A role for endothelial cells in radiation-induced inflammation.

PURPOSE
To unravel the role of the vasculature in radiation-induced brain tissue damage.


MATERIALS AND METHODS
Postnatal day 14 mice received a single dose of 10 Gy cranial irradiation and were sacrificed 6 h, 24 h or 7 days post-irradiation. Endothelial cells were isolated from the hippocampus and cerebellum using fluorescence-activated cell sorting, followed by cell cycle analysis and gene expression profiling.


RESULTS
Flow cytometric analysis revealed that irradiation increased the percentage of endothelial cells, relative to the whole cell population in both the hippocampus and the cerebellum. This change in cell distribution indicates that other cell types are more susceptible to irradiation-induced cell death, compared to endothelial cells. This was supported by data showing that genes involved in endothelial cell-specific apoptosis (e.g. Smpd1) were not induced at any time point investigated but that genes involved in cell-cycle arrest (e.g. Cdkn1a) were upregulated at all investigated time points, indicating endothelial cell repair. Inflammation-related genes, on the other hand, were strongly induced, such as Ccl2, Ccl11 and Il6.


CONCLUSIONS
We conclude that endothelial cells are relatively resistant to ionizing radiation but that they play an active, hitherto unknown, role in the inflammatory response after irradiation. In the current study, this was shown in both the hippocampus, where neurogenesis and extensive cell death after irradiation occurs, and in the cerebellum, where neurogenesis no longer occurs at this developmental age.


Background
Cranial radiotherapy is commonly used in the treatment of childhood cancers. However, this life-saving treatment is associated with several chronic debilitating effects, including cognitive decline, endocrine dysregulation and motor dysfunction (Lannering et al. 1990;Mulhern et al. 2004a;Ellenberg et al. 2009;Redmond et al. 2013). It has been proposed that the vasculature constitutes the primary target of irradiation (IR) and that for example swelling and necrosis of endothelial cells (ECs) in turn cause secondary effects such as tissue necrosis and ischemia (McDonald and Hayes 1967). This hypothesis proposes that it is crucial to minimize damage to cerebrovascular wall elements in order to reduce late side effects after cranial IR (O'Connor and Mayberg 2000).
However, despite intense research in the field of radiotherapy, the effect on the vasculature and its potential role in downstream damage is still an open question.
The hippocampus is important for memory function and is one of only two distinct neurogenic regions possessing proliferating neural stem and progenitor cells in the adult brain (Altman and Das 1965;Eriksson et al. 1998;Curtis et al. 2007;Kuhn and Blomgren 2011). Furthermore, the hippocampus is particularly susceptible to IR since it harbors highly proliferating cells and it has been proposed that loss of these cells contribute to the cognitive deficits observed after radiotherapy (Raber et al. 2004;Kuhn and Blomgren 2011). The hippocampus is a highly vascularized area where neural stem/ progenitor cells are born in close proximity to blood vessels (Palmer et al. 2000;Monje et al. 2002). Previous work by our group has proposed that the hippocampal vasculature adjusts to the changing needs of the surrounding neural and glial tissue after IR. In addition, no IR-induced caspasedependent cell death of ECs was observed in irradiated hippocampi 6 h post-IR after a single dose of 8 Gy (Bostrom et al. 2013). These findings, together with evidence of an intact neurovascular niche 7 weeks post-IR, failed to support the hypothesis of ECs being an upstream target in the damage of hippocampal neural stem cells (Bostrom et al. 2013).
The cerebellum was originally identified as important for motor function but has recently also been linked to higher cognitive functions (Rapoport et al. 2000;Konczak and Timmann 2007;Cantelmi et al. 2008;Armstrong et al. 2010;Buckner 2013;Merchant et al. 2014). Although the cerebellum exhibits prolonged postnatal neurogenesis, it is not considered as an adult neurogenic region. Most of the interneurons in the cerebellum are formed during the first two postnatal weeks in mice (Miale and Sidman 1961) and postnatal cerebellar neurogenesis also occurs in humans until the age of 11 months (Abraham et al. 2001). Medulloblastoma, the most common malignant brain tumor in children, is typically infratentorial, that is, located in the cerebellum (Mulhern et al. 2004b). Treatment of medulloblastoma includes not only surgery and craniospinal IR, but also a local IR boost to the tumor bed (i.e. part of the cerebellum) (Fossati et al. 2009). Considering the motor deficits observed post-IR in children (Redmond et al. 2013), and its recently discovered importance for higher cognitive functions (Rapoport et al. 2000), the cerebellum is central to the acute as well as late effects observed in children after cranial radiotherapy. Furthermore, it has been shown that prevention of cerebellar injury significantly attenuated cognitive dysfunction in mice (Manda et al. 2007).
Due to the observed cognitive deficits observed following cranial radiotherapy, we chose to investigate both the hippocampus and the cerebellum. We focused on ECs to further unravel the role of the vasculature in radiation-induced brain tissue damage. ECs were isolated from the hippocampus and cerebellum at different time points after IR and analyzed for the proportion of ECs, cell cycle distribution and gene expression, including several inflammatory markers. This study provides new important insights into the specific effects of IR on ECs and their possible contribution to shortand long-term radiation-induced effects, particularly neuroinflammation.

Animals
Male C57BL/6J mice were purchased from Charles River Breeding Laboratories (Sulzfeld, Germany) and delivered on postnatal day nine (P9). The animals were housed with a 12h dark-light cycle (dark between 21:00 and 09:00) at constant temperature (24 C) and relative humidity (50-60%), with ad libitum access to food and water. All experiments were approved by the Swedish Animal Welfare Agency (Gothenburg animal ethics application no. 361/11).

Irradiation procedure
A linear accelerator (Varian Clinac 1600 C-2, Radiation Oncology Systems LLC, San Diego, CA) with 6 MV nominal photon energy and a dose rate of 2.3 Gy/min was used to irradiate mice on P14. The mice were anesthetized with an intraperitoneal (i.p.) injection of tribromoethanol (Sigma, Stockholm, Sweden), placed on a polystyrene bed in prone position (head to gantry) and irradiated with a symmetrical 2 Â 2 cm 2 radiation field. A 15 mm tissue equivalent material covered the head to obtain an even radiation dose distribution in the underlying tissue. The source to skin distance was approximately 99.5 cm and the irradiated tissue received a single absorbed dose of 10 Gy with a dose variation of ±5%. Using the LQ model (Fowler 1989) and an alpha/beta ratio of 3 for late effects in the normal brain tissue, the acute exposure of 10 Gy is equivalent to approximately 26 Gy when delivered in repeated 2 Gy fractions. This dose represents a clinically relevant dose, that is, a 48% isodose in radical radiotherapy of malignant brain tumors in children. Animals were kept on a warm bed (36 C) both before and after IR to maintain body temperature. Control animals were anesthetized but did not receive any IR.

Isolation of cells from tissue
Mice were sacrificed by decapitation 6 h, 24 h or 7 days post-IR (10 controls and 10 irradiated for each time point) for the RT-qPCR TaqMan array and cell cycle analysis (see Figure 1 for experimental design), and at 6 h post-IR (six controls and six irradiated) for the Ccl11 RT-qPCR (see Supplementary  Figure 1(A) for experimental design). Brains were rapidly removed from the cranium, followed by rinsing in sterile 0.9% NaCl. The hippocampus and cerebellum were collected by microdissection and kept in ice cold Hibernate V R -A (Gibco/ Invitrogen, San Diego, CA, USA) solution until further processed. The left and right hippocampi from the same animal were placed in a tube as one single sample and further processed together in order to obtain a single cell suspension devoid of connective tissue. Each individual sample was thoroughly minced with a razor blade and then incubated for 15 min at 37 C with a papain/protease/DNase I enzyme solution consisting of 0.01% papain (Worthington, Lakewood, NJ), 0.1% Dispase II (SigmaAldrich, Saint Louis, MO, USA), 0.01% DNase I (Wortington, Lakewood, NJ, USA) and 12.4 mM MgSO 4 diluted in HBSS without Ca 2þ and Mg 2þ (Hank's Balanced Salt Solution; Wortington). After incubation, tissue pieces were gently triturated 5-10 times with a 1000 mL pipette, followed by 10 min incubation at 37 C and gently triturated 10 times once more before centrifuged at 100g for 2 min. The supernatant was then aspirated and the cell pellet was resuspended in 1 mL 37 C warm Neurobasal A medium (NMB; Gibco/Invitrogen, San Diego, CA) supplemented with Glutamax (2 mM), B27 with vitamin A (1:50) and penicillin-streptomycin (PEST; 100 U penicillin, 100 mg streptomycin) (all from Invitrogen, San Diego, CA). This was followed by gentle trituration of the samples $15 times until no pieces were visible and 3 mL NBM with Glutamax/B27/PEST was then added to the cell suspension. After one more centrifugation at 100g for 3 min, the supernatant was aspirated and the pellet was resuspended in warm medium. This single cell suspension was transferred to Eppendorf tubes for fixation and staining procedures.

Fixation and staining of cells before flow cytometry
Immediately after isolation, the cells were fixed and stained as previously described (Hellstr€ om et al. 2007), with some modifications. The cell suspension was centrifuged at 200g for 5 min and the supernatant was aspirated before fixation for 20 min in 200 mL Fix and Perm V R Fixation Buffer A (Invitrogen/Life Technologies). Cells were washed with 1 mL RNase-free phosphate-buffered saline (PBS), centrifuged at 200g for 5 min and the supernatant was aspirated. This was followed by incubation for 20 min with primary antibodies against either cluster of differentiation 31 (CD31; monoclonal rat anti-mouse CD31, 1:250, BD Biosciences Pharmigen, Figure 1. Experimental design of the study. (A) ECs were isolated from the hippocampus and cerebellum, respectively, at 6 h, 24 h or 7 days after 10 Gy cranial IR in 2 weeks old male mice (P14). Cells were stained with an endothelial-specific antibody (CD31) and sorted with FACS. The mRNA expression profiles of control and IR animals were then evaluated using RT-qPCR. In addition, a DNA cell cycle analysis was performed on both ECs and the whole cell population. (B) Gating strategy for the sorting of ECs at 6 h, 24 h and 7 days post-IR in the hippocampus and the cerebellum. The left panel shows how the 7-AAD staining was used to gate cells from debris, the middle panel shows a negative control staining where the primary antibody was omitted and the right panel shows a representative sample stained with CD31. ECs: endothelial cells; FACS: fluorescence-activated cell sorting; IR: irradiation; P: postnatal day; RT-qPCR: quantitative reverse transcription polymerase chain reaction.
Franklin Lakes, NJ, USA) alone, or CD31 together with rabbit anti-ionized calcium-binding adapter molecule 1 (Iba1, 1:500, Wako Pure Chemical Industries, ltd., Osaka, Japan) and mouse anti-Glial Fibrillary Acidic Protein (GFAP, 1:500, Merck Millipore, Darmstadt, Germany) diluted in 200 mL Fix and Perm V R Permeabilization Buffer B (Invitrogen/Life Technologies). After another wash with PBS (as above), the cells were incubated for 20 min in the dark with the appropriate secondary antibodies diluted 1:1000 in 200 mL PBS (donkey anti-rat Alexa 488, donkey anti-mouse Alexa 488, Invitrogen/Molecular Probes, Carlsbad, CA, USA; donkey antirat Cy5, Jackson Immunoresearch Laboratories, West Grove, PA, USA; donkey anti-rabbit PE, Santa Cruz Biotechnology Santa Cruz, CA, USA) and the nucleic acid stain Hoechst 33258 (50 mg/L, Sigma, Stockholm, Sweden). Finally, the cells were washed in PBS, resuspended in 400 mL of RNase-free PBS and transferred to FACS tubes.
Isolation of endothelial cells with FACS for downstream cell cycle analysis and RT-qPCR TaqMan array CD31 þ ECs for RT-qPCR TaqMan array analysis were isolated from the hippocampus and the cerebellum at 6 h, 24 h and 7 days post-IR using a FACS Aria II equipped with a 100 mm nozzle and DIVA software (BD Bioscience, Franklin Lakes, NJ, USA). The cell population was first identified by using forward scatter (FSC; size) and side scatter (SSC; granularity). To further distinguish cells from debris, 5 mL of 7-aminoactinomycin D staining solution (7-AAD, BD Bioscience, Franklin Lakes, NJ, USA) was added to the cell suspension 10 min prior to sorting. Primary gates were thereafter set to include cells but exclude debris, based on the 7-AAD signal ( Figure  1(B), left panel). In the current study, 7-AAD was hence not used as a viability marker. As controls, amorphous secondary gates were based on cells stained with the secondary antibody only (Figure 1(B), middle and right panel). The cells were sorted with FACS Flow Sheath Fluid (BD Bioscience, Franklin Lakes, NJ, USA) directly into 2 mL sterile micro tubes. Total cell number per sorted population was recorded, resulting in an average of 46,000 ± 1200 ECs and 118,000 ± 3700 ECs for the hippocampus and cerebellum, respectively, with similar numbers observed in control and irradiated animals. Representative data was captured for the first 20,000 cells.
Cell cycle analyses were performed based on the Hoechst 33258 DNA stain acquired in linear scale during the sorting procedure of ECs. Data were analyzed using the ModFit LT software (ModFit LT for Mac, version 3.2.1., Verity Software House, Topsham, ME, USA). Debris and cellular aggregates were automatically excluded during the analysis.

Isolation of endothelial cells and microglia with FACS for downstream Ccl11 RT-qPCR
In order to validate the inflammatory gene expression by ECs observed from the RT-qPCR TaqMan array and to compare this with the gene expression by microglia, a new round of cells (triple stained with: CD31, Iba1 and GFAP) were isolated from the hippocampus at 6 h post-IR with the FACS Aria II used above. Sorted cells were either CD31 þ ECs or Iba1 þ microglia. Staining for GFAP was included to exclude possible GFAP endfeet around the ECs. Samples were progressively gated based first on size (FSC) and granularity (SSC), followed by a pulse geometry gate (FSC-H Â FSC-A) where doublets were excluded from the analysis and further gated with 7-AAD to separate cells from debris. As above, amorphous secondary gates were based on cells stained with secondary antibodies only and two cell populations were sorted according to the following: CD31 þ Iba1 À GFAP À and CD31 À Iba1 þ GFAP À .

RNA isolation and cDNA synthesis
Immediately after the isolation of cells with FACS, the cell suspension was centrifuged for 20 min at 13,400 rpm, the supernatant was completely aspirated and the cells were frozen at À80 C until further processed. Total RNA from sorted cells was purified using the RNeasy kit for formalin-fixed paraffinembedded (FFPE) tissue sections according to the protocol provided by the manufacturer (Qiagen, Hilden, Germany). Total RNA was eluted in 15 mL RNase-free water and RNA concentrations were measured using a NanoDrop Spectrophotometer (Thermo Scientific, Wilmington, DE, USA). RNA integrity was assessed by Agilent 2200 TapeStation (Agilient Technologies, Santa Clara CA, USa). For hippocampal samples, average RIN was 7, and for cerebellum samples, average RIN was 2.2. A total amount of 100 ng RNA (RT-qPCR TaqMan array) or 50 ng (Ccl11 RT-qPCR) was converted into cDNA with the SuperScript VILO cDNA Synthesis Kit (Invitrogen/Life Technologies). For the RT-qPCR TaqMan array, some samples did unfortunately not reach the wanted amount of RNA (100 ng) and for them all available material was converted to cDNA. This did not differ between the treatment groups (controls and irradiated). The generation of cDNA was performed according to the protocol provided by the manufacturer in a total volume of 20 mL, with the incubation at 42 C prolonged with an extra 30 min (total of 90 min). The cDNA was stored at À20 C until downstream RT-qPCR.

Gene expression profiles of endothelial cells with RT-qPCR TaqMan gene expression assays
Custom RT-qPCR TaqMan array cards (384-well microfluidic cards) were designed by choosing predesigned real-time PCR 5 0 nuclease TaqMan gene expression assays (Applied Biosystems/Life Technologies). The 48-format was chosen and a total of 47 target assays were selected (plus 18S as a mandatory control always included in the arrays), designed to span introns when possible. A complete list of the gene assays included in the array can be found in Supplementary  Table 1. The total amount of converted cDNA (in most cases 100 ng/sample) was mixed with RNase-free water and TaqMan Gene Expression Master Mix into a total volume of 100 mL. Acquisition of data was performed with a 7900HT fast real-time PCR system, with thermal cycling conditions of 50 C for 2 min, 94.5 C for 10 min, 97 C for 30 s and 59.7 C for 1 min. A total number of 40 cycles were run before terminating the acquisition. Samples not displaying the common exponential curve prolife were excluded from the analysis. In addition, samples from both control and irradiated samples were run as À RT negative controls to detect genomic DNA contamination. Cutoff were set at 36 cycles and missing data were given the maximum numbers of cycles plus 1 (36 þ 1). All genes on the array were evaluated as potential reference genes with the GenEx software (MultiD Analysis, Gothenburg, Sweden). Nrp1, Pdgfb, Tie1, Tjp1 and Tnfrsf1a were identified as suitable reference genes by the NormFinder algorithm for all conditions. Gene expression data were normalized using the DDCt method (Livak and Schmittgen 2001). Data are presented as log2 fold change/ relative mRNA expression in irradiated samples compared to control samples at each respective time point and structure.

Validation of the gene expression of Ccl11 in endothelial cells and microglia
Primers for the reference gene (Tie1) and the gene of interest (Ccl11) were designed using PrimerBank (https://pga.mgh.harvard.edu/primerbank/) and tested for unintended targets on Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primerblast/). Finally, Primer3 (http://primer3.ut.ee/) were used to evaluate the probability for self-complementation and hairpin structures. All primers were designed to span introns and À RT negative controls were run to detect possible genomic contamination. Primer sequences are shown in Supplementary Table 2. PCR efficiency was calculated from a dilution curve and ranged between 95% and 100%. Melting curves were analyzed with the Roche LightCycler 480 Software 1.5 (Roche Life Science, Indianapolis, IN, USA).
SYBR-green based RT-qPCR analyses were performed using the Maxima SYBR green/ROX qPCR kit (Thermo Scientific, Waltham, MA, USA) in 96-well plates (Roche Life Science, Indianapolis, IN, USA). To each 20 mL reaction; 10 mL SYBR green, 4.88 mL nuclease-free-water and 0.6 mL of forward and reverse primer (0.3 mM) respectively, was added together with 5 mL of cDNA (6.25 ng). RT-qPCR was performed in a Roche LightCycler 480 (Roche Life Science) with the following program: 95 C for 10 min and then repeating 95 C for 15 s followed by 60 C for 30 s (at a maximum of 40 times totally and with a plate read between each cycle), a melting curve 65-95 C, increment 0.5 C for 5 s, and a final plate read. A total number of 40 cycles were run before terminating the acquisition. Obtained values were analyzed in Roche LightCycler 480 1.5, Microsoft Excel 2010 (Redmond, WA, USA) and GraphPad Prism 5 (La Jolla, CA, USA). Cut-off were set at 35 cycles and missing data were given the value of 36 (35 þ 1). The Ccl11 expression was normalized against Tie1 expression in control animals using the DDCt method. The relative gene expressions are presented in log2 scale.

CCL11 protein expression in CD311 ECs on histological sections
Animals were deeply anesthetized with sodium pentobarbital (Pentothal V R ; Electra-box Pharma, Tyres€ o, Sweden) and transcardially perfused with 0.9% saline, followed by 4% paraformaldehyde (PFA), 6 h post-IR. Brains were removed and post-fixed for 24 h in PFA at 4 C before being transferred to a 30% sucrose solution (30% sucrose in 0.1 M phosphate buffer). After equilibration in sucrose, the brains were fixed with a cryo-gel (Tissue-Tec V R O.C.T compound, Askim, Sweden) to a dry ice-cooled copper block and one hemisphere was sagittally cut into 25-lm sections with a sliding microtome (Leica SM2000R, Leica Microsystems, Nussloch, Germany). Serial sections were collected in a series of 12 tubes containing tissue cryoprotectant solution (TCS; 25% ethylene glycol, 25% glycerine and 0.05 M phosphate buffer), and stored at 4 C until used for immunohistochemistry.
Staining for immunofluorescence images were performed on every 12th section and rinsed in Tris-buffered saline (TBS; 0.08 mol/L Trizma-HCl, 0.016 mol/L Trizma-Base, 0.15 mol/L NaCl, pH 7.5) before staining. After rinsing in TBS, sections underwent two subsequent steps to avoid unspecific binding. First, endogenous peroxidase activity was blocked by incubating the sections in 30% hydrogen peroxide (H2O2) for 30 min, followed by rinsing in TBS. This was followed by 1 h incubation in TBS with 3% donkey serum and 0.3% Triton X-100 (blocking solution) to avoid unspecific antigen binding. Sections were incubated with primary antibodies diluted in blocking solution at 4 C for 72 h (monoclonal rat anti-mouse CD31, 1:500, BD Bioscience Pharmingen, Franklin Lakes, NJ, USA; polyclonal goat anti-mouse CCL11/Eotaxin, 1 mg/mL, R&D Systems, Minneapolis, MN, USA). After rinsing in TBS, sections were incubated at room temperature for 1 h with biotinylated secondary antibody diluted 1:1000 in blocking solution (donkey anti-rat IgG, Jackson ImmunoResearch Laboratories, West Grove, PA, USA, USA). After rinsing in TBS, a TSA Biotin Detection Kit (PerkinElmer, Waltham, MA, USA) was used to amplify the CD31 antibody according to the following procedure: sections were incubated with SA-HRP (1:200) diluted in blocking solution for 30 min, rinsed in TBS and incubated 8 min with Biotin Tyramide (1:50) diluted in amplification diluent. After another rinsing in TBS, the sections were incubated at room temperature for 2 h with secondary antibodies diluted 1:1000 in blocking solution:donkey anti-goat Alexa 488 (1:1000, Invitrogen/Molecular Probes, Carlsbad, CA, USA) and Alexa 555 conjugated Streptavidin (Invitrogen/Molecular Probes). After incubation with secondary antibodies, the sections were extensively washed in TBS and mounted on glass slides with ProLong Gold Antifade Reagent with DAPI (Molecular Probes/Invitrogen). Representative images of the staining were acquired at 40Â magnification using a Leica DM6000B microscope (Leica, Wetzlar, Germany).

Statistical analysis
Data from cell cycle analysis and flow cytometric quantification of cell populations were compared using two-way ANOVA with treatment and time after IR as main effects. If an interaction between treatment and time was significant it meant that the investigated parameters changed differently over time depending on the treatment. However, when no interaction was found, the two main factors (treatment and time) were further tested separately with a Bonferroni post hoc test. RT-qPCR data are presented as relative gene expression/fold change (log2 scale) compared to the control group at respective time and structure using independent samples t-test. Data from the RT-qPCR TaqMan array were further adjusted for multiple comparisons within each time point with Benjamini-Hochberg. Statistical analysis was performed using SPSS 19.0 (SPSS, Chicago, IL, USA). Significance values were set to: Ã p < .05, ÃÃ p < .01 and ÃÃÃ p < .001.

Irradiation alters the percentage of endothelial cells
Flow cytometric analyses can be used to detect and quantify cell populations and yield results that correlate well with quantitative immunohistochemistry (Hellstr€ om et al. 2007). With the purpose to investigate how ECs are affected by IR in relation to other hippocampal and cerebellar cell populations, we acquired data about the percentage of CD31 þ ECs within the entire heterogeneous cell population during the isolation of ECs with FACS. CD31 is constitutively expressed in all ECs (Minami and Aird 2005) and the protein expression has previously been shown to be unaffected by IR towards skin organ cultures (Heckmann et al. 1998).
Analysis of the percentage of ECs revealed a significant interaction between treatment and time for both the hippocampus (Figure 2(A), p < .001) and the cerebellum ( Figure  2(B), p ¼ .0011), meaning that the percentage of ECs changed differently over time depending on the treatment. In the hippocampus, control and irradiated animals had similar levels of ECs at 6 h post-IR. However, 24 h and 7 days post-IR irradiated animals displayed an increased percentage of ECs compared to control animals (15.3% and 13.2% in controls compared to 21.3% and 17.5% in irradiated animals, respectively). During normal development, the percentage of ECs in the hippocampus gradually decreased from P14 (6-hour time point) to P21 (7-day time point) (17.8%, 15.3% and 13.2% ECs for P14, P15 P21 respectively, Figure 2(A)). After IR, this change over time had been altered and was therefore not observed.
For the cerebellum (Figure 2(B)), an increased percentage of ECs in irradiated mice compared to control mice was apparent already 6 h post-IR (4.0% in controls and 6.2% in irradiated) and the difference remained at 24 h post-IR (5.3% in controls and 8.4% in irradiated), but not at 7 days post-IR. The changes in the cerebellum were however modest compared to the hippocampus. Unlike the hippocampus, the percentage of ECs increased from P14 (6-hour time point) to P21 (7-day time point) in both control and irradiated mice. This indicate that the hippocampus and cerebellum are in different developmental phases at this postnatal age.

Cell cycle arrest in endothelial cells after cranial irradiation
IR is known to cause acute cell death and reduce proliferation of cells in the neurogenic subgranular zone of the hippocampal dentate gyrus, clearly detectable 6 h post-IR (Fukuda et al. 2004;Roughton et al. 2012). With the purpose to assess if IR negatively influenced proliferation of hippocampal and cerebellar ECs in the young brain, we performed a cell cycle analysis where the ECs were grouped in two different phases: non-cycling (G0/G1) versus actively cycling (S/G2/M). Since all cells were grouped into these two phases, an up-regulation in one phase is always balanced by a corresponding down-regulation in the other phase. Data from the actively cycling ECs are presented and discussed below.
In the hippocampus, IR did not affect the percentage of ECs that were actively cycling at any of the time points investigated (Figure 2(C)). However, there was a drastic decrease in the percentage of actively cycling cells over time in both control (17.9% at P14 vs. 0.57% at P21, p < .001, Figure 2(C), left panel) and irradiated animals. In the cerebellum, an interaction between treatment and time was found for the percentage of ECs in the cycling phase (Figure 2(D), p ¼ .009). For example, IR resulted in fewer actively cycling ECs 6 h post-IR (Figure 2(D)). No difference was observed 24 h post-IR, but at 7 days post-IR irradiated ECs exhibited increased proliferation compared to controls (1.77% in controls vs. 5.05% in irradiated animals, Figure 2(D)).

Gene expression alterations in endothelial cells after irradiation
To further evaluate what role ECs have in IR-induced tissue damage and if IR changes the molecular pattern in hippocampal and cerebellar ECs differently, the relative expression of 48 selected genes involved in DNA damage, DNA repair, apoptosis, proliferation, angiogenesis and inflammation was analyzed using RT-qPCR. For a complete list of all the evaluated gene assays see Supplementary Table 1, and for a list with the observed significantly altered gene expressions, see Table 1.
A principal component analysis (PCA) of the gene expression data was performed in order to illustrate the changes in gene expression at different time points post-IR and between the two investigated brain structures. The PCA scatter plots demonstrate that each group clustered nicely and that the samples within each group were homogenous. In the controls (Figure 3(A)), it was clearly shown that the gene expression profile changed during normal development (from the age of P14 to P21) and that the hippocampal and cerebellar gene expression profiles were significantly different from each other. This supports the findings in Figure 2(A,B), showing that the hippocampus and cerebellum display different developmental profiles at these postnatal ages. Furthermore, in the hippocampus (Figure 3(B)), the controls and the irradiated animals were well separated at 6 h post-IR, converged at 24 h and at were no longer separated 7 days post-IR. In the cerebellum, the groups were also well separated at 6 h post-IR, but less clearly so at the later time points (Figure 3(C)). Hence, the radiation-induced damage responses in the hippocampus and cerebellum were similar, but not identical.
In the current study, we evaluated the role of ECs in IRinduced damage and observed that they appear to be important for the inflammatory response. The chemokine Ccl11 (also known as eotaxin) was the most strongly upregulated gene in the hippocampus (5.3-fold) and the second most upregulated gene in the cerebellum (3.3-fold) (Figure Data are presented as relative gene expression compared to controls in log2 scale, compared using independent samples t-test and adjusted for multiple comparisons within each group (Benjamini-Hochberg). n ¼ 6-10 per group. ECs: endothelial cells; FC: fold change; IR: irradiation; p values are in italics.

4(A)
). Since pro-inflammatory cytokines to a great extent are produced and released by activated microglia (Smith et al. 2012), we isolated a new round of ECs together with microglia acutely post-IR (6 h post-IR) in order to compare the gene expression in ECs and microglia, and to validate this novel finding of EC induction of the pro-inflammatory cytokine Ccl11 after in vivo cranial IR. We observed that expression of Ccl11 was induced in both ECs (3.4-fold, p ¼ .002) and microglia (2.0-fold, p ¼ .001) 6 h post-IR, thereby strengthening the results from the RT-qPCR TaqMan array (Supplementary Figure 1). Immunofluorescence double staining 6 h post-IR confirmed that CCL11 was expressed by CD31þ ECs and that the expression was stronger in irradiated animals compared to controls ( Figure 5). In line with these results, we observed an up-regulation on the RT-qPCR TaqMan array of the chemokine Ccl2 (gene alias Mcp-1) which was upregulated both 6 h (2.5-fold) and 24 h (1.4-fold) post-IR in the hippocampus and at 6 h (3.2-fold) post-IR in the cerebellum (Figure 4(B)). Related to the increased gene expression of the two chemokines, the cytokine interleukin 6 (Il6) was upregulated at 6 h post-IR in both the hippocampus (2.2-fold) and the cerebellum (2.7-fold, Figure 4(C)). Similarly, gene expression of the adhesion molecule Icam1 was induced at 6 h post-IR in both structures (1.2-fold, Figure 4(D)). In line with the inflammatory response of ECs, we also observed an increased expression of Sele (E-selectin), Selp (P-selectin), Vcam1 and Tnf (Tnfa) at 6 h post-IR in the cerebellum and of Tnf at 24 h post-IR in the hippocampus. This was followed by a reduction of E-selectin at 24 h post-IR in the cerebellum (Table 1).
The cell cycle inhibitor Cdkn1a (also known as p21) was the only transcript that stayed upregulated, with similar expression levels at all three time points investigated. Furthermore, the fold change was higher in the hippocampus (2.3-2.5) compared to the cerebellum (1.8-2.0), at all three time points (Figure 4(E)). Besides the induction of inflammatory gene expression and cell cycle arrest in irradiated ECs, we also observed significant changes in many of the other genes on the RT-qPCR TaqMan array (Table 1). For genes involved in angiogenesis/proliferation of ECs, we observed an up-regulation of Angpt2 at 6 h post-IR in both the hippocampus and the cerebellum and of Fgfr1 at 6 h post-IR in the only nicely showed that the gene expression differs between the hippocampus and the cerebellum. Samples from the hippocampus cluster to the left, while samples from the cerebellum cluster to the right. In addition, a gradual developmental change was observed from the 6-h time point to the seven-day time point (P14 ! P21). (B) In the hippocampus, the PCA analysis showed that irradiated and control animals were well separated at 6 h post-IR, but that they progressively adopt similar gene expression patterns until the two groups cannot be distinguished from each other at 7 days post-IR. (C) For the cerebellum, the pattern was not as clear as for the hippocampus, but we still observed that control animals and irradiated differ at 6 h post-IR. n ¼ 6-10 per group. circles: irradiation; squares: controls; blue: hippocampus (hip); grey: cerebellum (cer); IR: irradiation; P: postnatal day; PCA: principal component analysis; RT-qPCR: quantitative reverse transcription polymerase chain reaction.
hippocampus, consistent with vascular regression (in the absence of angiogenic inducers). This was followed by a down-regulation of Tek (Tie-2 receptor) at 24 h post-IR in the cerebellum. Furthermore, we observed an up-regulation of Acvrl1 (Alk1) in the hippocampus at 6 and 24 h post-IR and of Id1 at 6 h post-IR in the cerebellum.

Discussion
The primary aim of the current study was to unravel the role of the vasculature in radiation-induced tissue damage and investigate how IR affects the endothelium. The major findings were the following: (i) the hippocampus and the cerebellum display different developmental profiles and respond in a similar, but not identical, way to IR at the age studied (P14), (ii) ECs are less susceptible to IR than surrounding nonvascular cells in the hippocampus and the cerebellum, (iii) ECs contribute to the inflammatory response following cranial IR, a role normally attributed to microglia in the brain.
We observed that the proportion of ECs relative to the whole cell population decreased in the hippocampus during normal brain development, but increased in the cerebellum (Figure 2(A,B)). Further, PCA analysis of the gene expression data showed that the gene expression profiles in the hippocampus and cerebellum changed during normal development and that the profiles clearly differ between the two brain regions (Figure 3(A)). The two brain structures hence display different developmental profiles between the age of P14 and P21, and respond, in part, differently to cranial IR. This observation is further supported by a recent study of radiation-induced pathological changes in the cerebellum where the results indicated that the cerebellum was more sensitive to IR than the hippocampus and the white matter (Zhou et al. 2017).
In this study, we found an increased percentage of ECs in irradiated animals at 6 h and 24 h post-IR in the cerebellum and at 24 h and 7 days post-IR in the hippocampus ( Figure  2(A,B)). Hence, the effects of IR were observed earlier in the cerebellum than in the hippocampus. The results from the hippocampus are supported by a previous study where stereological quantifications showed an increased density of microvessels 1 week post-IR in the hippocampus, presumably because the surrounding non-vascular tissue exhibited a more pronounced cell death and growth arrest compared to the vasculature (Bostrom et al. 2013). A minimum IR dose of 8 Gy has been proposed for an EC apoptotic response in the intestine (Paris et al. 2001). We have previously investigated EC death in the hippocampus by immunohistochemistry after a single IR dose of 8 Gy. In that study, we did not observe any IR-induced cell death of ECs in the hippocampal dentate gyrus, as measured by caspase-3 activation (Bostrom et al. 2013), indicating that a dose of 8 Gy is not enough to induce cell death of ECs in the developing hippocampus. Previous data have shown that 10 Gy whole brain IR suppressed proliferation and increased apoptosis of ECs in the adult rat brain (Lee et al. 2011) and we therefore evaluated a cranial IR dose of 10 Gy in the current study. We performed a cell cycle analysis where the ECs were characterized as either 'resting' or 'cycling' and investigated genes involved in EC apoptosis. For the cell cycle analysis the following was found: (i) ECs in the cerebellum were mildly affected by IR, while no difference was found in the hippocampus, (ii) a large proportion of ECs was in cycling phases (S/G2/M) at the age of P14 and (iii) from P14 to P21 ECs gradually proliferated less frequently. These results are consistent with a previous study showing that EC proliferation peaks between P5 and P9 but gradually decreases to maintenance levels of a few percent at P25 (Robertson et al. 1985). The finding that irradiated animals had an increased percentage of hippocampal ECs at 24 h and 7 days post-IR (Figure 2(A)) despite the fact that no difference was observed in the fraction of proliferating cells between control and irradiated animals (Figure 2(C)) is at a first glance puzzling. A possible explanation for this is that other cell types are more affected by IR than the vascular tissue, resulting in a reduction of surrounding non-vascular cells and ultimately in an increased percentage of ECs, as observed in a previous study (Bostrom et al. 2013). In addition, we have shown that IR of the young, still growing rodent brain impairs growth and results in reduced size (Fukuda et al. 2004;Fukuda et al. 2005;Han et al. 2016), an effect observed already at 7 days post-IR (Kalm et al. 2016).
While apoptosis in a number of cell types is dependent on and mediated by p53 (e.g. apoptosis of neural progenitor cells), the apoptosis of ECs has been proposed to be p53independent. Instead, EC apoptosis could be attributed to the acid sphingomyelinase (ASMase) pathway (Garcia-Barros et al. 2003;Li et al. 2003;Fuks and Kolesnick 2005;Li et al. 2010). We therefore investigated the gene expression of ASMase (Smpd1) as well as neutral sphingomyelinase (NSMase or Smpd2), but did not observe any induction or repression at any time during the experiment. The lack of induced expression indicates that a dose of 10 Gy is not enough to induce apoptosis of ECs, at least not through the ASMase pathway. However, it has previously been shown that Figure 5. CCL11 protein expression was observed in ECs on histological sections. Representative micrographs were acquired using standardized, identical light conditions to visualize the proinflammatory CCL11 protein expression (green) in CD31þ microvessels (red) 6 h post-IR. CD31 expression in control (A) and irradiated (B) animals shown in the left panel, CCL11 expression in control (C) and irradiated (D) in the middle panel and double labeling of CD31 and CCL11 in control (E) and irradiated (F) animals shown in the right panel. The CCL11 staining was stronger in irradiated treated tissue in both CD31þ ECs and in the surrounding tissue. The higher staining intensity of CCL11 in and around CD31þ ECs was interpreted as a sign that CCL11 has been released into the surrounding tissue to chemotactic recruit immune cells. Scale bar ¼ 20 mm. CD31: cluster of differentiation 31; C: control; CCL11: chemokine (C-C motif) ligand 11; ECs: endothelial cells; IR: irradiation.
ASMase and NSMase are rapidly activated within minutes to hours post-IR , hence proposing that posttranslational modifications and subsequent activation might have occurred although no difference was detected on the level of gene expression, or that the 6-h time point was too late to detect any upregulation. Interestingly, Cdkn1a (p21) was the only transcript that was upregulated at all three time points investigated. These results are consistent with a study where 8 Gy IR delivered to the brains of P9 rats increased the expression of Cdkn1a at both 6 h and 7 days post-IR (Kalm et al. 2009). In addition, increased Cdkn1a expression was observed in hippocampal stem cells 16 h post-IR (Hellstr€ om et al. 2011). Cdkn1a is a cell cycle inhibitor that blocks progression from G1 to S phase (Phan et al. 2005) and ultimately G1 arrest (Canman et al. 1998). Moreover, Cdkn1a was recently shown to be directly involved in DNA repair (Cazzalini et al. 2010). The lack of apparent radiation-induced apoptosis in ECs in the current and our previous study (Bostrom et al. 2013), together with the consistent upregulation of Cdkn1a, suggest that ECs undergo extensive DNA repair following cranial IR. This provides a possible explanation to why ECs are less susceptible to IR than other cell populations.
The major IR-induced gene expression responses of ECs detected in the current study were related to inflammation. In the current study, several genes were upregulated acutely (6 h) post-IR and we then observed a gradual decrease in the expression until 7 days post-IR. For example, the chemokine Ccl11 (eotaxin) was the most strongly upregulated gene in the hippocampus and the second most upregulated gene in the cerebellum 6 h post-IR. Ccl11 has previously been shown to be induced by IR, at least in human dermal fibroblasts in vitro (Huber et al. 2000). However, to our knowledge no other study has shown that IR induces acute expression of Ccl11 in ECs after in vivo cranial radiation. Following in vitro IR, ECs adopt a senescence-associated secretory phenotype that is characterized by the upregulation of pro-inflammatory cytokines and chemokines such as CCL11, CCL2 and IL-6 (Ungvari et al. 2013), that is, three of the genes that were acutely upregulated in ECs after in vivo IR in the current study. Further, an upregulation of CCL11 plasma levels in aged mice has been coupled to the natural decline in hippocampal neurogenesis. Importantly, increased levels of CCL11 in plasma and cerebrospinal fluid during normal aging of healthy human individuals have also been observed. These findings were further supported by an observed decreased neurogenesis following i.p. injections of CCL11 (Villeda et al. 2011). Nevertheless, a recent study showed reduced levels of CCL11/Eotaxin three months after cranial proton radiation, proposing that radiation-induced effects do not imitate the processes occurring in the aged brain (Raber et al. 2016). This study was however performed in adult mice, and not in the juvenile brain as the current study.
The finding in the current study that ECs express Ccl11 and that the expression of Ccl11 in ECs was even higher compared to the expression in microglia, demonstrates that ECs play an active and not only a bystander role in initiating the inflammatory response post-IR. Further, the inflammatory response has shown to have long-term deleterious effects on for example neurogenesis (Monje et al. 2002). The immunofluorescence staining verified that ECs express CCL11 and that the expression was higher in irradiated animals compared to controls ( Figure 5). Moreover, the results from the current study also support previous reports of a transient inflammatory response in the young rodent brain after cranial IR (Kalm et al. 2009), but describe a previously unknown contribution by ECs. The receptor for CCL11, called CCR3 is expressed by both microglia (Xia et al. 1998) and neural stem cells (Mendelsohn and Larrick 2011), thereby proposing that ECs could be involved in the direct activation of microglia and/or direct detrimental effects on neural stem cells following cranial IR. These results together proposes that ECs adopt a senescent phenotype following in vivo cranial IR, but the role of ECs in the inflammatory response is however to our knowledge an unexplored area. Further experiments are needed in order to determine the exact role that ECs play in this complicated process.

Conclusions
The formerly widespread belief that vascular damage is the primary cause of radiation-induced brain injury and subsequent cognitive so called late effects in patients has been challenged. Nevertheless, there is still ample evidence that changes to ECs and the vasculature can, at least partly, explain some of the damage caused by radiotherapy. In the current study, we demonstrate that ECs are less susceptible to IR than surrounding cells in the hippocampus and the cerebellum, and demonstrate that they are actively involved in the inflammatory response observed after cranial IR. Hvitfeldtska Foundation. The funding agencies had no influence on the study design.

Notes on contributors
Martina Bostr€ om is postdoctoral fellow at the Department of Oncology at the University of Gothenburg, Sweden. Her research is focusing on how vascular changes are involved in both the acute, as well as the chronic complications from cancer treatment.
Marie Kalm is an assistant professor at the Department of Pharmacology at the University of Gothenburg, Sweden. Her research is focused on foreseeing and preventing the chronic neurodegenerative state following cranial radiotherapy.
Yohanna Eriksson is a PhD student at the Department of Pharmacology at the University of Gothenburg, Sweden. Her PhD project aims at better understanding the late effects of radiotherapy to the juvenile brain.
Cecilia Bull is a researcher at the Division of Cancer Epidemiology, Department of Oncology at the University of Gothenburg, Sweden. She has a PhD in Neurobiology and her research is focused on preventing radiation-induced injury to the gastrointestinal system.
Anders Ståhlberg is principal investigator at the Sahlgrenska Cancer Center, University of Gothenburg, Sweden. He is experienced in tumor biology, liquid biopsies and single-cell analysis.