Hypoxia-Independent Overexpression of Hypoxia-Inducible Factor 1α as an Early Change in Mouse Hepatocarcinogenesis

Hypoxia-inducible factor 1 (HIF-1) is involved in tumor progression/metastasis and activated in various cancers. Here we show that HIF-1α, which plays a major role in HIF-1 activation, is overexpressed in preneoplastic hepatocytic lesions from a very early stage during hepatocarcinogenesis in mice and man. Transcriptional targets of HIF-1, such as vascular endothelial growth factor, glut-1, c-met, and insulin-like growth factor II (IGF-II), were also overexpressed in mouse lesions. Oxygen tension within the lesions was not different from that of the normal hepatic tissues, indicating that HIF-1α expression was independent of hypoxia. On the other hand, Akt, the pathway of which can up-regulate HIF-1α expression, was activated in the mouse lesions, whereas HIF-1α was markedly down-regulated in the mouse hepatocellular carcinoma (HCC) cell lines after treatment with a phosphatidylinositol 3-kinase (PI3K) inhibitor, LY294002, indicating that HIF-1α expression is dependent on PI3K/Akt signaling. Conversely, HIF-1α knockdown by short interfering RNA in the HCC cell line resulted in decreased expression of activated Akt together with the HIF-1 target genes, indicating that Akt activation is reversely dependent on HIF-1 activation. Treating the HCC cells with IGF-II or epidermal growth factor (EGF) up-regulated both phospho-Akt and HIF-1α, whereas inhibition of IGF-II or EGF signaling down-regulated them both, suggesting that IGF-II and EGF can, at least in part, mediate the activation of Akt and HIF-1α. However, Akt was not activated by IGF-II or EGF in the HIF-1α knockdown cells, indicating that expression of the HIF-1 target genes is necessary for the Akt activation. These findings suggest that the reciprocal activation of PI3K/Akt signaling and HIF-1α may be important in the progression of hepatocarcinogenesis. (Cancer Res 2006; 66(23): 11263-70)

Although hepatocellular carcinomas (HCC) are one of the leading causes of cancer death in the world, the molecular mechanism of hepatocarcinogenesis, especially in the early stage, is still not sufficiently understood. During the early stage of hepatocarcinogenesis, focal lesions called hyperplastic or dysplastic hepatocytic foci emerge (1, 2). These lesions exhibit a number of altered gene expression and higher proliferating capacity compared with the surrounding normal hepatic tissue and grow into grossly visible lesions, called hepatic adenomas or dysplastic nodules, in the later stage. It is thought that, during the development of these lesions, the gradual accumulation of genetic changes may occur in the preneoplastic hepatocytes, leading to the final development of HCC cells. However, how the cellular signaling enabling the early preneoplastic hepatocytes to exhibit the growth advantage occurs is not exactly known.

Hypoxia-inducible factor 1 (HIF-1) is a transcription factor that enhances many types of gene expression including those involved in angiogenesis, cell proliferation, glucose metabolism, erythropoiesis, and cell survival (3). HIF-1 is composed of α and β subunits, where the β subunit is constitutively expressed whereas the α subunit is degraded under normoxic conditions despite the fact that it is continuously synthesized. Under normoxic conditions, HIF-1α undergoes hydroxylation at specific proline residues by HIF hydroxylase, is immediately ubiquitinized by binding with the von Hippel-Lindau (VHL) tumor suppressor protein, and then degraded by proteasomes (4–7). Under hypoxic conditions, however, HIF-1α is hydroxylated less and imported into the nucleus where it binds with HIF-1β and other transcriptional factors and coactivators to transactivate a variety of genes containing the hypoxia response element (8).

Recent studies have further shown that HIF-1 can be activated independent of hypoxia by various mechanisms such as oncogene activation (9, 10), inactivation of tumor suppressor genes (6, 11, 12), and activation of growth factor signaling (13–15). In particular, phosphatidylinositol 3-kinase (PI3K)/Akt and mitogen-activated protein kinase (MAPK) signaling are important for HIF-1α activation (13–15). The PI3K/Akt and MAPK signaling pathways activate mammalian target of rapamycin and the protein-synthesizing machinery, which in turn up-regulates HIF-1α expression, leading to HIF-1 activation. Because HIF-1 transactivates growth factor genes, such as vascular endothelial growth factor (VEGF), insulin-like growth factor 2 (IGF-II), and transforming growth factor α (Tgf-α), the products of which can activate PI3K/Akt and MAPK signaling, HIF-1 can also be activated by autocrine mechanisms (3).

Although HIF-1α has been shown to be frequently overexpressed in HCCs as well as various other types of cancers (16, 17), it is still not clear whether it occurs in the early stages of hepatocarcinogenesis. To clarify the role of HIF-1 in hepatocarcinogenesis, we investigated HIF-1α expression in chemically induced preneoplastic hepatocytic lesions in mice and dysplastic hepatocytic foci in man. Because IGF-II and TGF-α, transcriptional targets of HIF-1 that can activate PI3K/Akt signaling, have been reported to be frequently overexpressed in preneoplastic hepatocytic lesions and HCC tissues and cell lines (18–20), we also investigated the possibility that these autocrine factors are related to the activation of HIF-1α and PI3K/Akt signaling.

Tissue samples. To induce hepatocarcinogenesis in mice, male B6C3F1 mice (Clea Japan, Tokyo, Japan) were administered diethylnitrosamine i.p. at a dose of 5 μg/g body weight at age 2 weeks (21) and sacrificed 5 to 15 months after diethylnitrosamine treatment. Hepatic tumors were removed from the surrounding normal liver tissue, frozen in liquid nitrogen, and stored at −80°C until use. For some mice, the livers were perfusion fixed with 4% paraformaldehyde dissolved in PBS, further immersed in the same solution for 12 hours at 4°C, and processed for paraffin embedding. All experimental procedures were approved by the institutional committee according to the guidelines for animal protection. Paraffin blocks of human liver biopsy samples containing dysplastic hepatocytic foci were also used after obtaining the informed consent of patients.

Measurement of tumor oxygen tension. Tumor oxygen partial pressure (pO₂) was measured with polarographic needle electrode. Briefly, the mice bearing hepatic tumors were anesthetized by diethyl ether, kept on a heating pad to maintain body temperature between 35°C and 36°C, and laparotomized. After stabbing the O₂ needle probe into the tumors, pO₂ was monitored using a pO₂ monitor (POG-500, MT-Giken, Tokyo, Japan). The electrode-stabbed tissues were histologically examined after the measurement.

Tissue hypoxia analysis. Mice with hepatic lesions were administered pimonidazole hydrochloride (Chemicon, Temecula, CA) i.p. dissolved in PBS at a dose of 60 μg/g body weight. After 2 hours, the livers were perfusion fixed with the paraformaldehyde solution and embedded in paraffin, followed by immunohistocehmical detection of the pimonidazole protein adducts as detailed below.

Immunohistochemistry. For immunohistochemistry, serial 3.5-μm-thick paraffin sections were sequentially treated before application with the primary antibodies in the following way: deparaffinization, rehydration, endogenous peroxidase quenching, and antigen retrieval. Antigen retrieval involved incubation with citrate buffer [0.1 mol/L citrate (pH 6.0), 0.1% NP40] in an autoclave for HIF-1α and pimonidazole, and treatment with proteinase K (Dako, Carpinteria, CA) for phospho-Akt. To detect pimonidazole, the sections were incubated with the FITC-conjugated monoclonal antibody against pimonidazole (Chemicon) and then with horseradish peroxidase (HRP)–conjugated secondary antibody against FITC (1:50 dilution; Chemicon), followed by development of color with diaminobenzidine. The immuhistochemistry for HIF-1α and activated Akt was carried out with anti-HIF-1α monoclonal antibody (1:50; Sigma-Aldrich, St. Louis, MO) and anti–phospho-Akt (Ser⁴⁷³) polyclonal antibody (1:50; Cell Signaling, Beverly, MA), respectively, followed by detection of antibody binding using the Catalyzed Signal Amplification System II (Dako). For all these procedures, staining without the primary antibodies was done as a negative control.

Cell culture. HCC cell lines, which had previously been established from primary mouse HCCs in our laboratory (22), were cultured in Williams' E medium supplemented with 10% fetal bovine serum (FBS), 10⁻⁷ mol/L epidermal growth factor (EGF; Toyobo, Japan), 10⁻⁷ mol/L transferrin, 10⁻⁷ mol/L insulin, 10⁻⁷ mol/L dexamethasone, 10⁻⁵ mol/L aprotinin, 100 units/mL penicillin, and 100 μg/mL streptomycin. All cultures were maintained at 37°C in a humidified atmosphere containing 5% CO₂. For IGF-II or EGF stimulation, the HCC cells were exposed to starvation medium (Williams' E medium containing the antibiotics alone) for 6 hours and then incubated with Williams' E medium supplemented with 5% FBS and recombinant mouse IGF-II (R&D Systems, Minneapolis, MN) or EGF for 24 hours. Goat anti-mouse IGF-II polyclonal antibody (R&D Systems) and nonimmune goat serum were added to the medium for IGF-II neutralization and control, respectively. To inhibit the growth factor signaling, LY294002 (PI3K inhibitor, 40 μmol/L; Sigma-Aldrich), AG1024 [IGF-I receptor (IGF-IR) inhibitor, 3 μmol/L; Calbiochem, San Diego, CA], or AG1478 [EGF receptor (EGFR) inhibitor, 3 μmol/L; Calbiochem] was added to the culture medium.

Immunocytofluorescence. For immunocytofluorescence, the HCC cells were cultured on collagen-coated cover glass and fixed with paraformaldehyde solution. The cells were then incubated first with the anti-HIF-1α monoclonal antibody and then with the Cy3-conjugated secondary antibody (Sigma-Aldrich), followed by nuclear staining with 4′,6-diamino-2-phenylindole dihydrochloride (DAPI). The numbers of HIF-1α-positive nuclei were counted in five randomly selected microscopic fields (×200), each containing ∼200 cells, and the percentage of the HIF-1α-positive nuclei against total DAPI-positive nuclei was calculated.

Western blot analysis. The tissue and cell samples were lysed in SDS sample buffer, separated with 10% SDS-acrylamide gel, and electrotransferred to nitrocellulose membranes. After blocking with 5% nonfat dry milk in TBST buffer [10 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 0.05% Tween 20], the membranes were probed with anti-HIF-1α (1:1,000), anti-VEGF (1:500; Santa Cruz Biotechnology, Santa Cruz, CA), anti-glut-1 (1:500; Santa Cruz Biotechnology), anti-IGF-II (1:2,500; R&D Systems), anti-c-met (1:500; Santa Cruz Biotechnology), anti-α-tubulin (1:1,000; Santa Cruz Biotechnology), anti-Akt (1:1,000; Cell Signaling), and anti–phospho-Akt (Ser⁴⁷³, 1:1,000; Cell Signaling) antibodies, followed by incubation with HRP-conjugated antirabbit or antimouse immunoglobulin G secondary antibodies (1:2,000; Amersham Biosciences, Buckinghamshire, United Kingdom). The antibody binding was then visualized with enhanced chemiluminescence reagent (Amersham Biosciences), and the band images detected with the LAS3000 system (Fuji Film, Tokyo, Japan) were densitometrically analyzed using the Image Gauge (Fuji Film).

Reverse transcription-PCR. Total RNA was isolated from each cell line with Trizol reagent (Invitrogen, Carlsbad, CA) and then reverse transcribed with the oligo dT primer and SuperScript RT (Invitrogen). The PCR primers used were as follows: HIF-1α, 5′-TCAAGTCAGCAACGTGGAAG-3′ (forward) and 5′-TATCGAGGCTGTGTCGACTG-3′ (reverse); IGF-II, 5′-GAGTTCAGAGAGGCCAAACG-3′ (forward) and 5′-CCTGCTCAAGAGGAGGTCAC-3′ (reverse); TGF-α, 5′-AGCATGTGTCTGCCACTCTG-3′ (forward) and 5′-TGGATCAGCACACAGGTGAT-3′ (reverse); and β-actin, 5′-AGCCATGTACGTAGCCATCC-3′ (forward) and 5′-CTCTCAGCTGTGGTGGTGAA-3′ (reverse). The thermal cycle profile consisted of a 1-minute extension at 72°C, 1-minute denaturation at 95°C, 1-minute annealing at 56°C, and a final 7-minute extension. Quantitative real-time reverse transcription-PCR (RT-PCR) was done for IGF-II and TGF-α mRNA using the Roche Molecular Biochemicals (Basel, Switzerland) LightCycler with LightCycler3 Run software (version 5.10). Fluorescence was generated using Platinum SYBR Green qPCR SuperMix UDG (Invitrogen) and data were collected with LightCycler3 Data Analysis software (version 3.5.28).

Establishment of HIF-1α knockdown HCC clones. Referring to the technical information of Ambion (Austin, TX), short interfering RNA (siRNA) was designed to interfere with mouse HIF-1α expression. A set of 19-mer oligonucleotides was selected from the mouse HIF-1α mRNA sequence as a target, and it was confirmed that these sequences would not interfere with other genes by analyzing the homology to other mouse genes by a BLAST search. The sense (5′-AGATGAGTTCTGAACGTCG-3′) and antisense sequences (5′-CGACGTTCAGAACTCATCT-3′) were linked to the nine-nucleotide spacer (5′-TTCAAGAGA-3′) as a loop and six T bases were added to the 3′ end of the oligonucleotides. In addition, reverse oligonucleotides corresponding to the forward sequences were also designed. The BamHI (GATCC) and HindIII (AGCTT) restriction site sequences were also added to the 5′ end of the forward and 3′ end of the reverse oligonucleotides, respectively. The DNA was ligated with linearized pSilencer 3.1-H1 hygro siRNA expression vector (Ambion) at the BamHI and HindIII sites to construct a mouse HIF-1α siRNA vector (pSH1a-1). An unrelated control siRNA vector (pSGFP), which targets the green fluorescence protein (GFP) DNA sequence (5′-GGTTATGTACAGGAACGCA-3′) with no significant match to any mouse gene, was also prepared. The HCC cells were transfected with 1 μg each of the control (pSGFP) and HIF-1α knockdown (pSH1a-1) vectors that contained the hygromycin resistance gene using Lipofectamine (Invitrogen). The transfected cells were maintained in the culture medium containing hygromycin B (0.5 mg/mL) for 7 days and the stable control (sGFP) and HIF-1α knockdown (sH1a-cl2 and sH1a-cl3) cell lines were established.

In vitro and in vivo growth assay. For the in vitro growth assay, 2 × 10⁵ original HCC cells, control cells (sGFP), or HIF-1α knockdown cells (sH1a-cl2 and sH1a-cl3) were seeded on a 35-mm dish, harvested with trypsin-EDTA solution on days 2, 4, and 6, and cell numbers were microscopically examined with a hemocytometer. For the in vivo growth assay, 1 × 10⁶ cells suspended in 200 μL of PBS were injected into the s.c. tissues of the back skin of B6C3F1 mice. Tumor diameters (d short and d long) were measured with calipers, and the tumor volume was calculated using the following formula: volume = d short² × d long / 2 (23). The tumors on day 7 after inoculation were histologically examined. Both the in vitro and in vivo growth assays were done in triplicate and repeated twice. The data were presented as the values ± SD.

Statistics. Statistical analysis of the data was done with the Student paired t test; P < 0.05 was considered to be significant.

Expression of HIF-1α and the HIF-1 target genes in preneoplastic lesions. In the mice treated with diethylnitrosamine at age 2 weeks, livers contained preneoplastic foci and adenomas, which were respectively smaller and larger than the hepatic lobule size (24) 5 to 9 months later. These lesions were found to be round in shape with weak compression against the surrounding hepatic tissue and were composed of preneoplastic hepatocytes with slightly basophilic cytoplasm (Fig. 1A,-a). HIF-1α immunohistochemistry detected positive staining in the nucleus and cytoplasm of most preneoplastic hepatocytes in all the lesions of the five mice (Fig. 1A,-b). HCCs developed in the livers 12 to 15 months after diethylnitrosamine treatment (Fig. 1B,-a). These HCCs also showed positive HIF-1α staining in the nucleus and cytoplasm (Fig. 1B,-b). The human liver biopsy samples occasionally contained small lesions called dysplastic hepatocytic foci, likely to be of a preneoplastic nature as were the mouse lesions (1), which consisted of altered hepatocytes with clear cytoplasm (Fig. 1C,-a). Positive HIF-1α staining was also shown in the cytoplasm and nucleus of the dysplastic hepatocytes in the three human lesions examined (Fig. 1C,-b). Western blot analysis revealed that HIF-1α and the products of the HIF-1 target genes, VEGF, glut-1, IGF-II, and c-met, were expressed more abundantly in the hepatic adenomas than in the normal liver tissue in mice (Fig. 1D).

Lack of hypoxia in the mouse preneoplastic lesions. Measurement of oxygen tension using the needle electrode showed that the pO₂ in five adenomas and three HCCs were not different from that in the normal liver tissues (Fig. 2A). To further confirm this, 2 hours before sacrifice, the mice bearing the hepatic lesions were administered pimonidazole, which creates adducts with thiol-containing proteins in hypoxic (pO₂ < 10 mm Hg) cells (25), followed by immunohistochemical detection of pimonidazole. In the normal hepatic tissue, although pimonidazole staining was weakly positive in the hepatocytes around the terminal hepatic vein as previously described (26), it was negative in other parts (Fig. 2B,-a). The staining was negative in the preneoplastic foci or adenomas (Fig. 2B,-b), whereas HIF-1α was positively stained in the contiguous sections (data not shown). On the other hand, pomonidazole staining was mostly negative within the HCC tissues (Fig. 2B,-c) but strongly positive in the focal areas (Fig. 2B -d). These observations indicated that HIF-1 expression in the preneoplastic hepatic lesions is independent of hypoxia, whereas it may be mostly independent of and partly dependent on hypoxia in the HCC tissues.

Regulation of HIF-1α expression by PI3K/Akt signaling. Because PI3K/Akt signaling can activate HIF-1α expression independent of hypoxia in a cell type–specific manner (13–15), we then explored whether HIF-1α expression in the mouse preneoplastic hepatic lesions was dependent on the activation of PI3K/Akt signaling. The results in Fig. 3A showed that all mouse hepatic adenomas exhibited higher levels of the activated form of Akt [phospho-Akt (Ser⁴⁷³)] than normal livers without any change in the total amounts of Akt and α-tubulin presented as the internal control. Immunohistochemistry for activated Akt showed positive staining in the nucleus and cytoplasm of preneoplastic hepatocytes in all mouse lesions examined (Fig. 3B). We then investigated whether the high HIF-1α expression correlated to the activation of PI3K/Akt signaling using the mouse HCC cell lines. These cell lines showed active cell proliferation after treatment with growth factors (EGF and insulin) plus 10% FBS (GF/FBS+), whereas cell growth was stopped in the absence of the growth factors and FBS (GF/FBS−; ref. 27). These HCC cells exhibited high HIF-1α and activated Akt levels under GF/FBS+ conditions, whereas both HIF-1α and activated Akt were down-regulated under GF/FBS− conditions (Fig. 3C). Moreover, both HIF-1α and activated Akt levels were markedly decreased by treatment with LY294002, a PI3K inhibitor, under GF/FBS+ conditions (Fig. 3C). The HIF-1α immunocytofluorescence showed that HIF-1α was strongly positive in cytoplasm and in ∼40% nuclei under the GF/FBS+ condition, but it was very weakly positive mainly in the cytoplasm under the GF/FBS− condition or in the presence of LY294002 (Fig. 3D -a, b). These observations strongly support the hypothesis that high HIF-1α expression in the HCC cells, possibly in preneoplastic hepatic lesions as well, may be due to activated PI3K/Akt signaling.

Effects of HIF-1α knockdown in HCC cell lines. In an effort to further investigate the role of HIF-1α activation in hepatocarcinogenesis, we established two HIF-1α knockdown clones (sH1a-cl2 and sH1a-cl3) from the mouse HCC cell lines by stable expression of the mouse HIF-1α siRNA. The densitometric analysis of Western blotting revealed that the two HIF-1α knockdown clones showed an ∼70% reduction of HIF-1α protein expression compared with the original HCC cells and the control cell line introduced with the GFP siRNA (Fig. 4A). In accordance with the reduction of HIF-1α expression, expression of the HIF-1 target genes (VEGF, glut-1, c-met, and IGF-II) was also reduced in HIF-1α knockdown cells. Furthermore, activated Akt was also down-regulated in HIF-1α knockdown cells without any changes in the total amount of Akt (Fig. 4A), suggesting that the activation of PI3K/Akt signaling is reversely dependent on the high HIF-1α expression. In the in vitro growth assay, proliferation of the HIF-1α knockdown cells was slower than the original HCC cells or the control GFP siRNA–introduced cells in the GF/FBS+ medium (Fig. 4B). Moreover, although s.c. inoculation of the original HCC cells or the control cells (sGFP) resulted in tumor formation in B6C3F1 mice, the HIF-1α knockdown cells temporally formed small tumors during the 4- to 6-day period after inoculation, but these tumors disappeared by day 10 (Fig. 4C). Histologic examination showed that the small tumors of the HIF-1α knockdown cells contained a large necrotic area in their central part (Fig. 4D,-c, d), whereas such change was not seen in the tumors of the original and control cells (Fig. 4D -a, b). These findings indicate that down-regulation of HIF-1α resulted in remarkable growth retardation of HCC cells both in vitro and in vivo.

Increased expression of activated Akt and HIF-1α by growth factors. Down-regulation of HIF-1α expression by inactivation of PI3K/Akt signaling and down-regulation of activated Akt by HIF-1α knockdown suggest a possible cause-and-effect relationship between the two phenomena. Because IGF-II is known to activate the PI3K/Akt pathway via binding to the cognate IGF-IR and hybrid IGF-I/insulin receptor (28) and also because IGF-II has been shown to be one of the transcriptional targets of HIF-1 (29–31), it is possible that activation of HIF-1, IGF-II, and PI3K/Akt signaling may create an autocrine loop (3). Furthermore, because IGF-II is frequently overexpressed in HCC tissue and cell lines (18, 19), it may behave like an autocrine growth factor. As shown in Figs. 1D and 4A, respectively, most mouse hepatic adenomas and the HCC cell line overexpressed IGF-II protein. In the HIF-1α knockdown HCC cells, IGF-II protein (Fig. 4A) and mRNA levels (Fig. 5A,-a) were markedly down-regulated. Quantitative real-time RT-PCR showed that the IGF-II mRNA level was decreased to ∼50% in the HIF-1α knockdown cells compared with the original and control cells (Fig. 5A,-b). To investigate whether IGF-II contributed to the expression of both activated Akt and HIF-1α, the HCC cells were cultured in the starvation medium (GF/FBS−) for 6 hours and then in medium containing various concentrations of mouse recombinant IGF-II together with 5% FBS. The results in Fig. 5B,-a show that the expression of HIF-1α, VEGF, and activated Akt proteins was up-regulated by IGF-II in a dose-dependent manner in the original HCC cells. On the other hand, when the cells were treated simultaneously with IGF-II and the neutralizing IGF-II antibody or the AG1024 IGF-IR inhibitor, expression of both activated Akt and HIF-1α was reduced without any change in the total amount of Akt (Fig. 5C,-a,b). Surprisingly, however, when the HIF-1α knockdown cells were treated with IGF-II, Akt activation did not occur at all (Fig. 5B,-a). We also addressed whether TGF-α, another possible autocrine growth factor that is one of the HIF-1 target genes (3) and is frequently overexpressed in preneoplastic hepatic lesions (20), has a similar effect on IGF-II. RT-PCR revealed that TGF-α mRNA was decreased in HIF-1α knockdown cells (Fig. 5A,-a), the level of which was 45% to 65% as compared with the original cells and 55% to 75% as compared with the control cells (Fig. 5A,-b). The effect of TGF-α is thought to be mimicked by that of EGF because both TGF-α and EGF bind to the EGFR to activate the downstream signaling (32). HIF-1α, VEGF, and Akt were activated by EGF in a dose-dependent manner in the original HCC cells (Fig. 5B,-b), whereas the activation by EGF was reduced in the presence of the EGFR inhibitor AG1478 (Fig. 5C,-c). In addition, by treatment with EGF, Akt was only very weakly activated in HIF-1α knockdown cells (Fig. 5B -b). These observations suggest that the overexpression of activated Akt and HIF-1α in the original HCC cells may have been, at least in part, mediated by IGF-II, and possibly by TGF-α as well, whereas they could no longer activate Akt in HIF-1α knockdown cells where the HIF-1 target genes had been suppressed.

There is increasing evidence suggesting that HIF-1 activation occurs in the early stages of carcinogenesis. It has been shown that in the kidneys of patients with von Hippel-Lindau's disease, which is frequently associated with renal carcinomas, HIF-1α and the HIF-1 target genes are overexpressed in morphologically normal single renal tubular cells and epithelial cells, forming multicellular foci or microcysts similar to overt renal carcinoma cells (33). HIF-1α was also shown to be expressed in a few cells in ductal hyperplastic areas adjacent to invasive breast cancer (34), dysplastic uterine cervical epithelium (35), colorectal adenomas (36), and prostate intraepithelial neoplasia (37) like their malignant counterparts. Furthermore, HIF-1α and HIF-1 target genes were shown to be overexpressed in hyperplastic and dysplastic skin lesions during multistage epidermal carcinogenesis in K14-HPV16 transgenic mice where the oncogenic proteins of human papillomavirus 16 (HPV16) were overexpressed in epidermal cells under the control of the keratin 14 gene promoter (38). In the present study, to the best of our knowledge, we are the first to show that HIF-1α is activated in preneoplastic hepatocytes during the early stages of hepatocarcinogenesis and long before the development of HCCs.

pO₂ in mouse hepatic tumors was the same as that in normal liver tissues, and pimonidazole staining was not detected in the preneoplastic lesions, indicating that HIF-1α expression was independent of hypoxia. Because the early preneoplastic hepatic lesions examined were generally small in size, with dilated sinusoids covered with the endothelial cells with characteristic features such as loss of fenestrations³

and not showing any necrosis, they may be refractory to hypoxia. In contrast, positive pimonidazole staining was seen in the focal areas within HCCs, which occasionally showed necrosis, suggesting that HIF-1α expression in HCCs may be mostly independent of and partly dependent on hypoxia. HIF-1α expression independent of hypoxia has been reported for human cancer tissues (39) and cell lines (40). It has been shown that HIF-1α can be activated not only by hypoxia but also by other various mechanisms such as the activation of oncogenes (9, 10), loss of function in tumor suppressor genes such as PTEN (11), VHL (6), and p53 (12), and activation of growth factor signaling (13–15).

In the present study, preneoplastic hepatic lesions showed increased levels of phospho-Akt compared with the normal livers, although total Akt levels were unchanged. On the other hand, when HCC cells were treated with LY294002, a PI3K inhibitor, not only the levels of phospho-Akt but also those of HIF-1α were down-regulated, suggesting that HIF-1α expression is dependent on PI3K-Akt signaling. Akt can activate mammalian target of rapamycin, which phosphorylates p70 S6 kinase and the eukaryotic translation initiation factor 4E binding protein (41), leading to activation of protein synthesis (42). The increased activity of protein-synthesizing machinery up-regulates HIF-1α protein synthesis without affecting HIF-1α mRNA levels, HIF-1α ubiquitination, and the capability of HIF-1α to interact with pVHL (13, 43). Our observations strongly suggest that activated PI3K/Akt signaling increases HIF-1α expression in HCC cell lines and presumably preneoplastic hepatic lesions as well.

Conversely, when HIF-1α was knocked down by siRNA in the HCC cells, not only the expression of the HIF-1α target genes but also the phospho-Akt levels were down-regulated without any change in total Akt, suggesting that the activation of Akt is reversely dependent on HIF-1α activation. Among HIF-1 targets, growth factors such as IGF-II, TGF-α, and VEGF and growth factor receptors such as c-met may activate PI3K/Akt signaling by activating the receptor-associated tyrosine kinases (3). In the present study, IGF-II was overexpressed in most preneoplastic lesions and the HCC cell lines. Furthermore, treatment of the HCC cells with IGF-II in the 5% serum condition up-regulated both phospho-Akt and HIF-1α, whereas simultaneous treatment with IGF-II and the neutralizing anti-IGF-II antibody or AG1024 IGF-IR inhibitor down-regulated them both, indicating that, at least in part, IGF-II plays a role in the activation of PI3K/Akt signaling and subsequent HIF-1α activation in HCC cells. On the other hand, TGF-α mRNA was increased in the original and control HCC cells but was decreased in HIF-1 knockdown cells. TGF-α, which is also one of the target genes of HIF-1 (3), is frequently activated during hepatocarcinogenesis, presumably acting as an autocrine growth factor (20). Because TGF-α interacts with the EGFR to activate intracellular signaling, the observation that EGF up-regulated both HIF-1α and phospho-Akt expression in the HCC cells indicates that TGF-α may also have a role in the activation of HIF-1α by activating the PI3K/Akt signaling. These observations strongly suggest that HIF-1, IGF-II, TGF-α, and PI3K/Akt signaling may create an autocrine loop and may have an important role in the progression of hepatocarcinogenesis. However, it remains to be investigated which is the initiating change in such autocrine mechanism.

On the other hand, the capacity of IGF-II and EGF to activate Akt was abrogated in HIF-1α knockdown cells, suggesting that some factors directly or indirectly activated by HIF-1 may have a role in the activation of PI3K/Akt signaling. For example, suppression of IGF binding proteins, which are also targets of HIF-1 (3) and can potentiate or suppress IGF-II activity (44), may modulate the responsiveness to IGF-II in HIF-1α knockdown cells. Recently, Calvani et al. (45) reported that basic fibroblast growth factors (bFGF), which induce the survival and sprouting of human umbilical vascular endothelial cells in vitro, were abrogated by HIF-1α knockdown with siRNA. Furthermore, it was shown that HIF-1 enhances the responsiveness to bFGF in endothelial cells by increasing the binding of bFGF to heparan sulfate on the cell surface (46). It is then thought that HIF-1 may amplify the growth factor signaling by inducing not only the autocrine growth factors but also those factors that modulate the responsiveness to the growth factors.

c-Met, a receptor of hepatocyte growth factor (HGF) and also one of the targets of HIF-1 (47, 48), was up-regulated in preneoplastic hepatic lesions and HCC cell lines. HIF-1α knockdown resulted in the down-regulation of c-met expression, suggesting that c-met expression is regulated by HIF-1 in HCC cell lines. Recently, LeCouter et al. (49) showed that sinusoidal endothelial cells produce HGF by VEGF stimulation via the VEGF receptor-1 (Flt-1) activation in mice. It is then possible that HIF-1 activation stimulates VEGF production by preneoplastic hepatocytes, which, in turn, may stimulate the sinusoidal endothelial cells to produce HGF. On the other hand, HIF-1 may increase c-met expression in preneoplastic hepatocytes, leading to sensitization to HGF. Furthermore, it was shown that HGF/c-met can activate PI3K/Akt signaling, which activates HIF-1α and HIF-1 target genes in HepG2 cells (50). It is then possible that HIF-1, VEGF, and c-met may create a paracrine loop involving preneoplastic hepatocytes and sinusoidal endothelial cells, although further studies are required.

In conclusion, overexpression of HIF-1α and activation of its target genes are changes in the early stages of hepatocarcinogenesis. In particular, activation of PI3K/Akt signaling, IGF-II, and possibly TGF-α as well, under the influence of HIF-1 activation may be important for the proliferation of preneoplastic hepatocytes. Therefore, HIF-1 activation may be crucial in the progression of hepatocarcinogenesis by expanding preneoplastic hepatocyte populations, which in turn increases the chance for accumulation of oncogenic mutations within the populations. Intervention of the HIF-1 pathway may then be effective to prevent the development of HCCs.

Grant support: The Japanese Ministry of Education, Culture, Sports, Science and Technology and the Japanese Ministry of Health, Welfare and Labour.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Masanobu Kobayashi (Division of Cancer Biology, Institute for Genetic Medicine, Hokkaido University, Sapporo, Japan) for critically reading the manuscript and giving helpful suggestions, and Professor Akira Takai and Associate Professor Masaaki Hashimoto (Department of Physiology, Asahikawa Medical College, Asahikawa, Japan) for their support on measurement of tumor oxygen tension.