β-amyloid (Aβ) is a peptide, 38 to 43 amino acids long, that derives from the proteolytic processing of amyloid precursor protein (APP) by the γ-secretase; Aβ40 and Aβ42 are the most studied Aβ peptide species (Figure 1). The β-amyloid cascade model of Alzheimer’s disease pathogenesis consists in the hypothesis that an imbalance between production and clearance of Aβ42 and related Aβ peptides is a very early, often initiating factor in Alzheimer’s disease [13].

Figure 1. Generation of the Aβ peptide from the processing of APP by BACE and γ-secretase. APP is a type I trans-membrane glycoprotein. The β-secretase BACE has a single cleavage site on APP and generates the N-terminus of Aβ peptides. The γ-secretase has multiple cleavage sites on APP, which leads to the generation of Aβ peptides of variable length that differ for their C-terminus. The most abundant peptides are Aβ40 and Aβ42. Aβ42 is particularly prone to aggregation.

Strong evidence of the amyloid cascade model comes from the genetics of the ‘dominantly inherited forms of Alzheimer’s disease’ (DIAD). DIAD is caused by mutations that impact on the Aβ pathway, affecting either the amount of Aβ production or the ratio between the different forms of the peptide, in particular, by mutations within and immediately flanking the Aβ region of APP and by missense mutations in presenilin 1 and presenilin 2, the catalytic subunit of γ-secretase, that results in a relative increase in the production of Aβ42/43 peptides. In agreement with the concept that an excessive dose of Aβ causes Alzheimer’s, an enhanced gene dose of the precursor protein APP can cause the disease. In fact, APP is located on chromosome 21, and people with Down’s syndrome, who harbor three copies of APP, develop the typical neuropathology of Alzheimer’s disease at a young age [13].

The physiological functions of APP, the closely related APP-like proteins (APLPs), and their multiple processing products are still not well understood; there is evidence for a role in the development of the central nervous system, the formation and function of synapses, and neuroprotection following brain injury [15]. Aβ42, when excised from its precursor, is prone to undergo a conformational change that renders it able to self-aggregate into oligomers that can further assemble into fibrils and amyloid plaques (Figure 2). Amyloid fibrils are a structure that several proteins can adopt, characterized by a cross-β-sheet conformation in which β-strands run transversely to the main fiber axis and form an intermolecular network of hydrogen bonds [16].

Figure 2. Process of Aβ aggregation and β-amyloid plaque formation. The Aβ peptide, once excised from APP, is prone to misfolding and self-aggregation. Misfolded Aβ monomers aggregate into small soluble oligomers. The oligomers interact to form protofibrils, which grow to form mature fibrils. Eventually the fibrils aggregate, forming the β-amyloid plaques.

Aβ plaque deposition is followed by neuritic and glial cytopathology in surrounding areas. A sequential association has been reported between brain β-amyloid accumulation, subsequent tau change, and resulting cognitive decline in individuals with dominant inherited AD [17]. While the amyloid plaques have been the target of many therapeutic approaches, the Aβ oligomers, which are diffusible and neurotoxic and exist in equilibrium with plaques, play the major role in neurodegeneration [18]. Aβ42 oligomers induce tau hyperphosphorylation and cause neuritic dystrophy in cultured neurons. In animal models, Aβ oligomers decrease synapse density, inhibit long-term potentiation, enhance long-term synaptic depression, and impair memory [19]. The most likely sequence of events leading to the disease comprises a neocortical Aβ accumulation, followed by a microglial inflammatory reaction to Aβ, neuritic dystrophy and spread of tau from the limbic system to the neocortex, and progressive tau accumulation and spread resulting in neurodegeneration [8,17]. Different therapeutic targets could be required for different stages of the disease process: Aβ for primary prevention, microglia for secondary prevention, and tau for established disease [8]. The β-amyloid peptide displays a highly complex self-assembly behavior, and it is challenging to define the aggregation process in terms of molecular events. The process is modeled as comprising primary and secondary pathways, that is, pathways that generate aggregates at a rate dependent on the concentration of monomers alone and independent of the concentration of existing fibrils (primary) and pathways that generate new aggregates at a rate dependent on the concentration of fibrils (secondary) [20]. Secondary pathways include mechanisms that depend only upon the concentration of fibrils, such as the fragmentation of fibrils, and mechanisms that depend on the concentration of both monomers and fibrils, such as secondary nucleation, whereby the surface of existing fibrils catalyzes the formation of oligomers from monomers. Secondary nucleation is a positive feedback loop in the aggregation process [20].

In agreement with this model of the aggregation process, oligomeric Aβ42 aggregates are particularly abundant around plaques; the neurodegeneration that is observed, in a gradient, around plaques is believed to derive mainly from the gradient of concentration of the diffusible, toxic oligomers [19,21]. Strategies to suppress the production of toxic oligomers need to consider both primary and secondary nucleation pathways of oligomer production. In this respect, while the oligomers and not the plaques are the direct culprit of the neurodegeneration, removing existing plaques makes sense as a step toward the reduction in the concentration of oligomers.

The early phases of the aggregation process reflect the fact that Aβ is a prion-like peptide, namely a peptide that adopts alternative conformations, which are self-propagating [22]. The self-propagating conformers, also known as ‘Aβ seeds’, can be quantified in cellular assays and in mouse models [22,23]. Aβ seeding potency is greatest early in the pathogenic cascade and diminishes as Aβ accumulates in the brain [22,24]. Interestingly, Aβ, as prions, can assemble into distinct strains of aggregates. Such strains may drive some of the phenotypic heterogeneity observed in Alzheimer’s disease [25].

3. Mechanisms of Action of Anti-β-Amyloid Antibodies

Many studies in vitro, in cell cultures and animal models, have shown that antibodies against β-amyloid can exert potentially useful effects to counteract β-amyloid dependent neurodegeneration, such as interfering with β-amyloid aggregation, blocking β-amyloid toxicity, or reducing the amount of β-amyloid in the brain (Figure 3).

Figure 3. Mechanisms of action of anti-Aβ-antibodies. Anti-Aβ-antibodies can interfere with the β-amyloid cascade at several levels, by interacting with the Aβ monomers, the Aβ oligomers, the Aβ-fibrils, or the Aβ-plaques. The epitope specificity and concentration of antibody required for the different mechanisms are probably different. In human clinical trials, evidence of phagocytosis of plaques by microglia has been reported. The ‘peripheral sink effect’ is a hypothesis that has been disproven.

The entry of antibodies in the brain is restricted by the blood–brain barrier [26]; the concentration of an antibody in the CSF is 0.1 to 0.2% of the concentration in blood plasma [27,28]. Peripherally administered antibodies against the β-amyloid peptide can enter the central nervous system and reduce amyloid load; this was first documented in mouse models of Alzheimer’s Disease [29], and more recently also observed in clinical trials of passive immunization [30].

Different mechanisms of action of anti-β amyloid antibodies have been hypothesized or documented, including allosteric effects, the induction of plaque phagocytosis by microglia, the promotion of efflux of Aβ from the CNS to the circulation, the neutralization of β-amyloid toxicity (Figure 3). It is reasonable to expect that different mechanisms can become relevant depending on the stage of the amyloid deposition process, and on the concentration, isotype, and epitope specificity of the antibodies.

Amyloid deposition appears to follow a sigmoidal trajectory over time [31]. The time window where the slope of the amyloid load versus time curve is greatest represents a potential therapeutic window for secondary preventive interventions [31]. On the other hand, therapeutic interventions designed to reduce the rate of new amyloid deposition, rather than removing previously deposited amyloid, may be less effective in patients who have already reached plateau levels of amyloid deposition [31].

In vitro, anti-Aβ monoclonal antibodies can prevent Aβ monomers from forming fibrillar aggregates, and can convert fibrillar aggregates into an amorphous state [32]. Interestingly, the efficacy of these mechanisms depends on the concentration of antibody. At low concentrations of AMY-33, a monoclonal antibody raised against β-amyloid fragment 1–28, only amorphous aggregates are formed; increasing the concentration of antibody to equimolar antigen/antibody ratios maintained the solubility of β-amyloid [32]. The solubilization of already formed aggregates also required an equimolar ratio of antigen/antibody ratio, and prevented the neurotoxicity of β-amyloid [33].

Some epitopes of Aβ are preferentially available within plaques, whereas other epitopes are only available on the soluble peptide. Therefore, the epitope specificity of an antibody affects its binding affinity against different β-amyloid species. For epitopes exposed both in the monomer and in the aggregate forms, the presence of two antigen-biding sites generates avidity effects for aggregates. Therefore, the preferential binding of an antibody to the aggregate forms does not necessarily imply the recognition of a conformational epitope only present in the aggregate species. For instance, the sera of mice immunized with a multimeric protein that displays the Aβ(1–11) peptide [34,35], recognize the synthetic Aβ(1–11) peptide in ELISA assay, but display a marked preference for oligomeric and fibrillar species of β-amyloid in dot blot assays [36].

It has been hypothesized that soluble β-amyloid levels in the brain and the peripheral blood are in equilibrium so that blocking or degrading β-amyloid in blood should increase its efflux from the brain, and reduce brain amyloid. This hypothesis, known as the peripheral sink hypothesis [37], has been refuted. In various experimental systems, treatments that substantially decreased peripheral Aβ levels failed to affect brain and cerebrospinal fluid levels of Aβ; these findings suggest a lack of a significant peripheral sink effect through which brain amyloid burdens can be therapeutically reduced [38,39,40]. In clinical trials, a monoclonal antibody that binds soluble Aβ in blood, solanezumab, did not affect brain levels of amyloid [41]. When Aβ in blood is bound to solanezumab, the half-life of the Aβ-antibody complex is much longer than the half-life of free Aβ; therefore, treatment with solanuzemab leads to an increase in of the concentration of Aβ in blood, but the increase is not due to an efflux of Aβ from the brain.

While the fragment of the antibody that binds the epitope determines the specificity of the antibody, the constant region of the antibody, or Fc, is responsible for the capacity of an antibody to activate the complement cascade or to bind the Fc receptors that are expressed on a wide variety of cell types of the immune system, including microglia in the central nervous system; different antibody isotypes differ in their affinity for Fc receptors and ability to activate complement. The antibody isotype, therefore, affects Fc- or complement-mediated phagocytosis of plaques by microglial cells. An analysis of the epitope and isotype specificity of antibodies against β-amyloid able to protect against Alzheimer’s disease concluded that epitopes within the N terminus of Aβ are important for plaque clearance and neuronal protection via an Fc-mediated mechanism, and that IgG2a antibodies against Aβ are more efficient than IgG1 or IgG2b antibodies in reducing neuropathology [42].

4. Effect of Anti-β-amyloid Vaccination in Mouse Models of β-Amyloid Deposition

Genetically engineered mouse models have been instrumental to Alzheimer’s disease research and preclinical drug development. Transgenic mouse strains that overexpress mutant human APP linked to familial AD progressively develop many of AD’s pathological hallmarks—including senile plaques, synaptic loss, astrocytosis, and microgliosis—and have been largely used as preclinical research models. Recently, new mouse models have been generated that contain humanized sequences and clinical mutations in the endogenous mouse APP gene [43].

The first report of a disease-modifying effect of immunization against β-amyloid was published over 20 years ago. A transgenic mouse model of Alzheimer’s disease, the PDAPP mouse, was immunized with Aβ42, either before the onset of neuropathology, or at an older age, when amyloid-β deposition and neuropathology were well established. Immunization of the young animals prevented the development of β-amyloid-plaque formation, neuritic dystrophy, and astrogliosis, and treatment of the older animals markedly reduced the neuropathology [44].

A study performed in a different transgenic mouse model of Alzheimer’s disease, the TgCRND8 mouse, reported that Aβ immunization reduced both the deposition of cerebral fibrillar Aβ and cognitive dysfunction without, however, altering total levels of Aβ in the brain. The authors suggested that either a 50% reduction in dense-cored Aβ plaques was sufficient to affect cognition, or vaccination may modulate the activity/abundance of a small subpopulation of especially toxic Aβ species [45].

In yet another transgenic mouse strain, the Tg 2576 APP transgenic mice, and in a double mutant including both the APP and a PSEN mutation, vaccination against Aβ afforded protection from memory impairment, in the presence of reduced—but still substantial—Aβ deposits. The authors hypothesized that the antibodies could neutralize Aβ in some restricted compartment or deplete a non-deposited form of Aβ (for example, a soluble form) responsible for the memory loss [46].

The perspective of injecting the Aβ1–42 peptide in healthy, young humans to vaccinate them against Alzheimer’s raises in principle a safety concern over the possibility that the injected peptide itself, being able to promote oligomerization and fibrillogenesis, may start the pathogenic cascade in some healthy vaccinees, causing an enhanced probability of incurring the disease decades later. This scenario still cannot be formally excluded. Therefore, it is important to bear in mind that, ideally, the molecules used to immunize against β-amyloid should not be able to initiate the amyloid cascade. So far, all human immunizations have been performed in individuals that already had aggregated β-amyloid in their body. Importantly, in mice, it is unnecessary to immunize with the entire, pre-aggregated Aβ1–42 to observe an effect on brain amyloid. Immunization of transgenic APP mice with a soluble nonamyloidogenic, nontoxic Aβ homologous peptide, consisting of the first 30 amino acid residues of Aβ with six additional lysine residues at the N-terminus, reduced cortical and hippocampal brain amyloid burden and brain levels of soluble Aβ1–42, and reduced neuroinflammation [47]. 

7. Anti-β-Amyloid Vaccination in Humans with Second Generation Vaccines

Despite the serious adverse events, the results of the AN1792 clinical trials encouraged further research into active anti-Aβ immunotherapy. Since the meningoencephalitis was attributed to the T cell response, mapped to the central part of the β-amyloid peptide, and the effects on disease progression were attributed to the antibody response, mapped to the N-terminus of Aβ, second generation vaccines mainly used the strategy of directing the immune response to the N-terminal B cell epitope, without inducing a concomitant T cell response to β-amyloid.

The active immunogens that have entered clinical trials are shown in Table 1. Four anti-Aβ vaccines are currently being tested in phase II trials: CAD106, ACI-24, UB-311, and ABVac40.

Table 1. Vaccines against β-amyloid that entered phase II clinical trials.

The T cell response is important for the generation of high affinity IgG; several second generation anti-Aβ vaccines include exogenous T cell epitopes. Some candidate vaccines rely on T cell epitopes present in a carrier, for instance vaccine CAD106 consists of the B cell epitope Aβ1–6 linked to the capsid of the Qβ bacteriophage, and vaccine ABvac40 consists of the B cell epitope Aβ33-40 conjugated to a carrier protein, keyhole limpet hemocyanine (KLH) (Table 1).

Other candidate vaccines consist of synthetic peptides that include T cell epitopes from pathogens: vaccine UB-311 is a mixture of two synthetic peptides, each including the B cell epitope Aβ1–14 and a T cell epitope either from the hepatitis virus or the measles virus; vaccine Lu AF20513 includes B cell epitope Aβ1–12 and T cell epitopes from tetanus toxin (Table 1).

None of the second-generation vaccines that entered clinical trials have induced meningoencephalitis, whereas antibody-mediated adverse effects, such as ARIA, were reported. No clinical benefit from the second-generation vaccines has been reported so far.

Although no direct comparison has been published, the duration of the anti-Aβ response elicited by second generation vaccines appears shorter than the duration of the response elicited by the AN1792; vaccine-induced antibodies become undetectable a few months after the last injection.

In the case of ACC-001 [72], substantial interindividual variability in anti-Aβ IgG titer was observed, and no correlation was found between IgG titer levels and cognitive or functional efficacy results, or biomarker results [72].

CAD106 induced an antibody response against Aβ in most patients. No meningoencephalitis was observed. CAD106-induced antibodies reacted with amyloid plaque cores and Aβ oligomers ex vivo. The best reactivity was seen with the entire Aβ1–6 epitope. No statistically significant differences were observed in CSF biomarkers [73].