The NeuroImmune System in Alzheimer’s Disease: The Glass is Half Full

Abstract

It is well established that microglia, the neuroimmune cells of the brain, are associated with amyloid-β (Aβ) deposits in Alzheimer’s disease (AD). However, the roles of these cells and other mononuclear phagocytes such as monocytes and macrophages in AD pathogenesis and progression have been elusive. Clues to mononuclear phagocyte involvement came with the demonstration that Aβ directly activates microglia and monocytes to produce neurotoxins, signifying that a receptor mediated interaction of Aβ with these cells may be critical for neurodegeneration seen in AD. Also, in AD brain, mononuclear phagocyte distribution changes from a uniform pattern that covers the brain parenchyma to distinct clusters intimately associated with areas of Aβ deposition, but the driving force behind this choreography was unclear. Here, we review our recent work identifying mononuclear phagocyte receptors for Aβ and unraveling mechanisms of recruitment of these cells to areas of Aβ deposition. While our findings and those of others have added significantly to our understanding of the role of the neuroimmune system in AD, the glass remains half full (or half empty) and a lot remains to be uncovered.

There is little doubt that Alzheimer’s disease (AD) involves the innate neuroimmune system. The cellular elements of the innate neuroimmune system that are likely to be involved are the mononuclear phagocytes, more specifically, the microglia, perivascular macrophages, and circulating monocytes.

THE CELLS

Microglia are the principal innate immune cells of the brain. They constitute about 5–12% of the total number of brain cells depending on the region studied [1]. For decades, microglia were grouped with other glial cells under the term mesoglia and were described as “capable of acting as phagocytes” but were not considered capable of “taking part in the formative processes of repair in the central nervous system” [2]. It was not until Del Rio-Hortega in 1932 that microglia were identified as separate cells distinct from other neuroglia. Del Rio-Hortega emphasized their phagocytic role “related to the elimination of substances resulting from metabolism of neuronal breakdown” and established their role in inflammatory and necrotizing processes, therefore providing the foundation for modern studies of these cells [3]. More recently, in vivo imaging with two photon microscopy of mouse brains showed that these cells constantly and actively survey their environment providing direct evidence for their dynamic role in the central nervous system (CNS) [4]. The origin of resident microglia remains a matter of intense investigation. It is almost universally accepted that they are derived from myeloid precursors. Microglia migrate into the CNS during embryonic stages, but a subset of these cells move into the brain in the post-natal period [5, 6]. Bone marrow transplantation (BMT) experiments done in the past two decades suggested that circulating peripheral blood monocytes can migrate into the adult brain and become microglia-like under non-inflammatory conditions [7]. However, doubt has been raised recently as to whether migration of monocytes into the brain in the absence of brain pathology following BMT is an artifact of irradiation, a required step necessary for bone marrow ablation before transplantation of new marrow [8]. Whether derived from monocytes or not, microglia perform macrophage functions “par excellence”. They can express all known phagocytic receptors [9–11] and are capable of phagocytosis of invading microorganisms [12], and apoptotic cells in the CNS [13]. Microglia are antigen presenting cells [14], and they respond to usual macrophage stimuli much as macrophages do with chemokine and cytokine production [9]. The second type of neuroimmune cells, perivascular macrophages, are as their name implies macrophages that surround blood vessels [15]. Perivascular macrophages differ from resident parenchymal microglia by their expression of several markers including the scavenger receptor CD163 [16, 17]. Based on BMT experiments, it is believed that perivascular macrophages are derived from circulating blood monocytes and undergo complete turnover every ~3 months [18, 19]. Perivascular macrophages can perform all the known functions of macrophages listed above. The third cell type in the innate neuroimmune system is the circulating blood monocyte. While it is not clear how often circulating blood monocytes enter the CNS under resting and non inflammatory conditions, under conditions that disrupt the blood brain barrier, and when appropriately stimulated, circulating monocytes cross the blood brain barrier and differentiate into microglia-like cells or perivascular macrophages morphologically and phenotypically. Whether they can shuttle back and forth between the perivascular space and the blood is not clear.

THE NEUROIMMUNE SYSTEM IN AD: THE EARLY PHASE

It is well established that microglia are a cellular component of the senile plaque in AD. While microglial distribution appears uniform in the normal brain, they become 2–5 fold more concentrated around plaques in AD patients and mouse models of this disease [20]. For this reason, most of the work done on microglia in AD accepts the amyloid hypothesis as the working model for this disease. In AD, microglia express major histocompatibility complex II and markers of inflammation [21, 22]. While we have definitive evidence for the accumulation of microglia around amyloid-β (Aβ) deposits, and for the phenotypic changes they undergo, when we became interested in this topic, little was known about the role of microglia in AD.

A breakthrough in identifying a role for microglia in AD occurred in 1995. A group led by Filippo Rossi in Verona showed that Aβ activates microglia to produce reactive nitrogen species and tumor necrosis factor-α (TNFα) [23]. The findings by Rossi and colleagues suggested that the interactions between Aβ and microglia may be critical for disease pathogenesis (at least for neurodegeneration). These findings also suggested that microglial-Aβ interactions are likely receptor mediated events. We were both postdocs at that time in Sam Silverstein’s group at Columbia University, and together with John Loike we asked the important question: what are the receptors involved in the interaction of Aβ with microglia? If we could find answers to this question, perhaps the nature of the role of microglia in AD would be revealed. We also asked a second question: since the concentration of microglia appears to increase around sites of Aβ deposition, what is the mechanism of accumulation of these cells? Since then, we have been attempting to answer these two questions and our work proceeded down these two parallel and eventually intersecting paths.

IDENTIFICATION OF THE FIRST RECEPTORS FOR Aβ

In the mid 1990 s, we started working on a family of receptors expressed on macrophages called scavenger receptors. These receptors are structurally heterogeneous but all share the ability to bind oxidized low density lipoproteins (ox-LDL) and are therefore believed to play a major role in the pathogenesis of atherosclerosis. We thought that Scavenger receptor A1 (Scara1) is a good candidate for an Aβ receptor for three reasons. First, binding of Scara1 to its ligands does not require divalent cations [24, 25]. Interestingly, binding of Aβ to microglia is also divalent cations independent. Second, Scara1 has a large extracellular collagen-like domain which is believed to be the ligand binding domain. Similarly, the complement protein C1q also has a collagen-like domain and work by Andrea Tenner’s group showed that this C1q collagen-like domain binds Aβ [26]. Third, Bradley Hyman and his group had found that Scara1 was expressed on microglia in AD [27]. For all these reasons, we tested if Scara1 can indeed bind Aβ and we went on to identify this receptor as the first known microglial receptor for Aβ [28]. We also found that Aβ binds to the collagen-like domain of Scara1 [28].

COULD OTHER SCAVENGER RECEPTORS BE INVOLVED?

Scara1 is not the only scavenger receptor expressed on mononuclear phagocytes. CD36, a class B scavenger receptor, is also expressed on these cells and Scara1 and CD36 share several ligands and mediate the interactions of the cells with these ligands in a non-redundant manner. Indeed, while Scara1 promotes binding and uptake of ox-LDL, CD36 promotes ox-LDL-induced activation of the cells and production of chemokines, cytokines, and reactive oxygen and nitrogen species [29]. We speculated that this may also be the case for Scara1 and CD36 binding to Aβ. Pursuing this line of investigation, we found that Scara1 is required for phagocytosis of Aβ by mononuclear phagocytes, but is not involved in activation. On the other hand, CD36 is required for activation and production of chemotactic factors and neurotoxins by microglia in response to Aβ, but is not necessary for phagocytosis of Aβ [9, 30].

The next question was whether CD36 alone was enough to promote activation or were other co-receptors involved? We had found that activation of macrophages by other ligands of CD36 such as the fungal pathogens Cryptococcus neoformans required the presence of Toll-like receptors (TLRs) [31]. Applying our observations to Aβ activation of microglia, in collaboration with Kathryn Moore at Massachusetts General Hospital, we found a unique Aβ receptor complex comprising CD36 and TLR4 and TLR6 [32]. Expression of each individual component of this receptor complex alone is not enough to mediate Aβ-induced activation of microglia. All three receptors have to be present to induce NFκB translocation and production of reactive oxygen species [32]. The need for co-receptors for TLRs is now becoming a recognized paradigm in immunology. Our working hypothesis is that TLRs alone cannot hold on to their ligands and require co-receptors such as CD36 to help “dock” these ligands and make them available to bind to the TLRs. In AD, this pathway is crucial for mediating Aβ-induced, microglia-mediated production of neurotoxins and subsequent neurodegeneration [32]. It is conceivable that blocking this pathway pharmacologically could be used as a therapeutic modality for AD, and we have recently embarked on studies to identify reagents that can be used for this purpose [33].

ACCUMULATION OF MONONUCLEAR PHAGOCYTES AT SITES OF Aβ DEPOSITION

In addition to trying to understand how microglia and other neuroimmune cells interact with Aβ, we have dedicated a significant effort in the past few years to understand how these cells accumulate at sites of Aβ deposition. Areas of Aβ depositions in the brain are studded with microglia and their density in these areas is 2–5 times higher than in neighboring areas of the brain [20]. We hypothesized that if we understand how neuroimmune cells accumulate in sites of Aβ deposition, then blocking such accumulation may give us insight into the role of these cells in the pathogenesis of this disease.

NEUROIMMUNE CELLS TRAFFIC TO THE BRAIN IN SEVERAL INFLAMMATORY CONDITIONS

In addition to microglia, monocytes and macrophages are important components of the neuroinflammatory response associated with many neurological disorders. For the sake of simplicity, but also reflecting the lack of complete understanding of how to clearly differentiate between the various neuroimmune cells, we will refer to these cells as a group by mononuclear phagocytes. Several infectious diseases of the CNS provide compelling examples that show that peripheral myeloid mononuclear cells can accumulate in the brain. In scrapie, a model of prion diseases, early and rapid engraftment of bone marrow-derived IBA + cells was observed [7]. Similarly, in models of bacterial meningitis, monocytes accumulate and contribute to the pool of microglia [34]. In some bacterial, fungal, and parasitic infections of the CNS, circulating monocytes act as cellular “Trojan horses” carrying intracellular organisms across the blood brain barrier thereby allowing translocation of these organisms into the brain and facilitating intracerebral infection [35–37]. In addition to CNS infections, peripheral monocytes can migrate into the CNS and initiate primary demyelination [38] and contribute to the inflammatory response in the MPTP mouse model of Parkinson’s disease [39]. A common denominator for all these conditions is the role of chemokines.

Indeed, chemokines are chemotactic cytokines that form a large family (50 members) of secreted and membrane bound, 8–10 kDa proteins that induce the recruitment of leukocytes to sites of acute and chronic inflammation [40]. Chemokines bind to specific G protein-coupled seven-transmembrane cell surface receptors on target cells [40]. Recent work from our lab using direct RNA sequencing confirmed that microglia and monocytes express a wide repertoire of chemokine receptors (Hickman and El Khoury, unpublished data). An important chemokine that has been implicated in the recruitment of monocytes in infectious and other inflammatory conditions is CCL2. CCL2 binds to its receptor CCR2 and promotes accumulation of these cells in acute and chronic inflammatory conditions such as peritonitis [40], atherosclerosis [41], and experimental autoimmune encephalomyelitis, a murine model of multiple sclerosis [42].

We decided to investigate the role of CCL2-CCR2 in accumulation of mononuclear phagocytes at sites of Aβ deposition in AD for several reasons. First, CCL2 is expressed in reactive microglia in senile plaques and in some neurons and astrocytes [43, 44]. Indeed, based on an immunohistochemical study of AD brains, logistic linear regression modeling determined that CCL2 was the most reliable predictor of disease [44]. Second, CCL2 is also induced by Aβ in microglia and monocytes and it induces chemotaxis of these cells in vitro [9]. In addition, we found that CCL2 is a potent chemoattractant for human fetal microglia, suggesting that these cells express CCR2. Based on the above, we hypothesized that CCL2/CCR2 interactions play a key role in recruitment and/or activation of mononuclear phagocytes to sites of Aβ deposition in AD.

CCR2 IS REQUIRED FOR THE ACCUMULATION OF MONONUCLEAR PHAGOCYTES IN APP TG2576 MICE BEFORE FORMATION OF VISIBLE Aβ DEPOSITS

To test the role of CCR2 in mononuclear phagocyte accumulation in an AD mouse model, we bred CCR2 null mice with transgenic mice expressing the human amyloid-β protein precursor with the Swedish mutation (APP Tg2576) and analyzed the resulting APP-CCR2^−/− mice for AD-like pathology [45]. We found that CCR2 deficiency significantly reduced the number of CD11b positive cells that accumulated in the brains of APP mice early in the disease process, before formation of senile plaques. CCR2 deficiency also abolished the accumulation of mononuclear phagocytes at sites of intracerebral injection of Aβ. Analysis of the phenotype of these recruited cells by flow cytometry, showed that they express high levels of CD45 in addition to CD11b, suggesting they are monocytes. This reduction in the number of mononuclear phagocytes was associated with increased mortality and higher Aβ levels in the brain, suggesting that early mononuclear phagocyte accumulation promotes the clearance of Aβ and protects mice from Aβ toxicity early in the disease process [45]. Analysis of the sites of Aβ deposition in APP-CCR2^−/− mice showed that such deposition occurred exclusively around blood vessels and no parenchymal Aβ deposits were observed (Fig. 1). Our data indicate that CCR2 plays a non-redundant role in the pathogenesis of AD and support the hypothesis that early accumulation of monocytes is protective and promotes Aβ clearance.

Fig. 1.

Increased brain Aβ levels in CCR2-deficient AβPP mice is associated with perivascular Aβ deposition. A) Aβ₄₂ levels were measured by ELISA (each data point is mean ± SEM, n = 3–6 per group, *p < 0.004). b) Immunohistochemistry with anti-Aβ and control antibodies showing perivascular Aβ deposition (arrowheads) in 65 day old APP-CCR2^−/− mice.

The absence of parenchymal Aβ deposition in young APP-CCR2^−/− mice and increased perivascular Aβ deposition could be interpreted in two different ways. On one hand it is possible that the mice did not have a chance to develop visible parenchymal Aβ deposits since they died at an early age before being able to do so [45]. It is also possible that the CCR2-dependent recruitment of “protective” monocytes is only relevant when Aβ is deposited around blood vessels; such perivascular deposits are found in a high percentage of AD patients and are a hallmark of cerebral amyloid angiopathy. This raises the possibility that accumulation of microglia at parenchymal sites of Aβ deposition involves a separate process that does not require CCR2, unlike what happens at perivascular sites of Aβ deposition. The latter argument is indirectly supported by a paper showing that selective depletion of perivascular macrophages exacerbates perivascular Aβ deposition, and stimulation of their turnover reduces perivascular load [46]. The distinction becomes less critical if, as proposed by Kumar-Singh and colleagues [47], all Aβ deposits develop initially around blood vessels then these vessels are obliterated leading to formation of the classic plaques.

IF MONONUCLEAR PHAGOCYTES CLEAR Aβ IN THE BRAIN, HOW CAN WE EXPLAIN PERSISTENT Aβ DEPOSITION IN SPITE OF THE LARGE NUMBER OF THESE CELLS AROUND PLAQUES?

The data supporting a role for mononuclear phagocytes in Aβ clearance are compelling, but these data also raise an important question. Why does Aβ continue to accumulate, and why does AD pathology progress despite continued mononuclear phagocyte recruitment? One possible explanation for the failure of mononuclear phagocytes to stop AD progression would be that these cells become overwhelmed by the excess amount of Aβ produced and cannot keep up with the pace of Aβ generation. Another possibility would be that, as AD progresses, the phenotype of accumulating mononuclear phagocytes changes and these cells become more proinflammatory and lose their Aβ-clearing capabilities, resulting in reduced Aβ uptake and degradation, and increased Aβ accumulation. To investigate this hypothesis, we developed a method to isolate fresh adult mouse microglia from the bigenic APPswe/PSEN1dE9 (PS1-APP) mice [48, 49], a robust mouse model of Aβ accumulation and microgliosis, and from their non-transgenic littermates at various ages and stages of AD-like pathology, and compared gene expression of Aβ- binding receptors and Aβ-degrading enzymes [50]. Our data show that, as PS1-APP mice age, their microglia become dysfunctional and exhibit a significant reduction in expression of their Aβ-binding receptors (Fig. 2) and Aβ-degrading enzymes [50], but maintain their ability to produce proinflammatory cytokines [50]. Therefore, in contrast to their protective role early in the disease process, as AD-like pathology progresses, a reduction in the expression of Aβ-receptors and Aβ-degrading enzymes indicates that microglia further contribute to disease progression. In support of this possibility, Fiala and colleagues found that monocytes and macrophages from AD patients were less effective in Aβ phagocytosis compared with monocytes and macrophages from age-matched control non-AD patients [51].

Fig. 2.

Reduced expression of Scara1 Aβ-binding receptor in microglia from old transgenic PS1-APP mice. Expression of Scara1 in CD11b⁺ cells was compared between transgenic PS1-APP mice and their age-matched WT littermates at 1.5, 3, 8, and 14 months of age. At 8 and 14 months of age, PS1-APP transgenic mice show significantly reduced expression of Scara1 compared with their WT littermates. (*p < 0.05).

Our data also suggest that failure of microglia to adequately clear Aβ may be a direct result of the Aβ-induced inflammatory response. TNFα, a major cytokine produced by microglia in response to Aβ stimulation, reduced expression of the Aβ receptors, similar to the reduction observed in microglia from aging PS1-APP mice [50]. Although the mechanism behind decreased expression of the degrading enzymes is not clear, upregulation of reactive oxygen species and a number of proinflammatory cytokines has been observed in transgenic mice with AD-like pathology. It is possible that one or more of these cytokines is responsible for downregulation of Aβ degrading enzymes. Interestingly, in addition to downregulating Aβ clearance pathways, proinflammatory cytokines may also contribute to Aβ accumulation by another mechanism. Recently, it was shown that TNFα and interferon-γ upregulate β-secretase (BACE1), an enzyme involved in Aβ production [52]. TNFα, IL-1β, and interferon-γ were also found to stimulate γ-secretase-mediated cleavage of AβPP [53]. Proinflammatory cytokines therefore may also contribute to AD-like pathology by promoting Aβ generation.

CONCLUSIONS

While our findings discussed above and those of many others have added significantly to our understanding of the role of the neuroimmune system in AD, the glass remains half full (or half empty) and a lot remains to be uncovered. What we can say is that our data support the model that the inflammatory response in AD is a “double-edged sword” and that mononuclear phagocytes appear to play a dichotomous role in AD. Microglia, which are constantly surveying their environment, sense extracellular Aβ after it is released from neurons and are able to phagocytose it, degrade it, and clear it and are initially neuroprotective. They also secrete chemokines to recruit cells from the periphery to aid in Aβ removal. However, as AD progresses, these cells either become overwhelmed, or their continued interaction with Aβ induces a phenotypic switch leading to increased Aβ production, inadequate Aβ clearance and continued production of neurotoxic inflammatory products thereby contributing to neurodegeneration. Any anti-inflammatory approach for treatment of AD should be designed to differentiate between these dichotomous functions of mononuclear phagocytes. Such therapy should sustain their ability to clear Aβ, while lessening their ability to produce proinflammatory cytokines.