Cardiovascular Protective Effect of Metformin and Telmisartan: Reduction of PARP1 Activity via the AMPK-PARP1 Cascade

Hyperglycemia and hypertension impair endothelial function in part through oxidative stress-activated poly (ADP-ribose) polymerase 1 (PARP1). Biguanides and angiotensin II receptor blockers (ARBs) such as metformin and telmisartan have a vascular protective effect. We used cultured vascular endothelial cells (ECs), diabetic and hypertensive rodent models, and AMPKα2-knockout mice to investigate whether metformin and telmisartan have a beneficial effect on the endothelium via AMP-activated protein kinase (AMPK) phosphorylation of PARP1 and thus inhibition of PARP1 activity. The results showed that metformin and telmisartan, but not glipizide and metoprolol, activated AMPK, which phosphorylated PARP1 Ser-177 in cultured ECs and the vascular wall of rodent models. Experiments using phosphorylated/de-phosphorylated PARP1 mutants show that AMPK phosphorylation of PARP1 leads to decreased PARP1 activity and attenuated protein poly(ADP-ribosyl)ation (PARylation), but increased endothelial nitric oxide synthase (eNOS) activity and silent mating type information regulation 2 homolog 1 (SIRT1) expression. Taken together, the data presented here suggest biguanides and ARBs have a beneficial effect on the vasculature by the cascade of AMPK phosphorylation of PARP1 to inhibit PARP1 activity and protein PARylation in ECs, thereby mitigating endothelial dysfunction.


Introduction
Endothelial function is impaired under pathophysiological conditions such as hyperglycemia and hypertension, in part because of an imbalanced redox state. Induced by oxidative stress, poly (ADP-ribose) polymerase 1 (PARP1) plays an important role in DNA repair and

Cell culture, adenovirus, and EC infection and transfection
Human umbilical vein endothelial cells (HUVECs) were isolated from human umbilical cords. With patients' consensus, the cords were obtained from Department of Obstetrics and Gynecology, the First Affiliated Hospital, Xi'an Jiaotong University. The protocol for the isolation of endothelial cells was reviewed and approved by the Ethic Committee of Xi'an Jiaotong University Health Science Center. HUVECs were cultured in medium M199 supplemented with 20% fetal bovine serum (FBS), 1 ng/mL recombinant human fibroblast growth factor, 90 μg/mL heparin, 20 mM HEPES (pH 7.4) and 100 U/mL penicillin-streptomycin. Bovine aortic endothelial cells (BAECs) were a gift from Department of Physiology, Peking University. BAECs were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and antibiotics. Cells were maintained in a humidified 95% air, 5% CO 2 incubator at 37°C. We used a recombinant adenovirus expressing a constitutively-activated form of AMPKα2, hereafter called Ad-AMPK-CA. The parental adenoviral vector, called Ad-null, was used as an infection control. Confluent ECs were infected with recombinant adenoviruses at the indicated multiplicity of infection (MOI) and incubated for 24 hr before experiments. ECs were transfected with various DNA plasmids with the use of Lipofectamine 2000 RNAi Max (Invitrogen).
RT-qPCR analysis of mRNA RNA was isolated from cultured cells or tissues by using TRIzol (Invitrogen). For mRNA quantification, total RNA was reverse-transcribed (RT) by use of the iScript cDNA synthesis kit (Invitrogen), followed by quantitative real-time PCR (qPCR) with SYBR Green (Promega) and a 7500 realtime PCR system (Applied Biosystems). The relative level of mRNA was calculated by the ΔΔ Ct method with GAPDH or other appropriate genes as an internal control.

Western blot analysis
Total proteins were extracted by use of RIPA buffer (6.5 mM Tris, pH 7.4, 15 mM NaCl, 1 mM EDTA, 0.1% SDS, 0.25% sodium deoxycholate, 1% NP-40). Bicinchoninic Acid reagents (Thermo Scientific) were used to measure the protein concentration. Equal amounts of proteins were separated by SDS-PAGE and transferred to PVDF membranes. The blots were immunoreacted with primary antibodies and secondary antibodies conjugated with horseradish peroxidase. Protein bands were visualized by enhanced chemiluminescence detection and the intensity was quantified by use of Scion Image software.

PARP1 activity measurement and NO generation detection
For the colorimetric PARP1 activity assay, EC nuclear extracts were incubated with histones and PARP cocktail in a strip well format. The PARP1 activity was then assessed by incorporation with biotinylated poly(ADP-ribose), which was quantified by the reading the absorbance at 450 nm. Serially diluted PARP-HSA (high specific activity) standards were used in parallel to generate the standard curve for subsequent calculation. EC-derived NO was measured in condition media by using Griess reagents following the standard protocol.

Animal experiments
The animal experiments were approved by the Institutional Animal Care and Use Committee of Xi'an Jiaotong University (No. XJTULAC2014-208). All animals were housed in colony cages (Animal Care Systems, Centennial, Colorado, USA) with a 12-hr light/12-hr dark cycle ad libitum. We monitored the mouse and rat body weight, teeth, fur, and behavior on the daily base. According to our close observation, there was no mortality during the experimental period. At the end of experiments, all animals were euthanized by CO 2 and tissues were collected afterwards for analyses. All treatments were by oral gavage. C57BL6 mice were purchased from Experimental Animal Center of Xi'an Jiaotong University. Eight-to twelve-weekold male C57BL6 mice were treated with metformin (200 mg/kg/day), glipizide (1.3 mg/kg/ day), telmisartan (10 mg/kg/day), metoprolol (30 mg/kg/day), or saline as a mock control for up to 24 hr. AMPKα2 -/mice were a gift from Institute of Vascular Medicine, Peking University Third Hospital. AMPKα2 -/mice and their wild-type littermates (AMPKα2 +/+ ) in a C57BL6 background were treated with metformin (200 mg/kg/day) for 12 hr. Ten-week-old db/db and age-matched db/m mice were purchased from the Model Animal Research Center of Nanjing University. db/db mice were treated with metformin (200 mg/kg/day) or glipizide (1.3 mg/kg/day) for 2 weeks. Fasting blood glucose was measured by use of the Accu-Check Glucose Meter. Spontaneously hypertensive rats (SHRs) and Wistar-Kyoto (WKY) rats were purchased from Vital River Laboratory Animal Technology (Beijing, China). Ten-week-old male SHRs and WKY rats were fed standard chow diet and treated with telmisartan (10 mg/kg/ day) or metoprolol (30 mg/kg/day) for 8 weeks. Systolic arterial pressure was measured by tailcuff plethysmography.

Statistical Analysis
Significance of variability was determined by unpaired Student's t test between two groups or ANOVA for multiple comparisons. Data were expressed as mean±SD from at least 3 independent experiments. p<0.05 was considered statistically significant.

High glucose and Ang II decrease PARP1 Ser-177 phosphorylation in ECs
Given that AMPK activity is inhibited under hyperglycemia and a high level of Ang II, both known to augment oxidative stress in ECs [21,22], we first sought to test whether phosphorylation of PARP1 Ser-177 was also decreased under these conditions with the use of the newly developed anti-phospho-PARP1. As expected, H 2 O 2 , high glucose and Ang II dose-and timedependently attenuated the phosphorylation of AMPKα Thr-172 and PARP1 Ser-177 in HUVECs (Fig 1). Notably, the use of 30 mM mannitol did not decrease PARP1 and AMPK phosphorylation as that of glucose did (S1 Fig). Further, although H 2 O 2 at a lower concentration (1 μM) slightly induced AMPK and PARP1 phosphorylation (Fig 1A), these phosphorylation events were decreased at higher concentrations (e.g., 100 and 1000 μM).

Metformin and telmisartan increase PARP1 Ser-177 phosphorylation
Although metformin and glipizide are first-line drugs to treat type II diabetes, metformin, but not glipizide, has an additional effect on vasculature [23]. Therefore, we examined whether metformin and glipizide could induce PARP1 phosphorylation in HUVECs. Metformin treatment indeed dose-and time-dependently increased the phosphorylation level of PARP1 Ser-177 and AMPK Thr-172 in cultured HUVECs (Fig 2A and 2E). However, cells treated with glipizide showed little increase in PARP1 and AMPK phosphorylation (Fig 2B and 2F). Regarding anti-hypertensive medication, telmisartan improves EC functions, but metoprolol does not seem to benefit the endothelium [17,18]. Similarly, telmisartan, but not metoprolol, induced PARP1 and AMPK phosphorylation (Fig 2C, 2D, 2G and 2H). To recapitulate findings from experiments with cultured HUVECs, we treated C57BL6 mice with metformin, glipizide, telmisartan, and metoprolol. Compared with aortas of control mice receiving saline, aortas of mice with metformin or telmisartan treatment showed increased PARP1 Ser-177 and AMPK Thr-172 phosphorylation (Fig 3A and 3C). However, glipizide and metoprolol administration did not significantly increase the phosphorylation of PARP1 or AMPK (Fig 3B and 3D). Together, results presented in Figs 1-3 suggest that phosphorylation of PARP1 Ser-177 agreed with AMPK activity in ECs under pathophysiological conditions (i.e., elevated oxidative stress) or with pharmacological interventions (i.e., metformin and telmisartan).

PARP1 Ser-177 Phosphorylation is AMPK-dependent
Because PARP1 Ser-177 is a putative AMPK phosphorylation site [20], we investigated whether PARP1 phosphorylation in HUVECs is AMPK-dependent. HUVECs were infected with Ad-AMPK-CA (a constitutively active form for AMPKα2) at different multiplicity of infection (MOI) and control cells were infected with Ad-null at 50 MOI. Ectopic expression of a constitutively active form of AMPKα2 increased the phosphorylation of AMPK Thr-172 and that of PARP1 Ser-177 ( Fig 4A). Furthermore, activation of endogenous AMPK by 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) increased phosphorylation of PARP1 and AMPK in HUVECs under high glucose and Ang II (Fig 4B and 4C). In a complementary experiment (Fig 4D), the increased phosphorylation of AMPK and PARP1 in response to metformin was attenuated by pretreatment with the AMPK inhibitor Compound C.

PARP1 Ser-177 phosphorylation affects endothelial function
Next, we examined the role of AMPK phosphorylation of PARP1 Ser-177 in modulating PARP1 activity and related EC function. Because both high glucose and Ang II can activate PARP1 and cause endothelial dysfunction [24,25], we cultured HUVECs under high glucose or Ang II with or without AICAR to examine whether activation of AMPK can inhibit PARP1 activation and consequent PARylation. With 30 mM glucose or Ang II treatment, the protein level of PAR was increased in HUVECs as compared with respective controls (Fig 5A and 5B). Co-incubation with AICAR significantly reduced the high glucose-or Ang II-induced protein  PARylation (Fig 5A and 5B). We then compared the effect of metformin versus glipizide and telmisartan versus metoprolol on protein PARylation. Metformin reduced PARylation, and glipizide had little effect on HUVECs under 30 mM glucose (Fig 5C). Similarly, telmisartan, but not metoprolol, decreased Ang II-increased PARylation (Fig 5D). These results are consistent with PARP1 activity assay (S2 Fig). To confirm further whether AMPK phosphorylation of PARP1 Ser-177 inhibited PARP1 activity and PARylation, we transfected BAECs with plasmids encoding the wild-type, dephospho-mimetic S177A PARP1, or phospho-mimetic S177D PARP1. High glucose or Ang II increased protein PARylation in BAECs transfected with wild-type PARP1, and overexpression of PARP1 S177A increased PARylation both at the basal level and with glucose or Ang II were treated with or without metformin or glipizide for 6 hr, then incubated with or without 30 mM glucose for 24 hr. (D) HUVECs were treated with or without telmisartan or metoprolol for 6 hr, then incubated with or without 100 nM Ang II for 24 hr. (E, F) BAECs were transfected with flag-tagged wild-type (WT), S177A, or S177D PARP1 plasmids for 24 hr, then incubated with 30 mM glucose (E) or 100 nM Ang II (F) for 6 hr. (G) BAECs were transfected with S177A or S177D PARP1 plasmid. RT-PCR analysis of mRNA level of eNOS, SIRT1, KLF4, MCP-1, and VCAM-1. (H, I) BAECs were transfected with S177A or S177D PARP1 or WT PARP1 treated with or without PJ-34 ( (Fig 5E and 5F). In contrast, protein PARylation was lower with PARP1 S177D than S177A or wild-type transfection at the basal level and under glucose or Ang II stimulation. Given the strong inhibition of protein PARylation by PARP1 S177D, we further examined its effect on the expression of genes critical to BAEC function. As expected, the mRNA level of eNOS, SIRT1 and Krüppel-like factor 4 (KLF4) was higher and that of monocyte chemoattractant protein-1 (MCP-1) and vascular cell adhesion molecule-1 (VCAM-1) was lower in BAECs with PARP1 S177D than S177A overexpression (Fig 5G). We also compared p-eNOS Ser-1177 and SIRT1 protein level in BAECs transfected with PARP1 S177A or S177D mutants. In a parallel experiment, wild-type PARP1-transfected BAECs were treated with PJ-34, a PARP1 pharmacological inhibitor. The effect of PARP1 S177D overexpression was similar with PJ-34 treatment, both of which decreased PARP1 activity and protein PARylation but increased eNOS Ser-1177 phosphorylation, SIRT1 expression level, and NO bioavailability (Fig 5H and  5I). Thus, AMPK phosphorylation of PARP1 Ser-177 inhibits PARP1 activity and hence improves endothelial function.

Metformin and telmisartan activate the AMPK-PARP1 cascade in diabetic and hypertensive animal models
To provide in vivo evidence for the AMPK-PARP1 cascade and its downstream function, we used diabetic and hypertensive rodent models and compared the drug effect of metformin versus glipizide and that of telmisartan versus metoprolol in the long-term (i.e., 2 and 8 weeks). At the basal level, namely untreated groups, db/db mouse and SHR aortas showed lower phosphorylation of AMPKα, PARP1, and eNOS and higher protein PARylation than db/m mouse and WKY rat aortas, respectively (left panels, Fig 6A, 6B, 6D and 6E). Consistent with previous reports, metformin and glipizide had similar glucose-lowering effect in db/db mice (S3 Fig). As well, telmisartan and metoprolol decreased blood pressure in SHR to the comparable level (S4 Fig). However, metformin but not glipizide, increased the phosphorylation of AMPKα, PARP1, and eNOS and decreased protein PARylation in db/db mouse aortas (right panel, Fig  6A and 6B). Telmisartan, but not metoprolol, increased the phosphorylation of AMPKα, PARP1, and eNOS and decreased protein PARylation in SHR aortas (right panel, Fig 6D and  6E). To associate the activation of AMPK-PARP1 with vascular function in various models, we assessed the expression of genes affecting EC function positively (e.g., eNOS, SIRT1, KLF4) or negatively (e.g., ICAM-1, VCAM-1). The mRNA levels of eNOS, SIRT1 and KLF4 were lower and those of ICAM-1 and VCAM-1 higher in db/db mouse and SHR aortas than db/m mouse and WKY rat controls, respectively. Metformin but not glipizide improved EC functions in diabetic db/db mouse aortas, as indicated by decreased expression of ICAM-1 and VCAM-1 and increased expression of eNOS, SIRT1, and KLF4 (Fig 6C). A similar vascular beneficial effect was found in aortas of SHRs receiving telmisartan but not metoprolol (Fig 6F).

Discussion
Herein, we demonstrated for the first time that AMPK phosphorylation of PARP1 Ser-177 in vivo, which contributes to the endothelial protection in the context of anti-hypertensive and anti-diabetic drugs. Although vasculature is directly benefited from the blood pressure and glucose lowering effect, additional therapeutic efficacy of some, but not all, of these drugs may build on the endothelial protection because of this AMPK-dependent PARP1 inhibition. Thus, the major translational implication of this finding is that the vascular protection of metformin and telmisartan occurs in part through activation of the AMPK-PARP1 cascade.
Clinical studies such as the UK Prospective Diabetes Study (UKPDS) have underscored the vasoprotective effect of metformin. In contrast, patients receiving sulfonylurea (e.g., glipizide) did not show a similar extent of cardiovascular protection despite glycemic control [23,26,27]. Vasoprotection with metformin is mediated in part, if not predominantly, by improved EC function [28]. At the molecular level, metformin increased AMPK activity in adipose tissue from diabetic patients, which was not observed with sulfonylurea treatment [29]. Additionally, metformin improves endothelial function in obese mice by inhibition of ER stress through AMPK/PPARδ pathway [30]. Hyperglycemia associated with diabetes mellitus is known to impair endothelial function, as manifested by PARP1 activation and resulting oxidative and inflammatory stresses [24]. Given our findings that metformin can induce AMPK phosphorylation of PARP1 Ser-177 and in turn inhibit PARP1 activation, the AMPK-PARP1 axis may be an important mechanism underlying the vasoprotective effect of metformin. Phenformin, Metformin is also used to treat polycystic ovary syndrome that is often associated with endothelial dysfunction [31,32]. AMPK phosphorylation of PARP1 Ser-177 may also contribute to the improved EC function in PCOS patients receiving metformin [33].
Although endothelial dysfunction is prevalent in hypertensive patients, the protective effect of anti-hypertensive drugs on the endothelium is less clear than that of biguanides. In the 2014 evidence-based guideline for management of hypertension in adults, the Eighth Joint National Committee (JNC 8) recommended ARBs as a first-line treatment, with β-blockers as later-line alternatives in part because of the higher rate of cardiovascular events associated with β-blockers as compared with ARBs [34]. Several small-scale clinical studies indicated that telmisartan improved FMD in patients with essential hypertension or metabolic syndrome [15,35,36]. As well, the ARBs losartan and irbesartan were found superior to atenolol (a selective β1 blocker) in improving the structure and EC function of resistant arteries [37,38]. Likewise, telmisartan, but not metoprolol, restored eNOS activity and ameliorated oxidative stress in diabetic rats [39,40]. Others showed that telmisartan activated PPARγ in macrophages and ECs [17,41,42], and we demonstrated that telmisartan, but not metoprolol, activated the AMPK-PARP1 axis, which provides additional evidence for the cardiovascular protective effect of telmisartan. This finding is consistent with a report by Sharma et al showing that metoprolol did not change AMPK activity or phosphorylation of acetyl-CoA carboxylase, a canonical substrate of AMPK [43].
Besides AMPK, several kinases have been shown to regulate PARP1. However, these phosphorylation events increase, rather than decrease, PARP1 activity. Extracellular signal-regulated kinase (ERK) can bind to PARP1 and phosphorylate Ser-372 and Thr-373 to enhance PARP1 activity [45,46]. In the context of EC biology, inhibition of ERK can decrease disturbed flow-induced PARP1 activity [47]. As well, JNK can interact with and phosphorylate PARP1 to increase PARP1 activity [48]. With the robust induction of PARP1 by oxidative stress, ERK and/or JNK activation under hyperglycemia and hypertension may contribute to increased PARP1 activity. In contrast, metformin and telmisartan decreased PARP1 activity via the AMPK-PARP1 cascade, as shown in the current study. Therefore, AMPK phosphorylation of PARP1 would counteract other phosphorylation events activating PARP1.
Noticeably, statins (known to activate AMPK) [49] and other ARBs may also activate this beneficial pathway because atorvastatin and losartan increased PARP1 phosphorylation in ECs. On the other hand, other sulfonylurea drugs and β-blockers lack this beneficial effect (S6 and S7 Figs). While our study provides molecular basis of additional vascular beneficial effect of metformin and telmisartan in hyperglycemic or hypertensive patients, these results by no means challenge the clinical use of metoprolol and glipizide. Rather, given that metoprolol has profound effect in treating patients with heart failure, ischemic heart disease, or arrhythmia, combinatory use of drugs such as telmisartan or statins may provide additional vascular protection. Likewise, the combined use of glipizide, metformin, and statins may show pleiotropic effect in the vessels of patient with diabetes through activation of the AMPK-PARP1 axis.