Microglia in the CNS: Immigrants from another world


  • Marco Prinz,

    Corresponding author
    1. Department of Neuropathology, University of Freiburg, Freiburg, Germany
    • Department of Neuropathology, University of Freiburg, Breisacher Str. 64, D-79106 Freiburg, Germany
    Search for more papers by this author
  • Alexander Mildner

    1. Department of Neuropathology, University of Freiburg, Freiburg, Germany
    Current affiliation:
    1. Alexander Mildner is currently at Department of Immunology, The Weizmann Institute of Science, Rehovot, Israel
    Search for more papers by this author


Microglia are immunocompetent cells of the brain that continuously survey their environment with highly motile extensions. Although first described almost a century ago, the developmental origin of microglia is still debated. Besides early myeolomonocytic cell populations that colonize the developing neuroectoderm already during embryogenesis, newly migrated myeloid cells are considered important disease modulators in the adult brain. Thus, understanding the mechanisms of myeloid cell recruitment from the periphery and their development into bone marrow-derived phagocytes in the adult brain is pivotal for manipulating immune cell entry into the CNS and potentially reducing disease burden. A major question is whether bone marrow-derived mononuclear phagocytes have similar features and fates as their endogenous counterparts. Significant advances in our understanding of microglia biology under physiological as well as under pathological conditions have been made in the last few years, especially by combining novel gene-targeting techniques that allow cell manipulation and trafficking analysis. These new aspects will be discussed in the present review. ©2010 Wiley-Liss, Inc.


Microglia comprise a distinct population of glia cells in the central nervous system (CNS) and were first described by del Rio-Hortega (1932). He defined the so-called “third element” consisting of oligodendroglia and microglia, which is distinct from neurons and astrocytes. Using silver staining methods and light microscopy, Rio-Hortega not only concluded that microglia originated from the invasion of mesodermal pial elements into the CNS but also speculated about blood mononuclear cells as a potential source.

Numerous recent reports suggest that microglia are specialized cells of the mononuclear phagocyte lineage. Indeed, it has been shown that they share many features with other myeloid cells. For example, microglia express Fc and complement receptors, CD11b and F4/80 epitopes typically found on myelomonocytes (Hanisch,2002; Perry et al.,1985). It hence appears that brain microglia, like other resident macrophages, derive from myeloid precursors which originate during embryonic development. Beside these similarities, microglia also display specific molecular differences when compared with their myeloid cousins. For instance, it has been shown that microglia produced significant lower levels of superoxide dismutase (SOD) as compared to splenic or BM macrophages (Enose et al.,2005; Glanzer et al.,2007). Furthermore, under lipopolysaccharide/interferon-γ-induced CNS inflammation microglia behave differently than CNS-infiltrating phagocytes, e.g., as is reflected in the expression levels of C1qA, Trem2, and CXCL14 (Schmid et al., 2009).

In addition to the microglia that invade the brain during early embryogenesis, it has been postulated that myeloid progenitors can penetrate into the brain even in normal adult mice to replace decaying microglial cells. Moreover, it has been reported that during CNS diseases phagocytes with morphological features of endogenous microglia can be derived from bone-marrow (BM) cells or from circulating monocytes that subsequently become an integral part of the pathology and can be incorporated into the local cellular networks (Mildner et al.,2007; Priller et al.,2001; Simard et al.,2006). However, conflicting data have been published in recent years which contribute to the complexity of microglia research. In this review, we will summarize and discuss the latest findings in the field to help uncover missing pieces in the puzzle with regard to the origin and developmental fate of myeloid cells in the CNS during health and disease.


Histological data indicate that the first microglia precursors colonize the CNS during the embryonic and fetal phases of development (Rezaie and Male,2002). The first macrophage-like cells with an ameboid shape appear in the rodent neuroepithelium as early as day 8.5–10 of embryogenesis (Ashwell,1990,1991; Chan et al.,2007) independent of a developed circulatory system (Kurz and Christ,1998). At this early time point, the first immature macrophages can already be found in the yolk sac (Takahashi et al.,1996) and may be the precursors of microglial cells (Alliot et al.,1999), which then develop through a nonmonocyte pathway (Lichanska and Hume,2000; Takahashi et al.,1996). At embryonic stage 13.5, when the fetal liver is the primary hematopoietic organ and the main site of hematopoietic stem cell (HSC) expansion and differentiation (Lichanska and Hume,2000), microglial precursors can be detected in significant numbers within the ventricular lining of the fourth ventricle (Chan et al.,2007). However, a 20-fold increase of CD11b+ F4/80+ microglia cell numbers can be observed during the early postnatal periods (P0–P11) in rodents (Alliot et al.,1999). It is still unclear whether this increase is due to proliferation of embryonic microglial precursors, a phenomenon that could frequently be observed in the developing brain (Cuadros and Navascues,1998; Mander and Morris,1996) or whether a new recruitment of monocyte-derived microglial precursors occurs. The latter hypothesis is supported by the fact that the absence of microglia in mice deficient for the transcription factor PU.1 can be rescued by the injection of wild-type bone marrow into newborns, leading to a complete repopulation of the CNS by donor-derived microglia (Beers et al.,2006). However, it remains to be proven whether this postnatal recruitment of microglial precursors also occurs in wild-type animals.

In the zebrafish, yolk sac-derived macrophages enter the developing brain where they develop into immature microglia with high-endocytosis activity (Herbomel et al.,2001). The origin of these cells from the progenitor cells in the yolk sac is still unknown, but evidence suggests that these primitive macrophages develop in a PU.1-independent pathway (Lichanska et al.,1999). In contrast, PU.1-deficient mice are devoid not only of circulating monocytes and tissue macrophages (McKercher et al.,1996) but also of parenchymal microglia in the brain (Beers et al.,2006). Similarly, colony-stimulating factor 1, also known as macrophage colony-stimulating factor (M-CSF), is believed to be involved in the development of microglia (Alliot et al.,1991). Indeed, a natural null mutation in M-CSF in rodents impairs but does not completely block microglia development and results in morphological changes of microglia (Wegiel et al.,1998). A comparable phenotype could be detected in mice deficient for an adapter protein for M-CSF/M-CSFR signalling, DAP12 (Otero et al., 2009). Further studies investigated microglia development in athymic mice and demonstrated a significant reduction of CD11b+ microglia in various brain regions (Htain et al.,1997). This effect was attributed to a disturbed production of the hormone thymosin by the thymus gland, which is thought to be essential for the maturation of monocytes and again suggesting hematopoietic origin of microglia. In contrast, some researchers believe that microglia are derived from neuroectodermal matrix cells that differentiate locally into microglia (Chan et al.,2007). Others suggested that microglia originate from pericytes (Baron and Gallego,1972) or from the subependyma adjacent to the lateral ventricles (Lewis,1968). Although the myelomonocytic origin of microglia has now been widely accepted, the neuroectodermal hypothesis remains interesting from an historical point of view.


The most burning question in the field of microglia during the last years is, whether “bone marrow-derived microglia” exist and if so, whether they are functional? The answer to this question could have tremendous clinical implications for the treatment of many important diseases of the human CNS such as amyotrophic lateral sclerosis (ALS), Alzheimer's (AD) and Parkinson's disease (PD), since specific microglia precursors could be used as Trojan horses to deliver neuroprotective or immune relevant genes into the diseased CNS to modulate pathology.

The first seminal cell transplantation experiments in rats demonstrated that only perivascular macrophages, but not cells with ramified microglia characteristics in the parenchyma, could be observed in the CNS after BM transplantation (Hickey and Kimura,1988; Hickey et al.,1992). Similar results were obtained in humans, in female patients who underwent sex-mismatched BM transplantation and were examined for the engraftment of Y-chromosome-positive microglial cells (Unger et al.,1993). Importantly, only donor-derived perivascular macrophages but no parenchymal microglia could be detected in this study. In contrast to these reports, Lawson et al. (1992) used an in vivo [3H]thymidine labeling approach in mice to determine microglia turnover in steady state. The agent was injected intraperitoneally into adult mice and brains were analyzed several hours after isotope injection for ramified F4/80+ radiolabeled microglial cells. One day after dye injection, double labeled cells already showed a microglia-like cell structure as well as resident morphology and could be found throughout the brain parenchyma, the meninges and among cells of the choroid plexus. The authors concluded that microglia might be recruited from the circulating monocyte pool through an intact blood-brain barrier (BBB) and rapidly differentiate into resident microglia (Lawson et al.,1992). It is important to emphasize that all these studies are based on immunohistochemical approaches and therefore lacked the sensitivity of cell transfer experiments with genetically labeled cells. Priller et al. (2001) were among the first who used green fluorescent protein (GFP)-marked hematopoietic cells transduced with retroviral vectors to examine the long-term fate of myeloid cells in the murine CNS after BM transplantation in an experimental setting including whole body irradiation. Similar to other groups they were able to demonstrate GFP-expressing parenchymal microglia deep in the cerebellum, striatum, and hippocampus several weeks after transplantation (Eglitis and Mezey,1997; Priller et al.,2001). Despite the differences to the other studies mentioned above, the concept of bone marrow-derived phagocytes in the CNS was firmly established. Subsequently, a plethora of publications appeared in the following years which examined the assumed function and fate of BM-derived mononuclear phagocytes (BM-DP) in different neurological models using similar experimental paradigms. Infiltration of BM-DP was remarkably demonstrated in animal disease models with no obvious BBB damage such as ALS (Solomon et al.,2006), AD (Malm et al.,2005), scrapie (Priller et al.,2006), and many more (Djukic et al.,2006; Priller et al.,2001; Wang et al.,1996). However, all these studies used irradiation of the recipients followed by whole BM transplantation to discriminate between the labeled hematopoietic cells from the donor and the resident microglia in the hosts. An alternative strategy was used by Massengale and colleagues. They investigated the fate of hematopoietic cells in the brain by using parabiotic mice in which the bloodstream of a GFP-positive partner was connected to a GFP-negative mouse (Massengale et al.,2005). Although comparisons of data from different laboratories using divergent protocols are problematic, the parabiosis model always resulted in dramatically less BM-DP compared to irradiated chimeras (Massengale et al.,2005). What might have been the reason for these discrepancies in microglia engraftment between the parabiosis model and BM transplantation studies using irradiation protocols? Two possibilities were conceivable. First, lethal irradiation could have significant influences on the CNS tissue. Indeed, it had been reported that irradiation affects the integrity of the BBB (Diserbo et al.,2002; Yuan et al.,2003) and the expression of tight junction proteins (Kaya et al.,2004), and also induced apoptosis of endothelial cells (Li et al.,2004). An alternative explanation could have been the injection of BM cells by which myeloid precursors gain nonphysiological access to the circulation and thereby facilitate phagocyte generation from the blood stream. To clarify these questions, we and others performed experiments to shed light into the mystery of BM-DP (Ajami et al.,2007; Mildner et al.,2007). An experimental setup was used in which the recipient mice received only partial irradiation by excluding the head from the irradiation field and thus circumventing any irradiation-induced changes of the brain (“protected” irradiation). Importantly, de novo generation of BM-DP from the circulation was strongly diminished in the brains of mice when the tissue was not irradiated prior to transplantation (Mildner et al.,2007). However, GFP-positive BM-DP with microglia morphology could be detected in unprotected irradiated parts of the CNS such as the spinal cord. These data clearly support the idea that irradiation itself has a significant influence on the engraftment of microglial precursors to the CNS under physiological as well as under pathological conditions. Ajami et al. (2007), on the other hand, investigated the recruitment of peripheral myeloid precursors to the CNS in the parabiosis model. The bloodstream of a GFP-positive mouse was connected with a GFP-negative animal and the presence of GFP-expressing mononuclear cells was examined during steady state as well as pathogenic conditions in the CNS of the GFP-negative recipient. Unlike Massengale et al. (2005), Ajami et al. (2007) failed to detect BM-DP cells in the CNS in the GFP-negative partner under any tested conditions. These findings indicate that the engraftment of marrow-derived myeloid cells in the CNS is an extremely rare event, which is strongly influenced by the experimental design, e.g., irradiation. Furthermore, these experiments also point to the fact that endogenous microglia exhibit a high potential for self-renewal and proliferation. However, there are also some discrepancies between the two studies and questions remain. In one experiment, Ajami and colleagues irradiated the GFP-negative partner of a parabiotic pair and analyzed the brain 6 weeks after irradiation for the presence of GFP-expressing, CNS mononuclear phagocytes. However, despite the irradiation and high levels of chimerism, they still did not detect any GFP-positive cells in the recipient brain. Was this due to the fact that the mice did not receive intravenous injection of BM cells, which contain potential myeloid precursors as in BM-transplanted mice? This is an interesting possibility, but it cannot explain the fact that we failed to detect donor-derived cells in the brains of protected animals that also received an intravenous injection of BM cells (Mildner et al.,2007). It is most likely that irradiation as a precondition together with the injection of BM cells synergistically facilitated BM-DP development. One concern that arose from these studies is the supposed low chimerism of the mice examined (Soulet and Rivest,2008). In fact, the parabiosis model results in a chimerism rate of about 40–50%, which is well comparable with the rate observed in protected (head shielded but body irradiated) animals, in which also the skull bone marrow contributes to the hematopoiesis. Nevertheless, both studies undoubtedly support the hypothesis that the rate of microglia turnover from the circulation under physiological conditions in experiments using irradiation is overestimated due to irradiation-induced changes and artificial BM cell injections into the bloodstream. On the other hand, an underestimation of microglia turnover in parabiosis animal models or head-shielded irradiation methods due to low chimerism is also possible. However, much of the previous work on chimeric mouse models which involved whole body irradiation has to be re-evaluated under these aspects.

Despite all this complexity one point is undisputable: Under specific conditions, BM-DP precursors can engraft into the brain and become an integral part of the cellular network in the CNS. These specific conditions include irradiation but might also occur in certain diseases with no obvious primary BBB damage, such as X-linked adrenoleukodystrophy (ALD). A recent publication indeed reports on successful autologous stem cell transplantation with gene-modified HSCs in nonirradiated humans (Cartier et al.,2009). ALD is a fatal demyelinating disease of the CNS caused by mutations of the adenosine triphosphate-binding cassette transporter (ABCD1) gene. Cartier et al. (2009) purified CD34+ HSCs of two patients with mutated ABCD1 genes. After isolation, the cells were transduced ex vivo with a lentivirus encoding for a functional ABCD1 gene and transplanted back into the patients after they received myeloablative treatment with cyclophosphamid and busulfan. Surprisingly, despite low levels of ABCD1-expressing peripheral blood cells (only 15% of leukocytes produced ABCD1), progressive cerebral demyelination stopped in both patients and the levels of very long-chain fatty acids in the plasma as indicators for disease activity decreased. Whether the BBB was intact in these patients even on microscopic level remains to be determined, but at least gross measurements by MRI indicated no major disturbances of this physiological barrier. However, the authors attributed the beneficial outcome of the treated patients to “the recruitment of BM-derived cells to CNS macrophages/microglia” as they could show in an animal model with irradiated mice, and therefore speculated that this approach “may provide a new avenue for cell-base gene therapy in […] CNS diseases” (Cartier et al.,2009). Consequently, the identification of a microglial precursor for the therapy of CNS diseases is of great importance. But which BM cells could act as microglia progenitors?


In the last 10 years, many efforts have been made to identify the origin of different tissue macrophages as well as dendritic cells in adult mice. However, none of these studies has yet directly addressed the detailed phenotypic characterization of fetal or embryonic myeloid cells. As a consequence, the development of monocytic cells in organs beside the adult BM remains unclear. For instance, the macrophage and dendritic cell progenitor (MDP), which can give rise to monocytic as well as dendritic cells, can neither be detected under steady-state conditions in the circulation nor in the adult murine spleen (Liu et al.,2009). We will therefore mainly focus on the development of mononuclear cells arising from the BM in adulthood and the possible precursor of BM-DP.

The group of Weissman was the first to isolate a common myeloid progenitor from the BM of mice, the granulocyte/macrophage lineage-restricted progenitor (GMP), which can differentiate into all myeloid cells (Akashi et al.,2000). More recently, several groups characterized the myeloid precursors in the BM of mice in more detail and defined further cell subsets. Fogg et al. (2006) identified the MDP, which lacks the capacity to generate neutrophils but supplies both monocytes/macrophages and dendritic cells. Another group discovered that the common dendritic cell precursor (CDP) is able to develop into classical dendritic cells but not into monocytes (Fig. 1) (Liu et al.,2009). However, since monocytes have the capacity to differentiate into tissue macrophages, they could also act as circulating intermediate progenitors. A powerful tool for the characterization and isolation of myeloid progenitors was created as early as 2000 when the chemokine receptor CX3CR1 was replaced by GFP (Jung et al.,2000). This ingenious device led to the subsequent identification of at least two distinct blood monocyte subsets in mice (Geissmann et al.,2003), which may resemble the two major monocyte subpopulations found in humans (Shen et al.,1983). Monocytes are thought to circulate immature myeloid cells, which can differentiate under certain conditions to intestinal macrophages, lung alveolar macrophages or eventually also phagocytes in the brain and are involved in the innate immune response under inflammatory conditions. Indeed, these functional diverse properties could be assigned to the different monocyte populations: “Inflammatory” or Ly-6C+ monocytes—characterized by high expression of Ly-6C and the chemokine receptor 2 (CCR2)—especially migrate to inflammatory stimuli (Geissmann et al.,2003; Mildner et al.,2008,2009) whereas “resident” or Ly-6C monocytes, which lack Ly-6C and CCR2, can differentiate in vivo into tissue phagocytes such as lung macrophages (Landsman and Jung,2007) and are involved in early immune responses upon vessel disruption (Auffray et al.,2007). Adoptive transfer experiments of purified monocyte subsets helped clarify the question of specific progenitors of resident tissue phagocytes in the peripheral tissues (Geissmann et al.,2003; Varol et al.,2009). However, these approaches cannot be used for the investigation of brain phagocyte precursors under physiological conditions. The reasons for this limitation are the presence of an intact BBB, the relatively short half-life of the injected cells and the quite long differentiation time to gain the features of endogenous microglia like ramified morphology with parenchymal and not perivascular location. Therefore, techniques to study cell engraftment into the healthy, adult CNS are restricted nowadays to irradiation procedures. Nevertheless, a strong defect on BM-DP development under steady-state conditions could be demonstrated in mice in which recipients were reconstituted with BM deficient for CCR2 (Mildner et al.,2007). CCR2 is essential for the extravasation of CCR2-expressing Ly-6C+ monocytes from the BM to the circulation (Serbina and Pamer,2006). Thus, reconstitution of mice with CCR2-deficient BM leads to a dramatic reduction of circulating Ly-6C+ monocytes. Under these circumstances no BM-DP could be observed in the brains of chimeric mice, which potentially argues for Ly-6C+ monocytes as direct circulating precursors of bone marrow-derived mononuclear phagocytes in the CNS. However, recently GMPs within the BM were shown to partially express CCR2 (Si et al.,2010). Reconstitution with CCR2-deficient BM may thus also affect the CCR2-dependent migration of microglial precursors (such as GMPs) into the CNS. Further experiments are urgently needed to clarify this point in more detail. As mentioned above, beside all these identified precursors with myeloid-restricted differentiation potential, it is also possible that the microglial progenitors may differ in their phenotypical characteristics. Langerhans cells (LCs), the skin's resident dendritic cell population, show some biological similarities to microglia. For instance, LCs remain of host origin after syngeneic bone marrow reconstitution (Merad et al.,2002) and are only replenished by BM-derived elements, e.g., Ly-6C+ monocytes after LC depletion (Ginhoux et al.,2006). Just recently, the precursors of LCs were identified in vivo (Chorro et al.,2009). Interestingly, the early myeloid precursors in the embryonic skin are phenotypically distinct from adult BM myeloid progenitors in regard to the cell surface markers investigated such as Flt3 receptor (CD135) and c-kit (CD117) expression. It is therefore tempting to speculate that also the putative embryonic microglial precursors might be distinct from the adult BM myeloid cells, but more experiments are needed to clarify this point in detail.

Figure 1.

Schematic network of mononuclear cell development and the relationship to microglia. In adulthood mononuclear cells derive from HSCs in the bone marrow. Further development into the myeloid cell lineage compromises the differentiation to granulocyte/macrophage lineage-restricted progenitors (GMP) and macrophage and dendritic cell progenitors (MDP). It is possible that in the presence of M-CSF, MDPs can give progeny to Ly-6C+ (and maybe Ly-6C) monocytes whereas FLT-3L facilitates the differentiation to CDPs. Ly-6C+ monocytes leave the bone marrow in a CCR2-dependent manner and are recruited to inflammatory stimuli within the CNS, where they can achieve proinflammatory or anti-inflammatory functions. Under certain conditions, e.g. after LC depletion or after irradiation of the CNS, Ly-6C+ monocytes can differentiate into LCs in the skin or into bone marrow-derived phagocytes in the CNS. The potential of GMPs to engraft the CNS is yet unclear. Note that only 30% of GMPs express CCR2 (indicated by *). Bone marrow-derived myeloid cells are recruited to Aβ deposits during AD and are involved in phagocytotic activities. Mutation in microglial SOD1-expression leads to activated microglia and the manifestation of ALS. Endogenous microglia as well as LC in the embryo are derived from embryonic precursors and are characterized by self-renewal. SR, self-renewal; M-CSF, macrophage-colony stimulating factor; FLT-3L, FMS-like tyrosine kinase 3 ligand.


As the key immunologic cell of the CNS, resting microglia are distributed throughout the whole CNS and act as sensors for pathologic conditions. The fine processes of microglia are highly motile and seem to survey the microenvironment, whereas the soma itself is static (Nimmerjahn et al.,2005). It has proposed that the high motility of the protrusions has immunological functions such as scanning the environment for pathological changes or inflammatory stimuli, but recent experiments indicate that microglia also support and monitor synaptic function (Wake et al.,2009), control synaptogenesis (Roumier et al.,2004) and induce apoptosis of developmental Purkinje cells (Marin-Teva et al.,2004). Hence, microglia also play an important role during the development and maintenance of the CNS. In the resting state, murine microglia express low levels of major histocompatibility complex (MHC) class II molecules (De Haas et al.,2008), membrane CD45 (Ford et al.,1995; Sedgwick et al.,1991), and Ly-6C (Mildner et al.,2009). These markers can be used for the differentiation of peripheral or perivascular myeloid cells and resting microglia (Mildner et al.,2009). However, after activation of microglial cells by pathological or inflammatory events, microglia undergo morphological and immunophenotypical changes like the upregulation of MHC class II molecules (Perry,1998). Under physiological conditions, microglia are separated from the peripheral immune system by the BBB. Certain diseases will cause damage of the BBB, which is in some cases obviously due to mechanical damage (e.g., spinal cord injury) or to massive leukocyte infiltration [e.g., multiple sclerosis or its animal model experimental autoimmune encephalomyelitis (EAE)]. Under these conditions, activated microglia share a similar cell surface marker expression profile with their blood-derived counterparts, and a clear discrimination of activated microglia from CNS-infiltrating macrophages is tricky. Recent work, however, indicated that the function of endogenous microglia and engrafted blood monocytes could be substantially different. For instance, in a spinal cord injury model only engrafted monocytes seem to play an important role for tissue remodelling and recovery, whereas resident microglia seem not to be required during this process (Shechter et al.,2009). On the other hand, in autoimmune, T-cell-mediated diseases such as EAE, infiltration of the CNS by Ly-6C+ monocytes leads to exacerbated symptoms, thus indicating a pathological role of this subset (Fig. 1) (King et al.,2009; Mildner et al.,2009; Prinz and Priller,2010). Because of the breakdown of the BBB in these models, infiltrating cells and their progeny represent blood monocytes rather than long-term, engrafted, marrow-derived phagocytes seen in neurological diseases with no obvious BBB damage such as the neurodegenerative disorders AD, ALS, and PD.

Neurodegenerative diseases in general are characterized by a loss of distinct neuronal cells in defined CNS regions. Despite heterogeneity in disease pathogenesis microglia are always activated during virtually all neurodegenerative disorders. The activation of microglia can be triggered by misfolded proteins (e.g., amyloid-β in AD, α-synuclein in PD) or genetic mutations (SOD1 mutation in ALS), which in turn lead to an activation-dependent release of reactive oxygen species and proinflammatory cytokines accompanied by a loss of neuronal support. Therefore, some features of microglia accompanying neuronal loss might be similar throughout disease entities. However, due to the fact that the literature in the field is quite diverse and sometimes even contradictory, and that the different disease models are difficult to compare, we will subsequently summarize the role of mononuclear phagocytes in the most common neurodegenerative disorders separately.


Besides tau pathology, one hallmark of AD is the accumulation of amyloid-β (Aβ) protein, a proteolyticaly processed part of the amyloid precursor protein (APP). Aβ can form fibrillar structures that aggregate to senile plaques. Senile plaques are densely surrounded by microglia and therefore a contribution of these cells to the etiopathology of AD was assumed (Itagaki et al.,1989; McGeer et al.,1988). Further examinations revealed that microglia at the site of the plaques secrete complement proteins, proinflammatory cytokines and reactive oxygen species (Rogers et al.,2007). These observations led to the hypothesis that AD shows characteristics of a neurodegenerative disorder in which activated microglia may play a central role. However, despite intense research in this field—about 800 research articles can be found on this topic to date, more than for any other neurological disease—the role of microglia in AD is still obscure. Because of the close location to the senile plaques, an Aβ-phagocytosing role of microglia was proposed which indeed could be shown by some groups (Ard et al.,1996; Paresce et al.,1997; Wisniewski et al.,1991) whereas others could not find any evidence for microglia phagocytosis (Stadler et al.,2001). Microglial removal of Aβ deposits could be beneficial for the pathology, but on the other hand, persistent activation by Aβ could lead to “dystrophic” microglia, characterized by abnormally structured, cytoplasmic processes (Streit et al.,2009). This overactivation could lead to microglial degeneration in which neurons lose microglial support or even are exposed to neurotoxic factors and subsequently undergo slow neurodegeneration. Interestingly, the dystrophic appearance of microglia is associated with aging as well and seems to indicate senescence of the cells (Streit and Xue,2009). The burning question whether peripheral leukocytes are involved in the pathology of AD is still a big matter of debate. Schenk and colleagues first provided evidence of a connection between the peripheral immune system and AD when they vaccinated human APP transgenic mice with synthetic human Aβ (hAβ) (Schenk et al.,1999). Repetitive vaccination of APP transgenic mice before onset of disease led to high production of hAβ-specific antibodies and nearly complete abolishment of plaque development (Schenk et al.,1999). Beside these promising results, vaccination studies in humans resulted in severe site effects: approximately 6% of patients developed brain inflammation consistent with aseptic meningoencephalitis (Check,2002) and the clinical phase IIa trial was stopped in January 2002. However, patients who responded with the generation of highly Aβ-specific antibodies showed reduced cognitive impairments (Hock and Nitsch,2005). Despite these facts, long-term examination of AD patients could not demonstrate a beneficial outcome after vaccination despite production of Aβ-specific antibodies in these patients (Holmes et al.,2008). Further experimental data suggested that peripheral leukocytes are recruited to the diseased brain and may therefore contribute to AD in mice (Malm et al.,2005; Simard et al.,2006). It was shown that BM-DP efficiently locates to the site of plaques in irradiated APP-transgenic mice and showed a higher potential to phagocytose Aβ compared to their endogenous, senescent counterparts (Fig. 1). Furthermore, the treatment of APP-transgenic mice with M-CSF resulted in a strong reduction of Aβ deposits as well as in cognitive improvement, which was accompanied by increased microglia numbers (Boissonneault et al.,2009). Similar results were obtained when APP-transgenic mice were treated with granulocyte colony-stimulating factor (Sanchez-Ramos et al.,2009). The authors of both studies attribute this effect to the enhanced recruitment of BM-DP into the AD brain. But it still remains to be proven that marrow-derived cells are indeed recruited to the AD brain in nonirradiated animals. A recent publication, however, argues against a beneficial role of endogenous microglia in AD. Microglia in the brains of APP-transgenic animals were physically depleted for 4 weeks virtually without affecting other myeloid cell populations (Grathwohl et al.,2009). Surprisingly, the authors could not detect any changes in amyloid plaque formation and maintenance during this relatively short time period. One caveat of this report is that microglia depletion was performed at very late disease stages when amyloid plaques are already established and glial cells already exhausted. It is therefore still possible that endogenous microglia contribute to the induction of amyloid plaques in status nascendi. Indeed, earlier data pointed to a role of microglia in AD. It was demonstrated that CCR2-deficient APP-transgenic mice die significantly earlier than their CCR2-competent littermates, and strongly increased Aβ deposition was observed in them (El Khoury et al.,2007). This phenotype was attributed to decreased microglia accumulation and defective recruitment of CCR2-expressing microglial precursor to the CNS. Interestingly, the group could not detect classical AD plaques in CCR2−/− APP-transgenic mice. Instead, amyloid deposits were discovered to be localized at small blood vessels and mice were post mortem found to have intracranial hemorrhage (El Khoury et al.,2007). It is therefore possible that other myeloid cells such as perivascular macrophages instead of microglia could be significantly involved in the pathology of AD. These cells may eliminate blood vessel-associated Aβ and therefore decrease the influx of the APP cleavage product into the brain. Increasing the number of vessel-associated macrophages could potentially be beneficial, whereas the reduction of this cell population could lead to an exacerbation of the symptoms as demonstrated recently (Hawkes and McLaurin,2009). In summary, several myeloid cells might be involved in the pathogenesis of AD and interaction with distinct cell subsets could be therapeutically useful in different disease stages.


ALS is an adult-onset, progressive neurodegenerative disorder that specifically affects the upper and lower motorneurons, leading to atrophy of skeletal muscle, spasticity, paralysis and subsequently to death within 4–6 years. Twenty percent of familial ALS cases are caused by a mutation in the ubiquitously expressed gene SOD1 encoding the free radical-scavenging metalloenzyme copper, zinc SOD. That microglia are actively involved in this disease was shown in an elegant study in SOD mutant mice (Beers et al.,2006). In this study, mice reflecting symptoms of ALS through expression of the mutated SOD1 gene (SOD1G93A) were bred to transcription factor PU.1-deficient mice, which are characterized by a lack of all myeloid cells (McKercher et al.,1996). Because PU.1−/− as well as PU.1−/− SOD1G93A mice die shortly after birth (Beers et al.,2006; McKercher et al.,1996), intraperitoneal BM transfer into newborns was performed. PU.1−/− SOD1G93A newborns received wild-type or SOD1G93A BM cells, respectively, and survival as well as motorneuron loss was analyzed. Interestingly, reconstituted PU.1−/− SOD1G93A mice showed a full reconstitution of the CNS with donor-derived cells although recipient mice did not receive any prior irradiation. Full CNS repopulation by donor cells is potentially explainable by the fact that the injected donor cells could colonize the free niche within the microglia deficient PU.1−/− CNS. However, PU.1−/− SOD1G93A pups that were reconstituted with wild-type BM showed a significantly longer survival as well as decreased motorneuron loss, indicating a pathology-promoting role of SOD1G93A microglial cells (Beers et al.,2006). Similar results were obtained by using a different approach: inactivation of mutated SODG37R specifically in CD11b+ microglia extended the survival of mice significantly, particularly during the late phase of disease (Boillee et al.,2006). Indeed, the neurotoxic nature of mutated SOD1-expressing microglia was shown previously (Xiao et al.,2008) and additional activation of microglial cells in ALS mice by M-CSF treatment resulted in exacerbated symptoms (Gowing et al.,2009). This treatment led to increased proliferation and an altered morphology of microglial cells, which led to enhanced expression of proinflammatory cytokines such as interleukin 1β and tumor necrosis factor α (Gowing et al.,2009). Importantly and in contrast to the long-term reconstitution experiments in neonates (Beers et al.,2006), BM transplantation of wild-type BM cells in adult 6-week-old SOD1G93A mice (Solomon et al.,2006) or even allogeneic BM transplantation in sporadic ALS patients did not result in any beneficial outcome (Appel et al.,2008). This discrepancy could be due to inefficient engraftment of the CNS by bone marrow-derived elements in adults as demonstrated previously (Ajami et al.,2007), which could lead to a nonsufficient exchange of mutated SOD1-expressing microglia by wild-type BM-derived cells. However, the experimental results obtained so far point to a disease-promoting role of mutant SOD1-expressing microglia in familial ALS. Therefore, therapeutic application of wild-type or gene-modified microglial precursors, equipped with a high capacity to infiltrate the brain, should be combined with a preconditioning regime to facilitate engraftment into the brain.


PD is the second most common, age-related neurodegenerative disorder after AD. One pathological hallmark of PD is the degeneration of dopaminergic neurons (DNs) of the substantia nigra and the presence of intraneural eosinophilic inclusions called Lewy bodies. The cause of PD is still not clear but a certain neurotoxin, methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), has been described that induces degeneration of DN (Langston and Ballard,1983; Langston et al.,1983) and leads to the development of a MPTP-induced, PD-like disease model in animals. Because of the presence of abundant reactive MHC class II-expressing microglia in the regions of dopaminergic neuronal loss (McGeer et al.,1988) a disease-promoting role of microglia was proposed. This hypothesis was further supported by the fact that increased levels of reactive oxygen species—potentially produced by activated microglia—were found in the area of neuronal loss (Liberatore et al.,1999; Wu et al.,2003). The potential pathway leading to microglial activation comprises the MPTP-dependent production and release of misfolded α-synuclein by DN (Dauer and Przedborski,2003), which in turn can activate microglia (Su et al., 2007; Zhang et al., 2005). Further support for the possible classification of PD as a neuroinflammatory disease came from the observation that mice deficient for lymphocytes, in particular CD4+ T cells, showed attenuated DA cell death (Brochard et al., 2009). Therefore, anti-inflammatory therapies may have potential for treatment of PD patients and are currently under investigation (McGeer and McGeer,2007). But how are endogenous microglia or marrow-derived myeloid cells involved in this process? It has been shown in BM transplantation experiments that a significant increase of BM-DP was detectable in regions enriched for DN (Rodriguez et al.,2007). However, the engraftment of PD-affected brains by BM cells could again be the result of irradiation-induced changes and would therefore require re-evaluation by the usage of nonirradiated conditions. Expression of CCR2 is crucial for chemotaxis of microglia/monocytes to Aβ deposits in AD (El Khoury et al.,2007) and the migration of CCR2-expressing myeloid precursors to the CNS (Mildner et al.,2007). In sharp contrast to these findings CCR2 and its ligand, CCL2, seem to be redundant in MPTP-induced PD, since the absence of the genes does not protect against DN loss (Kalkonde et al.,2007), which argues against the engraftment of BM-DP precursors to the CNS. Also, the question as to how and why microglia should be activated to attack DN is still a matter of debate (Rogers et al.,2007). In this context, a new hypothesis assumes a disease-dependent malfunction of senescent microglia—as in AD—leading to compromised neuroprotective functions and subsequently to the loss of DN support (Graeber and Streit,2010).


Although long underestimated, microglia nowadays comprise an attractive target for accessing the diseased CNS. Their presence has been confirmed extensively in countless reports describing their involvement in virtually all neuropathologies. However, the question as to their origin is still controversial.

Given the fact that endogenous microglia under normal CNS conditions are not replenished by the periphery at all and during neurodegenerative disease only sparsely if the BBB integrity is broken, it will be a major challenge for future research to facilitate the use of this ambivalent Trojan horse to access to the brain for putative therapeutic approaches. On the other hand, it is now believed that localized self-renewal can sustain microglia maintenance and probably also function throughout adult life as a main mechanism of microgliosis under conditions of preserved BBB. Therefore, further research is necessary to understand mechanisms of microglia self-renewal, proliferation and decay that normally occur in the brain during lifetime. Finally, despite strong similarities to their peripheral counterparts, microglia of the CNS may provide some unexpected insights due to their brain-specific anatomy and physiology.


The authors acknowledge current and previous members of the Prinz lab for their continuous support and stimulating discussions. They are especially grateful to Hauke Schmidt (Göttingen), Katrin Kierdorf (Freiburg), and Steffen Jung (Rehovot).


Alzheimer's disease


amyotrophic lateral sclerosis


blood-brain barrier


bone marrow


bone marrow-derived phagocyte


dopaminergic neurons


hematopoietic stem cell


macrophage-colony stimulating factor


Parkinson's disease