Microglia are not necessary for maintenance of blood-brain barrier properties in health, but PLX5622 alters brain endothelial cholesterol metabolism

Abstract

Microglia, the resident immune cells of the central nervous system, are intimately involved in the brain's most basic processes, from pruning neural synapses during development to preventing excessive neuronal activity throughout life. Studies have reported both helpful and harmful roles for microglia at the blood-brain barrier (BBB) in the context of disease. However, less is known about microglia-endothelial cell interactions in the healthy brain. To investigate the role of microglia at a healthy BBB, we used the colony-stimulating factor 1 receptor (CSF1R) inhibitor PLX5622 to deplete microglia and analyzed the BBB ultrastructure, permeability, and transcriptome. Interestingly, we found that, despite their direct contact with endothelial cells, microglia are not necessary for the maintenance of BBB structure, function, or gene expression in the healthy brain. However, we found that PLX5622 treatment alters brain endothelial cholesterol metabolism. This effect was independent from microglial depletion, suggesting that PLX5622 has off-target effects on brain vasculature.

Keywords: CSF1R; LDLR; PLX5622; blood-brain barrier; cholesterol; endothelial cells; genetic microglial depletion; microglia; off-target effects; vasculature.

Figure 1.. Microglia are not required for maintenance of BBB properties in healthy young adult mice.

(A) Representative images of cortical sections from adult C57BL/6 mice on control or PLX5622 diet for one month. Sections were stained for IBA1. Scale bar=50μm. (B) Quantification of cortical IBA1+ microglia after one month diet, presented as percentage of microglial density normalized to the average control density. Each data point represents one mouse. n=3 mice/group; p<0.0001, unpaired two-tailed t-test. (C) Representative TEM images of vessel cross-sections of cortical tissue from mice after one month diet. Images show examples of tight junctions (arrows, top row) and invaginating vesicles (arrows, bottom row). Scale bars=500 nm. (D) Quantification of tight junction length in EM images. n=3 mice/group, 20 vessel cross-sections/mouse; p=0.400, Mann-Whitney test. (E) Quantification of invaginating vesicles per vessel cross-section in EM images. n=3 mice/group, 20 vessel cross-sections/mouse; p>0.999, Mann-Whitney test. (F) C57BL/6 mice were tested for BBB permeability to sodium fluorescein after one month diet. n=9-12; p=0.183, unpaired two-tailed t-test. (G) C57BL/6 mice were tested for BBB permeability to rhodamine 123 after one month diet. n=12; p=0.150, unpaired two-tailed t-test. (H) Representative images of cortical sections from Csf1r^ΔFIRE/+ and Csf1r^{ΔFIRE/ΔFIRE} mice stained for IBA1. Scale bar=50 μm. (I) Adult Csf1r^{ΔFIRE/ΔFIRE} mice and Csf1r^ΔFIRE/+ or Csf1r^+/+ mice were tested for BBB permeability to sodium fluorescein. n=6; p=0.868, unpaired two-tailed t-test. (J) RNA sequencing was performed on ECs from C57BL/6 mice fed control or PLX5622 diet for one month. Expression is presented in transcripts per million. n=5 mice per group, p-adj>0.05 for all genes. (K) Representative images of cortex from C57BL/6 after one month diet. Sections were stained for CD31 (green) and either claudin 5 (CLDN5), occludin (OCLN), or p-glycoprotein (P-GP)(magenta). Scale bars=50 μm. (L) Western blot on homogenized cortex of mice after one month diet. Membrane was cut and stained for key BBB proteins: CLDN5, OCLN, and P-GP, with beta-actin as a loading control. All images are from the same membrane. (M) Quantification of band sizes in blot from (L). Each CLDN5 band was normalized to the respective beta-actin band in that lane. n=4 mice/group; p=0.709, unpaired two-tailed t-test. (N) Quantification of band sizes in blot from (L). Each OCLN band was normalized to the respective beta-actin band in that lane. n=4 mice/group; p=0.178, unpaired two-tailed t-test. (O) Quantification of band sizes in blot from (L). Each P-GP band was normalized to the respective beta-actin band in that lane. n=4 mice/group; p=0.986, unpaired two-tailed t-test. All error bars represent SEM.

Figure 2.. PLX5622 treatment increases brain endothelial expression of cholesterol-related genes.

(A) Cholesterol-related gene expression in brain ECs from C57BL/6 mice after one month on control or PLX5622 diet. Gene expression is presented as the log₂(FC) from the average of control expression. Each column is a sample from one male (m) or female (f) mouse. Asterisks indicate FDR adjusted p-value (*p-adj<0.05; **p-adj<0.01; ***p-adj<0.001; ****p-adj<0.0001). (B) Representative images of cortical sections from adult C57BL/6 mice after one month control or PLX5622 diet. Sections were stained for CD31 (green) and LDLR (magenta). Scale bar=50μm. (C) Quantification of data in (B) examining percentage of CD31+ vascular length that is also LDLR+. n=5; p=0.0002, unpaired two-tailed t-test with Welch’s correction. (D) Quantification of vascular LDLR in cortical sections from adult CBA/J mice after one week on control or PLX5622 diet. n=4; p=0.0006, unpaired two-tailed t-test with Welch’s correction. (E) RT-qPCR validation of PLX5622-induced expression changes in Ldlr. Each data point represents one sample of isolated vessels pooled from 3 mice. n=3 samples; p=0.0033, unpaired two-tailed t-test. (F) RT-qPCR validation of PLX5622-induced expression changes in Dchr24. Each data point represents one sample of isolated vessels pooled from 3 mice. n=3 samples; p=0.0009, unpaired two-tailed t-test. (G) UMAP plot showing 11 endothelial clusters from single-cell RNA sequencing of ECs isolated from adult C57BL/6 mice after one month on control or PLX5622 diet. (H) The UMAP plot shown in (G) pseudo-colored based on brain EC identity as venous (light green), capillary (turquoise), or arterial (dark blue). Identity was determined based on gene expression profiles. (I) Dot-plot of brain EC expression of cholesterol-related genes in control and PLX5622 diet conditions. Dot size represents percentage of ECs expressing that gene, and dot color represents average expression on a log scale (J) UMAP plots in which brain endothelial cells are colored based on their expression of Hmgcr, Dhcr24, Ldlr, and Abca1. All error bars represent SEM.

Figure 3.. PLX5622 increases expression of cholesterol synthesis and uptake genes specifically in central nervous system ECs.

(A) Gene expression of cell-type markers in a mixed population of astrocyte, oligodendrocyte, and neuronal mRNA isolated from brains after one month on control or PLX5622 diet. Expression shown in transcripts per million. n=3 mice. (B) Cholesterol-related gene expression in the mixed population of astrocytes, oligodendrocytes, and neurons after one month on control or PLX5622 diet. Expression is represented as log₂(FC) from the average of expression in control samples. (C) Representative images of brain, retina, and spinal cord from mice fed control or PLX5622 diet for one week. Sections were stained for CD31 (green) and LDLR (magenta). Arrowheads denote examples of LDLR+ vessels. Scale bars=50μm. (D) Representative images of thymus, kidney, and muscle from mice fed control or PLX5622 diet for one week. Sections were stained for CD31 (green) and LDLR (magenta). Scale bars=50μm. (E) Expression of pan-endothelial, BBB-enriched, and periphery-enriched genes in the control population of liver ECs shown in (F) compared to the control population of BBB ECs from Fig 1J and 2A. Expression shown in transcripts per million. n=3-5. (F) Cholesterol-related gene expression in liver ECs after one month of control or PLX5622 diet. Expression is represented as log2(FC) from the average of expression in control samples. p-adj>0.05 for all genes. All error bars represent SEM.

Figure 4.. PLX5622-mediated increase in LDLR expression is independent of microglial depletion.

(A) Expression of genes most strongly bidirectionally regulated by neuronal activity in activating, silencing, and PLX5622 diet conditions. Expression is presented as log2(FC) from the average of each condition’s respective controls. Of the neuronal activity-dependent EC genes, only Fam13a, Sqle, and Sema6d were significantly altered by PLX5622 (p-adj=0.0161; p-adj=3.39E-13; p-adj=0.0206). (B) Quantification of cortical microglial density in 8- to 12-week-old C57BL/6 mice fed control or PLX5622 diet for 6, 12, 24, 48, or 72 hours. Data are represented as a percentage of microglial density normalized to the average of control density. n=5 mice/group/timepoint; 6 hr: p=0.623; 12 hr: p=0.181; 24 hr: p=0.0138; 48 hr: p=0.00101; 72 hr: p<0.0001, unpaired two-tailed t-tests with Welch’s correction. (C) Quantification of vascular LDLR in the same samples from (B). n=5 mice/group/timepoint, 3 sections/mouse; 6 hr: p=0.404; 12 hr: p=0.000157; 24 hr: p<0.0001; 48 hr: p<0.0001; 72 hr: p=0.000864, unpaired two-tailed t-tests with Welch’s correction. (D) Representative images of cortex from mice on control or PLX5622 diet for 6 hrs. Sections were stained for CD31 (green) and LDLR (magenta). Scale bar=50μm. (E) Quantification of microglia contacting CD31+ vasculature in the cortex at the 6 hr diet timepoint (images in D). Number of microglia with soma and/or processes in contact with vasculature was divided by the total number of microglia per image to calculate the percentage of microglia making vascular contact. n=3 mice/group, p=0.0162, unpaired two-tailed t-test with Welch’s correction. (F) LDLR− and LDLR+ vasculature was analyzed for microglial contacts after 12 hours of PLX5622. n=4 mice/group, p=0.610, unpaired two-tailed t-test with Welch’s correction. (G) Quantification of cortical microglial density during microglial repopulation. n=5 mice/group. Control vs depleted, control vs 24 hr, and control vs 48 hr: all p<0.0001. Depleted vs 24 hr: p=0.978; depleted vs 48 hr: p=0.183; 24 vs 48 hr: p=0.336; one-way ANOVA with Tukey’s multiple comparisons test. (H) Quantification of vascular LDLR in the samples described in (G). n=5 mice/group. Control vs depleted: p<0.0001; control vs 24 hr: p=0.0038. Depleted vs 24 hr: p=0.0032; depleted vs 48 hr: p<0.0001; control vs 48 hr: p=0.999. One-way ANOVA with Tukey’s multiple comparison test. (I) Linear regression analysis of vascular LDLR and microglial density at 24 hours from data shown in (G-H). n=5, p=0.103, simple linear regression. (J) Linear regression analysis of vascular LDLR and microglial density at 48 hours from data shown in (G-H). n=5, p=0.780, simple linear regression. (K) Representative images of cortex from 3- to 4-month-old Cx3cr1-CreER^T2; Csf1r^f/f mice and Csf1r^f/f controls. Sections stained for IBA1 (cyan), CD31 (green) and LDLR (magenta). Scale bar=50μm. (L) Representative images of cortex from P17-20 Csf1r^−/− and wildtype littermate controls. Sections stained for IBA1 (cyan), CD31 (green) and LDLR (magenta). Scale bar=50μm. (M) Representative images of cortex from 2- to 4.5-month-old Csf1r^{ΔFIRE/ΔFIRE} mice and littermate controls 22 hours after treatment with PLX5622 or vehicle. Sections stained for IBA1 (cyan), CD31 (green) and LDLR (magenta). Scale bars=50μm. (N) Quantification of microglial density in 3- to 4-month-old Cx3cr1-CreER^T2; Csf1r^f/f mice and Csf1r^f/f controls. n=3 mice/group, p=0.0049, unpaired two-tailed t-test with Welch’s correction. (O) Quantification of percent CD31+ vascular length that was also LDLR+ in Cx3cr1-CreER^T2; Csf1r^f/f mice and Csf1r^f/f controls. n=3 mice/group, p=0.511, unpaired two-tailed t-test with Welch’s correction. (P) Quantification of percent CD31+ vascular length that was also LDLR+ in Csf1r^−/− mice and Csf1r^+/+ controls. n=5 mice/group, p=0.364, unpaired two-tailed t-test with Welch’s correction. (Q) Quantification of percent CD31+ vascular length that was also LDLR+ in Csf1r^{ΔFIRE/ΔFIRE} mice and littermate controls 22 hours after treatment with PLX5622 or vehicle. n=5-6 mice/group. Control vs FIRE, vehicle: p-adj.>0.999; control vs FIRE, PLX5622: p-adj.>0.999; vehicle vs PLX5622 in control: p-adj.=0.0034; vehicle vs PLX5622 in FIRE: p-adj.=0.0164, 2-way ANOVA with Bonferroni correction for multiple comparisons. (R) Published single-cell RNA sequencing data (McNamara et al.) from brains of Csf1r^{ΔFIRE/ΔFIRE} and Csf1r^+/+ mice. For several cholesterol-related genes, average raw counts across all cells in each genotype condition were compared. n=6-7 samples/genotype, p-adj.>0.05 for all genes shown. (S) Representative images of middle cerebral artery (MCA)-adjacent parenchymal arterioles isolated from C57BL/6 mice and cultured with PLX5622 or vehicle stained for DAPI (blue), CD31 (green) and LDLR (magenta). Scale bar=50μm. (T) Quantification of LDLR in cultured arterioles shown in S. Each symbol type (circle, square, triangle) denotes data from a particular mouse. n=9 vessels from 3 mice/group, p<0.0001, unpaired two-tailed t-test. (U) Representative images of cortex from Cdh5-CreER^T2,· Csf1r^f/f and Csf1r^f/+ littermate controls stained for CD31 (green) and LDLR (magenta). Scale bars=50 μm. (V) Quantification of vascular LDLR in Cdh5-CreER^T2; Csf1r^f/f mice and Csf1r^f/+ littermates. n=3 mice/group, p=0.924, unpaired two-tailed t-test. All error bars represent SEM.