Nrf2 controls homeostatic transcriptional signatures and inflammatory responses in a cell-type specific manner in the adult mouse brain

Summary

Nrf2 is a promising therapeutic target for neurological disorders, but its mechanisms of action remain unclear. While often linked to anti-inflammatory effects, this is mostly based on studies involving pharmacological activation. In the brain, Nrf2 is highly expressed in microglia, astrocytes, and endothelial cells. As yet, the brain cell type-specific role of Nrf2 in regulating the basal transcriptome and controlling neuroinflammation is unknown. To address this, we employed three inducible conditional Nrf2 knockout mice in which Nrf2 is deleted in microglia, astrocytes, and brain endothelial cells, respectively. We discovered that in healthy brains, Nrf2 controls distinct transcriptional profiles in the three brain cell types under study. Surprisingly, after systemic inflammation, microglia, and astrocytes lacking Nrf2 showed reduced inflammatory responses, unlike endothelial cells. Additionally, even without any insult, Nrf2 in microglia supported the expression of pro-inflammatory genes. Our findings reveal a pro-inflammatory role for endogenous Nrf2 in specific brain cell types.

Graphical abstract

Highlights

• •Nrf2 controls distinct transcriptional profiles in microglia, astrocytes, and BECs
• •Nrf2 loss in microglia and astrocytes reduces inflammatory responses to systemic insult
• •Microglial Nrf2 supports basal expression of inflammation-related genes
• •Study reveals unexpected pro-inflammatory role of endogenous Nrf2 in brain glial cells

Biological sciences; Neuroscience; Molecular neuroscience

Introduction

The transcription factor Nrf2 (encoded by the Nfe2l2 gene) is a widely-expressed stress-responsive master regulator of genes involved in important aspects of homeostatic physiology.^1,2,3 Under normal conditions, Nrf2 is targeted for ubiquitin-mediated degradation by Keap1, but in response to stress, this interaction is inhibited, leading to Nrf2 accumulation in the nucleus where it regulates the expression of genes that contain antioxidant response elements (AREs) in their promoter.^1,2 In the brain, Nrf2 is highly expressed in microglia, astrocytes, and brain endothelial cells (BECs), with low expression in neurons.^4,5,6 However, the cell-type specific roles of Nrf2 in the brain were not well-understood, partly due to a historical reliance on global Nrf2 knockout mice to probe the roles of Nrf2.⁷ This knowledge gap is important to fill because cell type-specific roles of Nrf2 are likely to be influenced by the function of that cell type, and from a translational point of view, one ideally needs to know the locus of action of any Nrf2-targeting therapeutic to optimize bioavailability.

Previous studies have highlighted an anti-inflammatory role of Nrf2 in the brain, via possible mechanisms such as elimination of reactive oxygen species (ROS), blocking proinflammatory cytokine transcription, or reducing expression of adhesion molecules.^8,9,10,11 However, most of these studies were based on prior pharmacological activation of Nrf2 using natural or synthetic compounds of Nrf2 activators.¹¹ The role of the endogenous activity of Nrf2 in controlling inflammatory responses in individual brain cell types is not clear. The global Nrf2 knockout mouse has pleiotropic phenotypes in development and maturity,¹² which, in addition to the non-cell autonomous effects between different cell types, makes it difficult to dissect the cell-type specific role of Nrf2 in the brain under inflammatory conditions. In this study, we hypothesized that Nrf2 plays cell type-specific roles both in controlling the basal transcriptome as well as determining those cell types’ responses to inflammatory insults.

Results

Generation of inducible-conditional Nrf2 knockout mice in microglia and astrocytes

To investigate the cell-type specific roles of Nrf2 in the brain, we used Nfe2l2^fl/fl mice which possess loxP sites flanking exon 5 of the Nfe2l2 gene which encodes the DNA binding domain of Nrf2. Previously, we generated an inducible-conditional EC-specific knockout (Nfe2l2^Endo) mouse by crossing Nfe2l2^fl/fl mice with Cdh5^CreER that expresses tamoxifen-inducible CreERT specifically in endothelial cells. In the current study, we crossed Nfe2l2^fl/fl mice with either an inducible astrocyte-specific Cre mouse (Aldh1l1^CreER) or an inducible monocyte-specific Cre mouse line (Cx3cr1^CreER). Following tamoxifen injection, exon 5 of the Nrf2 gene is deleted exclusively either in astrocytes to generate an inducible astrocyte-specific knockout line (Aldh1l1^CreER:Nfe2l2^fl/fl, hereafter Nfe2l2^Astro), or exclusively in microglia to generate an inducible monocyte-specific knockout line (Cx3cr1^CreER:Nfe2l2^fl/fl, hereafter Nfe2l2^Micro). The Nfe2l2^fl/fl nomenclature refers to control mice in the study not crossed onto any Cre lines in which exon 5 of Nfe2l2 remains intact. Tamoxifen injection will cause deletion of exon 5 of Nrf2 in both microglia and blood monocyte-derived macrophages, but within 4 weeks blood monocytes turn over,¹¹ leaving only microglia (half-life approximately 2 years) lacking exon 5 of Nrf2. Sorting of brain cells by FACS followed by qPCR and RNA-seq reads mapping confirmed the purity of microglia, astrocytes, and BECs (Figure S1 and Table S1) and the successful deletion of Nrf2 exon 5 in both microglia in Nfe2l2^Micro mice (Figures 1A and 1C) and astrocytes in Nfe2l2^Astro mice (Figures 1B and 1D) without affecting Nrf2 exon 5 levels in other brain cell types in either mouse line (Figures 1A and 1B).

Figure 1.

Generation of inducible-conditional Nrf2 knockout mice in microglia and astrocytes

(A) RT-qPCR confirms gene expression of exon 5 of Nrf2 is specifically deleted in microglia, but not in astrocytes, oligodendrocytes, or neurons in Nfe2l2^Micro mice. ∗∗: p < 0.01, from left to right p = 0.0014, 0.7955, 0.9989, and 0.9988, n = 4, unpaired t test.

(B) qRT-PCR confirms gene expression of exon 5 o1f Nrf2 is specifically deleted in astrocytes, but not in microglia, oligodendrocytes, or neurons in Nfe2l2^Astro mice. ∗: p < 0.05, from left to right p = 0.049, 0.9890, 0.9875, and 0.9862, n = 4, unpaired t test.

(C) Sashimi plot for Nrf2 exon 2, exon 3, exon 4, and exon 5 (from right to left) in Nfe2l2^fl/fl mice and Nfe2l2^Micro mice. Per-base expression is plotted on y axis of Sashimi plot, genomic coordinates on x axis.

(D) Sashimi plot for Nrf2 exon 2, exon 3, exon 4 and exon 5 (from right to left) in Nfe2l2^fl/fl mice and Nfe2l2^Astro mice. Per-base expression is plotted on y axis of Sashimi plot, genomic coordinates on x axis.

Nrf2 controls distinct transcriptional signatures in microglia, astrocytes, and BECs

Previously, using the Nfe2l2^Endo mice, we found that Nrf2 is required to maintain BEC homeostatic transcriptional signature in the adult mouse brain.¹³ In the current study, we have extended this analysis to microglia from Nfe2l2^Micro mice and astrocytes from Nfe2l2^Astro mice. RNA-seq analysis of differentially regulated genes in Nrf2-deficient microglia from Nfe2l2^Micro mice revealed 24 upregulated and 33 downregulated genes (padj<0.05; Figure 2A, Table S2). NRF2-deficient microglia exhibited a down-regulation of known Nrf2 target genes (Hmox1 and Srxn1, Figure 2A). Functional categorization using the Enrichr Bioinformatic database for GO term enrichment analysis of biological process (BP), molecular function (MF), and KEGG pathway enrichment indicate that Nrf2 controls key transcriptional signatures in microglia (Table 1). For example, Nrf2-deficient microglia have dysregulated immunity and iron ion homeostasis, reduced response to oxidative stress, chemical stress, cytokine stimulus and bacterium, and dysregulated protein folding. Cellular compartment (CC) enrichment analysis revealed reduced expression of secretory granule and phagocytic vesicle membrane. KEGG pathway suggested downregulation of genes involved in infection and neurodegenerative diseases.

Figure 2.

Nrf2 maintains homeostatic transcriptional signatures in microglia and astrocytes

Scatterplot of analysis was generated for genes with average expression >0.1 FPKM across the datasets. Highlighted with red and blue crosses are the genes whose expression are significantly increased or decreased, respectively (DESeq2 P_adj < 0.05, n = 4–6).

(A) The influence of Nrf2 specific KO in microglia on microglial transcriptome in Nfe2l2^Micro mice at basal conditions.

(B) The influence of Nrf2 specific KO in astrocytes on astrocyte transcriptome in Nfe2l2^Astro mice at basal conditions.

Table 1.

Genes downregulated by Nrf2-deficiency in microglia in Nfe2l2^Micro mice under basal condition

	Overlap	p value	Adjusted p value	Odds ratio	Genes
Term_BP
Cellular Response To Oxidative Stress (GO:0034599)	4/117	1.1E−05	2.9E−03	35.16	Srxn1; Hmox1;Hspa1B; Nfe2l2
Multicellular Organismal-Level Iron Ion Homeostasis (GO:0060586)	2/12	9.0E−05	5.1E−03	181.51	Slc11A1;Hmox1
Positive Regulation Of Proteasomal Ubiquitin-Dependent Protein Catabolic Process (GO:0032436)	3/83	1.3E−04	6.5E−03	35.53	Cebpa;Hspa1B;Nfe2l2
Cellular Response To Chemical Stress (GO:0062197)	3/89	1.6E−04	7.1E−03	33.04	Srxn1;Hspa1B;Nfe2l2
Chaperone Cofactor-Dependent Protein Refolding (GO:0051085)	2/27	4.8E−04	1.9E−02	72.55	Dnajb1;Hspa1B
Integrated Stress Response Signaling (GO:0140467)	2/29	5.5E−04	2.0E−02	67.17	Cebpa;Nfe2l2
'De Novo' Post-Translational Protein Folding (GO:0051084)	2/32	6.7E−04	2.2E−02	60.44	Dnajb1;Hspa1B
Response To Hydrogen Peroxide (GO:0042542)	2/46	1.4E−03	3.6E−02	41.18	Hmox1;Nfe2l2
Intracellular Iron Ion Homeostasis (GO:0006879)	2/50	1.6E−03	4.0E−02	37.74	Slc11A1;Hmox1
Defense Response To Gram-negative Bacterium (GO:0050829)	2/75	3.6E−03	5.6E−02	24.79	Slc11A1;Camp
Negative Regulation Of Myeloid Leukocyte Mediated Immunity (GO:0002887)	1/5	6.0E−03	5.6E−02	217.09	Cx3Cr1
Positive Regulation Of I-kappaB Phosphorylation (GO:1903721)	1/5	6.0E−03	5.6E−02	217.09	Cx3Cr1
Positive Regulation Of Antigen Processing And Presentation (GO:0002579)	1/5	6.0E−03	5.6E−02	217.09	Slc11A1
Term_MF
Ubiquitin Protein Ligase Binding (GO:0031625)	3/271	4.03E−03	7.6E−02	10.51	Tubb2A;Hspa1B;Nfe2l2
Heme Binding (GO:0020037)	2/87	4.85E−03	7.6E−02	21.27	Hmox1;Nfe2l2
Gap Junction Channel Activity Involved In Cell Communication By Electrical Coupling (GO:1903763)	1/5	5.99E−03	7.6E−02	217.09	Gjb6
Iron Ion Transmembrane Transporter Activity (GO:0005381)	1/7	8.37E−03	7.6E−02	144.71	Slc11A1
MHC Class I Protein Binding (GO:0042288)	1/13	1.55E−02	7.6E−02	72.33	Tubb2A
Purine Ribonucleoside Triphosphate Binding (GO:0035639)	3/476	1.87E−02	7.6E−02	5.89	Tubb2A;Tuba4A;Hspa1B
Chemokine Receptor Activity (GO:0004950)	1/19	2.26E−02	7.6E−02	48.21	Cx3cr1
Metal Ion Transmembrane Transporter Activity (GO:0046873)	1/29	3.42E−02	7.9E−02	30.98	Slc11A1
Term_CC
Polymeric Cytoskeletal Fiber (GO:0099513)	4/265	2.6E−04	1.6E−02	15.11	Tubb2A;Kif3A;Gjb6;Tuba4A
Cytoskeleton (GO:0005856)	4/599	5.3E−03	4.1E−02	6.51	Tubb2A;Kif3A;Dynll1;Tuba4A
Mitotic Spindle (GO:0072686)	2/143	1.3E−02	7.1E−02	12.79	Tubb2A;Dynll1
Secretory Granule Membrane (GO:0030667)	2/279	4.4E−02	1.3E−01	6.47	Slc11A1;Dynll1
Phagocytic Vesicle Membrane (GO:0030670)	1/45	5.3E−02	1.4E−01	19.70	Slc11A1
KEGG
Protein processing in endoplasmic reticulum	3/171	1.1E−03	3.4E−02	16.84	Dnajb1;Hspa1B;Nfe2l2
Salmonella infection	3/249	3.2E−03	3.4E−02	11.46	Tubb2A;Dynll1;Tuba4A
Parkinson disease	3/249	3.2E−03	3.4E−02	11.46	Tubb2A;Tuba4A;Nfe2l2
Pathways in cancer	4/531	3.4E−03	3.4E−02	7.38	Cebpa;Hmox1;Pim2;Nfe2l2
Phagosome	2/152	1.4E−02	7.0E−02	12.02	Tubb2A;Tuba4A
Pathogenic Escherichia coli infection	2/197	2.3E−02	9.4E−02	9.22	Tubb2A;Tuba4A
Ferroptosis	1/41	4.8E−02	1.4E−01	21.67	Hmox1

Analysis of differentially regulated genes in Nrf2-deficient astrocytes in Nfe2l2^Astro mice revealed 30 upregulated and 29 downregulated genes (p < 0.05; Figure 2B and Table S2). As expected, Nrf2-deficient astrocytes in Nfe2l2^Astro mice exhibited a down-regulation of several known Nrf2 target genes (e.g., Gstm1, Fth1, Ttl1, Gclc, and Txnrd1; Figure 2B). Functional categorization using GO term enrichment analysis of BP, MF, and KEGG pathways suggest that Nrf2-deficient astrocytes have reduced expression of genes involved in glutathione metabolism, vitamin B6 metabolism, nicotinate and nicotinamide metabolism, and tyrosine metabolism. Nrf2-deficient astrocytes exhibited a reduced expression of genes involved in the response to oxidative stress, ferroptosis, and axon regeneration (Table 2). CC enrichment analysis revealed decreased expression of genes in lysosome, peroxisome, and perisynaptic extracellular matrix regions (Table 2).

Table 2.

Genes downregulated by Nrf2-deficiency in astrocytes in Nfe2l2^Astro mice under basal condition

	Overlap	p value	Adjusted p value	Odds ratio	Genes
Go_Term_BP
Cellular Response To Reactive Oxygen Species (GO:0034614)	2/73	5.0E-03	9.7E-02	20.76	Mpv17L;Nfe2l2
Glutathione Biosynthetic Process (GO:0006750)	1/6	8.7E-03	9.7E-02	142.61	Gclc
Intracellular Sequestering Of Iron Ion (GO:0006880)	1/6	8.7E-03	9.7E-02	142.61	Fth1
Glycine Metabolic Process (GO:0006544)	1/7	1.0E-02	9.7E-02	118.84	Rida
Hepoxilin Metabolic Process (GO:0051121)	1/7	1.0E-02	9.7E-02	118.84	Gstm1
Go_Term_MF
Glutathione Transferase Activity (GO:0004364)	2/28	7.5E-04	4.0E-02	56.82	Gstm1;Mgst1
Flavin Adenine Dinucleotide Binding (GO:0050660)	2/55	2.9E-03	5.4E-02	27.84	Txnrd1;Aox1
Iron Ion Binding (GO:0005506)	2/56	3.0E-03	5.4E-02	27.32	Fth1;Aox1
Ferroxidase Activity (GO:0004322)	1/5	7.2E-03	6.5E-02	178.28	Fth1
Acid-Amino Acid Ligase Activity (GO:0016881)	1/6	8.7E-03	6.7E-02	142.61	Gclc
Go_Term_CC
Autolysosome (GO:0044754)	1/9	1.3E-02	1.1E-01	89.12	Fth1
Peroxisome (GO:0005777)	2/129	1.5E-02	1.1E-01	11.57	Rida;Mgst1
COPI Vesicle Coat (GO:0030126)	1/12	1.7E-02	1.1E-01	64.81	Copg2
KEGG
Glutathione metabolism	3/57	7.6E-05	1.6E-03	42.56	Gclc;Gstm1;Mgst1
Ferroptosis	1/41	1.6E-03	1.1E-02	37.86	Gclc;Fth1
Metabolism of xenobiotics by cytochrome P450	2/76	5.4E-03	3.0E-02	19.92	Gstm1;Mgst1
Vitamin B6 metabolism	1/6	8.7E-03	3.6E-02	142.61	Aox1

Previously, using an inducible conditional-endothelial Nrf2 KO mice (Nfe2l2^Endo), we found that basal Nrf2 activity controls key transcriptional signatures in BECs.¹³ In this study, we compared the differentially regulated genes under basal conditions in microglia in Nfe2l2^Micro with those in astrocytes in Nfe2l2^Astro, and in BECs in Nfe2l2^Endo (Figures 3A and 3B), we found only a few genes common between these three cell types. These data suggest that Nrf2 controls distinct transcriptional programmes in these three cell types under basal conditions.

Figure 3.

Nrf2 controls distinct transcriptional signatures in microglia, astrocytes, and BECs

(A) The significantly downregulated genes due to Nrf2 KO in microglia in Nfe2l2^Micro mice, in astrocytes in Nfe2l2^Astro mice, and in BECs in Nfe2l2^Astro mice, compared to the respective littermate controls, were compared using Venn diagram.

(B) The significantly upregulated genes due to Nrf2 KO in microglia in Nfe2l2^Micro mice, in astrocytes in Nfe2l2^Astro mice, and in BECs in Nfe2l2^Endo mice, compared to the respective littermate controls, were compared using Venn diagram.

Nrf2 controls inflammatory responses in microglia, astrocytes, and BECs following systemic inflammation

We next wanted to determine the impact of Nrf2 deficiency on cellular inflammatory responses. It is well-established that systemic inflammation can lead to blood brain barrier (BBB) compromise and neuroinflammation, so we employed a model of systemic inflammation to determine how Nrf2 status affects different brain cells’ response to inflammatory insults. We induced systemic inflammation in mice by intraperitoneal injection of bacterial endotoxin LPS in Nfe2l2^fl/fl mice, chosen as a “conditional-ready” control to compare with conditional knockout mice used in the study. Following LPS exposure in Nfe2l2^fl/fl mice, microglia, astrocytes and BECs display widespread transcriptional changes (Table S3), affecting multiple signaling pathways in each cell type (Table S4). When we took the activated pathways in microglia to look at how they were altered in astrocytes and BECs following LPS exposure, we found that majority of these pathways were also activated in astrocytes in a higher level overall. However, in BECs, several of them were either not altered or inhibited (Figures 4A and 4B). For example, canonical inflammatory pathways including Interferon alpha/beta signaling, TNFR2 non-canonical NF-kB pathway, cell surface interactions at the vascular wall, necroptosis signaling pathway, and interleukin-4 and interleukin-13 signaling were all activated in three cell types, whereas KEAP1-NFE2L2 pathway and role of hypercytokinemia/hyperchemokinemia in the Pathogenesis of Influenza were hardly altered and Inhibition of ARE-mediated mRNA degradation pathway, beta-catenin independent WNT signaling, oxidative stress induced senescence, leukocyte extravasation signaling were inhibited in BECs (Figure 4A). Thus, microglia and astrocytes respond to LPS in a more similar manner, whereas BECs behave more differently.

Figure 4.

LPS elicits similar but different transcriptional responses in microglia, astrocytes, and BECs

The Z scores (Table S2) of LPS-activated pathways in microglia were compared with those in astrocytes and BECs.

(A) Heatmap showing the Z score of each activated pathway in microglia and the Z scores of the same pathways in astrocytes and BECs, respectively.

(B) Violin plot showing the z-scores of the activated pathways in microglia and the z-scores of the same pathways in astrocytes and BECs, respectively.

We then wanted to know how Nrf2 deficiency influenced the transcriptome of microglia, astrocytes and BECs under these pro-inflammatory conditions. RNA-seq analysis of Nrf2-deficient microglia in Nfe2l2^Micro mice revealed 22 upregulated and 100 downregulated genes compared to Nfe2l2^fl/fl microglia in mice treated with LPS insults (p < 0.05; Figure 5A, Table S2). Functional categorization using KEGG pathway and GO term enrichment analysis of BP (Table 3) suggests that pathways involved in leukocyte transendothelial migration, response to cytokine and response to tumor necrosis factor were reduced in microglia from Nfe2l2^Micro mice. Analysis of Nrf2-deficient astrocytes in Nfe2l2^Astro mice revealed 103 upregulated and 128 downregulated genes compared to astrocytes from Nfe2l2^fl/fl mice (p < 0.05; Figure 5B). Functional categorization using KEGG pathway and GO term enrichment analysis of BP (Table 4) suggests that inflammatory pathways including MAPK signaling pathway, JAK-STAT signaling pathway, and p53 signaling pathway were downregulated due to Nrf2-deficiency in astrocytes following systemic LPS insults. In addition, Wnt signaling pathway, HIF-1 signaling pathway, and ubiquitin mediated proteolysis were also downregulated, suggesting Nrf2-deficiency astrocytes reduced responses to inflammatory stimulus. In contrast to Nrf2-deficient microglia and astrocytes, Nrf2-deficient BECs in Nfe2l2^Endo mice had modest number of differentially regulated genes in response to peripheral LPS compared to Nfe2l2^fl/fl mice (Figure 5C). Only 9 genes were differentially expressed in Nfe2l2^Endo BECs compared to Nfe2l2^fl/fl following LPS treatment. Thus, endogenous Nrf2 activity has a stronger influence on inflammatory responses in microglia and astrocytes than in BECs.

Figure 5.

Nrf2 controls inflammatory responses in microglia, astrocytes, and BECs under systemic inflammation

Mice were given i.p. injection of either saline or LPS for 24 h. The brain cells were isolated and then sorted with FACS for microglia, astrocytes and BECs, RNA extracted and RNA-seq performed. Scatterplot was generated for genes with average expression >0.1 FPKM across the datasets. Highlighted with red and blue crosses are the genes whose expression are significantly increased or decreased, respectively (DESeq2 P_adj < 0.05, n = 6–8). “n” refers here and throughout as an independent replicate (i.e., a mouse).

(A) Scatterplot of RNA-seq analysis showing the microglial transcriptome modified by Nrf2 KO in Nfe2l2^Micro mice compared to Nfe2l2^fl/fl mice in response to peripheral LPS insults.

(B) Scatterplot of RNA-seq analysis showing the astrocyte transcriptome modified by Nrf2 KO in astrocytes in Nfe2l2^Astro mice compared to Nfe2l2^fl/fl mice in response to peripheral LPS insults.

(C) Scatterplot of RNA-seq analysis showing the BEC transcriptome modified by Nrf2 KO in BECs in Nfe2l2^Endo mice compared to Nfe2l2^fl/fl mice in response to peripheral LPS insults.

Table 3.

Genes downregulated by Nrf2-deficiency in microglia in Nfe2l2^Micro mice under inflammatory condition

	Overlap	p value	Adjusted p value	Odds ratio	Genes
KEGG
Cell adhesion molecules	6/148	2.3E-04	3.4E-02	7.58	Cdh5;Cldn5;Esam;Nrcam;Ncam2;Sele
Fatty acid elongation	2/27	1.1E-02	3.4E-02	13.94	Echs1;Acot1
Butanoate metabolism	2/28	1.1E-02	3.4E-02	13.40	Echs1;Abat
beta-Alanine metabolism	2/30	1.3E-02	3.4E-02	12.44	Echs1;Abat
ECM-receptor interaction	3/88	1.5E-02	7.0E-02	6.18	Col4A2;Col4A1;Spp1
Leukocyte transendothelial migration	3/114	2.9E-02	9.4E-02	4.73	Cdh5;Cldn5;Esam
Pyruvate metabolism	2/47	3.0E-02	1.4E-01	7.73	Pkm;Akr1A1
Term_BP
Response To Cytokine (GO:0034097)	8/125	6.7E-07	3.9E-04	12.51	Csf3;Sphk1;Mx1;Timp3;Ccr7;Timp1;Sele
Response To Tumor Necrosis Factor (GO:0034612)	5/108	4.2E-04	2.4E-02	8.65	Dhx9;Sphk1;Gbp2;Sele;Gbp3
Response To Lipid (GO:0033993)	5/110	4.5E-04	2.4E-02	8.49	Acer2;Spp1;Sox9;Trim16;Sele
Positive Regulation Of Phosphatidylinositol 3-Kinase Signaling (GO:0014068)	4/89	6.9E-04	3.1E-02	10.89	Csf3;Nedd4;Sox9;Dcn

Table 4.

Genes downregulated by Nrf2-deficiency in astrocytes in Nfe2l2^Astro mice under inflammatory conditions

	Overlap	p value	Adjusted p value	Odds ratio	Genes
KEGG
MAPK signaling pathway	7/294	1.82E-03	4.2E-02	4.31	Irak1;Myc;Tgfa;Igf1;Fos;Dusp8;Mapk8Ip1
JAK-STAT signaling pathway	5/162	2.75E-03	4.3E-02	5.56	Socs2;Cdkn1A;Socs1;Myc;Aox1
p53 signaling pathway	3/73	9.23E-03	8.6E-02	7.38	Cdkn1A;Igf1;Gtse1
Wnt signaling pathway	4/166	1.69E-02	1.2E-01	4.27	Daam2;Myc;Csnk1E;Ppard
HIF-1 signaling pathway	3/109	2.67E-02	1.5E-01	4.87	Cdkn1A;Hmox1;Igf1
Ubiquitin mediated proteolysis	3/140	5.01E-02	2.1E-01	3.76	Socs1;Ubc;Ube2Ql1
Term_BP
Cellular Response To Hypoxia (GO:0071456)	4/72	8.7E-04	7.1E-02	10.22	Pink1;Myc;Hmox1;Ndrg1
Positive Regulation Of Carbohydrate Metabolic Process (GO:0045913)	3/34	1.0E-03	7.1E-02	16.70	App;Irs2;Igf1
Negative Regulation Of JNK Cascade (GO:0046329)	3/34	1.0E-03	7.1E-02	16.70	Per1;Pink1;Mapk8Ip1
Negative Regulation Of Stress-Activated MAPK Cascade (GO:0032873)	3/37	1.3E-03	7.1E-02	15.23	Per1;Pink1;Myc
Regulation Of Chemokine Production (GO:0032642)	3/41	1.8E-03	7.7E-02	13.62	App;Egr1;Hmox1
MyD88-dependent Toll-Like Receptor Signaling Pathway (GO:0002755)	2/12	2.2E-03	8.1E-02	34.26	Irak1;Irak2
Apoptotic Process (GO:0006915)	6/228	2.3E-03	8.4E-02	4.74	App;Csrnp1;Myc;Mal;Gzmb;Ppard

After a systemic insult, Nrf2 supports the proinflammatory responses in microglia and astrocytes, but not in BECs

We next wanted to determine whether the groups of genes up or downregulated following LPS treatment show any pattern of perturbation due to Nrf2 deficiency. We firstly assessed whether Nrf2 was able to modulate microglial overall responses to LPS in Nfe2l2^Micro mice. We took microglial genes up- and downregulated by LPS and studied the impact of Nrf2-deficiency in microglia. We found that the gene set induced by LPS in microglia was overall repressed as the result of Nrf2-deficiency in microglia (Figures 6A and 6G), and the gene set repressed by LPS in microglia was overall elevated as the result of Nrf2-deficiency in microglia (Figures 6B and 6H). This suggests that Nrf2-deficiency in microglia dampened down microglial proinflammatory responses to peripheral LPS insults. Using the same strategy, we assessed the influence of basal Nrf2 activity on modulating astrocyte and BEC responses to LPS in Nfe2l2^Astro mice and Nfe2l2^Endo mice, respectively. We found that similar to Nrf2-deficient microglia, the gene set induced by LPS in astrocytes was overall repressed as the result of Nrf2-deficiency microglia (Figures 6B and 6G), and the gene set repressed by LPS in astrocytes was overall elevated as the result of Nrf2-deficiency in astrocytes (Figures 6C and 6H). This suggests that like microglia, Nrf2-deficiency in astrocytes reduced astrocytes’ overall proinflammatory responses to peripheral LPS insults. In contrast, the gene sets induced or repressed by LPS in BECs were not overall changed as the result of Nrf2-deficiency in BECs in Nfe2l2^Endo mice (Figures 6E, 6G, 6F, and 6H). Thus, basal Nrf2 activity in microglia and astrocytes overall enhances the responses of these cell type to peripheral inflammation, contrary to the classical view of Nrf2 being solely anti-inflammatory.

Figure 6.

Nrf2 promotes proinflammatory responses in microglia and astrocytes, but not in BECs under systemic inflammation

(A) The influence of Nrf2 KO in microglia on the expression of microglial LPS-induced gene set under inflammatory conditions. For microglial LPS-induced genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in microglia in response to peripheral LPS insults were plotted.

(B) The influence of Nrf2 KO in microglia on the expression of microglial LPS-repressed gene set under inflammatory conditions. For microglial LPS-repressed genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in microglia in response to peripheral LPS insults were plotted.

(C) The influence of Nrf2 KO in astrocytes on the expression of astrocyte LPS-induced gene set under inflammatory conditions. For astrocyte LPS-induced genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in astrocytes in response to peripheral LPS insults were plotted.

(D) The influence of Nrf2 KO in astrocytes on the expression of astrocyte LPS-repressed gene set under inflammatory conditions. For astrocyte LPS-repressed genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in astrocytes in response to peripheral LPS insults were plotted.

(E) The influence of Nrf2 KO in BECs on the expression of BEC LPS-induced gene set under inflammatory conditions. For BEC LPS-induced genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in BECs in response to peripheral LPS insults were plotted.

(F) The influence of Nrf2 KO in BECs on the expression of BEC LPS-repressed gene set under inflammatory conditions. For BEC LPS-repressed genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in BECs in response to peripheral LPS insults were plotted.

(G) From left to right, ∗∗∗p = 4.9E-12, F (1,3859) = 47.81 relates to main effect of Nrf2-KO on the expression of microglial LPS-induced gene set under inflammatory conditions, two-way ANOVA; ∗∗∗p = 1.5E-05, F (1,4761) = 18,76 relates to main effect of Nrf2-KO on the expression of astrocyte LPS-induced gene set under inflammatory conditions, two-way ANOVA; p = 0.5382, F (1,3594) = 0.3789 relates to main effect of Nrf2-KO on the expression of BEC LPS-induced gene set under inflammatory conditions, two-way ANOVA.

(H) From left to right, ∗∗∗p = 1.1E-31, F (1,6993) = 138.6 relates to main effect of Nrf2-KO on the expression of microglial LPS-repressed gene set under inflammatory conditions, two-way ANOVA; ∗∗∗p = 6.1E-18, F (1,3357) = 75.34 relates to main effect of Nrf2-KO on the expression of astrocyte LPS-repressed gene set under inflammatory conditions, two-way ANOVA; p = 0.5382 F (1,5570) = 0.9470 relates to main effect of Nrf2-KO on the expression of BEC LPS-repressed gene set under inflammatory conditions, two-way ANOVA.

Nrf2 mediates proinflammatory gene expression in microglia, not in astrocytes and BECs at basal condition

Finally, we wanted to know whether the impact of Nrf2-deficiency controls the previously described LPS response genes under basal, non-inflammatory conditions. We found that the gene set induced by LPS in microglia was overall repressed as the result of Nrf2-deficiency in microglia under basal conditions (Figures 7A and 7G), and the gene set repressed by LPS in microglia was overall elevated as the result of Nrf2-deficiency in microglia under basal conditions (Figures 7B and 7H). In contrast, when we took the gene sets of either up- or downregulated by peripheral LPS insults in astrocytes or BECs and studied the impact of Nrf2-deficiency on expression of these genes in these two cell types under basal conditions, we found that Nrf2-deficiency had no impact on the expression of these gene sets in either astrocytes (Figures 7C, 7G, 7D, and 7H) or BECs (Figures 7E–7H). Overall these data suggest that Nrf2 controls the basal expression of inflammatory response genes in a cell-type specific manner.

Figure 7.

Nrf2 mediates proinflammatory gene expression in microglia, but not in astrocytes and BECs at basal conditions

(A) The influence of Nrf2 KO in microglia on the expression of microglial LPS-induced gene set at basal conditions. For microglial LPS-induced genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in microglia at basal conditions were plotted.

(B) The influence of Nrf2 KO in microglia on the expression of microglial LPS-repressed gene set at basal conditions. For microglial LPS-repressed genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in microglia at basal conditions were plotted.

(C) The influence of Nrf2 KO in astrocytes on the expression of astrocyte LPS-induced gene set at basal conditions. For astrocyte LPS-induced genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in astrocytes at basal conditions were plotted.

(D) The influence of Nrf2 KO in astrocytes on the expression of astrocyte LPS-repressed gene set at basal conditions. For astrocyte LPS-repressed genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in astrocytes at basal conditions were plotted.

(E) The influence of Nrf2 KO in BECs on the expression of BEC LPS-induced gene set at basal conditions. For BEC LPS-induced genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in BECs at basal conditions were plotted.

(F) The influence of Nrf2 KO in BECs on the expression of BEC LPS-repressed gene set at basal conditions. For BEC LPS-repressed genes (x axis, FPKM>1, DESeqP_adj < 0.05, Log₂FC > 1) in WT mice, the Log₂FC (y axis) of each gene (x axis) due to Nrf2 KO in BECs at basal conditions were plotted.

(G) From left to right, ∗∗∗p = 5.0E-12, F (1,7700) = 47.82 relates to main effect of Nrf2-KO on the expression of microglial LPS-induced gene set at basal conditions, two-way ANOVA; p = 0.3441, F (1,4761) = 0.8952 relates to main effect of Nrf2-KO on the expression of astrocyte LPS-induced gene set at basal conditions, two-way ANOVA; p = 0.4823, F (1,5990) = 0.4936 relates to main effect of Nrf2-KO on the expression of BEC LPS-induced gene set at basal conditions, two-way ANOVA.

(H) From left to right, ∗∗∗p = 6.0E-09, F (1,3885) = 34.00 relates to main effect of Nrf2-KO on the expression of microglial LPS-repressed gene set at basal conditions, two-way ANOVA; p = 0.3791, F (1,3357) = 4.311 relates to main effect of Nrf2-KO on the expression of astrocyte LPS-repressed gene set at basal conditions, two-way ANOVA; p = 0.4693 F (1,6126) = 0.5236 relates to main effect of Nrf2-KO on the expression of BEC LPS-repressed gene set at basal conditions, two-way ANOVA.

Discussion

Previous studies have pointed to multiple cytoprotective functions of Nrf2 in the brain,¹ however, cell^14,15-type specific role of Nrf2 in adult brain was not clear. Using cell-type specific deletions of Nrf2 in adult mouse brain, our study found that Nrf2 controls expression of slightly overlapping, but mainly distinct, set of genes in microglia, astrocytes and BECs, suggesting that Nrf2 controls distinct transcriptional programmes in different brain cell types. This may be at first glance surprising, particularly for genes positively regulated by Nrf2, since these genes often contain ARE elements that in theory should be responsive to Nrf2 in all cell types. However, not all ARE-containing genes are Nrf2 responsive in all cell types since Nrf2 deletion has different effects on gene expression in different cell types. The few Nrf2 target genes affected by Nrf2 deletion could be due to the low activity of Nrf2 under basal conditions in vivo, especially for microglia which are said to exist in a homeostatic state at rest, and it is likely that the un-stressed cellular conditions mean efficient degradation of Nrf2 via Keap1. As such, deletion of Nrf2 in resting-state microglia does not impact strongly on gene expression due to the already extremely low level of basal Nrf2 activity. Some Nrf2 target genes may be epigenetically silenced in certain cell types, making their promoters inaccessible to Nrf2. Moreover, since Nrf2 target genes can be controlled by other transcription factors (such as AP-1)^14,15 and so the impact of Nrf2 deficiency may depend on the expression of other transcription factors in that cell type.

Global Nrf2 KO mice are susceptible to various pathogen infections,^16,17,18,19 suggesting a protective role of Nrf2 in inflammatory and infectious diseases. However, other studies have suggested a pathogenic role of Nrf2 in certain disorders, including metabolic disorders associated with chronic inflammation, with certain aspects of Nrf2 function (e.g., in lipid metabolism) outweighing the effects of other functions (e.g., antioxidant production).^20,21,22 Our study showed that Nrf2-deficient microglia had reduced expression of genes involved in responses to cytokine stimulation, and Nrf2-deficient astrocytes had reduced expression of genes involved in classical pathways of inflammation including MAPK signaling pathway, JAK-STAT signaling pathway, and p53 signaling pathway. In line with this functional categorization, comparisons of our RNA-seq data show that Nrf2-deficiency in microglia and in astrocytes caused a decrease in overall inflammatory responses to a peripheral LPS insult. Taken together, our study suggests that endogenous Nrf2 activity partly mediate responses in microglia and astrocytes under inflammatory conditions.

It is therefore possible that Nrf2 plays a pro-inflammatory role under certain conditions, although the precise mechanism behind this warrants further investigation. Indeed, a certain level of Nrf2 activity is beneficial and essential in supporting inflammatory responses and controlling pathogenic processes in inflammatory and infectious diseases. Moreover, the absence of Nrf2 may lead to an insufficient inflammatory response, leading to susceptibility to pathogens in global Nrf2 KO mice, whereas in chronic inflammatory diseases, such as metabolic disorders, a sustained high level of Nrf2 becomes pathological. Therefore, it is possible that Nrf2’s role in inflammatory responses, and the consequences of these responses, is highly sensitive to both the cell type and the nature of the inflammatory insult.

In addition to mediating an inflammatory response, our RNAs-seq data also finds that pathways in response to oxidative stress, were downregulated in microglia and astrocytes at basal conditions due to the absence of Nrf2. Increased oxidative stress contributes to several hallmarks of aging²³ and Nrf2 activity declines with aging and in neurodegenerative diseases.^24,25 Cyano-3,12-dioxooleana-1,9-Dien-28-Oic acid (CDDO) and analogues are by far the most potent small molecule of Nrf2 activators with anti-inflammatory and antioxidative properties.²⁶ Over the last two decades, these triterpenoid compounds have exhibited a broad range of applications in animal disease models and clinical trials.²⁷ Compared to other triterpenoid compounds, RTA-404 has greater capacity to cross the BBB²⁸ and has shown neuroprotective effects in preclinical models of neurodegenerative diseases, including autoimmune encephalomyelitis,²⁹ Huntington’s disease,³⁰ and amyotrophic lateral sclerosis.³¹ Our group has recently found that RTA-404 acts on endothelial cells in vivo to prevent immune cell infiltration into the brain.¹³ RTA-408 was recently approved by the FDA for treating Friedreich’s Ataxia, a genetic neurodegenerative disease associated with oxidative stress.³² This provides proof-of-concept that Nrf2 can be safely activated by the triterpenoid compounds in humans and restoring or boosting Nrf2 activity via pharmacological Nrf2 activator could represent a promising therapeutic target in other neurodegenerative diseases, even when Nrf2 deficiency is not the primary cause of the disease.

IPA analysis shows that KEAP1-NFE2L2 pathway was strongly activated in microglia and astrocytes, however, hardly altered in BECs following LPS stimulation in Nfe2l2^fl/fl mice. Compared to Nfe2l2^Micro and Nfe2l2^Astro mouse lines, fewer genes are differentially altered in BECs in Nfe2l2^Endo mice following peripheral LPS stimulations. It suggests that under inflammatory conditions, signaling within endothelial cells does not activate Nrf2 substantially. ROS generation is a key inducer of Nrf2-dependent gene expression, and ROS production can occur in any cell type including microglia and endothelial cells. However, microglia are far more sensitive to LPS than endothelial cells, therefore LPS-induced ROS generation via NADPH/NOX2^33,34 leading to Nrf2 activation may only take place in microglia and not endothelial cells. However, the effects might be different in a model of cardiovascular disease or other challenges with a higher component of oxidative stress (e.g., hypoperfusion), which is worthy of further investigation. Indeed, our lab is currently investigating the effect of hypoperfusion on BECs and downstream effectors using the conditional Nrf2 endothelial KO mice.

It has been suggested that in certain situations, Cre expression can influence cellular phenotypes.^35,36,37 The three CreER lines, Aldh1l1^CreER,³⁸ Cx3cr1^CreER(Jung),³⁹ and Cdh5^CreER40 used in our current study have been well-characterized and used extensively. The specificity, efficiency and reliability of these CreER lines, especially Cx3cr1^CreER(Jung) line⁴¹ have been well-studied, with no Cre toxicity or off-target effects reported. One recent study which performed RNA-seq analysis to look at the effects of Cre activation on the microglial transcriptome in the Cx3cr1^CreER(Jung) mouse found negligible changes in gene expression.⁴² Future studies are necessary to determine whether Cre alone influences the transcriptional state of microglia, astrocytes, and endothelial cells at baseline and in inflammatory states.

To conclude, further studies are required to fully understand the mechanism by which Nrf2 supports inflammatory responses in a cell type-specific manner, enabling any therapeutic strategy to target the right cell at the right time to combat disorders of neuroinflammation and neurodegeneration.

Limitations of the study

One limitation of this study is that our genetic strategy to delete Nrf2 in Nfe2l2^Endo mice results in knockout across all endothelial cells as opposed to just BECs, as discussed in our recent study.¹³ Another potential limitation is that the Nfe2l2^fl/fl mice used to generate conditional Nrf2 knockouts target only exon 5 of the Nrf2 gene, which encodes the DNA-binding and transcriptional activation domains of the Nrf2 protein.⁴³ To our knowledge all mouse models targeting the Nfe2l2 gene (conditional and constitutive) use this approach, which leaves other exons intact and potentially capable of expressing a truncated N-terminal portion of the protein. Although this truncated protein will have no direct transcriptional regulation capacity, it could theoretically exert other effects through interaction with Keap1 via it’s N-terminus. Future studies will be needed to determine the existence or otherwise of Nrf2 N-terminal domains in these conditional knockout models. If present, alternative approaches, such as generating a new Nfe2l2^fl/flmouse line or employing a cell type-specific AAV-CRISPR-Cas9 system, may be required to achieve complete Nrf2 deletion.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jing Qiu (jing.qiu@ed.ac.uk).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All RNA-seq data that support the findings of this study is available at the European Bioinformatics Institute (ArrayExpress: E-MTAB-14572). All other data are available from the lead contact upon reasonable request.

Acknowledgments

We would like to thank the flow cytometry facility, led by Dr. Fiona Rossi, at the Center for regenerative medicine at the University of Edinburgh for their help with the flow cytometry work. The research leading to these results received funding from Ann Rowling Regenerative Neurology Clinic, United Kingdom; British Heart Foundation, United Kindgham; UK Dementia Research Institute, United Kingdom.

Author contributions

J.Q. conceived the study, performed the experiments, analyzed the data and wrote the manuscript. O.D. and X.H. analyzed the data.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Tamoxifen (TAM)	Merck Life Sciences UK	Cat. #T5648
Corn oil	Merck Life Sciences UK	Cat. #C8267
LPS	InvivoGen	Cat. # tirl-3pelps
Bovine serum albumin (BSA)	Merck Life Sciences UK	Cat. #A9647
RNAprotect Cell Reagent	Qiagen	Cat. #76526
Deposited data
RNA-seq datasets	European Bioinformatics Institute-ArrayExpress	E-MTAB-14572
Experimental models: Organisms/strains
Nrf2flox mouse	The Jackson Laboratory	Cat. #025433
CDH5CreERT2 mouse	N/A	N/A
Cx3cr1CreER(Jung) mouse	The Jackson Laboratory	Cat. #020940
Aldh1l1CreER mouse	The Jackson Laboratory	Cat. #029655
Oligonucleotides
Gapdh Fwd primer 5′-GGGTGTGAACCACGAGAAAT-3′	Merck Life Sciences UK	N/A
Gapdh Rev primer 5′-CCTTCCACAATGCCAAAGTT-3′	Merck Life Sciences UK	N/A
Nefe212 exon5 Fwd primer 5′-TCCATTTACGGAGACCCACC-3′	Merck Life Sciences UK	N/A
Nefe212 exon5 Rev primer 5′-GGATTCACGCATAGGAGCAC-3′	Merck Life Sciences UK	N/A
Software and algorithms
Prism	Graphpad	https://www.graphpad.com/scientific-software/prism
FCS Express	De Novo Software	https://denovosoftware.com
Ingenuity Pathway Analysis	Qiagen	N/A

Experimental model and study participant details

Mice

The animals involved in this study were all on C57BL/6 background. The Nfe2l2^fl/fl mouse line (Mouse Strain Number – 025433), Cx3cr1^{CreER(Jung)39} mouse line (Mouse Strain Number – 020940) and Aldh1l1^CreER38 mouse line (Mouse Strain Number – 029655) were imported from Jackson Laboratories. The Nfe2l2^fl/fl mice carrying loxP sites flanking exon 5 of the Nfe2l2 gene were crossed to the Cx3cr1^CreER(Jung) mouse and Aldh1l1^CreER mouse to generate Cx3cr1^CreER(Jung): Nfe2l2^fl/fl mice (hereafter Nfe2l2^Micr) and Aldh1l1^CreER2: Nfe2l2^fl/fl mice (hereafter Nfe2l2^Astro) respectively. The Nfe2l2^fl/fl nomenclature refers to the controls in the study not crossed onto any Cre lines in which exon 5 of Nfe2l2 remains intact. Male mice aged 12–18 weeks have been used throughout the study. All procedures described were performed at the University of Edinburgh in compliance with the UK Animals (Scientific Procedures) Act 1986 and University of Edinburgh regulations and carried out under project license numbers P2262369. Mice were group-housed in environmentally enriched cages within humidity and temperature-controlled rooms, with a 12-h light dark cycle with free access to food and water. Mouse genotypes were determined using real-time PCR with transgene specific probes (Transnetyx, Cordova, TN) unless otherwise stated.

Method details

Tamoxifen and LPS treatments

For induction of Cre recombinase activity, 6-8-week-old Cdh5^CreER: Nfe2l2^fl/fl mice were given orally with 8 mg tamoxifen (TAM, T5648, Sigma-Aldrich, UK) solved in corn oil (C8267, Sigma-Aldrich, UK) at 4 time points with 24hrs apart. For all experiments, littermates carrying the respective loxP-flanked alles but lacking expression of Cre recombinase (+/+ TAM) were used as controls. Male mice have been used throughout the study. The Cx3cr1^CreER: Nfe2l2^fl/fl mice, Aldh1l1^CreER: Nfe2l2^fl/fl mice, and the respective control mice were injected peritoneally with 3 mg/kg of LPS (tlrl-3pelps, Invivogen, US) for 24 h.

Isolation of brain cells, cell sorting, and RNA extraction

Single brain cells were isolated using Adult Brain Dissociation kit (Miltenyie, 130-107-677, Germany) as per the manufacturer’s instructions. The single brain cell suspension was then incubated with antibodies against CD31 (Biolegend, 102523, UK) CD11b (Biolegend, 101205, UK), ACSA2 (Miltenyie, 130-116-245, Germany), O4 (Miltenyie, 130-117-357, Germany), CD45 (Biolegend, 103125, Germany) for 15 min in ice, washed with PBS and then immediately FACS sorted into RNAprotect Cell Reagent (Qiagen, 76526, UK) under the gates of CD31⁺CD45⁻ACSA2^- for BECs, LY6G⁻CD11b⁺CD45^low for microglia, ACSA⁺O4^- for astrocytes and O4⁺ for oligodendrocytes. RNA extraction was carried out using the RNeasy Plus Micro Kit (Qiagen) as per the manufacturer’s instructions.

Quantitative RT-PCR

cDNA was generated using the Transcriptor First Strand cDNA Synthesis Kit (Roche). Seven microlitres of RNA was added to the RT and buffer mixture prepared with random hexamers and oligoDT primers as per kit instructions, and qRT-PCR carried out with the following program: 10 min at 25 °C, 30 min at 55 °C and 5 min at 85 °C. qPCRs were run on a Stratagene Mx3000P QPCR System (Agilent Technologies) using SYBR Green MasterRox (Roche) with 6 ng of cDNA per well of a 96-well plate, using the following program: 10 min at 95 °C, 40 cycles of 30 s at 95 °C, 40 s at 60 °C and 30 s at 72 °C, with a subsequent cycle of 1 min at 95 °C and 30 s at 55 °C ramping up to 95 °C over 30 s (to measure the dissociation curve).

RNA-seq and analysis

For RNA-seq analysis, libraries were prepared by Edinburgh Genomics using the Illumina TruSeq stranded mRNA-seq kit, according to the manufacturer’s protocol (Illumina). The libraries were pooled and sequenced to 75 base paired-end on an Illumina NovaSeqTM 6000 to a depth of approximately 50 million paired-end reads per sample. For read mapping and feature counting, genome sequences and gene annotations were downloaded from Ensembl version 94. Differential expression (DGE) analysis on datasets was performed using DESeq2 (R package version 1.18.1) using a significance threshold set at a Benjamini-Hochberg-adjusted p-value of 0.05.

Quantification and statistical analysis

Statistical testing of the RNA-seq data is described in that section. Other testing was performed in Prism or Excel and involved a 2-tailed paired Student’s t test, or a one- or two-way ANOVA followed by Sidak’s post hoc test, as indicated in the legends. For t tests, variance was generally found to be similar, abrogating the need for Welsh’s Correction. Error bars indicate the SEM throughout. Throughout the manuscript, independent biological replicates are defined as independently performed experiments on material derived from different animals. All the statistical details can be found in the figure legends.

Published: July 25, 2025