Therapeutically targeting mitochondrial redox signalling alleviates endothelial dysfunction in preeclampsia

Aberrant placentation generating placental oxidative stress is proposed to play a critical role in the pathophysiology of preeclampsia. Unfortunately, therapeutic trials of antioxidants have been uniformly disappointing. There is provisional evidence implicating mitochondrial dysfunction as a source of oxidative stress in preeclampsia. Here we provide evidence that mitochondrial reactive oxygen species mediates endothelial dysfunction and establish that directly targeting mitochondrial scavenging may provide a protective role. Human umbilical vein endothelial cells exposed to 3% plasma from women with pregnancies complicated by preeclampsia resulted in a significant decrease in mitochondrial function with a subsequent significant increase in mitochondrial superoxide generation compared to cells exposed to plasma from women with uncomplicated pregnancies. Real-time PCR analysis showed increased expression of inflammatory markers TNF-α, TLR-9 and ICAM-1 respectively in endothelial cells treated with preeclampsia plasma. MitoTempo is a mitochondrial-targeted antioxidant, pre-treatment of cells with MitoTempo protected against hydrogen peroxide-induced cell death. Furthermore MitoTempo significantly reduced mitochondrial superoxide production in cells exposed to preeclampsia plasma by normalising mitochondrial metabolism. MitoTempo significantly altered the inflammatory profile of plasma treated cells. These novel data support a functional role for mitochondrial redox signaling in modulating the pathogenesis of preeclampsia and identifies mitochondrial-targeted antioxidants as potential therapeutic candidates.

Preeclampsia is a pregnancy-specific syndrome that complicates 5% of nulliparous pregnancies and worldwide affects approximately 4 million women per annum 1 . Globally, preeclampsia is a leading cause of maternal mortality and it is responsible for occupancy of approximately 20% of neonatal intensive care unit cots. The syndrome is characterized clinically by maternal hypertension accompanied by proteinuria or haematological or biochemical abnormalities 1 . Despite intense research, the precise pathophysiological mechanisms underlying this syndrome remain poorly elucidated.
However, there is substantial evidence that defective placentation in early pregnancy is a pivotal event in the aetiology of this condition 2 . A consequential reduction in placental perfusion provokes an ischemic placental microenvironment due to fluctuations in oxygen delivery to the placenta and fetus, which results in oxidative stress 3 . Elevated placental oxidative stress is evident in preeclampsia as early as ~8-10 weeks' gestation 4 . Placental ischemia is inherently linked to elevated production and secretion of placental-derived deleterious mediators that induce widespread maternal endothelial dysfunction. Uncomplicated pregnancy is itself a state of oxidative stress 5 as a result of amplified maternal metabolism and the subsequent metabolic activity of the placenta 6 . However, during preeclampsia the mitigative systems normalizing the placental oxidative state are distorted, leading to elevated generation of pathogenic factors and subsequent vascular dysfunction. During preeclampsia, oxidative stress manifests in both the placenta and maternal circulation 7 , with evidence of diminished antioxidant defences 8 , increased free radical formation and isoprostanes 9 .
There are several sources of reactive oxygen species (ROS) within the cell; however mitochondria are the dominant cellular producers of ROS. Recent evidence has established that mitochondrial-ROS (mROS) have evolved as raconteurs directing mitochondrial function and other critical physiological signalling roles to preserve homeostasis and stimulate adaption to deleterious stressors 10 . There is a growing body of evidence implicating mitochondrial dysfunction as a pathogenic mediator of oxidative stress in preeclampsia 11 . There is substantial mitochondrial content in the placenta, in part to mediate the elevated metabolic activities during pregnancy 12 . Excessive production of mitochondrial-ROS is intrinsically linked to mitochondrial dysfunction 13,14 . Furthermore, there is a higher incidence of preeclampsia in a family with pre-diagnosed mitochondrial dysfunction 15 .
MitoTempo is a mitochondria-targeted superoxide dismutase antioxidant mimetic. Recent work demonstrated that MitoTempo accumulates in the mitochondria by increasing mitochondrial O 2 − dismutation while not affecting cytoplasmic dismutation in endothelial cells 16 . Additionally, MitoTempo decreased mitochondrial O 2 − in intact endothelial cells. Furthermore, MitoTempo improved endothelial function and reduced mROS production in an in vivo model of hypertension 16 . In the current studies, we identified that deleterious plasma mediators present in preeclampsia generate increased mitochondrial-specific superoxide production, by treating HUVEC with pooled plasma from preeclampsia pregnancies and matched uncomplicated controls from the SCOPE study (www.scopestudy.net) and non-pregnant controls. We show that increased mROS production evoked vascular dysfunction. Finally, we determine that antioxidants (MitoTempo) directly targeting mitochondrial superoxide scavenging prevent increased mROS production and elucidate the potential novel therapeutic pathway that may treat the syndrome more effectively.

Results
Preeclampsia plasma mediators modulate mitochondrial metabolism in HUVEC. The emerging role of mitochondrial dysfunction in mediating the pathogenesis of preeclampsia, led us to investigate a potential link between deleterious plasma mediators in preeclampsia and a subsequent dysregulation of mitochondrial function in HUVEC. In our first set of experiments we examined mitochondrial respiration by measuring oxygen consumption in plasma-treated HUVEC using the MitoXpress assay containing an oxygen sensitive fluorescent probe. Rates of oxygen consumption (OCR) are calculated from the changes in fluorescence signal over time using fluorescence plate reader. We established that after 4 hrs incubation of HUVEC with 3% plasma from women with preeclampsia significantly reduced OCR (40.61 ± 18.10 RFU, n = 5, P < 0.05) when compared with treatment with 3% plasma from uncomplicated pregnant (94.12 ± 34.9 RFU, n = 5) and non-pregnant women (52.12 ± 18.8, n = 5) (Fig. 1).

Effect of preeclampsia-plasma mediators on endogenous endothelial antioxidant and inflammatory markers.
Given the proficiency of mROS in regulating various signalling pathways and consequent pathologies, the effect of plasma mediators on antioxidant gene expression was determined by quantitative real-time PCR in HUVEC. Incubation with preeclampsia plasma induced a significant increase in SOD1 gene expression (1.32 ± 0.10 fold, n = 8, P < 0.05), SOD2 gene expression (1.35 ± 0.21 fold, n = 8, P < 0.05) and HO-1 gene expression (1.82 ± 0.42 fold, n = 8, P < 0.05) respectively compared to uncomplicated pregnancy (Fig. 4a).

Cytoprotective effects of MitoTempo on H 2 O 2-induced cell death.
The aim of the next series of experiments was to analyse the protective properties of mitochondrial targeted antioxidant, MitoTempo in HUVEC in response to cellular stressors. Initially we determined the optimal concentration of MitoTempo that did not adversely affect cell viability. HUVEC were treated with increasing concentrations of MitoTempo and cell viability was determined using an MTT assay. In summary, 5 uM MitoTempo was used in subsequent experiments (Fig. 5a). Additionally, using an MTT assay we determined that 5 uM of non-mitochondrial targeted N-acetylcysteine (NAC) was the optimal concentration that did not adversely affect HUVEC cell viability (Fig. 5b).
Furthermore, H 2 O 2 is frequently used as a cellular stressor to mimic oxidative stress in in vitro cellular systems and is a potential pathogenic mediator present in the preeclampsia plasma milieu. To determine the optimal dose of H 2 O 2 that reduces cell viability in HUVEC, cells were treated with increasing concentrations of H 2 O 2 and cell viability was recorded by MTT assay. A significant reduction in cell viability was seen in cells treated with a range of 20 uM-500 uM H 2 O 2 compared to control (*P < 0.05 and **P < 0.01). 200 uM H 2 O 2 was chosen as the optimal dose to use in subsequent experiments as it significantly reduced HUVEC cell viability by approximately 35% (Fig. 5c).

MitoTempo reduces preeclampsia plasma mediated increase in mitochondrial-specific ROS generation in HUVEC.
We examined if pre-incubation with MitoTempo could scavenge the volume Figure 3. Detection of mitochondrial-specific superoxide in plasma treated HUVEC. HUVEC were incubated with 3% plasma from women with preeclampsia, uncomplicated pregnancies and non-pregnant women for 4 hrs and mitochondrial-specific superoxide was detected using fluorogenic MitoSox Red dye. (a) Confocal microscopy of MitoSox (red fluorescence, 1st panel) and DAPI (blue fluorescence, 2nd panel) at 20X. Merged image localizes mitochondrial superoxide production (3rd panel). (b) MitoSox Red generation was quantified using Image J software. Data is the mean of 10 independent experiments and are expressed as difference in percentage pixel intensity between the study groups ± SEM. **P < 0.01. of mROS production in HUVEC exposed to preeclampsia plasma mediators. To elucidate the proficiency of mitochondrial-specific antioxidants, we additionally included non-mitochondrial targeted N-acetylcysteine (NAC) in our next experiments. Cells were pre-treated with 5 μ M MitoTempo or 5 μ M NAC for 2 hrs prior to exposure to 3% plasma from women with preeclampsia and levels of mitochondrial superoxide were detected by fluorogenic MitoSox Red dye and analysed using Image J software. Pre-treatment with MitoTempo significantly reduced mROS generation compared to untreated cells (61.23% ± 15.42% vs 100% ± 0%, n = 9, P < 0.01). Pre-treatment with non-targeted antioxidants reduced mROS production (79.73% ± 13.97% vs 100% ± 0%, n = 9, P < 0.05) but its effects are not as potent as targeted mitochondrial antioxidants (Fig. 6).

MitoTempo mediates redox and inflammatory signals in response to cellular stressors.
To specifically elucidate the protective effects of mitochondrial targeted antioxidant, MitoTempo in regulating pathogenic cellular pathways in endothelial cells, HUVEC were pre-treated with MitoTempo prior to exposure to stressors including H 2 O 2 (oxidative stress) and LPS (inflammation). Firstly, we examined the expression of inflammatory markers in response to 24 hrs stimulation with LPS (100 ng/ml) with/without 2 hrs MitoTempo (5 μ M) pre-treatment. There was a significant decrease in TNF-α gene expression (2.09 ± 0.18 fold vs 0.99 ± 0.15 fold, n = 3, P < 0.01) (Fig. 7a) in cells pretreated with MitoTempo compared to untreated cells.
Additionally, we determined the expression of redox signalling markers in response to 24 hrs stimulation with H 2 O 2 (200 μ M) with/without 2 hrs MitoTempo (5 μ M) pre-treatment. There was a significant decrease in UCP-1 gene expression (12.23 ± 0.39 fold vs 0.95 ± 0.43 fold, n = 3, P < 0.01) in cells pretreated with MitoTempo compared to untreated cells. Uncoupling proteins mediate the electrochemical potential across the inner mitochondrial membrane, which can subsequently regulate mROS production. Furthermore, there was a significant decrease in TLR-9 gene expression (4.2 ± 0.73 fold vs 1.02 ± 0.19 fold, n = 3, P < 0.05), demonstrating a novel role for MitoTempo in reducing mROS-mediated innate immune response (Fig. 7b).

Discussion
In excess of 800 peer-reviewed publications over the past 25 years have corroborated the hypothesis that oxidative damage is involved in the pathophysiology of preeclampsia yet current antioxidant interventions are not clinically effective. One possible explanation is that these antioxidant regimens have failed to reach the intracellular location, namely the mitochondria; hence they have failed to ameliorate the pathological oxidative damage.  . MitoTempo scavenging reduces mROS production. HUVEC were pre-treated with 5 μ M MitoTempo or 5 μ M NAC for 2 hrs prior to incubation with 3% plasma from women with preeclampsia for 4 hrs and mitochondrial-specific superoxide production was detected using fluorogenic MitoSox Red dye. MitoSox Red generation was quantified using Image J software. Data is the mean of 9 independent experiments and are expressed as difference in percentage pixel intensity compared to untreated ± SEM, *P < 0.05, **P < 0.01. Mitochondrial pharmacology has recently greatly advanced with a number of different pharmacology strategies in development to address mitochondrial dysfunction.
Mitochondrial dysfunction is a pathogenic mediator of oxidative stress in preeclampsia with increased mitochondrial lipid peroxidation and enhanced susceptibility to oxidation evident in mitochondria in the placenta of pregnancies complicated by preeclampsia 18 . Endothelial cells are the primary targets of the circulating factors and preeclampsia is characterised by aberrant vascular dysfunction 19 . There are a number of different instigators that distort mitochondrial function, including altered oxygen consumption, decreased ATP production, increased mROS production and mtDNA damage 20 . We explored the pathogenic mechanisms of preeclampsia plasma mediators on mitochondrial function by assessing oxygen consumption rates. We showed that preeclampsia plasma mediators significantly reduced mitochondrial respiration compared to uncomplicated pregnancy. This corroborates with recent work that showed alterations in mitochondrial morphology in preeclampsia and highlighted that miR210 (increased in preeclampsia) is potentially responsible for repression of mitochondrial respiration in preeclampsia 21 .
In the present study we show that mitochondrial-ROS is markedly elevated in HUVEC treated with plasma from pregnancies complicated by preeclampsia compared to uncomplicated pregnancies. Increased mROS production may be caused by decreased cellular respiration observed in OCR assays as impaired cellular respiration can lead to backward flux of electrons in the oxidative phosphorylation chain 22 . Likewise analysis of the mitochondrial placental proteome in preeclampsia reported increased abundance of proteins involved in oxidative stress and ROS generation 23 . Furthermore, recent work in a transgenic murine model overexpressing storkhead box 1 (preeclampsia susceptibility gene) showed exaggerated placental mitochondrial activity 24 .
PGC-1α is a well characterised pleiotropic orchestrator of mitochondrial biogenesis and antioxidant activity 25 . We showed increased PGC-1α protein expression in HUVEC treated with plasma form preeclampsia pregnancies. The dynamics of mitochondrial biogenesis and function is a complex system of cellular and molecular processes. Mitochondrial mass signifies the equilibrium between rates of biogenesis and degradation 26 . We measured mitochondrial mass and found no significant difference in mitochondrial mass in HUVEC treated with preeclampsia plasma when compared with plasma from uncomplicated pregnancy. The lack of a change in mitochondrial mass could be related to increased rate of degradation; interestingly recent data has shown increased autophagy in preeclampsia and in HUVEC's exposed to oxidative stress 27 . These results would suggest that preeclampsia plasma mediators alter mitochondrial metabolism and provoke mitochondrial dysfunction through multiple mechanisms.
Mitochondrial-ROS production is stringently regulated by numerous antioxidant systems in order to maintain redox-signalling homeostasis. To determine if these antioxidant pathways were modulated in HUVEC treated with plasma from preeclampsia pregnancies, we analysed the expression of both mitochondrial and non-mitochondrial antioxidants in the endothelium. Preeclampsia plasma significantly increased mitochondrial SOD1, SOD2 and non-mitochondrial HO-1 gene expression respectively. Previous studies have shown elevated decidual HO-1 protein expression and increased HO-1 in maternal serum from preeclampsia pregnancies 28 . Previous work has reported reduced SOD1 mRNA and SOD activity in isolated trophoblasts from preeclampsia patients 29 .
Recent reports have demonstrated that mitochondrial DAMPs (mROS and mtDNA) act as ligands for TLR-9 14 . Activation of TLR-9 signaling via mtDNA induces a subsequent inflammatory response with the synthesis of pro-inflammatory cytokines including TNF-α 14 . We showed increased levels of mtDNA in preeclampsia plasma. We also reported a significant increase in endothelial TLR-9 gene expression with a consequent increase in pro-inflammatory cytokine TNF-α gene expression respectively in HUVEC treated with preeclampsia plasma. TLR-9 expression has been shown to be significantly increased in the placenta in patients with preeclampsia 30 . Furthermore we showed a significant increase in ICAM-1 gene expression (marker of vascular dysfunction) in HUVEC treated with preeclampsia plasma. This correlates with previous work, which reported elevated ICAM-1 expression in preeclampsia 31 and in HUVEC's exposed to pathogenic necrotic trophoblast debris 32 . Furthermore, a recent transcriptomic study in HUVEC exposed to preeclampsia plasma established perturbation in pathways mediating endothelial homeostasis 33 . These findings implicate cross-talk between cellular stressors present in the maternal plasma milieu in preeclampsia.
MitoTempo is a mitochondria-targeted superoxide dismutase antioxidant mimetic. We observed that 5 μ M MitoTempo rescued cell viability following exposure to oxidative stress (200 μ M H 2 O 2 ). Importantly we showed that MitoTempo significantly reduced mROS generation following exposure to plasma from preeclampsia pregnancies. Intriguingly, mitochondrial-targeted antioxidant pre-treatment was more effective than general antioxidant (NAC) at similar concentrations, highlighting the importance of a direct-targeted therapeutic approach.
In order to specifically characterise the potential signalling pathways mediated by MitoTempo, HUVEC were pre-treated with the MitoTempo prior to exposure to two recognised cell stressors (oxidative stress and inflammation) hypothesised to be present as pathogenic mediators in the preeclampsia plasma milieu. We identified that MitoTempo exerted an anti-inflammatory response following LPS stimulation with a significant reduction in TNF-α gene expression. Interestingly, MitoTempo significantly reduced TLR-9 gene expression only in response to H 2 O 2, implicating mROS in provoking inflammatory response in plasma-treated HUVEC. MitoTempo has previously been shown to reduce expression of pro-inflammatory cytokines in in vivo models of hypertension 34 . Furthermore MitoTempo significantly normalised UCP-1 gene expression. Uncoupling reduces ROS generation by decreasing the electrochemical potential across the inner mitochondrial membrane, reducing the half-life of the most reactive steps in the electron transport chain 35 . Hence our results have identified potential mechanistic pathways for mitochondrial scavengers in mediating oxidative damage.
Here we provide evidence for the first time that plasma mediators of preeclampsia dysregulate mitochondrial function, generate increased production of deleterious mROS in the endothelium and ultimately provoke inflammatory-induced vascular dysfunction. We describe a mechanism for mediating the aberrant production of these pathogenic regulators using mitochondrial-targeted antioxidants that directly scavenge mitochondrial superoxide production. Our findings delineate that these mitochondrial antioxidants facilitate this outcome by adapting mitochondrial metabolism. Thus, our study shows that MitoTempo restrains production of mROS-mediated deleterious inflammatory cellular signaling pathways and provides evidence that therapeutic strategies directly targeting mitochondrial superoxide scavenging should be actively pursued in future therapeutic studies of preeclampsia.

Study Subjects. Subjects were recruited from the Screening for Pregnancy Endpoints (SCOPE) study
Ireland, an international multicentre prospective cohort study of nulliparous singleton pregnancies. Further details of this study have been published previously 36 . Preeclampsia was defined as systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥ 90 mm Hg on at least 2 occasions 4 hrs apart after 20 weeks' gestation but before the onset of labor or postpartum, with proteinuria (24 hour urinary protein ≥ 300 mg, or urine dipstick protein ≥ 2+ ) or any multisystem complication of preeclampsia. Time-of-disease samples (n = 12) for preeclampsia were taken when women had these criteria to diagnose preeclampsia. Control blood samples were taken from healthy pregnant women with uncomplicated pregnancies (n = 12) in the SCOPE study and matched for age, body mass index (BMI), and gestational age and from non-pregnant women matched for BMI and maternal age (Supplementary Table S1). The SCOPE study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures were approved by the Clinical Research Ethics Committee of the Cork Teaching (ECM5(10)05/02/08), and all women provided written informed consent. Sample Collection. Plasma samples were thawed on ice, centrifuged at 3000rpm at 4 °C, and the soluble component was removed, the remainder of sample was agitated. An equal volume (20-50ul per sample) was pooled with other samples in a single Falcon tube and agitated thoroughly. Aliquots were divided into 30ul volumes for storage at − 80 °C. In preliminary experiments, cell viability was determined using a range of plasma concentrations, 3% plasma concentration reduced the cytotoxic effects of plasma while maximizing relative differences between subject groups (Supplementary Figure S1). This concentration was used in the remainder of the study.  the RNeasy mini-kit (Qiagen). Superoxide dismutase 1 (SOD1), SOD2, uncoupling protein-1 (UCP-1), heme oxygenase-1 (HO-1), toll-like receptor 9 (TLR-9), tumour necrosis factor-α (TNF-α ) and intracellular adhesion molecule-1 (ICAM-1) gene expression was quantified by Real-Time PCR using the StepONE Plus Detection system. Taqman assays (Applied BioSystems) and Sybr Green primers (Supplementary Table S2 online) were used for quantification. The amounts of target gene, normalized to geometric mean of two internal controls (18S and Tata Binding Protein) were determined using the 2 −ΔΔCT . mtDNA Quantification. DNA was extracted from 200 μ l of plasma from both preeclampsia (n = 12) and uncomplicated pregnancies (n = 12) respectively with QIAmp DNA mini kit (Qiagen) according to manufacturer's instructions. Real-Time PCR was performed with 10 ng total DNA using the StepOne Plus Detection system. Mitochondrial DNA primers (hMitoF5, hMitoR5) and β 2M nuclear genome primers (β 2MF2, β 2MR2) were used for quantitation and are provided in Supplementary Table S2 online. Relative quantification of mitochondrial DNA (mtDNA) over nuclear DNA levels were determined using the using the 2 −ΔΔCT method.
Immunofluoresence. HUVEC were serum starved for 4 hrs and treated with 3% plasma for 24 hrs. Cells were initially fixed in 3% paraformaldehyde for 15 mins, prior to incubation in 0.2% Triton X-100 (Sigma-Aldrich) for 5 mins. Cells were blocked with 5% Bovine Serum Albumin (BSA) at room temperature for 30 mins prior to incubation with PGC-1α (1:200) (Novus Biologicals) overnight at 4 °C. Cells were then incubated with Alexa Fluor 488 goat anti-rabbit fluorescent secondary antibody (Invitrogen) at a 1:200 dilution. Cells were counterstained with 4,6-Diamidino-2-phenylindole dihydrochloride (DAPI) (10 μ g/ml) to identify nuclei. Cells were visualized by fluorescent microscopy (Zeiss AxioImager M2). Mean fluorescent intensity was analyzed using Image J software in at least 10 random fields of view and compared to DMSO controls.
Determination of mitochondrial mass. Mitochondrial mass was measured using MitoID Green fluorescent marker (Enzo Life Sciences) according to manufacturer's instructions. MitoID Green is a cell-permeable small organic probe that spontaneously localizes to mitochondria regardless of their membrane potential. Briefly, cells were serum starved for 1hour and incubated with 3% plasma for 4hrs. Media was removed and cells were loaded with 500 μ l 1X Assay buffer containing 0.5 μ l MitoID Green Reagent and 0.5 μ l Hoechst 33342 Nuclear Stain for 30 mins at 37 °C. Cells were then washed with PBS and fixed in 3% paraformaldehyde for 15 mins, prior to mounting. Mean fluorescent intensity was analyzed using Image J software in at least 10 random fields of view and compared to non-pregnant controls.
Detection of mitochondrial superoxide by fluorescent microscopy. Intramitochondrial superoxide production was measured in treated HUVEC using the MitoSOX Red fluorescent reagent (Invitrogen). This fluorogenic dye selectively enters mitochondria within living cells where it is oxidized by superoxide anions. This oxidation reaction then emits a red fluorescence when bound to mitochondrial DNA. Cells were serum starved for 1hour and incubated with 3% plasma for 4hrs. Alternatively, cells were serum starved for 1hour, pre-treated with 5 μ M Mito-Tempo, 5 uM N-acetylcysteine (NAC), or DMSO control for 2 hrs and incubated with 3% plasma for 4 hrs. Media was removed and cells were loaded with 0.5 uM MitoSox Red for 30 mins at 37 °C. Cells were then fixed and permeabilized prior to nuclear localization with DAPI. Mean fluorescent intensity was analyzed using Image J software in at least 10 random fields of view and compared to DMSO controls.

Measurement of Mitochondrial O2 consumption. MitoXpress ® Xtra-Oxygen Consumption Assay
(Luxel Biosciences) was used for the direct, real-time analysis of cellular respiration and mitochondrial function.
MitoXpress ® Xtra is quenched by O 2 through molecular collision, and thus the amount of fluorescence signal is inversely proportional to the amount of extracellular O 2 in the sample. HUVEC were serum starved for 1 hour and incubated with 3% plasma for 4 hrs. After incubation cells were washed with clear respiration media. The MitoXpress-Xtra-HS probe was added to cells in accordance with manufacturer's instructions. Oxygen consumption was measured using time-resolved fluorescence (TR-F) with a dual delay of 30 μ s and 70 μ s using a VarioSkan fluorescence plate reader (Thermo-Scientific). Rate of oxygen consumption was determined from the slope of fluorescence vs. time for each sample using relative fluorescence units/hour. Statistical Analysis. Data are shown as mean ± SEM, or fold change relative to vehicle control of at least 10 independent experiments. Mann Whitney U test or analyses of variance (ANOVA) were used where appropriate to determine statistical significance between groups in in vitro studies unless otherwise specified. Values of **P ≤ 0.01 and *P ≤ 0.05 were considered significant