypoxia-induced CD147 upregulation. In addition, previous reports have shown that hypoxia-inducible MCT4 regulates maturation and trafficking of CD147 to the plasma membrane in the metastatic breast cancer cell line MDA-MB-231 ( 42 ). These data suggest the presence of a cooperative mechanism at transcriptional and post-translational levels for hypoxia-induced CD147 upregulation. Therefore, it is not difficult to understand that CD147 is not as specific as a hypoxia sensor like HIF-1α.

We further explored the biological functions of upregulated CD147 in hypoxia adaptation. Various lines of strong experimental evidence have demonstrated that HIF-1 target genes are highly enriched for proteins that facilitate glycolytic switch under hypoxia ( 10 , 43) . Our data showed that the glycolytic activity of cancer cells was significantly inhibited by the knocking down of CD147 expression with subsequent decrease in glucose uptake, acid production and LDH activity, which is strongly supported by a recent study that CD147 silencing inhibits lactate transport and reduces malignant potential of pancreatic cancer cells ( 17 ). Mounting evidence indicated that hypoxia-induced tumor cell metabolic reprogramming is mainly orchestrated by HIF-1 through the transcriptional activation of key genes encoding metabolic enzymes, including LDHA, PDK1, PKM2, CA IX and Glut-1 ( 10 , 44) . Other important molecules such as Myc, p53 have also been reported to play central roles in this process ( 45 ). These reports provide reasonable explanation for our finding that si-CD147 inhibited the lactate production by 20% under hypoxia, whereas glycolysis decreases by 100% under normoxia. Moreover, our in vivo data further indicated that the inhibition of CD147 expression reduced ¹⁸ F-FDG uptake in a nude mouse xenograft model and thus confirmed that CD147 plays a vital role in constitutively upregulated glucose uptake within tumors. In short, our findings provide a novel molecular basis for altered tumor glycolysis that is mediated by a critical role of CD147 in promoting glycolytic switch under hypoxic microenvironmental stress.

Accumulating evidences have demonstrated that CD147 plays a pivotal role as an ancillary protein required for the expression and function of MCTs in regulating lactate efflux ( 16 , 17 , 46) . Consistently, our results indicated that knocking down of CD147 under hypoxia reduced the protein levels of both MCT1 and MCT4, suggesting the contribution of CD147 to the hypoxia-induced glycolytic phenotype of cancer cells is partly through the regulation of MCT1 and MCT4 expression. It is reasonable to expect that the expression level of Glut-1 may be reduced by silencing of CD147, while we did not observe a notable expression change of Glut-1 after CD147 knockdown, which indicated that CD147-mediated glycolytic switch might not be a direct outcome of the reduction of glucose transporter expression. Nonetheless, our preliminary data cannot rule out the possibility that other important mechanisms are involved in the CD147-mediated regulation of glycolysis function and thus future investigations are warranted.

Considering that hypoxia-induced ‘glycolytic switch’ within solid tumors is largely associated with tumor malignancy, our study provides further insights into the critical role of CD147 in promoting cell survival and invasion under hypoxic stress. A wide array of previous studies have demonstrated that CD147 serves as a pluripotent molecule involved in multiple stages of tumor progression with the abilities to promote tumor cell proliferation, invasiveness, metastasis and VEGF production ( 47–49 ). It has also been demonstrated that CD147 confers resistance of breast cancer cells to anoikis through inhibition of Bim ( 30 ). A recent study also revealed the anti-apoptotic effect of CD147 through coexpression with MCTs ( 50 ). Similarly, we observed that the knockdown of CD147 significantly increased the late-stage apoptosis of tumor cell under hypoxia, suggesting that hypoxia-induced CD147 upregulation may represent a novel adaptive response that increases cell survival. In addition, in earlier reports, it has been widely accepted that CD147 promotes tumor cell invasion by stimulating the secretion of matrix metalloproteinases ( 11 ). In consistency, our study found that the relative amount of invading cells decreased in transwell invasion assay after CD147 was inhibited under hypoxic condition. Taken together, we provide further solid evidence that CD147 serves as an important player in hypoxia adaptation for tumor cell survival and invasion.

In summary, our data demonstrate for the first time that hypoxia upregulates the expression of CD147 by a mechanism of transcriptional activation that involves a combined effect of HIF-1α and Sp1. Furthermore, our findings support the idea that hypoxia-induced CD147 upregulation significantly promotes glycolytic switch and promotes tumor cell survival and invasion. These results reveal a novel molecular mechanism for hypoxia adaptation of tumor cells and indicate the potential of CD147 as a promising target for cancer therapy, especially in selective killing of hypoxia-tolerated cancer cells.

Glossary

Abbreviations:

¹⁸ F-FDG

¹⁸ F-fluorodeoxyglucose

CA

carbonic anhydrase

ChIP

chromatin immunoprecipitation

Ct

threshold cycle

HIF

hypoxia-inducible factor

MMA

mithramycin

A; PI

propidium iodide

Sp1,

specificity protein 1

SNC

scramble negative control

YC-1

3-(5ʹ-hydroxy-methyl-2ʹ-furyl)-1-benzylindazole