Interferon regulatory factor 8/interferon consensus sequence binding protein is a critical transcription factor for the physiological phenotype of microglia

Background Recent fate-mapping studies establish that microglia, the resident mononuclear phagocytes of the CNS, are distinct in origin from the bone marrow-derived myeloid lineage. Interferon regulatory factor 8 (IRF8, also known as interferon consensus sequence binding protein) plays essential roles in development and function of the bone marrow-derived myeloid lineage. However, little is known about its roles in microglia. Methods The CNS tissues of IRF8-deficient mice were immunohistochemically analyzed. Pure microglia isolated from wild-type and IRF8-deficient mice were studied in vitro by proliferation, immunocytochemical and phagocytosis assays. Microglial response in vivo was compared between wild-type and IRF8-deficient mice in the cuprizon-induced demyelination model. Results Our analysis of IRF8-deficient mice revealed that, in contrast to compromised development of IRF8-deficient bone marrow myeloid lineage cells, development and colonization of microglia are not obviously affected by loss of IRF8. However, IRF8-deficient microglia demonstrate several defective phenotypes. In vivo, IRF8-deficient microglia have fewer elaborated processes with reduced expression of IBA1/AIF1 compared with wild-type microglia, suggesting a defective phenotype. IRF8-deficient microglia are significantly less proliferative in mixed glial cultures than wild-type microglia. Unlike IRF8-deficient bone marrow myeloid progenitors, exogenous macrophage colony stimulating factor (colony stimulating factor 1) (M-CSF (CSF1)) restores their proliferation in mixed glial cultures. In addition, IRF8-deficient microglia exhibit an exaggerated growth response to exogenous granulocyte-macrophage colony stimulating factor (colony stimulating factor 2) (GM-CSF (CSF2)) in the presence of other glial cells. IRF8-deficient microglia also demonstrate altered cytokine expressions in response to interferon-gamma and lipopolysaccharide in vitro. Moreover, the maximum phagocytic capacity of IRF8-deficient microglia is reduced, although their engulfment of zymosan particles is not overtly impaired. Defective scavenging activity of IRF8-deficient microglia was further confirmed in vivo in the cuprizone-induced demyelination model in mice. Conclusions This study is the first to demonstrate the essential contribution of IRF8-mediated transcription to a broad range of microglial phenotype. Microglia are distinct from the bone marrow myeloid lineage with respect to their dependence on IRF8-mediated transcription.


Introduction
Microglia, the resident mononuclear phagocyte (MP) system of the CNS, play principal roles in the innate immune response to tissue damage and invasion of foreign organisms, and also in initiation and modulation of the adaptive immune response [1][2][3]. Although microglia share many common features of MPs with bone marrow (BM)-derived monocyte lineage cells and tissue macrophages of monocyte origin, recent studies have demonstrated that the contribution of BM-derived myeloid cells to microglial homeostasis in the steady state is a far less frequent event than was previously believed [4,5]. Only under some abnormal conditions such as tissue alteration following irradiation to the CNS, do a subset of circulating monocytes (Ly-6C/Gr-1 hi CCR2 + ) become engrafted as BM-derived microglia (BMDM) through the CNS vasculature [5]. Moreover, in vivo fate mapping studies have established that microglia are an ontogenically distinct population in the MP system. Microglial precursors arise from primitive extra-embryonic hematopoiesis in the yolk sac prior to the onset of blood circulation. These primitive MPs migrate into the CNS through blood vessels independently of the second wave of hematopoiesis within the embryo proper, which leads to the adult (definitive) hematopoiesis in the BM [6]. Once having colonized the CNS, embryonic microglia are highly proliferative and strongly dependent on colony stimulating factor 1 receptor (CSF1R)-mediated signaling for development during embryogenesis. In mice lacking CSF1R, the number of microglia is greatly reduced [6,7], whereas circulating monocytes are present, indicating that, in contrast to microglia, the BM-derived monocytes are less dependent on CSF1R-mediated signaling for their development [6]. These accumulating pieces of evidence further support the notion that microglia are maintained by self-renewal as a distinct MP population without continuous replenishment by systemic BM-derived precursors or myeloid cells, although it remains to be clarified to what extent microglia share common molecular mechanisms with BM-derived MPs in their development, maintenance, and functioning.
The molecular basis of myeloid development in the BM hematopoiesis has been intensively studied. Among the molecules known to be involved in myeloid development to date, interferon regulatory factor-8 (IRF8, also known as interferon consensus sequence binding protein (ICSBP)), a member of the interferon regulatory factor family, acts as an essential transcription factor for differentiation and maturation of BM-derived MPs in the myeloid lineage as well as for B cell development from hematopoietic stem cells [8][9][10]. IRF8 limits the size of the bipotential granulocytemacrophage progenitor pool, and directs these progenitor cells to differentiate into the monocyte lineage by stimulating expression of genes critical for MP differentiation and by repressing a series of genes required for granulocytic differentiation [11]. IRF8-deficient BM-derived MPs are not generated efficiently in IRF8-deficient mice, and are also defective in production of cytokines and reactive oxygen species [12][13][14]. IRF8 is constitutively expressed in microglia as well [15]. Since microglia arise from mesodermal myeloid progenitors in the primitive hematopoiesis, which is distinct from the development of BM-derived MPs, it is essential to determine how IRF8 regulates development of microglia to understand the molecular basis of distinct development of microglia and BM-derived MPs. Moreover, given the essential roles for IRF8 in the functional phenotypes of BMderived MPs, constitutive expression of IRF8 could also be critical for microglial functions. Indeed, a recent study has demonstrated an essential role for IRF8 in activation of microglia in the spinal cord following peripheral nerve injury [16].
In this study, we performed a comprehensive study of the microglia of constitutional IRF8-null mice, and further demonstrate essential roles for IRF8 in physiological phenotype and function of microglia.

Animals
The mouse strain with targeted null mutation in the Irf8 gene in this study was reported previously [17]. The strain was backcrossed onto the C57BL/6J strain for more than 5 generations. Animals were housed in standard laboratory cages with unrestricted access to food and water, and maintained under 12 h light/dark cycles. All procedures using the animals were approved by the Institutional Animal Care and Use Committee of the University of California, Davis.

Reagents and chemicals
All reagents and culture media used in this study were purchased from SIGMA (St. Louis, MO, USA) and Invitrogen (Carlsbad, CA, USA), respectively, except for the following products. Mouse anti-β-actin antibody was from Cell Signaling Technology (Danvers, MA, USA). Rabbit anti-IBA1/AIF1 antibody was from Wako Chemicals USA (Richmond, VA, USA). Rabbit anti-IRF8 antibody was reported previously [18].

Cell culture Mixed glial culture
Whole brains from 0 to 2-day-old mice were dissected and diced into 2-mm cubes. After cleaning off meninges, and vessels including choroidal plexus, the brain chunks were digested by 20 units/ml papain in Earle's buffered salt solution supplemented with 1 mM L-cystaine, 0.5 mM EDTA, 0.36% (w/v) D-glucose, and 250 units/ml DNase at 33°C for 90 min. Papain was inactivated by 1.5 mg/ml ovomucoid in PBS containing 1.5 mg/ml bovine serum albumin (BSA) and 250 units/ml DNase. The softened chunks were gently triturated by passing through a 1-ml serological pipette several times. The resulting suspension was left for 1 minute to allow undissociated tissues to settle down on the bottom, and then the supernatant containing dissociated cells was collected to another tube. After repeating this trituration step three times, the obtained cell suspension was spun at 220 × g for 15 minutes. Cells were resuspended in the MG medium; 5% (v/v) heat-inactivated fetal bovine serum, 1× N2 NeuroPlex supplement (Gemini Bio-Products, West Sacramento, CA), 100 units/ml penicillin, and 100 μg/ml streptomycin in high glucose Dulbecco's modified Eagle's medium (DMEM), and plated on a poly-D-lysine-coated T75 tissue culture flask. In some experiments, 20 ng/ml of either macrophage colony stimulating factor (M-CSF) or granulocyte-macrophage colony stimulating factor (GM-CSF) was added to the culture medium at 5 days in vitro, by which time most cultures had reached confluency. Cultures were fed with fresh medium every other day.

Magnetic cell sorting (MACS) to isolate CD11b + mouse microglia
Mixed glial cells in a T75 flask were maintained in the presence of 20 ng/ml M-CSF from day 5 to day 9 in vitro. Then cells were washed with 5 ml Ca ++ -and Mg ++ -free Hanks' balanced salt solution (HBSS-) and then 1 ml trypsin-EDTA (0.05% (v/v) and 0.53 mM, respectively, in HBSS-) was added. After a 3 minutetrypsinization at 37°C, 0.5 ml of trypsin inhibitor-BSA-DNase (0.25 mg/ml, 1.44 mg/ml and 60 units/ml, respectively, in DMEM.) was added, and detached cells were collected into a 50-ml tube containing 3.5 ml MG medium in addition to the medium saved from the cultures and from the HBSS-that had been used for wash. Cells were collected by centrifugation at 520 × g for 5 minutes, resuspended into a mixture of 40 μl CD11b microbeads (catalog number: 130-093-634, Miltenyi Biotech, Auburn, CA, USA) and 160 μl ice-cold MACS buffer (0.5% BSA, 2 mM EDTA in PBS, pH 7.2), and incubated at 4°C for 15 minutes. After a wash with 20 ml ice-cold MACS buffer, cells were resuspended into 500 μl ice-cold MACS buffer, and then applied to a MACS separation column LS (Miltenyi Biotech) in a magnetic field. The column was washed with 3 ml icecold MACS buffer three times and removed from the magnetic field. Magnetic labeled cells were eluted with 5 ml MACS buffer. After centrifugation at 520 × g for 5 minutes, cells were plated on to poly-D-lysine-coated culture ware at 1.25 × 10 3 cells/cm 2 in MG medium.

Immunoblots
Protein lysates were prepared in the lysis buffer as described previously [19]. Twenty μg of protein from each sample were size-fractioned by SDS-polyacrylamide gel electrophoresis, transferred onto a nitrocellulose membrane (Schleicher & Schnell, Keene, NH, USA) and probed with primary antibodies for IRF8 (1:5000) and AIF1/IBA1 (1:5000) for 1 h. Full range recombinant Rainbow Molecular Weight Markers (Amersham Biosciences, Piscataway, NJ, USA) were used as a reference for molecular sizes. Immunoreactive signals were detected by enhanced chemiluminescence according to the manufacture's protocol (Amersham Biosciences). Equal protein loading was confirmed by subsequent probing with the mouse monoclonal antibody against β-actin (1:1000) in each experiment.

Flow cytometry
Mixed glial cells were collected by trypsinization and stained with anti-CD45 and anti-CD11b antibodies together with anti-F4/80, anti-CD11c, or anti-Ly-6G antibody for 30 minutes at room temperature. Each antibody was diluted at 1:100 in PBS containing 1% (w/v) BSA and Fc-blocker (1:40; BD Bioscience). Samples were acquired by Cyan-ADP flow cytometry (DakoCytomation, Carpinteria, CA, USA) and analyzed with Summit software (DakoCytomation). Antibodies used for flow cytometry are listed in Table 1.

EdU (5-ethynyl-2′-deoxyuridine)-incorporation assay
Mixed glial cultures were incubated with 10 μM EdU for 6 h. Then, trypsinized cells were collected and stained with FITC-conjugated rat anti-CD11b antibody (1:100; BD Bioscience) in PBS containing 1% (w/v) BSA and Fcblocker (1:40) for 30 minutes at room temperature. After washing with PBS containing 1% (w/v) BSA three times, cells were fixed with 4% (w/v) paraformaldehyde and permeabilized with saponin-based permeabilizing buffer. EdU was detected using a pacific blue-conjugated azide (Invitrogen) according to the manufacturer's instruction. Samples were acquired by Cyan-ADP flow cytometry, and the number of EdU-positive cells in either CD11bpositive or negative cell population was separately counted by gating with Summit software. For purified microglial cultures, EdU (10 μM as a final concentration) was added to pure microglia plated on poly-D-lysine-coated 12 mm-diameter round cover slips at 5×10 4 cells/cover slip at 18 h after addition of MG alone, or MG medium supplemented with 20 ng/ml M-CSF or GM-CSF. After incubation for 6 h, cells were fixed and EdU incorporated into the nucleus was detected with azide conjugated with Alexa Fluor ™ 594 according to the manufacturer's instruction (Click-iT ™ EdU cell proliferation assay, Invitrogen). All nuclei were counted with the aid of DAPI.

Phagocytosis assay by flow cytometry
Microglia plated on a 24-well plate were incubated with Alexa Fluor ™ 488-conjugated zymosan A particles (Invitrogen) at 2×10 6 particles/ml in the MG medium. After wash with 0.5 ml HBSS-twice, cells were detached by trypsinization and collected into a 15-ml tube. After centrifugation at 520 × g for 5 minutes, cells were resuspended into 0.5 ml 0.1% (w/v) BSA in PBS, and then analyzed by CyAn-ADP flow cytometer. Mean fluorescent intensity (MFI) was calculated with Summit software.

Cuprizone-induced demyelination in mice
Demyelination was induced by feeding wild-type (Irf8 +/+ ) and IRF8-deficient (Irf8 -/-) male mice at 8 to 10 weeks of age a diet of pellet feed (Harlan Laboratories, Madison, MI, USA) containing 0.25% (w/w) cuprizone (Sigma) up to 6 weeks. Animals were sacrificed at the indicated time points for histological analyses.

Myelin staining and Oil Red O staining
After fixation with paraformaldehyde as described above, brains were embedded in paraffin. Five μm-thick coronal brain sections between bregma -1.2mm and -2.2mm were stained with Luxol fast blue (LFB) (Sigma) for myelin together with periodic acid Schiff (PAS) for microglia and demyelinated axons. For Oil Red O staining, mice were perfusion fixed by 10% formalin. Brains were removed and postfixed, and then 6 μm-thick cryosections were prepared. NovaUltra ™ Oil Red O Stain Kit (IHCWORLD, Woodstock, MD, USA) was used to stain undigested myelin lipids according to the manufacture's instruction.

Statistical analysis
Unless otherwise noted, data are presented as mean ± SD, and statistical significance was determined by twotail ANOVA followed by Student-Newman-Keuls post hoc test. Results were considered significant when P < 0.05.

Results
IRF8-deficient microglia populated the CNS normally, but demonstrated less complex morphology with less IBA1 expression than wild-type microglia Our immunohistochemical analysis of microglia in the CNS tissues of wild-type (Irf8 +/+ ) and IRF8-deficient (Irf8 -/-) adult mice first revealed that most Irf8 -/microglia could not be identified as cells positive for AIF1/ IBA1, a common marker protein for microglia. In contrast to AIF1/IBA1, Irf8 -/microglia expressed other microglial markers such as CD11b, CD68 and F4/80 at levels comparable with those of Irf8 +/+ microglia. Among these markers, we found that double immunolabeling for membrane CD11b and nuclear PU.1/SFPI1 (SPI1 in humans), a transcription factor exclusively expressed in hematopoietic lineage cells, is the most reliable method to identify microglia in both Irf8 +/+ and Irf8 -/-CNS tissues particularly for quantitative analysis ( Figure 1A). In the corpus callosum and the dorsal column of the spinal cord, nearly 100% of CD11b-positive microglia were positive for nuclear PU.1/ SFPI1. Quantitative analysis of microglia in the dorsal column of the spinal cord using these microglial markers revealed that the proportion of microglia to total cells was not different between Irf8 -/and Irf8 +/+ mice ( Figure 1B). These results indicate that, in contrast to compromised development of the BM myeloid lineage in Irf8 -/mice, development and colonization of microglia are not obviously affected by loss of IRF8.
In addition to the reduced expression of AIF1/IBA1 in Irf8 -/microglia, we also found a clear difference in their morphology in vivo. Generally, Irf8 +/+ microglia in the normal CNS tissues extend elaborate and highly-branched processes which sometimes reach a distance of more than 50 μm from the cell soma and cover a broad area. As a good example, transverse sections of the Irf8 +/+ dorsal column contain numerous CD11b-positive processes, some of which were from the microglia outside of the sections. In contrast, CD11b-positive Irf8 -/microglial processes in the dorsal column were apparently sparser than those in the Irf8 +/+ counterpart ( Figure 1C, Figure 2). This difference was statistically significant when total CD11bpositive areas in the whole cross-sectional area of the dorsal column were compared quantitatively between Irf8 +/+ and Irf8 -/mice. We observed a similar difference in other CNS regions such as the corpus callosum ( Figure 2). Confocal z-stack images demonstrated that the processes of Irf8 -/microglia were generally thick and less extended compared with those of Irf8 +/+ microglia ( Figure 1C), suggesting that, since the density of microglia was similar between Irf8 +/+ and Irf8 -/mice, Irf8 -/microglia were . Note that few Irf8 -/microglia demonstrated detectable immunoreactivity for AIF1/IBA1. Scale bar: 100 μm. **P < 0.01. (C) 20 μm-thick z-stack confocal images of the microglia immunolabeled for CD11b (green) and PU.1 (red) in the dorsal columns at the Th1 spinal level. CD11b-positive microglial processes were much sparser in the dorsal column of Irf8 -/mice than those in Irf8 +/+ mice, and processes of Irf8 -/microglia were thicker and less elaborated than those of Irf8 +/+ microglia. PU.1-positive microglial nuclei are indicated by arrows, demonstrating that similar numbers of microglia are distributed in the Irf8 +/+ and Irf8 -/spinal cords in contrast to the different density of the microglial processes. Scale bar: 100 μm. defective in forming normal extended processes with complex arborization.
IRF8-deficient microglia did not proliferate well in mixed glial cultures, but demonstrated a hyperproliferative response to GM-CSF We next studied Irf8 -/microglia in vitro. We initially tried to isolate Irf8 -/microglia by the standard 'shaking-off' method from the mixed glial cultures [21]. Compared with Irf8 +/+ microglia, however, we found that Irf8 -/microglia did not proliferate well in the mixed glial cultures, resulting in a much lower yield of pure Irf8 -/microglia than that of Irf8 +/+ microglia. Reduced proliferation rates of Irf8 -/microglia in the mixed glial cultures were confirmed by EdU incorporation assay ( Figure 3J), suggesting that Irf8 -/microglia did not respond normally to undetermined mitotic signal(s) present in the mixed glial cultures.
Since Irf8 -/-BM myeloid progenitors are known to exhibit altered growth response to GM-CSF, G-CSF (granulocyte colony stimulating factor), and M-CSF, the cytokines involved in proliferation and differentiation of the myeloid lineage [22], we added M-CSF and GM-CSF to the Irf8 +/+ and Irf8 -/mixed glial cultures to examine the growth response of CD11b-positive microglia. Unlike Irf8 -/-BM myeloid progenitors, addition of M-CSF significantly enhanced proliferation of Irf8 -/microglia in mixed glial cultures. Importantly, the rates of EdU incorporation were nearly equivalent between Irf8 +/+ and Irf8 -/microglia in the presence of M-CSF. In contrast, Irf8 -/microglia in mixed glial cultures exhibited a hyperproliferative response to GM-CSF compared with Irf8 +/+ microglia ( Figure 3J). The lineage progression of Irf8 -/-BM myeloid progenitors  is skewed toward cells of the granulocytic lineage even in the presence of M-CSF [22]. Therefore, we further examined alteration in the myeloid phenotype of Irf8 -/microglia in response to either M-CSF or GM-CSF. Our results confirmed that M-CSF did not induce CD11c and Ly-6G, the dendritic cell (DC) and granulocyte markers, respectively, in both Irf8 +/+ and Irf8 -/microglia. On the other hand, GM-CSF reduced expression of F4/80 and increased CD11c-positive population in Irf8 -/microglia ( Figure 4).
Based on these results, we isolated microglia from both Irf8 +/+ and Irf8 -/mixed glial cultures pretreated with M-CSF (20 ng/ml) for further in vitro analyses. To ensure the purity and maximum yield of microglia, the cells that were surface positive for CD11b were sorted out using MACS anti-CD11b microbeads. In the Irf8 +/+ and Irf8 -/microglial cultures which we obtained by this method, CD11b-positive cells were 96.2 ± 3.0% and 90.6 ± 2.3% of total cells, respectively, at 24 h after isolation. These Irf8 +/+ and Irf8 -/microglia were difficult to distinguish by morphology in vitro under a phase contrast microscope ( Figure 5A-D). As reported by Campbell's group [15], constitutive expression of IRF8 mRNA and protein in the isolated microglia in vitro was confirmed by qPCR and immunoblotting, respectively. IRF8 mRNA levels were comparable to those of PU.1/SFPI1 mRNA in Irf8 +/+ microglia, and further upregulated by interferon-γ but not by lipopolysaccharide (LPS). IRF8 protein levels were regulated in accordance with mRNA levels. No immunoreactivity for IRF8 was detectable in the isolated Irf8 -/microglia ( Figure 5E, F). Purified Irf8 -/-CD11b + microglia demonstrated significantly lower EdU incorporation rates than Irf8 +/+ CD11b + microglia in control cultures and even in the presence of exogenous M-CSF. Interestingly, both Irf8 +/+ and Irf8 -/-CD11b + microglia no longer demonstrated proliferative responses to exogenous GM-CSF in the absence of concomitant other glial cells ( Figure 5G).

IRF8-deficient microglia demonstrated a reduced maximum phagocytic capacity in vitro
Like their myeloid cousins, microglia are characterized by their motility and phagocytic activity. Motility of Irf8 +/+ and Irf8 -/microglia was analyzed in vitro by the same time-lapse video imaging as used in our previous study [23]. In vitro motility analysis failed to reveal significant differences between the two genotypes (data not shown).
Alexa Fluor ™ 488-conjugated zymosan particles were added to the pure microglial cultures to assess the phagocytic activity of Irf8 +/+ and Irf8 -/microglia by flow cytometry. Soon after addition of the particles, both Irf8 +/+ and Irf8 -/microglia started taking them up actively, and there was no difference in the percentages of the cells containing the fluorescent particles between Irf8 +/+ and Irf8 -/microglia throughout the experiment ( Figure 6C), indicating that the molecular mechanism necessary for internalization of zymosan functioned normally in the absence of IRF8. On the other hand, we observed that Irf8 -/microglia contained less zymosan particles than Irf8 +/+ microglia when their phagocytic activities reached a plateau (compare Figures 6A and  6B). In support of this observation, the averaged fluorescence intensity was significantly reduced in Irf8 -/microglia compared with Irf8 +/+ microglia at 24 hr after addition of zymosan particles ( Figure 6D). As the fluorescence emission or the profile of the excitation/emission spectra of the Alexa Fluor ™ dye is unchanged over the pH range 4 to 9 [24], the averaged fluorescence intensity is quite likely to correlate with the amount of phagocytosed zymosan, thereby confirming that the maximum phagocytic capacity of Irf8 -/microglia was reduced. IRF8 was necessary to maintain the basal transcriptional level of AIF1/IBA1 in microglia, but not for its induction by interferon-γ To explore the mechanisms underlying reduced IBA1/ AIF1 expression in Irf8 -/microglia in vivo, we examined IBA1/AIF1 expression in Irf8 +/+ and Irf8 -/microglia in vitro. In agreement with our in vivo results and another  study [16], Irf8 -/microglia expressed far less IBA1/AIF1 protein than Irf8 +/+ microglia in the control cultures ( Figure 7B). The qPCR results also demonstrated that the basal levels of AIF1/IBA1 mRNA in Irf8 -/microglia were approximately five-fold lower than those in Irf8 +/+ microglia ( Figure 7A). However, both AIF1/IBA1 mRNA and protein were upregulated by IFNγ even in the absence of IRF8. Interestingly, this IFNγ-mediated induction of AIF1/ IBA1 was significantly inhibited by simultaneous stimulation with LPS, particularly in Irf8 -/microglia (Figure 7).

Altered cytokine induction in IRF8-deficient microglia in response to LPS and IFNγ
In BM-derived myeloid cells, IRF8 is critically involved in transcriptional regulation of various cytokines which are essential for initiation and modulation of the adaptive immune response [14,25,26]. Among those cytokines, we asked whether IRF8 regulates the transcriptional induction of Interferon-β1 (IFNB1) and Interleukin-12b (IL12B) in response to LPS and IFNγ in microglia in the same manner as reported in preceding studies using BM-derived myeloid cells. As shown in Figure 8A, quantitative and kinetic analysis of IFNB1 mRNA revealed that IRF8 contributes to the transcriptional regulation of IFNB1 in microglia in at least two ways; First, IRF8 positively regulates acute transcriptional induction of IFNB1 in response to LPS, which is a distinctive mode of induction in myeloid cells [27]. In the absence of IRF8, however, this rapid induction of IFNB1 still occurs at approximately ten-fold lower levels. Second, IRF8 represses IFNγ-mediated induction of IFNB1. The qPCR results of IL12B mRNA demonstrated that IRF8 acted as a strong transcriptional enhancer in both IFNγ-and LPS-mediated transcriptional induction of IL12B in microglia ( Figure 8B).  IRF8-deficient microglia demonstrated a delayed response to tissue damage and a defect in clearing damaged tissues in the cuprizone demyelination model We finally examined whether Irf8 -/microglia are defective in their functions in vivo. As we consider microglia as a myeloid population distinct from BM-derived myeloid cells, it was essential to evaluate their functions with minimum contributions of BM-derived myeloid cells from the systemic circulation. Cuprizone-induced demyelination in mice is an established rodent model in which cuprizone in the diet selectively kills myelinating oligodendrocytes particularly in the corpus callosum. Importantly, the blood brain barrier remains intact during this intoxication [28], and contributions of BM-derived MPs and T lymphocytes from the circulation are quantitatively small [29,30]. We first confirmed that numbers of microglia, myelinating oligodendrocytes, and their precursors that are recognized by surface expression of NG2 chondroitin sulfate proteoglycan (CSPG4) were not altered quantitatively in the corpus callosum of Irf8 -/mice before cuprizone intoxication ( Figure 9A). Six week-feeding with 0.25% (w/w) cuprizone induced demyelination in the corpus callosum in both Irf8 +/+ and Irf8 -/mice, although more myelin remained in the corpus callosum of Irf8 -/mice than in Irf8 +/+ mice. However, surviving myelinating oligodendrocytes labeled with CC1 monoclonal antibody were reduced similarly in both Irf8 +/+ and Irf8 -/mice during cuprizone feeding, indicating the same cytotoxicity of cuprizone to Irf8 +/+ and Irf8 -/oligodendrocytes ( Figure 9B). Nevertheless, we found delayed accumulation of microglia double labeled with CD11b and PU.1/SFPI1 in the demyelinated corpus callosum of Irf8 -/mice. Interestingly, the activated Irf8 -/microglia had characteristic CD11b-positive oval cell soma with a clear margin, whereas activated Irf8 +/+ microglia extended CD11bpositive cell soma diffusely, which made it difficult to distinguish each Irf8 +/+ microglia by single label staining for CD11b ( Figure 9C). More notably, oil red O staining which visualized myelin lipids revealed accumulation of myelin debris in Irf8 -/mice, indicating defective scavenging activity of Irf8 -/microglia ( Figure 9D).

Discussion
The first finding in this study is that, even in the absence of IRF8, microglia can develop and colonize the CNS. This contrasts clearly to inefficient development of BMderived MPs and accumulation of myeloid progenitor cells in Irf8 -/mice [11], and indicates that IRF8-mediated transcriptional regulation is not necessary for There was no quantitative difference in corpus callosal microglia, myelinating oligodendrocytes, and NG2-positive oligodendroglial precursor cells between Irf8 +/+ and Irf8 -/mice before cuprizone feeding. Microglia and myelinating oligodendrocytes were identified by immunoreactivity for nuclear PU.1 and for the cytosolic CC1 epitope, respectively. (B) Cuprizone-mediated demyelination in the medial corpus callosum in Irf8 +/+ and Irf8 -/mice. Representative LFB-PAS-stained paraffin sections of the medial corpus callosi of Irf8 +/+ and Irf8 -/mice at 4 and 6 weeks of cuprizone diet (left panel). Scale bar: 100 μm. Reduction of myelinating oligodendrocytes in the corpus callosum. CC1 + oligodendrocytes per unit area of 6 μm-thick coronal sections were counted. At least three mice were used for each data point (right graph). (C) Microglial accumulation in the corpus callosum was delayed in Irf8 -/mice compared with Irf8 +/+ mice. CD11b + microglia with PU.1 + nuclei were quantified in 6 μmthick coronal sections as in B (right graph). Note that the abnormal oval cell shape of activated Irf8 -/microglia at 6 weeks of cuprizone feeding. Scale bar: 100 μm. (D) Oil red O plus hematoxylin staining of the corpus callosum of Irf8 +/+ and Irf8 -/mice at 0, 4 and 6 weeks of cuprizone feeding. Lipid-rich myelin debris visualized by Oil red O staining (red arrows) rapidly accumulated in the corpus callosum of Irf8 -/mice, as demyelination progressed. Scale bar: 25 μm. Data are presented as mean ± standard error. *P < 0.05 by Mann-Whitney U test. development of microglia in the primitive hematopoiesis and for their migration into the embryonic CNS. On the other hand, loss of PU.1/SFPI1 blocks MP development as early as the primitive hematopoiesis in the yolk sac, resulting in lack of tissue residential MPs including microglia, liver Kupffer cells, and lung alveolar macrophages [31]. IRF8 is known to associate with PU.1/SFPI1 to activate transcription of some genes by interacting with the composite Ets/IRF-cis elements in the promoter region [32,33]. However, our result indicates that PU.1/ SFPI1 does not require IRF8 to exert its critical roles in development of microglia. After migration into the CNS, a large proportion of the adult microglial cell population arises during the first two postnatal weeks by an intense in situ proliferation in proportion to the growth of the CNS tissues [34]. Based on the observations on the mutant mice lacking CSF1R [7,35] or one of its ligands, M-CSF [36,37], this perinatal expansion of the microglial population size is largely dependent on mitotic signals mediated by the M-CSF receptor CSF1R/CD115. We can conclude from our in vivo and in vitro experiments that, even in the absence of IRF8, microglia can respond to M-CSF and proliferate.
However, colonized IRF8-deficent microglia are not phenotypically and functionally equivalent to wild-type microglia. An obvious difference is that the processes of Irf8 -/microglia in the steady state are less extended into the tissues compared to those of Irf8 +/+ microglia. Recent two-photon in vivo or ex vivo imaging has revealed that, although ramified microglia in the steady state have been referred as 'resting' microglia, their processes are highly dynamic to probe changes actively in their microenvironment [38] and to monitor neuronal synaptic activities [39]. If microglia detect tissue damage, their processes can quickly converge to form a barrier to contain a damaged area from the surrounding healthy tissue [40,41]. Our observation suggests that Irf8 -/microglia may be less capable of sensing tissue damage due to their less extended processes. IRF8-deficent microglia will provide a unique model for further understanding of process dynamics and sensing capability of microglia.
Reduced expression of AIF1/IBA1 is also a prominent feature of Irf8 -/microglia [16]. As far as we could examine, however, AIF1/IBA1 is the only microglial marker affected by loss of IRF8, indicating again that a major part of microglial development occurs normally in the absence of IRF8. Our analysis of AIF1/IBA1 mRNA and protein levels demonstrated that IRF8 is required to maintain the constitutive expression of AIF1/IBA1 in microglia, whereas AIF1/IBA1 is still inducible in Irf8 -/microglia by IFNγ. AIF1/IBA1, a 17-kDa protein bearing two EF-hand Ca 2+ binding motifs, is known to contribute to the plasma membrane and cytoskeleton dynamics tightly linked to motility and phagocytosis of microglia [42]. It is thereby reasonable to speculate that impaired process formation in Irf8 -/microglia might be a consequence of reduced AIF1/IBA1 expression. Despite the report that AIF1/IBA1 is involved in the formation of phagocytic cups in a microglial cell line [42], however, engulfment of zymosan particles occurred in Irf8 -/microglia in the same time frame as that in Irf8 +/+ microglia, indicating that the molecular mechanisms underlying a series of microglial responses from chemotaxis to phagocytosis [43] are not impaired by the reduction of AIF1/IBA1 protein. A hypothetical explanation for this discrepancy would be that the trace amount of AIF1/IBA1 protein in Irf8 -/microglia might be sufficient for the phagocytic responses. Alternatively, AIF1L/IBA2, a homolog of AIF1/ IBA1, might compensate the reduced AIF1/IBA1 [44]. Basal expression of AIF1/IBA1 mRNA in Irf8 -/microglia, though it is five-fold reduced, is likely to be attributable to PU.1/SFPI1 that was expressed at the levels similar to those in Irf8 +/+ microglia, because PU.1/SFPI1 has been shown to bind to the promoter region of the aif1/iba1 gene and act as a transcriptional activator [45]. Further studies will be necessary to determine whether IRF8 enhances expression of AIF1/IBA1 directly or indirectly in cooperation with PU.1/SFPI1 in microglia.
In contrast to normal zymosan engulfment, the maximum phagocytic capacity of Irf8 -/microglia is reduced as determined by the amount of zymosan particles engulfed in each microglia. After engulfment, phagosomes containing internalized materials such as bacteria and apoptotic cell corpse undergo a maturation process along with sequential changes in integral membrane proteins and progressive acidification, and eventually fuse with lysosome structures to form phagolysosomes, leading to degradation of the contents [46,47]. IRF8-mediated transcriptional regulation might be involved in this highly organized maturation process of phagosomes, or cell structural changes associated with the phagocytic activities. Defective scavenging activity of Irf8 -/microglia was also observed in our in vivo experiment using the cuprizone-induced demyelination model. More myelin debris rich in lipids remained in the corpus callosum of Irf8 -/mice. However, this in vivo result may be attributable not only to reduced phagocytic capacity of Irf8 -/microglia, but also to delayed accumulation of Irf8 -/microglia in the demyelinating lesions.
After development of the CNS, microglia continue to cycle slowly [48], and are capable of accelerated proliferation particularly in response to tissue damages [30,49,50]. Although the molecular basis of this proliferative response associated with microglial activation might differ depending on the pathological conditions, CSF1R/CD115mediated signaling is principally involved in microglial proliferation in some types of CNS lesions such as facial motoneuron death after axotomy [51,52]. Microglia also proliferate in vitro in response to various cytokines including M-CSF, GM-CSF, interleukins-3, -4, and -5 [53][54][55][56][57]. An interesting finding in our in vitro experiments was that Irf8 -/microglia demonstrated significantly reduced proliferation in mixed glial cultures, whereas exogenous M-CSF restored proliferation of Irf8 -/microglia at a rate comparable to that of Irf8 +/+ microglia. Exogenous M-CSF also enhanced proliferation of purified CD11b + Irf8 -/microglia, but to a lesser extent compared with purified CD11b + Irf8 +/+ microglia, suggesting that IRF8dependent transcription in microglia is required for their normal proliferative response to M-CSF at a physiological concentration. Since a previous in vitro study using M-CSF-deficient mice indicated that M-CSF derived from astroglia principally contributes to microglial proliferation in mixed glial cultures [58], however, IRF8-dependent mechanisms in astroglia might be involved in secretion of colony stimulating factors including M-CSF from astroglia. In addition, the reduced proliferative response of Irf8 -/microglia in the mixed glial cultures could at least partly account for delayed accumulation of Irf8 -/microglia in the cuprizone-induced demyelinated lesions.
GM-CSF is known to be a potent mitogen for wildtype microglia in vitro as well [53,56,59]. In mixed glial cultures, Irf8 -/microglia demonstrated a hyperproliferative response to GM-CSF compared with Irf8 +/+ microglia, which is quite likely to be identical to that observed with Irf8 -/-BM-derived myeloid progenitor cells [22]. However, our results indicated that this hyperproliferative response of Irf8 -/microglia requires the presence of other glial cells, because exogenous GM-CSF failed to enhance proliferation of Irf8 -/microglia after isolation from other glial cells. A straightforward explanation for these findings is that GM-CSF acts indirectly on microglia through GM-CSF-mediated production of other mitogenic factor(s) from astroglia, and that IRF8 is involved in the proliferative response of microglia to these factors. Since some prior studies have pointed out that, in the absence of astroglia, GM-CSF failed to induce proliferation of microglia particularly from adult animals [60,61], it is also conceivable that GM-CSFinduced proliferation of microglia is dependent on their maturational and activation stages. Further studies will determine whether microglia share the same IRF8dependent intracellular signaling pathways leading to the mitotic response to GM-CSF as BM-derived myeloid progenitor cells [62,63]. GM-CSF also induces a phenotypic skew of microglia towards DC-like cells, but not towards granulocytes [60]. We observed that this phenotypic skew was more apparent in Irf8 -/microglia than in Irf8 +/+ microglia. Given that IRF8 positively regulates development of diverse BM-derived DC subsets in a subset-selective manner [64], roles for IRF8 in microglial differentiation towards a DC phenotype in the presence of GM-CSF might be distinct from those in BM-derived common DC progenitors.
We also confirmed that Irf8 -/microglia have the same abnormality in the transcriptional induction of IFNB1 and IL12B as that reported in Irf8 -/-BM-derived myeloid lineage cells. Our quantitative and kinetic analysis of IFNB1 mRNA clearly demonstrated that IRF8 positively regulates LPS-mediated (Toll-like receptor 4-mediated) acute induction of IFNB1 mRNA, whereas it suppresses IFNγ-mediated delayed transcriptional induction in microglia. Recently, Li et al. reported that, using human monocytes, IRF8 positively regulates rapid and robust induction of IFNβ in cooperation with IRF3 and PU.1, and that this mode of induction is characteristic to monocytes [27]. Our result is in good agreement with their findings, indicating that microglia utilize the same transcriptional machinery in the IFNB1 promoter as BM-derived monocytes. IRF8 is also an essential transcription factor for induction of IL12 p40 subunit in BMderived MPs [14,26]. Although our result confirmed the same positive regulatory role for IRF8 in microglia, we also noticed that LPS-mediated IL12B mRNA induction still occurred in the absence of IRF8 at ten-fold lower levels, suggesting the presence of a compensatory mechanism in microglia.
Our study provides further evidence for differences between microglia and BM-derived MPs with respect to the dependence of their development and physiological phenotypes on IRF8-mediated transcriptional regulation. Detailed gene expression analysis of Irf8 +/+ and Irf8 -/microglia, and Irf8 +/+ and Irf8 -/-BM-derived MPs will be necessary to clarify the downstream events leading to the difference in gene expression profiles between microglia and BM-derived MPs [65,66].

Conclusions
In conclusion, our comprehensive analyses of IRF8-deficient microglia demonstrate that IRF8 is a critical transcription factor for a broad range of microglial functions, some of which are unique in microglia. In line with our findings, recent studies have revealed direct association of IRF8 with human disease conditions [67,68]. This study warrants future studies to enhance understanding of critical contributions of IRF8-mediated transcriptional regulation to physiological and pathological roles of microglia.