Aberrant oligodendroglial-vascular interactions disrupt the blood-brain barrier, triggering CNS inflammation

Abstract

Disruption of the blood-brain barrier (BBB) is critical to initiation and perpetuation of disease in multiple sclerosis (MS). We report an interaction between oligodendroglia and vasculature in MS that distinguishes human white matter injury from normal rodent demyelinating injury. We find perivascular clustering of oligodendrocyte precursor cells (OPCs) in certain active MS lesions, representing an inability to properly detach from vessels following perivascular migration. Perivascular OPCs can themselves disrupt the BBB, interfering with astrocyte endfeet and endothelial tight junction integrity, resulting in altered vascular permeability and an associated CNS inflammation. Aberrant Wnt tone in OPCs mediates their dysfunctional vascular detachment and also leads to OPC secretion of Wif1, which interferes with Wnt ligand function on endothelial tight junction integrity. Evidence for this defective oligodendroglial-vascular interaction in MS suggests that aberrant OPC perivascular migration not only impairs their lesion recruitment but can also act as a disease perpetuator via disruption of the BBB.

Fig. 1:. Perivascular clustering of OPCs in multiple sclerosis.

(a) Schematic of human Multiple sclerosis (MS) showing lesions throughout the brain. MS cases assessed are outlined in Suppl. Table 1. (b) Low magnification of MS Case 1 active lesion, with luxol fast blue (LFB) to assess demyelination and LN3 immunohistochemistry to assess inflammatory cell activity. Boxed area shows area of magnified view in (c), and from where images (c-g) were taken. (d) Magnified region from Case 1 showing a cluster of cells (unfilled arrow) which do not co-localize with inflammatory marker LN3 (filled arrows). (e, f) Clustered perivascular OLIG2+ cells (arrows) and (g) clustered RNF43+ cells (arrow) in Case 1 active lesion. (h-i) Clustered OLIG2+ cells (arrows) on blood vessels in active lesions of MS Cases 2 (h) and 3 (i). (j) MS Case 4 chronic active lesion, with boxed areas showing regions where images from (k) and (l) were taken. (k) Chronic core of lesion showing hypocellularity and limited LN3+ activity. Example of isolated OLIG2+ cell (arrow) in close association with blood vessel. (l) Active edge of chronic active lesion with significant LN3+ inflammatory activity and perivascular clustering of NKX2.2+ cells (arrows). (m-n) Clustering of OLIG2+ cells is not seen in normal appearing grey matter (NAGM) or normal appearing white matter (NAWM) from these patients. (o) Frequency of OLIG2+ clusters seen in NAGM, NAWM, and areas of chronic inactive and active lesions from the patients described in Suppl. Table 1. Data were analyzed by one-way ANOVA. The measure of center represents the mean. Active vs. Chronic p= 0.04, Active vs. NAWM p=0.003, Active vs. NAGM p=0.003. Scale bars, 2.2mm (b), 90μm (c, k left panel, l left panel), 20μm (d-i, m,n, k right panel, l right panel), 750μm (j). * P < 0.05, ** P < 0.01.

Fig. 2:. OPC perivascular clusters following demyelination represent a defective single cell migration.

OPCs use single cell perivascular migration for their recruitment to areas of demyelination. (a-b) OPCs show increased association with vasculature at lesion edges following focal demyelination in mice. Double staining for Olig2+ (a) or PDGFRα+ (b) cells and CD31+ vasculature before lysolecithin lesioning (“No lesion”), and at lesion edges at 2 days post lesioning (dpl) in mouse corpus callosum and in spinal cord (Suppl. Fig. 3a). (c) Percentage of PDGFRα+ OPC cell bodies directly in contact with blood vessels at lesion edges at different days post lesioning in mouse spinal cord dorsal funiculus (n=6 animals at each time, all statistical analyses compared to 0d, 1d vs. 0d p=9.66 E-7, 2d vs. 0d p=6.94 E-5, 3d vs. 0d p=0.0003, 5d vs. 0d p= 0.0161, 7d vs. 0d ns p=0.1477, 14d vs. 0d ns p=0.5022). Data were analyzed by unpaired two-sided Student’s t test. (d) Time lapse imaging in slice cultures of adult spinal cord dorsal funiculus from 1.5dpl lysolecithin lesioned NG2creERT:TdTomato mice, following intracardiac infusion of fluorescein-lectin for vessel labeling. A TdTomato-expressing cell (red, arrow) migrates along a vessel (green, outlined by dotted line) at the edge of a 1.5dpl lesion (Suppl. Video 1). (e) Frequency of PDGFRα+ OPC clusters at 10dpl lesion edges in PDGFRα-creERT2:APC fl/fl mice compared to APCfl/fl controls (n=4 animals, p=0.0240). Data were analyzed by unpaired two-sided Student’s t test. (f-h) Perivascular clustering is mediated by high Wnt tone in OPCs during murine remyelination, as evidenced by PDGFRα+ and Olig2+ cell clusters in PDGFRα-creERT2:APC fl/fl (f) and Olig2cre:APC fl/fl (g) mice (which are not seen in control mice, e), and by GFP+/ PDGFRα+ double positive cell clusters in PDGFRα-creERT2:APC fl/fl:Rosa-GFP (h), at 10dpl lesion edges following focal lysolecithin injection into spinal cord dorsal funiculus white matter. Scale bars, 20μm in all panels. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Values are mean ± s.d.

Fig. 3:. OPC perivascular clusters physically displace astrocyte end feet from vessels.

(a) GFP labels astrocytes and their end feet (arrow, in close association with CD31+ vasculature) in P9 corpus callosum of APCfl/fl:Aldh1l1-GFP control mice, but when crossed into Olig2cre: APC fl/fl mice (b), end feet are displaced (arrow in b) from vessels in areas with OPC perivascular clusters (stained with dapi). (c) Percentage of blood vessels with OPC clusters (P12 vs. P9 p=2.51 E-8, P16 vs. P9 p=1.93 E-8, P20 vs. P9 p=9.92 E-8, P30 vs. P9 p=1.04 E-7) and percentage without astrocyte endfeet (P12 vs. P9 p=5.85 E-8, P16 vs. P9 p=2.44 E-8, P20 vs. P9 p=1.96 E-8, P30 vs. P9 p=1.81 E-8) at different postnatal times in CC of Olig2cre:APCfl/fl:Aldh1l1-GFP (n=6 animals at each time, all statistical analyses compared to P9). Data were analyzed by unpaired two-sided Student’s t test. d) 3D reconstruction of astrocytes (in green) covering vessels (in blue) in CC of P9 Olig2cre:APC+/+:TdTomato:Aldh1l1-GFP control mice. (e) P9 CC from Olig2cre:APCfl/fl:TdTomato:Aldh1l1-GFP showing perivascular OPC clusters (red), vessels (white) and astrocytes (green). (f) 3D reconstruction of images in (e) showing OPCs (red), astrocytes (green) and vessels (blue). Astrocyte processes are completely displaced from areas with OPC perivascular clusters (arrows). (g) Quantitative analysis of percentage loss of Aldh1L1-GFP+ astrocytic endfeet coverage of vessels in P9 CC of Olig2cre:APCfl/fl:Aldh1l1-GFP versus APCfl/fl:Aldh1l1-GFP controls (n=6 animals, p=1.56 E-7). Data were analyzed by unpaired two-sided Student’s t test. (h-i) PDGFRα+ OPC perivascular clusters are seen in P9 spinal cord of Olig2-cre:APC fl/fl mice (i), but not in APCfl/fl controls (h), and lead to gaps in astrocyte end foot marker Aquaporin 4 (Aqp4)(arrows in i, quantified in Suppl. Fig. 8e). Scale bars, 20μm in all panels. **** P < 0.0001. Values are mean ± s.d.

Fig. 4:. OPC perivascular clusters disrupt the BBB.

(a-b) Staining for plasmalemma vesicle-associated protein (PLVAP) and claudin5 (Cldn5) in P9 spinal cord of Olig2-cre:APC fl/fl and APCfl/fl control mice around vessels in longitudinal (a) and transverse (b) views. A PLVAP+/claudin5- state around OPC clusters in Olig2-cre: APC fl/fl mice (arrows) identifies areas of immature BBB. (c) Ratio of PLVAP:Cldn5 expression on endothelium in areas with OPC perivascular clusters in Olig2-cre:APC fl/fl spinal cord compared to comparable regions without clusters in APCfl/fl controls (n=6 animals, p=3.29 E-4). Data were analyzed by unpaired two-sided Student’s t test. (d-e) Staining shows fibrinogen only within the lumen of CD31+ blood vessels in P9 APCfl/fl control spinal cord (SC) and brain (d), but leaking into parenchyma (arrows) around OPC clusters (stained with dapi) in Olig2-cre:APC fl/fl (e). (f) Quantification of area of fibrinogen staining outside blood vessels in CNS parenchyma per mm² in SC of Olig2-cre:APC fl/fl and APCfl/fl controls (n=6 animals, p=0.0026). Data were analyzed by unpaired two-sided Student’s t test. (g-h) Dextran- tetramethylrhodamine (rhodamine) injection into P9 Olig2-cre:APC fl/fl tail vein (h) shows extravasation into brain parenchyma (arrows in h) around OPC perivascular clusters (stained with dapi), which is not seen in APCfl/fl control mice (g). (i) Quantification of area of rhodamine staining outside blood vessels in CNS parenchyma per mm² in SC of Olig2-cre:APC fl/fl and APCfl/fl controls (n=6 animals, p=6.46 E-6). Data were analyzed by unpaired two-sided Student’s t test. Scale bars, 20μm in all panels. ** P < 0.01, *** P < 0.001, **** P < 0.0001. Values are mean ± s.d.

Fig. 5:. OPC perivascular clusters trigger a CNS inflammation

(a-b) Marked upregulation of CD11c expression in P9 Olig2-cre:APC fl/fl brain compared to controls around OPC perivascular clusters (stained with dapi), quantified in (b)(n=4 animals; p=0.009). Data were analyzed by unpaired two-sided Student’s t test. (c) As Cd11c can detect some other immune cells in addition to activated microglia and macrophages, we made use of CX3CR1-GFP: CCR2-RFP mice (which label microglia green, and macrophages red) crossed into Olig2cre:APC fl/fl, identifying these cells as predominantly activated microglia around OPC clusters. (d) Frequency of F4/80 expressing cells at different postnatal times in CC in Olig2-cre:APC fl/fl versus controls (n=6 animals at each time, all statistical analyses compare Olig2cre:APC fl/fl to controls at each time, P9 p=2.28 E-6, P12 p=1.94 E-5, P16 p=1.39 E-4, P30 p=0.5490). Data were analyzed by unpaired two-sided Student’s t test. (e-f) Extravasation of small numbers of CD3+, CD4+ and CD8+ T cells within OPC perivascular clusters in P9 Olig2 cre: APC fl/fl mouse brain (f) which are not seen in APCfl/fl controls (e). (g) Quantification of CD3+ cell numbers per perivascular cluster in Olig2-cre:APC fl/fl versus controls (n=6 animals, p=0.0006). Data were analyzed by unpaired two-sided Student’s t test. (h-i) SMI-32+ (non phosphorylated neurofilament) axonal spheroids co-localize with the neurofilament marker NF200 (NF), indicating swellings of damaged axons adjacent to an OPC perivascular cluster (arrows in i, stained with dapi) in Olig2cre:APC fl/fl P9 brain, which are not seen in controls (h). (j) Quantification of SMI32+/NF200+ axonal dots within 100μm distance of clusters in Olig2cre:APCfl/fl P9 CC versus similar areas in APCfl/fl (n=4 animals, p=0.0286). Data were analyzed by unpaired two-sided Student’s t test. (k) Quantification of OPC perivascular cluster size in Olig2 cre: APC fl/fl: TdTomato corpus callosum at times between P9 and P30 (n=6 animals, all statistical analyses compared to P9, P12 vs. P9 p=0.3846, P16 vs. P9 p=0.0007, P20 vs. P9 p=3.37 E-6, P30 vs. P9 p=2.55 E-8). Data were analyzed by unpaired two-sided Student’s t test. (l-m) Tunel staining in P9 mouse brain co-localizes with (l) PDGFRα (in Olig2cre:APC fl/fl) and with (m) TdTomato (in Olig2cre:APC fl/fl: TdTomato, arrows) indicating OPC death within perivascular clusters. (n) Percentage of TdTomato+ cells that are TUNEL+ in perivascular clusters versus non clusters in Olig2cre:APCfl/fl:TdTomato (n=6 animals, p=0.0009). Data were analyzed by unpaired two-sided Student’s t test.Scale bars, 10μm (l), 20μm (all other panels). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Values are mean ± s.d.

Fig. 6:. Wif1 production by Wnt activated perivascular OPCs disrupts EC tight junctions.

(a-c) Wif1 mRNA is highly upregulated by Wnt activation in OPCs in Olig2cre:APCfl/fl mice in vivo in P9 mouse (a) spinal cord and (b) corpus callosum and (c) in vitro (fold change by qPCR) in isolated OPCs from Olig2cre:APCfl/fl compared to APCfl/fl (n=3 independent experiments, ****p=5.43E-5), and in isolated WT OPCs treated with Wnt3a compared to untreated CTL (n=3 independent experiments, **p=0.0093). Data were analyzed by unpaired two-sided Student’s t test (c). (d) Significantly upregulated Wif1 mRNA in vivo co-localizes with Olig2 protein in clustered cells in Olig2cre:APCfl/fl corpus callosum. (e-f) WIF1 is expressed by clustered OLIG2+ cells in MS active lesion white matter (f). (g) WIF1 protein expression by cells around CD31+ vasculature (arrows) in MS lesion white matter. (h) Tight junctions stained for Claudin5 (Cldn5) protein in isolated CNS endothelial cells (ECs) treated with control PBS, Wif1, Wnt3a, or Wif1+Wnt3a for 2 days, and (i) Claudin5 fluorescent intensity sums (as fold change compared to PBS control) in EC cultures under these treatment conditions (n=4 independent experiments; PBS vs. Wif1 *p=0.0235; PBS vs. Wnt3a ****p=2.59 E-5; Wnt3a vs. Wif1+Wnt3a ####p=3.03 E-6). Data were analyzed by one-way ANOVA. (j) Claudin5+ tight junctions in isolated ECs treated for 2 days with conditioned medium from APCfl/fl or Olig2cre:APC fl/fl OPCs (WT-CM+BSA and Olig2cre:APC fl/fl-CM+BSA respectively) or conditioned medium from the same OPCs depleted of Wif1 protein by overnight anti-Wif1 antibody (+Ab) treatment and agarose bead pull down, and (k) Claudin5 fluorescent intensity sums (as fold change compared to WT-CM+BSA control) in EC cultures under these treatment conditions (n=4 independent experiments; WT-CM+BSA vs. APC-CM+BSA ****p=3.13 E-7; APC-CM+BSA vs. APC-CM+Ab ####p=5.49 E-5). Data were analyzed by one-way ANOVA. (l) ELISA OD values for Wif1 protein concentrations in conditioned medium from these groups (n=4 independent experiments; WT-CM+BSA vs. APC-CM+BSA ****p=6.79 E-8; WT-CM+AB vs. APC-CM+Ab ###p=0.0002). Data were analyzed by one-way ANOVA. (m-n) Fold change of mRNA in cultured ECs (by qPCR relative to control) for tight junction marker Claudin5, endothelial fenestrae marker PLVAP, and Wnt pathway activation marker Axin2 in the Wif1/Wnt3a treatment groups from Fig.4h (m), and the OPC conditioned medium treatment groups from Fig. 4j (n). (m) n=4 independent experiments; Claudin5: PBS vs. Wif1 *p=0.0381; PBS vs. Wnt3a ****p=3.86 E-7; Wnt3a vs. Wif1+Wnt3a ####p=6.16 E-5; PLVAP: PBS vs. Wif1 ****p=1.46 E-11; PBS vs. Wnt3a *p=0.0137; Wnt3a vs. Wif1+Wnt3a ##p=0.0017; Axin2: PBS vs. Wif1 *p=0.0338; PBS vs. Wnt3a ****p=8.97 E-9; Wnt3a vs. Wif1+Wnt3a ####p=8.07 E-5. Data were analyzed by one-way ANOVA. (n) n=4 independent experiments; Claudin5: WT-CM+BSA vs. APC-CM+BSA **p=0.0019; APC-CM+BSA vs. APC-CM+Ab #p=0.018; PLVAP: WT-CM+BSA vs. APC-CM+BSA ****p=3.93 E-6; APC-CM+BSA vs. APC-CM+Ab ##p=0.00217; Axin2: WT-CM+BSA vs. APC-CM+BSA **p=0.00164; APC-CM+BSA vs. APC-CM+Ab ##p=0.00103; In these three biomarkers, WT-CM+BSA vs. WT-CM+Ab has no significant difference. Data were analyzed by one-way ANOVA. (o) In vitro BBB endothelial permeability assay with control PBS, Wif1, Wnt3a, or Wif1+Wnt3a (n=3 independent experiments; PBS vs. Wnt3a **p=0.0063; PBS vs. Wif1 ***p=0.0004; Wif1 vs. Wif1+Wnt3a ####p=8.22 E-5). Data were analyzed by one-way ANOVA. Scale bars, 90μm (a, b), 20μm (d left panel, g), 10μm (d right panel, e, f, h, j). Values are mean ± s.d.