Background
Neuroinflammation is a cellular and molecular response of the CNS to a variety of cues, including pathogens, abnormal protein deposits, toxic metabolites, trauma, autoimmunity or massive cell damage, which has an inflammatory character. Microglia, the innate immune cells of the brain, are the main cell type involved in this response, with astrocytes also playing a prominent role. Chronic and/or exacerbated neuroinflammation is associated with the production of potentially cytotoxic molecules, such as oxygen and nitrogen free radicals, pro-inflammatory cytokines or proteases [
1], and this is thought to contribute to neurodegeneration in a growing list of neurological and psychiatric disorders [
2].
In neuroinflammation, microglia undergoes massive changes in gene expression that are regulated by a reduced number of transcription factors. There is strong evidence showing that pro-inflammatory gene expression by activated microglia is regulated by the transcription factors NF-κB, AP-1, CREB, STATs, C/EBPβ and C/EBPδ, whereas PPARγ and Nrf2 are two of the most important transcription factors regulating anti-inflammatory programs in microglia [
3]. Because of their bottleneck position, transcription factors are potential targets to pharmacologically modulate whole cellular programs such as the neuroinflammatory one in activated microglia.
The present study is focused on the role of C/EBPβ, a transcription factor of the b-zip class, in microglial activation. The C/EBPβ gene codes for a single mRNA which can be translated into three protein isoforms, named Full, LAP and LIP, by use of alternative in-frame translation initiation codons [
4]. C/EBPβ levels increase in activated microglia [
5] where it regulates the expression of key pro-inflammatory genes [
3,
6‐
8]. In vitro and in vivo data suggest a neuroprotective potential for C/EBPβ inhibition in microglia. Thus, neurotoxicity elicited by activated microglia in neuronal–microglial co-cultures is abolished by the absence of C/EBPβ in microglia [
3]. Also, C/EBPβ-deficient mice show reduced neuronal death and neurological deficits caused by ischemic [
9] or excitotoxic [
7] damage in vivo. Even though these in vivo findings are promising, ubiquitous C/EBPβ inhibition is probably undesirable as a therapeutic strategy because C/EBPβ plays important roles in other cell types, such as adipocytes, hepatocytes or neurons [
4], that could be compromised. Our hypothesis is that microglia-targeted C/EBPβ inhibition could be a safe and effective approach to attenuate neuroinflammation-driven neurodegeneration. In order to test this hypothesis, we generated transgenic mice with myeloid-specific C/EBPβ deficiency using a Cre-LoxP system and transcriptomically profiled C/EBPβ-deficient microglia. To explore the effects of C/EBPβ myeloid deficiency in a pathological context, we analyzed the clinical response of these animals to experimental autoimmune encephalomyelitis (EAE), an animal model of multiple sclerosis. The data here presented show that microglial C/EBPβ absence results in remarkable effects on pro-inflammatory gene expression programs and in the amelioration of EAE symptomatic phenotype.
Methods
Human samples
Postmortem human temporal cortex samples were supplied by the Neurological Tissue Bank of the Biobanc-Hospital Clínic-IDIBAPS (Barcelona, Spain): healthy controls (
n = 4; 2 ♀, 2 ♂; age, 66–81 years; postmortem delay, 3.5–23.5 h) and patients with a diagnosis of primary progressive MS (
n = 5; 2 ♀, 3 ♂; age, 46–68 years; postmortem delay, 3–8 h). Protein was extracted from frozen tissue blocks [
10], and it was processed for Western blot analysis as described below.
Animals
Mice with myeloid C/EBPβ deficiency were generated by crossing transgenic mice expressing Cre-recombinase under the lysozyme M (LysM) promoter (B6.129P2-Lyz2
tm1(cre)Ifo/J 004781, Jackson Laboratories) with mice carrying a C/EBPβ gene flanked by LoxP sites [
11] kindly donated by Prof Esta Sterneck (National Cancer Institute, Frederick, MD). These mice are referred to as LysMCre-C/EBPβ
fl/fl in the manuscript. LysMCre, C/EBPβ
fl/fl and LysMCre-C/EBPβ
fl/fl colonies, all on a C57BL/6 background, were intercrossed every three generations to diminish inbreeding. Mouse tail samples were used for genotyping the presence of the Cre transgene and the floxed C/EBPβ alleles. DNA was extracted and purified using KAPA Mouse Genotyping Kit (Kapa Biosystems). One microliter of the supernatant was used for polymerase chain reaction (PCR) amplification with specific primers for the LysMCre (5′-CCCAGAAATGCCAGATTACG-3′ mutant, 5′-CTTGGGCTGCCAGAATTTCTC-3′ common, 5′-TTACAGTCGGCCAGGCTGAC-3′ wild type) as described in the JAX mouse database (The Jackson Laboratory) and the C/EBPβ floxed allele (forward: 5-GAGCCACCGCGTCCTCCAGC-′3′, reverse: 5′-GGTCGGTGCGCGTCATTGCC-3′). PCR products were loaded on 2% agarose gels to check the presence (700 bp) or absence (350 bp) of the LysMCre transgene and the wild-type (240 bp) or floxed (320 bp) C/EBPβ allele. The mice were breed and housed under specific pathogen-free conditions in the animal facilities at the School of Medicine, University of Barcelona.
Primary microglial cultures
Microglial cells were isolated from primary mixed glial cultures prepared from P1-P3 C/EBPβ
fl/fl and LysMCre-C/EBPβ
fl/fl mice. The brains were dissected, the meninges removed and the cortices digested with 0.25% trypsin for 30 min at 37 °C. Trypsinization was stopped by adding an equal volume of culture medium (Dulbecco’s modified Eagle’s medium-F-12 nutrient mixture, fetal bovine serum 10%, penicillin 100 U/mL, streptomycin 100 μg/mL and amphotericin B 0.5 μg/mL) with 160 μg/mL deoxyribonuclease I (all from Invitrogen or Sigma) and brought to a single cell suspension by repeated pipetting followed by passage through a 100-μm pore mesh. The solution was pelleted (7 min, 200 g) and resuspended in culture medium. Glial cells were seeded at a density of 3.5 × 10
5 cells/mL and cultured at 37 °C in humidified 5% CO
2–95% air. Medium was replaced once a week. Microglial cultures were prepared from DIV19-21 mixed glial cultures by the mild trypsinization method [
12] and used 24 h after isolation.
Ex vivo isolation of adult microglia
Microglial cells were acutely isolated from male adult mouse brains as described [
13]. Briefly, mice were deeply anesthetized with isofluorane, perfused transcardially with ice-cold PBS + 10 U/ml heparin and the brains dissected and dissociated by trypsin digestion as described in the “
Primary microglial cultures” section. Dissociated cells were then resuspended in 30% Percoll (GE Healthcare) and centrifuged for 10 min at 700
g. The myelin-containing supernatant was removed, and the pelleted cells were washed twice with ice-cold PBS. Cells were then incubated with 10 μl CD11b Microbeads (Miltenyi Biotec) and 90 μl buffer (PBS supplemented with 0.5% BSA and 2 mM EDTA) for 15 min at 4 °C. After washing, CD11b+ cells were separated in a magnetic field using MS columns (Miltenyi Biotec). The CD11b+ fraction was collected and used for further analyses. The yield was 379.415 ± 79.073 (SD,
n = 48) microglial cells per brain.
In vivo systemic lipopolysaccharide (LPS) injection
Systemic LPS injection was used as an in vivo model of acute neuroinflammatory response. Eight-week-old male mice were injected i.p. with 4 mg/kg LPS (055:B5, Sigma-Aldrich; 100 μl per animal) or vehicle (PBS), and microglia was isolated 16 h after LPS injection as described above.
Experimental autoimmune encephalomyelitis
EAE, an animal model of multiple sclerosis, was used to analyze the effect of microglial C/EBPβ deficiency in an in vivo model of a neurological disorder. LysMCre, C/EBPβfl/fl and LysMCre-C/EBPβfl/fl 7–8-week-old female mice were used. Under isofluorane anesthesia, the animals were subcutaneously injected at two sites into the flanks with 200 μl of a freshly prepared immunization cocktail containing myelin oligodendrocyte glycoprotein peptide (MOG35-55; Sigma; 100 μg/mouse), complete Freund’s adjuvant (Sigma) and Mycobacterium tuberculosis (H37R; Difco; 1 mg/mouse). Control mice received immunization cocktail without MOG35-55. Immediately after immunization and 2 days later, mice were injected i.p. with Bordetella pertussis toxin (Sigma, 500 ng/mouse). Mice were weighed and scored daily from day 8 post-immunization in a blind manner according to the following scale: 0, no deficit; 1, tail paralysis; 2, hind limb paresis; 3, incomplete hind limb paralysis; 4, complete hind limb paralysis; and 5, moribund state or death. Wet food was supplied when score 2 was reached and 200 μL of saline was administrated subcutaneously when animals scored 4. For comparison of EAE progression among genotypes, mice were scored until 52 days post-immunization at which point they were sacrificed. For analysis of C/EBPβ messenger RNA (mRNA) and protein expression in EAE, CFA- and MOG35-55-treated wild-type mice were sacrificed at various time-points post-immunization (9, 14, 21 and 28 days). The spinal cords were dissected into cervical, thoracic and lumbar regions, whereas the brains were dissected into hindbrain, midbrain and forebrain. Tissue samples were quickly frozen and stored at −80 °C.
Immunocytochemistry
Cultured cells were fixed with 4% paraformaldehyde in PBS for 20 min at RT. After permeation with chilled methanol for 7 min and three PBS rinses, cells were incubated overnight at 4 °C with the primary antibody diluted in 7% normal goat serum (Sigma) in PBS containing 0.01% sodium azide. After rinsing in PBS, cells were incubated for 1 h at RT with secondary antibody and DAPI (5 μg/mL). The primary antibodies were monoclonal mouse anti-C/EBPβ (1:500, Abcam, ab-18336), polyclonal rabbit anti-nitric oxide synthase 2 (NOS2) (1:400, BD Transduction Laboratories, 610333), polyclonal rabbit anti-GFAP (1:1000, DakoCytomation, Z0334), polyclonal rabbit anti-Iba1 (1:1000, Wako, 019-19741) and monoclonal rat anti-CD11b (1:300, Serotec, MCA711G, clone 5C6). The secondary antibodies were goat anti-mouse Alexa 546 (1:1000, Molecular Probes, A-11018), Alexa 488 (1:1000, Molecular Probes, A11017), goat anti-rabbit, Alexa 488 (1:1000, Molecular Probes, A-11070) and goat anti-rat Alexa 546 (1:1000, Molecular Probes, A11081). Microscopy images were obtained with an Olympus IX70 microscope and a digital camera (CC-12, Soft Imaging System GmbH).
Cytospin
25.000 microglial cells from one adult mouse brain isolated as described in the “
Ex vivo isolation of adult microglia” section were centrifuged for 5 min at 1000 rpm at RT using Shandon Cytospin 4 (Thermo Scientific) and collected in gelatinized slides. Cells were then fixed with 4% paraformaldehyde in PBS for 20 min at RT and washed three times in PBS. After blocking for 30 min at RT with PBS containing 1% BSA, 0.03% Triton and 10% normal donkey serum (Gibco), cells were incubated overnight at 4 °C with primary antibodies diluted in blocking solution. After washing in PBS, slides were incubated for 1 h at RT with secondary antibodies and DAPI (5 μg/mL) diluted in blocking solution. The primary antibodies were monoclonal mouse anti-C/EBPβ (1:1111, AbCam, ab-18336) and monoclonal rat anti-CD68 (1:1000, Serotec, MCA1957). The secondary antibodies were donkey anti-mouse Alexa 546 (1:1000, Molecular Probes, A21203) and goat anti-rat Alexa 488 (1:1000, Molecular Probes, A21208). Microscopy images were taken with a Nikon Eclipse E 1000 microscope and a digital camera Olympus DP72.
Histology
C/EBPβfl/fl and LysMCre-C/EBPβfl/fl 10-week-old female mice were deeply anesthetized with isofluorane and perfused transcardially with ice-cold 4% paraformaldehyde in PBS. The brain, lung, heart, mammary gland, spleen, liver, lung, kidney, femur bone and lumbar vertebrae were carefully dissected. Initial examination of macroscopic appearance of the organs was performed by reporting any gross anormality. Tissue specimens were formalin-fixed, paraffin-embedded using routine procedures and then cut into 5-μm semi-serial sections. Every fifth slide was stained with haematoxylin and eosin and examined under light microscope.
NO determination
Primary microglial cultures from C/EBPβ
fl/fl and LysMCre-C/EBPβ
fl/fl mice were treated with LPS (100 ng/mL) with or without IFNγ (0.1, 1, 10 or 30 ng/mL) for 48 h. NO production was assessed by detecting accumulation of nitrites in the conditioned medium by the Griess colorimetric assay, as described [
12].
Isolation of total proteins and Western blot
Total microglial proteins were isolated by lysing pelleted cells from one 75 cm
2 flask of primary microglial cultures per treatment condition (LPS 100 ng/mL with or without IFNγ 1 ng/mL for 24 h) and CD11b+ cells isolated from one adult brain as described in the “
Ex vivo isolation of adult microglia” section with 50 μL of RIPA buffer (containing Igepal CA-630 (10 μL/mL), sodium deoxycholate (5 mg/mL), SDS (1 mg/mL) and protease inhibitor cocktail Complete®, in PBS. Protein amount was determined by the Bradford assay, and protein samples (20 μg) were denatured (5 min 95 °C), resolved by SDS-PAGE on 12% gels and transferred to a PDVF membrane. Membranes were incubated overnight at 4 °C with primary anti-C/EBPβ antibodies of mouse origin (Abcam, ab18336; for primary cultured microglia proteins) or rabbit origin (Abcam, ab32358; for acutely isolated adult microglia proteins) diluted in both cases 1:500 in immunoblot buffer (Tris-buffered saline (TBS) containing 0.05% Tween-20 and 5% nonfat dry milk). Following three TBS-0.05% Tween-20 rinses, HRP-labelled anti-mouse (1:2000, Santa Cruz, sc-2055) and anti-rabbit (1:15000, GE Healthcare, NA934) secondary antibodies were incubated for 45 min at RT. To label the loading control protein, monoclonal mouse anti-β-actin-peroxidase (1/40000, Sigma, A3854) was incubated for 20 min at RT. Proteins were visualized with WesternBright™ Quantum-ECL (Advansta) and membranes exposed to Agfa Curix RP2 Plus films. Data are expressed as the ratio between the intensity of the C/EBPβ band and the loading control protein band (β-actin). For Western blot with EAE and human postmortem samples, the same protocol was used with the exception that the primary antibody was co-incubated with an enhancer solution (SignalBoost™Immunoreaction Enhancer Kit, Merck).
Total RNA extraction and qRT-PCR
Isolation of total RNA from microglial cell cultures was performed by lysing pelleted cells from one 75 cm
2 flask per treatment condition (LPS 100 ng/mL with or without IFNγ 1 ng/mL for 6 h) with 1 mL of TriReagent (Sigma) and 100 μl of 1-bromo-3-chloropropane (BCP, Sigma). The aqueous phase containing total RNA was recovered after centrifugation for 15 min at 12.000
g at 4 °C, mixed with an equal volume of ice-cold 70% ethanol and loaded onto a PureLink™ Micro Kit column (Invitrogen). Total RNA was then purified following manufacturer’s instructions. Isolation of total RNA from adult microglial cells was carried on after pelleting CD11b+ cells isolated from one adult brain as described in the “
Ex vivo isolation of adult microglia” section, using PureLink™ Micro Kit (Invitrogen) following manufacturer’s instructions. Isolation of total RNA from spinal cord (cervical, thoracic and lumbar) and brain (mesencephalon plus diencephalon, rhombencephalon and telencephalon) samples of EAE mice was performed using Trizol method (Tri®Reagent, Sigma-Aldrich). Total RNA was quantified spectrophotometrically with Nano Drop ND-1000 (Thermo Scientific). Reverse transcription reactions were carried out from 300 ng (cultured cells) or 1 μg (tissue) of total RNA with random primers using Transcriptor Reverse Transcriptase (Roche). Complementary DNA (cDNA) was diluted 1/30, and 3 μl were used to perform qRT-PCR with qPCRBIO Sygreen Mix Lo-ROX (PCB-P20.11-50, Vitro) in 15 μl of final volume reaction using CFX96 Thermal Cycler equipment (Bio-Rad). Measurements were performed in duplicates. Primers, shown in Table
1, were used at final concentration of 300 nM. Samples were run for 45 cycles (95 °C for 30 s, 60 °C or 62 °C for 1 min and 72 °C for 30 s). Amplification specificity was confirmed by the analysis of melting curves, and relative gene expression values were calculated using Bio-Rad CFX Managing software (Bio-Rad) with the comparative Ct or ΔΔCt method.
Table 1
Primers used in this study
Cebpb | AAG CTG AGC GAC GAG TAC AAG A | TCA GCT CCA GCA CCT TGT G |
Il23a | TGT GCC CCG TAT CCA GTG TG | AAA AGC CAG ACC TTG GCG GA |
Cybb | ACT CCT TGG GTC AGC ACT GGC | GCA ACA CGC ACT GGA ACC CCT |
Tnfα | TGA TCC GCG ACG TGG AA | ACC GCC TGG AGT TCT GGA A |
Ptges | AGG CCA GAT GAG GCT GCG GA | AGC GAA GGC GTG GGT TCA GC |
Csf3 | AGA GCT GCA GCC CAG ATC ACC | AGC TGC AGG GCC ATT AGC TTC A |
Rn18s | GTA ACC CGT TGA ACC CCA TT | CCA TCC AAT CGG TAG TAG CG |
Hprt1 | ATC ATT ATG CCG AGG ATT TGG | GCA AAG AAC TTA TAG CCC CC |
Sdha | TGG GGA GTG CCG TGG TGT CA | CAT GGC TGT GCC GTC CCC TG |
Viability
Microglial cell cultures from C/EBPβ
fl/fl and LysMCre-C/EBPβ
fl/fl mice were treated with LPS 1 μg/mL with or without IFNγ 1 or 30 ng/mL. Cells were fixed (0, 3, 5 or 7 days post-treatment) and probed as described in the “
Immunocytochemistry” section. DAPI and Iba1-positive cells were manually counted using ImageJ software. In every experiment (
n = 3), three wells per condition and three fields per well were analyzed. Results are represented as percentage of live cells on day 0.
Phagocytosis
The
Salmonella enterica serovar Typhimurium (
S. typhimurium) SV5015 strain, a His+ derivative of the SL1344 strain (mouse-virulent) [
14], was transformed with the pBR.RFP.1 plasmid [
15] to render red autofluorescent bacterial cells. To test microglia phagocytic capacity, first, microglial cell cultures from C/EBPβ
fl/fl and LysMCre-C/EBPβ
fl/fl mice were treated with LPS 100 ng/mL with or without IFNγ 1 ng/mL in antibiotic-free culture medium. Twenty-four hours after treatment, microglia were infected with
S. typhimurium for 30 min at multiplicity of infection of 5, defined as the ratio of bacteria per cultured cell. Non-ingested bacilli were eliminated by washing three times with PBS. Cells to assess phagocytic activity were then fixed as described in the “
Immunocytochemistry” section, whereas to study resolution of infection, microglial cells were further incubated for 1 h in medium containing 100 μM gentamicin (Sigma-Aldrich) to kill extracellular bacteria and then switched to medium with a lower dose of gentamicin (10 μM) for 3 h before fixation. Cells were immunostained for Iba1 as described in the “
Immunocytochemistry” section. Intracellular autofluorescent bacteria were manually counted using the ImageJ software along with DAPI and Iba1-positive cells. In every experiment (
n = 3), two wells for each condition and three fields per well were analyzed. Results are represented as percentage of infected microglial cells and number of bacteria phagocytosed at 0.5 and 4 h post
Salmonella infection.
RNA sequencing (RNAseq)
Total RNA from primary microglia was isolated as described in the “
Total RNA extraction and qRT-PCR” section. Total RNA integrity and quality were assessed with the Bioanalyzer 2100 system (Agilent). Library preparation and ultrasequencing were performed following Illumina’s (San Diego, CA) protocols. First, transfer RNA (tRNA) and ribosomal RNA (rRNA) were removed from 1 μg of total RNA using TruSeq Stranded Total RNA Sample Prep Kits (Illumina). Then, the RNA pool (mRNA + miRNA + lncRNA + other RNAs) was fragmented into pieces of approximately 200 bp using divalent cations under elevated temperature. The cleaved RNA fragments were reverse-transcribed into first-strand cDNA using reverse transcriptase and random primers. Next, the second strand was synthesized using DNA polymerase I and RNAse H. These double-stranded cDNA fragments were end-repaired by T4 DNA polymerase and Klenow DNA polymerase, and phosphorylated by T4 polynucleotide kinase. The cDNA products were incubated with Klenow DNA polymerase to generate 39 adenine overhangs, therefore allowing ligation to Illumina indexing adapters to the double-stranded cDNA ends. The adapter-ligated products were purified with Ampure XP magnetic beads (Agencourt Bioscience Corporation, Beverly, MA, USA), and libraries were amplified by 15 cycles of PCR with Phusion DNA polymerase (Finnzymes Reagents, Vantaa, Finland). Constructed libraries were validated and quantified using Bio-Rad’s automated electrophoresis system Experion and qRT-PCR, respectively. Pools of six indexed libraries were mixed (multiplexed) at equimolar ratios to yield a total oligonucleotide mix concentration of 10 nM. Finally, the resulting libraries were sequenced on the Genome Analyzer IIx platform (Illumina) to generate 150 bp single reads. Six pooled indexed libraries were sequenced in each flow cell lane. Raw sequence (FASTQ format) were processed through a series of sequential steps: (1) aggressive adapters removal; (2) alignment/mapping of RNA sequences to the mouse genome reference (Mus_musculus.mm10) using
tophat software; and (3) sorting and cataloguing of the results by using
Samtools software [
16] in Bam files.
Bam files of RNA readings were processed using the Rsubread package [
17] in the R environment and aligned to the mm10 version of mouse genome with the function
featureCounts [
18]. Summarized readings by gene were then normalized using
voom normalization [
19] to fit the count matrix into linear modelling with the package
limma [
20]. Normalized whole read counts were used to cluster samples with package
hcluster using standard hierarchical cluster with average linkage. A linear fit model algorithm was used to obtain differentially expressed genes (DEGs) that complied with
p < 0.01 and fold change >2. DEGs by comparison were summarized using the
VennDiagram package. Heatmap visualization and clustering of DEGs was done with Genesis software after expressing gene values in standard deviation and using hierarchical clustering with average linkage by genes and samples. Finally, Genecodis [
21] tools were used to obtain Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway terms enrichment using the adjusted
p value <0.01 after a hypergeometric test. Reads per kilobase per million mapped reads (RPKM) filtered values (sum of RPKM > 2 for each gene) were introduced to the Weighted Correlation Gene Network Analysis (WGCNA) package in R, to perform WGCNA as previously described [
22]. Combined samples were surveyed and measured with the pickSoftThreshold function to obtain a correct power β considering the smallest value to give a free-scale topology. Using 9 as the power value, the blockwise function was used with parameters mergeCutHeight = 0.001 and detectCutHeight = 0.995 to obtain 34 modules codenamed by color, containing genes with high co-expression similarity. Module relationship with traits: treatment (Vehicle = 0; LPS = 1; LPS + IFNγ = 2) and genotype (C/EBPβ
fl/fl = 0; LysMCre-C/EBPβ
fl/fl = 1) was also calculated for each module to obtain a correlation. Gene symbols and normalized values were appended, and heatmap visualization of individual modules was obtained within the WGCNA package in R. Finally the Metacore™ platform was used to obtain the described interactions of proteins among the different modules to produce “external knowledge networks”, while VisAnt software was used with the correlation values in our experiments to produce the specific network of correlation within our data. For the separate analysis of DEGs in each experimental condition, samples of each group (control, LPS, LPS + IFNγ) were used in a separate pipeline, then a linear fit algorithm with Bayesian correction was applied grouping biological replicates of samples and designing a contrast matrix that compared genotypes selecting genes which complied with criteria fold change > 2, adjusted
p < 0.05 (Benjamini–Hochberg procedure) as sample number was lower and power diminished for these analyses. DEGs in these lists were mostly in agreement with the whole comparison matrix DEGs.
Statistics
Statistical analyses of RNAseq data are described in the “
RNA sequencing (RNAseq)” section. For other parts of the study, statistical analyses were performed using one-way ANOVA followed by Newman–Keuls post hoc test when three or more experimental groups were compared. When the effect of treatment on genotype or the opposite were studied, two-way ANOVA followed by Bonferroni post-test was used. Non-parametric measures (EAE scores) were analyzed with Kruskal–Wallis test and Dunn’s multiple comparison post-test. Values of
p < 0.05 were considered statistically significant. Results are represented as mean ± standard error of the mean (SEM). Experimental data were analyzed using GraphPad Prism 5.01 software.
Discussion
C/EBPβ-deficient mice show reduced neuronal damage induced by excitotoxic or ischemic insults in vivo [
7,
9]. Since C/EBPβ is expressed by many cell types, the question of which is the cell type(s) responsible for the neuroprotective effects of C/EBPβ absence in these models is of interest. In neuron–microglia co-cultures, the absence of C/EBPβ only in microglial cells completely abrogates their neurotoxic effects elicited by LPS + IFNγ activation [
3]. This led us to hypothesize that C/EBPβ inhibition in the microglia could have therapeutic potential as a target to attenuate deleterious effects of neuroinflammation. As a proof of concept, we have generated mice with C/EBPβ deficiency in myeloid cells. The results here presented show a marked attenuation of clinical symptoms of EAE in these mice. Besides, RNAseq analysis of cultured microglia shows that C/EBPβ plays a key role in the regulation of gene transcription in microglial activation and provides possible explanations for the neuroprotective effects of C/EBPβ absence in myeloid cells.
In order to produce C/EBPβ deletion in microglia, we have used the Cre-LoxP approach with Cre expression under the control of the LysM promoter [
27]. Unlike astrocytes, for which GFAP promoter is considered the gold standard, there is not an ideal promoter to drive Cre expression in microglia. Promoters of LysM, CD11b, F4/80, CSF1 receptor or Iba1 genes have been used successfully, and in the last few years, CX3CR1 is becoming the most widely used promoter in this respect (reviewed by [
28]), but in all cases, Cre expression is induced not only exclusively in microglia but also in other subsets of myeloid cells. The LysMCre mouse line has been used extensively to induce Cre expression in myeloid cells, particularly in macrophages and microglia. Data from crossing LysMCre mice with reporter mice have shown functional Cre in 30–45% of microglial cells in control CNS [
29,
30] and strong recombination in primary microglial cultures [
30‐
32]. Various studies have obtained positive results by using LysMCre in in vivo models that have been attributed primarily to microglia [
33‐
35] or to microglia and macrophages [
31,
36]. Our data clearly show an efficiency of recombination of LysMCre close to 100 and 90% in primary microglial cultures and in microglia in vivo in LPS-treated mice, respectively, supporting the use of this mouse line to drive Cre expression in microglia in these experimental models. In our opinion, the most likely interpretation of the neuroprotective effects of LysM-driven C/EBPβ deletion in EAE is that the absence of C/EBPβ in microglia results in an attenuated neuroinflammatory response, milder neurodegeneration and less severe EAE symptoms. However, macrophages and granulocytes show also LysM-driven Cre expression in this mouse line [
27], and in both cell types, C/EBPβ expression has been reported [
37]. We cannot therefore discard that C/EBPβ deletion in macrophages and granulocytes accounts, at least partly, for the neuroprotective effects observed in LysMCre-C/EBPβ
fl/fl mice in EAE. Unfortunately, specific inhibition of C/EBPβ in microglia is not feasible at present. Strategies aimed at targeting microglia, be it nanoparticles, vectors or other approaches, cannot avoid targeting also macrophages and often other myeloid cells [
38]. Although we would favor a microglial-specific strategy, these results suggest that, if unavoidable, C/EBPβ inhibition not only in microglia but also in microglia-related cells would not be necessarily undesirable.
In contrast to mice with full C/EBPβ deficiency, LysMCre-C/EBPβ
fl/fl mice show normal fertility and survival. This is important in this study because it allowed us to obtain genetically homogenous litters of LysMCre-C/EBPβ
fl/fl pups from which to prepare primary microglial cultures in sufficient amounts to thoroughly characterize the role of C/EBPβ in microglial activation in vitro. In agreement with data from microglial cultures of C/EBPβ-deficient mice [
3], LysMCre-C/EBPβ
fl/fl microglia in culture showed normal growth and proliferation and reduced NOS2 expression and NO production upon LPS + IFNγ challenge. Since NO production is an important contributor to AICD in microglia [
25], we hypothesized that AICD, a mechanism of elimination of overactivated cells best described in T cells [
39], could be attenuated in LysMCre-C/EBPβ
fl/fl microglia. Strong AICD was observed in LPS + IFNγ-treated C/EBPβ
fl/fl microglial cells, but this was unaffected by C/EBPβ absence indicating that C/EBPβ-independent factors play important roles in microglial AICD. Since AICD is a protective mechanism that prevents overactivation [
39], its maintenance in the absence of C/EBPβ could be a positive outcome of an eventual pharmacological strategy based on the inhibition of microglial C/EBPβ.
The RNAseq analysis is the greatest leap forward this study provides on the characterization of the role of C/EBPβ in microglia. Various studies have analyzed the role of C/EBPβ on gene expression by microarrays after genetic deletion, inhibition by RNAi or overexpression in a variety of cell types. Particularly relevant are the studies that have shown important effects of C/EBPβ on LPS + IFNγ- [
40] or IFNγ- [
41] induced gene expression in macrophages. Studies in other cell types such as anaplastic large cell lymphoma have demonstrated a role or C/EBPβ in the expression of immune response genes [
42]. The present study is the first to analyze the role of C/EBPβ in microglia with a transcriptomic approach and also the first one to analyze the role of C/EBPβ by RNAseq in any cell type. It shows massive changes in gene expression in microglia caused by the absence of C/EBPβ. One thousand sixty-eight genes show significant differences in expression in C/EBPβ-deficient microglia. Most of these genes, 867, were also affected by LPS ± IFNγ confirming the role of C/EBPβ in microglial activation, but interestingly, C/EBPβ absence affected the expression of 201 genes that were not affected by treatment, indicating a role for C/EBPβ also in the biology of non-activated microglia. Gene Ontology enrichment analysis of the 1068 genes affected by microglial C/EBPβ deficiency identified five GO terms related to immune/inflammatory response among the 25 most significant GO terms. This is a strong confirmation of a key role for C/EBPβ in the regulation of the inflammatory gene program in microglia that others and we have proposed on the basis of the analysis of a limited number of genes [
3,
7,
8,
24]. Relevant C/EBPβ-dependent genes among these classes include toll-like receptors and related proteins (Tlr1, Tlr7, Tlr8, Tlr9, Cd14, Cd180, Ly86, Nod2), cytokines (Il1a, Il12a, Il23a, Mif, Cxcl2, Cxcl3), cytokine receptors (Tnfrsf1a, Tnfrsf1b, Ccr1, Ccrl2, Csf1r) and enzymes such as inflammation-related kinases (Rps6ka4, Mapkapk2), prostaglandin synthetic enzymes (Ptges, Ptgs2 (=Cox2)) or NADPH oxidase subunits (Cybb, Ncf1). A remarkable effect of C/EBPβ deficiency was observed in the expression of Nlrp3, a key component of the inflammasome [
43]. Altogether, these findings show that the phenotype of LPS ± IFNγ-treated microglia is markedly altered by the absence of C/EBPβ, showing reduced responsiveness and attenuated responses in most arms of the pro-inflammatory program. KEGG pathway enrichment analysis also identified pathways related to inflammatory responses, such as chemokine signaling or cytokine–cytokine receptor interaction, but somewhat unexpectedly, three pathways related to phagocytosis among the five most significant KEGG pathways. Particularly interesting is the group of lysosome-related genes. The significant genes in this group had a remarkably homogenous pattern of LPS ± IFNγ-induced downregulation in C/EBPβ
fl/fl microglia that was blunted in LysMCre-C/EBPβ
fl/fl microglia. These genes included degradative enzymes such as proteases (Ctss, Lgmn), nucleases (DNAse2a), sulfatases (Ids, Gns, Arsb) and glycosylases (Naga) and genes needed for lysosome acidification and assembly (Atp6v0a1, Atp6v0b, Ap1g2, Ap1s1, Ap3m2). The higher expression of lysosome genes may explain the improved ability of C/EBPβ-deficient microglia to eliminate phagocytosed bacteria that we have observed. This pattern of attenuated pro-inflammatory gene expression and improved phagocytic and digesting capacity could be of interest in a neurodegeneration context where phagocytosis of cell debris and abnormal protein deposits is necessary, whereas chronic production of pro-inflammatory mediators can be detrimental.
The possibility to acutely isolate microglial cells from the adult CNS has allowed us to estimate the efficiency of C/EBPβ deletion in LysMCre-C/EBPβ
fl/fl microglia in vivo, and the data obtained shows it to be very high, close to 90%. This is a critical point because it opens the possibility to use these mice to study for the first time the functional role of C/EBPβ in microglia in vivo which was the initial goal when generating this colony. In order to analyze the role of C/EBPβ on transcription of microglial genes in vivo, we selected a group of five pro-inflammatory genes (Tnfa, Il23a, Csf3, Ptges and Cybb) that (1) are regulated by C/EBPβ in microglia in culture as shown by RNAseq data; (2) have a well-established role in microglial activation; and (3) play a pathogenic role in EAE. Thus, genetic deletion of Ptges [
44] and pharmacological inhibition of Tnfa [
45] or Cybb [
46] result in a significant attenuation of clinical symptoms of EAE, whereas genetic deletion of Il23a [
47] or Csf3r [
48] completely prevent the appearance of clinical EAE. The expression of these five genes in acutely isolated microglia was induced by systemic LPS injection in accordance with their pro-inflammatory character. Intriguingly, this expression was markedly attenuated in microglia isolated from LysMCre-C/EBPβ
fl/fl mice. This demonstrates that also in vivo C/EBPβ plays a major role in the regulation of pro-inflammatory gene expression in microglia.
Finally, robust neuroprotection from EAE was observed in LysMCre-C/EBPβ
fl/f mice in three independent experiments. This neuroprotective effect is probably caused by the absence of C/EBPβ not only in microglia but also in monocyte-derived macrophages since both cell types are important in EAE pathogenesis [
49,
50] and LysMCre will recombine in both [
28]. The abovementioned pro-inflammatory genes Tnfa, Il23a, Csf3, Ptges and Cybb are candidates to mediate this effect, but given the role of C/EBPβ as transcription factor, we favor the idea that this effect is mediated by a higher number of genes directly or indirectly regulated by C/EBPβ and that C/EBPβ absence results in a global alteration of the phenotype of activated microglia/macrophages, as demonstrated by the RNAseq analysis in cultured microglia, rather than by an effect of one or a few genes. These data strongly support the search for strategies to selectively inhibit C/EBPβ in microglia or microglia/macrophages as potential therapies in neurological disorders with a strong neuroinflammatory component. The neuroprotection observed in LysMCre-C/EBPβ
fl/fl mice in EAE together with the increased C/EBPβ expression in human multiple sclerosis samples point to this disease as a particularly suitable indication for such therapies.
Acknowledgements
The authors thank Esta Sterneck (National Cancer Institute, USA) for providing the C/EBPβfl/fl mice, the staff of the Animal Facilities of the School of Medicine (University of Barcelona) for the professional care of the mouse colonies, Ellen Gelpí (Banc de Teixits Neurològics, Hospital Clinic-Universitat de Barcelona, IDIBAPS) for providing the human samples, Pep Tusell (IIBB, CSIC) for the technical assistance, Maria Iñíguez and Alberto de Luis (Genomic Plataform, CIBIR) for the sequencing of RNA samples, John H. Brumell (Hospital for Sick Children, Toronto, Canada) for the pBR.RFP.1 plasmid and Antonio Juárez (University of Barcelona) for the Salmonella typhimurium strain.