Differential responses of blood-brain barrier associated cells to hypoxia and ischemia: a comparative study

Undisturbed functioning of the blood–brain barrier (BBB) crucially depends on paracellular signaling between its associated cells; particularly endothelial cells, pericytes and astrocytes. Hypoxic and ischemic injuries are closely associated with disturbed BBB function and the contribution of perivascular cells to hypoxic/ischemic barrier regulation has gained increased attention. Regardless, detailed information on the basal hypoxic/ischemic responses of the barrier-associated cells is rare and the outcome of such cell-specific responses on BBB modulation is not well understood. This study investigated crucial parameters of hypoxic/ischemic adaptation in order to characterize individual perivascular cell responses to stress conditions. The brain microvascular endothelial cell line RBE4 (EC cell line) as well as primary rat brain endothelial cells (ECs), pericytes (PCs) and astrocytes (ACs) were exposed to 24 and 48 hours of oxygen deprivation at 1% and 0.2% O2. All primary cells were additionally subjected to combined oxygen and glucose deprivation mimicking ischemia. Central parameters of cellular adaptation and state, such as HIF-1α and HIF-1 target gene induction, actin cytoskeletal architecture, proliferation and cell viability, were compared between the cell types. We show that endothelial cells exhibit greater responsiveness and sensitivity to oxygen deprivation than ACs and PCs. This higher sensitivity coincided with rapid and significant stabilization of HIF-1α and its downstream targets (VEGF, GLUT-1, MMP-9 and PHD2), early disruption of the actin cytoskeleton and metabolic impairment in conditions where the perivascular cells remain largely unaffected. Additional adaptation (suppression) of proliferation also likely contributes to astrocytic and pericytic tolerance during severe injury conditions. Moreover, unlike the perivascular cells, ECs were incapable of inducing autophagy (monitored via LC3-II and Beclin-1 expression) - a putative protective mechanism. Notably, both ACs and PCs were significantly more susceptible to glucose than oxygen deprivation with ACs proving to be most resistant overall. In summary this work highlights considerable differences in sensitivity to hypoxic/ischemic injury between microvascular endothelial cells and the perivascular cells. This can have marked impact on barrier stability. Such fundamental knowledge provides an important foundation to better understand the complex cellular interactions at the BBB both physiologically and in injury-related contexts in vivo.


Background
Brain neurons require a stable environment and high nutrient supply for proper functioning [1]. In this regard, the well-coordinated teamwork of the cells at the blood-brain barrier (BBB) is indispensable. Endothelial cells that line barrier vessels have intimate contact with supporting pericytes and astrocytes, as well as microglia, neurons and a well-defined extracellular matrix to fulfill their gatekeeper function at the BBB, crucially controlling nutrient supply and homeostatic balance [2]. It has become very apparent that these endothelial cells depend on inductive signaling of the surrounding cell types to fully develop and maintain their unique barrier phenotype [3,4]. Recently the important contribution of perivascular astrocytes and pericytes to BBB induction and maintenance has received increasing attention (for review see [5]).
Hypoxia (reduced tissue oxygenation) severely impairs BBB function in vivo and in vitro [6][7][8] causing pathophysiological changes including disturbed energy balance and water/ion homeostasis [5], inflammatory events and leakage of blood-proteins into the brain [5]. Similarly ischemia, being characterized by impaired nutrient and O 2 supply, also results in significant BBB breakdown [9,10]. Clearly, during injury conditions specific responses of astrocytes and pericytes impact BBB maintenance and brain function. Both astrocytes and pericytes have been shown to support barrier function via tight junction retention and reduction of endothelial cell death under hypoxic conditions [6,[11][12][13][14]. On the contrary, both cell types up-regulate barrier permeabilizing factors in response to hypoxia/ischemia such as vascular endothelial growth factor (VEGF), cytokines, and matrix metalloproteinases (MMPs) [15]. Currently we do not understand well how, or from where, the differential signals that alter BBB characteristics originate, although some evidence suggests that injury severity and duration determines the functional outcome of hypoxic/ischemic insults on BBB tightness [5,16,17]. In this regard hypoxia-inducible factor 1 (HIF-1), a central regulator of hypoxic/ischemic responses, is of particular importance. HIF-1 constitutes a heterodimeric transcription factor composed of the hypoxia-inducible HIF-1α subunit and the constitutively expressed HIF-1β subunit [18]. HIF-1 mediates hypoxic adaptation through transcriptional control of more than 100 genes participating in angiogenesis, proliferation and cellular metabolism to reduce O 2 consumption and increase O 2 supply but can also trigger cell death [18,19]. Notably, the ability of different cell types to adapt to hypoxic/ischemic stress varies considerably. Neurons for example are very sensitive to such insults, whereas other brain cells such as astrocytes are far more resistant [1]. Our previous study showed that astrocytes and pericytes differentially modulate BBB permeability at varying severities of hypoxic insults [6], implying that insult severity likely triggers individual cell-specific responses with subsequent consequences on barrier modulation.
To gain better knowledge of cell-specific behavior in terms of overall tolerance as well as potential signaling to surrounding cells, we directly compared basal responses of individual BBB associated cells (brain microvascular endothelial cell line and primary rat endothelial cells (ECs), pericytes (PCs) and astrocytes (ACs) to hypoxic and ischemic insults in vitro. We analyzed different parameters that influence hypoxic and ischemic responses including HIF-1α stabilization and target gene expression, actin cytoskeletal alterations, proliferation, and cell survival. Notably the responsiveness and tolerance of the three cell types differed considerably and likely underlies their ability to maintain and/or modulate BBB integrity.

Material and methods
Cell culture and isolation of primary cells All cell culture media and reagents were obtained from Gibco® (Life Technologies, Zug, Switzerland). A rat brain endothelial cell line, RBE4 (EC cell line) [20] was cultivated from passage 35-45 in 1:1 α-MEM/Ham's F-10 medium mixture supplemented with 10% FBS, 300 μg/ml geneticin, and 1 ng/ml basic fibroblast growth factor (Pepro Tech, Rocky Hill, NJ, USA) on rat tail collagen coated petri dishes. Rat tail collagen was isolated as described in Roll et al. [21]. Primary rat brain endothelial cells (primary ECs) were isolated from 8-10 week old male Wistar rats according to Coisne et al. [22], with slight modifications. Isolated cortices were homogenized after removal of meninges in a 40 ml Dounce homogenizer. Microvessels were isolated from the homogenate by adding an equal volume of 30% Dextran solution and subsequent centrifugation at 3000 g for 25 min at 4°C. The microvessels were then filtered through a 60 μm nylon mesh to remove larger vessels and digested in HBSS buffer supplemented with 10 mM HEPES, 0.1% BSA, 2 mg/ml collagenase-dispase, 10 μg/ml DNAse I and 147 ng/ml TLCK, for 45 min at 37°C in a water bath. Digested microvessels were re-suspended in endothelial media (DMEM Glutamax II supplemented with 10& calf serum, BME amino acids, vitamin solution, 2 mM L-glutamine, 1 ng/ml bFGF, 50 μg/ml gentamycin sulfate) and plated on collagen-IV (Sigma-Aldrich)-coated culture dishes. One day after isolation primary ECs were treated with 3 μg/ml puromycin-containing media for 24 h to prevent pericyte contamination and then used in experiments on day 5 after isolation (i.e. without passaging). Primary rat astrocytes (ACs) were isolated from neonatal pups as described previously [23] and cultured in DMEM supplemented with 10% FBS and 50 μg/ml gentamycin sulfate. Primary rat brain pericytes (PCs) were isolated from male adult rats following an established protocol [24] with slight modifications. Briefly, cortices were homogenized after removal of meninges. Subsequently microvessels were isolated by centrifugation (20 min at 3000 g at 4°C) using dextran (final concentration 16%) and digested in DMEM + 0.1% Collagenase/Dispase + 125 ug/ml DNAse I (both Roche Diagnostics, Rotkreuz, Switzerland) in a water bath for 2 h at 37°C. After centrifugation (700 g, 5 min) freshly isolated pericytes were plated on rat tail collagen coated dishes and cultured in DMEM supplemented with 20% FBS, 50 μg/ml gentamycin sulfate and 2.5 μg/ml amphotericin B until confluency. Subsequent passages were maintained on uncoated dishes in astrocyte medium. For all experiments astrocytes and pericytes were used at passage 2. The purity of all primary cultures was ≥95% as analyzed by immunostaining for standard cell markers namely GFAP (astrocytes), PECAM-1 (endothelial cells), NG-2 (pericytes) and PDGFR-β (pericytes).

Hypoxic and ischemic exposures
O 2 deprivation experiments were carried out in purposebuilt hypoxic glove-box chambers (InVivO 2 400, Ruskinn Technologies, Pencoed, UK) maintained at 37°C with 5% CO 2 . O 2 concentration was constantly monitored with an internal O 2 sensor. Cells were exposed to hypoxia (1% O 2 ) and near anoxia (0.2% O 2 ) for up to 48 h. Normoxic controls were maintained at 21% O 2 , 5% CO 2 at 37°C. Ischemia was simulated in vitro by oxygenglucose deprivation (OGD). OGD exposures were carried out on all primary cells under hypoxia and near anoxia using glucose-free media.

Quantitative real-time PCR
Total RNA was isolated directly from culture dishes using TRIzol® Reagent (Life Technologies, Zug, Switzerland) according to the manufacturer`s description. One μg of RNA per sample was reverse transcribed using the ImProm-II ReverseTranscriptase kit (Promega, Dübendorf, Switzerland) according to the manufacturer's instructions.

F-actin staining and microscopy
The EC cell line was grown on rat tail collagen coated coverslips, primary ECs were grown on coverslips coated with commercially available collagen IV, ACs on gelatin-coated and PCs on uncoated coverslips until confluency. After hypoxic and ischemic exposure cells were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 in PBS and stained for filamentous actin (F-actin) using rhodamine-conjugated phalloidin (Life Technologies). Cell nuclei were counterstained with DAPI (4′,6-Diamidin-2-phenylindol). Pictures were taken using an inverted fluorescence microscope coupled to an 8-bit CCD camera (Axiocam HR, Carl Zeiss, Feldbach, Switzerland) and processed using ImageJ software.

BrdU incorporation assay
Cell proliferation was measured via BrdU incorporation using the colorimetric BrdU Cell Proliferation ELISA (Roche Diagnostics) according to the manufacturer's instructions. Briefly, 80-90% confluent primary ECs, ACs and PCs were subjected to OGD and O 2 deprivation for 24 h and 48 h. For 24 h treatments BrdU reagent was added simultaneously with onset of OGD/O 2 deprivation, whereas for 48 h exposures BrdU was added 24 h before the end of the treatment. After exposure cells were fixed for 30 min and incubated for 1.5 h at 37°C with a peroxidase conjugated BrdU antibody (anti-BrdU-POD). After three PBS washes, peroxidase substrate was added and plates incubated to reach an OD 405/520 of 0.4-0.5. The colorimetric reaction was terminated by adding 1 M H 2 SO 4 and absorbance was read at 450/620 nm using a Multiskan RC Microplate Photometer (Thermo Labsystems, Helsinki, Finland).

MTT assay
Cell viability was measured using MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. Briefly all cells were cultured in 96-well plates. After ischemic/hypoxic exposures MTT solution (Sigma-Aldrich) was added to the medium to a final concentration of 0.5 mg/ml and the plates were incubated for 1 h at 37°C. Subsequently the media was removed and cells were lysed by adding equal volume of dimethyl sulfoxide. Optical density was measured at 560/670 nm using a Multiskan RC Microplate Photometer (Thermo Labsystems).

Statistical analysis
All results are expressed as mean ± SD of three or more independent experiments. Statistical significance was assessed by one-way ANOVA for comparison within one group or two-way ANOVA for comparison between different groups using GraphPad Prism 5 software (La Jolla, CA). Bonferroni's post hoc test was used for the majority of analyses. Kolmogorov-Smirnov test was used to assess the qPCR data for normal distribution and subsequently the Kruskal-Wallis non-parametric test was used to evaluate significance. A P-value <0.05 was considered significant.

HIF-1α induction profiles in BBB associated cells
Stabilization of HIF-1α protein levels in response to O 2 deprivation is a central mechanism by which cells adapt to reduced oxygenation [25]. HIF-1α expression profiles of EC cell line, ACs and PCs were analyzed after exposure to hypoxia (HX, 1% O 2 ) or near anoxia (AX, 0.2% O 2 ) for up to 24 h and compared to normoxic baseline expression (NX + Glc) indicated by the dashed line ( Figure 1A). In addition, ACs and PCs were subjected to oxygen-glucose deprivation (OGD) at 1% and 0.2% O 2. Since the EC cell line was extremely sensitive to O 2 deprivation alone and OGD led to significant cell death, additional OGD treatment was not performed. In all cell types HIF-1α protein was rapidly stabilized within 2 to 6 h of exposure and then reduced again after 24 h ( Figure 1A). Notably the induction profiles differed across the cell types; in the EC cell line HIF-1α levels peaked already at 2 h whereas in ACs peak induction occurred at 6 h. Moreover, AC and PC HIF-1α levels remained elevated after 24 h for most exposures, whereas expression levels in EC cell line had already returned to baseline. Importantly, the EC cell line and PCs did not show any large difference in total HIF-1α expression between HX and AX at any time point and similarly OGD had no additional effect on HIF-1α stabilization in PCs compared to O 2 deprivation alone. In contrast HIF-1α expression was clearly increased in ACs depending on the severity of insult i.e. HX versus AX and O 2 deprivation versus OGD ( Figure 1A). To directly compare the amount of HIF-1α protein across the three cell types, equal amounts of protein from cells exposed to HX ( Figure 1B) or AX ( Figure 1C) in presence of glucose, were simultaneously analyzed on the same Western blot membrane. Intriguingly the EC cell line consistently showed 3-5 times higher HIF-1α expression than ACs and PCs.

Differential regulation of HIF-1 and ischemia-inducible gene expression in BBB associated cells
To further characterize the individual BBB cell responses to reduced oxygenation, mRNA expression profiles of the HIF-1 targets PHD2 (prolyl hydroxylase 2), VEGF and GLUT-1 as well as the hypoxia/ischemia-inducible MMP-9 [26] were analyzed by quantitative real-time PCR after 24 and 48 h of treatment ( Figure 2). A table of the basal normoxic threshold cycles (CT) for the individual cell types and analyzed genes is presented in Figure 2A Figure 2D). Similar to our findings with VEGF, OGD was a stronger inducer of GLUT-1 mRNA in ACs and PCs compared to O 2 deprivation alone.
Interestingly, basal expression of MMP9 mRNA (CT 33, Figure 2A) was very low in the EC cell line, unaltered after 24 h of O 2 deprivation and even decreased after 48 h ( Figure 2E). ACs and PCs showed comparable severity-dependent up-regulation of MMP9 mRNA after 24 and 48 h in presence of glucose and OGD differed only in HX-Glc where ACs showed a remarkable increase of MMP-9 compared to PCs.
Notably, tests on primary ECs for basal expression of the above targets revealed similar CT values compared to the cell line (data not shown) highlighting similar responses of the cell line and primary cells. In conclusion, these findings underline the differential regulation of hypoxic/barrier-modulating genes in cells at the BBB and the consequence of cell-specific responses.

Injury induced actin cytoskeletal rearrangements in BBB associated cells
Reduced oxygenation causes dynamic reorganisation of the actin cytoskeleton resulting in changes of cell contractility, shape and motility [27] thereby providing important information on cell state and activation. Thus we monitored alterations in F-actin organisation under different conditions via phalloidin staining, and used such changes to make conclusions of the sensitivity of our cells to injury (Figure 3). Under normoxic conditions, parallel bundles of F-actin were observed in both the EC cell line ( Figure 3A) and primary ECs ( Figure 3B) that either localized to cell-cell borders and/or formed a cortical actin rim, structures crucial for maintaining endothelial tightness [28]. In the EC cell line, within 6 h HX-and AX-mediated disruption of the cortical actin rim was observed and characterized by disassembled F-actin (asterisk, Figure 3A). After 24 h of either treatment the cortical actin rims were completely disassembled, suggesting major disruption of endothelial integrity (white arrow). Furthermore large inter-endothelial gaps were observed after 24 h of AX (red arrows, Figure 3A) and by 48 h of AX a large number of cells were detached or dead (data not shown). Using primary ECs 24 h of HX (data not shown) and AX also caused disorganisation of the actin cytoskeleton in the presence of glucose (asterisk, Figure 3B) compared to normoxic controls. In HX-Glc, this disruption was further characterised by condensation of actin near the membrane (data not shown) and such observations were exaggerated by 24 h of Ax-Glc with the formation of inter-endothelial gaps (red arrow, Figure 3B) -suggesting major disruption of endothelial integrity.
ACs and PCs have a different F-actin architecture with parallel bundles arranged across the cells in so called stress fibers [28]. In contrast to the EC cell line and EC primary cultures, both ACs and PCs proved to be very insensitive to O 2 deprivation in presence of glucose as no alterations in F-actin organization were observed after 24 h in either HX (data not shown) or AX ( Figure 3B) and even prolonged treatment for 48 h had no effect (data not shown). Although HX-Glc was ineffective, AX-Glc caused appearance of condensed and disorganized stress fibers in both cell types after 24 h of treatment ( Figure 3B) that was further pronounced after 48 h (data not shown). Importantly, the effects on ACs were less severe as only a small fraction were affected and cell shape was largely conserved after 24 h AX-Glc, whereas nearly all PCs displayed significant F-actin disruption and chaotic cell shape changes resulting in very long and thin processes ( Figure 3B). These results underline our previous observation that both EC cell line and primary cultures are sensitive to O 2 deprivation, whereas ACs and PCs can tolerate prolonged and severe O 2 deprivation.

Only ACs and PCs adapt their proliferation during O 2 deprivation and ischemia
Modulation of proliferation is a crucial factor to survive hypoxic/ischemic events [29]. Proliferation of primary ECs, ACs and PCs was measured using a BrdU incorporation ELISA. Exposure for 24 h in presence or absence of glucose did not alter the proliferation of any of the cells ( Figure 4A) although a trend towards reduced BrdU incorporation under AX-Glc was measured, albeit slightly stronger in PCs ( Figure 4A). Overall AC and PC BrdU incorporation correlated inversely with injury severity at 48 h. Indeed prolonged oxygen deprivation for 48 h significantly reduced proliferation in both groups and substantial inhibition was observed after 48 h of OGD ( Figure 4B). Again a trend of PCs being more affected by glucose deprivation than ACs was apparent. Surprisingly, OGD exposure even up to 48 h did little to suppress primary EC proliferation and similar effects were observed using the EC cell line (data not shown) suggesting ECs maintain their proliferative rate during injury conditions.

Hypoxic/ischemic exposure differentially affects metabolic activity of BBB associated cells
To assess cellular viability, we measured metabolic activity using the MTT assay that allows overall evaluation of the degree of cellular impairment. After 24 h of HX or AX exposure, the metabolic activity of all cells remained unchanged, and even moderately increased in primary ECs ( Figure 5A). Additional glucose withdrawal lowered the metabolic activity of all cells by 20-40% although AC Figure 4 Adaptation of proliferation by barrier associated cells during hypoxic and ischemic insults. Proliferation of primary ECs, ACs and PCs was analyzed using BrdU incorporation assay after exposure to NX, HX and AX in presence (solid bars) and absence (hatched bars) of glucose for 24 h (A) and 48 h (B). Data values were normalized to normoxic baseline (NX + Glc). N = 3-5. *P < 0.05, **P < 0.01, ***P < 0.001; 1-way ANOVA compared to normoxia + glucose. § § § P < 0.001; 1-way ANOVA compared to normoxia-glucose. # P < 0.05, ## P < 0.01, ### P < 0.001; 2-way ANOVA comparing + to -glucose. EC: primary endothelial cells, AC: primary astrocyte, PC: primary pericyte, NX: normoxia, HX: hypoxia, AX: near anoxia, +Glc: with glucose, −Glc: without glucose. performed slightly better in AX-Glc. In the presence of glucose, with the exception of reduced activity of AC in AX, prolonged treatment for 48 h had no effect on the different cells ( Figure 5B). Interestingly, during prolonged normoxic glucose withdrawal PCs were significantly less affected than ACs and primary ECs that lost more than 50% of their initial activity. However, 48 h of HX-Glc gave similar degrees of AC and PC metabolic impairment that was accentuated in AX-Glc. It is noteworthy that we observed the RBE4 cell line to be significantly more sensitive than primary EC with their metabolic activity severely reduced by about 40% after 24 h HX or AX exposure and up to 70% after 48h (data not shown).

Differential regulation of autophagy and cell death in BBB associated cells
The above assays highlighted interesting differences in cell status and viability during different injury paradigms. However, as these assays do not allow direct conclusion of survival and cell death pathway activation, prominent apoptotic and autophagy markers were analyzed ( Figures 6, 7 and 8). Since BNIP3 (BCL2/adenovirus E1B 19 kDa protein-interacting protein 3) is a HIF-1 target gene implicated in regulation of survival promoting autophagy, autophagy preceding cell death, as well as apoptosis and necrosis [30] its mRNA and protein expression levels were analyzed by qPCR and Western blot respectively ( Figure 6). All three cell types expressed very low basal levels and upregulated BNIP3 transcription in response to O 2 deprivation although their fold induction and temporal expression profiles differed strongly ( Figure 6A). An initial 8-10 fold increase in EC BNIP3 mRNA after 24 h of HX and AX was measured but this was down-regulated after prolonged exposure to a 3-fold induction ( Figure 6A). In total contrast, PCs and particularly ACs were highly responsive showing a severity dependent-induction that was largely evident during HX-Glc ( Figure 6A). BNIP3 protein levels in the EC cell line showed a similar trend at 24 h with a significant Figure 5 Hypoxic/ischemic exposure differentially modulates metabolic activity of EC and perivascular cells. Viability of primary ECs, ACs and PCs was measured by MTT conversion after exposure to NX, HX and AX in presence (solid bars) and absence (hatched bars) of glucose for 24 h (A) and 48 h (B). Data was normalized to normoxic baseline (NX + Glc). N = 3-5. **P < 0.01, ***P < 0.001; 1-way ANOVA compared to normoxia + glucose. § § P < 0.01, § § § P < 0.001; 1-way ANOVA compared to normoxia-glucose. # P < 0.05, ## P < 0.01, ### P < 0.001; 2-way ANOVA comparing + to -glucose. EC: primary endothelial cell, AC: astrocyte, PC: pericyte, NX: normoxia, HX: hypoxia, AX: near anoxia, +Glc: with glucose, −Glc: without glucose.
reduction after 48 h of AX ( Figure 6B). AC and PC protein levels correlated well with the qPCR induction profiles and the reduced fold-change induction of BNIP3 protein in PCs compared to ACs was also detected.
The consequences of increased hypoxic BNIP3 expression are controversial as both cell death/apoptosis promoting [30] and survival-promoting functions via induction of autophagy have been reported [31]. Cell death promoting functions of BNIP3 have been attributed to activation of Bax (Bcl-2 associated protein X) -a crucial regulator of mitochondria-dependent apoptosis that forms pores in the outer mitochondrial membrane allowing release of apoptotic proteins such as cytochrome C [32]. However, we did not observe Bax induction in any of the cells, and levels stayed constant up to 24 h of O 2 deprivation and/or OGD ( Figure 7A). Furthermore prolonged treatment for 48 h significantly reduced Bax levels in all cells suggesting that in our model BNIP3 does not induce Bax-mediated signaling. BNIP3 also regulates autophagy via the activation of Beclin-1 with subsequent formation of LC3-II [31]. Beclin-1 participates in early autophagosome formation by coordinating localization of autophagic proteins and membrane trafficking [33]. Surprisingly, similar to Bax we did not observe increased representative Western blots (lower panels). N = 4. *P < 0.05, **P < 0.01, ***P < 0.001; 1-way ANOVA compared to normoxia + glucose. § P < 0.05, § § P < 0.01, § § § P < 0.001; 1-way ANOVA compared to normoxia-glucose. # P < 0.05, ## P < 0.01, ### P < 0.001; 2-way ANOVA comparing + to -glucose treatment. EC cell line: RBE4, AC: astrocyte, PC: pericyte, NX: normoxia, HX: hypoxia, AX: near anoxia, +Glc: with glucose, −Glc: without glucose.
Beclin-1 expression in any of the cell types during injury, but a reduction in total protein levels particularly at AX-Glc treatments for ACs and PCs and at all hypoxic/ anoxic conditions for the EC cell line ( Figure 7B).
well with the effects observed for BNIP3 expression, the EC cell line had significantly down-regulated LC3-II protein levels suggesting inhibition of autophagy during oxygen deprivation ( Figure 8B). In presence of glucose, ACs and PCs only showed minor changes in LC3-II levels during O 2 deprivation with a mild but significant reduction at 48 h ( Figure 8B). However, an OGDdependent increase in LC3-II protein levels was observed in both ACs and PCs ( Figure 8B).
reason for this is that our understanding of individual BBB cellular component responses in terms of proliferation, survival, target gene induction and barrier modulation during injury is very limited. This study is the first to directly compare hypoxic and ischemic responses of all three major cells types at the BBB and to address various aspects of cellular adaptation and survival. We show that the BBB-associated cells differ considerably in sensitivity to hypoxic and ischemic insults. All our data points to the fact that endothelial cells experience O 2 and glucose deprivation more severely than the other cell types and show a higher susceptibility. Two publications by Ceruti et al. and Redzic et al. recently compared survival of primary rat ECs, ACs and PCs under OGD (0% O 2 ) conditions and similarly concluded that ECs are more susceptible than ACs and PCs [36,37]. Importantly, our data further suggests that ACs tolerate ischemic injury better than PCs. In this regard, it is noteworthy that the differential tolerance of these cells correlates well with the degree of O 2 they experience under physiological conditions in vivo. Indeed, in brain capillaries an O 2 content of roughly 8% (pO 2 58 mm Hg) has been recorded [38], whereas O 2 concentrations in the brain parenchyma are about 4% (pO 2 35 mm Hg) [39] thus based on their localization ECs are accustomed to higher physiological O 2 concentrations than PCs and ACs.
The HIF-1 signaling pathway crucially controls hypoxic/ischemic gene expression for cell survival and energy metabolism but also the expression of genes regulating apoptosis and autophagy [18,19]. Growing evidence suggests that the severity as well as the duration of an insult play a crucial role in the pathways activated [5,6,16,17]. Indeed, we observed the more sensitive ECs stabilize 3-5 times more HIF-1α protein than the relatively resistant ACs and PCs, suggesting that HIF-1 induction correlates to sensitivity and may determine functional outcome for the different cells. In addition, HIF-1α was rapidly induced and O 2 deprivation severity did not alter the amount of HIF-1α stabilized in either ECs or PCs. This "stereotype" maximal HIF-1α response of ECs and PCs to both mild and more severe insults indicates an inability to respond in a more fine-tuned manner as observed with ACs. The ECs also sense low O 2 very rapidly and respond with a strong short lasting HIF-1α stabilization highlighting a greater responsiveness to O 2 deprivation than the perivascular cells. Other studies have also shown that HIF-1α induction is strongly cell-type dependent. A study by Bracken et al. showed that whereas some cells readily stabilize HIF-1α at 5% O 2 , others require O 2 levels below 1% and that HIF-1α transactivation is also dependent on the severity of the hypoxic insults in different cell types [40]. As a direct consequence of differential HIF-1α stabilization, mRNA levels of select HIF-1 target genes were also cell-specifically regulated. Pronounced induction of VEGF (particularly during mild hypoxia) and GLUT-1 in EC compared to the other cell types correlated well with substantial HIF-1α stabilization, again underlining their strong responsiveness to hypoxic events. Interestingly, despite early HIF-1α stabilization similar to ECs, PCs did not induce VEGF expression in comparable strength. Indeed, reduced HIF-1α stabilization in PCs compared to ECs suggests the amount of HIF-1 accounts for the differential regulation of these genes. In contrast, PHD2 and BNIP3 induction levels were reduced in ECs compared to ACs and PCs although both are HIF-1 target genes, suggestive of either a cell type-specific regulation/activation or a severity-dependent regulation of HIF-1 target gene expression. In agreement it was shown previously that basal mRNA expression of hypoxia-inducible genes varies considerably amongst cell types and that cell-type specific induction of target genes occurs [41]. Similar to our conclusions, the authors correlated the amount of HIF-1α transcripts and protein abundance to the fold induction of hypoxia responsive genes and, importantly, to the susceptibility of the cells to hypoxic injury [41].
Since appropriate adaptation of energy metabolism is a crucial factor for surviving injurious events like O 2 deprivation and ischemia, transcriptional regulation of the glucose transporter GLUT-1 was of particular interest as glucose represents the major brain energy source [42]. During O 2 deprivation, glycolysis is crucial for ATP production and our observation of strong induction of GLUT-1 by EC during O 2 deprivation agrees with literature showing that cerebral vascular glucose consumption and transport increases in response to hypoxia [43]. This raises the question whether ECs are more dependent on glucose uptake to feed their ATP consumption during O 2 deprivation than ACs and PCs. Possible reasons for this can either be that PCs and ACs have different energy metabolism and/or less reliance on glucose per se due to efficient use of alternate energy sources. Interestingly GLUT-1 induction was lower in ACs than PCs underlying the exceedingly versatile metabolism of ACs, i.e. the ability to exhibit high rates of oxidative metabolism and glycolysis, as well as the use of alternate energy fuels such as glycogen or glutamate [1,44,45]. Indeed it has long been known that ACs are the major glycogen store within the brain, whereas capillary endothelial cells and PCs have only minor reserves [46,47]. Although further study is required, it is possible that PCs could employ mechanisms similar to ACs to support their resistance to hypoxic/ischemic injuries.
A more global interpretation of cellular well-being was obtained by monitoring actin cytoskeletal architecture and metabolic activity using MTT assay -both of which emphasized swift alteration and impairment of ECs. Notably, rapid injury-mediated cytoskeletal reorganisation in ECs coincided with loss of BBB function under similar conditions in our previous studies [2,6]. Interestingly, changes in cytoskeletal structure preceded metabolic suppression in all cell types but marked alterations in AC and PC organisation occurred only during prolonged OGD correlating well with cell survival and highlighting their ability to cope well with hypoxia/ischemia. Similar conclusions can be drawn from our BrdU proliferation assays showing that both cells adapted their proliferation depending on duration as well as insult severity to similar levels -likely a crucial response to reduce energy consumption and foster cell survival. In line with this, a microarray study showed that many hypoxia downregulated genes are linked to proliferation [41]. In contrast, increased AC proliferation during hypoxia and ischemia has also been reported in vivo and in vitro during acute (6 h) OGD and hypoxic exposure [48-50]. We did not observe increased AC proliferation under any condition probably due to the more chronic time points selected suggesting time-dependency for the different severities. Perplexingly, despite being more impaired than ACs or PCs, ECs maintained their proliferation regardless of the severity of injury. Maintaining high proliferation, and therefore high energy consumption, during adverse conditions only confounds cellular adaptation and has an overall negative impact on survival. Although it remains somewhat a mystery why the ECs do not suppress/reduce proliferation, a link with angiogenesis-mediated responses during oxygen deprivation could be suggested [5].
The HIF-1 target gene BNIP3 represents a point of intersection in regulation of cell viability during O 2 deprivation. For BNIP3, both pro-apoptotic/cell death promoting functions via Bax activation [30], and survival-promoting functions via autophagy-related proteins including Beclin-1 and LC3-II [31], have been observed. Our data suggest that BNIP3 does not promote Bax-mediated cell death in our model system but observations of increased LC3-II expression points towards BNIP3-mediated induction of autophagy. This observation is further supported by the resistant ACs and PCs showing strong BNIP3 expression, whereas the more sensitive EC cells did not. Unexpectedly, the expression profile of LC3-II did not match that of BNIP3, indeed we frequently observed an inverse correlation. It is notable that LC3-II is suggested to be a good indicator of the number of autophagosomes formed but not necessarily autophagic flux, as LC3-II itself is degraded during autophagosome maturation [34]. For this reason, strong activation of autophagy can lead to low LC3-II levels because of its rapid degradation, which may then be misinterpreted as a lack of autophagy. Although our results point towards activation of autophagic pathways by ACs and PCs to modulate their cell survival, additional experiments are required to fully understand the underlying mechanism. Nevertheless it is tempting to speculate that autophagy may represent an adaptive mechanism of ACs and PCs to cope with hypoxic/ischemic injury by facilitating removal of damaged mitochondria and regeneration of ATP by catabolism of existing cytoplasmic precursors [51]. It remains unclear why brain ECs do not employ similar mechanisms in response to hypoxia although it must be noted that hypoxia-mediated autophagy induction in peripheral endothelial cells has been demonstrated [52]. We hypothesize that absence of hypoxic BNIP3 induction underlies insufficient activation of survival-promoting autophagy and contributes to increased EC sensitivity to hypoxic stress.

Conclusion
This study confirms that BBB cell types vary considerably in their sensitivity to O 2 deprivation and ischemia. We showed that on different levels of cellular responses ECs are more susceptible to O 2 deprivation whereas ACs and PCs are very tolerant to sole O 2 deprivation. It is also noteworthy that although the EC cell line proved to be more susceptible to injury compared to the primary EC cells, similar overall responses and pathway signaling mechanisms were seen in both. Our data further establishes that perivascular cells are predominantly glucose responsive with ACs withstanding extreme ischemic exposure (48 h, 0.2% O 2 ) better than PCs. Overall we propose that the physiological oxygen concentrations experienced by cells underpin their sensitivity to insult. Since differential susceptibility can directly and/or indirectly impact BBB stability, a better understanding of such causative relationships is crucial to combat BBB disturbance during injury. Submit your manuscript at www.biomedcentral.com/submit