Introduction

Alzheimer disease (AD) is the most common age-related dementia, with a prevalence that increases exponentially with age. About 1% of the population has AD by 65 years of age, increasing to almost 50% in people over 85 years of age. A prophylactic or curative treatment for AD would, therefore, have enormous benefits. Currently approved drugs for treatment of AD mainly target symptoms, and new effective and safe treatments are needed to address this issue.

This article discusses recent clinical trials of potentially disease-modifying immunotherapy for AD, including the limited drug efficacy and associated adverse effects that led to early trial termination in some cases. We discuss the possible mechanisms of action of both endogenous and exogenous amyloid-β (Aβ)-specific autoantibodies, and use this knowledge to propose possible strategies to optimize immunotherapy for AD. The ultimate goal is to develop Aβ-targeted therapeutic antibodies that lack serious adverse effects and have increased efficacy.

Immunotherapy clinical trials

According to the amyloid cascade hypothesis, Aβ has a pivotal role in the pathogenesis of AD.1 The theory posits that accumulation of Aβ in the brain leads to formation of senile plaques and neurofibrillary tangles, and causes neurite degeneration, neuronal cell death and, ultimately, dementia. Aβ is, therefore, considered to be the most important therapeutic target for AD.

Aβ-targeted immunotherapy showed promise in animal models of AD, but human trials have had disappointing results. In clinical trials of active Aβ immunization with target antigens, or passive immunization via direct administration of Aβ-specific antibodies, Aβ levels in the brain were substantially reduced, but without improvement in cognition.2,3 Long-term follow-up of patients in a phase I–II trial who were actively immunized against Aβ with the drug AN1972 showed increased titres of antibodies against Aβ plaques, but no overall differences in cognitive performance were found between the placebo and treatment groups.4 In a phase II trial, anti-Aβ antibody improved cognition in patients with AD who were not carriers of the AD risk allele apolipoprotein E ɛ4. However, no significant differences were found in the primary efficacy analysis of data from all participants.3

The lack of efficacy of Aβ-targeted immunotherapy on cognition has led to a revision of the Aβ cascade hypothesis. Specifically, although Aβ is still thought to have a crucial role in initiating the cascade, secondary pathological processes, such as neuroinflammation and tau hyperphosphorylation, are also involved in AD pathogenesis and, once initiated, are less dependent on Aβ.5 Treatments that target Aβ should, therefore, be administered before the onset of clinical symptoms or in the mild cognitive impairment stage.6,7 This requirement might explain the failure of past immunotherapy trials, as antibodies were administered after the development of full-blown AD.

Adverse effects

An important factor that has compromised the efficacy of immunotherapy in clinical trials is the adverse effect profile of vaccines or antibodies against Aβ. Such adverse effects include meningoencephalitis and brain microhaemorrhage, which notably led to suspension of the clinical trial of AN1972.4

Innate immune response

An existing inflammatory state can impair production of antibodies in response to vaccination.8 In patients with AD, chronic inflammation is already established around neuritic plaques, thereby potentially presenting an obstacle to elicitation of a vaccine-mediated immune response. In AD, an innate immune response is triggered by local production of Aβ protein. Activation of the innate immune system is evident from elevated levels of complement proteins and activation of microglia, which together result in release of proinflammatory cytokines and chemokines (as has been reviewed elsewhere9). Additionally, Aβ fibrils can be modified by endogenous sugars to form advanced glycation end products (AGEs), which in turn activate proinflammatory signal transduction pathways involving upregulation of receptors for AGEs and overproduction of reactive oxygen species. Aβ and AGEs activate transcription factors, leading to upregulation of neurotoxic cytokines including IL-1, IL-6 and tumour necrosis factor.10 This inflammatory pathology is thought to be a secondary response to early accumulation of brain Aβ.

Importantly, both passive and active Aβ immunization elicit CNS inflammation, and can also induce cerebral microhaemorrhage and vasogenic oedema in the already inflamed milieu.3,11 Although the underlying mechanisms remain unclear, the formation of the Aβ–antibody complex is known to activate both microglia and the complement system as outlined above, yielding inflammatory responses in the CNS, and vessel damage.

Adaptive immune response

In addition to an innate immune response to inoculation with Aβ, autoimmune T lymphocytes can become activated by the vaccine and infiltrate the brain parenchyma. Damage elicited by such T cells is the main cause of aseptic meningoencephalitis that has been seen in clinical trials of immunotherapy.12 Increased T-cell reactivity against Aβ has not been demonstrated in amyloid precursor protein (APP)-transgenic mice, which might explain the lack of adverse effects of Aβ vaccination in preclinical animal studies. This difference between mouse and human responses may be secondary to high levels of peripheral Aβ and resulting induction of T-cell tolerance in APP mice.13

Another proposed explanation for the encephalitis that occurred in the AN1972 trial is that pathology resulted from antigen spreading; that is, an expansion of T-cell clones that are specific to myelin antigens and not Aβ.14

Vascular involvement

A substantial change that is elicited by immunotherapy is the shift of Aβ from the brain parenchyma to the vasculature—an event that promotes cerebral amyloid angiopathy (CAA) and microhaemorrhage.15,16 Antibody binding to vascular Aβ has been suggested to form a 'seed' that enhances CAA by trapping soluble Aβ during efflux from the brain. A higher burden of CAA might, in turn, sequester more anti-Aβ antibodies in the vessel walls, thereby triggering a positive-feedback loop of inflammation in the vessel wall, and increasing vulnerability to blood–brain barrier (BBB) damage, microhaemorrhage, and vasogenic oedema (Figure 1).17,18

Figure 1: Adverse effects associated with Alzheimer disease immunotherapy.
figure 1

Abbreviations: Aβ, amyloid-β; APP, amyloid precursor protein; BBB, blood–brain barrier; CAA, cerebral amyloid angiopathy.

Full size image

The 'dust-raising' effect

In addition to the above adverse effects that have been observed in both preclinical and clinical studies, a possible problem associated with immunotherapy is mobilization of insoluble Aβ from plaques, and the potential neurotoxicity of these Aβ species as they traverse the BBB from brain to periphery.19 Aβ has a high aggregating capacity and exists in different forms throughout the brain. Once generated by neurons, Aβ monomers rapidly aggregate into soluble oligomers, then into insoluble fibrils with β-pleated sheet structures surrounding neurons.

Increasing evidence suggests that oligomeric Aβ is the most neurotoxic form of the various Aβ species.20,21 Clearance of already well-established parenchymal plaques might, therefore, have limited potential to improve cognitive decline in patients with AD. Aggregation of Aβ into insoluble Aβ fibrils has been proposed as a way to reduce Aβ toxicity and, as such, Aβ oligomers are likely to be a better target for AD therapies in early-stage development than are plaques. Nevertheless, Aβ-directed immunotherapies that are currently in experimental and clinical trials still largely target Aβ deposits, and raise, or do not reduce, the levels of soluble Aβ.16,19 In fact, in several patients who received active immunization in the AN1972 study, reduced Aβ plaque load was associated with increased brain tissue concentrations of water-soluble and detergent-soluble forms of Aβ,16 suggesting that antibodies that entered the brain from the periphery solubilized Aβ deposits, leading to plaque mobilization. Such a change would favour the formation of Aβ oligomers, which might cause further damage to neurons during the clearance of Aβ from the brain to the periphery.

Soluble oligomers could, therefore, be envisioned as dust in the atmosphere, and Aβ deposits as sediment on the ground. Aβ clearance would be analogous to sweeping the ground sediment with a broom, which clears the ground but raises more dust into the atmosphere, where it can easily be inhaled and cause harm. This scenario could explain the greater brain volume decrease in antibody responders who received Aβ vaccination relative to patients in the placebo group in the clinical trial of AN1792.22 This 'dust-raising' effect of immunotherapy is a potential safety concern, and will need to be addressed in the future.

Autoimmunity and Aβ production

Endogenous antibodies to Aβ exist in healthy individuals, and can inhibit Aβ aggregation and block Aβ neurotoxicity.23,24 Thus, Aβ autoantibodies might be a physiological mechanism to prevent AD development. A recent study suggests that autoantibodies from healthy individuals are mainly directed against middle and C-terminal epitopes of Aβ, and are able to prevent Aβ oligomerization and fibrillization.23 This phenomenon is not universal, however, and the levels and functions of autoantibodies to Aβ in patients with AD remain unclear. Recently, we have found that autoantibodies to Aβ isolated from patients with AD stimulate Aβ production.25 Furthermore, we showed that autoantibodies to the N-terminal domain of Aβ are the most potent of Aβ-targeted antibodies in promoting Aβ production via β-secretase-mediated cleavage of APP.25 The exact mechanism remains to be elucidated, although binding of autoantibodies to the N-terminal domain of APP, which is exposed on the neuronal membrane surface, might initiate endocytosis and enzymatic proteolysis of the APP–antibody complex, thereby promoting Aβ generation. These findings raise two important points that are relevant to AD immunotherapy and pathogenesis.

First, autoantibodies generated following Aβ vaccination might stimulate Aβ production (Figure 1), as they are mainly specific for the N-terminal portion of Aβ.26 This specificity results from vaccination with Aβ fibrils, on which the N-terminal region of Aβ is exposed on the fibril surface and is available for antibody generation.27 By contrast, the middle and C-terminal regions of Aβ are 'masked' in the fibrils. It is intriguing to speculate whether antibodies against the Aβ N-terminus might promote Aβ production in the brain while also clearing Aβ deposits. The design of new generations of Aβ vaccines is based largely on the N-terminal sequence of Aβ, as other epitopes can trigger a T-cell response and result in aseptic meningoencephalitis (Figure 2). A potential problem is that such an approach could have an amyloidogenic effect. In addition, binding of antibodies to membrane APP could, in theory, trigger immune-mediated destruction of neurons.

Figure 2: Epitopes of the amyloid-β (Aβ) peptide sequence.
figure 2

Therapeutic antibodies with different Aβ domain targets that are currently in clinical trials are shown. Most such agents target the N-terminus of Aβ. See ClinicalTrials.gov for further details.41

Full size image

The second implication of a potential amyloidogenic effect of Aβ-targeted antibodies is that, at least in a subset of cases, AD might be an autoimmune syndrome, given that Aβ autoantibodies are observed in patients with AD. Some individuals might produce antibodies to Aβ that could also cross-react with APP and shift APP proteolysis in favour of amyloidogenesis. For example, a recent study suggests that brain-reactive autoantibodies, which are prevalent in serum from patients with AD, increase intraneuronal Aβ42 deposition in the brain.28 As such, screening of epitope specificity and functions of autoantibodies in patients with AD could be a future approach to understanding the pathogenesis, and possibly the aetiology, of sporadic AD.

Optimal immunotherapy

Adverse effects of Aβ-targeted immunotherapy have a substantial impact on treatment efficacy, but the underlying mechanisms remain unclear. Strategies to preserve the effectiveness of Aβ clearance while avoiding such adverse effects are a key focus of current immunotherapy research. New approaches to address this issue include vaccines that do not contain a T-cell epitope, so as to avoid autoimmune T-lymphocyte activation; non-Aβ vaccines that mimic the conformational structure of Aβ; and use of antibody fragments (Fab and single-chain antibodies) or full Aβ antibody with its Fc fragment deglycosylated, to avoid Fc-mediated inflammatory responses (reviewed elsewhere29).

On the basis of the findings outlined in this article, we propose that therapeutic Aβ-specific antibodies should have the following three properties: first, the capacity to remove both soluble and insoluble Aβ without increasing soluble Aβ levels and promoting oligomer production; second, the capability to block Aβ neurotoxicity during Aβ clearance; and last, the capacity to specifically bind to Aβ or Αβ oligomers, but not to APP or other metabolites of APP. Here, we propose possible ways to develop such a vaccine.

Peripheral Aβ clearance

The 'peripheral sink' hypothesis states that anti-Aβ antibodies in the blood bind and sequester plasma Aβ in the periphery and prevent its influx into the brain, thereby disrupting the free equilibrium of Aβ across the BBB and facilitating efflux of this peptide from the brain. Efflux of soluble Aβ in turn alters the equilibrium between soluble and insoluble Aβ in the brain, favouring transformation of Aβ fibrils into soluble Aβ, and finally yielding a reduction in brain Aβ load. To expand on our previous analogy, this process would be akin to using a vacuum cleaner rather than a broom, preventing the dust from rising into the atmosphere. As a result, levels of soluble Aβ would not increase. In addition, these anti-Aβ antibodies cannot enter the brain, which avoids formation of antibody–APP immune complexes on the neuron membrane or of antibody–Aβ aggregate complexes in the parenchyma and vasculature. Such antibodies should not, therefore, induce autoimmunity-related adverse events such as Aβ overproduction, neuronal membrane attack, and inflammatory responses (Figure 1). Thus, as a therapeutic framework, the peripheral sink mechanism is a potentially safe and promising strategy.

In some immunotherapy studies in AD, CAA—which occurred in response to treatment owing to redistribution of solubilized parenchymal Aβ to the vasculature15,30—compromises Aβ clearance.31 Whether peripheral sink clearance of Aβ can aggravate this effect remains unclear. A recent study found that chronic administration of antibody m266, which is thought to work as a peripheral sink as it recognizes the middle region of Aβ and is devoid of reactivity for soluble oligomers and insoluble Aβ plaques in the brain, had no effect on vascular Aβ levels.32 By contrast, administration of antibody 3D6, which recognizes the N-terminal region of Aβ and binds insoluble Aβ plaques in the brain, was associated with CAA exacerbation in animals.18 However, the increase in vascular Aβ is thought to be transient, as in clinical trials of Aβ immunization vascular Aβ was eventually cleared.30

Clinical trials of solanezumab (humanized m266) are considered a test of the peripheral sink hypothesis. In phase II trials, solanezumab treatment did not cause significant adverse effects in the vasculature such as vasogenic oedema and microhaemorrhage,33 whereas these adverse effects occurred in one-fifth of patients treated with bapineuzumab (humanized 3D6).34 However, entry of solanezumab into the CNS may stabilize brain Aβ monomers, retarding Aβ efflux to the periphery.35 Use of larger Aβ-binding molecules that cannot enter the brain yet can still promote peripheral metabolism of Aβ via efflux from the brain should be pursued in the future.

The C-terminus-specific antibody ponezumab is also suggested to work through a peripheral sink mechanism, but this antibody was withdrawn from phase II trials, for reasons that have not been reported to date.

Targeting the toxic domain of Aβ

Consideration of the domain responsible for Aβ neurotoxicity is important for the development of a safe and effective vaccine. Evidence suggests that this domain is mainly located in the amino acid 25–35 (AA25–35) region, which exists physiologically as a truncated form of Aβ in the brains of elderly people and plays a part in AD pathogenesis. The aggregation capacity of Aβ is proposed to correlate with its neurotoxicity and, importantly, the AA25–35 region is crucial for Aβ aggregation.36 Colocalization of the neurotoxic and aggregating domains provides an excellent target for development of antibodies that could improve Aβ clearance and reduce neurotoxicity. The target of future Aβ vaccination might, therefore, shift from the current N-terminus to the middle or C-terminus of Aβ.

In this regard, neurotrophin receptor p75 (p75NTR) could be an ideal therapeutic target. As an Aβ receptor, p75NTR mediates Aβ-induced neuronal death and neurite degeneration, and promotes Aβ production. We found that the recombinant extracellular domain of p75NTR, which binds to the AA25–35 region of Aβ, can inhibit Aβ oligomerization and fibrillation, as well as Aβ-induced neurotoxicity, suggesting the therapeutic potential of targeting this receptor.37

Improving antibody specificity

Improvement of antibody specificity for the Aβ peptide, as opposed to other APP products or full-length APP, is another approach to optimization of immunotherapy for AD. A recent study by our group has found that some antibodies against the N-terminal region of Aβ can also bind APP and substantially increase Aβ production by favouring β-secretase while dampening α-secretase cleavage of APP.25 Another potential concern related to antibody specificity for Aβ is that binding of antibodies to APP could trigger neuron-targeted immune responses via activation of the complement cascade. Understanding the mechanisms of these off-target effects is necessary for future development of anti-Aβ immunotherapies that do not promote Aβ generation or initiate autoimmune responses. Toward this goal, the conformational structure and amino acid sequence of Aβ-targeted autoantibodies should be fully characterized.

Another aspect of antibody specificity is the form of Aβ that is targeted. Antibodies should be developed that target oligomeric forms of Aβ, which are thought to be the main cause of neuronal death in AD. Selective removal of Aβ oligomers and Aβ-derived diffusible ligands is expected to protect against Aβ-mediated neurotoxicity and preserve cognitive functions without altering the normal physiology of monomeric Aβ.38 Conformation-specific antibodies that recognize these Aβ species are currently under development.39,40

Conclusions

Immunotherapies that target Aβ are promising candidates to be the first disease-modifying treatments for AD. However, adverse effects that compromise efficacy, including interaction of anti-Aβ antibodies with APP, inadvertent mobilization of insoluble Aβ, CAA-associated microhaemorrhage and vasogenic oedema, need to be resolved. Progress in understanding Aβ clearance mechanisms and immunotherapy-induced autoimmunity will enable the development of more-refined therapeutic antibodies, allowing completion of clinical trials without serious adverse effects, and producing immunotherapies with increased efficacy.