Nrf2-regulated redox signaling in brain endothelial cells adapted to physiological oxygen levels: Consequences for sulforaphane mediated protection against hypoxia-reoxygenation

Ischemic stroke is associated with a surge in reactive oxygen species generation during reperfusion. The narrow therapeutic window for the delivery of intravenous thrombolysis and endovascular thrombectomy limits therapeutic options for patients. Thus, understanding the mechanisms regulating neurovascular redox defenses are key for improved clinical translation. Our previous studies in a rodent model of ischemic stroke established that activation of Nrf2 defense enzymes by pretreatment with sulforaphane (SFN) affords protection against neurovascular and neurological deficits. We here further investigate SFN mediated protection in mouse brain microvascular endothelial cells (bEnd.3) adapted long-term (5 days) to hyperoxic (18 kPa) and normoxic (5 kPa) O2 levels. Using an O2-sensitive phosphorescent nanoparticle probe, we measured an intracellular O2 level of 3.4 ± 0.1 kPa in bEnd 3 cells cultured under 5 kPa O2. Induction of HO-1 and GCLM by SFN (2.5 μM) was significantly attenuated in cells adapted to 5 kPa O2, despite nuclear accumulation of Nrf2. To simulate ischemic stroke, bEnd.3 cells were adapted to 18 or 5 kPa O2 and subjected to hypoxia (1 kPa O2, 1 h) and reoxygenation. In cells adapted to 18 kPa O2, reoxygenation induced free radical generation was abrogated by PEG-SOD and significantly attenuated by pretreatment with SFN (2.5 μM). Silencing Nrf2 transcription abrogated HO-1 and NQO1 induction and led to a significant increase in reoxygenation induced free radical generation. Notably, reoxygenation induced oxidative stress, assayed using the luminescence probe L-012 and fluorescence probes MitoSOX™ Red and FeRhoNox™-1, was diminished in cells cultured under 5 kPa O2, indicating an altered redox phenotype in brain microvascular cells adapted to physiological normoxia. As redox and other intracellular signaling pathways are critically affected by O2, the development of antioxidant therapies targeting the Keap1-Nrf2 defense pathway in treatment of ischemia-reperfusion injury in stroke, coronary and renal disease will require in vitro studies conducted under well-defined O2 levels.


Introduction
Ischemic stroke is a leading cause of death and adult morbidity worldwide [1]. The critical reduction of blood flow within a major cerebral artery leads to reduced oxygen and nutrient delivery to the brain [2], time-dependent neuronal cell death and the development of neurological deficits [3]. The initiation of a pathophysiological cascade, involving oxidative stress and inflammation [4], is further exacerbated by the generation of reactive oxygen species (ROS) and mitochondrial dysfunction during reperfusion [5,6]. Timely restoration of cerebral blood flow is currently the only effective pharmacological treatment for acute ischemic stroke. Treatment with tissue plasminogen activator (rt-PA) improves reperfusion and functional outcomes, yet is limited to a ~4.5 h window after the onset of stroke due an increased risk of hemorrhagic transformation [7,8]. Recent trials of endovascular thrombectomy in stroke patients with large vessel occlusion report significant improvements in functional outcomes [1]. However, increased generation of reactive oxygen species in ischemic brain regions may compromise potentially rescuable penumbral tissue surrounding the infarct core [5,9].
Disruption of the blood-brain barrier (BBB) in ischemic stroke leads to extravasation of blood-borne inflammatory cells and fluid into the brain parenchyma which underlies dysregulation of neurovascular function [10][11][12][13]. Our previous studies in a rodent model of ischemic stroke established that pretreatment of animals with the dietary isothiocyanate sulforaphane (SFN) [14], an electrophilic activator of the redox sensitive transcription factor Nuclear factor-erythroid 2 p45-related factor 2 (Nrf2) [15,16], significantly reduces BBB permeability, infarct volume and behavioral deficits [17,18]. Our MRI studies further demonstrated that prophylactic SFN delivery reduced lesion volume, consistent with reduced BBB permeability to IgG and improved neurological outcome [17]. Notably, SFN rapidly enters the brain [19] and upregulates Nrf2 and HO-1 expression in brain perivascular astrocytes and endothelial cells [17,18].
The majority of studies in endothelial and other cell types are conducted during culture under atmospheric O 2 (18 kPa), whereas most cells experience much lower levels in vivo, with brain endothelial cells exposed to ~3-7 kPa [20]. Hyperoxic conditions create a pro-oxidation environment, reducing replicative lifespan [21] and enhancing cellular antioxidant defenses [22,23], thereby potentially limiting the clinical relevance of in vitro findings. We recently reported that SFN mediated induction of select Nrf2 target genes in umbilical vein endothelial cells (HUVEC) is attenuated under physiological normoxia (5 kPa O 2 ) compared to atmospheric O 2 levels [22]. Moreover, we reported that adaptation of HUVEC to 5 kPa O 2 enhances nitric oxide bioavailability, modulates agonist-induced Ca 2+ signaling [24] and protects against Ca 2+ overload due to increased SERCA activity [25].
In this study, we further explore the mechanisms underlying SFN afforded protection in ischemic stroke by investigating redox signaling in mouse brain microvascular endothelial cells (bEnd.3) subjected to hypoxia-reoxygenation following adaptation to defined O 2 levels. Our findings demonstrate that SFN induces Nrf2-regulated defense enzymes in bEnd.3 cells to protect against reoxygenation induced reactive oxygen species generation. These findings together with our study in of ischemic stroke in vivo [17,18] suggest that SFN may be a prophylactic therapeutic for targeting the Keap1-Nrf2 defense pathway in stroke and potentially coronary and renal disease.
treatments and experiments are conducted within the O 2 -regulated workstation and/or plate reader (CLARIOstar, BMG Labtech, Germany). All experiments were conducted using bEnd.3 cells in passages 7-15.

Phosphorescence lifetime measurements of O 2 levels in bEnd.3 cell cytosol and medium
Intracellular O 2 levels were monitored in live cells using a cellpenetrating phosphorescent platinum-porphyrin based nanoparticle probe, MitoXpress®-INTRA (Agilent, USA) [26]. A time-resolved fluorescence plate reader (CLARIOstar, BMG Labtech), equipped with an atmospheric control unit, enabled us to measure cytosolic O 2 levels under defined ambient O 2 levels. bEnd.3 cells were seeded into 96-well black microtitre plates and loaded with MitoXpress®-INTRA (10 μg/ml) for 16 h in complete DMEM. The probe emits a phosphorescence signal at 655 ± 55 nm when excited at 355 ± 55 nm [22,24]. Molecular oxygen quenches the phosphorescence signal, and the signal decay is inversely proportional to the concentration of O 2 . Phosphoresence intensity after excitation was measured after 30 μs (t 1 ) and 70 μs (t 2 ) with a 30μs window and converted to probe lifetime (τ) using the formula: τ=(t 2 -t 1 )/ln (f 1 /f 2 ), where f 1 and f 2 represent phosphorescence intensities at respective timepoints [27]. Averaged lifetime measured at 7 ambient O 2 tensions was plotted against the known O 2 tension and subjected to an exponential fit analysis. Lifetime values were then interpolated from this curve (see Fig. 2B) to give the dissolved intracellular O 2 level in live bEnd.3 cells. Dissolved O 2 culture medium was also measured in parallel by diluting MitoXpress®-INTRA (2.5 μg/ml) in DMEM medium.

Immunoblotting
Cell lysates were extracted with SDS lysis buffer containing protease and phosphatase inhibitors (pH 6.8) on ice for 10 min. Denatured samples (10 μg) were separated by gel electrophoresis, electrotransferred onto polyvinylidene difluoride membranes and then probed with primary and HRP-conjugated secondary antibodies, using Lamin A/C (Santa Cruz, USA), α-tubulin (Millipore, UK) or β-actin (Sigma-Aldrich, USA) as reference proteins for nuclear and cell protein, respectively [22,28]. Nuclear protein was extracted using a nuclear extraction kit (Active Motif). Membranes were probed for HO-1 (Cell Signaling Technology), GCLM (gift from Prof. T. Kavanagh, University of Washington, WA, USA), NQO1 (Santa Cruz, USA), HIF-1α (Abcam, UK), catalase (Calbiochem, UK) and Nrf2 (Santa Cruz, USA). Protein expression was determined by enhanced chemiluminescence with images captured in a gel documentation system (G-BOX, Syngene Ingenius Bioimaging) and analysed by densitometry using Image J software (NIH, USA).

Mitochondrial reactive oxygen species measured using MitoSOX™ Red
Mitochondrial reactive oxygen species generation was measured using a mitochondrial targeted fluorogenic probe MitoSOX™ Red [34], and we previously confirmed that MitoSOX fluorescence in endothelial cells is attenuated by scavenging superoxide [28,35]. bEnd.3 cells seeded in black-walled, clear-bottomed 96-well plates were cultured under 18 or 5 kPa O 2 for 5 d and then incubated in serum-free DMEM in the absence or presence of rotenone (1 μM, complex 1 inhibitor) or L-NAME (100 μM, eNOS inhibitor). Cells were exposed to hypoxia (1 kPa O 2 , 1 h) and loaded with MitoSOX™ Red (5 μM, Invitrogen) for 5 min before the start of reoxygenation under 18 or 5 kPa O 2 , respectively. Cells were washed twice with ice-cold PBS and fixed with 4% paraformaldehyde for 10 min before staining nuclei with DAPI (2 μg/ml, Sigma). MitoSOX™ Red fluorescence (Ex 545 nm/Em 602 nm) was detected using a Nikon Diaphot microscope, with images captured using an ORCA-03G (Hamamatsu, Japan) camera with 0.89 s exposure. Fluorescence quantification was conducted using image analysis software (ImageJ, NIH, USA), measuring the integrated intensity of fluorescence, area of field of view and mean grey value.

Statistics
Data denote the mean ± S.E.M. of at least 3-5 different bEnd.3 cell cultures and were processed using Graphpad Prism 8, with some preliminary handling steps performed using MARS data analysis software (BMG Labtech). Significance was assessed using either a paired Student's t-test or one-or two-way ANOVA followed by a Bonferroni Post Hoc test where appropriate, with significance confirmed by P < 0.05. we established that physiological concentrations of SFN (0.5-2.5 μM) [14] significantly upregulated Nrf2 mediated HO-1 protein levels (12-24 h, Supplementary Fig. S1A) and mRNA expression of HO-1, NQO1, Bach1 and Keap1 (4 h, Supplementary Fig. S1B).

Real-time measurement of intracellular O 2 level in bEnd.3 cells
We and others have emphasized the importance of monitoring O 2 gradients between culture medium and cell cytosol [20,22,39,40], and here report the first real-time measurement of intracellular O 2 in brain microvascular endothelial cells (bEnd.3) using the cell-penetrating phosphorescent nanoparticle probe MitoXpress®-INTRA [26]. Phosphorescence lifetime in bEnd.3 cells and medium was measured during stepwise reductions of O 2 (18 kPa-0 kPa) within an O 2 -regulated, time-resolved fluorescence plate reader ( Fig. 2A). The relationship between ambient O 2 levels in the plate reader and phosphorescence lifetime is illustrated in Fig. 2B. An intracellular O 2 level of 3.4 ± 0.1 kPa (inset, Fig. 2B) was measured in cells cultured under 5 kPa O 2 , recapitulating levels in the cortex of awake mice [41], with dissolved O 2 in the medium (5.2 ± 0.2 kPa) similar to the O 2 level (5 kPa) in the plate reader.

Adaptation of bEnd.3 cells to 5 kPa O 2 does not induce a hypoxic phenotype
To determine whether adaptation of bEnd.3 cells under 5 kPa O 2 induces hypoxic responses, we examined stabilization of HIF-1α, a key modulator of transcriptional responses to hypoxia. In the presence of oxygen, HIF-1α is degraded via prolyl hydroxylation [42], involving HIF-1α association with von Hippel-Lindau protein E3 ubiquitin ligase complex to promote degradation [42,43]. As intracellular O 2 availability decreases, these enzymes are no longer able to hydroxylate HIF-1α subunits, resulting in stabilization and upregulation of HIF-1α protein levels. When bEnd.3 cells were adapted to 18, 5 or 1 kPa O 2 , HIF-1α stabilization was only detected under hypoxia (Fig. 2C), confirming the absence of a hypoxic phenotype in cells maintained long-term under physiological normoxia (5 kPa O 2 ).

Effects of ambient O 2 levels on cell viability, ATP and GSH content and proliferation
Adaptation of bEnd.3 cells to 5 kPa O 2 had no effect on cell viability, as evidenced by negligible changes in mitochondrial dehydrogenase activity (data not shown) or intracellular ATP levels (5 kPa O 2 : 24.3 ± 2.2 vs 18 kPa O 2 : 20.7 ± 1.7 nmol/mg.protein) (Fig. 2D). Intracellular GSH (Fig. 2E) and catalase (Fig. 2F) levels were significantly lower in bEnd.3 cells adapted to 5 kPa O 2, consistent with our previous findings in airway epithelial cells [23] and other studies in epidermoid carcinoma cells [40]. Total intracellular GSH levels were similar in bEnd.3 cells in passages 7-15 (data not shown). Moreover, bEnd.3 cell proliferation was decreased under 5 kPa O 2 compared to 18 kPa O 2 (Fig. 2G). The implications of these findings are that the enhanced oxidative stress during standard cell culture under hyperoxia (18 kPa O 2 ) is attenuated in cells adapted to physiological normoxia 5 kPa O 2 ).

Physiological normoxia attenuates sulforaphane induced Nrf2 regulated enzyme expression
To determine whether Nrf2 redox signaling was affected by changes in ambient O 2 levels, bEnd.3 cells were adapted to 18 or 5 kPa O 2 and Nrf2 induced antioxidant enzyme expression determined by immunoblotting. Although basal levels of HO-1 and GCLM expression were affected negligibly following adaptation to physiological normoxia (5 kPa O 2 ), upregulation of Nrf2 regulated enzyme expression by SFN (2.5 μM, 24 h) was significantly attenuated in cells adapted to 5 kPa O 2 (Fig. 3). These findings are consistent with reports of diminished induction of antioxidant enzymes by Nrf2 in HUVEC [22], airway epithelial cells [23], RAW264.7 macrophages [44] and epidermoid carcinoma cells [40] under physiological normoxia.

Reoxygenation induced superoxide production in bEnd.3 cells under 18 kPa O 2
bEnd.3 cells adapted to 18 kPa O 2 were incubated with the chemiluminescent probe L-012 to investigate reactive oxygen species generation during hypoxia-reoxygenation. As shown in Fig. 4A, reoxygenation induced free radical generation was significantly inhibited by SOD (100U/ml) and PEG-SOD (PSOD, 50U/ml), whereas PEG- CAT (PCAT, 200U/ml) led to a non-significant decrease in L-012 luminescence, suggesting that reoxygenation most likely increases superoxide generation. Fig. 4B summarizes the changes in luminescence induced by reoxygenation in the absence and presence of scavengers of reactive oxygen species.
As NADP(H) oxidases (NOX) have been implicated as a source of free radical generation in cerebral ischemia-reperfusion [45,46], bEnd.3 cells adapted to 18 kPa O 2 were pre-treated with the NOX inhibitor VAS2870 (5 μM) for 30 min, incubated with L-012 and then exposed to hypoxia (1 kPa O 2 , 1 h) and reoxygenation. As shown in Supplementary  Fig. S2, reoxygenation induced increases in L-012 luminescence were unaffected by VAS2870, suggesting that NADPH oxidases are an unlikely source of acute reoxygenation induced free radical generation in bEnd.3 cells.

Sulforaphane pretreatment protects against reoxygenation induced superoxide generation
To determine whether upregulation of Nrf2 target genes attenuates reoxygenation induced free radical generation, bEnd.3 cells were adapted to 18 kPa O 2 and pre-treated with either vehicle (DMSO 0.01%) or SFN (2.5 μM) for 24 h. SFN significantly diminished reoxygenation induced L-012 luminescence ( Fig. 4C and D) and moreover, in cells transfected with scrambled or Nrf2 siRNA, we confirmed that silencing Nrf2 significantly enhances the L-012 luminescence signal during reoxygenation ( Fig. 4E and F).

Reoxygenation-induced increases in MitoSOX red fluorescence
To further validate reoxygenation mediated changes in L-012 luminescence, we examined mitochondrial reactive oxygen species generation in bEnd.3 cells adapted to 18 or 5 kPa O 2 . Hypoxia-reoxygenation increased MitoSOX fluorescence in cells under 18 kPa O 2 with negligible changes detectable under 5 kPa O 2 (Fig. 6A and B). As shown in Supplementary Fig. S3

Reoxygenation-induced increases in FeRhoNox fluorescence
Based on caveats associated with the specificity of L-012 and Mito-SOX Red, such as non-specific oxidation of the probes [49,50], further indirect measurements of reoxygenation induced superoxide production were conducted using an Fe 2+ -specific fluorescent indicator, FeRho-Nox™-1, in bEnd.3 adapted to 18 or 5 kPa O 2 . Inhibition of superoxide dismutase by ammonium tetrathiomolybdate has been reported to prolong the cytosolic Fe 2+ signal in dermal fibroblasts and endothelial cells challenged with UVA radiation [37]. As shown in Fig. 7A, increases in FeRhoNox-1 fluorescence during reoxygenation under 18 kPa O 2 were inhibited by PEG-SOD and potentiated by inhibition of SOD with ammonium tetrathiomolybdate. Previous studies have shown that mitochondria exposed to superoxide release iron from iron-sulphur clusters, indicating that increased radical production will lead to an increase in free iron [51,52]. In the present study, we exploited the fact that increases in free iron would increase FeRhoNox™-1 fluorescence, and thus changes in fluorescence served as an indirect measure of superoxide generation in bEnd.3 cells subjected to reoxygenation. Notably, reoxygenation induced FeRhoNox-1 fluorescence was attenuated in bEnd.3 cells adapted to 5 kPa O 2 (Fig. 7B). Together with our findings of reoxygenation induced changes in L-012 luminescence and MitoSOX fluorescence, FeRhoNox-1 measurements suggest that superoxide is the most likely free radical generated during acute reoxygenation induced oxidative stress in bEnd.3 cells.

Discussion
Changes in ambient O 2 levels during cell culture in vitro alter (i) ion channel and kinase activities [53][54][55], (ii) endothelial Ca 2+ signaling, nitric oxide bioavailability and their sensitivity to Ca 2+ overload [24,25], and (iii) induction of Nrf2-targeted antioxidant defenses [22,23,56]. We here further demonstrate that the redox phenotype of mouse brain microvascular endothelial cells is critically affected by ambient oxygen levels. Endothelial cells lining the blood-brain barrier in vivo are exposed to O 2 levels ranging between ~3 and 7 kPa, yet the majority of studies in brain endothelial and other cell types in vitro have employed standard culture conditions in which cells are exposed to hyperoxia (18 kPa O 2 ) and therefore sustained oxidative stress [20].
Using the O 2 -sensitive nanoparticle probe MitoXpress®-INTRA, we obtained the first measurements of intracellular O 2 (3.4 kPa) in bEnd.3 endothelial cells, recapitulating O 2 levels measured in brain endothelium in vivo. Importantly, long-term adaptation of bEnd.3 cells to 5 kPa O 2 was not associated with HIF-1α stabilization, confirming the absence of a hypoxic phenotype under physiological normoxia. Moreover, as gradients exist between ambient O 2 levels in a Scitive workstation, medium and cytosol, it is critical that medium and intracellular O 2 levels are measured simultaneously [20,22]. MitoXpress®-INTRA has been used to measure intracellular O 2 in umbilical vein endothelial cells [22,24], mouse embryonic fibroblasts [26], cortical neurons [57] and now in brain microvascular endothelial cells.
Adaptation of bEnd.3 cells under 5 kPa O 2 did not affect cell viability or intracellular ATP levels, but significantly decreased levels of intracellular GSH and catalase, suggesting that cells under physiological normoxia experience less oxidative stress [20,58]. Basal expression of Nrf2-regulated antioxidant enzymes was similar in bEnd.3 cells cultured under 18 or 5 kPa O 2 , however SFN mediated induction of HO-1 and GCLM was significantly attenuated in cells adapted to 5 kPa O 2 . Our finding of diminished HO-1 induction is in agreement with our previous studies in human umbilical vein and coronary artery endothelial cells  [22] and other studies in lung epithelial cells [23] and human dental pulp stem cells [56] cultured under relevant physiological O 2 levels. Notably, electrophile and nitric oxide mediated induction of HO-1 in human endothelial cells adapted 5 kPa O 2 is attenuated, but reversible on re-exposure of cells to 18 kPa O 2 or following silencing of the Nrf2 repressor Bach1 [22].
Our previous studies of reperfusion injury in a rodent model of transient ischemic stroke established that activation of Nrf2 antioxidant defenses by SFN affords neurovascular and neurological protection [17,18]. To mimic ischemia-reperfusion injury in stroke at a cellular level, bEnd.3 cells were adapted to either 18 or 5 kPa O 2 and subjected to hypoxia (1 kPa O 2 ) and reoxygenation under 18 or 5 kPa O 2 , respectively. Reoxygenation-induced increases in L-012 luminescence in cells adapted to 18 kPa O 2 was abrogated by SOD and polyethylene glycol SOD, implicating superoxide as the most likely free radical species generated during reoxygenation. Although polyethylene glycol catalase led to a non-significant decrease in reoxygenation-induced free radical production, we cannot exclude that inhibition of L-012 luminescence signal by SOD or polyethylene glycol SOD may be due to generation of superoxide from molecular oxygen during L-012 oxidation by H 2 O 2 /peroxidase [59]. We further demonstrated that upregulation of Nrf2-regulated antioxidant enzymes by SFN led to a significant decrease in reoxygenation induced free radicals ( Fig. 4C and D), whilst silencing Nrf2 transcriptional activity enhanced reoxygenation induced free radical generation (Fig. 4E and F).
Reoxygenation induced changes in L-012 luminescence were lower in bEnd.3 cells adapted to 5 kPa O 2 (Fig. 5), although L-012 signals trended to decrease in the presence of PEG-SOD or following SFN pretreatment. In this context, studies in macrophages [44], epidermoid carcinoma cells [40] and dental pulp stem cells [60], as well as, our experiments in human endothelial cells (data not shown) confirm that oxidative stress is lower in cells adapted to physiological normoxia. Thus, under standard, hyperoxic cell culture conditions, the redox phenotype of cells is characterized by an upregulation of Nrf2-regulated gene transcription to counteract enhanced reactive oxygen species generation and sustained oxidative stress [20].
We further investigated the redox status of bEnd.3 cells exposed to hypoxia-reoxygenation by assaying MitoSOX fluorescence as an index of mitochondrial reactive oxygen species generation. Reoxygenation significantly increased MitoSOX fluorescence in bEnd.3 cells adapted to 18 but not 5 kPa O 2 , and notably SFN pretreatment only attenuated reoxygenation-induced MitoSOX fluorescence in cells adapted to 18 kPa O 2 , further supporting our finding that SFN inhibits acute reoxygenation induced increases in L-012 luminescence. Reoxygenation induced increases MitoSOX fluorescence were unaffected by L-NAME or a pan-NADPH oxidase inhibitor (VAS2870), suggesting that free radical generation in bEnd.3 cells was unlikely due to eNOS or NOX. Although VAS2870 had no effect on cell viability or reoxygenation induced free radical generation, we cannot exclude that VAS2870 and other NOX inhibitors may have off-target effects via thiol alkylation, inhibition of mitochondrial respiration and cytotoxicity [61]. Furthermore, although undetectable in our study, we cannot exclude the possibility of enhanced reactive oxygen species generation during hypoxia, as it has recently been suggested that acute hypoxia drives the import of Na + into the mitochondrial matrix, reducing inner mitochondrial fluidity and consequently concentrating the production of superoxide at complex III [62].
To further characterize reoxygenation-induced free radical generation in bEnd.3 cells, release of intracellular Fe 2+ was monitored as an indirect measure of intracellular superoxide generation. By using PEG-SOD and a SOD inhibitor, we demonstrated for the first time that changes in FeRhoNox-1 fluorescence provide a useful measure of reoxygenation-induced superoxide generation in brain microvascular endothelial cells. Release of labile iron is closely associated with reactive oxygen species generation, and iron accumulation occurs in stroke [63], traumatic brain injury [64] and neurodegenerative disorders [65,66]. Furthermore, mitochondria exposed to superoxide anions release iron from iron-sulphur clusters, such that increased free radical generation will result in increased free iron [51,52]. Increases in superoxide in the presence of SOD inhibition reduces Fe 3+ in the ferritin core to Fe 2+ , releasing Fe 2+ into the cytoplasm [37,67].
Our study establishes that bEnd.3 cells adapted to long-term to hyperoxia (18 kPa O 2 ) exhibit heightened sensitivity to hypoxiareoxygenation, resulting in increased reactive oxygen species generation on reoxygenation. Studies in vivo have reported that following ischemia-reperfusion injury in the heart and brain, accumulation of succinate in mitochondria drives reactive oxygen species generation via reverse electron transport at mitochondrial complex I, and that oxidative damage can be decreased by reducing succinate accumulation [68]. In the present study, activation of Nrf2 by SFN significantly diminished reoxygenation induced free radical generation while silencing of Nrf2 exacerbated free radical generation, implicating Nrf2 in protection against reoxygenation/reperfusion injury. In this context, Nrf2 has been shown to significantly affect the mitochondrial membrane potential, fatty acid oxidation and the availability of substrates including succinate [69,70]. As Nrf2 deficient cells and mice in vivo are more sensitive to oxidative damage [71,72], activation of Nrf2 by SFN not only upregulates antioxidant defense enzymes but importantly also influences mitochondrial substrate utilization and respiration [69,70].
As generation of reactive oxygen species was attenuated in bEnd.3 cells adapted to physiological normoxia, it is possible that the probes L-012 and MitoSOX Red used in this study and other studies lack sufficient sensitivity to monitor low levels of radical generation in response to acute reoxygenation. Although recent advances in multiphoton redox and pO 2 imaging have enabled elegant quantification of metabolic processes under different ambient O 2 levels [73], we consider it important to ensure that decreasing ambient oxygen levels from 18 kPa O 2 does not result in HIF-1α stabilization and activation hypoxic signaling pathways. In this context, we previously reported that long-term (~5 d) culture of vascular cells under physiological O 2 levels is required to exclude a hypoxic phenotype [22,24].
In view of the caveats concerning luminescence and fluorescence indicators [49,50], further studies are warranted using novel genetic biosensors for high-resolution, real-time imaging of reactive oxygen and nitrogen species in single cells and subcellular compartments [74,75]. Conducting such experiments in cells adapted long-term under controlled and physiologically relevant O 2 levels will prove challenging, but we are convinced that such in vitro cell culture models, in particular targeting biosensors to mitochondria in live cells, will provide insights for the design of novel therapeutics for treatment cerebral, coronary, renal and hepatic ischemia-perfusion injury.

Declaration of competing interest
The authors declare no competing interests.