Ethics statement

All experiments were approved by the ethical committee of the “Landesungersuchungsamt Rheinland-Pfalz” and the authority “Landesuntersuchungsamt Rheinland-Pfalz”, protocol number: “Aktenzeichen 23 177 – 07 A12-1-004 and 23 177 – 07 A12-1-005”. Principles of laboratory animal care (European laws 86/609/EEC, national laws and NIH publication No. 86-23, revised 1985) were followed. All efforts were undertaken to minimize the number of animals used and their suffering.

Preparation of COSC

Cortical organotypic slice cultures were prepared as described in detail before with slight modifications[11–13]. All efforts were made to minimize the number of animals used and their suffering. All experimental manipulations were carried out according to the European and national laws (86/609/EEC) on animal handling. In brief C57/BL6 or heterozygous EYFP-NG2 knockin P3/4 mouse pups were rapidly decapitated. The head was disinfected with a drop of ethanol 70% and the brain was quickly removed and transferred into 4°C cold medium. Beneath a benchtop microscope under laminar flow bulbi and the cerebellum were dissected. Hemispheres were carefully separated and meninges were removed with forceps. Afterwards hemispheres were cut into 350 µm thick coronal slices using a chopper. Cortices were transferred onto Millicell membrane filters (Merck Chemicals, Schwalbach, Germany) that were pre-equilibrated with medium over night at 37°C, 5 % CO₂, humidified atmosphere. Medium was exchanged 1 day after preparation, thereafter every 2-3 days. Culture medium consisted of 50% MEM HEPES GlutaMax, 25% heat inactivated horse serum, 25% Hanks balanced salt solution supplemented with 1 mmol/l magnesiumchloride, 2 mmol/l calciumchloride. Glucose was added to a total concentration of 6-9 mg/ml, pH was adjusted to 7.2. We would like to stress the fact that pH was most important for the preparation and great care should be taken to make sure the pH is correctly adjusted to 7.2. For culture purposes Millicell filter membranes were pre-incubated in 6 Well plates containing 1 ml medium per well over night. COSC were carefully examined with light microscopy for intact morphology ahead of experimental manipulation. Only COSC that displayed an intact morphology were used for experimental procedures.

5-bromo-2'deoxyuridine (BrdU) application

BrdU was applied similarly as documented elsewhere[14]. In brief COSC were exposed towards medium containing 10 µmol/l BrdU for 3 hours on the third day in vitro. Subsequently COSC filter membranes were washed gently once and medium was switched to BrdU-free medium. COSC were fixed the following day in ice cold acetone (about 10 minutes) for immunohistochemistry.

Cardiac perfusion

C57/BL6 or heterozygous EYFP-NG2[15] mice were anesthetized by intraperitoneal (i.p.) injection of ketamine (120 mg/kg bodyweight) and xylazine (16 mg per KG bodyweight). Body temperature was kept at around 37°C with temperature controlled heating pads until the stadium of surgical tolerance according to Guedel[16] was reached. This was assessed by absence of a pain reflex upon toe pinch. The thorax was carefully opened with a microsurgical pair of scissors and animals were cardially perfused via the left ventricle with ice cold Ringer solution containing heparin (1 IU/ml). After removal of blood the perfusion solution was switched to 2% PFA and perfusion was continued for another 10 minutes. Brains were collected and post-fixed in 2 % PFA for 2 hours at 4°C. 100 µm thick coronal sections were cut with a vibratome and used for immunohistochemical processing. For some analyses brains were collected after blood was removed by saline perfusion and snap frozen in tissue tek with the help of liquid nitrogen. Here, 20 µm thick coronal sections were cut with a cryotome and fixed in ice cold acetone for 4-5 minutes.

Live Cell Imaging of pericytes

P3/4 heterozygous EYFP-NG2 mouse pups were deeply anesthetized by i.p. injection of ketamine (120 mg/kg bodyweight) and xylazine (16 mg per KG bodyweight). After surgical tolerance stadium was reached (tested as documented above) 50–100 µl tomato lectin were injected into the left ventricle and after 2 minutes animals were perfused with ice cold heparinized (1 IU/ml) Ringer solution. Thereafter brains were collected for COSC procedure as afore (see above) documented. COSC were examined with an upright confocal spinning disk system at 37°C in colorless Hanks balanced saline solution maintaining 1 mmol/l magnesiumchloride, 2 mmol/l calciumchloride and 6-9 mg/ml glucose.

Immunohistochemistry

Immunohistochemical stainings were carried out according to standard procedures described in detail elsewhere[11,17]. Tables 1 and 2 give an overview on used antibodies, dilutions and purposes of the staining. In brief fixed probes were washed with 0.01 mol/l PBS. Subsequently tissue was blocked with 7% normal donkey serum (017-000-121, Dianova, Hamburg) and permeated with 0.3% triton in PBS 0.01 mol/l for two hours at room temperature. Only for BrdU staining samples were then incubated for 90 minutes at room temperature with 2 mol/l HCl. To verify the specific binding of the secondary anti-mouse antibody (used in Claudin-5 and NeuN stains) mouse IgG were blocked using a Fab (1:20, 2 hours in PBS 0.01 mol/l, AffiniPure Fab Fragment Donkey Anti-Mouse IgG (H+L), Jackson, Dianova) blocking technique. Primary antibodies were incubated in 2% bovine serum albumin (001-000-161 Diana, Hamburg) containing 0.05% azide and 0.1% triton in PBS 0.01 mol/l (overnight, room temperature). After incubation with primary antibodies probes were washed with PBS 0.01 mol/l and incubated with secondary antibodies and DAPI in 2% bovine serum albumin with 0.05% azide for another 2 hours at room temperature. After a final washing step in 0.01 mol/l PBS probes were embedded in Fluoromount. The sensitivity of the nitrotyrosine antibody was confirmed by positive controls with peroxynitrite as recommended by the manufacturer.

Elecytrophysiological Recordings in Multi-Electrode Arrays

To evaluate spontaneous neuronal network activity of COSC after five days in vitro, MEA recordings were performed. COSC on Millicell filter membranes were cut out with a small scalpel and transferred on an 8 x 8 MEA chip (multichannel systems, Reutlingen, Germany) consisting of 60 electrodes with a diameter of 30 µm and an interelectrode distance of 200 µm. To avoid dislocation of COSC, a small weight was put on top of the slices. Prior to use the chip was heated up to a temperature of 37°C. COSC on MEA chips were kept at 20% O₂ and 5% CO₂ in a humidified atmosphere. Recordings were initiated 10 minutes after placing COSC on MEA chips and lasted 30 minutes. MC-Rack software version 4.1.1 (multichannel system, Reutlingen, Germany) was used for acquisition of the data. Sigma Plot version 7.0 (Systrat Software, Erkrath, Germany) was utilized for analysis.

Image analysis

All image analysis were processed and performed using Leica Application Suite Advanced Fluorescence (Leica Microsystems), Metamorph (Molecular Devices Corp., Downington, CA, USA), NIH Fiji is just ImageJ and IMARIS imaging software (Bitplane, Switzerland). Caspase-3 positive pericytes were identified by visual identification of PDGFR beta positive pericytes and cleaved caspase-3 co-localization. The ratio of pericytes positive for caspase-3 related to all pericytes was calculated within randomly chosen fields of view (RFV) of confocal images in COSC in layers II-IV. Images were obtained with a Leica SP5 confocal. The ratio of groups treated with IL-1 beta, hypoxia or both stimuli was set in relation to the mean ratio of respective control groups. Relative changes in fold change are presented. For this purpose 7-41 COSC preparations from 6-28 animals (up to 2812 pericytes) were analyzed for each condition. Microvascular pericyte coverage was determined by measuring total vessel length with the Fiji measurement tool in pixels similar as described before[11]. Total pericyte length was related to vessel length and relative microvascular pericyte coverage was obtained. For evaluation of caspase-3 inhibition on pericytes loss, pericytes in RFV of cortical layers II-IV were identified by DAPI, pan-laminin and PDGFR beta co-staining and the total number of pericytes per volume was determined via confocal z-stacks. Here the minimum spatial resolution was: 1.5 µm X * 1.5 µm Y * 2 µm Z voxel size. Values were related to control groups and differences in fold change are displayed. At least 3 COSC per group were analyzed.

If not stated otherwise all images shown in this manuscript are confocal images. 3 dimensional reconstructions of z-stacks were performed with IMARIS imaging software.

Image acquisition

We used a Leica SP5 confocal microscope for image analysis. Excitations were 405 nm, 488 nm, 561 and 633 nm. In addition we used an upright microscope with confocal spinning disk system (QLC10 Visitech, Sunderland, UK) equipped with a temperature controlled chamber for live cell imaging (excitation: 488 nm, 568 nm).

Induction of hypoxia and inflammation, DPI and Z-DEVD-fmk treatment

Hypoxia was induced by placing COSC membranes into medium which was bubbled with a gas mixture of 8% O₂, 5 % CO₂, rest N₂. Then probes were kept in a hypoxic incubator for 24 hours. Atmosphere in the incubator was adjusted to reach levels of about 75 mmHg O₂ in culture medium. O₂ levels were monitored ahead and after hypoxia using an Oxylite pO₂ sensing probe (Oxford Optronics, Oxford, UK). For inflammation probes were incubated for 24 hours with medium containing 10 or 100 ng/ml Interleukin 1 beta (IL1B R & D Systems, Catalog number 401-ML). The inhibitor DPI was preincubated for 1 hour at a concentration of 50 µmol/l ahead of experimental manipulation. We chose this concentration because we found DPI to be protective at 50 µmol/l on brain endothelial cells during moderate hypoxic conditions (Zehendner et al. unpublished data). In addition DPI was successfully used at 50 µmol/l by other groups[18]. To elucidate impact of caspase-3 inhibition on pericyte cell death cells were treated with Z-DEVD-fmk (20 µmol/l) during 24 h of moderate hypoxic and inflammatory (100 ng/ml IL1B) conditions. Respective controls were treated with solvent (PBS, DMSO <0.1%) only. Hypoxic and / or inflammatory treatment was started on DIV 2 and COSC were fixed for analyses on DIV 3.

Statistics

Results are documented as Mean + standard error of the mean (SEM). All data were analyzed using Graphpad Prism for Windows (version 4.02, Graphpad, San Diego, CA, USA). Datasets were checked for normalization by using D’Agostino and Pearson omnibus test. Groups that passed the D’Agostino and Pearson omnibus test[19] were therefore analyzed by two-sided unpaired student’s t-test. Groups that did not pass the test for normalization were analyzed with a two sided Mann-Whitney U test. Differences were regarded to be significant at alpha < 0.05.

Perivascular cells within cortical organotypic slice cultures

In the CNS the PDGFR beta receptor is exclusively expressed by pericytes and therefore represents a sensitive pericyte marker [4,20]. To identify perivascular PDGFR beta positive cells in COSC we performed immunohistochemical analyses with different vessel markers (CD105, Claudin 5), PDGFR beta and DAPI. Our COSC preparations displayed a well preserved cytoarchitecture of cortical layers after 5 days in vitro (DIV) as demonstrated by DAPI staining (Figure 1 A). The transmembraneous tight junctional protein Claudin 5 (Cl-5) is expressed by brain endothelial cells [17] in COSC and is crucial for blood brain barrier integrity in vivo [21] and in vitro [12,17]. Cl-5 staining revealed a high number of Cl-5 positive microvessels within COSC that were in close contact with PDGFR beta positive cells (Figure 1 B). Another protein expressed by pericytes is NG2. However NG2 is not solely restricted to pericytes but also expressed e.g. by oligodendrocyte precursor cells [22]. Co-stainings of PDGFR beta and NG2 demonstrated a partial co-localization of both pericyte markers (arrowhead Figure 1 C) that was confirmed by 3 D reconstructions (Movies S1, S2). Pericytes in native P3 mice (Figure 1 D) demonstrated a similar perivascular localization as found in COSC (Figure 1 E, insets), demonstrating that our in vitro model recapitulates the anatomical in vivo situation. PDGFR beta positive cells appeared as cells with a DAPI positive cell soma, their processes following microvessels (Figure 1 D, E insets). Note that the cell nucleus of a pericyte is mostly spared by PDGFR beta staining because PDGFR beta is a tyrosine kinase situated in the membrane of a pericyte, not in the nucleus [22]. The pericyte nucleus is surrounded by PDGFR beta staining that continues along pericyte processes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Pericytes in cortical organotypic slice cultures.

Microvessels and pericytes in COSC were immunolabelled with different vascular markers (CD105, Cl-5), the pericyte marker PDGFR beta and DAPI. COSC preparations displayed a well preserved cytoarchitecture of cortical layers after 5 DIV (DAPI, epifluorescent image, A). A high number of microvessels covered by pericytes remained in COSC (epifluorescent image, B). PDGFR beta (green) partly co-localized (yellow) with the proteoglykan NG2 (red) in COSC which is another pericyte marker (arrowhead, confocal Z-stack with orthogonal section views, C). Similar to the perivascular localization of PDGFR beta positive cells in newborn mice at the age of postnatal day 3 (D) we observed PDGFR beta positive cells to be in close contact with microvessels in COSC after 5 DIV (E). Note the DAPI positive pericyte cell soma in which the nucleus is mostly spared by PDGFR beta staining (marked by asterisks) and its processes that follow the microvessel (D, E insets). Further the pericyte marker Desmin was found to be expressed by perivascular PDGFR beta positive cells in COSC (Panel F) as well as in native P3 mice (Panel G). Pericyte coverage in cortical microvessels was not significantly different in native P3 mice compared with COSC (H).

https://doi.org/10.1371/journal.pone.0081637.g001

Another antigen reported to be expressed by pericytes is Desmin [22]. Desmin positive cells were found to co-localize with PDGFR beta next to cortical microvessels in COSC (Figure 1 F). Desmin also co-localized with PDGFR beta in native P3 mice (Figure 1 G). In addition microvascular pericyte coverage, a hallmark for neurovascular integrity [4], was not significantly different at 5 DIV within COSC compared with native P3 mice (COSC DIV5: 0.9 + 0.02 vs. P3: 0.89 + 0.02; n = 93 - 101 microvessels [>10 COSC preparations, 6 native P3 mice], P = 0.7821, Figure 1 H).

The neurovasular unit in the novel in vitro model

Immunohistochemical quadruple stainings show that in the presented in vitro approach almost all of the cell types of the neurovascular unit are present and display a preserved morphology in COSC after 3 DIV. Using high power imaging of cortical microvessels, pericytes and astrocytes we found that the endothelium is enwrapped by pericytes that are further in close contact with astrocytes (Figure 2 A, Movie S3). NeuN immunoreactivity revealed the presence of neurons with an intact cellular morphology within the COSC preparation (Figure 2 B). It has been reported that the basement membrane (BM), which consists of laminins, proteoglykans, nidogens as well as collagen IV [23], is of high relevance for neurovascular integrity [24,25]. Therefore we were interested in the question if a BM is present in the cortical neurovascular in vitro model. We found that cortical microvessels and pericytes in COSC possess a BM as demonstrated by pan-Laminin stainings (antibody that binds to Laminin chains alpha-1, beta-1, alpha-2 and gamma-1, and thereby many laminin isoforms that contain at least one of these chain types) in COSC [24,26]. Laminins are heterotrimers composed of an alpha, beta and gamma chain [23]. They are a hallmark for a proper formation of a BM and therefore are a marker of the BM in microvessels in the CNS [25,27]. We observed a preservation of the BM in the neurovascular unit of our in vitro preparation after 4 DIV (representative images from more than 10 COSC preparations, Figure 3 A). High power confocal images revealed that cortical microvessels and pericytes are embedded in the BM (Figure 3 B, Movie S4). Animated three-dimensional reconstructions with laser scanning confocal imaging confirmed these findings on an intact NVU (Movies S3, S4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. The neurovascular unit in COSC.

Immunohistochemical stainings show a preserved morphology of astrocytes and pericytes that cover cortical microvessels in COSC (A). Stainings with the neuronal marker NeuN demonstrated a preserved neuronal morphology within the neurovascular unit of COSC (B).

https://doi.org/10.1371/journal.pone.0081637.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Basement membrane in the neurovascular unit of COSC.

Co-labeling of cortical microvessels (Cl-5, green), pericytes (PDGFR beta, red) and laminins with a pan-Laminin antibody (white) reveals the presence of a basement membrane (BM) in the neurovascular unit in COSC after 4 DIV (confocal Z-stacks, maximum projection A). High power confocal magnification visualizes the BM (arrowheads, B) that encloses microvessels (Cl-5, green) and pericytes (PDGFR beta, red). Note the DAPI positive cell nuclei of the Cl-5 positive microvessel (asterisk) and the PDGFR beta positive pericyte (marked by #).

https://doi.org/10.1371/journal.pone.0081637.g003

Live cell imaging of pericytes in EYFP-NG2 knockin mice

To address the question whether the in vitro model can be used for live cell imaging of pericytes, we labeled cortical microvessels of P4 heterozygous EYFP-NG2 knockin mouse mutants with tomato lectin. This mouse lineage carries an EYFP label inserted in exon 1 of the NG2 gene [28]. Here, in line with our immunohistochemical NG2 analyses, the EYFP-NG2 signal was partly present next to cortical microvessels in living COSC (Figure 4 A). However other NG2 cells were not associated with microvessels indicating that they are not pericytes. Pericytes could be identified through the perivascular EYFP-NG2 signal for up to 3 days after COSC preparation (Figure 4 B). To verify that the obtained perivascular EYFP-NG2 signals were of pericyte origin we performed immunohistochemical co-stainings of EYFP-NG2 cortices with PDGFR beta, NG2 and Cl-5, which demonstrated a co-localization of EYFP-NG2 and PDGFR beta in a perivascular manner (Figure 4 C, D, insets). Co-stainings with NG2 and Cl-5 confirmed this observation (Figure 4 E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Live cell imaging of pericytes in COSC from EYFP-NG2 mice.

Lectin stained microvessels in COSC appeared as red labeled vascular structures that were surrounded by EYFP-NG2 expressing pericytes (A, insets). Pericytes could be identified for up to 3 DIV 3 (B). Co-stainings with PDGFR beta (C, D), Cl-5 and NG2 demonstrate that perivascular EYFP-NG2 expressing cells are indeed pericytes (E, insets).

https://doi.org/10.1371/journal.pone.0081637.g004

Pericytes in COSC are capable of proliferation

It has recently been shown that pericytes from human origin can be reprogrammed into induced neuronal cells by the use of vesicular stomatitis glycoprotein-pseudotyped retroviruses to transduce certain differentiation factors e.g. Mash1 and Sox2 [29]. However, this process is crucially dependent on the capacity of pericytes to proliferate because a successful transduction via retroviral vectors (except for lentiviruses) is dependent on the ability of the targeted cells for cell division. We found that BrdU (exposition for 3 hours on DIV 3, 10 µmol/l) was incorporated by some pericytes within 24 hours after exposure (Figure 5 A, Movies S5, S6). In addition some pericytes were positive for Ki67 (Figure 5 B), a marker for cell proliferation, after 4 DIV [30]. Ki67 and BrdU stainings were performed in more than 4 COSC preparations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Pericytes in COSC are capable of cell division.

Confocal analyses revealed that BrdU (3 hours exposition, 10µmol/l) was incorporated by pericytes within 24 hours on DIV 4 (arrowheads and asterisks in A mark a pericyte cell nucleus positive for BrdU). Ki-67 is a marker for cell proliferation. Here, a pericyte cell nucleus immunoreactive for Ki-67 on DIV 4 is shown (arrowheads, asterisks in B).

https://doi.org/10.1371/journal.pone.0081637.g005

Neuronal network activity and longterm survival of pericytes in vitro

To elucidate if COSC maintain spontaneous network activity during culture conditions (5% CO₂, 20% O₂, rest N₂, humidified atmosphere, 37°C) multi-electrode array (MEA) recordings were performed as described previously with slight modifications (Heck et al., 2008). Here we observed spontaneous synchronized neuronal network activity (Figure 6 A, representative image from n = 3 independent recordings, DIV 5). Pericytes could be identified in COSC for up to 3 weeks. Figure 6 B shows a pericyte with characteristic perivascular morphology after 14 days in vitro.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Spontaneous neural network activity and longterm persistence of pericytes in COSC.

MEA recordings demonstrated that spontaneous synchronized neural network activity is preserved in COSC under culture conditions on DIV 5 (A). In addition pericytes were found to be present in COSC for weeks. Here, a 2 week old pericytes situated next to a Cl-5 positive microvessel is shown (B).

https://doi.org/10.1371/journal.pone.0081637.g006

Pericytes express elevated levels of caspase-3 after inflammation and hypoxia

Chronic hypoxia and inflammation have been proposed to be key factors leading to preterm brain injury [31]. In addition we and others have shown that caspase-3 is involved in neuronal and BBB pathology during ischemia and inflammation [17,32,33]. Therefore we were interested in the question if inflammation induced by interleukin 1 beta (IL1B), a cytokine which has been shown to disrupt proper white matter formation in the developing brain [34], and moderate hypoxia result in caspase-3 activation in pericytes. Thus we subjected COSC towards prolonged moderate hypoxia and different concentrations of IL1B. The cultures were exposed to a medium maintaining 71 + 2 mmHg pO₂ for 24 hours. This pO₂ level is about 58% of control medium kept in normoxic cell incubators (hypoxia 71 + 2.2 mmHg vs. normoxia 123 + 0.5 mmHg pO₂, P < 0.0001, n = 15-48 measurements).

IL1B treatment resulted in significantly higher amounts of caspase-3 positive pericytes (representative image of a caspase-3 positive pericyte treated for 24 hours with IL1B 100 ng/ml, Figure 7 A) after 24 hours compared with control (control: 1 + 0.1 vs. IL1B 10 ng/ml 1.42 + 0.15, P = 0.0269; control: 1 + 0.1 vs. IL1B 100 ng/ml 1.98 + 0.19, P < 0.0001, n = 7 - 8 COSC preparations per group [535-668 pericytes], Figure 7 B). We detected significantly more caspase-3 positive pericytes in COSC treated with IL1B 100 ng/ml than in those treated with 10 ng/ml (IL1B 10 ng/ml 1.42 + 0.15 vs. IL1B 100 ng/ml 1.98 + 0.19, P = 0.0237, n = 8 COSC per group [535 - 668 pericytes], Figure 7 B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. IL1B and moderate hypoxia induce caspase-3 cleavage in pericytes independently from peroxynitrite.

Exposition of COSC towards pathologic conditions led to an increase of cleaved caspase-3 positive pericytes as shown in a representative confocal image in panel A. Stimulation of COSC with different concentrations of IL1B resulted in significantly elevated levels of cleaved caspase-3 in pericytes (B). Furthermore 24 hours of moderate hypoxia and a combination of IL1B and 24 hours of moderate hypoxia resulted in caspase-3 cleavage in pericytes (C). Of note, combination of IL1B and hypoxia did not significantly increase caspase-3 cleavage compared with probes treated with 24 hours of moderate hypoxia alone. Immunohistochemical stainings for nitrotyrosine did not reveal any significant formation of this ROS reaction product in pericytes (D, E). The NADPH oxidase inhibitor DPI [50µmol/l, 1 h pre-incubation] was not able to block caspase-3 cleavage under moderate hypoxic conditions. *P<0.05, ***P<0.001.

https://doi.org/10.1371/journal.pone.0081637.g007

Moderate hypoxia for 24 hours resulted in a significantly higher amount of cleaved caspase-3 positive pericytes (control: 1 + 0.05 vs. hypoxia: 2.48 + 0.23, P < 0.0001, n = 11-41 COSC preparations [571 - 2800 pericytes], Figure 7 C). Exposure of combined hypoxia and IL1B 100 ng/ml significantly elevated cleaved caspase-3 levels in pericytes (control: 1 + 0.05 vs. yhpoxia+IL1B: 2.38 + 0.15, P < 0.0001, n = 18-41 COSC preparations [677-2800 pericytes], Figure 7 C). However, a combination of IL1B and 24 h of moderate hypoxia did not increase caspase-3 cleavage compared with 24 h of moderate hypoxia alone (hypoxia: 2.48 + 0.23 vs. hypoxia+IL1B 2.38 + 0.15, P = 0.9763, n = 11 - 18 COSC preparations [571 - 677 pericytes], Figure 7 C).

It has been suggested that oxidative stress in form of reactive oxygen species (ROS) plays a major role in immature brain damage [31]. Therefore we were interested in the question if ROS may contribute to the observed enhancement of caspase-3 cleavage in pericytes after 24 h of hypoxia and whether an inhibition of oxidative stress may prevent caspase-3 cleavage. Stainings for nitrotyrosine, a product that arises from the reaction of the ROS peroxynitrite and the amino acid tyrosine [35] did not show any significant levels of nitrotyrosine in pericytes (representative image from at least 3 COSC preparations, Figure 7 D, E). The NADPH oxidase is a major source of reactive oxygen species [36] in cerebral ischemia and diphenyliodonium chloride (DPI) is a widely used inhibitor of the NADPH oxidase [18]. We therefore tested if a pharmacological inhibition of NADPH oxidase by DPI has an effect on blockade of caspase-3 cleavage in pericytes during hypoxia. We found that DPI was not able to block caspase-3 cleavage in pericytes. Despite DPI treatment caspase-3 levels in the hypoxic group were significantly higher than in the normoxic control (hypoxia+DPI: 3.049 + 0.16 vs. normoxic control: 1 + 0.07, P < 0.0001, n = 24 COSC preparations per group [1597-2812 pericytes], Figure 7 F).

Caspase-3 inhibition ameliorates pericyte loss in hypoxia and inflammation

Because we observed enhanced levels of cleaved caspase-3 in pericytes upon ischemia and inflammation we were interested in the question if caspase-3 activation may result in pericyte loss. To address this question COSC were treated with the selective and irreversible caspase-3 inhibitor Z-DEVD-fmk [17] during hypoxia and inflammation and pericyte numbers in confocal z-stacks were compared with solvent treated normoxic and non-inflammatory control COSC. We found that hypoxia resulted in a significant decrease of pericytes/volume compared with controls (control: 1 + 0.16 vs. hypoxia 0.31 + 0.02, P = 0.0034, n = 5 - 6 RFV from more than 3 COSC). After IL1B (100 ng/ml) treatment pericytes were also significantly reduced in cortical layers II-IV (control: 1 + 0.16 vs. IL1B 0.46 + 0.05, P = 0.0142, n = 5-6 RFV from more than 3 COSC preparations per group). COSC that were treated with Z-DEVD-fmk sustained significantly higher pericyte numbers after hypoxic (hypoxia 0.31 + 0.02 vs. hypoxia/Z-DEVD-fmk 0.52 + 0.08, P = 0.0324, n = 5 RFV from more than 3 COSC) and inflammatory conditions compared with solvent treated groups (IL1B 0.46 + 0.05 vs. IL1B/ZDEVD-fmk 0.95 + 0.16, P = 0.0406, n 5 - 8 RFV from more than 3 COSC preparations per group). However Z-DEVD-fmk was not able to completely block pericyte loss under hypoxic conditions (control: 1 + 0.16 vs. hypoxia/Z-DEVD-fmk 0.52 + 0.08, P = 0.0301, n = 5 - 6 RFV from more than 3 COSC).