A novel hypoxia-inducible factor-independent hypoxic response regulating mammalian target of rapamycin and its targets.

Hypoxia triggers a reversible inhibition of protein synthesis thought to be important for energy conservation in O2-deficient environments. The mammalian target of rapamycin (mTOR) pathway integrates multiple environmental cues to regulate translation in response to nutrient availability and stress, suggesting it as a candidate for O2 regulation. We show here that hypoxia rapidly and reversibly triggers hypophosphorylation of mTOR and its effectors 4E-BP1, p70S6K, rpS6, and eukaryotic initiation factor 4G. Hypoxic regulation of these translational control proteins is dominant to activation via multiple distinct signaling pathways such as insulin, amino acids, phorbol esters, and serum and is independent of Akt/protein kinase B and AMP-activated protein kinase phosphorylation, ATP levels, ATP:ADP ratios, and hypoxia-inducible factor-1 (HIF-1). Finally, hypoxia appears to repress phosphorylation of translational control proteins in a manner analogous to rapamycin and independent of phosphatase 2A (PP2A) activity. These data demonstrate a new mode of regulation of the mTOR pathway and position this pathway as a powerful point of control by O2 of cellular metabolism and energetics.

Hypoxia triggers a reversible inhibition of protein synthesis thought to be important for energy conservation in O 2 -deficient environments. The mammalian target of rapamycin (mTOR) pathway integrates multiple environmental cues to regulate translation in response to nutrient availability and stress, suggesting it as a candidate for O 2 regulation. We show here that hypoxia rapidly and reversibly triggers hypophosphorylation of mTOR and its effectors 4E-BP1, p70 S6K , rpS6, and eukaryotic initiation factor 4G. Hypoxic regulation of these translational control proteins is dominant to activation via multiple distinct signaling pathways such as insulin, amino acids, phorbol esters, and serum and is independent of Akt/protein kinase B and AMP-activated protein kinase phosphorylation, ATP levels, ATP:ADP ratios, and hypoxia-inducible factor-1 (HIF-1). Finally, hypoxia appears to repress phosphorylation of translational control proteins in a manner analogous to rapamycin and independent of phosphatase 2A (PP2A) activity. These data demonstrate a new mode of regulation of the mTOR pathway and position this pathway as a powerful point of control by O 2 of cellular metabolism and energetics.
In mammalian systems, low O 2 (hypoxia) has two types of effects on cellular metabolism and gene expression. Rapid and reversible effects on cell signaling, contractility, ion flux, and redox state (for review, see Ref. 1) are critical for neural, cardiovascular, and pulmonary function and serve to balance energy supply and demand in the face of reduced capacity for oxidative metabolism. Slower transcriptional responses are largely dependent on the hypoxia-inducible factors (HIFs), 1 a family of O 2 -sensitive transcription factors (for review, see Ref. 2) whose target genes enable long term cellular survival and adaptation to hypoxic conditions. Among the immediate effects of hypoxia is a rapid inhibition of mRNA translation. Translation, especially at the initiation step, is highly regulated and exquisitely sensitive to cellular stress. Osmotic (3)(4)(5) and oxidative stress (5), UV irradiation (6 -8), heat shock (9 -12), elevated AMP levels and AMP-activated protein kinase (AMPK) activity (13)(14)(15), viral infection, heme deficiency, and ER stress (for review, see Ref. 16) are known to regulate translation initiation factors, as are deficiencies in glucose, amino acids (for review, see Ref. 17), and ATP (18). The two central mechanisms for regulating translation initiation are the assembly of (i) active eukaryotic initiation factor (eIF) 4F and (ii) eIF2⅐GTP⅐Met⅐tRNA i ternary complexes (for review, see Ref. 19), whose assembly and activity can be rapidly and reversibly modulated by changes in the phosphorylation states of subunits or interacting proteins. Functional eIF4F complex binds to the m 7 GTP cap of mRNA, unwinds mRNA secondary structure, and recruits the 40 S ribosome and other translation factors to mRNA. The eIF2⅐GTP⅐Met⅐tRNA i complex is required for initiation codon recognition and 60 S ribosome recruitment and is locked in its inactive GDP-bound form by stress-induced phosphorylation of the eIF2␣ subunit by its kinases (for review, see Ref. 19). The mammalian target of rapamycin (mTOR), a highly conserved serine/threonine kinase, is central to the control of translation in response to stress and nutrient deprivation. In growth-promoting conditions, mTOR sustains translation by phosphorylating the eIF4E-binding proteins (4E-BPs) and ribosomal protein S6 kinases (S6Ks) and activating the phosphorylation of eIF4G, although the mechanism of the last is not well understood (for review, see Ref. 20). Hierarchical phosphorylation of 4E-BPs at several conserved serine and threonine residues renders them unable to inhibit eIF4F activity, thereby activating translation initiation. Phosphorylation of p70 S6K at Thr 389 of its linker domain is closely correlated with in vivo activity, as is phosphorylation at Thr 421 and Ser 424 of the pseudosubstrate domain, which is thought to relieve autoinhibition by this domain (21). Activated p70 S6K phosphorylates both eukaryotic elongation factor 2 kinase, increasing the activity of elongation factor 2 (22) and the ribosomal protein S6. Although the phosphorylation of eIF4G and rpS6 are strongly correlated with activated translation, the precise role of phosphorylation of these two regulatory proteins is unclear. Both the phosphatidylinositol 3-kinase (PI3K) and mTOR pathways appear to be required for full phosphorylation of mTOR substrates (20), and there is clearly extensive, although incompletely understood, cross-talk between PI3K/Akt signaling and mTOR. Recent reports propose that PI3K/Akt signaling upregulates mTOR activity by relieving repression of mTOR by the TSC2 complex (23).
Many animals possess the ability to alter their metabolism in responses to changes in O 2 , a phenomenon known as oxygen conformance (24,25) in which cells down-regulate nonessential anabolic processes in advance of energetic crisis. In particular, reversible inhibition of translation has been shown in many organisms to play a role in conformance. In mammals, models of ischemic disease demonstrate both eIF4F (26 -30) and eIF2 (26, 30 -36) regulation in ischemia and/or reperfusion, and it is thought that translational arrest is an important component of the cellular damage caused by such diseases. Severe hypoxia (0.02%) results in the phosphorylation and inhibition of eIF2␣ by the ER-resident kinase PERK (37), implicating ER stress in the translational response to pathophysiological hypoxia. Mammalian tissues can encounter O 2 concentrations anywhere from Ϸ0 to 10%, and little is known about the effects of physiological hypoxia on translation initiation and whether it differs from more severe ischemic conditions. Reversible inhibition of translation occurs between 0.5 and 1% O 2 (38,39), and a study comparing 95% O 2 with 5% O 2 showed differences in the phosphorylation state of 4E-BP1 and its association with eIF4E (40).
In this report we demonstrate for the first time coordinated regulation of the mTOR pathway by O 2 . Hypoxia, defined here as 1.5% O 2 , caused rapid and reversible hypophosphorylation of mTOR and its effectors 4E-BP1, p70 S6K , rpS6, and eIF4G. Hypoxic regulation of these translational control proteins was independent of Akt phosphorylation and was dominant to leucine-, phorbol ester-, and serum-induced phosphorylation of mTOR targets. This suggests that hypoxia acts at or downstream of mTOR, the putative point of convergence of multiple stress-and nutrient-sensing signaling pathways. Hypoxia effects were also independent of HIF-1 and did not correlate with changes in cellular ATP levels or AMPK signaling, demonstrating that O 2 regulation of the mTOR pathway is not secondary to hypoxic effects on HIF-1 or cellular energetics. Hypoxic regulation of mTOR targets was also independent of PP2A. Together, these data show that hypoxia triggers molecular events that result in the coordinate inhibition of central effectors of the cellular translational apparatus.
Western Blots-Extracts were electrophoresed on SDS-PAGE gels, transferred, and immunoblotted according to standard protocols or manufacturers' instructions using 5% nonfat dry milk (Carnation) in Tris-buffered saline with Tween. Blots were stained with Ponceau S to ensure equal loading. Anti-human HIF-1␣ was purchased from BD Transduction Labs (#610958), anti-eIF4G (sc-11373) was from Santa Cruz Biotechnology, and all others including mouse and rabbit secondary antibodies were from Cell Signaling Technologies. ECL reagents were purchased from Amersham Biosciences. Blots were stripped in 61.50 mM Tris, pH 6.8, 2% SDS, and 100 mM ␤-mercaptoethanol at 55°C for 1 h before being blocked and reprobed according to standard protocols.
ATP and ATP:ADP Measurements-293 cells were plated at 3 ϫ 10 5 /well of a 6-well plate (total ATP) or 1 ϫ 10 4 /well of a 96-well plate (ATP:ADP) in D/10% and allowed to adhere overnight. Cells were then exposed to 30 min of hypoxia, 30 min of hypoxia followed by 30 min of reoxygenation, or medium containing 100 mM 2-deoxy-D-glucose (Sigma) for 30 min. Overall ATP levels were determined using the ATP bioluminescence assay kit CLS II (Roche Applied Science) after lysis with buffer from ATP bioluminescence assay kit HS II (Roche Applied Science). Values are the means Ϯ S.E. of three independent experiments of four wells of cells for each condition tested. ATP:ADP ratios were determined using the ApoGlow assay (BioWhittaker) according to the manufacturer's instructions. Values are means Ϯ S.E. of three independent experiments of five wells of cells for each condition.

Hypoxia Causes Hypophosphorylation of mTOR and Its
Targets-Serum-starved HEK293 cells were incubated under hypoxic (1.5% O 2 ) or normoxic (20% O 2 ) conditions for 30 min before a 45-min insulin stimulation during which the indicated O 2 concentrations were maintained. Hypoxic treatment resulted in dephosphorylation of mTOR at Ser 2481 (Fig. 1a), which was rapidly reversed after 1 h of reoxygenation. This hypoxia-induced hypophosphorylation was not observed at Ser 2448 of mTOR. Ser 2481 has been shown to be a site of mTOR autophosphorylation (43), whereas Ser 2448 is phosphorylated upon activation of the PI3K/Akt pathway, suggesting that hypoxia reversibly inhibits mTOR activity but not PI3K or Akt activity.
Insulin stimulated the phosphorylation of mTOR effectors 4E-BP1, p70 S6K , and eIF4G as well as the mTOR-independent phosphorylation of Akt/protein kinase B (second lanes of Figs.  1, b-e). As a positive indicator of hypoxic responses in this and other experiments, Western blots for HIF-1␣ were performed (Fig. 1e). After 30 min of hypoxia, insulin-stimulated phosphorylation of 4E-BP1 at Thr 37 and Ser 65 (third lane, Fig. 1b) was suppressed to below base-line levels. Overall 4E-BP1 phosphorylation can also be visualized by changes in mobility; the shift to faster mobility (i.e. hypophosphorylated) bands upon hypoxic treatment is evident in Western blots for total 4E-BP1. Hypoxic hypophosphorylation of 4E-BP1 was reversed after 1 h of reoxygenation at 20% O 2 (fourth lane).
Similar to 4E-BP1, insulin-induced phosphorylation of the p70 and p85 isoforms of the p70 S6K gene at Thr 389 , Thr 421 , and/or Ser 424 (Fig. 1c) was suppressed by hypoxic treatment and restored upon reoxygenation. Although the phospho-specific antibodies are reactive against both isoforms, the antibody against total p70 S6K only appears to detect the p70 isoform. This may reflect the relative abundance and phosphorylation states of these two isoforms or may be a result of differential sensitivity of the antibodies. Phosphorylation of eIF4G at Ser 1108 (Fig. 1d) is also induced by insulin and is thought to be mediated indirectly by mTOR activity (20). Phosphorylation at this site, like 4E-BP1 phosphorylation, was repressed below base-line levels by hypoxic treatment and return to 20% O 2 relieved this repression. Akt phosphorylation at Ser 473 , which is induced by insulin in a PI3K-dependent and mTOR-independent fashion, is unaffected by hypoxic treatment (Fig. 1e), demonstrating that hypophosphorylation of mTOR and its effectors is not a result of either repression of PI3K/Akt signaling or global repression of protein phosphorylation. Although recent results demonstrate that more severe or longer hypoxic treatments can result in PERK-dependent phosphorylation of eIF2␣ (37), this was not observed under the less severe, short term conditions used here (Fig. 1e).
Hypoxia Enhances 4E-BP1 Association with eIF4E and Suppresses Phosphorylation of rpS6 -The phosphorylation data from Fig. 1 predict changes in the activities of downstream effectors of mTOR signaling. 4E-BP1 hypophosphorylation results in an increased association with its target, eIF4E, and we used m 7 GTP-conjugated Sepharose beads to affinity-purify eIF4E-containing, m 7 GTP binding complexes from 293 cells (Fig. 2a). Insulin stimulation, as expected, strongly inhibited the association of 4E-BP1 with eIF4E compared with serumstarved cells, indicative of a translationally active state. By contrast, hypoxia prevented the dissociation of 4E-BP1 and eIF4E to the same degree as rapamycin, whereas reoxygenation restored their insulin-stimulated dissociation. The formation of active eIF4F translation initiation complexes is therefore tightly regulated by O 2 .
rpS6 is a p70 S6K target that is phosphorylated at serines 235, 236, 240, and 244 in response to insulin stimulation (Fig. 2b). This stimulation was completely suppressed by hypoxia and was rapidly restored upon return of hypoxic cells to normal O 2 concentrations. rpS6 phosphorylation closely reflects the phosphorylation state of p70 S6K and, taken together with the 4E-BP1/eIF4E association data, demonstrates that hypoxia suppresses the activity of 4E-BP1 and the phosphorylation of a p70 S6K target.
Hypoxia Suppresses Leucine-, Phorbol Ester-, and Serumstimulated Phosphorylation of mTOR Targets-Amino acids, most potently leucine (for review, see Ref. 44), strongly activate phosphorylation of mTOR substrates, as do phorbol esters like PMA, the latter through extracellular signal-regulated kinasedependent and PI3K-independent mechanisms (45,46). To ascertain whether hypoxia regulates translational control by multiple activating stimuli, we tested the ability of hypoxia to suppress translational activators other than insulin. Leucine (Fig. 3a) or PMA (Fig. 3b) stimulation of serum-starved cells induced the phosphorylation of 4E-BP1, p70 S6K , and rpS6. In both cases, this induction was suppressed by hypoxia. In exponentially growing cells in standard (10% FBS) conditions, 4E-BP1, p70 S6K , and rpS6 were all hypophosphorylated after 15 min of hypoxia (Fig. 3c). After 1 h of hypoxia, phosphorylation of all 3 proteins returned to basal levels after as little as 15 min of reoxygenation. Insulin, leucine, and PMA activate distinct upstream signaling cascades, and it seems unlikely that hypoxia specifically inhibits proximal signaling events in each of the pathways tested. Together with the mTOR and Akt phosphorylation data from Fig. 1, these data suggest that hypoxic regulation of mTOR targets is a generalized mechanism capable of opposing a variety of stimuli and that hypoxia exerts its control at potential sites of convergence for the signaling pathways tested.

HIF-1 Is Neither Necessary nor Sufficient for Hypoxic
Regulation of the mTOR Pathway-HIF-1 is the critical transcriptional regulator of mammalian O 2 and glucose homeostasis. We therefore determined whether HIF-1 activity was necessary for hypoxia effects on translational control proteins using BpRc1 cells, a murine hepatoma line that lacks ARNT/HIF-1␤, and thus, functional HIF-1 complex (41,42). In BpRc1 cells expressing either vector alone or ARNT/HIF-1␤, hypoxia resulted in hypophosphorylation of both rpS6 and 4E-BP1, indicating that HIF-1 is not necessary for hypoxia effects on mTOR targets (Fig. 4a).
HIF-1 is primarily regulated through changes in the stability of its regulated ␣ subunit, which is constitutively transcribed, translated, ubiquitinated, and degraded under normoxic conditions. HIF-1␣ is stabilized by hypoxia in a complex, multistep process, the regulation of which is dependent on hydroxylation of multiple proline residues by a family of O 2 -dependent HIFprolyl hydroxylases (2). These enzymes are rapidly regulated by changes in O 2 concentrations, and we investigated whether their activity is involved in hypoxic regulation of translational control proteins. CoCl 2 and desferrioxamine induce HIF-1␣ stabilization and HIF-1 activity independently of O 2 levels by inhibiting HIF-prolyl hydroxylase activity. As Fig. 4b shows, substantial induction of HIF-1␣ by these chemicals failed to replicate the effects of hypoxia on phosphorylation of rpS6 and 4E-BP1, demonstrating that neither HIF-1␣ stabilization nor inhibition of HIF-prolyl hydroxylase activity is sufficient to induce hypophosphorylation of translational control proteins.
Hypoxic Regulation of Translational Control Proteins Does Not Correlate with Changes in Total ATP, ATP:ADP Ratios, or AMPK Phosphorylation-A recent report proposes a role for mTOR as a direct sensor of cellular ATP (18). AMPK, a kinase that inhibits anabolic processes in the cell in response to elevated AMP, can cause hypophosphorylation of mTOR targets and has been proposed to inhibit mTOR signaling (13)(14)(15). We therefore determined whether cellular energy state is directly or indirectly (via AMPK) involved in the hypoxic regulation of translation. Neither ATP (Fig. 5a) nor ATP:ADP ratios (Fig. 5b) showed a significant change after 30 min of hypoxia or after hypoxia/reoxygenation compared with normoxic controls. This was not surprising given the relatively short and mild treatments used here and is consistent with previous reports (47,48). In contrast, treatment with 2-deoxy-D-glucose, a non-hydrolyzable form of glucose that limits cellular glycolysis, resulted in 44 and 59% decreases in total ATP and ATP:ADP ratios, respectively. Consistent with these data, phosphorylation of AMPK at Thr 172 , a marker for its activation, was not induced after hypoxic treatment (Fig. 5c), in contrast to 2-deoxy-D-glucose treatment. These data make two important points as follows. 1) Relatively short exposure to 1.5% O 2 does not result in a decline in the overall energetic health of the cell; 2) the effects of hypoxia on translational control proteins are not correlated with changes in adenine nucleotide metabolism.
Hypoxic Regulation of mTOR Targets Is, Like Rapamycin, Resistant to OA-Protein protein phosphatase 2A has been implicated in the regulation of mTOR substrates and translational control (49 -52), and we used OA, a potent inhibitor of PP2A, to ascertain its role in hypoxia. Both hypoxia and rapamycin resulted in hypophosphorylation of p70 S6K and 4E-BP1 in untreated cells (Fig. 6, first three lanes). Although increasing doses of OA, as predicted, elevated basal levels of phosphorylation, both hypoxia and rapamycin caused decreased phosphorylation of p70 S6K and 4E-BP1 at all OA concentrations tested. This demonstrates that PP2A is not directly required for hypoxic regulation of mTOR targets. Because rapamycin functions by inhibiting some, if not all, aspects of mTOR kinase activity and hypoxia performed similarly to rapamycin in this assay with respect to mTOR target phosphorylation, these data suggest that hypoxia and rapamycin may inhibit translation in a mechanistically similar fashion, i.e. by inhibiting mTOR activity. DISCUSSION An emerging model of translational control posits mTOR as a central integrator of multiple environmental and nutritional cues. In this model, signals from the extracellular milieu like growth factors, along with intracellular metabolites like glucose, ATP, and amino acid levels, all feed into mTOR, which integrates these signals and modulates rates of translation to reflect and respond to the prospects for cellular growth. The ability to now include O 2 as a modulator of mTOR pathway signaling provides new support for the role of mTOR as an integrator of cellular metabolic state and adds to the list of nutrients to which this pathway can respond. Our findings also elucidate a novel HIF- mTOR has been proposed to be a direct sensor of cellular ATP (18) and to be inhibited by the AMP-mediated activation of AMPK (15), suggesting that mTOR can be regulated both directly and indirectly by cellular energy state. Although severe O 2 depletion results in a decrease in cellular ATP and an increase in AMPK activity (14), the less severe conditions used here do neither, suggesting that O 2 levels can regulate the mTOR pathway independently of cellular energy homeostasis. This suggests that hypoxia can initiate cellular energy conservation strategies (such as inhibition of translation), thus decreasing ATP consumption prior to O 2 becoming metabolically limiting. This phenomenon, known as oxygen conformance, is widely observed in anoxia-and hypothermia-tolerant aquatic animals (for review, see Ref. 24) and has also been observed in mammalian cells (47,53). The degree of oxygen conformance in different glioma cell lines has been linked to their ability to survive hypoxic treatment (48) and is thus of importance in thinking about the role of hypoxia in cancer cell survival, aggressiveness, and resistance to therapy (54).
Insight into the mechanisms of hypoxic inhibition of translation have come mostly in studies of oxygen conformant, anoxia-tolerant organisms such as fish (55), turtles (25), and shrimp (56) and in studies of anoxic or ischemic injury in mammalian systems (for review, see Ref. 57). What has been less well studied is the role of more moderate, physiological levels of hypoxia in modulating protein synthesis. We have demonstrated hypoxic regulation of the mTOR pathway that is independent of ATP levels, AMPK and Akt/protein kinase B phosphorylation, HIF-1, and HIF-prolyl hydroxylase activity and is dominant to mTOR pathway activation by insulin, serum, amino acids, and phorbol ester. How hypoxia regulates the mTOR pathway and whether hypoxia directly regulates mTOR activity remain unclear. That three independent effectors of mTOR in addition to a site of autophosphorylation in mTOR itself (43) are hypophosphorylated in response to hypoxia is highly suggestive of modulation of mTOR activity.
Many recent reports have investigated PI3K/Akt/mTOR activation of HIF-1 and, through it, cellular glucose metabolism and angiogenesis in the context of cancer progression (58,59), although the role of these events in the normal regulation of HIF-1 is unclear (60,61). Hypoxia opposes the phosphorylation of mTOR targets in all cell lines we have tested, including human glioblastoma and embryonic kidney cells and murine fibroblast, hepatoma, and pro-B cell lines (data not shown). Although it seems likely that oncogenic regulation of HIF-1 occurs in certain cellular and pathological contexts, our results here suggest that under hypoxia, mTOR is unlikely to activate HIF-1.
HIF-1␣ and its targets are required for cell long term response to hypoxia, necessitating the selective translation of specific mRNAs despite global inhibition of translation. A partial resolution of this paradox has become apparent recently with the description of internal ribosome entry site elements in the 5Ј-untranslated regions of both HIF-1␣ (38,39) and vascular endothelial growth factor, a HIF-1 target gene (62,63). These internal ribosome entry site elements function by allowing cap-independent translation initiation of mRNAs, rendering them resistant to inhibition of eIF4F and, thus, enabling uninhibited translation under hypoxic conditions.
Internal ribosome entry site elements have also been found in a variety of cellular mRNAs with other stress-response functions (64), suggesting that selective translation of mRNAs based on cis-acting untranslated region sequences is a common mode of ensuring efficient translation of stress-response proteins under conditions that otherwise inhibit protein synthesis. Such a mode of regulation has recently been elucidated for the amino acid transporter cat-1 (65,66), the translation of which is maintained during amino acid withdrawal by an internal ribosome entry site element in its 5Ј-untranslated region. The presence of such cis-acting control elements allows for a level of fast, reversible regulation that is independent of de novo transcription and which is able to respond to environmental, tem- FIG. 5. Hypoxic regulation of the mTOR pathway does not correlate with changes in total ATP, ATP:ADP ratios, or AMPK phosphorylation. 293 cells in DMEM, 10% FBS were exposed to normoxia (N), hypoxia (H), or 2-deoxy-Dglucose (2-DOG) for 30 min or 30 min of hypoxia followed by 30 min of reoxygenation (Reox) before measurement of total ATP (a) or ATP:ADP ratios (b) by bioluminescence. ATP and ATP:ADP ratios were collected in independent experiments. Overall ATP values are the means Ϯ S.E. of three independent experiments of four wells of cells for each condition tested. ATP:ADP ratios are means Ϯ S.E. of three independent experiments of five wells of cells for each condition. c, 293 cells were subjected to hypoxia, reoxygenation, or 2-deoxy-D-glucose as above before lysis and immunoblotting for total and phospho-AMPK and total mTOR as a loading control. Results in c were reproduced in at least three independent experiments. poral, or spatial cues resulting in changes in stability, localization, and translation of associated transcripts. These "posttranscriptional operons" (67) define a mechanism of gene regulation that is refractory to genomic or array-based study and should prove increasingly important in our understanding of how stress regulates gene expression.