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 BM-derived MPs and accumulation of myeloid progenitor cells in Irf8
-/- mice , and indicates that IRF8-mediated transcriptional regulation is not necessary for 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 . 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 . 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  and to monitor neuronal synaptic activities . 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 . 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 Ca2+ binding motifs, is known to contribute to the plasma membrane and cytoskeleton dynamics tightly linked to motility and phagocytosis of microglia . 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 , 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  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 . 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 . 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 , 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/CD115-mediated 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 [53–57]. 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+
-/- microglia, but to a lesser extent compared with purified CD11b+
+/+ microglia, suggesting that IRF8-dependent 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 , 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 wild-type 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 . 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-CSF-induced proliferation of microglia is dependent on their maturational and activation stages. Further studies will determine whether microglia share the same IRF8-dependent 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 . 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 , 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 . 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 BM-derived 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].