Introduction
Post-stroke inflammation is a double-edged sword in brain injury and involves various cellular players within the brain and also infiltrating immune cells [
1]. Microglia cells are the main professional immune cells resident in the brain and promptly react to cerebral ischemia, as identified by changes in morphology [
2] and inflammatory cytokine release [
3]. However, depletion of microglia exacerbates stroke outcome in adult [
4,
5] and neonatal mice [
6], suggesting a beneficial role for these cells, in part by removal of dead cells by phagocytosis and production of neurotrophic factors [
1]. Infiltrating immune cells, in turn, are diverse after stroke, with discrepant spatial–temporal effects [
7]. Despite decades of studies, the route and mechanisms of leukocyte infiltration into the brain remain to be fully understood. Moreover, we and others have shown major differences in inflammatory response [
8] and blood–brain barrier (BBB) function [
9] early after neonatal and adult stroke. However, data on cellular and molecular mechanisms of neonatal stroke remain scarce.
CD36, a type B scavenger receptor, is a multifunctional transmembrane protein expressed in many cell types including, microglia, monocytes, macrophages, epithelial and endothelial cells [
10]. CD36 exerts multiple biological functions by mediating innate immunity, fatty acid transport and lipid signalling, assembly of inflammatory pathways in lipid rafts and ROS production [
11]. CD36 also mediates several steps in phagocytosis of apoptotic material, including recognition of apoptotic cells, engulfment and digestion of apoptotic bodies [
12‐
14]. Furthermore, CD36 serves as an anti-angiogenic factor and is involved in enhancement of platelet aggregation [
15]. CD36-mediated effects are ligand-specific and context-dependent [
10,
16] and are mediated via its cooperation with other receptors, including vitronectin and Toll-like receptors (TLR), TLR2, TLR4 and TLR6 [
17,
18]. CD36 utilizes multiple ligands, such as phospholipids, advanced glycation end products, TSP-1 and oxidized low-density lipoprotein (OxLDL). In stroke and under neurodegenerative conditions in the adult and ageing brain, CD36 has been found harmful (reviewed in [
19]). While the presence of CD36 exacerbates injury after acute transient middle cerebral artery occlusion (tMCAO) in the adult [
13], in neonatal mice we observed much higher incidence of severe injury after acute tMCAO in mice that lack CD36 [
20]. Many CD36 ligands are low at birth, likely contributing to the differing CD36 effects between neonatal and adult brain. Thus far we identified several divergent aspects of CD36 signalling in neonates compared to adults after tMCAO. While superoxide accumulation was largely responsible for injury in the adult [
13], we observed similar superoxide accumulation in the vessels and in macrophages in ischemic-reperfused regions of acutely injured WT and CD36 KO neonates [
21]. At the same time, we showed that lack of CD36 in injured neonatal mice affects microglial morphology, limits phagocytosis and reshapes intracellular lipid signalling dependent on a Src kinase Lyn [
21].
Knowing that CD36 plays a key role in shaping macrophage phenotypes after injury [
14,
22,
23] and that the choroid plexus (CP) serves as a gateway for early leukocyte accumulation after neonatal stroke [
24], we examined effects of global CD36 knockout on leukocyte trafficking and the transcriptome in the CPs after neonatal stroke. We demonstrate that CD36 mediates recruitment of multiple leukocyte subtypes to the CP ipsilateral to tMCAO early after reperfusion, potentially contributing to increased number of CD45
highCD11b
+ cells in ischemic-reperfused brain parenchyma. tMCAO triggers marked changes in gene expression in the CP of WT pups, including activation of genes involved in inflammatory, metabolic and extracellular matrix signalling, changes that are further modified by lack of CD36.
Materials and methods
Animals
All research conducted on animals was approved by the University of California San Francisco Institutional Animal Care and Use Committee and in accordance with the Guide for the Care and Use of Laboratory Animals (U.S. Department of Health and Human Services). Animals were given ad libitum access to food and water, housed with nesting material and shelters, and kept in rooms with temperature control and light/dark cycles. The data are in compliance with the ARRIVE guidelines (Animal Research: Reporting in Vivo Experiments). Block litter design and randomization within individual litters were used. Blinded data analysis was used where possible.
Transient middle cerebral artery occlusion (tMCAO)
tMCAO was performed on postnatal day 9 (P9)-P10 C57BL/6 wild type (WT; purchased from Charles River) mice and CD36 KO mice of both sexes, as we previously described [
21]. Briefly, a midline cervical incision was made under isoflurane anesthesia, the common carotid artery and internal carotid artery (ICA) exposed, single threads from a 7-0 silk suture used to temporary tie a knot below the origin of the ICA to prevent retrograde bleeding from the arteriotomy. A coated 8-0 nylon suture was advanced 4–5 mm and removed 3 h later. In sham-operated pups suture was inserted but not advanced. Mice from the same litters were randomized to receive tMCAO or sham surgery. Temperature was maintained with temperature controlled blanket and overhead lamp. Based on our historic diffusion-weighted MRI data in the model and the presence of recirculation upon suture removal, as evident using intra-jugular injection of FITC–isolectin B4 [
6,
21,
25], the incidence of injury is > 70% and no bleeding associated with reperfusion. Data for male and female pups were combined based on our published data on similar injury in male and female pups at 72 h [
25] and unpublished data for outcomes at 1–4 weeks of reperfusion in several mouse lines on C57BL/6 background.
Histology and immunofluorescence
Animals were perfused and post-fixed with 4% PFA. Post-fixed, cryoprotected and flash frozen brains were sectioned on a cryostat (12 μm thick serial sections, 360 μm apart). Double-immunofluorescence was performed on adjacent sections blocked in 10% NGS/PBST and incubated overnight in 2% NGS/PBST with rabbit anti-mouse GLUT-1 (1:500, AbCAM), rabbit anti-Iba1 (1:500, WAKO), anti-mouse Timp1 (1:200, TFS) followed by appropriate secondary antibodies purchased from Invitrogen and DAPI. Z-stacks of 10–12 images were captured at 1.0 μm intervals (25 ×/100 × oil objectives, Zeiss Axiovert 100 equipped with Volocity Software, Improvision/Perkin Elmer) and analysis performed in four–five fields of view (FOV) per hemisphere/region in the ischemic-reperfused cortex and matching contralateral tissue using automated protocols for signal intensity threshold (> 2SD background in each channel) in a 1 × 106μm3 voxel.
Myeloid cell isolation
Mice deeply anesthetized with isoflurane were transcardially perfused with cold PBS, red blood cells were lysed using ACK lysis buffer, and cells were washed with RPMI. Brain and CP tissue were minced with razor blades and pushed through 70 μm nylon cell strainers, placed in 1 mg/ml collagenase for 45 min at 37 °C inverted frequently. Cells from brain tissue were washed, resuspended in 70% Percoll and overlaid with 30% Percoll. The cells were centrifuged at 2400 rpm for 30 min at 4 °C without brake. The interface was removed and washed before plating.
Flow cytometry
Single-cell myelin-free suspensions from contralateral and injured regions were plated (5 × 10
5/96 well), centrifuged, pellet resuspended in 100 μl blocking buffer containing CD16/32 (1:70, Biolegend) and incubated in 150 μl FACs staining buffer containing 2% FBS. For intracellular cytokine staining, cells were fixed (Fixation and Permeabilization kit, BD Bioscience), incubated with antibody mixture on ice for 20 min, washed, centrifuged, resuspended in staining buffer and evaluated on BD LSRII flow cytometer (BD Biosciences). Fluorescence Minus One (FMO) samples, a commonly used strategy to prevent false positive results through overlap of fluorophores [
26], was applied. The following combinations of antibodies diluted 1:200 in FACS staining buffer were used: anti-CD45-Pacific Blue (Biolegend), anti-CD11b-APC-Cy7 (Biolegend), Ly6g (IA8)-AF700 (Biolegend), Ly6c (Hk1.4)-APC (Biolegend), CD206-FITC (Biolegend), IL-10-PE-Cy7 (Biolegend), CD86-FITC (Biolegend), IL-1β-PE (Biolegend). Compensation beads (BD CompBeads) were incubated in Fixation and Permeabilization solution (100 μl, 4 °C, 20 min), incubated with antibody mixture (4 °C, 30 min) and resuspended in staining buffer. Gating and data analysis were performed using FlowJo software (Tree Star).
Western blot
Western blot analysis was performed in cortical lysates using anti-Timp1 (1:200), TLR4 (1:200) and β-actin (1:5000 Sigma Aldrich) antibodies diluted in 5% milk in 0.2% Tween 20/TBS, 4 °C, overnight.
RNA sequencing of the CP
The RNA extraction and RNA sequencing process was performed as described previously [
27]. Briefly, the CP of mice subjected to tMCAO were collected from ipsilateral and contralateral side of the brain 3 h after reperfusion. After collection, CPs were placed in RNAlater (Ambion) for 24 h at 4 °C to preserve the RNA. Following a brief centrifuge, the RNAlater was removed and the CPs were stored at − 80 °C. Total RNA was extracted using Trizol (Qiazol,Qiagen) and miRNAeasy Kit (Qiagen) following manufacturer protocols. RNA quality was assessed using Experion Automated Electrophoresis System (Bio-Rad) and 500 ng of RNA was used for library preparation using TruSeq Stranded mRNA and Total RNA Sample Preparation Kit with Rib-Zero Gold (Illumina). The CPs from 5 mice were pooled together before RNA extraction. In sham-operated mice, ipsilateral and contralateral CP were mixed and samples from 3 mice were pooled. The sequencing was performed on a NovaSeq 6000 (Illumina) with a read length of 2 × 100. Mouse assembly mm10 from UCSC was used as reference genome.
RNA sequencing data analysis
Differential expression was obtained using DEseq2 with Benjamini–Hochberg correction for multiple comparisons. Venny diagram v.2.1 was used for visualizing the number of differentially regulated genes in different groups. Gene cluster analysis was performed using the Qlucore Omics Analyser v.3.6 on top 1000 genes differentially regulated between groups (
q < 0.05,
F test and ANOVA). Enrichment analysis of gene clusters was performed using g:Profiler [
28]. Ingenuity Pathway Analysis (IPA,Qiagen) was used to identify canonical pathways related to the regulated genes.
Statistical analysis
Block litter design was used to avoid litter-to-litter variability, randomization for tMCAO/sham was used where possible, data for all figures were obtained on unsexed neonatal mice. Each dot on all graphs represents an individual mouse. Two-way ANOVA with post-hoc Turkey’s Multiple Comparison test was performed for comparing groups with multiple variables, as described in legends to individual figures. GraphPad Prism 8 software was utilized to generate statistical data. Differences were considered significant at p < 0.05. Results are shown as mean ± SD.
Discussion
We demonstrate for the first time that CD36 mediates step-wise leukocyte recruitment to the CP early after tMCAO in neonatal mice, first with a predominant increase in inflammatory monocytes and neutrophils, followed by a delayed increase in the accumulation of patrolling monocytes. These effects parallel early increase in the number of CD45hiCD11b+ cells in ischemic-reperfused brain parenchyma. RNA sequencing analysis in the CPs reveals that lack of CD36 attenuates the inflammatory gene profile induced by ischemia–reperfusion and alters the makeup of the CP ECM. Together, these data suggest cooperation between peripheral and brain CD36—mediated signalling following neonatal stroke via the CP.
CD36 was shown to contribute to stroke, neurodegenerative diseases and dementia and is being considered as a therapeutic target during adulthood and aging [
13,
17,
30‐
34]. CD36 deficiency led to protection of adult mice from injury after tMCAO via several mechanisms that include reduced superoxide accumulation in activated microglia/macrophages and in the vasculature [
13], a shift toward anti-oxidative [
30] and anti-inflammatory [
32] brain status and attenuation in inflammasome activation [
34]. In the adult, synergy of CD36 signalling between the periphery and the injured brain after stroke [
35], along with induced BBB leakage [
36], has been demonstrated. Conditional deletion of CD36 either in microglia or endothelium reduced ischemic injury in adult mice, demonstrating the injurious involvement of CD36 in both cell types [
34]. While these findings attest to an injurious role of CD36, its role in monocytes and macrophages in mediating phagocytosis during the resolution phase of stroke has also been demonstrated [
37], suggesting potentially beneficial aspects of CD36 signalling. Interestingly, only preventative, but not post-stroke inhibition of CD36 attenuated brain swelling in hyperlipidemic stroke [
38].
In a neonatal tMCAO model, we previously showed that while the magnitude of injury on diffusion-weighted MRI during MCAO was similar in WT and CD36 KO, lack of CD36 substantially increased incidence of severe acute injury, in part by omitting CD36-dependent phagocytosis of dead neurons 24 h after reperfusion [
21]. More severe injury in CD36 KO pups was associated with increased gene and protein levels of monocyte/microglial chemoattractant MCP-1 and increased number of CD11b
+TLR2
+ cells in injured regions but, interestingly, several aspects of CD36 inflammatory signalling reported in adult stroke were not present. For example, NfkB activation and superoxide accumulation in injured regions of neonates were unaffected by lack of CD36 [
21], suggesting brain maturation-dependent inflammatory responses of CD36.
Here, in WT mice, we report rapid accumulation of multiple myeloid subtypes at 3 and 13 h after reperfusion in the CP ipsilateral to tMCAO as well as parallel early increase in the number of CD45+/CD11b+ cells in ischemic-reperfused regions. In the CP of CD36 KO pups, in turn, accumulation of CD45hiCD11b+ cells is significantly attenuated, including CD45hiCD11b+Ly6chi inflammatory monocytes and CD45+11b+Ly6cintLy6g+ neutrophils. Reduction of leukocyte trafficking might be due to inhibition of certain inflammatory and chemotactic pathways required for leukocyte recruitment to the inflammation site, as suggested by IPA data. Examination of the phenotypes of cells accumulated in the ipsilateral CP, inflammatory vs. anti-inflammatory, also showed largely unaffected numbers of CD45+CD11b+CD206+IL-10+ cells but lower numbers of inflammatory CD45+CD11b+CD86+IL-1β+ cells in CD36 KO pups compared to WT pups at both post-reperfusion timepoints, suggesting that lack of CD36 is anti-inflammatory in the CP early after injury. In the ischemic-reperfused parenchyma, attenuation of CD45hiCD11b+ cells in CD36 KO pups is transient at 3 h, as the overall numbers of myeloid cells in the ipsilateral cortex at 13 h do not depend on CD36 presence. Nonetheless, the accumulation of both inflammatory monocytes and neutrophils are lower at both timepoints in the injured cortex of WT mice compared to CD36 KO mice.
Considering that the dynamic interactions at the CSF-brain interface early after reperfusion can contribute to the forming/evolving inflammatory response, we examined gene expression in the CP at this early timepoint after tMCAO, at 3 h reperfusion. Although the CP is not directly affected by cerebral blood flow reduction in this model, tMCAO itself induced rapid profound changes in gene expression in the CP ipsilateral to the occlusion. Most notably, we demonstrate downregulation of genes related to DNA repair and activation of inflammation and metabolic processes. Furthermore, while under more basal conditions (sham), gene expression is similar in both WT and CD36 KO CPs, the magnitude of significantly regulated genes in ipsilateral CP is muted in CD36 KO, i.e., ~ threefold lower than in WT pups (1228 vs. 4135 genes), consistent with attenuated leukocyte accumulation in CP of CD36 KO mice after tMCAO. The inflammatory and metabolic pathways are among those that were most prominently affected. Gene clustering analysis and IPA suggested that genetic deletion of CD36 diminishes several key inflammatory pathways including expression of one of the key upstream inflammatory receptors, TLR4, which signals in part by formation of TLR4-TLR6 complex in response to sterile inflammation [
17,
18]. Consistent with lower gene expression, protein expression of TLR4 was significantly lower in CP in ischemic-reperfused CD36 KO than in WT, not only validating RNAseq data but also demonstrating that lack of its co-receptor, CD36, attenuates TLR4 upregulation and TLR4–mediated effects. These data expand the importance of CD36-TLR2/4/6 interactions for inflammatory signalling from adult [
17] to neonatal brain. Increased expression of anti-inflammatory genes and corresponding proteins, such as a classic anti-inflammatory cytokine IL-10, may serve as another modulator of leukocyte trafficking, especially when associated with reduced IL-1β levels in cells of monocyte lineage.
Recruitment of leukocytes to the inflammation site requires rearrangements in the cytoskeleton of leukocytes and the ECM of endothelium and epithelium of affected tissue [
40]. We previously showed that TLR2-mediated leukocyte trafficking through the CP leads to structural changes in basement membrane of CP epithelium and transcriptomic changes in cytoskeleton genes [
27,
41]. Here, we show that lack of CD36 markedly alters expression of several genes related to the ECM following tMCAO, including an MMP inhibitor Timp1, as well as key endothelial adhesion molecules. While ECM can act as a barrier to invading cells, it can catalyse leukocyte adhesion via interaction of adhesion molecules expressed on leukocytes (e.g., integrins) as well as chemokines [
42,
43]. Several enzymes with peptidase activity (MMPs and ADAMs) in the ECM were upregulated after tMCAO in WT CP, but not in the CD36 KO, which coincides with upregulation of the MMP inhibitor, Timp1, in the CP of WT but not CD36 KO mice, suggesting a negative feedback loop.
A number of findings warrant further investigation. For example, CD36 itself can act as a pattern recognition receptor that responds to ligands such as OxLDL and activates inflammasome pathways [
34,
44]. It is possible that oxLDL and other endogenous ligands are released after stroke, activating CD36-inflammasome signalling in the CP and contributing to inflammatory response [
19]. In support, expression of Nlrp3, a key inflammasome pathway component, was reduced in CD36 KO mice. Our data show very dynamic changes in the patterns of myeloid cell accumulation between 3 and 13 h after reperfusion, which may depend on the magnitude of building inflammatory response in the CSF and in the parenchyma, therefore, relationships between early events in the CP and more severe injury in the parenchyma of CD36 KO are not trivial. In addition, intriguingly, expression of some genes within the inflammatory response GO term with relation to coagulation were higher in CP of CD36 KO mice. Recently, platelets’ CD36 has been shown to promote thrombosis [
39]. Therefore, upregulation of some coagulation genes in CD36 KO may be a compensatory mechanism [
45]. Another unexplored aspect in our study is cell-type specific effect of CD36 signalling in multiple cell types. In adult brain injury models, microglial CD36 has been proposed as a key determinant of post-ischemic IL-1β production in mice by regulating caspase-1 activity, whereas endothelial CD36 is supposedly needed for the full expression of the endothelial activation induced by IL-1β [
34]. While this study does not allow distinguishing effects in individual cell types due to global CD36 deletion, future studies entail understanding how the presence of CD36 on specific cell types influences neonatal neuroinflammation. Translational impact of our study is demonstration of brain maturation-dependent effects of CD36 after stroke. Further implications of CD36 in neonatal stroke as compared to adult stroke are to be revealed by the use of pharmacological inhibitors/activators of CD36 in the neonatal stroke model.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.