Abstract
Cerebral small vessel disease and resulting white matter pathologies are worsened by cardiovascular risk factors including obesity. The molecular changes in cerebral endothelial cells caused by chronic cerebrovascular risk factors remain unknown. We developed a novel approach for molecular profiling of chronically injured cerebral endothelial cells using cell-specific translating ribosome affinity purification (RiboTag) with RNA-seq in Tie2-Cre:RiboTag mice. We used this approach to identify the transcriptome of white matter endothelial cells after the onset of diet-induced obesity (DIO). DIO induces an IL-17B signaling pathway that acts on the cerebral endothelia through IL-17Rb leading to increased endothelial expression of CXCL5 in both the DIO mouse model and in aged humans where cerebral small vessel disease is common. In the white matter, endothelial CXCL5 acts as a chemoattractant and promotes the association of oligodendrocyte progenitor cells (OPCs) with cerebral endothelia increasing vessel-associated OPC cell number and surface area. Targeted blockade of IL-17B with peripheral antibody administration reduced the population of vessel-associated OPCs by reducing endothelial CXCL5 expression. CXCL5-mediated sequestration of OPCs to white matter vasculature impairs OPC differentiation after a focal white matter ischemic lesion. DIO promotes a unique white matter endothelial-to-oligodendrocyte progenitor cell signaling pathway that compromises brain repair after stroke.
Significance Statement Chronic cardiovascular risk factors damage cerebral small vessels yet the molecular pathways induced in injured cerebral endothelial cells that lead to tissue injury are unknown. Obesity-induced endothelial expression of CXCL5 in brain white matter disrupts normal myelination by forcing oligodendrocyte progenitor cells to associate with the cerebral vasculature and impedes brain repair after stroke.
Introduction
Cerebral small vessel disease is an age-related entity affecting brain white matter. The resulting white matter lesions accumulate over time (1) and contribute to disability (2), dementia (3–5), and death (6). Cerebral small vessel injury is significantly worsened by chronic cardiovascular risk factors such as hypertension, diabetes, and obesity (7–10). In particular, abdominal obesity and its associated metabolic disturbances in blood pressure, lipids, and blood sugar control increase the risk of developing white matter lesions on MRI (11–14) and increase the likelihood of lacunar brain infarction or stroke (15). While the pathologic changes associated with cerebral small vessel disease are well known (16, 17), the molecular pathways that drive small vessel injury in the brain are largely unknown.
Emerging data suggests that an interaction between cerebral vessels and cells of the oligodendrocyte lineage play a key role in maintaining white matter homeostasis (18–20). A subset of platelet-derived growth factor receptor alpha-positive (PDGFRα+) oligodendrocyte progenitor cells (OPCs) closely associate with the vasculature (21) and use it to migrate in the brain during development (22). Proteins secreted by endothelial cells promote OPC migration and proliferation in vitro (23, 24). In the spontaneously hypertensive rat model of cerebral small vessel disease, the OPC population is increased in association with vascular changes and delays in OPC maturation may be mediated by endothelial secretion of HSP90α (25). Both the diagnosis and treatment of cerebral small vessel disease would be advanced by identifying additional molecular pathways active in cerebral endothelia and driven by chronic cardiovascular risk factors (26).
To identify the molecular changes in white matter endothelia in the setting of chronic cardiovascular risk factors, we used a mouse model of diet-induced obesity (DIO) (27) that recapitulates a number of features of human cardiovascular risk (28). We demonstrate that DIO is associated with a loss of white matter vasculature, increases in the number of OPCs in brain white matter, thinner myelin, and disrupted axons. We used cell-specific translating ribosome affinity purification and RNA-sequencing in Tie2-Cre:RiboTag mice to isolate the endothelial-specific transcriptome after the onset of DIO. Pathway analysis of the top up-regulated genes indicates that DIO induces an IL-17B signaling pathway that acts on the cerebral endothelia through the up-regulation of IL-17Rb leading to increased endothelial expression of CXCL5. Endothelial expression of CXCL5 is increased in cerebral white matter vessels of aged humans. In the mouse model, endothelial over-expression of CXCL5 directly signals to OPCs acting as a chemoattractant in vivo. DIO-induced endothelial expression of this immune signaling pathway exacerbates the white matter injury response to a focal white matter ischemic lesion and restricts the maturation of OPCs during the repair phase after stroke. These findings indicate that DIO promotes cerebral endothelial inflammation and through the expression of CXCL5 exerts vascular regulation of myelination by sequestering the main progenitor cell of the adult brain, the OPC, to blood vessels. This pathway has direct implications for the understanding of human cerebral small vessel disease and repair of cerebral white matter.
Results
Diet-induced obesity as a model of white matter and vascular injury
Obesity is a significant risk factor for the development of small vessel disease and white matter injury (10, 11, 13, 14). We used a well-established model of diet-induced obesity (DIO) (27) to model the effects of chronic cardiovascular risk on brain white matter and the vasculature. Beginning at 8 weeks of age, mice were fed a control fat diet (CFD) or a high fat diet (HFD) for 12 weeks. After 12 weeks on the dietary intervention, mice exhibit 84% weight gain and metabolic disturbances in cholesterol and blood sugar (Fig. S1) broadly consistent with the diagnostic criteria for metabolic syndrome (29). After the development of obesity, we examined the vasculature and cellular makeup of the white matter.
Using a Tie2-Cre;tdTomato (Ai14) strain that robustly labels the vasculature throughout the brain, we characterized the effects of DIO on the white matter vasculature throughout the corpus callosum (Fig. 1A). DIO reduces the volume of tdT+ vessels by 26.0% and the branch complexity of the vasculature (Fig. 1B). DIO also resulted in an increase in the PDGFRα+ oligodendrocyte progenitor cells (OPCs) within the corpus callosum and a concordant increase in OPCs associated with vessels, measured as OPCs per unit vessel length (Fig. 1C-D). Consistent with a DIO-induced increase in OPCs, we observed fewer and shorter axonal paranodal segments (Fig. S2A) as well as thinner myelin with an increased g-ratio as measured by electron microscopy (Fig. S2B), indicating compromise of white matter integrity in DIO mice. We developed a direct RNA hybridization assay for oligodendrocyte staging based on overall marker gene expression patterns in white matter, clustered by the three main stages of oligodendrocyte development. The top 40 genes marking OPCs, pre-myelinating oligodendrocytes (PMO), and myelinating oligodendrocytes (MO) (30) were used to indicate oligodendrocyte stages in mice on Using this 120 gene expression platform (Supplemental Data File 1; Fig. 1E, Fig. S2C), we found an increase in OPC gene expression in DIO white matter, with animals on HFD clustering more closely with genetically defined OPCs, consistent with light microscopy findings. Together, these findings suggested that DIO increases OPC proliferation, increases the number of OPCs associated with vessels, and biases the oligodendrocyte lineage towards immaturity.
Molecular profiling of white matter endothelia
Systemic cardiovascular risk factors such as DIO exert their effect on white matter by primarily damaging the cerebrovasculature. To identify the molecular pathways induced in white matter endothelia, we developed an approach using Tie2-Cre:RiboTag mice together with translating ribosome affinity purification using the RiboTag method (31) (Fig. 2A). This transgenic approach leads to robust HA labeling in the cerebrovasculature (Fig. 2B). HA-immunoprecipitated RNA using this approach shows endothelial specificity, with a specific enrichment of endothelial transcripts (Fig. 2C) compared to established marker genes for pericytes and OPCs (30). DIO results in a specific gene expression profile compared to endothelial cells from normal weight mice (Fig. S3A; SI Data File 2) within white matter endothelia: 112 genes are up-regulated and 60 genes are down-regulated (FDR<0.1, Fig. 2D). Gene ontology of the up-regulated genes suggests enrichment of immune signaling pathways including C-X-C chemokine signaling and interleukin receptor activation (Fig. S3B). Among the top differentially regulated genes, interleukin-17 receptor b (IL17Rb) (8.83-fold increased, FDR=0.090) and its effector chemokine Cxcl5 (11.35-fold increased, FDR=0.064) were two of the most strongly up-regulated genes when comparing DIO vs. control animals (Fig. 2E).
IL-17Rb and CXCL5 up-regulation in human and murine white matter vasculature
IL-17 signaling involves five interleukin ligands (A-E) and five cognate receptor isoforms that hetero and/or homo-dimerize to effect downstream signaling (32). Within our transcriptional dataset, the only IL-17 receptor isoform that was significantly differentially regulated in DIO-affected cerebral endothelial cells was IL-17Rb (Table S1). Among a number of diverse functions, IL-17 receptor signaling activates effector chemokine signaling, including CXCL5 (33) as a mechanism of identifying tissue injury. CXCL5 is a member of the C-X-C chemokine family (34) acts a chemoattractant in other tissues and has been reportedly up-regulated in white matter after peri-natal hypoxia (35). Guided by our RNA-seq data, we hypothesized that DIO may induce IL-17B signaling acting through IL-17Rb resulting in increased endothelial expression of CXCL5 (Fig. 3A). To confirm RNA-seq results, we performed TRAP-qPCR using independent Tie2-Cre:RiboTag biologic replicates for a subset of differentially regulated genes (Glut-1, Itgb3, Cd180, Hsd3b3, Tnfrsf10b, Il17rb, Cxcl5, and Ttc21a ) (Fig. 3B). Similar degrees of up-regulation for Il17rb and Cxcl5 using qPCR were seen. Retro-orbital venous blood sampling confirmed increased serum detection of CXCL5 in DIO mice (Fig. 3C). Immunofluorescent labeling for IL-17Rb (Fig. 3D) and CXCL5 (Fig. 3E) in Tie2-Cre;tdTomato (Ai14) mice demonstrated a marked increase in detection of both molecules within white matter cerebral vessels in DIO mice.
To verify the relevance of this DIO-induced cerebrovascular molecular pathway to human cerebral small vessel disease, we examined endothelial CXCL5 expression in a series of older (86±8 years of age) human post-mortem specimens with (n =5) and without (n =5) a pathologic diagnosis of cerebral vascular disease sufficient to influence cognition in the setting of low levels of Alzheimer’s disease pathology. All subjects with vascular disease had combined evidence of macro- and microscopic cerebral vascular disease (Table S2). Sections containing frontal peri-ventricular white matter were immunolabeled for CXCL5 (Fig. S4). In this older cohort, CXCL5 is robustly detected in cerebral vessel segments within white matter, with 80% demonstrating at least some CXCL5 staining within white matter vasculature, while the mean percentage of vessel segments showing CXCL5 staining was 71.2±0.08% (17.2±3.4 vessel segments/subject) (Fig. 3F). In this small aged cohort, white matter vessel CXCL5-positivity did not associate with a clinical diagnosis of vascular dementia. However, most, if not all individuals living to his age have some degree of frontal white matter injury indicating that small vessel disease is nearly ubiquitous at advanced age. In fact, 70% of these cases had pathologic evidence of cerebral vascular injury (Table S2).
The IL-17/CXCL5 pathway as a novel vessel-to-OPC signaling paradigm
With the known role of chemokine receptor (CXCR) signaling on OPC migration (22), we reasoned that endothelial up-regulation of CXCL5 in DIO mice may function to promote OPC migration to the vasculature. To this end, we observed a notable increase in OPCs that were in close apposition to CXCL5+ vessel segments (Fig. 4A) in DIO mice. In vitro exposure of O4+ OPCs to increasing doses of recombinant murine CXCL5 resulted in a dose-dependent increase in OPC cell area with cytoskeletal changes suggesting motility (Fig. 4B). To determine the ability of endothelial CXCL5 to signal to OPCs in vivo, we used a combined transgenic and targeted viral gene expression approach (Fig. 4C). We designed a pCDH-FLEX-CXCL5-T2A-GFP lentiviral construct and injected this virus into the subcortical white matter of Tie2-Cre;tdTomato mice resulting in targeted gene expression specifically in white matter vasculature (Fig. S5). After 6 weeks of endothelial up-regulation of CXCL5-GFP or GFP in normal weight mice, we measured the distance of individual OPCs from vessels and the cell area of vessel-associated OPCs (Fig. S6), as well as the number of OPCs per unit vessel length (Fig. S7). The average distance of OPCs from tdT+ vessels was reduced in CXCL5-GFP injected animals compared to GFP injected animals while the number of PDGFRα+ OPCs in apposition to tdT+ vessels were increased (upper panels Fig. 4C, 4E, 4F) supporting a chemoattractant role for CXCL5 on OPCs. Similar to the effects of recombinant CXCL5 on OPCs in vitro, endothelial over-expression of CXCL5 in vivo resulted in increased OPC cell area (lower panels Fig. 4C, 4G).
To block DIO-induced endothelial CXCL5 expression resulting from IL-17Rb activation, we employed repetitive peripheral injections of a function-blocking anti-IL-17B antibody or isotype control IgG for 6 weeks in Tie2-Cre;tdTomato mice on HFD (Fig. 4D, S1D). Endothelial CXCL5 expression within the tdT+ vasculature was reduced by 60.4% using this approach (Fig. 4H) while IL-17Rb levels were not changed (Fig. S8) indicating that DIO-induced increases in endothelial CXCL5 can be at least partially regulated through IL-17B signaling at the endothelial cell surface. Peripheral blocking of IL-17B signaling significantly reduced the association of OPCs with the cerebral vasculature in DIO mice (upper panels Fig. 4D, 4E, 4F). Vessel-associated OPC cell area was not significantly different in HFD mice administered anti-IL-17B antibody (lower panels Fig. 4D, Fig. 4G), suggesting additional HFD-associated pathways drive the regulation of OPC cell area.
Endothelial CXCL5 exaggerates the OPC response to focal white matter ischemia and impedes remyelination after stroke
To determine the consequence of endothelial CXCL5 on injury response after focal white matter ischemia, we used an established model of white matter stroke (36, 37) (Fig. 5A) that produces a distinct population of stroke-responsive PDGFRα+ OPCs (38). At 7d after white matter stroke, there was no significant difference in the stroke lesion volume comparing animals on CFD vs. HFD (Fig. S9). To determine the role that DIO-induced endothelial CXCL5 expression influences injury response after focal white matter ischemia, we labeled for GLUT-1 and CXCL5 at 7d post-stroke. We measured the percentage of CXCL5+ voxels that co-localized with GLUT-1 within the peri-infarct tissue surrounding the stroke (Fig. S10). As in uninjured white matter, the percentage of CXCL5+/GLUT-1+ voxels were significantly increased within the peri-infarct tissue in animals on HFD (Fig. 5B). Immunofluorescent labeling for PDGFRα+ OPCs identified an increase in stroke-responsive OPCs per lesion in DIO mice compared to control (Fig. 5C) at 7d post-stroke. Spatial mapping of stroke-responsive OPCs coupled with nearest neighbor comparative analysis (Fig. S11) indicates a greater distribution of stroke-responsive OPCs that specifically occurs at the lesion margins in DIO mice compared to control (Fig. 5D) where endothelial CXCL5 levels were increased. This increase is at least partially accounted for by an increased association of OPCs with CXCL5+ vessel segments within the peri-lesional tissue. This DIO-induced OPC-vessel interaction in the early phase impacts repair after stroke. At 28d after stroke, we compared PDGFRα+ OPC and GST-π+ mature oligodendrocyte cell counts in three regions of interest spanning the ischemic white matter lesion. This analysis revealed a significant change in oligodendrocyte cell populations 28d after stroke (p=0.0011, two-way ANOVA, F=14.47) (Fig. 5E). An increased number of residual stroke-responsive PDGFRα+ OPCs were present at 28d post-stroke in animals on HFD compared to CFD (adjusted p=0.0114). The number of GST-π+ mature oligodendrocytes within the lesion at 28d post-stroke was variable and generally reduced in animals on HFD compared to CFD (adjusted p=0.0654).
Discussion
Cerebral small vessel disease is increasingly recognized as a substantial contributor to stroke risk and dementia (39). Microvascular injury in the brain is driven by cardiovascular risk factors yet molecular factors that link systemic vascular risk factors with molecular pathways in the brain are lacking. Obesity is a major cardiovascular and cerebral vascular risk factor, is growing prevalence (40), is associated with white matter changes in humans (11, 13–15), and has a reliable animal model (27). Using a diet-induced obesity model, we show that obesity reduces white matter vasculature and increases OPCs in chronically injured white matter, as reported in other models (25). Our results showing ultrastructural changes in myelin in adult onset diet-induced obesity are similar to those seen in genetically obese (ob/ob) mice with reductions in myelin (41) and increases in OPCs in leptin-deficient ob/ob mice (42) validating this model for the study of chronic white matter injury. The present findings are the first to identify the transcriptome of chronically injured cerebral endothelia. We used that dataset to identify disordered vascular signaling that acts to regulate OPCs and impairs remyelination after stroke.
Cell-specific transcriptional profiling using ribosomal tagging is a valuable tool in parsing out molecular signals from a complex tissue such as the brain (31, 43). Here, we developed a methodology to profile cerebral endothelial cells in vivo. By combining this vascular Ribotag mouse with a chronic vascular risk factor model, we identified novel endothelial pathways that appear relevant to human cerebral small vessel disease. A similar vascular profiling approach could be easily applied to identify microvascular injury signals in other organs such as the kidney or retina. It could also be applied to other chronic or acute neurologic conditions that feature microvascular injury including aging, diabetes, or isolated hypertension. Such efforts to better characterize the molecular pathways relevant to human cerebral small vessel disease is crucial for the development of diagnostic and therapeutic interventions (26). While we focused on the paracrine effects of chronically injured endothelial cells and signaling into the white matter, the identification of the endothelial response to chronic cerebrovascular risk factors may also facilitate the development of novel fluid-based biomarkers to track the response of the cerebral endothelium to chronic risk factors.
Vessels and OPCs are known to interact both during development and to maintain white matter homeostasis (44). During CNS development, OPCs migrate extensively to distribute throughout the entire CNS and this migration requires the physical vascular scaffold (22). Cerebral endothelial cells secrete trophic factors that activate Src and Akt signaling pathways to support the survival and proliferation of OPCs (18). However, the full spectrum of molecular pathways that drive the vessel-OPC interaction remain largely unknown. The present data in disease and studies in the developing brain indicate that chemokines are critical. In-vivo time lapse imaging reveals that in the developing mouse brain, OPCs interact with vasculature and migrate along the vessels to the destined cerebral regions dependent on CXCR4 activation in OPCs, which binds to endothelial secreted ligand CXCL12, and promotes their attraction to cerebral vasculature (45). Our study illustrates a similar phenomenon, with DIO-induced endothelial expression of CXCL5 promoting the association of OPCs to the vasculature within adult white matter in vivo. These results further imply that chemokine signaling pathways play a significant role in regulating the interaction between endothelial cells (ECs) and OPCs that is critical to white matter maintenance and the response to injury.
What is the consequence of sequestering OPCs to the vascular bed? In the spontaneously hypertensive rat model of cerebral small vessel disease in which endothelial cells are progressively injured, OPCs are increased and fail to mature properly. In this model, dysfunctional endothelial cells impair the maturation of OPCs in vitro and promote their proliferation through the production of HSP90α (25). In the DIO mouse model, we observed a reduction in white matter microvascular complexity with the surviving endothelial cells responding with a specific transcriptional response implicating growth factor and immune signaling. In part, this response appears to cause OPCs to respond to vascular injury through CXCL5 signaling to OPCs. These OPCs are elongated along vessels with a long leading edge or display hypertrophied cell bodies and processes compared to those OPCs that are not associated with blood vessels, similar to those previously reported for migratory or reactive OPCs (22, 23, 46). These migratory or reactive OPCs are likely responding to disrupted vascular integrity but as a result of their new restriction to the vascular bed, fail to properly differentiate, ultimately compromising remyelination after stroke. This disordered vascular regulation of myelination provides a new concept in understanding cellular signaling in cerebral small vessel disease. Though we clearly demonstrate that CXCL5 can serve as this vascular injury regulatory signal in rodents and humans, further studies are needed to definitively prove whether this response serves some partially protective role on the blood-brain barrier through reciprocal OPC to endothelial signaling (47).
White matter ischemic lesions are characterized by a robust early loss of axons, myelin and oligodendrocytes (48–50). Similar to inflammatory white matter lesions (51), OPCs respond early and robustly to white matter ischemic lesions common to the aging human brain (38, 49, 52). The peri-infarct white matter at the margin of the ischemic lesion, often referred to as the white matter penumbral region (53), is where cross-talk between axons and oligodendrocytes is compromised (55, 56). Here, we used a novel approach to identify the spatial relationship of stroke-responsive OPCs to a focal white matter stroke lesion. This approach directly informs our data by showing that the increase in stroke-responsive OPCs produced by obesity occurs precisely in the peripheral margins of the white matter stroke lesion where tissue repair and the stimuli for remyelination would be maximal. In DIO mice, the stroke-responsive OPC lesion area is 30% larger and this expanded penumbral region is marked by increased endothelial CXCL5 expression explaining why more stroke-responsive OPCs are seen at the lesion periphery. The consequence of this is apparent at 28 days after stroke; DIO-induced endothelial expression of CXCL5 leaves behind a population of activated, injury-responsive OPCs whose maturation is inhibited. This progenitor restricted state could indicate that remyelination is simply delayed after stroke or it could lead to a progressive dysfunctional OPC response as the ability of NG2+ OPCs to differentiate into oligodendrocytes declines with chronic insults (57).
From our data, an emerging concept places the cerebral endothelial cell at the center of the pathophysiology relevant to cerebral small vessel disease. As the conduit between the brain and systemic insults such as hypertension, diabetes, and the metabolic disturbances of obesity, identifying molecular pathways in the cerebral endothelia represent a yet incompletely-realized target for understanding disease pathogenesis. With the tools described here, the molecular pathways activated in cerebral endothelia can now be deeply characterized. Changes in normal white matter homeostasis that result from chronic cerebrovascular risk factors can directly alter injury response and repair after stroke by acting through vascular regulation of myelination. That the pathway we characterized is functionally absent from normal young adult mice yet ubiquitous in aged human samples suggests greater efforts to appropriately model co-morbid conditions in animal models of stroke and cerebrovascular injury may pave a smoother path to translation into human trials.
Material and Methods
Animals
Mice were housed under UCLA regulation with a 12-hour dark-light cycle. All animal studies presented here were approved by the UCLA Animal Research Committee ARC#2014-067-01B, accredited by the AAALAC. Diet-induced obesity was induced in mice by ad lib feeding with 60%kCal from fat chow (HFD) or 10%kCal from fat chow (CFD) (Research Diets, Inc.). Weights (g) were measured weekly. Mice strains used in this study are described in SI Materials and Methods.
Translating ribosome affinity purification and RNA-sequencing
HA-tagged ribosomal associated RNAs from cerebral white matter endothelia were isolated and purified by Nucleospin miRNA kit (Machary-Nagel). RNA-sequencing was run using 69 bp paired end reads. Reads were aligned to the mouse genome using STAR (v.mm10). Differential gene expression analysis was performed using EdgeR assuming an FDR <0.1 as significant. Gene ontology analysis was performed using Enrichr (58). Additional information of RNA isolation, RNA sequencing and analysis are described in SI Materials and Methods.
Immunofluorescence and Confocal Imaging
Animals were euthanized with a lethal dose of isoflurane, transcardially perfused with PBS followed by 4% paraformaldehyde in 0.1 M sodium phosphate buffer, brains removed, post-fixed for 24 hrs and cryoprotected for 48 hrs in 30% sucrose in PBS. Forty microns coronal cryosections and immunostaining was performed essentially as described (36). Details regarding antibodies and microscopic imaging are available in SI Materials and Methods. Human post-mortem brain sections were selected from the UC Davis ADC Neuropathology Core samples based on a priori selection criteria and stained for CXCL5/6 using standard immunohistochemistry.
White matter stroke
Subcortical white matter ischemic injury was induced as previously described (37) using three stereotactic injections of L-Nio (L-N⁵-(1-Iminoethyl) ornithine, dihydrochloride; Calbiochem) into the subcortical white matter under sensorimotor cortex. Animals (n = 4/grp) were sacrificed at 7- or 28-days post-stroke and analyzed for tissue outcomes.
Lentiviral Injection
A dual promoter lentiviral backbone was created through sequential subcloning to place either GFP or murine CXCL5 between loxP sites. Control GFP and CXCL5-GFP lentivirus were packaged in human 293 cells (ATCC cat. no. CRL-11268) and concentrated by ultracentrifugation on a sucrose column. 200 nL of concentrated virus was injected into the subcortical white matter and allowed to express for 6 weeks.
Anti-IL-17B Antibody Administration
Anti-mIL-17B function blocking antibody (R&D, AF1709) was diluted with 0.9% saline to a concentration of 1mg/ml. Normal Goat isotype-matched IgG (R&D, AB-108-C) was used as control. Tie2-Cre;tdTomato mice were fed with high fat diet starting at 8 weeks old and weighed weekly. Aliquots of 50μg of anti-mIL-17B IgG or control IgG were prepared and administered in a blinded fashion every 72 hours by intraperitoneal injection from 14 weeks old and analyzed 48 hours after the last injection at 20 weeks old.
Statistical Analysis
The number of animals used in each experiment is listed in the Results section or Figure Legend. Statistical analysis was performed using GraphPad Prism 7 software. Unless otherwise stated, statistical significance was determined using α=0.05, corrected for multiple comparisons. Data are shown as mean ± SEM.
Other Methods
Further details regarding procedures related to gene expression and analysis, immunohistochemistry, electronic microscopy, spatial analysis, and statistical analysis are explained in detail in SI Materials and Methods. All DNA sequences, primers, plasmids and packaged viruses are available upon request. Gene expression data is available in the SI Data Files.
Author contributions
G.X., Ro.K., and J.D.H. designed research and collected data; G.X., Ri.K., Y.K., and J.D.H. performed gene expression analysis; G.X. and J.B. performed spatial analysis; G.X., L.A.H., and J.D.H. performed electron microscopy and analysis; A.B., M.M, S.T.C., and J.D.H performed lentiviral cloning and packaging; C.K.W, X.R.Z, H.V.V., V.K, G.X, and C.D. performed human brain tissue selection, staining, and analysis; G.X. and J.D.H. wrote the paper.
Materials and methods
Animals
Wild-type C57Bl/6 mice fed ad lib on 60%kCal from fat chow (HFD) (Strain #380050) or 10%kCal from fat chow (CFD) (Strain #380056) were purchased directly from Jackson Labs at 17 weeks of age and allowed to acclimate for 2 weeks prior to experimental use. The Tie2-Cre;tdTomato mice were generated by crossing Tie2-Cre mice with flox-stop tdTomato mice (Jackson Labs Strain #007908 – B6;129S6-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J).
Immunofluorescence
Animals were euthanized with a lethal dose of isoflurane, transcardially perfused with PBS followed by 4% paraformaldehyde in 0.1 M sodium phosphate buffer, brains removed, post-fixed for 24 hrs and cryoprotected for 48 hrs in 30% sucrose in PBS. Forty microns coronal cryosections and immunostaining was performed essentially as described [1]. The following primary antibodies were used: mouse anti-NF200 (1:200, Sigma), rabbit anti-MBP (1:500, Calbiochem), goat anti-PDGFRα (1:500; Neuromics), mouse anti-HA (1:1000, Biolegend), rabbit anti-AnkG (1:1000, Dako), rabbit anti-NaV1.6 (1:250), rabbit-Gst-π (1:1000, Millipore), rat anti-IL-17Rb (1:500, Santa Cruz Biotech), rat anti-CXCL5 (1:250, R&D) in PBS containing 5% goat or donkey serum and 0.3% Triton-X 100 (Sigma) overnight at 4°C. Secondary antibody labeling was performed using donkey anti-mouse, donkey anti-rabbit, or donkey anti-goat Fab2-Alexa conjugated antibodies (Jackson Immunoresearch, Inc.).
Immunohistochemistry for human brain samples
Case selection was made from a subset of 950 UC Davis ADC Neuropathology Core samples based on a priori selection criteria. In this case, we selected 5 cognitively normal individuals with low Braak and Braak scores and 5 demented subjects with low Braak and Braak scores that also had evidence of cerebrovascular disease sufficient to cause dementia. To be most consistent with current diagnostic criteria, we selected the most recent cases available based on the selection criteria. Immunohistochemistry was performed using formalin (Medical Chemical Corporation, 575A) fixed paraffin embedded tissue sections cut at 6μm. Sections where placed on positive charged slides (Fisherbrand, 12-550-15) then incubated overnight at 60°C. De-paraffinization was accomplished with three 5min xylene (Fisher Scientific, X3P) washes. The samples were rehydrated with graded concentrations of alcohol (American MasterTech, ALREACS) diluted with deionized water. Endogenous peroxidase was blocked with a 3% solution of hydrogen peroxide (Fisher Scientific, H325-500) 20min incubation. Heat-induced epitope retrieval used a citrate buffer (BioCare Medical, CB910M). The slides incubated in the buffer at 90°C for 45min. Blocking used 2.5% normal horse serum (Vector, S2012) for 60min. Antigen specificity was elucidated by incubating the slides for 90min in CXCL5/6 (1:100, Abcam, ab198505). Primary antibody detection was amplified with a 45min incubation using a secondary antibody (Vector, MP-7401). A 5 seconds counterstain used hematoxylin (Richard Allan Scientific, 7221). The samples were dehydrated with graded alcohols and three xylene washes before being coverslipped.
Microscopy and microscopic analysis
All microscopic images were obtained using a Nikon C2 confocal microscope. Tie2-Cre;tdTomato expressing vessels were used for the analyses of vessel volume, vessel length and junction point. Vessel volume was measured by Imaris software with automated “Add surface” function. Volume of small particle less than 30μm3 was subtracted to eliminate the background interference. Vessel length and junction point were analyzed by AngioTool [2]. The parameters for AngioTool measurement were set as “Diameter 5-40”, “Intensity 40-255” and “Particles less than 10000”.
The size of PDGFRα+ OPC was measured individually by Imaris automated “Add surface” function. Voxel of small particle less than 800 was subtracted to eliminate the background interference. OPC-vessel distance was measured by Imaris with “Add spot” function. For PDGFRα+ OPC location, nucleus with Dapi staining was used as a reference. The distance of Tie2-Cre;tdTomato vessel to PDGFRα+ OPC was measured with the function of “Spot to Spot closest distance”.
OPC number and x,y,z positional coordinates were analyzed by Imaris “Add spot” function. OPCs stroke responsive areas were generated by Imaris “Add Surface” function. Measurements of stroke areas were generated using Fiji [3]. Representative images were selected for presentation.
The levels of CXCL5/IL-17Rb in anti-IL-17B/IgG treated mice white matter were measured by Imaris “Coloc” function. The percentages of CXCL5/IL-17Rb that colocalized with Tie2cre;tdTomato positive vessels were measured as voxel areas. For GLUT-1/CXCL5 colocalization measurement, GLUT-1 positive vessels were masked by Imaris with “Add surface” to create new GLUT-1 and CXCL5 channels. The percentages of GLUT-1/CXCL5 colocalization in new channels were measured as voxel areas by Imaris “Coloc” function.
Electron microscopy
Wild-type C57Bl/6 mice (n =6/grp) on CFD or HFD were transcardially perfused with a 2% glutaraldehyde solution, post-fixed for 24 hrs, hemisected in the sagittal plane and 2 mm cubes including the corpus callosum were dissected and embedded in plastic resin for ultrastructural analysis as previously described [4]. One micron, plastic embedded toluidine blue stained sections were used to select transcallosal fibers underneath sensorimotor cortex by light microscopy. Three electron micrographs were obtained at a primary magnification of 7200X using a JEOL 100 CX transmission electron microscope and a representative electron micrograph of high technical quality from each animal was used for quantitation of fiber diameter, axon diameter, myelin thickness, and g-ratio.
Translating ribosome affinity purification and RNA-sequencing
After 12 weeks on CFD or HFD, Tie2-Cre:RiboTag mice (n = 3 per group) were sacrificed and the subcortical white matter was freshly dissected and immediately placed into buffer with RNase inhibitors. A tissue homogenate was generated using a 1mL sequential glass homogenizer. HA-tagged ribosomes from cerebral white matter endothelia were isolated following the established RiboTag protocol [5] using equal volumes of tissue homogenate. Total RNA was isolated from both input and IP samples using the Nucleospin miRNA kit (Machary-Nagel). Normalized RNA amounts (ng) underwent cDNA library generation using the TrueSeq with Ribozero kit preparation (Illumina), pooled and sequenced using 69 bp paired end reads. Samples were sequenced over 4 lanes for an average of read count of 62.1 ± 10.7 million per sample. Reads were aligned to the mouse genome using STAR (v.mm10). Differential gene expression analysis was performed using EdgeR assuming an FDR <0.1 as significant. Gene ontology analysis was performed using Enrichr [6].
Nanostring gene expression
Using established oligodendrocyte stage marker genes [7], we designed a custom Nanostring gene expression array (XT_GX CodeSet Oligostg #116000651) using gene-specific probes for each the 40 genes marking each oligodendrocyte stage (OPC, pre-myelinating oligodendrocyte (PMO), and myelinating oligodendrocyte (MO)). C57Bl/6 mice (n=4/group) were maintained on 12 weeks of CFD or HFD. At 20 weeks of age, the corpus callosum was isolated, homogenized, and RNA using the Nucleospin miRNA kit (Clontech) to collect both large and small RNA species. RNA was quantitated and 100 ng of RNA from each animal was provided as input for the direct RNA detection assay. Manufacturer protocol was followed and the results were analyzed using nCounter software. Results were normalized and differentially expressed genes were compared individually between groups using a Student’s t-test (p<0.05). The number of DEGs (p<0.05) per stage was determined. log2FC values were calculated as a function of averaged housekeeping gene expression levels (actb, b2m, gapdh, pgk1, rpl19). To generate an oligodendrocyte stage cell type index, raw gene expression values were combined with FPKM values from Zhang et al. (2014) and normalized by rank. Hierarchal clustering analysis was performed using hclust and tSNE plots (perplexity=3, max iteration = 5000) generated using standard algorithms. The resulting classification and 2-dimensional representation were highly reproducible.
Lentiviral preparation, packaging, and injection
A plasmid containing the open reading frame of the murine CXCL5 sequence with a 3’ stop codon was purchased from Origene (#MR200761). The pCDH-EF1-FLEX-EGFP-CMV-2A-TagBFP2-SC dual promoter lentiviral backbone was created by subcloning the FLEX-GFP sequence between the loxP sites from the pAAV-FLEX-GFP vector (Addgene #28304) into the pCDH-EF1-MCS-CMV-2A-pTagBFP2-SC dual promoter lentiviral construct using restriction digestion. The pCDH-EF1-FLEX-EGFP-CMV-2A-TagBFP2-SC backbone was linearized by removing the GFP sequence between the loxP sites using restriction digestion with XhoI and EcoRI (New England Biolabs). The murine CXCL5 sequence was PCR amplified in a HiFi DNA Assembly reaction (New England Biolabs) such that it was subcloned in the 3’>5’ position in between the loxP sites. The resulting reaction was transformed into Stbl3 E.coli cells and positive clones were identified by restriction digestion and verified by DNA sequencing. Subsequently, a 3’>5’ T2A-copGFP sequence was added 5’ to the murine CXCL5 sequence. The donor T2A-copGFP vector (pCDH-EF1-MCS-copGFP; System Biosciences) was PCR amplified and subcloned into pCR-Blunt II TOPO (ThermoFisher Scientific) for amplification and utilized in a HiFi DNA Assembly reaction. The resulting reaction was transformed as above and positive clones were identified by restriction digestion and DNA sequencing. DNA amplification was performed using an Endotoxin-Free PureLink Plasmid Midiprep Kit (ThermoFisher Scientific). Resulting DNA was quantified and used in lentiviral packaging. All DNA sequences, primers, plasmids and packaged viruses are available upon request.
T175 flasks were coated with Poly-L-Lysine for about 20min and allowed to air dry completely. 2.3×107 293T/17cells (ATCC cat. no. CRL-11268) were seeded into each T175 flask and cultured with 30 ml DMEM (Gibco cat.no. 11960-044)/10% FBS (Thermo Fisher HyClone SH30071.03) overnight with 3% CO2 at 37°C. Fresh DMEM/10% FBS was exchanged 4 hours before transfection. Four plasmids (60 μg pCDH-FLEX-CXCL5-GFP or pCDH-FLEX-GFP expression vector, 39 μg pMDLg/pRRE packing, 15 μg pRSV-REV plasmid and 21 μg pMD2.G envelope) were mixed with 0.750 ml of 1M CaCl2 and water to a total volume of 3 ml/tube. After filtering through a 0.2μm syringe filter, 2.8 ml of 2xBES (50 mM N,N-bis[2-hydroxyethyl]-2-aminoethanesulfonic acid, 280 mM NaCl, 1.5 mM Na2HPO4; pH 6.96) was added to a 2.7 ml of DNA/plasmids mixture in 14ml polystyrene tube and incubated on a vortex mixer for 15 minutes at room temperature. After incubation, DNA/plasmids mixture were transferred to a T175 flask, swirled gently and incubate at 37°C for 16 hours with 3% CO2. After overnight incubation, cells were washed with 1XPBS and incubated with fresh 28 ml DMEM/10%FBS and 10 mM sodium butyrate for 24 hours with 3% CO2. Supernatant was collected, centrifuged at 1500 RPM for 10 min and filtered through 0.45 μm filter. Virus supernatant was centrifuged at 20000 RPM for 2 hours 25 min and resuspended in 2 ml DPBS (Fisher cat.no. MT-21-031-CM). The resuspended supernatant was then loaded on 20% sucrose column and centrifuged at 32000 RPM for 2 hours 25 min. After centrifugation, the supernatant was discarded and the viral pellet was gently resuspended in 40 – 100 μl of DPBS. The virus suspension was transferred to a sterile Eppendorf tube and centrifuged at 7000 RPM for 5 min. Virus was aliquoted without disturbing the pellet and stored at −80°C.
Two stereotactic injections (A/P:0.55, M/L1.95, D/V:-1.30; A/P1.10, M/L:1.75, D/V: - 1.35) of 200 nL pCDH-FLEX-CXCL5-GFP or pCDH-FLEX-GFP were injected into subcortical white matter at 14 weeks old Tie2-Cre;tdTomato mice and allowed to express for 6 weeks. Animals were maintained with control fat diet and analyzed at 20 weeks old.
Spatial analysis
Analysis of the spatial distribution of stroke-responsive OPCs was performed as follows. The boundary of increased PDGFR-α-+ cells and the loss of GST-π-+ cells was identified in each of three sections per animal (n=3 animals/group). Using Imaris software, the x,y,z position of each PDGFR-α-+ cell relative to the user defined center point (x=0, y=0, z=0) of the elliptical stroke region was determined. Because the z-axis was limited (10 μm), a two-dimensional grid analysis was performed using a 2D modification of the previously reported 3D spatial density estimator [8] using a smoothing parameter of k=8. The local cell density in each position within the overlaid grid is compared statistically as previously described [9]. Therefore, a p-value map is generated for each position in the grid and thresholded (p<0.05) to reveal regions with significant density differences.
Statistics
The number of animals used in each experiment is listed in the Results section. Oligodendrocyte population cell counts as a fraction of total cells were determined by averaging counts from 5 fields of view throughout the corpus callosum across a minimum of three sections 240 μm apart. Per animal averages were generated and significance between groups determined using an unpaired Welch’s t-test (α=0.05). Analysis of nodal/paranodal complexes was performed using an unpaired Welch’s t-test (α=0.05), while paranodal length varies based on axonal diameter and was therefore analyzed using a distribution analysis and Chi-square statistic (df=12). Measurements of white matter ultrastructural features were averaged across animals and each feature was compared separately using Mann-Whitney U test between groups (α=0.05). Gene expression differences generated by Nanostring were determined at the individual gene level using unpaired Welch’s t-test (α=0.05). Determination of stroke lesion area was performed by sampling lesion area (n=3-5 40 μm sections) across groups (n=4/grp) and using the sampled distribution to create bootstrapped area distribution (n=25) representing a full area sampling of the approximate 1 mm lesion created by the stroke model. This area distribution was averaged across animals in each group and compared using a Mann-Whitney U test between groups (α=0.05). Spatial analysis of stroke-responsive OPCs were determined as detailed above. Cell counts at 28d post-stroke were determined across three sections 240 μm apart with lesion core and edge analyses determined using a two-way ANOVA (α=0.05) with post-hoc Holm-Sidak test to adjust p-values for multiple t-tests. Statistical analysis was performed using GraphPad Prism 7 software. Data are shown as mean ± SEM.
Study approval
All animal studies presented here were approved by the UCLA Animal Research Committee ARC#2014-067-01B, accredited by the AAALAC. All human subjects were involved in the UCD ADC longitudinal study IRB # 215830-46. Written informed consent for autopsy was obtained from the subject or next of kin prior to examination.
Data and materials availability
All DNA sequences, primers, plasmids and packaged viruses are available upon request. Gene expression data is available in the Supplemental Data Files.
Acknowledgments
The authors are grateful to Kelsey Ericson for preparation of human brain tissue specimens. This work was graciously supported by grants from the Larry L. Hillblom Foundation (TLLHF 2014-A-014), the American Heart Association Grant-in-Aid (16GRNT31080021). LAH and STC receive support from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. CD receives support from UC Davis Alzheimer’s Disease Center grant P30 AG 010129. JDH receives support from the National Institute for Neurological Disorders and Stroke (K08 NS083740) and the United States Department of Veterans Affairs Greater Los Angeles Healthcare System.