Oxygen Tension Regulates the Stability of Insulin Receptor Substrate-1 (IRS-1) through Caspase-mediated Cleavage*

The insulin and insulin-like growth factor-1 (IGF-1) receptors mediate signaling for energy uptake and growth through insulin receptor substrates (IRSs), which interact with these receptors as well as with downstream effectors. Oxygen is essential not only for ATP production through oxidative phosphorylation but also for many cellular processes, particularly those involved in energy homeostasis. The oxygen tension in vivo is significantly lower than that in the air and can vary widely depending on the tissue as well as on perfusion and oxygen consumption. How oxygen tension affects IRSs and their functions is poorly understood. Our findings indicate that transient hypoxia (1% oxygen) leads to caspase-mediated cleavage of IRS-1 without inducing cell death. The IRS-1 protein level rebounds rapidly upon return to normoxia. Protein tyrosine phosphatases (PTPs) appear to be important for the IRS-1 cleavage because tyrosine phosphorylation of the insulin receptor was decreased in hypoxia and IRS-1 cleavage could be blocked either with H2O2 or with vanadate, each of which inhibits PTPs. Activity of Akt, a downstream effector of insulin and IGF-1 signaling that is known to suppress caspase activation, was suppressed in hypoxia. Overexpression of dominant-negative Akt led to IRS-1 cleavage even in normoxia, and overexpression of constitutively active Akt partially suppressed IRS-1 cleavage in hypoxia, suggesting that hypoxia-mediated suppression of Akt may induce caspase-mediated IRS-1 cleavage. In conclusion, our study elucidates a mechanism by which insulin and IGF-1 signaling can be matched to the oxygen level that is available to support growth and energy metabolism.

Cellular oxygen tension is governed by a number of factors including the oxygen tension of the blood, the rate of perfusion, the distance from the nearest capillary, and the rate of oxygen consumption by neighboring cells (37). Although the partial pressure of oxygen in the air at sea level is 149 mm Hg (20%), it is considerably lower in vivo. The oxygen tension in arterioles ranges from 10 to 80 mm Hg depending on the diameter of the arteriole and the location. In capillaries, the tension ranges from 10 to 70 mm Hg and in veins from 10 to 40 mm Hg. The oxygen tension in the tissues decreases exponentially as a function of the distance from the blood vessels and varies depending on the tissue. The mean oxygen tension of the spleen, thymus, and myocardium are ϳ20, 10, and 5-15 mm Hg, respectively (38). The tissue oxygen tension can be further diminished in pathological conditions. For example, the oxygen tension can be severely reduced in tissues distal to an obstructed blood vessel. These types of severe hypoxia can lead to cell death and to tissue or organ damage. A more subtle fluctuation of cellular oxygen tension can occur in the setting of capillary recruitment and derecruitment. In endothelial cells, Akt phosphorylates and activates endothelial nitric-oxide synthase (eNOS), increasing the production of the potent vasodilator nitric oxide, which in turn recruits capillaries and increases the delivery of oxygen and nutrients to cells (39,40). Individuals suffering from sleep apnea develop hypoxemia and are insulin-resistant (41). Although individuals suffering from sleep apnea tend to be obese, hypoxemia is an independent risk factor for insulin resistance (42).

MATERIALS AND METHODS
Transfection and Hypoxic Incubation-Plasmids were transfected into 293T cells by using Effectene. After 40 h, the confluent transfected cells were placed into a hypoxic chamber (Hypoxia Work Station, Biotrace Inc.) for 6 h or left in the tissue culture incubator. Cells were then harvested by scraping into radioimmune precipitation assay buffer containing protease and phosphatase inhibitors. Hypoxic samples were harvested in the hypoxic chamber. Proteins were analyzed by SDS-PAGE followed by immunoblotting. The hypoxic samples from mouse embryonic fibroblast (MEF), 3T3L1, and MCF7 cells were produced in the same manner.
Degradation of IRS-1 in Vitro by Caspases-293T cells were transfected with a human IRS-1 expression vector. After 48 h, IRS-1 was immunoprecipitated using anti-IRS-1 antibody (Upstate catalog no. 06-248) and protein A/G beads (Pierce). The beads were washed three times with phosphate-buffered saline and once with 50 mM Hepes, pH 7.2, and divided equally into nine tubes. Digestion with caspases (Calbiochem) was performed according to the manufacturer's directions using 1 unit of each caspase at 37°C for 1 h in a final volume of 20 l. Digested samples were analyzed by electrophoresis using 4 -12% acrylamide gels followed by immunoblotting.
Treatment of Cells with Caspase 3/7 Inhibitor-293T cells transfected with plasmid expressing human IRS-1 were treated with 20 M Z-DEVD-fmk (Calbiochem, stock, 2 mM in Me 2 SO) 30 min prior to hypoxic treatment (2 h). HeLa cells were treated with 17 M Z-DEVD-fmk beginning 30 min prior to hypoxic treatment (3 h).
DNA Fragmentation Assay-Confluent HeLa or 293T cells transfected with human IRS-1 were treated or not with 2 M staurosporine overnight or for the indicated times and harvested by scraping into phosphate-buffered saline. The samples were divided in two so that both protein and genomic DNA could be extracted. Genomic DNA was harvested by using a DNA ladder kit (Biovision) and was analyzed by electrophoresis on a 1% agarose gel containing ethidium bromide. Simultaneously, confluent HeLa or 293T cells transfected with human IRS-1, which had been incubated in normoxia or 1% oxygen for 5 or 2 h, respectively, as well as untreated controls were scraped into phosphate-buffered saline. Each sample was divided in two and analyzed as above.
Soluble and Insoluble Fractionation-293T cells were ϳ98% confluent. Following normoxia or hypoxia (6 h), cells were scraped into radioimmune precipitation assay buffer. They were centrifuged for 20 min. The supernatant was the soluble fraction. The pellet was washed three times with radioimmune precipitation assay buffer and then extracted with urea buffer (20 mM Tris-HCl, pH 8, 8 M urea) to obtain the insoluble fraction.
Insulin and IGF-1 Treatment-Unless indicated otherwise, serum-starved cells were treated with 10 Ϫ5 units/ml insulin (65 pM) or 125 pM IGF-1 for 5 min immediately before harvesting. Cells were treated with 60 nM insulin for 5 min immediately before harvesting (see Fig. 7c).

RESULTS
Although the functions and properties of IRSs have been extensively characterized, virtually nothing is known about how oxygen tension affects the IRS proteins. To study the effect of mild hypoxia on the IRS proteins, we incubated a confluent culture of MEFs in serum-free media overnight, after which they were incubated either in a hypoxic chamber containing 1% oxygen or in a tissue culture incubator in room air (normoxia). To ensure that the oxygen tension did not change during the experiment, the lysate preparation after hypoxia treatment was also performed inside the hypoxic chamber. We discovered that within 1-4 h of hypoxia, the IRS-1 protein level was drastically reduced (Fig. 1a). The protein level of IRS-2 and of the insulin receptor did not change significantly with hypoxia. The IRS-1 level was also drastically reduced in other cell types including 3T3-L1 and MCF7 cells (Fig. 1a). Similar results were seen in other cell lines including NIH3T3, HeLa, and 293T cells (see below), suggesting that hypoxia-induced reduction in IRS-1 levels may be a general phenomenon. However, it was possible that chamber conditions unrelated to the oxygen tension might have been responsible for the reduction in the IRS-1 level. To rule out this possibility, we exposed MEFs to 5% oxygen in the same chamber (Fig. 1b). Whereas IRS-1 was almost undetectable after 4 h in 1% oxygen, the IRS-1 level was only slightly reduced in 5% oxygen, confirming that the reduction of the IRS-1 level was oxygen-dependent. We measured the IRS-1 mRNA level with real-time PCR and detected no significant decrease with hypoxia (data not shown), suggesting that the hypoxia-induced reduction in IRS-1 levels occurred at the level of translation or post-translation.
It has been shown previously that prolonged treatment with insulin also leads to degradation of IRS-1 protein (43)(44)(45). Ser/ Thr hyperphosphorylation of IRS-1, which is induced by prolonged insulin treatment, is required for degradation (46), and inhibition of PI3K blocks IRS-1 degradation (44,45,47,48). Ser/Thr hyperphosphorylation of IRS-1 and -2 can be induced with the phosphatase inhibitor okadaic acid in the absence of insulin treatment, indicating that there is basal Ser phosphorylation. Treatment of cells with okadaic acid resulted in an electrophoretic mobility shift of FIGURE 2. Hypoxia-induced reduction of IRS-1 level is not mediated by proteasomal degradation. a, After 2 h in hypoxia (H), before the IRS-1 level was decreased, 293T cells were treated with two different doses of okadaic acid for 1 h. Cells exposed to normoxia (N) were used as controls. Cell extracts were subjected to SDS-PAGE and immunoblotted with antibody specific for IRS-1, phosphorylated Ser-302 and Ser-612 in IRS-1, IRS-2, and actin. b, treatment with okadaic acid restores IRS-1 level in hypoxia. 293T cells exposed to hypoxia for 4 h were treated with okadaic acid for 1 h. Cell extracts were processed as described for panel a. c, treatment with the proteasome inhibitor MG132 (MG) did not block the reduction of the IRS-1 level. Cleavage of procaspase 3 to caspase 3 in hypoxia is detected by immunoblotting with procaspase 3 antibody. b, recombinant caspases were incubated with IRS-1 immunoprecipitated from 293T cells that had been transiently transfected with the human IRS-1 expression vector. c, 293T cells were exposed to hypoxia for 4 h with or without the caspase inhibitor Z-DEVD-fmk. 293T cells exposed to normoxia (N) were used as a control. Hypoxia-induced IRS-1 degradation was blocked by the caspase inhibitor.
IRS-1, a consequence of Ser/Thr hyperphosphorylation. Indeed, okadaic acid treatment dramatically increased the phosphorylation of Ser-302 and Ser-612 in IRS-1 (Fig. 2a). If reductions of the IRS-1 level by hypoxia and by chronic insulin treatment were mediated by the same biochemical pathway, phosphorylation of Ser-302 and Ser-612 and gel mobility shift should occur prior to the reduction of IRS-1 level by hypoxia. However, phosphorylation of Ser-302 and Ser-612 and gel mobility shift were not induced by hypoxia. It is possible that Ser-302 and Ser-612 were not phosphorylated because the cells were not exposed to hypoxia long enough. However, this is unlikely to be the case because the presence of okadaic acid during exposure (4 h) to hypoxia prevented reduction of IRS-1 levels (Fig. 2b). Taken together, these results indicate that the reduction of IRS-1 level by hypoxia and by chronic insulin treatment occurs through different pathways. The reduction in IRS-1 level after prolonged insulin (45) or IGF-1 (44) treatment is mediated by proteasome-dependent degradation. To determine whether proteasome-dependent degradation decreases IRS-1 levels in hypoxia, hypoxic cells were treated with the proteasome inhibitor MG132. As shown in Fig. 2c, MG132 did not prevent IRS-1 degradation in hypoxia. Taken together, these findings indicate that the hypoxia-induced decrease in IRS-1 levels is not mediated by Ser/Thr hyperphosphorylation and proteasome-mediated degradation and therefore is mechanistically different from the insulin-induced degradation of IRS-1.
We then asked whether the caspase pathway (49), which is induced by apoptotic signals but suppressed by survival signals, might be responsible for IRS-1 degradation. First, we examined whether transient hypoxia affected caspase 3 activation. As shown in Fig. 3a, cleavage of procaspase 3 to active caspase 3 increased with hypoxia. Then, we tested a panel of recombinant caspases to determine whether caspases are capable of degrading IRS-1 directly (Fig. 3b). IRS-1 was immunoprecipitated from cells with IRS-1-specific antibody and incubated with recombinant caspases 1-10. Caspases 3 and 7 had very potent IRS-1-degrading activity, and caspase 10 had significant but weaker IRS-1-degrading activity. To determine whether caspases 3 and 7 mediated hypoxia-induced cleavage of IRS-1 in cells, we incubated hypoxic cells with Z-DEVD-fmk, an inhibitor of caspases 3 and 7 (50). As shown in Fig. 3c, inhibition of caspases 3 and 7 prevented hypoxia-induced IRS-1 degradation.
Because caspases are normally involved in executing apoptosis, we isolated genomic DNA from HeLa cells exposed to hypoxia for 4 h and electrophoresed it on an agarose gel (Fig.  4a). As a positive control, we isolated genomic DNA from HeLa a, no DNA laddering in HeLa cells exposed to hypoxia for 4 h. Genomic DNA from cells exposed to normoxia (N) or hypoxia (H) for 5 h or to staurosporine (St) for 24 h was isolated and electrophoresed in an agarose gel, which was stained with ethidium bromide. b, IRS-1 is degraded in staurosporine-treated apoptotic cells. The IRS-1 level in human IRS-1 transfected 293T cells that were treated with staurosporine for 6.5 h or with hypoxia for 4 h is shown. c, 293T cells exposed to hypoxia for 4 h were returned to atmospheric oxygen for 1 or 5 min. IRS-1 level in cells exposed to normoxia was used as a control. d, soluble fraction and insoluble fraction, which was solubilized by extraction into urea buffer, were prepared from 293T cells exposed to normoxia or hypoxia for 6 h. All IRS-1 was in the soluble fraction of cells exposed to normoxia and hypoxia. FIGURE 5. Multiple sites in IRS-1 regulate IRS-1 degradation. a, IRS-1 mutants containing the deletions (gaps) shown below were transiently expressed in 293T cells. IRS-1 degradation was detected by immunoblotting with IRS-1 antibody specific for the N terminus after cells were exposed to normoxia (N) or hypoxia (H) for 4 h. Hypoxia-induced degradation of deletion mutants 1-3 (del 1-3) are shown at the top. The other deletion mutants were either strongly degraded (ϩ) or weakly degraded (ϩ/Ϫ) in hypoxia. The amino acid residues deleted in each deletion construct are indicated in parentheses. b, epitope-tagged IRS-1/2 chimeric proteins containing the corresponding IRS-1 (black) and IRS-2 (white) portions were transiently expressed in 293T cells. The amino acid residue of IRS-1 at which IRS-1 was joined to IRS-2 is indicated in parentheses. Hypoxia-induced degradation was detected by immunoblotting after cells were exposed to normoxia or hypoxia for 4 h. cells that had been treated with staurosporine to induce apoptosis. Although there was DNA laddering in staurosporinetreated cells, there was no DNA laddering in hypoxic cells, indicating that apoptosis was not induced during 4 h of hypoxia. Consistent with this, cells that were previously exposed to hypoxia for 4 h had a normal proliferation rate when returned to atmospheric oxygen (data not shown). Significantly longer incubation (e.g. more than 16 h) in hypoxia did induce some cell death. If caspases degraded IRS-1 in hypoxia, activation of caspases during staurosporine-induced apoptosis should lead to IRS-1 degradation. Indeed, IRS-1 was degraded in staurosporine-treated cells (Fig. 4b).
To determine whether IRS-1 cleavage is irreversible, we examined IRS-1 levels after transfer of cells from hypoxia to atmospheric oxygen. IRS-1 levels started to recover in less than 1 min, suggesting that the IRS-1 level can fluctuate very rapidly with oxygen tension (Fig. 4c). Because IRS-1 levels rebounded so rapidly, we considered the possibility that compartmentalization of IRS-1, rather than stability of IRS-1, was affected by hypoxia. For example, it is possible that IRS-1 moved from the cytosol to the cell compartments that form the insoluble fraction, which is discarded during lysate preparation, but returned to the cytosol with reoxygenation. The insoluble fraction, which had been discarded in past experiments, was solubilized and immunoblotted with IRS-1 antibody after electrophoresis (Fig. 4d). With or without hypoxia, there was no IRS-1 in the insoluble fraction, suggesting that the fluctuation in IRS-1 level with hypoxia and reoxygenation was not a result of compartmental shift.
Because IRS-1, but not IRS-2, was cleaved in hypoxia, we sought to determine the site of regulation of IRS-1 cleavage (Fig. 5a). 293T cells were transiently transfected with IRS-1 expression vectors containing various deletions and exposed to hypoxia for 4 h. As shown in Fig. 5a, no single deletion abrogated hypoxia-induced IRS-1 degradation. We transiently transfected expression vectors for IRS-1/2 chimeric proteins in which different regions of IRS-1 were replaced with the corresponding regions from IRS-2. All three chimeric proteins were degraded with hypoxia. Thus, it appears that multiple regions in IRS-1 can regulate or mediate caspase-dependent degradation. This finding is in contrast to prolonged insulin-induced degradation of IRS-1, which requires the N-terminal region of IRS-1 (51). To determine whether the IRS-dependent pathway was affected by hypoxia, we examined insulin-induced tyrosine phosphorylation of the IR (␤-subunit) and of IRS-1 in NIH3T3 cells exposed to hypoxia for only 1 h. Although there was not enough time for the IRS-1 level to decline, tyrosine phosphorylation of IR and IRS-1 was decreased (Fig. 6a). Consistent with this finding, the amount of the p85 subunit of PI3K (18) coimmunoprecipitated with IRS-1 was decreased in hypoxic NIH3T3 and MEF cells (Fig. 6b). In hypoxic cells, Akt activation by IGF-1 (Fig. 6c) or insulin (Fig. 6d) was modestly, but significantly, reduced compared with normoxic cells. Previously, Akt had been shown to suppress caspase activity (52). To determine whether suppression of Akt could affect cleavage of IRS-1, we coexpressed dominant-negative (kinase-dead) Akt or constitutively active Akt with IRS-1. Expression of dominant-negative Akt resulted in IRS-1 degradation even in normoxia, whereas expression of constitutively active Akt suppressed IRS-1 degradation modestly (Fig. 6e). Taken together, these results suggest that Akt, although downstream of IRS, may also regulate IRS-1 cleavage in hypoxia.
Protein tyrosine phosphatases (PTPs), which negatively regulate insulin and IGF-1 signaling by dephosphorylating tyrosine residues on IR, IGF-1R, and IRS proteins, are transiently suppressed by the PI3K pathway (see Fig. 7e) (53). It is thought that phosphatidylinositol trisphosphate recruits guanine nucleotide exchange factors to the cell membrane for the small GTP-binding protein Rac1. Activated Rac1 in turn stimulates NADPH oxidase, converting oxygen to superoxide, which upon dismutation is converted to H 2 O 2 (54 -56). PTPs such as PTP1B (57,58), which reverses the tyrosine phosphorylation stimulated by insulin, contain an active site cysteine residue that is susceptible to oxidation by oxidants such as H 2 O 2 . Once oxidized, PTP is inactivated, and the balance shifts toward tyrosine phosphorylation until H 2 O 2 is degraded and PTP is reactivated by electron donors. This negative control of PTP is necessary to prevent a futile phosphorylation-dephosphorylation cycle during insulin and IGF-1 signaling. We show in Fig. 6a that tyrosine phosphorylation of IR and IRS-1 is reduced in hypoxia. It is possible that reduced production of H 2 O 2 in hypoxia may have increased the PTP activity, which in turn led to suppression of Akt and its ability to suppress caspases. If this were the case, PTPs would be important for hypoxia-induced degradation of IRS-1. To test this possibility, we treated hypoxic cells with the PTP inhibitor vanadate (Fig. 7a). The IRS-1 level was partially restored with vanadate treatment in hypoxia. Treatment with H 2 O 2 also restored the IRS-1 level in hypoxia (Fig. 7b). Moreover, strong stimulation of the PI3K-Akt pathway with a higher concentration of insulin (60 nM) also inhibited IRS-1 degradation in hypoxia (Fig. 7c). Taken together, these findings are consistent with the notion that hypoxia-induced activation of PTPs, which suppresses Akt and its ability to suppress caspases, is important for IRS-1 degradation in hypoxia (Fig. 7e). In an attempt to identify the PTP important for IRS-1 cleavage, we examined IRS-1 cleavage in PTP1B Ϫ/Ϫ and PTP1B ϩ/ϩ MEFs. Previous work had shown that PTP1B is a major inhibitor of insulin signaling (57,58). However, IRS-1 cleavage occurred in both hypoxic PTP1B Ϫ/Ϫ and PTP1B ϩ/ϩ MEFs, suggesting that PTP1B is either not important for hypoxia-induced IRS-1 cleavage or that in the absence of PTP1B, other PTPs can take its place. We cannot distinguish between these two possibilities at this point.

DISCUSSION
The oxygen tension in the microenvironment of a cell in vivo is significantly different from that generally used in most in vitro experiments. Depending on the tissue, the oxygen tension can vary widely, but it is significantly lower than the atmospheric oxygen tension. The intracellular oxygen tension is controlled by a number of factors including the efficiency of oxygenation in the lungs, the blood flow in the arteries, arterioles, and capillaries that supply the tissue, the rate of oxygen consumption, and the distance from the nearest capillaries. Oxygen is essential for growth and for energy generation as well as for many other cellular processes, especially those involved in energy homeostasis. In this report, we have examined the effect of mild hypoxia on IRS proteins. Our findings indicate that IRS-1 is cleaved by caspases during transient hypoxia. Although caspases were involved in IRS-1 cleavage, transient exposure to hypoxia did not cause apoptosis, and IRS-1 cleavage was reversed within a few minutes of reoxygenation. PTPs may be important for the hypoxic effect because tyrosine phosphorylation of the IR is decreased in hypoxia and also because inhibition of PTPs with H 2 O 2 or vanadate blocks IRS-1 cleavage. It is difficult to prove directly that PTPs are activated in hypoxia, because the redox state of the PTPs is largely determined by the redox potential of the buffers used in the PTP reaction. Although NADPH oxidase is the best known candidate to be the oxygen sensor in the regulation of IRS-1 degradation, we have not been able to demonstrate that NADPH oxidase is important for IRS-1 degradation. However, based on the current knowledge of the role of H 2 O 2 in PTP inhibition, our find-FIGURE 7. Inhibition of protein tyrosine phosphatase inhibits hypoxia-induced IRS-1 degradation. a, NIH3T3 cells were exposed to normoxia (N) or hypoxia (H) for 4 h with increasing doses of vanadate. The levels of IRS-1 (top) and actin (bottom) are shown. b, cells were exposed to normoxia or hypoxia for 4 h and then treated with 100 M H 2 O 2 for 15 min. The levels of IRS-1 and actin are shown. c, cells were exposed to normoxia or to hypoxia with or without insulin (60 nM). The levels of IRS-1 and actin are shown. d, MEF derived from PTP1B ϩ/ϩ and PTP1B Ϫ/Ϫ embryos were exposed to normoxia or hypoxia for 4 h. The levels of IRS-1, PTP1B, and actin are shown. e, diagram of the IRS-1-dependent signaling pathway and the proposed role of oxygen in the regulation of this pathway are shown. MAP, mitogen-activated protein.
ings are consistent with the notion that hypoxia activates PTPs. Consistent with PTP activation and IRS-1 cleavage in hypoxia, Akt activity is suppressed in hypoxia. Akt, which is known to inhibit caspases, also appears to be an important regulator of IRS-1 cleavage, because overexpression of dominant-negative Akt induces IRS-1 cleavage even in normoxia. Conversely, overexpression of constitutively active Akt partially blocks hypoxia-mediated IRS-1 cleavage.
Controlling the IRS-1 level is important for energy homeostasis. For example, prolonged insulin stimulation, which leads to Ser/Thr phosphorylation of IRS-1, leads to proteasome-mediated IRS-1 degradation (43)(44)(45). This negative feedback protects against excessive insulin signaling. We suspect that controlling the IRS-1 level by caspase-mediated cleavage in cells exposed to transient hypoxia may also be important for energy homeostasis by matching insulin and IGF-1 signaling with the available oxygen tension. The net result of matching insulin and IGF-1 signaling to oxygen levels is that it restricts the growth signal and supply of nutrients to the cells that have sufficient oxygen supply, thereby allowing for more efficient energy distribution.