Hypoxia Induces c-fos Transcription via a Mitogen-activated Protein Kinase-dependent Pathway*

Hypoxia is a pathophysiological condition that occurs during injury, ischemia, and stroke. It is characterized by a decrease of reactive oxygen intermediates and a change of the intracellular redox level. In tumors hypoxia is regarded as a trigger for enhanced growth and metastasis. Here we report that in HeLa cells, hypoxic conditions induce the transcriptional activation of c-fos transcription via the serum response element. Mutations in the binding site for the ternary complex factor Elk-1 and the serum response factor abolished this induction, indicating that a ternary complex at the serum response element is necessary for the induction of the c-fos gene under hypoxia. The transcription factor Elk-1 was covalently modified by phosphorylation in response to hypoxia. Furthermore this hyperphosphorylation of Elk-1, the activation of mitogen-activated protein kinase (MAPK), and the induction of c-fos transcripts were blocked by PD98059, a specific inhibitor of mitogen-activated protein kinase kinase/extracellular signal-regulated protein kinase kinase 1. Anin vitro kinase assay with Elk-1 as substrate showed that MAPK is activated under hypoxia. The activation of MAPK corresponds temporally with the phosphorylation and activation of Elk-1. Thus, a decrease of the intracellular reactive oxygen intermediate level by hypoxia induces c-fos via the MAPK pathway. These results suggest that the intracellular redox levels may be directly coupled to tumor growth, invasion, and metastasis via Elk-1-dependent induction of c-Fos controlled genes.

Hypoxia is a pathophysiological condition that occurs during injury, ischemia, and stroke. It is characterized by a decrease of reactive oxygen intermediates and a change of the intracellular redox level. In tumors hypoxia is regarded as a trigger for enhanced growth and metastasis. Here we report that in HeLa cells, hypoxic conditions induce the transcriptional activation of c-fos transcription via the serum response element. Mutations in the binding site for the ternary complex factor Elk-1 and the serum response factor abolished this induction, indicating that a ternary complex at the serum response element is necessary for the induction of the c-fos gene under hypoxia. The transcription factor Elk-1 was covalently modified by phosphorylation in response to hypoxia. Furthermore this hyperphosphorylation of Elk-1, the activation of mitogen-activated protein kinase (MAPK), and the induction of c-fos transcripts were blocked by PD98059, a specific inhibitor of mitogen-activated protein kinase kinase/extracellular signal-regulated protein kinase kinase 1. An in vitro kinase assay with Elk-1 as substrate showed that MAPK is activated under hypoxia. The activation of MAPK corresponds temporally with the phosphorylation and activation of Elk-1. Thus, a decrease of the intracellular reactive oxygen intermediate level by hypoxia induces c-fos via the MAPK pathway. These results suggest that the intracellular redox levels may be directly coupled to tumor growth, invasion, and metastasis via Elk-1-dependent induction of c-Fos controlled genes.
The activation of c-fos by mitogens and changes of the intracellular redox level is mainly mediated by the serum response element (SRE). 1 The SRE is a regulatory element found in many growth factor-regulated promotors that directs the rapid induction of gene expression (for review see Ref. 1). The best studied SRE is that of the c-fos gene. Two kinds of transcription factors are required for SRE activity: the ubiquitous transcription serum response factor (SRF) (2) and the ternary complex factors (TCFs), which form a ternary complex with the SRF. The human TCFs include Elk-1, SAP-1, and SAP-2 and constitute a subfamily within the Ets family of transcription factors (for review see Ref. 3). These proteins need SRF to bind tightly to the SRE (4). The N-terminal domains of Elk-1 and SAP-1 mediate DNA contact and ternary complex formation. The transactivation domain at the C terminus contains several conserved MAPK phosphorylation sites. Growth factor-stimulated activation of the MAPK pathway results in phosphorylation of the Elk-1 C terminus (5), which then cooperates with the SRF C-terminal activation domain to activate transcription (6,7).
The transcription factor AP-1 is composed of members of the Fos family (c-Fos, Fos-B, Fra-1, and Fra-2) and the Jun family (c-Jun, JunB, and JunD) that form restricted homo-or heterodimers (for review see Ref. 8). With the exception of c-Jun homodimers, AP-1 is predominantly induced at a transcriptional level by novel synthesis of its subunits. This induction is controlled by cis-acting SRE elements and AP-1 binding sites in the promotors of several AP-1 genes. A large number of mitogenic and proinflammatory signals lead to activation of AP-1. Likewise, changes in the cellular redox status were reported to activate AP-1 (9,10). Changes of the cellular redox status are directly coupled to the intracellular level of ROIs. For example, hypoxia and antioxidants reduce the intracellular level of ROIs. A pathophysiological role for reduced levels of ROIs is evident in tumor initiation, growth, and metastasis. We have previously reported that treatment of cells with antioxidants such as pyrolidinedithiocarbamate or N-acetyl-L-cysteine (11), as well as overexpression of thioredoxin (12), strongly activates AP-1. Recently we and others could show that hypoxia, a physiological correlate to antioxidants, also activates AP-1 (13,14) and that this depends on novel gene transcription.
The activation of AP-1 by hypoxia prompted us to investigate the transcription factors responsible for AP-1 gene induction. The c-Fos protein is a component of the hypoxia-induced AP-1 complex (14). The major cis-acting element responsible for the induction of c-fos transcription in response to a variety of mitogens is the SRE (15,16). Here we show that hypoxia induces c-fos transcription via the SRE. Both the binding site for the ternary complex factor as well as the binding site for SRF are required for this induction. Hypoxia rapidly induces a hyperphosphorylated ternary complex at the c-fos SRE, which contains the ternary complex factor Elk-1. Furthermore, the activation of MAPK ERK2 but not JNK/SAPK parallels the activation of Elk-1. We therefore conclude that the activation of Elk-1 by the MAPK pathway represents an important link between hypoxia-induced cellular redox imbalance and the activation of the transcription factor AP-1 via c-fos transcription.

EXPERIMENTAL PROCEDURES
Cell Culture and Treatment-The human cervical carcinoma cell line HeLa (ATCC, CCL2) was cultured as described in Ref. 14. Hypoxic conditions and PMA stimulation were applied as described in Ref. 14. Preparation of cell extracts cells were cultivated in the absence of serum. For transfections cells were kept in 10% fetal calf serum (Sigma) 14 h before stimulation. In the experiments using PD98059 [2-(2Јamino-3Ј-methoxyphenyl)-oxanaphthalen-4-one] (New England Biolabs Inc.), cells were incubated with 10 M PD98059 for 1 h prior to hypoxia.
Electrophoretic Mobility Shift Assays (EMSAs)-3 ϫ 10 5 cells/60-mm dish were serum-deprived overnight in Dulbecco's modified Eagle's medium and treated with hypoxic medium for the indicated time as described in Ref. 14. The hypoxic medium was aspirated after the treatment, and the culture dish with the adherent cells was immediately frozen on liquid nitrogen. Cells were lysed in whole cell extract buffer during thawing (17). EMSA binding reactions were performed as described in Refs. 18 and 19). For Fig. 3, 6 l of Elk-1 antibody (Santa Cruz Biotechnology Inc.) or a control antibody (␣-glutathione S-transferase, Santa Cruz Biotechnology Inc.) was added to 0.5 g of cell extract and incubated for 40 min on ice. Phosphatase inhibitors present in the extracts used in Fig. 3B were removed. To achieve this whole cell extracts were microdialyzed on VS filters (0.025 m; Millipore) against phosphatase buffer (18). Phosphatase assays were performed as described (18). Gels were vacuum-dried and exposed to Kodak XAR5 films at Ϫ80°C for 12-48 h.
In Vitro MAPK Kinase Assay-HeLa cells (2 ϫ 10 6 cells) were treated as described (14). 56 g of total cellular extract were immunoprecipitated with a phospho-MAPK antibody, washed, and incubated with a glutathione S-transferase-Elk-1 fusion protein as substrate. The reaction was separated by SDS gel electrophoresis, blotted, and incubated with an antibody directed against phosphorylated Ser 383 of Elk-1 according to the instructions of the manufacturer (New England Biolabs Inc.).
Northern Blot Analysis-HeLa cells (1 ϫ 10 6 /100 mm dish) were treated with hypoxic medium for 15-90 min. Total cellular RNA was isolated as described in Ref. 20. Separation of 10 g total RNA/lane on a formaldehyde-agarose gel, blotting, hybridization, and washing steps were performed as described (21), except that hybridization was performed in Rapidhyb buffer (Amersham Corp.) with a 32 P-labeled c-fos cDNA probe. RNA integrity and equal amount were confirmed using UV shadowing and methylene blue staining of the membrane (not shown).
Transfection-Transient transfections were performed essentially as described (19,22). The reporter plasmids pSRE 2 -tk80-luc (22), pEL 2 -tk80 (⌬TCF), and pM 2 -tk80 (⌬SRE) were gifts of A. Nordheim. pEL 2 -tk80 and pM 2 -tk80 were identical to pSRE 2 -tk80-luc except they contained mutations in the SRE that drastically reduce the DNA binding of TCF/Elk-1 and SRF, respectively (4). Briefly 10 5 HeLa cells were plated in 60-mm dishes and transfected after changing the medium the next day using the calcium phosphate method. 16 h later the cells were supplied with fresh medium. The medium was aspirated after 16 h, and the cells were supplied with hypoxic medium and incubated in the hypoxic chamber. Control cells were transfected in parallel and cultivated under normal oxygen pressure. After 6 h cells were harvested and luciferase assays were performed as described (19). Luciferase-generated light units were normalized to protein contents and are shown as relative luciferase activation versus untreated cells. The means of four experiments performed in duplicate are shown. PMA stimulation was used as a control in the experiments and gave approximately double the stimulation of the hypoxia treatment (not shown).

Transcriptional Induction of the c-fos Gene by Hypoxia-To
test c-fos gene activation under hypoxia, HeLa cells were subjected to hypoxia. A rapid and transient increase of c-fos transcripts was observed (Fig. 1A). c-fos mRNA became detectable as early as 15 min following hypoxia (Fig. 1A, lane 2). Maximal induction was observed after 30 min of stimulation (lane 3). No c-fos mRNA was detected after 90 min of stimulation (lane 5). Similar kinetics have been observed for the induction of the c-fos gene by treatment with mitogens (15,16).
Transcriptional Induction by Hypoxia Requires the Intact c-fos SRE-We therefore investigated whether the c-fos SRE is responsible for the induction under hypoxia. Transient transfection experiments using a luciferase reporter plasmid driven by two c-fos SRE binding elements showed an 8.4-fold induction upon hypoxic stimulation (Fig. 1B). Using mutations in the binding sites for the ternary complex factor (⌬TCF) or the serum responsive factor (⌬SRF), we analyzed the requirement of the different binding sites for this induction. These mutations abolish binding of the TCF and SRF, respectively (4). No significant induction could be detected using the ⌬TCF and ⌬SRF constructs (Fig. 1B, columns 2 and 3). These data show that the intact binding site for the ternary complex factor as well as the binding site for the serum response factor are required for the induction of the c-fos gene during hypoxia.
Hypoxia Leads to Phosphorylation of Elk-1-We subsequently studied the effect of reduced oxygen pressure on the phosphorylation status of TCF in HeLa cells. Signal-induced TCF hyperphosphorylation can be assayed in native gels by a phosphatase-sensitive retarded migration of the TCF-DNA complexes (18,22,23). Cells were treated with hypoxia between 15 min and 6 h ( Fig. 2A). Stimulation of cells with the phorbolester PMA was used as a positive control (Fig. 2A, lane  12). Extreme care was taken to avoid reoxygenation of the cells.
Hypoxia rapidly induced a more slowly migrating TCF-DNA complex on the c-fos SRE as seen by EMSA (complex 2, Fig. 2A). This complex was indistinguishable in its mobility from that induced upon stimulation with PMA, which is known to lead to hyperphosphorylated TCFs (16) (Fig. 2A, compare lanes 2-5  with lane 12). The retarded ternary complex was observed within 5 min (see Fig. 4B, lane 2, upper panel), reached a peak at 15 min, and decreased after 40 min (Fig. 2A, compare lanes  2-5). Fig. 2B clearly shows that the observed retarded complex is a ternary complex that depends on the presence of the SRE binding site, the SRF core, and extracts containing TCFs (4). Neither SRF core alone (lane 14) nor extract without the SRF core (lane 17) can form this ternary complex with the c-fos SRE.
Formation of the modified protein-DNA complexes induced under hypoxia in HeLa cells was inhibited by the addition of a specific antibody against Elk-1 to the DNA binding reaction (Fig. 3A, lanes 1-4). We conclude that these complexes contain either bona fide Elk-1 or antigenically closely related TCFs. Consistent with our previous data it has been previously reported that Elk-1 constitutes the majority of HeLa cell TCF activity (24). To show that hyperphosphorylation of the TCFs is the cause of the altered mobility, a phosphatase treatment was performed. Incubation of the cell extracts with potato acid phosphatase prior to EMSAs abolished complex 2 formation (Fig. 3B, lanes 5-10), indicating that the retardation of the band upon hypoxia was caused by hyperphosphorylation of the TCFs TCF/Elk-1. In contrast to the activation of Elk-1 by hypoxic conditions, no activation was observed in gel shifts upon re-exposure of cells to atmospheric oxygen (data not shown).
Hypoxia Activates the MAPK Pathway-We next explored if the phosphorylation of TCF/Elk-1 under hypoxic conditions was due to activation of MAPK. Preincubation of cells with PD98059, a specific inhibitor of MEK1, resulted in a complete loss of complex 2 formation (Fig. 4A, compare lanes 2 and 3  with lanes 5 and 6, lower panel). This shows that MEK1 activity is necessary for the phosphorylation of TCF/Elk-1 during contain labeled c-fos SRE probe. Binding reactions contained either in vitro translation SRF core complex (amino acids 132-222) in lanes 14 and 16 or unprimed in vitro translation (lanes 15 and 17). In addition to lanes 16 and 17 cell extract of HeLa cells stimulated for 5 min with hypoxia was included in the binding reaction. Ternary complexes are labeled with 1 (uninduced) and 2 (induced) (see also part A).  1 and 2) was not observed consistently and is due to experimental variation (see also Fig. 4B, lower panel). B, complexes 1 and 2 are ternary complexes consisting of TCF, the SRF core complex, and the c-fos SRE. IVT, in vitro translation reaction. Lanes 13-17 1 and 3) or cell extracts after 5 min of hypoxic treatment (Hyp) were used (lanes 2 and 4). 0.5 g of total cell extract was preincubated with control antibody (lanes 3 and 4) or an ␣-Elk1 antibody (lanes 1 and 2) and analyzed in EMSAs. Only the ternary complexes are shown and labeled with 1 (uninduced) and 2 (induced). B, extracts of cells induced with hypoxia (15 min) or PMA (30 min) were incubated in the absence or the presence of potato acid phosphatase with or without phosphatase inhibitors and then analyzed by electrophoretic mobility shift assay as described in the legend to Fig. 2. hypoxia. In Western blots a transiently hyperphosphorylated form of ERK2 could be detected after 5-15 min, whereas JNK/ SAPK was not altered (data not shown). To investigate whether this hyperphosphorylated ERK2 correlated with activation of MAPK, an in vitro MAP-kinase assay was performed. After immunoprecipitation of active MAPK, recombinant Elk-1 was used as substrate, blotted, and detected by a specific phosphospecific anti-Elk-1 antibody. Phosphorylated Elk-1 can be detected after 5 min of hypoxia (Fig. 4B, lane 2, upper panel), indicating MEK activity. This already decreased after 15 min of hypoxia (Fig. 4B, lane 3, upper panel). The active form of MAPK temporally corresponds to the appearance of the hyperphosphorylated complex in EMSA (Fig. 4B, compare upper and  lower panels). Furthermore, the MEK1 inhibitor PD98059 blocked activation of MAPK (Fig. 4B, lane 5, upper panel) and appearance of the phosphorylated complex 2 (lane 5, lower panel). To test whether the induction of c-fos transcripts was also inhibited by the MEK1 inhibitor PD98059, a Northern analysis was performed. The induction of c-fos transcripts was blocked as well as the appearance of the phosphorylated complex 2 (Fig. 4A, lanes 5 and 6, compare upper and lower panels). These results show that the MAPK pathway is responsible for the activation of TCF/Elk-1 and the induction of c-fos under hypoxic conditions. DISCUSSION The molecular response of tissue to hypoxia is a complex and poorly understood process. Hypoxic cells in tumors show a reduced sensitivity toward radiation and chemotherapy (25)(26)(27)(28)(29). We and others have previously shown that hypoxia activates the transcription factor AP-1 in HeLa cells (13,14). This activation is dependent on novel AP-1 protein synthesis.

FIG. 3. The hypoxia-induced ternary complex is phosphatasesensitive and mainly contains the TCF Elk-1. A, control cell extracts (lanes
However, the regulation of AP-1 transcription upon hypoxic conditions remained to be clarified. Different modes of redox regulation have been described for a variety of mammalian transcription factors including steroid receptors (30), c-Myb protein (31), and the transcriptional activators NF-B (10, 11) and AP-1 (32,33). At the post-translational level, Ref-1, a redox-factor and a DNA repair enzyme (33,34), facilitates Fos-Jun DNA interaction when cells are treated with thioredoxin. Ref-1 is activated under hypoxic conditions as well (13). However, the pathways leading to the transcriptional activation of the AP-1 subunits under hypoxic conditions are still not understood. The recently described protein kinases N-terminal c-Jun kinase and stress-activated protein kinase, a subfamily of the ERKs that can phosphorylate c-Jun, are not activated by hypoxia (35). In this study we have analyzed the regulation of c-fos gene upon hypoxia.
Here we show that the ternary complex factor TCF/Elk-1 is activated via the MAPK pathway under hypoxia. In addition to our previous results, where we have shown that treatment of HeLa cells with the oxidant H 2 O 2 or with different antioxidants activated Elk-1 (19), these data underline the relevance of Elk-1 as a physiological sensor for the intracellular redox status. With an in vitro kinase assay we could show that MAPK is activated under hypoxia and is able to phosphorylate Elk-1. Preincubation of the cells with PD98059 as specific inhibitor of MEK1 (36) blocked the activation of MAPK in the in vitro kinase assay, the induction of c-fos transcripts in Northern analysis, and the retardation of the Elk-1 complex in EMSA. Therefore, hypoxia activates Elk-1 by the MAPK signal-transduction pathway. The same pathway is used when the c-fos gene is activated by antioxidants (19). This study indicates the molecular mechanism by which the c-fos gene is transcriptionally up-regulated upon hypoxia.
Why does reoxygenation not activate Elk-1? In our previous study we observed a phosphorylation of Elk-1 after treatment of HeLa cells with H 2 O 2 (19). However we could not observe any activation of Elk-1 by reoxygenating the cells following treatment with hypoxia (data not shown). Although reoxygenation increases the intracellular level of ROIs (37), the increase of H 2 O 2 seems to be neither high nor fast enough to activate Elk-1 under reoxygenation. It is possible that other ROIs than H 2 O 2 are predominantly generated under reoxygenation. Superoxide anions, which are generated under reoxygenation as well (37), are not able to activate Elk-1 (19). However, it cannot be excluded that other forms of oxidative stress occurring under physiological conditions can activate Elk-1. The activation of c-fos transcription in response to both oxidant and antioxidant conditions leads to the formation of active AP-1. The maintenance of intracellular redox homeostasis may be an important biological function of AP-1 conserved from yeast to mammals. Many genes induced by redox stress contain AP-1 binding sites that confer this responsiveness (38 -42).
Another transcription factor that was identified and cloned as a hypoxia inducible factor is HIF-1 (for review see Ref. 43). HIF itself is redox-regulated (44). There is no evidence so far that the MAPK pathway or Elk-1 play a role in the induction of the HIF-1 gene. In the enhancer of the glucose transporter-1, a serum reponse element seems to cooperate with HIF-1 binding sequences (45). It is tempting to speculate that the serum response factor and adjacent ternary complex factors such as Elk-1 can cooperate with HIF-1 to activate hypoxia-inducible genes. Protein-DNA interactions of AP-1 and HIF-1 have been described for the enhancer of the tyrosine hydroxylase gene that is dependent on functional AP-1 binding sites for activation under hypoxia (46).
The data presented in this paper provide an explanation for the appearance of c-Fos proteins in AP-1 under hypoxia (14). Firstly, c-fos gene expression is turned on by hypoxia-activated TCF/Elk-1 via the MAPK pathway. Then the newly synthesized c-Fos in the AP-1 complex can lead to activation of other genes. What are the target genes of c-Fos activated by hypoxia in tumor cells? It is well established that c-Fos controls the expression of a variety of genes that are up-regulated during malignant progression, including those encoding the tumor metalloproteases stromelysin and type I collagenase (47)(48)(49). Metalloproteases are a group of secreted enzymes that can degrade the extracellular matrix, thereby facilitating tumor growth, invasion, and metastasis (reviewed by Ref. 50). Increased metalloprotease expression has been associated with malignant progression in many models of in vivo cancer development (reviewed by Ref. 51). A variety of human and rodent tumors are known to overexpress c-fos (e.g. Refs. 52 and 53). C-Fos overexpression in transgenic mice induces bone tumors (54). In human breast cancer, c-fos overexpression is associated with malignant progression (55,56). Hypoxia on the other hand is a known inducer of metastasis (57). Therefore the activation pathway shown in this study provides a mechanism how c-fos activated by hypoxia leads to tumor growth and metastasis.
In conclusion our results demonstrate how hypoxia as a physiological stimulus is linked to the activation of c-fos. The data presented here provide evidence that a nuclear transcription factor, TCF/Elk-1, is phosphorylated and thereby activated in response to reduced oxygen pressure. TCF/Elk-1 is in turn activated by the MAPK ERK2. Future studies will need to address the pathway(s) by which hypoxia activates MEK1 and MAPK such as ERK2.