Brain pericytes acquire a microglial phenotype after stroke

The in vivo study was conducted using reporter Rgs5 ^gfp/+ mice [39]. In this mouse, GFP is expressed in the cytoplasm under the pericyte-specific RGS5 promoter which makes it possible to track pericytes. The Rgs5 ^gfp/+ mouse we use is a knock-out/knock-in mouse and is retaining the full promoter region [39]. Mice were housed under standard conditions, with free access to water and food. All animal experiments followed “Principles of laboratory animal care” (NIH publication No. 86-23, revised 1985) and experiments were approved by the Ethical Committee at Lund University (M90-12 and M479-12).

Focal cerebral ischemia was induced by permanently occluding the distal part of the left middle cerebral artery (MCA) as previously described [26]. In brief, the mice were placed on an operating table at 37.5 °C under a dissecting microscope. Under isoflurane anesthesia, an incision was made between the left lateral part of the orbit and the left ear. The parotid gland and the temporal muscle were displaced in the distal direction with a pair of anatomical forceps. A small craniotomy was made with a 0.8-mm burr above the anterior distal branch of the MCA using a high-speed microdrill. The MCA was exposed and occluded permanently by electrocoagulation using an electrosurgical unit (ICC50; Erbe, Germany). Marcain was used as local analgesia before suturing the wound.

Mice were injected intraperitoneally for 2 weeks with bromodeoxyuridine (BrdU, 50 mg/kg, Sigma-Aldrich) twice daily, starting 2 h after injury.

Human brain tissues were obtained from patients who suffered from stroke and age-matched controls for this study. Before the investigation, the entire collection of brain sections, was subjected to a neuropathological whole-brain analysis for clinical diagnostic purposes, according to routine procedures at the Department of Pathology, Division of Neuropathology, Lund University Hospital. Human brain pericyte lines were previously derived from brain biopsies as described [43].

All procedures involving human tissue were performed with informed written consent by the patient for the donation of brain tissue. The use of human brain tissue was approved by the Regional Ethical Review Board in Lund, Sweden (Dnr 196/2010) and according to the declaration of Helsinki.

Mice were killed at 1, 7 and 21 days after pMCAO and transcardially perfused with phosphate buffered saline (PBS) followed by 4 % paraformaldehyde (PFA), placed in 30 % sucrose in PBS and sectioned in the coronal plane at 40 μm. Sections were incubated with blocking solution (5 % normal donkey serum in PBS, with 0.25–2.5 % Triton-X100) for 1 h at room temperature (RT), then incubated at 4 °C or RT overnight with primary antibodies diluted in 3 % normal donkey serum. For BrdU detection, sections were pre-treated with 2 N HCl at room temperature for 30 min before staining. The following primary antibodies were used: alpha smooth muscle actin (α-SMA) (1:100, rabbit, Abcam), GFP (1:1,000, chicken, Abcam), PECAM-1 (CD31) (1:400, rat, BD), BrdU, (1:200, rat, Accurate), CD11b (1:1,000, rat, Sigma), CD13 (1:200, rat, Serotec), CD45 (1:100, rat, BD), CD45 (1:100, mouse, BD), CD68 (1:100, mouse, DAKO), collagen-IV (1:200, rabbit Serotec), fibronectin (1:300, rabbit, Abcam), fibroblast surface protein (FSP) (1:100, mouse, Sigma), galectin-3 (1:500, rat, kindly provided by Hakon Leffler), laminin (1:200, rabbit, Abcam), NG2 (1:500, rabbit, Chemicon), IBA1 (1:2,000, rabbit, Wako), anti-Ki67 (1:100, rabbit, Novacostra), anti-PDGFRβ (1:200, rabbit, cell signaling), RGS5 (1:200, mouse, ORIGENE). After washing, antibody staining was revealed using species-specific fluorophore-conjugated (Daylight 488, Cy3, Cy5 from Jackson, Alexa 488 from molecular probes) or biotin-conjugated secondary antibodies (Invitrogen). Biotinylated secondary antibodies were revealed using the ABC kit (Vector labs). DAPI (1 μg/mL, Sigma) and TOTO-3 iodide 642/660 (1:500, Invitrogen) were used for counterstaining. Control sections were stained with secondary antibody alone.

For immunocytochemistry, cells were plated on glass coverslips that were coated with vitronectin 0.50 μg/mL. Cultured cells were fixed with 2 % PFA at RT, washed in PBS and incubated for 1 h in PBS, 5 % donkey serum. Cultured cells were then stained for both primary and secondary antibodies in the presence of 0.1 % Triton X-100 and 2 % donkey serum.

Immunofluorescent stainings were visualized using an epifluorescence microscopy system (Olympus BX53) and processed using cellSens digital imaging software. Confocal microscopy was performed using a Zeiss LSM510 confocal microscope (Carl Zeiss) equipped with a GreNe and a HeNe laser, using the following lines of excitation: 488, 594, and 647 nm. Figures were composed using Photoshop CS5 software. To quantify the number of GFP⁺ pericytes over time, GFP⁺ cells were counted in both, the infarct area and the corresponding contralateral area using cellSens digital imaging software (3–4 optical fields of 0.18 mm² for each mice, bregma −1.2 to 1.5 mm, n = 5, per group). To determine proliferation in RGS5-expressing pericytes, the number of GFP⁺ cells and GFP⁺/BrdU⁺ cells was counted on three 40× confocal pictures spanning randomly the infarct area including both infarct core and peri-infarct area 7 days after pMCAO. Approximately 200 cells were counted per field in the ipsilateral part of stroke brains. The percentage of BrdU/GFP⁺ cells was determined for each animal and the average for five animals was calculated. To quantify GFP⁺/CD11b⁺, GFP⁺/IBA1⁺, and GFP⁺/GAL-3⁺ cells at 7 day after stroke, Photoshop software was used. Double-labeled cells were counted on 40× confocal pictures spanning the peri-infarct area. Approximately 300 cells were counted per animal. The percentage of cells was determined for each animal and the average for three animals was calculated.

For immunohistochemistry, the bone marrow was processed as previously described [38]. In brief, bones were post fixed for 2  h in 4 % PFA at 4  °C, then they were decalcified for 48  h in 250  mM EDTA pH  7.4 at 4  °C in a shaker. Decalcified bones were cryoprotected, sectioned in the coronal plane at 12 μm and stained for GFP, CD11b, and CD45. Stroke mouse brain sections were used as a positive control.

Primary microglial cells were isolated from mixed glial cell cultures prepared from forebrain of postnatal day 1 (P1) Rgs5 ^gfp/+ pups as described previously [12]. In brief, upon reaching confluence (9–13 days after isolation), microglia adhering to the astrocytes monolayer were dislodged by shaking for 1 h and resuspended in Dulbecco’s modified Eagle Media (DMEM), supplemented with 1 % penicillin/streptomycin and 10 % fetal bovine serum (FBS), and plated on uncoated 6-well plates. For cell sorting, cells were incubated with anti-CD11b-Alexa 647 (BD) and 7-AAD sorted by FACS (FACSAria; Becton–Dickinson) or DIVA software (Becton–Dickinson) using a low stream speed to ensure a high level of cell survival. Sorted microglia cells were then exposed to oxygen–glucose deprivation (OGD) for 1 h.

Human brain pericytes were prepared according to the method as described previously [43]. In brief, tissue samples were stored in Leibowitz-15 media (Invitrogen) at 4 °C, cut and enzymatically digested in enzyme solution (Collagenase 1 mg/mL (Sigma); Dispase 1.6 mg/mL (Roche); Trypsin 0.25 mg/mL (Sigma); DNase I 80 U/mL (Sigma) in DMEM and 4.5 mg/mL glucose (Invitrogen) at 37 °C for 20 min. Cells were plated on 24-well culture dishes and incubated at 37 °C/5 % CO₂ in DMEM/F-12/Glutamax/B27 (Invitrogen). Sorted cells were maintained in culture medium consisting of Stemline medium (Sigma-Aldrich), 2 % FBS, 1 % penicillin/streptomycin and 1 % Glutamax. For flow cytometry, cultured cells at high passages were labeled with the following commercial antibodies: anti-CD140b-PE anti-CD13-PECY7 anti-CD11b-PE, anti-CD14-FITC, and CD45-APC (all from BD). For OGD, the cells were first rinsed with PBS and OGD was induced by a deoxygenated basic salt solution without glucose, in a hypoxic chamber (oxygen below 1  mm Hg; Electrotek, Shipley, UK) at 37 °C. After 2 h, the basic salt solution was removed and replaced by differentiation medium consisting of DMEM (Gibco Laboratories, USA), 1 μg/mL fibronectin (Sigma), 5 ng/mL basic fibroblast growth factor (bFGF) (R&D), 50 nM cAMP (Sigma) and cells were placed immediately back into the incubator (5 % CO₂) at 37 °C. The cells were collected after 4 days for qPCR analysis or immunocytochemistry.

Cells were homogenized in RNeasy mRNA kit (Qiagen) and stored at −80 °C before total RNA extractions according to the manufacturer’s protocol. cDNA synthesis was performed using Script™cDNA Synthesis Kit (Bio-Rad). cDNA was analyzed using real-time PCR SsoAdvanced™ SYBR^® Green Supermix from Bio-Rad using appropriate primers and run on a Bio-Rad CFX96 real-time quantitative PCR (qPCR) system. Gene expression was normalized to the housekeeping gene GAPDH and calculated using the 2^−ΔCt method. Melt curve analyses were performed to ensure the specificity of qPCR product. Primer sequences can be provided on request. Values are mean ± SEM of three independent experiments, and within each experiment, triplicate samples were assessed.

Data were analyzed using GraphPad Prism version 5.04 software. Results of cell counts are reported as mean ± SD and gene expression values are expressed as mean ± SEM. The significance of differences between means was assessed by student t test or ANOVA with p < 0.05 considered statistically significant.

We analyzed brain section of control animals from Rgs5 ^gfp/+ mice. GFP⁺ pericytes were located around microcapillaries (Fig. 1a) where they co-expressed the pericyte marker platelet-derived growth factor receptor β (PDGFRβ) and CD13 (Fig. 1b, c), but not α-smooth muscle actin (α-SMA), a marker that was typically expressed by larger vessels (Fig. 1d). We did not detect GFP expression in any other cell type including microglia or astrocytes (Fig. 1e, f). Therefore, we could confirm previous reports that GFP is solely and exclusively expressed in pericytes in Rgs5 ^gfp/+ mice [39].

Next, we induced permanent focal brain ischemia [25] in Rgs5 ^gfp/+ mice (Fig. 2a). The ischemic injury induced an almost twofold increase in the number of GFP⁺ cells in the infarct area as compared to the non-injured contralateral side 1 day after ischemia (24.33 ± 3.6 and 14.33 ± 0.88, respectively, n = 5). The number of GFP⁺ pericytes peaked reaching a tenfold increase at day 7 (95.33 ± 7.42 and 9.66 ± 1.20, n = 5), and then gradually decreased by 21 days (34.00 ± 3.30 and 9.33 ± 1.85, n = 5) (Fig. 2b, c).

We then analyzed BrdU incorporation in the GFP⁺ cell population to address whether the increase in the number of pericytes is due to cell proliferation. BrdU immunolabeling revealed that proliferation of GFP⁺ pericytes started during the first days after stroke. At 7 days following the ischemic injury, 74 % ± 8.99 of the GFP⁺ pericytes had incorporated BrdU, indicating a significant increase in mitotic activity. GFP⁺ pericytes incorporating BrdU also expressed Ki67, a nuclear marker that is strictly associated with cell proliferation (Fig. 2d). None of the pericytes in the contralateral side incorporated BrdU or expressed Ki67 at any time point after the ischemia, indicating that pericytes in the adult brain remain quiescent under physiological conditions (Fig. 2e).

Both, in the brain of intact mice and in the contralateral hemisphere of stroke mice, pericytes showed low GFP expression and a flat morphology consistent with a quiescent state [39] (Fig. 3a). One day after the injury, GFP⁺ cells were still found in close proximity to blood vessels in the infarct area. However, they displayed an activated morphology with a prominent cell body and elongated processes as previously described [15] (Fig. 3b). To further elucidate that GFP⁺ cells in the infarct area are activated pericytes, we stained sections with an antibody for Neuron-Glial 2 chondroitin sulfate proteoglycan (NG2). NG2 is expressed on the surface of activated pericytes during the development and in the adult under both physiological and pathological conditions [41, 42, 47]. All GFP⁺ pericytes associated with capillaries were positive for NG2 at 1 day after the injury (Fig. 3b, c).

Seven days post-injury, GFP⁺ cells had left the microcapillary wall and migrated into the surrounding parenchyma (Fig. 3d). GFP expression remained in cells that had migrated into the parenchyma at day 7, but was weaker compared to day 1. GFP⁺ cells that had lost contact with blood vessels displayed an ameboid morphology (Fig. 3d), similar to the morphology that has been described for pericytes leaving the blood vessels in response to traumatic brain injury [16]. We therefore decided to further investigate the phenotype of the parenchymal GFP⁺ cells to elucidate whether pericytes give rise to another cell type.

Pericytes have been suggested to have phagocytic activities in vitro [5] and to give rise to follicular dendritic cells in lymphoid tissue in vivo [34]. To examine whether pericytes give rise to microglial cells in stroke, we stained brain sections for the microglial markers GAL-3, IBA1 and CD11b [30, 36]. In the infarct area, parenchymal GFP⁺ cells expressed GAL-3 and IBA1, whereas pericytes still in contact with microvessels were negative for these microglial markers at 7 days after the injury (Fig. 4a–c). GFP⁺ cells in the parenchyma had an ameboid and intermediate form as characteristic of activated microglia/macrophages or an arborized shape (Fig. 4d–f).

We then quantified GFP⁺ cells in the peri-infarct area. At 7 days after stroke, 75 % ± 5.5 of GFP⁺ cells expressed GAL-3 (Fig. 4d). GFP⁺ cells also co-labeled with CD11b (Fig. 5a–d). At 7 days, 88 % ± 6.1 of GFP⁺ cells expressed CD11b. Similar to GFP⁺/GAL-3⁺ cells, GFP⁺/CD11b⁺ had the morphology of activated microglia (Fig. 5b, c) and co-localized with IBA1 (Fig. 5b). Interestingly, some activated pericytes that still localized around the capillaries were positive for CD11b (Fig. 5d). In the peri-infarct area, 44 % ± 30.08 of GFP⁺ cells expressed IBA1.

Out of the microglial population, 25.6 % ± 2.5 of the GAL-3⁺ cells, 49 % ± 4.5 of CD11b⁺ cells and only 16 % ± 11.7 of IBA1⁺ cells were positive for GFP at 7 days after stroke suggesting that only a limited number of microglia were derived from pericytes. We did not observe any expression of IBA1, CD11b (Fig. 5e, f) and GAL-3 (data not shown) in quiescent GFP⁺ pericytes in the contralateral side of the stroke brains or in intact mice.

We also tested if Rgs5 mRNA is upregulated in stimulated microglial cells and thus may lead to unspecific GFP expression in activated microglia. We isolated microglial cells from Rgs5 ^gfp/+ mice and sorted them positively for CD11b (Fig. 5g). We then exposed these microglial cells to ischemic conditions in vitro, using OGD, and analyzed the activated microglia using qPCR. We did not detect any upregulation of Rgs5 mRNA levels in activated microglia upon OGD exposure (Fig. 5h).

We next examined pericytes in the infarct area for co-expression of CD68, a macrophage/microglia marker. CD68⁺ cells were found closely associated to blood vessels and in the parenchyma of the infarct area. Perivascular CD68⁺ cells had a similar morphology to pericytes but were negative for GFP, suggesting that these cells may be perivascular macrophages [4, 24] (Fig. 6a). Seven days after the injury, CD68⁺ cells stained positive for CD11b, but none of the CD68⁺/CD11b⁺ cells overlapped with GFP⁺ pericytes (Fig. 6b). We also examined GFP⁺ pericytes for co-expression of CD45 in the infarct area. Cells only positive for CD45 were found close to pericytes at 1 and 7 days following injury; however, we could never detect CD45⁺ cells that co-labeled with GFP (Fig. 6c–e).

Next, we investigated whether GFP⁺ pericytes had infiltrated the stroke area from the bone marrow as it had been previously suggested in stroke [33]. First, we analyzed the bone marrow of intact Rgs5 ^gfp/+mice. GFP⁺ pericytes were found around capillaries in the bone marrow, albeit at low frequency (Fig. 7a, d). When we analyzed the bone marrow of Rgs5 ^gfp/+ mice at 1 and 7 days post-stroke, we did not detect any increase in GFP⁺ cells. GFP⁺ cells did not express CD45 or CD11b at 1 (Fig. 7b, c) and 7 days (Fig. 7e, f) after stroke. In addition, we also harvested both, central and endosteal bone marrow from stroke-injured Rgs5 ^gfp/+ mice at day 1 and day 7 and analyzed the bone marrow using flow cytometry. Consistent with our immunohistochemical results, we did not observe any GFP⁺ cells that expressed CD45 or CD11b (Fig. 7g).

PDGFRβ⁺ pericytes have previously been shown to contribute to scar tissue formation after spinal cord injury and these particular pericytes have been named as type A pericytes [23]. We found that both PDGFRβ⁺ and α-SMA⁺ cells were mainly located in and around the infarct core (Fig. 8a, b), but displayed differences in morphology and tissue distribution compared to GFP⁺ pericytes. Although activated GFP⁺ pericytes in the infarct core were positive for PDGFRβ 1 day after the injury (Fig. 8c), at 7 days after the injury there was no overlap between parenchymal pericyte-derived cells and PDGFRβ⁺ or α-SMA⁺ cells in the infarct core (Fig. 8d, e).

Next, we examined other markers expressed in stromal cells that are associated with scar tissue formation. Interestingly, in contrast to previously described type A pericytes [23], GFP⁺ pericyte-derived cells were negative for the fibroblast markers fibronectin, collagen-IV and the fibroblast surface protein (Fig. 8f–h), suggesting that GFP⁺ pericytes are a subpopulation of pericytes that do not give rise to scar tissue in the brain following stroke.

We next analyzed sections of human postmortem stroke brain and control samples. RGS5 staining identified pericytes that lined the microvessels and expressed PDGFRβ (Fig. 9a, b). Consistent with our results in rodents, we found that pericytes expressed RGS5 in human stroke samples and showed an activated morphology with a round cell body (Fig. 9b, c). A proportion of pericytes co-expressed RGS5 and GAL-3 along the capillaries in the peri-infarct area (Fig. 9c).

Finally, we examined whether ischemic conditions can induce pericytes to acquire a microglial phenotype in vitro. We have previously reported that human brain-derived pericytes have mesenchymal stem cell characteristics and are able to differentiate into several different cell lineages in vitro [43]. Consistent with our previous data, flow cytometry analysis showed that human-derived pericytes highly express PDGFRβ/CD140b and CD13 (99.1 %). Human-derived pericytes did not express monocyte/macrophage (CD45, CD11b; 0.166 %) or monocyte markers (CD14, CD11b; 0.0528 %) as analyzed by flow cytometry (Fig. 9d). Importantly, using qPCR, we did not detect mRNA for microglial markers such as CD11B, GAL-3, IBA1, tumor necrosis factor alpha (TNFA) and major histocompatibility complex class II (MHC11) (Fig. 9k). To mimic ischemia-like conditions in vitro, human brain-derived pericytes were exposed to OGD and analyzed using immunocytochemistry and qPCR. Pericytes had the typical flat morphology in vitro and expressed PDGFRβ and α-SMA (Fig. 9e–g). Pericytes exposed to OGD changed their morphology compared to control cultures and expressed CD11b and GAL-3 (Fig. 9h–j). mRNA levels of typical genes for microglial cells CD11B, GAL-3, IBA1, TNFA and MHC11 were upregulated following OGD, but not expressed in control cells (Fig. 9k). Consistent with our findings in vivo, whilst mRNA for microglial markers was upregulated in cultures induced after OGD, the expression of the mRNA for pericyte markers PDGFRB, CD13 and A-SMA, was significantly decreased in the differentiated cells compared to controls (Fig. 9k). These data support that ischemic conditions can induce a phenotypic switch in pericytes toward a microglial phenotype.