Up-regulation of NADPH oxidase-mediated redox signaling contributes to the loss of barrier function in KRIT1 deficient endothelium

The intracellular scaffold KRIT1/CCM1 is an established regulator of vascular barrier function. Loss of KRIT1 leads to decreased microvessel barrier function and to the development of the vascular disorder Cerebral Cavernous Malformation (CCM). However, how loss of KRIT1 causes the subsequent deficit in barrier function remains undefined. Previous studies have shown that loss of KRIT1 increases the production of reactive oxygen species (ROS) and exacerbates vascular permeability triggered by several inflammatory stimuli, but not TNF−α. We now show that endothelial ROS production directly contributes to the loss of barrier function in KRIT1 deficient animals and cells, as targeted antioxidant enzymes reversed the increase in permeability in KRIT1 heterozygous mice as shown by intravital microscopy. Rescue of the redox state restored responsiveness to TNF-α in KRIT1 deficient arterioles, but not venules. In vitro, KRIT1 depletion increased endothelial ROS production via NADPH oxidase signaling, up-regulated Nox4 expression, and promoted NF-κB dependent promoter activity. Recombinant yeast avenanthramide I, an antioxidant and inhibitor of NF-κB signaling, rescued barrier function in KRIT1 deficient cells. However, KRIT1 depletion blunted ROS production in response to TNF-α. Together, our data indicate that ROS signaling is critical for the loss of barrier function following genetic deletion of KRIT1.


Results
Endothelial-targeted antioxidant enzymes reverse increased baseline microvessel permeability in KRIT1 deficient mice. Our previous studies demonstrated that KRIT1 hemizygous (Krit1 +/− ) mice exhibit a 2-fold increase in microvessel permeability, in both arterioles and venules 2 . This difference is exaggerated in animals with a complete loss of KRIT1 (data not shown). It is well described that several permeability inducing stimuli, including TNF-α, interleukin-1β (IL-1β), and lipopolysaccharide (LPS), stimulate ROS production and that scavenging ROS can limit permeability in vitro and in vivo. As it had also been previously reported that loss of KRIT1 expression can increase ROS production 12 , we tested whether scavenging ROS from the endothelium could reverse the increased permeability in KRIT1 deficient mice.
Loss of KRIT1 blunts TNF-α-induced increased ROS. Increased ROS production downstream of TNF-α stimulation is known to mediate changes in vascular permeability, gene expression, and other cell and tissue responses 27,28 . Given that the absence of KRIT1 caused high baseline levels of ROS, we asked whether this could potentially explain why KRIT1-deficient cells and animals have an abnormal response to TNF-α. Dihydroethidium (DHE) staining, a fluorescent indicator of oxidation, was used to detect ROS in KRIT1 siRNA-transfected cells treated with and without 100 ng/ml TNF-α. As previously reported, knockdown of KRIT1 in human arterial endothelial cells, and treatment of negative control (NC)-siRNA transfected cells with TNF-α, increased DHE staining. However, treatment of KRIT1-depleted cells with TNF-α was unable to increase DHE staining (KRIT1 siRNA, 2.05 +/− 0.31 vs. KRIT1 siRNA + TNF−α, 1.48 +/− 0.16, Fig. 2A,B). These data suggest that in the absence of KRIT1, TNF-α is unable to stimulate ROS production, which may contribute to the lack of permeability response.

Loss of KRIT1 increases production of ROS via up-regulation of Nox4. Endothelial cells contain
several sources of ROS, of these, the NADPH oxidases (Nox) are common targets of cytokine and chemokine signaling, can be regulated by molecules that also regulate cell adhesion, and have a well described role in vascular pathophysiology 16 . We therefore tested whether NADPH oxidases were involved in the increased ROS production in KRIT1 depleted cells by treating KRIT1-siRNA transfected cells with the pan-Nox inhibitor VAS2870. Ten µM VAS2870 reduced ROS levels in KRIT1 siRNA transfected cells to control levels (siKRIT1 +veh, 9.38 +/− 1.43 vs. siKRIT1 +VAS2870, 5.5 +/− 0.76, Fig. 3A), as shown by reduced fluorescence intensity per cell of the ROS sensor CellROX ® Green. This was similar in magnitude to the effect of this inhibitor on LPS-induced CellROX staining, which is known to be driven by Nox enzymes, strongly suggesting that loss of KRIT1 increases the activity of one or more Nox isoforms.
We measured Nox2 and Nox4 mRNA expression in both KRIT1 deficient (KO) and in KRIT1 overexpressing (9/6) MEFs (Fig. 3B) and found that loss of KRIT1 expression did not significantly effect Nox2 or Nox4 mRNA expression. In KRIT1 siRNA-transfected human pulmonary artery endothelial cells (HPAEC), Nox2 levels were undetectable, while Nox4 mRNA levels were modestly higher compared to the control (HPAEC siKRIT1 1.2 +/− 0.1 vs. HPAEC siNC 1.0 +/− 0.05, Fig. 3B). To examine Nox expression in wild type and KRIT1 heterozygous animals, we isolated RNA from mouse lung tissue, which is enriched in endothelial cells. Nox4 mRNA was strongly up-regulated in KRIT1 (Krit1 +/− ) heterozygous tissue, while SOD2 expression was slightly lowered (Fig. 3C). These data suggest that Nox4 mRNA expression is affected by loss of KRIT1 in a cell-type specific manner.
We then examined Nox4 expression at the protein level. Indeed, Nox4 protein and activity levels are tightly correlated, as this enzyme is constitutively activated. As shown in Fig. 3D,E, Nox4 protein was strongly up-regulated in KRIT1 null MEFs (KO) and KRIT1 siRNA transfected (siKRIT1) HPAEC. Taking advantage of endothelial KRIT1 null (Krit1 ECKO ) and heterozygous (Krit1 +/− ) mice derived from a conditional endothelial-specific KRIT1 knockout mouse model (PDGFBiCreER T2 Krit1 flox/flox 6 , see Methods), we also found that Nox4 was up-regulated in the endothelium of CCM-like lesions in Krit1 ECKO animals (Fig. 3F, i), as compared to non-lesion vessels (not shown), or to heterozygous controls (Krit1 +/− , Fig. 3F, iii). Taken together, these data strongly suggest that the loss of KRIT1 causes a significant up-regulation of this constitutive, Nox4-mediated, pro-oxidant pathway, both in vitro and in vivo.

Discussion
The regulation of vascular permeability is most often considered in the context of increased passage of water and plasma proteins across the vascular wall (edema) and leukocyte extravasation, processes involved in the inflammatory response. Our previous study suggested that the modification of the vascular barrier following decreased KRIT1 expression is independent from inflammatory factors, yet these changes can modify the vascular response to inflammatory mediators. On the other hand, previous findings showed that loss of KRIT1 increases ROS levels in endothelial cells and suggested that KRIT1 expression could modulate the molecular machinery that limit ROS accumulation 12,18 . As ROS are known to contribute to vascular permeability and are involved in inflammatory signaling, the goal of this study was to investigate this linkage in our model system.
We have now established that ROS signaling is a major contributor to increased endothelial permeability in KRIT1 deficient animals. Perhaps more importantly, we have demonstrated that ROS scavengers targeted specifically to the endothelium can block the increase in vascular permeability in KRIT1 deficient animals. These data agree with previous studies using Tempol and vitamin D3, both pleiotropic intracellular antioxidants, which increased endothelial barrier function in CCM2 deficient cells 30 . Our findings suggest that specifically targeting endothelial cells is as effective as systemic intervention, and that increased superoxide/H 2 O 2 is responsible for the observed enhanced permeability. Furthermore, we have demonstrated that a newly developed antioxidant and NF-κB inhibitor, yeast avenanthramide I (YAv1), is also capable of limiting vascular permeability in the absence of KRIT1 expression. This molecule shows a strong structural similarity to avenanthramides 19 , a group of anti-inflammatory compounds found in oats, which have been shown to inhibit IL-1β-induced NF-κB activity in endothelial cells 31 and IL-8-induced NF-κB activation in keratinocytes 32 . Intriguingly, it was previously observed that cell treatment with YAv1 can revert molecular phenotypes caused by KRIT1 loss-of-function, including the down-regulation of FoxO1 and SOD2 19 . Overall, these findings provide important insight into the functional mechanisms behind KRIT depletion-dependent increased permeability and into the role of KRIT1 in maintaining vascular quiescence.
It has long been held that post-capillary venules are the major site for fluid exchange, though more recently it has been accepted that permeability/filtration occurs throughout the microcirculation, including in arterioles. Mouse models suggest CCM lesions arise from capillaries and veins 33 , though this specificity is unverified in human patients. In our study, venule permeability was exquisitely dependent on ROS; indeed, ROS scavenging dampened venular permeability completely in both wild type and KRIT1 deficient animals (Fig. 1), and in the presence and absence of TNF-α. The baseline permeability of wild type arterioles, on the other hand, was unchanged by treatment with PECAM-AOE (Fig. 1C), though the increased baseline permeability in KRIT1 deficient arterioles was reversed (Fig. 1A). Tellingly, TNF-α-induced arteriole permeability was insensitive to ROS scavenging in either genotype, suggesting that permeability regulation downstream of TNF-α occurs through ROS-independent and ROS-dependent pathways (in arterioles and venules, respectively), and highlighting the heterogeneous behavior of endothelial cells from different vascular beds. Thus, while the mechanisms that govern arteriole vs. venule permeability in vivo remain ill-defined, our data suggest that venule permeability may be more sensitive to regulation by ROS, which is consistent with several other studies 34,35 .
We recently discovered that loss of KRIT1 affected the ability of TNF-α to increase permeability, but not to stimulate leukocyte activation 2 . Here, we found that TNF-α could not increase ROS levels in KRIT1 deficient cells (Fig. 2), suggesting that a component of the TNF-α-driven ROS production pathway is affected by the loss of KRIT1, an idea supported by the fact that antioxidant treatment could restore responsiveness to TNF-α in arterioles (Fig. 1A). Subsequently, we found that loss of KRIT1 increased Nox4 expression and up-regulated NF-κB activity (Figs 3 and 4), pathways known to be required for the endothelial response to TNF-α. The connection between TNF-α signaling and ROS is well established, though not well understood. TNF-α receptor activation leads to increased ROS production through activation of Nox2 and Nox4 36 , though the exact mechanism(s) remains unclear. Notably, ROS are also generated downstream of VEGF/VEGFR2 stimulation 37 . We recently reported that KRIT1 deficient cells secrete VEGF and exhibit constitutive activation of VEGFR2 38 . However, in our current study, the antioxidant and NF-κB inhibitor YAv1 was able to block endothelial leak in KRIT1 deficient cells (Fig. 5A), but not in VEGF-treated wild type cells (Fig. 5B). As VEGF increases endothelial ROS, but does not stimulate NF-κB activation, this suggests that the anti-NF-κB function of YAv1 may be more important than its antioxidant activity in the context of permeability regulation. Alternatively, these data could suggest that although the expression of VEGF by KRIT1 deficient cells contributes to the loss of barrier function, it is unlikely that it acts via increased ROS.
Nevertheless, the up-regulation of ROS following TNF-α stimulation is considered a critical component of the endothelial response to TNF-α, and has been suggested to lie upstream of both TNF−α induced increased permeability and NF-κB activation 39,40 . A substantial portion of ROS production downstream of TNF-α occurs via NADPH oxidase activity, and most studies of ROS generation in response to TNF−α have focused on the activity of Nox2. Nox2, normally expressed at low levels in quiescent endothelial cells, is stimulated by several stimuli, including TNF-α, which increase Nox2 expression or regulate the association of the Nox2 enzyme (gp91phox) with several regulatory subunits (p40, p47 and p67phox, and Rac1) 16 . In contrast, how TNF-α stimulation increases Nox4 activity remains largely unknown. Nox4 does not associate with the regulatory subunits that regulate the other Nox isoforms. Nox4 is thought to be regulated mainly at the transcriptional level, and several mechanisms have been implicated in controlling Nox4 expression, including NF-kB transcriptional activity 16,41 .
As both KRIT1 depletion and TNF-α increase expression of Nox4, it is possible that they act via the same mechanism, and that this contributes to the lack of permeability response to TNF-α in KRIT1 deficient animals. Herein, we reported an increase in Nox4 protein levels in HPAEC, an increase in Nox4 mRNA expression in Krit1 ECKO mouse lung tissue, and an increase in Nox4 staining in the endothelium of CCM-like vascular lesions (Fig. 3). This was accompanied by an increase in NF-κB (p65) expression and reporter activity (Fig. 4). These data suggest that the increase in Nox4 could be mediated transcriptionally by increased NF-κB activation. However, knockdown of KRIT1 in MEFs did not consistently increase Nox4 mRNA (Fig. 3B), while it did stimulate an increase in Nox4 protein in these cells (Fig. 3D,E). Thus, these data do not completely support a transcriptional mechanism for increased Nox4 in KRIT1 depleted cells. Moreover, inhibition of NADPH oxidases by VAS2870 was able to block KRIT1 depletion-dependent NF-κB reporter activity (Fig. 4C), which suggests that conversely, NF-κB may be downstream of Nox4. Both of these signaling relationships have precedence in the literature [42][43][44][45][46] . Therefore, given the lack of clarity regarding Nox4 regulation, and the large number of transcriptional and non-transcriptional signaling pathways active in KRIT1 depleted cells, additional studies to examine how loss of KRIT1 increases Nox4 expression will be required.
These data have clear implications for our understanding of vascular biology and the inflammatory response, yet, as KRIT1 is linked to an inheritable genetic disorder, the implications for the pathogenesis of CCM and for potential therapeutic development should also be considered. CCM lesions may remain clinically silent or unpredictably result in various clinical signs and symptoms at any age 47 . These lesions may be either single or multiple (even hundreds) with variable size (from a few millimeters to a few centimeters). The pleiotropic effects of ROS signaling in the endothelium suggest that oxidative stress could contribute to the development and progression of CCM disease, putatively contributing to several phenotypic characteristics including destabilization of endothelial cell-cell junctions, increased vascular permeability, and altered inflammatory responses 48 . Consistently, recent studies by Choquet et al. showed that genetic variants of modifier genes related to oxidative stress and inflammatory responses are associated with phenotypic markers of CCM disease severity 49,50 , suggesting that inter-individual variability in susceptibility to either oxidative stress or inflammation may contribute to CCM disease pathogenesis.
As we noted in our introduction, recent discoveries based on cell and animal models of KRIT1, CCM2, and CCM3 protein deficiency have highlighted a wide, interconnected network of signaling pathways that are regulated by these proteins. Interestingly, several of these mechanisms are regulated by-or regulate-cellular redox signaling 48 . Increased ROS production downstream of loss of KRIT1 causes the up-regulation of c-Jun, a basic component of the redox-sensitive transcription factor AP-1, in both cells and human CCM tissue samples 18 . ROS can also stimulate the activation of the TGF-β pathway, which has been observed in KRIT1 knockout mice 51 , and may be related to the up-regulation of c-Jun transcriptional activity 52 . Mitogen-activated protein kinase kinase kinase 4 (MEKK4) and Kruppel-like transcription factors (KLF) are both known targets and regulators of redox signaling [53][54][55] , as is Toll-like receptor 4 (TLR4) 56 , suggesting that ROS signaling may also be involved in the MEKK4/KLF4/TLR4 signaling axis recently identified as important in CCM pathogenesis 57,58 . Recent findings have also shown that KRIT1 deficient cells are defective in autophagy 59 , which is up-regulated in response to oxidative stress 60,61 . Thus, KRIT1 deficiency may lead to alterations in both ROS generation, as we show in this study, and in the adaptive response to increased oxidative stress.
It follows that targeting ROS/redox signaling may be a viable treatment strategy for CCM. Given that antioxidant treatment was particularly effective in venules, where CCM lesions develop 33 , our data supports the idea that controlling the redox state at the blood brain barrier could be a promising therapy for CCM. However, as studies in other redox-sensitive conditions have shown, reducing ROS generation below a certain threshold can be deleterious. In addition, though we have shown that antioxidant treatment is effective at reducing the increase in endothelial permeability, this may not be sufficient to prevent lesion formation. However, Gibson et al. reported that treatment with the antioxidants Tempol and vitamin D3 were able to reduce lesion burden in CCM2 deficient mice 30 . As a result, additional studies to examine the effectiveness of antioxidant treatment on reducing lesion formation in KRIT1 null animals, including the targeting of antioxidant treatment to the endothelium, are warranted.
In sum, our findings strongly support a role for increased ROS signaling in promoting increased baseline permeability in the absence of KRIT1. The increase in ROS likely prevents TNF-α from further increasing permeability in KRIT1 deficient cells and animals. This implies that the mechanism by which KRIT1 increases ROS is likely also utilized by TNF-α, which will help guide our future studies. As KRIT1 increased Nox4 expression and NF-κB activity, the nexus of Nox4, NF-κB and TNF-α signaling is a good starting point. Finally, we demonstrated that limiting ROS levels, either by introducing ROS scavenging enzymes into the blood, or by treating with recombinant yeast avenanthramide I, was able to block KRIT1 depletion-dependent increased permeability in vitro and in vivo, thus supporting the idea that antioxidant therapy could be a potential treatment for CCM.

Methods
Reagents. Conjugation via amino chemistry was used to prepare anti-PECAM/enzyme conjugates as described previously 62 . A rat monoclonal antibody against murine PECAM (clone MEC-13.3; BD Biosciences, San Jose, CA) was conjugated with catalase or SOD at 1:2 IgG to enzyme molar ratio, respectively. The effective diameter of the obtained conjugates was measured by a dynamic light-scattering apparatus (90Plus Particle Sizer). Conjugates in 7% sucrose were frozen by liquid nitrogen and stored at −80 °C until use. Recombinant mouse Mice. KRIT1 heterozygous (Krit1 +/− ) mice were obtained from Dr. Doug Marchuk (Duke University). These mice are on a clean C57BL/6 background after having been backcrossed 10 generations to C57BL/6NCrl. A conditional endothelial KRIT1 knockout mouse (PDGFBiCreER T2 Krit1 flox/flox ) was obtained from Dr. Dean Li (University of Utah). In these mice, tamoxifen injection stimulates PDGFB promoter-driven Cre-ER activity, causing the deletion of exons 4-8 of KRIT1. In this study, PDGFBiCreER T2 Krit1 flox/flox , and PDGFBiCreER T2 Krit1 flox/+ animals were treated with tamoxifen at P0 as in ref. 6 to yield endothelial KRIT1 null (Krit1 ECKO ) and heterozygous (Krit1 +/− ) control mice. Mouse lung tissue was harvested from both male and female mice at 4 months. Lung tissue was used instead of cremaster muscle as this tissue is enriched in endothelial cells, thus eliminating the overwhelming contribution of mRNA from skeletal muscle cells present in the cremaster. All mice were bred and maintained under standard conditions in the University of Rochester animal facility, which is accredited by the American Association for Accreditation of Laboratory Animal Care. All protocols were approved by the University of Rochester Committee on Animal Resources and all experiments were performed in accordance with relevant guidelines and regulations.
Intravital Microscopy. Mice were used between 8 and 12 weeks of age; littermate controls were used in all experiments. Male wild type (Krit1 +/+ ) or KRIT1 heterozygous (Krit1 +/− ) mice were anesthetized with sodium pentobarbital (65 mg/kg i.p.) and maintained on supplemental anesthetic via a jugular catheter. Preparation and visualization of the cremaster muscle, as well as measurement of permeability to Alexa-488 BSA was performed as previously described 2 . The vascular wall P s (cm/sec) was calculated from the measured flux per unit microvessel surface area at a known BSA concentration gradient 63 , with corrections for the source volume and surface area due to the confocal slice 64 . As ROS regulate arteriole tone, which has profound effects on permeability, we excluded from our analysis vessels that had a significant change in diameter during the measurement, which comprised <5% of all imaged vessels. A 1:1 mixture of anti-PECAM conjugated SOD and anti-PECAM conjugated catalase (100 µg each in 100 µl) was injected via tail vein 1hr before surgery. Yeast-avenanthramide I (45 mg/kg) and Tranilast (40 mg/kg) were administered by oral gavage in 15% ethanol 1hr prior to imaging. Images were acquired from at least 10 vessel sites, which are the appropriate determinant for n in these experiments. Data were given as arteriole and venule means +/− SEM.
Non-targeting negative control siRNA #1 and anti-KRIT1 siRNA (AM16708, Ambion/Invitrogen) were used as reported previously 1 . Knockdown efficiency was measured in each experiment by qPCR detection of KRIT1, as indicated in the appropriate figures. Transfection efficiencies ranged from 80-95% based on transfection of fluorescently labeled siRNAs (data not shown). Activity of this anti-KRIT1 siRNA against human protein has been reported previously 1 . Co-transfection of siRNA and cDNA was performed using Amaxa nucleofection, as previously described 4 .
RNA isolation and semi-quantitative RT-PCR. RNA was isolated using an RNeasy Purification Kit (Qiagen) or Trizol extraction according to the manufacturer's instructions. Complementary DNA was obtained with Transcriptor reverse transcriptase (Roche, Indianapolis, IN) using oligo dT primers. Amplifications were run in a 7000 real-time PCR system (Applied Biosystems/ThermoFisher) using mouse primer sets (Supplemental Information Table 1, and ref. 4) or Taqman gene expression assays (Krit1, Hs00184988; Gapdh, Hs99999905; Nox2 (CYBB), Hs00166163; Nox4, Hs01379108; ThermoFisher). Each value was calculated using the comparative Ct method 65 and normalized to Gapdh internal control. All samples were run in at least triplicate.
Immunoprecipitation and Western Blotting. Lysates were prepared as reported previously 4 . Nox4 was immunoprecipitated using 2 µg monoclonal anti-Nox4 antibody (Millipore/Sigma). Lysates were probed with polyclonal rabbit anti-Nox4 (Novus Biologicals, Littleton CO) at a dilution of 1:1000. Control lysates were immunoprecipitated with mouse non-immune IgG (Santa Cruz, Dallas, TX). Goat anti-rabbit Dylight-680 and goat-anti rabbit 800 (ThermoFisher) were used as secondary antibodies and the membranes imaged using an Odyssey Infrared Imaging System. NF-κB p65 was detected using a rabbit polyclonal anti-p65 (C-20) antibody from Santa Cruz. Tubulin was detected as a loading control (monoclonal anti-α-tubulin, Sigma). Goat anti-rabbit and goat anti-mouse HRP were used as a secondary antibodies and the blot was developed using Luminata ™ Forte Western HRP substrate reagent, then exposed to film.
Immunohistochemistry. Immunohistochemical staining of Nox4 was performed on cross-sections of paraffin-embedded mouse brain tissue. Antigen retrieval was performed by incubation in citrate buffer at 100 °C for 10 min. Slides were treated with 3% H 2 O 2 for 5 min to remove endogenous peroxidases, then blocked for 1hr in a solution of 2% normal goat serum, 1% BSA and 0.01% Triton X-100. Rabbit Nox4 polyclonal antibody was added at a 1:100 dilution in blocking buffer and slides were incubated 4hr at room temperature in a humidified chamber. Control serial sections were incubated with an equivalent concentration of normal rabbit IgG. Donkey anti-rabbit IgG-HRP (1:250) was used as secondary antibody. Staining was developed using the DAB + liquid stain reagent (diaminobenzidine, Dako, Denmark); slides were counterstained for 30 sec. with Mayer's hematoxylin.
Endothelial Monolayer Leak Assay. HPAECs were plated onto 3μm pore polyester transwell filters coated with 10 µg/ml human plasma fibronectin (gift from Dr. Denise Hocking, University of Rochester). Cells were transfected with siRNA 24 hours after plating and grown for 72 hours at 37 °C to full confluence. Cells were then incubated in DMEM with 0.5% FBS for 2 hours, then treated with or without YAv1 (100 µM) or TNF-α (25 ng/ml) for 24hrs, or with VEGF (50 ng/ml) for 40 minutes. Horseradish peroxidase (HRP, 1.5 µg/ml, Sigma) was added to upper wells, and the plates were then incubated for an additional 2 hours at 37 °C. The HRP content of the lower chamber medium was then measured using a 3,3′, 5,5′-tetramethylbenzidine colorimetric assay as described in ref. 38. NFκB reporter. Reporter assays were performed using the Dual-Glo luciferase assay system (Promega, San Luis Obispo, CA) according to the manufacturer's instructions. Firefly luciferase activity produced by the NFκB-dependent firefly luciferase reporter plasmid pNF-κB -Luc (Stratagene, La Jolla, CA) was measured using a Synergy H4 hybrid plate reader. All transfections included the constitutive Renilla luciferase plasmid pRL-TK, which was used to normalize for transfection efficiency. As appropriate, cells were treated with 25 ng/mL TNF-α for 4 hours prior to luciferase detection. DHE/CellROX staining. Qualitative measurement of intracellular ROS levels was performed using a well-established method based on the cell-permeable ROS-sensitive fluorogenic probe dihydroethidium, which has a sensitivity for superoxide anions. Briefly, cells grown to confluence in complete medium were washed twice with PBS, incubated with DHE at a final concentration of 5 mM in PBS at 37 °C for 20 min, treated +/− 25 ng TNF-α for 30 min, then fixed and analyzed by fluorescence microscopy on an Olympus IX70 fluorescence microscope equipped with a Hamamatsu CCD camera and MetaMorph software for hardware control and image acquisition. Quantitative analyses were performed using the CellROX ® Green Reagent (ThermoFisher) following the manufacturer's recommended protocol.
Statistics. Statistical analysis, including ANOVA (with appropriate post-hoc test for multiple comparisons), t-test, and Shapiro-Wilk test for normality was performed using PRISM software (version 4.0, GraphPad Software Inc., La Jolla, CA). In vivo permeability measurements exhibit a non-Gaussian distribution, therefore these data were analyzed by ANOVA using the Kruskal-Wallis correction and Dunn's post-hoc test. All other comparisons were analyzed by ANOVA and Tukey's post-hoc test, or by two-tailed t-test as indicated in the figure legend. Significance for all analyses was set at α = 0.05.