Protein Kinase A-dependent Translocation of Hsp90α Impairs Endothelial Nitric-oxide Synthase Activity in High Glucose and Diabetes*

Diabetes mellitus (DM) and high glucose (HG) are known to reduce the bioavailability of nitric oxide (NO) by modulating endothelial nitric-oxide synthase (eNOS) activity. eNOS is regulated by several mechanisms including its interaction with heat shock protein (Hsp) 90. We previously discovered that DM in vivo and HG in vitro induced the translocation of Hsp90α to the outside of aortic endothelial cells. In this report we tested the hypothesis that translocation of Hsp90α is responsible for the decline in NO production observed in HG-treated cells. We found that HG increased phosphorylation of Hsp90α in a cAMP-dependent protein kinase A-dependent manner, and that this event was required for translocation of Hsp90α in porcine aortic endothelial cells. Furthermore, preventing translocation of Hsp90α protected from the HG-induced decline in eNOS·Hsp90α complex and NO production. Notably, DM increased phosphorylation of Hsp90α and reduced its association with eNOS in the aortic endothelium of diabetic rats. These studies suggest that translocation of Hsp90α is a novel mechanism by which HG and DM impair eNOS activity and thereby reduce NO production.

Diabetes mellitus (DM) and high glucose (HG) are known to reduce the bioavailability of nitric oxide (NO) by modulating endothelial nitric-oxide synthase (eNOS) activity. eNOS is regulated by several mechanisms including its interaction with heat shock protein (Hsp) 90. We previously discovered that DM in vivo and HG in vitro induced the translocation of Hsp90␣ to the outside of aortic endothelial cells. In this report we tested the hypothesis that translocation of Hsp90␣ is responsible for the decline in NO production observed in HG-treated cells. We found that HG increased phosphorylation of Hsp90␣ in a cAMP-dependent protein kinase A-dependent manner, and that this event was required for translocation of Hsp90␣ in porcine aortic endothelial cells. Furthermore, preventing translocation of Hsp90␣ protected from the HG-induced decline in eNOS⅐Hsp90␣ complex and NO production. Notably, DM increased phosphorylation of Hsp90␣ and reduced its association with eNOS in the aortic endothelium of diabetic rats. These studies suggest that translocation of Hsp90␣ is a novel mechanism by which HG and DM impair eNOS activity and thereby reduce NO production.
Endothelium-derived nitric oxide (NO) 2 , a small, diffusible, lipophilic free radical gas, is a critical regulator of cardiovascular homeostasis (1,2). Vascular NO dilates all types of blood vessels. It also inhibits platelet aggregation and adhesion, and leukocyte adhesion to the vessel wall (3)(4)(5). A deficiency of NO (decreased NO production, increased degradation of NO or decreased NO sensitivity) can promote atherogenesis (1,6).
Endothelial NO synthase (eNOS) is responsible for most of the vascular NO produced. A functional eNOS oxidizes its substrate L-arginine to L-citrulline and NO (4,5). eNOS is highly regulated by post-translational modifications including myristoylation, palmitoylation, and phosphorylation (7)(8)(9)(10)(11). This enzyme is also tightly regulated by specific interactions with inhibitory and activating proteins such as caveolin-1 and Hsp90, respectively. The binding of heat shock protein 90 (Hsp90) to eNOS enhances eNOS activation, assists with the intracellular trafficking of eNOS, and helps activate eNOS by dissociating it from caveolin-1 (12,13). Furthermore, Hsp90 provides a scaffold for eNOS and Akt, and thereby enhances eNOS activation (9, 14 -16).
Several studies have demonstrated that hyperglycemia/diabetes mellitus (DM) causes a loss of endothelium-derived NO in both animals (17)(18)(19) and humans (20), but the underlining mechanism is poorly understood. DM does not influence the overall eNOS protein level or its mRNA level (21). One potential mechanism may be the reduced association of eNOS with Hsp90 (19).
Hsp90 is an ATP-dependent chaperone that interacts with over 100 proteins (22,23), and has been implicated in many physiological and pathological processes (24 -27). DM induces an increase in the amount of Hsp90␣ at the luminal surface of the aorta, and this event could be mimicked by treating cultured vascular endothelial cells with high glucose (HG) (28). Expression of Hsp90 on the cell surface has also been published by several other laboratories (29 -31).
In this report, we tested the hypothesis that HG reduced NO production by translocating Hsp90␣ out of the cell and thereby reducing the amount of Hsp90␣ available to interact with and activate eNOS. Our findings strongly support this hypothesis and suggest that a similar mechanism operates in DM.

EXPERIMENTAL PROCEDURES
Animals and Aorta Isolation-Sprague-Dawley male rats were made diabetic by injecting streptozotocin (55 mg/kg body weight) into the tail vein. Insulin was given as needed (28) to achieve a slow weight gain while maintaining hyperglycemia and glycosuria. Glycohemoglobin was measured (Glyc-Affin, Pierce, Rockford, IL) to assess the degree of hyperglycemia (17 Ϯ 1.9% in diabetic versus 4.5 Ϯ 0.5% in control rats, p Ͻ 0.00001). The thoracic aortas were isolated from sacrificed rats, which had endured 5 or 6 months of DM, or age-and sexmatched controls. Before sacrifice, the rats were anesthetized with ketamine/xylazine (70/9 mg/kg body weight), and then perfused with pre-warmed PBS (phosphate-buffered saline, pH 7.4) through the left ventricle using a catheter connected to a peristaltic pump (5 min, 100 mm Hg). The thoracic aorta was harvested, freed of periadventitial fat, dissected longitudinally, and then washed in ice-cold PBS. The protocol for the use of animals was approved by the Schepens Animal Care and Use Committee.
Cell Culture and Preparation of Cell Lysates-The porcine aortic endothelial cells (PAEC) were from Dr. Lena Claesson-Welsh (Uppsala University, Uppsala, Sweden). They were grown in a 1:1 mixture of Dulbecco's modified Eagle's medium (low glucose, Invitrogen), F-12 nutrient mixture (Ham's) (Invitrogen), which was supplemented with 10% fetal bovine serum (Gemini Bio-Products, Calabasas, CA). To achieve "HG," D-glucose was added to the medium to a final concentration of 30 mmol/liter; "normal glucose" was 5 mmol/liter. The "HG treatment" consisted of a 2-week incubation in the HGcontaining medium.
PKA Phosphorylation Assay-Full-length Hsp90␣ protein was purchased from Stressgen Biotechnologies (Victoria, BC, Canada), whereas the GST-H90N, GST-H90T89A, and GST-H90C fusion proteins were generated as described above. For each of the substrates, 100 ng was incubated with or without 1 l (2500 units) of the catalytic subunit of PKA (cAMP-dependent protein kinase (PKA) (New England Biolabs, Beverly, MA) in kinase buffer (50 mmol/liter Tris-HCl, pH 7.5, 10 mmol/liter MgCl 2 , 2.5 mmol/liter ATP) in a total volume of 25 l at 30°C for 1 h (34). The reactions were terminated by lowering the temperature to 0°C or by adding sample buffer. Following the kinase assay the proteins were subjected to SDS-PAGE and Western blot using a phospho-(Ser/Thr) PKA substrate antibody. The membranes were stripped and then reprobed (28,35) using anti-H90N and H90C antibodies. The Western blots were developed using the ECL Plus kit (Amersham Biosciences). The resulting data were scanned and quantified using Quantity One software (Bio-Rad).
For the stoichiometry experiments full-length Hsp90␣ was phosphorylated as described above, except that the concentration of unlabeled ATP was 20 mol/liter, 20 Ci of ␥-[ 32 P]ATP (6.6 ϫ 10 Ϫ6 mol/liter) was added and the reactions were run for 3 h. The proteins were separated by 10% SDS-PAGE, the gel was dried and subjected to autoradiography. The amount of radioactivity incorporated into Hsp90␣ was determined by scintillation counting of the excised Hsp90␣ band.
Overexpression of PKA-A pcDNA 3.1 (ϩ)-Flag-PKA plasmid containing the full-length mouse PKA (1047 bp) was kindly provided by Dr. Jiuyong Xie (University of Manitoba, Canada). The entire coding region of catalytic PKA (including the FLAG tag), was subcloned into pLPCX retroviral vector (BD Biosciences). The plasmids were transfected into 293GPG cells to obtain virus and the resulting viruses were used to infect PAECs as outlined previously (28).
Site-directed Mutagenesis of Hsp90␣-Mutants were introduced into Hsp90␣ using the QuikChange XL site-directed mutagenesis kit (Stratagene). The primers used for the Thr to Ala substitution at residue 89 were 5Ј-GATCGAACTCTTGC-AATTGTGGATACTGGAATTGGAATGACCAAG-3Ј and 5Ј-CTTGGTCATTCCAATTCCAGTATCCACAATTGCA-AGAGTTCGATC-3Ј. The mutation was confirmed by DNA sequencing and then subcloned into the pLNCX2 (BD Biosciences) retroviral vector.
Immunoprecipitation and Western Blotting-To immunoprecipitate the population of Hsp90␣ on the cell surface, PAECs were washed 3 times with cold PBS, and then the intact cells were incubated with an anti-Hsp90␣ antibody (Santa Cruz Biotechnology) or normal goat IgG (2 g/ml in 0.5% bovine serum albumin/PBS) for 30 min on ice with gentle shaking. The unbound antibody was removed by washing 3 times with cold PBS. Cells were then lysed by adding EB (1 ml/per dish), the dishes were scraped and the insoluble material was separated by centrifugation (13,000 ϫ g for 15 min at 4°C). The clarified supernatant was incubated with 50 l of protein G Plus-agarose (Santa Cruz Biotechnology) per sample for 2 h. The beads were washed 5 times with 1 ml of EB buffer each time, and resuspended in 1ϫ SDS sample buffer.
To immunoprecipitate Hsp90␣ from all cellular compartments, cells were first lysed in EB buffer, and the anti-Hsp90␣ antibody was added to the lysate. The remainder of the immunoprecipitation procedure was the same as described above. To immunoprecipitate Hsp90␣ from the aortic endothelium, lysates from the aortic endothelial layer of 5 diabetic or nondiabetic rats were obtained as described previously (28), and then incubated with 3 g of either an anti-Hsp90␣ antibody or non-immune IgG for 4 h at 4°C. The resulting immune complexes were captured on protein G Plus-agarose (Santa Cruz Biotechnology) for 2 h at 4°C. The agarose beads were washed 5 times with 1 ml of EB buffer each time, and resuspended in 1ϫ SDS sample buffer.
Immunoprecipitates from cultured cells were subjected to 10% SDS-PAGE and Western blotting using a phospho-(Ser/ Thr) PKA substrate antibody; the stripped membranes were reprobed using an anti-Hsp90␣ antibody. Immunoprecipitates from rat aortic endothelium were subjected to 10% SDS-PAGE and Western blotting using a phospho-(Ser/Thr) PKA substrate antibody; the stripped membranes were reprobed using anti-Hsp90␣ and anti-eNOS antibody.
To co-immunoprecipitate eNOS and Hsp90, the lysates prepared from PAECs cultured in normal glucose or HG for 2 weeks were incubated with anti-eNOS antibody (0.5 g) or non-immune IgG as described above. The resulting samples were subjected to an anti-eNOS Western blot followed by an anti-Hsp90 Western blot.
FACS Analysis-Following the desired experimental treatment, cells were washed twice with PBS and detached with cell dissociation solution (Sigma). The cells were incubated with a goat anti-Hsp90␣ antibody or non-immune goat IgG for 1 h on ice, washed twice with ice-cold PBS, incubated with a fluorescein isothiocyante-conjugated AffiniPure mouse antigoat secondary antibody for 30 min on ice, rinsed twice with ice-cold PBS, and then stained with propidium iodide (BD Biosciences). The cells were finally analyzed by flow cytometry (20,000 events) in Coulter Beckman XL (Beckman Coulter Inc., Miami, FL).
NO Assay-PAECs were cultured in normal glucose or HG for 2 weeks. The media were collected, and spun at 5,000 ϫ g for 5 min, and the supernatants were subjected to an NO assay following the protocol provided by the Griess reagent kit (Molecular Probes, Inc., Eugene, OR). This method involves the Griess diazotization reaction and spectrophotometric (at 548 nm) detection of nitrite formed by the spontaneous oxidation of NO under physiological conditions.
Statistics-Comparisons were made using unpaired and paired t test; a confidence level of p Ͻ 0.05 was considered statistically significant.

RESULTS
HG Increased Phosphorylation of Hsp90␣ in a PKA-dependent Manner-Whereas DM is known to decrease NO bioavailability (6,19), the underlying mechanism has not be elucidated. Our observation that DM (and HG treatment of cultured endothelial cells) promoted translocation of an eNOS activator (Hsp90) (28) suggested that NO declines because of the translocation of an eNOS activator. To test this hypothesis we determined if preventing translocation of Hsp90␣ prevented the decline in NO production caused by HG.
Because Hsp90␣ does not have a signal peptide, its translocation must proceed via a non-classical route used by other excreted proteins that also lack signal peptides such as basic fibroblast growth factor and galectin (36). Others have found that HG activates PKA in certain cell types (including vascular endothelial cells) (37)(38)(39)(40)(41), and that PKA-dependent phosphorylation of proteins promotes their translocation (34,40,42). Consequently, we speculated that HG promoted PKA phosphorylation of Hsp90␣, resulting in its translocation out of the cell. To test this hypothesis, we searched for potential PKA phosphorylation sites within Hsp90␣ using the Scansite software. The results showed a low stringency PKA phosphorylation site (NKQDRTLT 89 IVDTGIG) at position 89. This PKA site (underlined amino acids) was identical in all vertebrate species examined (human, pig, hamster, mouse, and zebra fish). The commercially available phosphospecific PKA substrate antibody recognizes proteins containing RRXpS or RXXpT, and this antibody should recognize Hsp90␣ phosphorylated at Thr-89. The first step to test our hypothesis that HG induced PKA-dependent translocation of Hsp90␣ was to determine whether HG induced phosphorylation of proteins that can be recognized by this antibody. Indeed, we found that HG stimulated the phosphorylation of numerous proteins at a consensus PKA phosphorylation site (Fig. 1).
To test if Hsp90␣ was one of these proteins, Hsp90␣ was immunoprecipitated and subjected to Western blotting using the phosphospecific PKA substrate antibody. We found that HG increased phosphorylation of Hsp90␣ by 2.2 Ϯ 0.2-fold ( Fig. 2A). Blocking PKA activity by preincubating the cells with a PKA inhibitor (0.5 mol/liter PKI for 12 h) reduced phosphorylation of Hsp90␣ in HG-treated cells (Fig. 2B). These data strongly suggested that the HG-induced phosphorylation of Hsp90␣ was PKA dependent.
To test if PKA is capable of directly phosphorylating Hsp90␣, we performed an in vitro kinase assay using purified PKA and either full-length Hsp90␣ or fusion proteins containing portions of Hsp90␣. Full-length Hsp90␣, or a GST fusion protein that included the N-terminal portion (1-284 aa) was readily phosphorylated in this in vitro setting (Fig. 3A). In contrast, we were unable to detect phosphorylation of the C-terminal fragment of Hsp90␣ (285-732 aa) (Fig. 3A). Thus PKA was capable of phosphorylating Hsp90␣, and the phosphorylation site(s) was(were) within the first 284 amino acids. The stoichiometry of phosphorylation of full-length Hsp90␣ was 15.6 Ϯ 0.7% (supplemental materials Fig. S1).
As mentioned above, there is a PKA phosphorylation site at position 89 of Hsp90␣, and Hsp90␣ phosphorylated at this site would be recognized by phosphospecific PKA substrate antibody. Consistent with this idea is the observation that mutating Thr-89 to alanine (T89A) abrogated recognition of the in vitro phosphorylated GST fusion containing the N-terminal portion of Hsp90␣ by the phosphospecific PKA substrate antibody (Fig.  3B). These findings indicated that PKA directly phosphorylated Hsp90␣ at Thr-89.  3-fold more phosphorylated Hsp90␣ in samples prepared from the HGtreated cells as compared with those cultured in normal glucose. In three independent experiments, the fold difference was 2.2 Ϯ 0.2. B, same as A except the HG-treated cells were exposed to the PKA inhibitor PKI (0.5 mol/ liter) for 12 h prior to lysis. The PKI treatment reduced the amount of Hsp90␣ recognized by the phospho-PKA substrate antibody by 2.6-fold. In three independent experiments, the average fold difference was 2.7 Ϯ 0.2. In the bottom half of this panel the membrane was stripped and reprobed using a combination of antibodies that recognized the N-and C-terminal fusion proteins. Although equivalent amounts of fusion protein was used in these experiments, the stronger signal for the full-length protein probably reflects the fact that the full-length protein was recognized by both antibodies, whereas only one of the antibodies reacted with the N-or C-terminal fusion proteins. B, PKA-dependent phosphorylation of Hsp90␣ requires Thr-89. Same as A except that the T89A mutant was used in the right-hand lane. The experiments shown in A and B were repeated on three independent occasions, and the results from each of these experiments was comparable with the data presented. MARCH 30, 2007 • VOLUME 282 • NUMBER 13

JOURNAL OF BIOLOGICAL CHEMISTRY 9367
Translocation of Hsp90␣ Required Phosphorylation at Thr-89-If phosphorylation of Hsp90␣ by PKA promoted its translocation to the cell surface, then the phosphorylated form of Hsp90␣ should predominate on the cell surface. To test this idea we prepared Hsp90␣ immunoprecipitates from the cell surface or from total cell lysates, which contained both the cell surface and intracellular pools. Consistent with our previous observation (28) less than 10% of the total Hsp90␣ was on the cell surface (data not shown), and hence we observed much more Hsp90␣ recovered from the total cell lysate samples (Fig.  4). After normalizing for the amount of Hsp90␣ present in each of the immunoprecipitates, we found that the cell surface samples contained 11.8 Ϯ 1.5-fold more PKA-phosphorylated Hsp90␣ (Fig. 4). There was also a greater fraction of the phosphorylated form of Hsp90␣ on the surface in the normal glucose setting (supplemental materials Fig. S2). Because these comparisons were between the surface pool and the total (intracellular ϩ surface populations), they probably underestimate the difference between the intra-and extracellular pools of Hsp90␣. Thus, whereas we are uncertain of the precise ratio between the phosphorylated Hsp90␣ that is translocated and that which remains within the cell, our data clearly show that the majority of phosphorylated Hsp90␣ was translocated out of the cell.
To test if elevating PKA activity was sufficient to induce Hsp90␣ translocation (in the absence of the many additional changes induced by HG), we stably overexpressed the catalytic domain of PKA (Fig. 5A). As expected, phosphorylation of Hsp90␣ (Fig. 5B) and many other proteins (supplemental materials Fig. S3A) was elevated in the cells overexpressing PKA. Furthermore, the PKA-overexpressing cells exhibited a 2.4 Ϯ 0.2-fold increase in surface Hsp90␣ (sHsp90␣) (Fig. 5C and  supplemental materials Fig. S3B). The idea that PKA promoted translocation of Hsp90␣ was further supported by the finding that blocking PKA activity with PKI reduced sHsp90␣ 3.3 Ϯ 0.3-fold ( Fig. 5D and supplemental materials Fig. S4A), without altering the total level of Hsp90␣ (data not shown). Treating cells with a second PKA inhibitor (H89) also reduced the level of sHsp90␣ (data not shown), whereas PKC inhibitors had no effect (supplemental materials Fig. S4B). The results from this combined pharmacological and molecular approach indicated that activation of PKA was sufficient to induce translocation of Hsp90␣.
To learn if phosphorylation of Hsp90␣ was requisite for its translocation to the cell surface, we investigated whether a mutant Hsp90␣ that could not be phosphorylated was unable to accumulate on the cell surface. The strategy was to reduce the endogenous porcine Hsp90␣ using porcine-specific siRNA, and then reconstitute the cells using a different species of Hsp90␣ to evade the porcine-specific siRNA oligos. We used human wild type (WT) or T89A Hsp90␣ and tested the level of sHsp90␣ in the resulting cells. Stable expression of a combination of siRNAs directed toward porcine Hsp90␣ reduced the endogenous level of Hsp90␣ by 71.5 Ϯ 2.2% (Fig. 6A); there was a corresponding decrease in the level of sHsp90␣ (Fig. 6B). Expressing human WT Hsp90␣ restored both total and surface Hsp90␣ to the control level ( Fig. 6 and supplemental materials Fig. S5). In contrast, sHsp90␣ remained low following expression of the human T89A Hsp90␣ mutant, despite restoration of total Hsp90␣ to the control level ( Fig. 6 and supplemental materials Fig. S5). Substitution of the threonine at position 89 to FIGURE 4. Phosphorylated Hsp90␣ was predominately on the cell surface. PAECs were treated with HG (30 mmol/liter) for 2 weeks. Hsp90␣ was immunoprecipitated (IP) with an anti-Hsp90␣ antibody from the cell surface (using 3 ϫ 10 7 cells/sample) or from total lysates (using 0.5 ϫ 10 7 cells/sample) as described under "Experimental Procedures." Immunoprecipitates were prepared in parallel with non-immune IgG. The resulting samples were subjected to a phosphospecific PKA substrate Western blot (top panel). The stripped membrane was reprobed with a Hsp90␣ antibody (bottom panel). The ratio of the signal from the bottom and top panels is presented. The results showed that there was 13.6-fold more phosphorylated Hsp90␣ in samples prepared from the cell surface as compared with total cell lysates. In three independent experiments, the average -fold difference was 11.8 Ϯ 1.5. other non-phosphorylatable residues (either aspartic or glutamic acid) resulted in the same outcome (data not shown), suggesting that the inability to phosphorylate residue 89 is the basis for the translocation defect of the T89A mutant. The experiment shown in Fig. 6 was done with HG-treated cells; comparable results were obtained in cells cultured in normal glucose (supplemental materials Fig. S5). Taken together, these data indicate that HG promoted PKA-dependent phosphorylation of Hsp90␣ at Thr-89, and that this event was a prerequisite for translocation of Hsp90␣ to the cell surface.
Preventing Translocation of Hsp90␣ Preserved the Hsp90␣⅐eNOS Complex and NO Production-We used the T89A mutant to test if translocation of Hsp90␣ was important for dissociation of the HSp90␣⅐eNOS complex and reduced production of NO that was observed following treatment with HG (2,43). In cells expressing WT Hsp90␣, HG induced a 3.7 Ϯ 0.2-fold reduction in the amount of Hsp90␣ that coprecipitated with eNOS (Fig. 7A), and reduced NO production by 113 Ϯ 16% (Fig. 7B). These HG-induced changes were smaller in the T89A cells; only a 1.3 Ϯ 0.1-fold reduction in the amount of Hsp90␣ that coprecipitated with eNOS and a 23 Ϯ 7% decline in NO production. The residual, endogenous Hsp90␣ expressed in the T89A cells (Fig. 7A) might account for the fact that HG was still able to reduce the Hsp90␣⅐eNOS complex and NO production in the T89A cells. Taken together, these findings indicate that translocation of Hsp90␣ was responsible for the HG-induced disruption of the Hsp90␣⅐eNOS complex and the decline of NO production.
DM Increased Phosphorylation of Hsp90␣ and Decreased the Hsp90␣⅐eNOS Complex-Because HG promoted translocation of Hsp90␣ in a PKA-dependent manner, we tested whether DM induced similar changes. Lysates from the aortic endothelium of 5 controls or diabetic rats were harvested and pooled. Western blot analysis using the phosphospecific PKA substrate antibody showed that DM increased PKA phosphorylation of numerous substrates (Fig. 8A). We immunoprecipitated Hsp90␣ from these lysates and observed 2.2 Ϯ 0.2-fold greater phosphorylation of Hsp90␣ in DM versus control samples (Fig.  8B). Finally, we addressed if there was a decline in the amount of Hsp90␣⅐eNOS complex in response to DM. Indeed, 2.4 Ϯ 0.2fold less eNOS coprecipitated with Hsp90␣ from aortic lysates prepared from a pool of diabetic rats as compared with the non-diabetic controls (Fig. 8C). Thus DM, which induces translocation of Hsp90␣ (28), also reduced the association between Hsp90␣ and eNOS. We propose that one of the mechanisms by which HG and DM attenuates NO production is by inducing the translocation of Hsp90␣ to the cell surface.

DISCUSSION
In this study we provide evidence that HG increased PKAdependent phosphorylation of Hsp90␣, promoted its translocation out of the cell and thereby reduced the amount available to complex with eNOS. These events contributed to the HGinduced decline in production of NO. Similar events were induced by diabetes in the aortic endothelium of rats, suggesting that translocation of Hsp90␣ was a contributing factor to the decline in NO production in the diabetic state.
Our findings reveal the existence of a previously unappreciated mechanism to control eNOS, namely reducing the availability of eNOS enhancers. Other mechanisms to suppress eNOS activity include phosphorylation (44), oxidation (45), and disruption of conformation (46,47). The diverse layers of negative regulation of eNOS are consistent with the physiological requirement to rapidly regulate NO levels.
Stalker et al. (19) reported that DM increases calpain activity, and that blocking calpain relieves the DM-induced decline in NO. Furthermore, calpain inhibitors increase the amount of the eNOS⅐Hsp90 complex in diabetic rats. Our findings, together, with the results of Stalker et al. (19), raises the possibility that calpain activates PKA, which phosphorylates Hsp90␣ and thereby reduces its association with eNOS. Calpain has been FIGURE 6. Phosphorylation of Hsp90␣ at Thr-89 was required for its translocation. Expression vectors harboring 2 different siRNA oligos directed against porcine Hsp90␣ were stably expressed (ϩ) to knock down the level of endogenous Hsp90␣. Control cells (Ϫ) received an empty expression vector. The wild type (WT) or mutant (T89A) human Hsp90␣ cDNA was subsequently expressed in these cells. A, total cell lysates of the indicated cells were subjected to Western blot analysis using either anti-Hsp90␣ or anti-RasGAP antibodies. The siRNA reduced endogenous Hsp90␣ by 72%, and re-expression of either the WT or mutant Hsp90␣ restored expression to the control level. In three independent experiments, the average reduction of Hsp90␣ was 71.5 Ϯ 2.2%. B, the cells characterized in A were cultured in HG for 2 weeks and analyzed for sHsp90␣ as described in the legend of Fig. 5. In three independent experiments, sHsp90␣ was 21.1 Ϯ 1.8, 10.3 Ϯ 1.1, and 19.6 Ϯ 1.0, for the EV, T89A, and WT cells, respectively. The difference between the sHsp90␣ level in T89A and WT cells was found to be statistically different using unpaired t test. *, p Ͻ 0.05.  Fig. 6A, were cultured in 5 or 30 mmol/liter glucose for 2 weeks, the total cell lysates were immunoprecipitated (IP) using an anti-eNOS antibody, and subjected to Western blot analysis using either eNOS or Hsp90␣ antibodies. In three independent experiments, HG decreased the association in WT cells by an average of 3.7 Ϯ 0.2-fold, but in T89A cells only by an average of 1.3 Ϯ 0.1-fold. B, the conditioned media were collected from cells as described in A, and subjected to indirect NO detection as described under "Experimental Procedures." In three independent experiments, nitrite concentrations were 2.06 Ϯ 0.11 and 1.67 Ϯ 0.05 mol/liter for the T89A in normal and high glucose, respectively. A much larger difference was observed in WT cells, 1.36 Ϯ 0.04 and 0.64 Ϯ 0.05 mol/liter for normal and high glucose, respectively. The high glucoseinduced drop in NO was greater in WT versus T89A cells. This difference was statistically significant. *, p Ͻ 0.05. WB, Western blot.
reported to cleave and activate PKA (48). The mechanism by which HG activates PKA and potential involvement of calpain are exciting areas of future investigation.
The fact that translocation of a small fraction of Hsp90␣ is sufficient to impact NO production begs the question of why the Hsp90␣ that remains in the cell does not compensate. It is possible that the Hsp90␣ that is detected in a Western blot of total cell lysates originates from separate pools within the cell, and that these pools are not functionally interchangeable. For instance, the Hsp90␣ that complexes with eNOS and promotes its activation is separate from Hsp90␣ that is performing other cellular functions. Consistent with this idea is the observation that the HG-induced translocation of 9% of Hsp90␣ results in a dramatic decrease (3.7 Ϯ 0.2-fold) in the amount of Hsp90␣ associated with eNOS (Fig. 7A).
As mentioned above, only a minority of cells express sHsp90␣, which raises the question of how a minority can have profound impact on the overall response of the population. The question is based on the assumption that all of the cells are behaving identically, for instance, making an equivalent contribution to the level of NO in the culture medium. Thus one would expect that inhibiting 10% of the population would lead to no more than a 10% decline in the level of NO. Our observation that there is a larger decline may indicate that the population is not behaving homogeneously such that only a fraction of the cells contribute to NO production. A second possibility is that 100% of the cells are translocating Hsp90␣, but we can detect translocation of Hsp90␣ in only 10% of them at any one moment in time (when the cells are harvested for FACS analysis). The other cells may have shed the Hsp90␣ and/or deposited it into exosomes (see below). The NO levels and amount of Hsp90␣ in the eNOS⅐Hsp90␣ complex may better reflect the cumulative effect of Hsp90␣ translocation.
In this report we focused on the novel finding that PKA phosphorylates and translocates Hsp90␣. Hsp90␣ may not be the only member of the eNOS complex that is a PKA substrate. PKA can also phosphorylate eNOS (at Ser-1177), and thereby activate it (7). Whereas we have not investigated whether HG triggered PKA-dependent phosphorylation of eNOS at Ser-1177, our work, as well as the findings of other labs (2,49) demonstrate that HG reduces NO production. Thus if eNOS is phosphorylated and activated by PKA in response to HG (Fig. 1) or diabetes (Fig. 8A), then this event may attenuate the extent of the drop in eNOS output.
Our finding that phosphorylation of Hsp90␣ reduced its association with eNOS suggests that phosphorylation either prevents binding or promotes dissociation of the two proteins. The interaction of Hsp90␣ with some of its partners is regulated by phosphorylation. For instance, only the unphosphorylated form of Hsp90 is complexed with the reovirus cell attachment protein 1 (50) and the formation of the inactive form of hemin-controlled inhibitor is prevented when hsp90 is phosphorylated (51,52). Additional studies are required to elucidate whether phosphorylation of Hsp90␣ is regulating its association with eNOS.
Our studies do not address the intriguing question of how Hsp90␣ translocates outside the cell. Hsp90␣ does not contain a traditional signal peptide that would direct it to the classical trafficking pathway by which proteins are targeted to the cell surface. The fact that the majority of Hsp90␣ is cytoplasmic further supports the idea that there is an alternative pathway by which a subpopulation of Hsp90␣ is translocated to the cell membrane. Our working hypothesis for how Hsp90␣ translocates to the surface is that its phosphorylation at Thr-89 regulates its association with proteins that shuttle it to the cell surface.
The observation that Hsp90␣ is translocated to the surface of cells raises the exciting possibility that Hsp90␣ has extracellular functions. There are several reports indicating that this is indeed the case. Extracellular Hsp90␣ activates matrix metalloproteinases 2 and thereby promotes the invasiveness of tumor cell lines (29). Similarly, we reported that Hsp90␣ promoted the translocation of annexin II to the surface of endothelial cells, and these events enhanced activation of plasmin (28). Thus in two different experimental settings, extracellular Hsp90␣ facilitated the activation of proteases. We are tempted to speculate that the increased translocation of Hsp90␣ to the cell surface not only reduces eNOS activity, but also contributes to diabetes-enhanced matrix metalloproteinase activity related to plaque rupture in atherosclerosis (53-55). Identifying addi- FIGURE 8. DM increased phosphorylation of Hsp90␣ at a PKA consensus site and decreased the Hsp90␣⅐eNOS complex. A, lysates prepared from the aortic endothelium of five diabetic or age-matched control rats were pooled, and analyzed by Western blotting using a phosphospecific PKA substrate antibody (top panel) and a RasGAP antibody (lower panel). B, the lysates described in A were immunoprecipitated (IP) with a Hsp90␣ antibody or non-immune IgG, and the resulting samples were analyzed by Western blot using a phosphospecific PKA substrate antibody (top panel). The stripped membrane was reprobed with a Hsp90␣ antibody (bottom panel). Quantification of Western blots showed a 2.2 Ϯ 0.2-fold increase in the amount of phospho-Hsp90␣ in the diabetic endothelium in three independent experiments. C, the Western blot described in B was reprobed using an anti-eNOS antibody. DM reduced the association of eNOS with Hsp90␣ in the endothelium of the diabetic rat aorta by 2.4 Ϯ 0.2-fold compared with the controls. The differences in panels B and C were statistically significant.
tional functions for sHsp90␣ and their physiological impact awaits further investigation.
Whereas this study did not address the fate of Hsp90␣ once it is translocated to the cell surface, we and others have considered the possibility that it is excreted. Oxidative stress triggers the release of Hsp90␣ from vascular smooth muscle cells (56). Furthermore, Hsp90␣ is one of the proteins that are found in exosomes, vesicles that are excreted from cells and accumulate in the medium of cells (29,57,58). In both normal and high glucose-treated cells we have found that Hsp90␣ accumulates in exosomes. 3 Thus translocation of Hsp90␣ may be an intermediate step in its exosomes-related export out of the cell.