The notion that microglia make up the brain's immune system has replaced the longstanding tenet that the brain is an immunologically privileged organ. However, this new concept of an endogenous immune system within the central nervous system (CNS) does not represent a fundamentally different view, but a modernized and expanded version of the same basic idea, namely, that the brain is different from most other organs in its immunological properties. The fact that there is a blood-brain barrier, which prevents unlimited entry of blood-borne leukocytes into the CNS parenchyma, speaks to the need for protecting the CNS from the potentially damaging consequences of a full-fledged immune response. Neurons are fragile and fastidious cells unable to withstand long-lasting exposure to many of the toxic molecules elaborated by activated peripheral leukocytes that defend against invading microorganisms. Under normal and especially under pathological conditions, neuronal well-being and proper functioning are highly dependent on the presence of large numbers of glial cells that sustain an abundance of neuron-supporting functions. The unique and fascinating point about microglia is that they are both supportive glia and immunocompetent defense cells. Functionally speaking, microglia are a hybrid between white cells and glial cells and, as such, their role is to protect and to support neurons within the CNS. At the same time, however, it needs to be stated that the immunological competence of microglia is not on the same scale as that of peripheral leukocytes, and that microglial immune functions are controlled by inhibitory factors intrinsic to the CNS. These attenuating factors are elaborated primarily by CNS neurons; recent years have seen considerable advances in the identification of both membrane-bound and secreted factors, such as OX2, neurotrophins, and cytokines/chemokines, which have the potential for controlling endogenous immune activity within the CNS (Neumann et al., 1998; Harrison et al., 1998; Hoek et al., 2000; Streit et al., 2000, 2001). The ongoing intensive search for molecular mediators and mechanisms of neuron-microglia interactions underscores the importance of these processes for improved understanding of CNS function and disease.
Major observations that have contributed to the idea that microglia represent the brain's internal immune system include the detection of MHC antigens, T- and B- lymphocyte markers, and other immune cell antigens on microglia, as well as the fact that microglia, in general, have properties similar to those of monocytes and other cells of the macrophage lineage (Table 1). Among such properties is their ability to secrete cytokines characteristic for immune accessory cells, and to serve as antigen-presenting cells (APCs) (Frei et al., 1987; Hickey and Kimura, 1988; Hetier et al., 1988). Although it is clear that of all the CNS parenchymal cell types, microglia are the most likely to function as APCs, studies from a number of different laboratories have shown that microglia are only weak APCs compared with APCs from peripheral organs (Ford et al., 1995; Carson et al., 1998, 1999; Flügel et al., 1999). This is, perhaps, not surprising, since the CNS also harbors other macrophages that, although not part of the parenchyma, populate the critical interface between the CNS parenchyma and the blood. These cells are known as perivascular cells because they reside in the perivascular spaces (Graeber and Streit, 1990); they may be considered “professional” macrophages and APCs, because they are bone marrow-derived and capable of presenting antigen strongly (Ford et al., 1995; Hickey and Kimura, 1988). These so-called “other CNS macrophages” (Ford et al., 1995) represent a subset that is phenotypically distinct (CD11b/c + and CD45hi) from parenchymal microglia (which are CD11b/c + and CD45low), and they have immunological properties of peripheral accessory cells. Because of their location in the perivascular spaces, these cells are likely to come into contact with and process CNS antigens that may seep from the parenchyma into the perivascular space, and they may present these CNS antigens to T lymphocytes. In contrast, microglia reside in the CNS parenchyma, where they do not usually encounter T cells; hence there may not be a need or an opportunity for microglia to present antigen within the CNS microenvironment. When the blood-brain barrier becomes disrupted after traumatic/ischemic injury or during autoimmune disease, white cells from the blood, including APCs and lymphocytes, gain access to the brain parenchyma and to brain antigens, which may then be presented to lymphocytes most effectively by the extravasating peripheral APCs.
Table 1. Types of Immunological Receptor Molecules Found on Microglia
In sum, while it seems appropriate conceptually to designate microglia as the brain's immune system, one needs to qualify this statement immediately and acknowledge that the immune competence of microglia within the CNS is limited. Nevertheless, the concept is a useful one because it reconciles the discrepancy between the absence of peripheral leukocytes in the brain and the brain's ability to defend itself against infections, injury, and disease. With microglia, evolution has found a way to achieve compatibility between the destructive power of the immune system and the vulnerability of the CNS to injury and disease.
ROLE OF MICROGLIA IN REGENERATION
Numerous studies of microglial activation have been conducted in the facial nerve axotomy paradigm and these are reviewed in detail elsewhere (Raivich, 2002). For purposes of the present discussion, two main points are related to the facial nerve axotomy paradigm that I would like to stress. First, the fact that a peripheral axotomy triggers microglial activation inside the CNS points to the existence of effective neuron-microglia signaling; thus, the facial nerve model is well suited for studying neuron-microglia interactions in vivo, a topic likely to receive continued attention in coming years. Second, axotomy of the motor axons that comprise the facial nerve triggers a regeneration program inside the facial motor nucleus, which ultimately results in the reinnervation of target muscles. The activation of microglia by injured motor neurons is an integral, and potentially crucial, component of this regeneration program.
The facial nerve axotomy paradigm provides unmistakable in vivo evidence that injured neurons recover from injury when activated microglial cells are present. Microglial activation in the facial nucleus is triggered within hours after the injury has occurred, and increased numbers of activated microglia remain in the area for at least 2 weeks, at which time some axons have already reinnervated their targets. The time course and morphology of microglial activation suggest that proregenerative and neuroprotective functions are being carried out by microglia. Cells become activated within 24 h, as evidenced by hypertrophy (Graeber et al., 1988); they are actively dividing by 48 h and, as their numbers are increased dramatically by 72 h, they have completely ensheathed most axotomized motor neurons with their cytoplasmic processes. Perineuronal ensheathment of neurons by microglia accomplishes at least two neuroprotective actions: removal of excitatory input through the displacement of afferent synapses (Blinzinger and Kreutzberg, 1968), and close physical proximity of axotomized neurons to several microglial cells, which may facilitate targeted delivery of growth factors from activated microglia to injured neurons. That these morphological relationships between injured motor neurons and microglia do, in fact, reflect proregenerative events is supported by the observation that, in contrast to a peripheral axotomy, a central axotomy, such as transection of the rubrospinal tract in the cervical spinal cord that does not result in regeneration, elicits only minimal microglial activation (Barron et al., 1990; Tseng et al., 1996; Streit et al., 2000). While rubrospinal tractotomy elicits some slight hypertrophy of microglia in the red nucleus, it does not stimulate microglial mitosis or perineuronal ensheathment of rubrospinal neurons by microglia. Tractotomy results in atrophy of rubrospinal neurons, suggesting that axotomized rubrospinal neurons generate insufficient signals to stimulate robust perineuronal microglial inflammation that could help prevent their atrophy. These in vivo observations clearly support the idea that microglial inflammation is required to facilitate neuronal regeneration. Additional and more direct evidence for this idea comes from transplantation studies showing that engraftment of cultured microglial cells into the injured spinal cord promotes neurite growth into such microglial grafts (Rabchevsky and Streit, 1997). It is thought that microglia-induced production of trophic factors and extracellular matrix molecules, such as laminin and thrombospondin, accounts for the observed neuritic ingrowth.
Collectively, the in vivo observations strongly support a neuroprotective and proregenerative role of microglia in the injured CNS. In addition, there are other in vivo observation that argue against the possibility that activated microglia cause bystander damage following CNS injury, most notably the fact that presence of activated microglia in areas of tissue damage does not cause additional neurodegeneration in adjacent areas. An excellent example is when microglial activation occurs after acute brain injuries where selective neuronal vulnerability plays a role, such as global cerebral ischemia. After four-vessel occlusion in the rat, CA1 neurons in the hippocampus undergo delayed neuronal death, whereas immediately adjacent CA2 neurons survive, creating an abrupt transition from dying to surviving neurons and from presence to absence of activated microglia at the CA1/CA2 border (Morioka et al., 1991). If activated microglia did, indeed, cause bystander damage one would expect to see evidence of microglial neurotoxicity in the CA2 region, which is not the case. Moreover, if delayed death of CA1 neurons is prevented by administration of the anti-excitotoxic agent, MK-801, microglial activation is attenuated showing that preventing neuronal damage results in reduced microglial activation (Streit et al., 1992). Thus, microglial activation after acute CNS injury is primarily a reactive and adaptive glial cell response, which is triggered by injured neurons and which is designed to ameliorate primary tissue damage and to promote subsequent repair.
MICROGLIAL ACTIVATION, NEUROTOXICITY, AND DYSFUNCTION
The last 10 years have seen plenty of discussion on whether activated microglia are beneficial or harmful to neurons, and I will not belabor this, almost historic, controversy. Suffice it to say that much of the debate has arisen from the fact that studies in cell culture have demonstrated ambivalent (both neuroprotective and neurotoxic) effects of microglia and/or microglia-conditioned media on cultured neurons, whereas in vivo studies support primarily neuroprotective and proregenerative roles, as discussed above. As I have pointed out before, the dispute can be resolved by accepting the idea that, in principle, microglia are capable of performing both neuroprotective and neurodestructive functions (Streit et al., 1999). Assuming that the primary purpose of microglial existence, like that of other glial cells, is to support neuronal function, the challenge now is to determine what sort of pathological scenarios could transform microglia into autoaggressive effector cells that attack healthy neurons and cause neurodegeneration. Clearly, this type of indiscriminate microglial neurotoxicity (also known as bystander damage) is unlikely to occur to any significant extent after acute injury, such as trauma or stroke, because it would be counterproductive to post-injury tissue repair. Instead, what could be a more likely scenario in acute injury situations is a process perhaps best described as facilitative neurotoxicity, a kind of cellular euthanasia. This refers to microglia-mediated elimination of neurons that have been compromised beyond viability and functionality by the primary traumatic or ischemic event. Although facilitative microglial neurotoxicity, like bystander damage, remains to be demonstrated conclusively by suitable in vivo experiments, it makes sense pathophysiologically, biologically, and teleologically. Continued presence of severely damaged neurons after injury would hardly be of benefit toward promoting functional neurological recovery and, since microglia are known to phagocytose dead cells, it is perhaps not unreasonable to propose that microglia take on an active role in speeding up the demise of neurons that are destined to die anyway. As discussed elsewhere, microglial facilitation of cell death is likely to play an important role also in the elimination of superfluous cells during CNS development (Streit, 2001).
In contrast to acute CNS injuries, chronic CNS disease is characterized by slowly progressive neurodegeneration that may take decades to develop. Alzheimer's disease (AD) represents the classic example of a slowly progressive neurodegenerative disease, and the hypothesis that neurodegeneration in AD is caused through bystander damage from autoaggressive microglial cells that produce neurotoxins in response to continued amyloid (Aβ) exposure has received widespread support (Akiyama et al., 2000; McGeer and McGeer, 2001). However, the bystander damage hypothesis claiming microglial autotoxicity leaves a number of unanswered issues, perhaps most significantly, that there is no clear-cut correlation between presence of Aβ and cognitive dysfunction; that is, significant deposits of Aβ can be found in nondemented individuals (Dickson et al., 1992). The bystander damage hypothesis also does not explain why Aβ accumulates in the first place or why the localization of neurodegenerative changes (neurofibrillary tangles) does not correlate with the localization of activated microglia, i.e., neurofibrillary tangles are not limited to the vicinity of senile plaques, where activated microglia are most prominent. To resolve these unanswered issues, I would like to offer an alternative way of looking at the role of microglia in AD pathogenesis, namely, that microglia become senescent and/or dysfunctional with normal aging and that such aging-related deterioration of microglia is aggravated by genetic and epigenetic risk factors that cause some individuals to progress to AD. To date, the concept of diseased microglial cells has received very little attention and virtually nothing is known about specific pathogenetic mechanisms that may underlie glial cell diseases in general, and microglial diseases specifically. Thus, in the absence of direct experimental evidence of glial cell dysfunction, the concept of diseased glial cells currently is primarily of theoretical merit, in that it could help better explain the pathogenesis of neurodegenerative disease. The apparent line of thinking would be that if microglia became disabled gradually, their functional capacity to support neurons would become diminished, and hence neurons would slowly degenerate. This idea may gain increasing importance and acceptance in the future, since microglia do show considerable ability in terms of producing trophic factors and extracellular matrix molecules that are essential for sustaining neuronal function (Chamak et al., 1994; Rabchevsky and Streit, 1997; Nakajima and Kohsaka, 2002). There are at least a couple of well-known disease states in which microglial pathology could play a decisive role. Human immunodeficiency virus (HIV), with its dual tropism for the immune system and the brain, is known to target primarily microglial cells in the CNS; this is because microglia share some of the same surface receptors with peripheral lymphocytes, such as CD4 and chemokine receptors (see Garden, 2002, this issue). Persistent infection of microglia with HIV could, over time, result in depleting and/or disabling microglia, just as lymphocytes are depleted and disabled peripherally. The obvious consequence would be a loss of endogenous CNS immune function and loss of microglial support for neurons, leading to opportunistic CNS infections and neurodegeneration/dementia. A second, somewhat less apparent, group of neurodegenerative syndromes in which microglial pathology may play a role are the prion diseases. Here, structural abnormalities in microglial cell shape, such as atypical, tortuous processes and intracytoplasmic vacuoles, suggest that microglia may be diseased (von Eitzen et al., 1998). It is thus conceivable that the neurodegenerative changes in prion diseases may occur as a consequence of and/or concomitantly with microglial dysfunction. Interestingly, structural abnormalities in microglia can also be observed in Alzheimer's disease (AD), as well as in the normally aged brain (Fig. 1). In AD, clustering of microglia around extracellular deposits of Aβ within senile plaques is perhaps the most readily apparent abnormality and this has been reported in numerous publications (Dickson et al., 1988; Itagaki et al., 1989; Perlmutter et al., 1990). Generally speaking, whenever microglial clustering occurs, it is an indication of ongoing pathology. It is seen in at least two well-known pathological scenarios: (1) during acute neuronal necrosis, when microglia form phagocytic clusters (Streit and Kreutzberg, 1988); and (2) during chronic neurodegenerative disease accompanied by the presence of extracellular deposits of Aβ. In age-matched “clean” human controls, which are free of Aβ deposits, microglial clustering does not occur. In addition to the clustering of microglia around senile plaques, formation of microglial spheroids can be observed. Microglial spheroids are bulbous swellings on cytoplasmic processes. Spheroids can be observed in microglia clustered within senile plaques, as well as in microglia that are not (Fig. 1). They occur in both “clean” control and in AD brain, suggesting that these structural abnormalities may be age-related. Thus, microglial spheroids may be indicative of senescent microglial cells. Other structural abnormalities of microglia include fragmentation of cytoplasmic processes, as well as formation of unusually long, stringy processes. Collectively, the observed irregularities in microglial cell shape support the possibility that structural deterioration of microglial cells may be accompanied by a loss of cell function with aging and, especially with AD, in which these morphological abnormalities appear to be more pronounced. It is tempting to conclude that a loss or deterioration of microglial cell function could contribute to the slow degeneration of neurons over time. Microglial dysfunction may become manifest in a number of ways, including a decreased ability to clear away amyloid-β protein (Aβ), decreased ability to produce neurotrophic factors, as well as increased neurotoxicity (Fig. 2).
MICROGLIA: PAST, PRESENT, AND FUTURE
Twenty years ago, many neuroscientists were unaware or uncertain about the very existence of microglial cells despite the fact that early neuroanatomists, like Nissl, Cajal, and del Rio-Hortega, had recognized them and, in del Rio-Hortega's case, provided very detailed descriptions of this glial cell type. As a fledgling graduate student during the early 1980s, I once showed my lectin-stained slides to a famous neuroanatomist visiting our institution and asked him, “What are these little cells?” He shrugged and said, “I don't know, but do find out.” This was the first time I encountered, and started to think about, microglial cells, and I was fortunate to become part of an ensuing renaissance of microglial research that continues on today. Interest in microglia has grown steadily since the 1980s, as evidenced by an exploding microglia literature that includes this second special issue of Glia devoted to microglia, and it is clear that microglia have moved to the forefront of neuroscience research. Issues that once dominated the field, such as the search for histological markers and the question of the origin of microglia, are now largely resolved, and other topics and priorities have taken over. Many neuroscientists now recognize the tremendous power bestowed in microglia, with their amazing ability to produce a battery of enzymes, cytokines, chemokines, growth factors, and other products that can decisively influence whether neurons live or die. Signal transduction mechanisms that regulate the production of these important molecules are currently being worked out, and it may become possible in the not-too-distant future to use this knowledge to manipulate microglial cell function for the treatment of neurological disease. That microglia are critically important for some of the most important neurological problems, such as trauma, stroke, and neurodegenerative diseases, is clear from this and many other publications, and perhaps we should be optimistic that continued microglial research will ultimately help improve public health.
One final comment relates to the notion that portrays activated microglia solely as autoaggressive villains attempting to destroy the CNS. I have great difficulty accepting this idea because if microglia became truly autoaggressive, there would be no neurons left in the brain. Of course, a counterargument could be that there is indeed a substantial loss of neurons in AD. However, this loss occurs gradually over many years, and there is no causal link between neurodegenerative changes and the prior presence of activated microglia, which, incidentally, are known to accumulate also in the normally aging brain (Streit and Sparks, 1997; Sheng et al., 1998). I believe it is reasonable to view microglial neurotoxicity as a process that, on one hand, is beneficial in terms of it aiding in the removal of nonviable cells after acute injury and during development, and on the other, is a byproduct of microglial senescence and/or microglial disease. Together with an impairment of the neuron-supporting functions of microglia, microglial neurotoxicity may be a contributing factor in the development of neurodegeneration. This hypothesis of microglial dysfunction represents a relatively new thought that is just beginning to make its way into the literature. I believe it could change the way pathogenesis of neurodegenerative diseases is viewed in the future, namely, that neurodegeneration does not result primarily from aggression, but from neglect.
The author thanks Dr. Joseph Rogers for providing human brain tissue specimens and Alli Stanton for staining sections shown in Figure 1. I also thank Dr. Gerry Shaw for providing the α-internexin antibody.