If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
* This study was supported in part by NHLBI Grants HL32646 and HL35440 from the National Institutes of Health.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Supported by an American Lung Association Research training fellowship.
During hypoxia, hypoxia-inducible factor-1α (HIF-1α) is required for induction of a variety of genes including erythropoietin and vascular endothelial growth factor. Hypoxia increases mitochondrial reactive oxygen species (ROS) generation at Complex III, which causes accumulation of HIF-1α protein responsible for initiating expression of a luciferase reporter construct under the control of a hypoxic response element. This response is lost in cells depleted of mitochondrial DNA (ρ0 cells). Overexpression of catalase abolishes hypoxic response element-luciferase expression during hypoxia. Exogenous H2O2 stabilizes HIF-1α protein during normoxia and activates luciferase expression in wild-type and ρ0 cells. Isolated mitochondria increase ROS generation during hypoxia, as does the bacterium Paracoccus denitrificans. These findings reveal that mitochondria-derived ROS are both required and sufficient to initiate HIF-1α stabilization during hypoxia.
reactive oxygen species
hypoxic response element
vascular endothelial growth factor
DCFH-DA, 2′,7′-dichlorofluorescein diacetate
Hypoxia initiates transcription of a number of gene products that help to sustain the supply of O2 to tissues and to enhance cell survival during severe O2 deprivation. Gene products that augment O2 supply at the tissue level include erythropoietin (Epo)1 which increases the proliferation of erythrocytes, tyrosine hydroxylase which is necessary for the synthesis of the neurotransmitter dopamine in the carotid bodies, and the angiogenic factor VEGF which stimulates growth of new capillaries (
). The significance of HIF-1 in transcriptional regulation was recently demonstrated by the marked decrease in mRNA expression of VEGF and glycolytic enzymes seen during hypoxia in HIF-1α- or ARNT-deficient murine embryonic stem cells (
The mechanism by which HIF-1 activation is initiated during hypoxia remains unclear. Both HIF-1α and ARNT mRNAs are constitutively expressed, indicating that functional activity of the HIF-1α·ARNT complex is regulated by post-transcriptional events. ARNT levels are not significantly affected by [O2], whereas HIF-1α protein is rapidly degraded under normoxic conditions by the ubiquitin-proteasome system (
). Hypoxia enhances HIF-1α protein levels by inhibiting its degradation, thereby allowing it to accumulate, to dimerize with ARNT, and to bind to the hypoxia-responsive element (HRE) in the promoter or enhancer regions of various genes. Thus, the functional HIF-1α·ARNT complex is primarily regulated by the abundance of the HIF-1α subunit.
Although much has been learned about the role of HIF-1 in controlling the expression of hypoxia-responsive genes, the underlying mechanism by which cells detect the decrease in [O2] and initiate the stabilization of HIF-1α is not known. Presently, four diverse O2-sensing mechanisms have been proposed to mediate the transcriptional response to hypoxia (
). Two of these models postulate the involvement of an iron-containing unit in the form of either a heme group or an iron/sulfur cluster, which undergoes a change in activity during hypoxia that triggers the transcriptional response. These models are supported by the observation that cobaltous ions, or alternatively the iron chelator desferrioxamine (DFO), stabilize HIF-1α under normoxic conditions (
). However, no specific proteins with this role have been identified in mammalian systems. Two other models involve the generation of reactive oxygen species (ROS) by a flavoprotein-containing NAD(P)H oxidase or by mitochondria. The NAD(P)H oxidase theory postulates that a decrease in ROS production triggers the transcriptional response to hypoxia (
), whereas the model would predict that DPI should activate the response during normoxia. We previously proposed that hypoxia partially inhibits mitochondrial electron transport, producing redox changes in the electron carriers that increase the generation of ROS. These oxidants then enter the cytosol and function as second messengers in the signaling pathway leading to stabilization of HIF-1α (
). In support of this model, hypoxia failed to increase ROS production or the expression of EPO, VEGF, and glycolytic enzymes in ρ0 cells, which lack mitochondrial DNA and electron transport activity. Also, the response to hypoxia was abolished by the DPI, which abrogates mitochondrial ROS generation by inhibiting electron transport at the flavin site in mitochondrial Complex I (
Our previous study tested whether ROS are required for the DNA binding of HIF-1 and the subsequent mRNA expression of Epo, VEGF, and glycolytic enzymes during hypoxia. However, protein levels of HIF-1α were not measured, so that study did not reveal whether mitochondrial ROS were required to trigger stabilization of HIF-1α during hypoxia. Moreover, although ROS were found to be necessary for the transcriptional response to hypoxia, it was not clear whether ROS by themselves were sufficient to initiate HIF-1α stabilization. Accordingly, the present study tested the following: (a) whether mitochondrial ROS are required for HIF-1α stabilization during hypoxia, cobalt treatment, or DFO; (b) whether cytosolic ROS are sufficient to trigger HIF-1α stabilization; and (c) whether mitochondrial cytochrome coxidase serves as the O2 sensor responsible for the redox changes underlying the increase in ROS generation during hypoxia.
ROS have been proposed to participate in the signal transduction process mediating the stabilization of the transcription factor HIF-1α during hypoxia. However, controversy exists regarding whether ROS levels increase or decrease under hypoxia. The present study extends our previous work by demonstrating that HIF-1α stabilization by hypoxia or CoCl2 requires an increase in ROS. Hypoxia and CoCl2 both elicited an increase in DCF fluorescence, revealing that oxidation of the probe increases prior to the stabilization of HIF-1α. The antioxidant PDTC abolished the stabilization of HIF-1α and oxidation of DCFH in response to hypoxia or CoCl2, further suggesting the involvement of ROS. Human 293 cells overexpressing catalase showed an attenuated HRE-luciferase expression response to hypoxia or CoCl2, demonstrating the requirement for H2O2 in a second cell line. Finally, administration of H2O2 caused HIF-1α stabilization and expression of HRE-luciferase under normoxic conditions in both cell types. Collectively, these findings support the conclusion that ROS are both required and sufficient to activate the signaling system resulting in the stabilization of HIF-1α. Our findings are consistent with recent reports demonstrating hypoxia-induced increases in ROS generation in mesenteric vesselsin vivo (
Previous studies using exogenous administration of H2O2 (0.1–1 mm) did not detect stabilization of HIF-1 under normoxia. In fact, H2O2 bolus treatment (1 mm) abolished HIF-1-mediated gene transcription during subsequent hypoxia (1% O2 for 8 h) (
). Our H2O2 measurements revealed that a bolus of peroxide is rapidly degraded by cells even in serum-free medium, suggesting that administration of repeated boluses might be required to sustain active signaling. Indeed, Hep3B cells responded to H2O2 boluses (every 15 min for 2 h) by demonstrating accumulation of HIF-1α, whereas 293 cells given H2O2 (40 μm every 15 min for 10 h) induced the expression of HRE-luciferase. These observations indicate that low levels of an oxidizing stimulus are sufficient to stabilize HIF-1α.
The iron chelator DFO also induces HIF-1α stabilization and HIF-1-dependent gene activation, but it appears to act at a more downstream site in the signaling cascade. Stabilization of HIF-1α by DFO did not appear to involve ROS, as DFO neither increased nor decreased DCF fluorescence during normoxia. The response to DFO was not abolished by either PDTC or catalase, both of which enhanced the scavenging of H2O2. The flavin inhibitor DPI and the Complex I inhibitor rotenone also failed to abolish HIF-1α stabilization during DFO, and ρ0 Hep3B cells were still able to respond to DFO. However, wortmannin and calyculin A both abolished the HIF-1α stabilization during DFO. These findings indicate that DFO does not require mitochondrial electron transport or ROS and appears to be acting at a site downstream from the oxidant-mediated step but upstream from the ubiquitin-proteasome degradation step.
What is the source of ROS that trigger stabilization of HIF-1α during hypoxia? Mitochondrial Complex III can generate superoxide anions during normoxia (
) (Fig. 7). Our data reveal that hypoxia increases the generation of ROS at Complex III to an extent sufficient to stabilize HIF-1α protein. In support of this, respiration-incompetent ρ0 cells failed to stabilize HIF-1α, were unable to express HRE-luciferase, and failed to increase oxidation of DCFH during hypoxia. However, ρ0cells were able to respond to CoCl2 and to DFO, indicating that they retained the ability to respond. Moreover, ρ0293 cells responded to exogenous H2O2 by inducing expression of HRE-luciferase. Collectively, these data indicate that an active electron transport chain is required for hypoxic stabilization of HIF-1α and that ROS generated during hypoxia are sufficient to effect this stabilization. The Complex I inhibitor rotenone diminished ROS generation by prohibiting the supply of electrons into Complex III. This compound abolished the increase in DCFH oxidation during hypoxia and prevented the increase in HIF-1α. Rotenone and antimycin A both inhibit electron transport and oxidative phosphorylation, but only the former compound abolished the increases in DCFH oxidation and HIF-1α protein during hypoxia. By inhibiting electron transport at the downstream end of Complex III, antimycin A augments superoxide generation, yet we were consistently unable to detect HIF-1α stabilization using that compound during normoxia. Because antimycin A caused only a modest increase in DCFH oxidation compared with hypoxia (
), we hypothesized that the failure of antimycin A to stabilize HIF-1α could be attributed to the smaller ROS signal. The amplitude of ROS signaling in a cell reflects a balance between the rate of oxidant generation and the efficacy of antioxidant systems. In Hep3B cells depleted of glutathione stores, the ability to metabolize H2O2 would be compromised, which should amplify the effects of a smaller oxidant source. Indeed, HIF-1α protein was stabilized in response to antimycin A during normoxia, whereas rotenone, DPI, and PDTC blocked that response in glutathione-depleted Hep3B cells. These observations provide further support for the conclusion that mitochondrial ROS generation at Complex III is required and sufficient for the stabilization of HIF-1α during hypoxia.
How do ROS generated in mitochondria escape to the cytosol? Superoxide generated in the mitochondrial matrix can be dismutated to H2O2 by Mn-superoxide dismutase. The H2O2 so produced could diffuse into the cytosol or could be degraded by the mitochondrial antioxidant system. Superoxide generated at Complex III could alternatively enter the cytosol via anion channels in the mitochondrial membranes (
). Our data show that anion channel inhibition by DIDS abolished the DCFH oxidation observed during hypoxia or with antimycin A during normoxia. By suppressing the egress of mitochondrial superoxide, anion channel inhibition should attenuate the cytosolic oxidant signals that are required for HIF-1α stabilization in hypoxia but have no effect on the response to CoCl2 or to DFO because these responses do not require mitochondria. Indeed, the data show that HIF-1α accumulation was abolished by DIDS during hypoxia in Hep3B cells and during antimycin A treatment in glutathione-depleted cells studied during normoxia. However, DIDS did not block the response to CoCl2 or to DFO. These observations further support the conclusion that an oxidizing stimulus localized to the cytosol is required for the stabilization of HIF-1α.
Stabilization of HIF-1α protein by CoCl2 or DFO does not require a functional mitochondrial electron transport system, as both were able to stabilize HIF-1α protein and induce expression of HRE-luciferase in ρ0 cells. However, CoCl2, unlike DFO, increased the oxidation of DCFH dye, and the CoCl2-induced expression HRE-luciferase was attenuated in ρ0 cells overexpressing catalase. Stabilization of HIF-1α by CoCl2 was not inhibited by DPI, suggesting that cobalt stimulates ROS generation by a non-enzymatic, non-mitochondrial mechanism. The observation that DFO does not require ROS or mitochondria led us to examine whether phosphorylation might serve as a common signaling event linking hypoxia, CoCl2, and DFO-induced stabilization of HIF-1α. Insulin and insulin-like growth factors have been shown to stabilize HIF-1α protein and HIF-1-dependent gene expression during normoxia (
). Our data indicates that both PI-3 kinases and phosphatases are required for the HIF-1α stabilization effected by hypoxia, CoCl2, and by DFO. We conclude that PI-3 kinases and phosphatases must be acting upstream of the ubiquitin/proteasome degradation step in the O2-sensing pathway, because H2O2-induced HIF-1α stabilization was abolished by wortmannin as well as by calyculin A. Protease and PI-3 kinase inhibitors are toxic to cells at higher concentrations, and toxicity could conceivably inhibit HIF-1α stabilization nonspecifically. However, accumulation of HIF-1α still occurred during normoxia in response to the proteasome inhibitorN-carbobenzoxyl-l-leucinyl-l-leucinyl-l-norvalinal in the presence of wortmannin or calyculin A, indicating that the cells retained the ability to respond in the presence of these inhibitors. To confirm further that wortmannin was not inhibiting HIF-1α stabilization by a nonspecific mechanism, the PI-3 kinase inhibitor LY 294002 was also tested. Like wortmannin, this compound inhibited the response to hypoxia, CoCl2, and DFO when used at a concentration of 20 μm, thereby supporting the conclusion that PI-3 kinase participates in the signaling pathway required for HIF-1α stabilization.
Does cytochrome oxidase act as the O2 sensor during hypoxia? We previously demonstrated that hypoxia causes a decrease in cytochrome c oxidase Vmax, which led us to hypothesize this as a mechanism affecting the redox state of cytochromes at more proximal locations. We then postulated that such redox changes might explain the stimulation of superoxide production at low [O2]. In support of this, we noted a similar temporal relationship between the changes in cytochrome c oxidaseVmax and the ROS generation in cardiomyocytes (
), yet these cells displayed HIF-1α protein accumulation within 30 min in the present study. Based on the discrepancy between the duration of hypoxia required to effect a change in Vmax and the duration required to stabilize HIF-1α, it is unlikely that cytochrome c oxidase could serve as the primary O2 sensor in hypoxia. Rather, it appears that Complex III must possess inherent sensitivity to [O2], allowing it to adjust its generation of ROS inversely with the O2 tension. Indeed, our previous observation that ROS generation increased at lower [O2] in Hep3B cells treated with antimycin A (
) is consistent with this conclusion, because electron transport inhibition by antimycin A would fully reduce Complex III, rendering its redox state insensitive to the redox state of cytochrome c oxidase. In the present study, increased oxidant generation by Complex III under hypoxia was confirmed in isolated mitochondria. When electrons were supplied exclusively to Complex III, ROS generation increased during hypoxia confirming the results in intact cells. By contrast, ROS generation tended to decrease during hypoxia when electrons were supplied to only to Complex I, demonstrating the specificity of this response for Complex III. We therefore conclude that Complex III functions as the O2sensor during hypoxia by regulating ROS generation inversely with [O2].
In a broader context, it is interesting to note that mitochondria are ubiquitous in eukaryotic cells and could provide a mechanism of O2 sensing in other physiological responses. Our observation of an increased DCFH oxidation by P. denitrificans, a bacterium that demonstrates important similarities to eukaryotic mitochondria, supports this notion by showing that [O2]-dependent ROS generation by Complex III is not unique to higher organisms. By contrast, E. coli oxidation of DCFH decreased at low O2 levels, indicating that not all organisms generate an increased oxidant signal under hypoxia. The failure to increase ROS generation by E. coli may relate to its lack of a bc1complex, allowing electron transfer to occur without the formation of a ubisemiquinone intermediate (
). This further supports the conclusion that univalent electron transfer from a semiquinone to O2acts as the source of superoxide in cells that contain a functioningbc1 complex. Future studies will need to address the possibility that mitochondria function as a O2sensors and ROS as signaling molecules in other systems.
We thank Dr. Jeffrey M. Leiden for generously providing the HRE-luciferase constructs used in this study.