Nitric Oxide-Induced Activation of the AMP-Activated Protein Kinase α2 Subunit Attenuates IκB Kinase Activity and Inflammatory Responses in Endothelial Cells

Background In endothelial cells, activation of the AMP-activated protein kinase (AMPK) has been linked with anti-inflammatory actions but the events downstream of kinase activation are not well understood. Here, we addressed the effects of AMPK activation/deletion on the activation of NFκB and determined whether the AMPK could contribute to the anti-inflammatory actions of nitric oxide (NO). Methodology/Principal Findings Overexpression of a dominant negative AMPKα2 mutant in tumor necrosis factor-α-stimulated human endothelial cells resulted in increased NFκB activity, E-selectin expression and monocyte adhesion. In endothelial cells from AMPKα2-/- mice the interleukin (IL)-1β induced expression of E-selectin was significantly increased. DETA-NO activated the AMPK and attenuated NFκB activation/E-selectin expression, effects not observed in human endothelial cells in the presence of the dominant negative AMPK, or in endothelial cells from AMPKα2-/- mice. Mechanistically, overexpression of constitutively active AMPK decreased the phosphorylation of IκB and p65, indicating a link between AMPK and the IκB kinase (IKK). Indeed, IKK (more specifically residues Ser177 and Ser181) was found to be a direct substrate of AMPKα2 in vitro. The hyper-phosphorylation of the IKK, which is known to result in its inhibition, was also apparent in endothelial cells from AMPKα2+/+ versus AMPKα2-/- mice. Conclusions These results demonstrate that the IKK is a direct substrate of AMPKα2 and that its phosphorylation on Ser177 and Ser181 results in the inhibition of the kinase and decreased NFκB activation. Moreover, as NO potently activates AMPK in endothelial cells, a portion of the anti-inflammatory effects of NO are mediated by AMPK.


Introduction
The AMP-activated protein kinase (AMPK) is a member of the Snf1/AMPK family of serine/threonine protein kinases and is an evolutionarily conserved sensor of the cellular energy status. Although the AMPK pathway is traditionally thought of as an intracellular fuel gauge and regulator of metabolism, recent evidence indicates that it may also be important for the maintenance of endothelial function and to redress the disturbed redox balance associated with vascular disease. Certainly, the AMPK can influence a number of signaling cascades that would be expected to result in anti-atherosclerotic effects, such as attenuated free radical generation and the activation of angiogenic factors (for review see [1]).
Although the link between cellular metabolism and AMPK activation has been repeatedly demonstrated in tissues such as skeletal and cardiac muscle [2], the precise role played by the AMPK in endothelial cell remains incompletely understood. Indeed, while there are some situations in which activation of the AMPK is reported to depend on an increase in the ADP/ATP ratio e.g. following cell stimulation with rosiglitazone [3], the activation of AMPK by Ca 2+ -elevating agonists such as bradykinin [4,5] and thrombin [6] has been attributed to the activity of an upstream activating kinase rather than to changes in AMP levels. There are two different isoforms of the catalytic a AMPK subunit (a1 and a2) that are differentially expressed in different tissues. For example, while the a1 isoform predominates in adipose tissue, skeletal muscle and cardiomyocytes express higher amounts of the AMPKa2 [7]. Interestingly, endothelial cells express both a subunits and different groups report the predominance of different isoforms, a finding that may explain the inconsistent dependence on changes in ADP/ATP for stimulation . We reported previously that the AMPK can be activated by fluid shear stress as well as by NO in endothelial cells, and that it can affect the expression of endothelial cell proteins including, the hydroxy-methylglutaryl coenzyme A reductase, cytochrome P450 2C8, and angiopoietin 2 [8][9][10][11]. Also the overexpression of dominant negative AMPKa2 in endothelial cells increases basal and tumor necrosis factor (TNF)-a-stimulated E-selectin expression [10]. While the latter findings imply the involvement of the transcription factor nuclear factor kB (NFkB) and there are reports of an attenuated NFkB activation following AMPK activation in different cell types [12][13][14][15], the molecular mechanisms involved are not clear. Therefore, the aim of the present study was to address the link between AMPK activation and NFkB inhibition as well as to determine whether or not the activation of the AMPK could at least partially account for the effects of NO on NFkB activity and thus adhesion molecule expression.

Effect of NO on the activation of AMPK and NFkB
Treatment of primary cultures of human endothelial cells with the NO donor DETA-NO (100 mmol/L) which has a t K of 16 hours, elicited the time-dependent phosphorylation of the AMPK on Thr172 ( Figure 1A). The phosphorylation of AMPK by exogenous NO was independent of the donor used as a substance with a markedly faster NO releasing kinetic i.e., DEA-NO, t K 2 minutes, resulted in the more rapid activation of the AMPK i.e., within 2 minutes ( Figure S1A). The effects were also concentration-dependent as indicated using a third NO donor with a more delayed NO release (DPTA NO, t K 5 hours; Figure  S1B). TNF-a (1 and 10 ng/ml, 30 min) elicited a marked and concentration-dependent increase in NFkB activity (EMSA; Figure 1B), that was attenuated by endothelial cell pretreatment with DETA-NO ( Figure 1B). TNF-a had however no acute effect on the activation of the AMPK in the absence of NO ( Figure 1C).

Effect of constitutively active and dominant negative AMPK mutant on NFkB activation
To determine the involvement of the AMPK in the prevention of NFkB activation, we assessed the effects of constitutively active and dominant negative AMPKa2 mutants on the TNF-a-induced activation of NFkB. While TNF-a enhanced the activity of NFkB in cells infected with a control virus, this response was blunted in cells overexpressing the constitutively active AMPK ( Figure 2A) and potentiated in cells expressing the dominant negative AMPK mutant ( Figure 2B). To estimate the influence of the AMPK in the NO-mediated inhibition of TNF-a-induced NFkB activation, cells overexpressing either the control virus or the dominant negative AMPK mutant were exposed to TNF-a in the absence and presence of DETA-NO. As before, pre-treatment with the NO donor attenuated the activation of NFkB ( Figure 2C). The effect of NO was however blunted in cells overexpressing the dominant negative AMPK mutant.

Effect of AMPK deletion on E-selectin and VCAM-1 expression in murine endothelial cells
Theoretically a dominant negative AMPKa2 mutant could affect the activity of the AMPKa1 by, for example, sequestering AMPKb and c subunits from the endogenous a1 subunit. To ensure that the effects observed could indeed be attributed to the AMPKa2, we performed additional experiments in endothelial cells from AMPKa2 -/mice or their wild-type littermates (AMPKa2 +/+ ).
First, to determine whether or not the loss of one AMPKa subunit could result in the compensatory up regulation of the other we assessed AMPKa1 expression in aortae from a2 +/+ and a2 -/mice as well as AMPKa2 expression in a1 +/+ and a1 -/mice. We found that the deletion of the AMPKa2 did not affect a1 expression while the a2 subunit was upregulated in tissue from AMPKa1 -/mice ( Figure S2).
To determine the effect of AMPK deletion on the response to inflammatory mediators, murine microvascular endothelial cells from AMPKa2 +/+ or AMPKa2 -/mice were exposed to IL-1b (10 ng/mL, 6 hours) or to LPS (10 ng/mL, 6 hours) in the absence and presence of DETA-NO and the surface expression of E-selectin was assessed. The basal expression of E-selectin was slightly (but not significantly) enhanced in cells isolated from the AMPKa2 -/compared to the AMPKa2 +/+ animals. While the basal expression of the adhesion molecule was attenuated by NO, levels remained slightly elevated in the AMPKa2 -/cells. Cell stimulation with IL-1b ( Figure 3A) or LPS ( Figure 3B) increased the expression of E-selectin with significantly higher expression levels being detected in AMPKa2 -/cells. TNF-a failed to result in the activation of NFkB in either AMPKa2 +/+ or AMPKa2 -/cells studied. Pre-treatment of endothelial cells with the NO donor attenuated adhesion molecule expression but again expression was higher (2.561.0-fold for IL-1b and 1.760.4-fold for LPS) in the cells from AMPKa2 -/mice than in cells from their wild-type littermates. Effect of constitutively active and dominant negative AMPK mutants on cell adhesion Next, we studied the consequences of AMPK mutant overexpression on the adhesion of mononuclear cells to endothelial cells. Consistent with the results obtained using the AMPKa2 -/murine endothelial cells, significantly more mononuclear cells attached firmly to human endothelial cells overexpressing the dominant negative AMPKa2 mutant than those expressing the constitutively active AMPK or treated with the control virus. Cell stimulation with TNF-a (1 ng/mL, 6 hours) resulted in a significant increase in mononuclear cell attachment ( Figure 4) that was potentiated in cells expressing the dominant negative, but attenuated in cells expressing the constitutively active AMPK mutants.

Consequences of AMPK activation/inactivation on IkB
The activation of NFkB mainly occurs via the phosphorylation and degradation of inhibitory molecules, including IkB. Interestingly, the activation of the AMPK was associated with the increased expression of IkB protein, a phenomenon that was evident on comparing pulmonary endothelial cells from AMPKa2 +/+ and AMPKa2 -/mice ( Figure 5A) or COS-7 cells overexpressing the constitutively active or dominant negative AMPKa2 mutants ( Figure 5B). Both these findings support the above evidence indicating that activation of the AMPK results in NFkB inhibition. No difference in IkB expression was detected between AMPKa1 +/+ and AMPKa1 -/endothelial cells (data not shown).  COS-7 cells expressing constitutively active or dominant negative AMPKa2 were used to determine the consequences of AMPK activation and inhibition on the IkB kinase (IKK)mediated phosphorylation of IkB independent of potential interference by endogenously generated NO. Neither of the AMPKa2 mutants studied affected the basal IkB phosphorylation. However, following stimulation with TNF-a (10 ng/mL), IkB phosphorylation was enhanced in cells expressing the dominant negative mutant ( Figure 5C). The lack of effect of the constitutively active AMPK mutant in these cells can most probably be attributed to the high expression level and activity of the endogenous AMPKa isoforms in this cell line.
The TNF-a-activated IKK complex also phosphorylates p65 on Ser536, a step thought to be required for enhanced p65 transactivation potential and the optimal induction of NFkB target genes [16]. The AMPK also seems to play a role in this pathway as we observed that the expression of the constitutively active AMPK led to a decrease in the TNF-a-stimulated phosphorylation of p65 ( Figure S3). Similar results were obtained using human endothelial over expressing the AMPKa2 but not the AMPKa1 subunit (data not shown).

Consequences of AMPK activation/inactivation on IKK
Given that AMPK activation attenuated the phosphorylation of IkB as well as p65 we proposed that the AMPK is able to attenuate NFkB activation by phosphorylating and inhibiting the IKK. In in vitro kinase assays using the purified AMPKa2 (together with b1 and c1) we identified IKKb as an AMPK substrate ( Figure 6A).
To determine the site(s) in IKKb phosphorylated by AMPK we replaced its major phosphorylation sites i.e. Ser177 and Ser181 with alanine by site-directed mutagenesis, and overexpressed these mutants in COS-7 cells. The wild-type and mutant IKKb were then immunoprecipitated and incubated with AMPKa2. We found that the mutation of IKKb on Ser177 and Ser181 resulted in a lower level of phosphorylation than the wild-type enzyme. Moreover, phosphorylation was barely detectable in the S177A/ S181A double mutant. When IKKb was replaced with GFP, no phosphorylation was observed in the presence of AMPKa2 ( Figure 6B). Also in COS-7 cells overexpressing a constitutively active AMPKa2 mutant the phosphorylation of IKK (the antibody used recognizes phospho Ser177 and 181) was increased compared to that detected in cells treated with a control virus ( Figure 6C). A similar, approximately 2-fold increase in pIKK levels was observed in murine endothelial cells stimulated with IL-1b (30 ng/mL, Figure 6D) in that the phosphorylation of IKK was consistently greater in cells from AMPKa2 +/+ versus a2 -/mice. Signals from solvent-treated cells were too low to quantify. There was no difference in the IL-1b -induced phosphorylation in AMPKa1 +/+ and AMPKa1 -/mice ( Figure S4).
To assess IKKb activity we next determined the phosphorylation of GST-IkBa in the presence of wild-type IKKb and either AMPKa1 or a2. We found that the phosphorylation of IkBa was reduced in the presence of AMPKa2 but not AMPKa1 ( Figure 6E).

Discussion
The results of the present investigation indicate that the AMPKa2 subunit plays an important role in regulating inflammatory responses, adhesion molecule expression (E-selectin and VCAM-1) and monocyte adherence to endothelial cells. These effects could be related, at least partly to the AMPK-mediated phosphorylation of IKK and subsequent inhibition of IkB and p65 phosphorylation as well as DNA binding (see Figure 7). Moreover, three different NO donors were able to activate the AMPK and the NO-mediated inhibition of NFkB activation was attenuated by a dominant negative AMPK in human endothelial cells as well as in endothelial cells from AMPKa2 -/mice. Thus, it appears that the NO-mediated inhibition of NFkB activity is, at least partially, dependent on the activation of the AMPKa2 subunit.
The possibility that the AMPK can regulate the DNA binding activity of NFkB was initially indicated by the fact that overexpression of a dominant negative AMPK increased, while a constitutively active AMPK decreased the TNF-a-induced expression of E-selectin [10]. There have since been reports linking AMPK with NFkB in endothelial cells, but the overall outcome of kinase activation is controversial as both NFkB inhibition [12,[17][18][19] and activation [20][21][22][23] have been reported. Whether or not this discrepancy can be attributed to the parallel activation/ inactivation of associated regulatory mechanisms and signal  transduction pathways, remains to be determined. One important consideration is that much of the published data was generated using the AMPK activator, 5-aminoimidazole-4-caboxymide-1-b-D-ribofuranoside (AICAR), which can elicit AMPK-independent effects [24], and substances such as compound C and iodotubercidin which are by no means specific inhibitors of the AMPK. Therefore, one focus of the current investigation was to make use of the available constitutively active and dominant negative AMPK mutants and endothelial cells from AMPKa2 +/+ and AMPKa2 -/mice to determine which of the catalytic subunits was most likely to affect transcription factor activity. The results obtained using murine lung endothelial cells indicate that the AMPKa2 plays a prominent role in modulating NFkB phosphorylation in endothelial cells.
Which point in the NFkB activation cascade could be affected by AMPK? NFkB is a dimer consisting of the transcription factors p65 (RelA) or p50, and under resting conditions it is associated with its inhibitory protein IkB which retains the complex in the cytosol. The activation of NFkB mainly occurs via the phosphorylation of IkB, which results in its degradation -leaving the NFkB dimer free to translocate into the nucleus and stimulate transcription [16]. Using dominant negative and constitutively active AMPK mutants we found that the phosphorylation of IkB (residue Ser32 or Ser36) was increased when the AMPK was inhibited but decreased when AMPK activity was increased. We also observed the attenuated basal expression of IkB in the AMPKa2 -/versus AMPKa2 +/+ endothelial cells. As the expression of IkB is known to be regulated by the activity of NFkB [25], these data further highlight the importance of the AMPKa2 subunit in the regulation of transcription factor activity.
The serine phosphorylation of IkB is mediated by a large multiunit complex containing two catalytic subunits (IKKa and IKKb) as well as the regulatory subunit IKKc or NEMO, which has no kinase domain. The p65 subunit is also a IKK substrate and p65 phosphorylation on Ser536 is thought to be required for enhanced p65 transactivation potential and the optimal induction of NFkB target genes [16]. Given that we observed that both AMPK activation and a constitutively active AMPK mutant decreased the phosphorylation of IkB and p65, it seemed logical to conclude that the AMPK is able to modulate the activity of the IKK complex.
To date there has only been circumstantial evidence to suggest a role of the AMPK in the modulation of IKK activity. For example, AICAR has been reported to be without effect on IKK in rat skeletal muscle cells [26] but to attenuate IkB phosphorylation in IL-18-stimulated cardiac microvascular endothelial cells [27]. Moreover, the overexpression of a dominant-negative AMPK in TNF-a-stimulated mouse macrophages resulted in the accelerated and exaggerated degradation of IkB [28]. Similarly, during the preparation of this manuscript the loss of AMPK activity was reported to result in increased IkBa degradation in cultured endothelial cells [23], a finding we have confirmed in native endothelial cells. The results of the present study clearly indicate that the AMPK can directly phosphorylate and attenuate the activity of the IKK in native and cultured endothelial cells. In in vitro assays we identified IKKb as an AMPK substrate and were able to detect the elevated phosphorylation of IKK in intact AMPKoverexpressing cells. The phosphorylation of IKKb on Ser 177 and/or 181 causes the rapid phosphorylation of a C-terminal serine cluster which in turn elicits the auto-inhibition of kinase activity and is thought to be an effective means of limiting the duration of IKK activation [29]. Our data suggest that the AMPKa2 is able to phosphorylate both of these serine residues as mutation of a single site reduced phosphorylation which was only barely detectable in a double S177A/S181A mutant. That the AMPK-mediated phos-phorylation of the IKK most likely results in NFkB inhibition was demonstrated by the fact that the IKK-dependent phosphorylation of a IkB-GST construct was markedly attenuated in the presence of the AMPKa2 but not the AMPKa1 subunit. Moreover, the expression of IkB was significantly higher in human endothelial cells overexpressing the constitutively active AMPKa2 and in AMPKa2 +/+ versus AMPKa2 -/mouse endothelial cells.
The AMPK has been implicated in the phosphorylation and activation of the endothelial NO synthase (eNOS), at least in vitro, and thus one further mechanism by which the AMPK could affect NFkB would be by modulating basal NO output (reviewed in [1]); which is known to decrease NFkB activity [30]. However, using COS-7 cells which do not express endogenous eNOS, we found that AMPK activation and the constitutively active AMPK were sufficient to attenuate the TNF-a-induced phosphorylation of IkB and p65. Thus in our hands, the AMPK-dependent inhibition of NFkB was not dependent on NO production. In fact, we were able to confirm that NO donors are effective activators of the AMPK a finding that fits well with previous reports in endothelial cells that AMPK activation is markedly attenuated in the presence of a NOS inhibitor or in cells from eNOS -/mice [5,9]. How NO activates the AMPK remains to be elucidated but may well be related to the activation of the Ca 2+ /calmodulin-dependent protein kinase kinase b, a well known AMPK kinase [5].
Taken together, the results of the present investigation indicate that the activation of the AMPK in endothelial cells can limit inflammatory responses via the phosphorylation and inhibition of IKK activity. Moreover, as the AMPK could be activated by NO, the AMPK dependent inhibition of IKK activity may contribute to the anti-inflammatory and anti-atherosclerotic actions of the endothelium-derived autacoid.

Cell culture
Human umbilical vein endothelial cells were isolated and cultured as described [31]. The use of human material in this study conforms to the principles outlined in the Declaration of Helsinki [32] and was confirmed in written form by the members of the Ethic-Commission of the Medical Faculty of the Goethe-University (Frankfurt am Main, Germany) and the donors gave verbal consent. Murine lung endothelial cells were isolated and cultured as described [8], from AMPKa1 -/or AMPKa2 -/mice or the respective littermate wild-type animals [33,34]  For the isolation of the pulmonary endothelial cells, mice were sacrificed using 4% isoflurane in air and subsequent exsanguination. COS-7 cells were purchased from the American Tissue Culture Collection (LCG Standards, Wesel, Germany) and cultured in Minimal Essential Medium (Invitrogen, Karlsruhe, Germany) supplemented with 8% fetal calf serum, pyruvate, non essential amino acids and gentamycin.

Adenoviral transduction of endothelial cells
Subconfluent endothelial cells (1 st passage) were infected with adenoviruses (provided by Ken Walsh, Boston and Benoit Viollet, Paris, France) to over-express constitutively active AMPKa2 [35], or dominant-negative AMPKa2 [36], as described [10]. As the viral backbones of the constitutive active and dominant negative viruses were not identical, the observed effects were always analyzed to the respective control virus.

Immunoblotting
Protein samples were heated with SDS-PAGE sample buffer and separated by SDS-PAGE as described [8]. Proteins were detected using their respective antibodies and enhanced chemiluminescence using a commercially available kit (Amersham, Germany). To assess the phosphorylation of proteins, either equal amounts of protein from each sample were loaded twice and one membrane incubated with the phospho-specific antibody and the other with an antibody recognizing total protein, or blots were reprobed with the appropriate antibody.

Electrophoretic Mobility Shift Assay (EMSA)
Nuclear and cytosolic proteins were isolated, and binding to c[ 32 P]-ATP (Hartmann Analytic, Braunschweig, Germany)-labeled double-stranded oligonucleotides containing the consensus sequence of the binding site for transcription factor NFkB (5-AGT TGA GGG GAC TTT CCC AGG C-3, Santa Cruz) was assessed as described [37].

Flow cytometry
Following stimulation endothelial cells were harvested using accutase (PAA laboratories, Coelbe, Germany) and washed twice with PBS. After blocking (1% BSA, 15 minutes, 4uC) cells were incubated with the specific conjugated antibodies (30 minutes, 4uC), washed twice with PBS and fixed in 1% paraformaldehyde. Antibody binding was analyzed using a flow cytometer (FACSCalibur, BD) by counting 20,000 cells per sample. The values are presented following subtraction of the isotype-matched, control IgG.

Adhesion assay
Peripheral blood mononuclear cells were freshly isolated from blood obtained from healthy volunteers using a BiocollH (1.077 g/ ml) gradient (Biochrom AG, Berlin, Germany) according to the manufacturer's protocol. Confluent cultures of human endothelial cells were stimulated with TNF-a for 6 hours, and were washed with culture medium before freshly isolated mononuclear cells (400,000 per 12 well) were added. After 10 minutes of incubation the non-adherent cells were removed the number of firmly adherent cells was quantified.

IKKb kinase assays
The phosphorylation of IKKb by AMPKa2 and the activity of IKK b were assessed in in vitro kinase assays. Ser177 and Ser181 in a flag-tagged IKK plasmid [38] (Addgene, Cambridge, MA) were mutated to alanine by site-directed mutagenesis (Quick exchange Kit, Agilent, Böblingen, Germany) using specific oligonucleotides (Biospring, Frankfurt, Germany). COS-7 cells were transfected with a plasmid encoding either the wild-type flagtagged IKKb or one of the S177A, S181A or S177A/181A mutants, which were then immunoprecipitated using a Flag antibody (Invitrogen, Karlsruhe, Germany). The AMPKa2 subunit was also immunoprecipitated from COS-7 cells and the in vitro kinase reaction performed as described [9]. In some experiments the activity of the IKKb was detected by monitoring the phosphorylation of a GST-IkB fusion protein (kindly provided by Fumiyo Ikeda, Frankfurt).

Statistical analysis
Values are expressed as the mean 6 SEM and statistical evaluation was performed using Student's t test for unpaired data and one-way ANOVA or ANOVA for repeated measures followed by followed by a Bonferroni t test where appropriate. Values of P,0.05 were considered statistically significant.