Special Reviews

Immunotherapeutic Approaches to Alzheimer's Disease

See allHide authors and affiliations

Science  31 Oct 2003:
Vol. 302, Issue 5646, pp. 834-838
DOI: 10.1126/science.1088469

This article has a correction. Please see:


Although neurodegenerative diseases such as Alzheimer's disease are not classically considered mediated by inflammation or the immune system, in some instances the immune system may play an important role in the degenerative process. Furthermore, it has become clear that the immune system itself may have beneficial effects in nervous system diseases considered neurodegenerative. Immunotherapeutic approaches designed to induce a humoral immune response have recently been developed for the treatment of Alzheimer's disease. These studies have led to human trials that resulted in both beneficial and adverse effects. In animal models, it has also been shown that immunotherapy designed to induce a cellular immune response may be of benefit in central nervous system injury, although T cells may have either a beneficial or detrimental effect depending on the type of T cell response induced. These areas provide a new avenue for exploring immune system–based therapy of neurodegenerative diseases and will be discussed here with a primary focus on Alzheimer's disease. We will also discuss how these approaches affect microglia activation, which plays a key role in therapy of such diseases.

Amyloid beta-peptide (Aβ) is a cleavage product of neuronal amyloid precursor protein (APP) (1, 2). Cleavage can yield either Aβ1-40 or Aβ1-42, and accumulation and aggregation of these products in cognitive brain regions during aging is a hallmark of Alzheimer's disease (AD) pathology (3). Aβ1-42 is the more aggregated form and is more highly correlated than Aβ1-40 with disease and neurotoxicity. Extensive studies have demonstrated that these Aβ plaques are co-localized with activated microglia and astrocytes, implicating additional neurotoxicity (46). The accumulation of Aβ in the brain may be part of the aging process in all individuals, and its accumulation to toxic levels may take years. The appearance of clinical symptoms may be due to increased levels of Aβ above a certain threshold that is no longer controlled by endogenous clearance mechanisms. This hypothesis is supported by the observation that genetic factors such as APP or presenilin mutations or the carrying of the ApoE4 allele results in earlier accumulation of Aβ and early onset of clinical symptoms (3, 7). Nonetheless, it is still unclear what are the crucial factors in aging that trigger the accumulation and toxicity of Aβ. Therapeutic approaches include targeting Aβ either by affecting its production (e.g., secretase inhibitors) or by using immune based strategies to enhance Aβ clearance (e.g., Aβ immunization, astrocytic or microglial activation).

Innate Immune Responses and AD

A prominent innate immune response occurs in the central nervous system (CNS) in association with Aβ deposition and plaque formation (Fig. 1). This innate immune response includes the activation of complement, secretion of proinflammatory cytokines such as interleukin (IL)–1β and tumor necrosis factor (TNF)–α; expression of the chemokines MIP-1α, MIP-1β, and MCP-1; and the secretion of nitric oxide (NO) (5, 812). Recent studies in transgenic mice that overexpress an AD-causing mutant form of human APP and develop amyloid deposits have revealed that crossing such mice with mice overexpressing a natural inhibitor of complement C3 results in a worsening of Aβ plaque load and more neuronal loss (13). This result suggests that the innate immune response found in AD and mouse models, including activation of the classical complement cascade, may in part represent a beneficial response (13). In contrast, dampening the innate immune response may also be beneficial, as a number of clinical studies suggest that anti-inflammatory drugs such as those used in arthritis may delay or slow the progression of AD (6), though these anti-inflammatory drugs may also work by targeting γ-secretase (14). Thus, it appears that Aβ or its fibrillated form is recognized in the CNS as a molecule that needs to be cleared and provokes activation of microglia and astrocytes. If microglial or astrocytic activation fails to clear the toxic forms of Aβ, the innate immune response becomes chronic and neurotoxic. On the basis of these observations, microglia or astrocytes can be modulated in two opposing ways to yield beneficial effects. First, one may downregulate their chronic activation and resultant injurious inflammatory response by treating with anti-inflammatory drugs. Second, one may activate them such that they clear Aβ more effectively.

Fig. 1.

Pathways of microglia activation in Alzheimer's diseases. Microglia are bone marrow–derived cells that acquire ramified morphology in the intact CNS. In response to Aβ deposition in AD, microglial cells are activated and differentiate into phagocytic cells (CD11b+) (left), which induce a proinflammatory environment and secrete IL-1β, TNF-γ, NO, free radicals, chemokines, and activate complement. The NO secreted by CD11b+ cells may enhance T cell apoptosis in the CNS. A second pathway for microglial cells is to differentiate into APCs (right), which are induced in the presence of GM-CSF and/or IFN-γ secreted by microglia, astrocytes, or other immune cells (T cells, macrophages) that infiltrate the CNS. As a result, microglia cells differentiate to dendritic-like cells that then may function as APCs for both TH1 and TH2 cells. These cells also may migrate from the CNS to secondary lymph nodes and induce T cell activation. In AD, this pathway could suppress the toxic innate immune response and can be enhanced by TH2 immunization.

Microglia as Antigen-Presenting Cells

Although Aβ-mediated innate activation of microglia has been studied extensively, the function of microglia as antigen-presenting cells (APCs) for adaptive immune responses [Aβ presentation to T cells in the context of major histocompatiblity complex (MHC)] has not been characterized in AD. The professional APCs of the CNS are the parenchymal microglia and the perivascular cells (15, 16), both of which are of myeloid origin and migrate to the CNS during development (17). The parenchymal microglia have recently been characterized as myeloid progenitor cells that can differentiate into macrophage-like cells expressing CD11b (integrin αM chain) or dendritic-like cells expressing CD11c (integrin αx chain) if stimulated with macrophage–colony-stimulating factor (M-CSF) or granulocyte-macrophage–colony-stimulating factor (GM-CSF), respectively (18). In another study, CD11c+ cells were isolated from inflamed and normal brain and supported the proliferation of allogeneic T cells that have characteristics different from allogeneic T cells stimulated by CD11b+ cells (19). It has been suggested that myeloid-derived microglia progenitor cells might differentiate into immature dendritic cells (DCs) in the presence of GM-CSF or sequentially to fully mature DCs with CD40 ligation and that they differentiate into macrophages in the absence of GM-CSF (19). Nonetheless, it is still unclear whether certain microglia progenitors in the CNS differentiate more effectively to CD11c+ cells and whether they migrate to secondary lymph nodes to induce brain-specific T cell activation.

It has become clear that microglia can serve as APCs for Aβ-reactive T cells and, in turn, T cells themselves can influence microglial differentiation (20, 21). This “crosstalk” between microglia and T cells has important implications for both immunopathology and immunotherapy in AD. We have recently found that both CD11b+ and CD11c+ cells are localized with Aβ plaques in APP transgenic (Tg) mice. Although these cells express the costimulation molecule CD86, they only express low levels of MHC II, which suggests that their function as APCs is limited (22). The increased levels of GM-CSF that have been reported in AD (23) may contribute to differentiation of microglia into “immature” DC-like cells, but their full maturation may require γ-interferon (IFN-γ) (Fig. 1) (24, 25). If microglia are pretreated with IFN-γ, they differentiate into CD11b+ and CD11c+ cells and serve as Aβ APCs for both Aβ1-40 and Aβ1-42 as well as support CD86-dependent proliferation of Aβ-reactive T cells (26).

Aβ stimulation of microglia and astrocytes leads to increased production of NO, which is toxic for neurons. We have recently found that Aβ-stimulated microglia can also be toxic to T cells via an NO-mediated pathway (26). This NO-mediated toxicity is enhanced in vitro by IFN-γ–producing T helper cells (TH1) and down-regulated by IL-4–, IL-10–, and TGF-β–producing TH2 and TH3 cells (2628). Nonetheless, increased levels of NO in the brain of patients with AD may effectively induce apoptosis of infiltrating T cells and prevent differentiation of microglia to effective APCs.

Microglia play a central role in CNS immune responses in Alzheimer's disease. Type of cross-talk that occurs between them and T cells can have an impact on the neurotoxicity of Aβ mediated by stimulation of the innate immune response (Fig. 1).

Aβ-Reactive T cells

The adaptive immune system can be broadly classified into cellular and humoral (antibody) types of responses. Among cellular responses, different types or classes of cellular immune responses, e.g., TH1 (IFN-γ) in contrast to TH2 or TH3 (IL-4, IL-10, TGF-β) have been identified that play a crucial role in understanding the mechanisms of inflammatory process regulation. The different classes of T cell responses have important implications not only for microglia activation and microglial–T cell cross-talk but also for the attempts to develop a vaccination strategy for Alzheimer's disease.

Aβ may be considered a self-antigen. Self-reactive T cells of low to moderate binding affinity are not necessarily deleted during negative selection in the thymus (29, 30), and some self-reactive T cells are positively selected and maintained in the normal immune repertoire (31). Although such self-reactive T cells can mediate autoimmune diseases such as multiple sclerosis, they may also play a beneficial physiological role in immune regulation and maintenance of normal tissues (3138). Although the CNS has been described as immunologically privileged, it is now known that activated T cells routinely penetrate the CNS (20, 39, 40); however, under inflammatory conditions, T cells undergo pronounced apoptosis in the CNS (4143).

We have recently demonstrated that cellular immune responses to Aβ occur in middle-aged and elderly healthy subjects and patients with AD (44). A significantly higher portion of healthy elderly and AD subjects had strong Aβ-reactive T cell responses than in middle-aged adults. Aβ-reactive T cells were detected in almost all individuals tested, suggesting that these cells either escape central and peripheral tolerance or are positively selected to maintain the normal T cell repertoire. The cytokine repertoire of these T cells was of TH1 (proinflammatory), TH2 (regulatory), and TH0 (TH1 and TH2) phenotypes. CD4 T cell epitopes were identified primarily in the Aβ15-42 peptide, which is segregated from the dominant B cell epitopes identified in Aβ1-15 (4546). We found that monoclonal antibody to DR inhibited virtually all T-cell lines tested. The analysis of Aβ T-cell epitopes and their restriction to human leukocyte antigen (HLA)–DR class II further demonstrate that Aβ is processed and presented by APCs in the context of MHC and that Aβ-specific T cell proliferation is mediated via MHC–T cell receptor (TCR) interactions. Thus, Aβ induces adaptive immune responses in the periphery in addition to innate immune responses in the CNS.

It is possible that the activation and expansion of Aβ-reactive T cells in the elderly and AD subjects indicate that Aβ is captured by local APCs in the brain such as CD11b or CD11c microglia in the context of Aβ deposition and that these APCs migrate to secondary lymph nodes and induce T cell activation. Although Aβ deposition occurs in elderly humans that do not have overt signs of AD, there appears to be increased T cell reactivity to Aβ in AD patients as, in contrast to elderly subjects, all AD patients tested had some Aβ reactivity. Such reactivity could reflect an endogenous reaction to Aβ deposition in the brain in the context of the local innate immune response that occurs in AD.

The expansion of Aβ-reactive T cells in the periphery may be another indication of the pathogenic role of Aβ in AD, although it is unclear whether these T cell responses affect susceptibility and course of AD. Nonetheless, these endogenous immune responses may have a beneficial role if properly boosted. Such a neuroprotective role for brain-specific T cells has recently been demonstrated in animal models of brain injury (4752) and stroke (53, 54) (Table 1).

Table 1.

Immunotherapeutic approaches in AD and other neurodegenerative CNS conditions. MOG, myelin oligodendrocyte glycoprotein; MBP, myelin basic protein; GA, glatiramer acetate; NMDA-R, N-methyl-d-aspartate receptor; PrPSc, the abnormal disease-causing isoform of prion protein.

DiseaseAntigensImmune approachesRouteMouse modelSuggested mechanism
Alzheimer's Immunization with Aβ or a B cell epitope, or passive immunization Parenteral, mucosal, or intravenous APP-Tg mice (View inline, View inline, View inline) Antibody-mediated clearance of Aβ
Brain injury MOG, MBP, GA Injection of MBP-reactive T cells, immunization with MBP, MOG, GA Parenteral or Mucosal Optic nerve and spinal cord injury (View inline, View inline, View inline) Clearance of toxic molecules, regulatory cells and cytokines, and neurotrophins or neuronal metabolic rest
Stroke MOG or MBP Oral and nasal administration of MBP and MOG Mucosal Cerebral ischemia (View inline, View inline) Regulation of immune response
Epilepsy NMDA-R Oral administration of an adeno-associated virus Mucosal Kainate-induced seizure (View inline) Blockage of glutamate toxicity
Prion Prion protein Passive immunization Parenteral Induction with scrapie brain homogenate (View inline) Reduction in PrPSc formation

Immunotherapy Approaches in AD

Although AD is associated with local innate immune responses, the induction of systemic adaptive immune responses to Aβ in mouse models of AD has been found to be beneficial for both the neuropathological and behavioral changes that these mice develop (5561). Here we describe the immunotherapy approaches taken thus far in mouse models of AD and in patients with AD.

Parenteral immunization with TH1 adjuvant. Parenteral immunization of APP Tg mice with synthetic Aβ in complete Freund's adjuvant can markedly decrease the number and density of Aβ deposits in the brain, with concomitant improvements in neuritic dystrophy and gliosis (55). The immunization procedure consisted of parenteral injection of Aβ in complete Freund's adjuvant and then boosting with Aβ in IFA (incomplete Freund's adjuvant). Neither T cell infiltrates nor brain inflammation were observed in these animal studies.

It appears that the induction of antibodies to Aβ plays the primary role in the vaccine-mediated clearance of Aβ from the brain, as passive transfer of Aβ antibodies has shown similar beneficial neuropathological effects (61). Remarkably, a single parenteral administration of a monoclonal antibody to Aβ has recently been shown to produce rapid (within hours) benefits on behavioral measures of cognitive function in a mouse model, apparently by interfering with diffusible, putatively synaptotoxic forms of Aβ (e.g., Aβ oligomers) without lowering the overall amount of Aβ deposits in the brain (60). Two broad theories about the mechanisms by which Aβ antibodies work in mice have emerged. First, evidence for an Fc-mediated uptake and clearance of Aβ-antibody complexes by local activated microglia has been obtained (61). Second, evidence of a net movement of Aβ peptide out of the brain as a result of its binding and mobilization by Aβ antibodies, in the serum and in the cerebrospinal fluid (CSF), has been provided. (62). These two proposed mechanisms are not mutually exclusive, and there may be additional ways in which antibodies decrease Aβ-mediated synaptic and neuronal dysfunction. It should be noted however, that passive immunization using monoclonal antibodies to Aβ resulted in brain hemorrhage (63).

Mucosal immunization. Positive effects have also been found after repetitive mucosal (intranasal) administration of the Aβ peptide to transgenic mice (58). Administration of Aβ intranasally to APP-Tg mice induced antibodies to Aβ and partial clearance of Aβ plaques, accompanied by infiltration of small numbers of mononuclear cells with anti-inflammatory properties, i.e., cells secreting IL-4, IL-10, and TGF-β in the CNS (58). It was recently shown that overexpression of TGF-β in the CNS of APP-Tg mice resulted in a significant reduction of Aβ plaque burden by promoting microglial clearance of the peptide (64). Adult mouse astrocytes also showed profound capabilities of degrading Aβ in vitro and in situ (65). Thus immune approaches targeted to induce Aβ antibodies and TH2 immune responses may result in activation of microglia and astrocytes with a beneficial effect on AD pathology. Further research is required to understand the effect of T cell–derived cytokines on microglia activation and their function in neurodegenerative diseases.

Human trials of Aβ vaccination. The finding that active vaccination with Aβ had profound Aβ-lowering effects in an animal model led to clinical trials in which an Aβ1-42 synthetic peptide was administered parenterally with a previously tested adjuvant (QS21) to patients with mild to moderate AD. Although a Phase I safety study with a small number of patients failed to reveal significant side effects, a subsequent Phase II trial was discontinued shortly after its initiation when approximately 5% of the treated patients developed what appeared to be a CNS inflammatory reaction. A recent postmortem case report has shown some macrophages and T cell infiltrates in the CNS after Aβ immunization (66). Such cellular reactions were not detected in mice and other mammals exposed to the vaccine during preclinical safety and efficacy testing. According to our recent finding that some elderly subjects and patients with AD have remarkably increased T cell reactivity to Aβ (44), it is possible that AD patients with high Aβ T cell reactivity develop severe T cell reactions in the CNS when immunized and boosted with Aβ emulsified in the TH1 adjuvant QS21. Nonetheless, titers of Aβ antibodies did not correlate with the occurrence of severity of symptoms or relapses (67), suggesting that the meningoencephalitis was not necessarily directly linked to Aβ-reactive T cells. Aβ-reactive T cells presumably contributed to effective production of Aβ antibodies, but it is yet not clear whether they contributed to severe brain inflammation. Therefore, it is possible that the Aβ-reactive T cells induced in the context of TH1 adjuvant mediated the first wave of T cell infiltrate to the CNS, which was then followed by T cell responses to myelin antigens caused by the inflammatory response in the brain. It is also possible that the prolonged exposure to the adjuvant QS21 induced the stimulation of pathogenic myelin-specific T cells, independent of Aβ-reactive T cells. The possibility that pathogenic myelin reactive T cells were involved is supported by magnetic resonance imaging (MRI) analysis which showed lesions in white matter of the brain where Aβ is not normally expressed.

There are several reasons why Aβ immunization of APP-Tg mice did not cause severe inflammatory responses in the CNS: (i) Aβ is not an encephalitogenic antigen; (ii) animals were boosted with IFA, which could result in increased antibody production in the context of Th2 type T cell responses; (iii) increased production of Aβ in APP-Tg mice could induce central and peripheral T cell tolerance (46); (iv) Aβ was not sufficiently immunogenetic for T cell responses in the strains tested; and (v) NO-mediated T cell apoptosis was enhanced in the CNS of APP-Tg mice (26).

Clinical evaluation after Aβ immunization showed that patients who develop effective titers of plaque-binding Aβ antibodies showed slower rates of decline of cognitive functions (68). The data observed so far in this cohort of subjects suggest that high titers of plaque-binding Aβ antibodies are important to achieve immune-mediated beneficial effects in AD. In the postmortem case, there was clear evidence of decreased amounts of Aβ plaques in neocortex regions as compared with nonimmunized patients with AD (66). In some regions that were devoid of Aβ plaques, Aβ immunoreactivity was associated with T cell infiltrates in the CNS, activated microglia, and low titer of Aβ antibodies (66). Despite the adverse effects that occurred after Aβ vaccination in humans, the results help define the path for future immunotherapy approaches.

Future Immunotherapeutic Strategies in AD

Other immunotherapeutic strategies may be used to obtain a beneficial effect without untoward side effects. First it is possible to induce Aβ antibodies with no Aβ T cell response, because Aβ1-15 contains the dominant B cell epitopes that bind to Aβ plaques and no T cell epitopes. Aβ antibodies could be induced by vaccinating with sequences in this region coupled to a carrier protein as previously shown (46, 48). Another approach is passive administration of Aβ antibodies, which, presumably, would need to be repeated periodically (Table 1).

Another approach is to vaccinate in a fashion that induces Aβ antibodies and a nonpathogenic or even beneficial T cell responses. Mucosal immunization with Aβ1-42 induces antibodies to Aβ and T cells that may have regulatory properties. Because almost all human Aβ-reactive T cell lines we studied also showed a TH2 phenotype, it is possible that mucosal immunization that preferentially induces TH2 or TH3 responses could boost this lineage and enhance clearance of Aβ by stimulating Aβ antibody production and by modulating microglial activation at sites of Aβ plaques, with a minimal risk of harmful T cell response in the CNS. This approach may also be applied using one of the T cell epitopes identified in humans (44). Furthermore, these epitopes could be modified to generate altered peptide ligands (APLs) that induce TH2 responses (69) that, upon specific interaction with microglia, enhance clearance and suppression of innate neurotoxic responses.

In conclusion, immunotherapy of AD involves both humoral and cellular limbs of the adaptive immune response and their interaction with innate immunity in terms of microglial responses within the CNS. Furthermore, the immunotherapeutic approaches described above for AD have also been applied to other neurodegenerative CNS conditions (Table 1). Despite the untoward side effects observed in recent trials of Aβ immunization, the opportunity for effective inmmunotherapy of AD will be enhanced as more is learned about the basic mechanisms of both the innate and adaptive immune responses to Aβ.


Stay Connected to Science

Navigate This Article