Protein aggregation diseases: pathogenicity and therapeutic perspectives

Key Points

• Amyloids can be broadly defined as insoluble protein aggregates containing a characteristic highly ordered, β-sheet-rich structural motif. The latter can be identified histologically with dyes such as Congo Red and thioflavin. Some amyloids may have an adaptive physiological function; however, most amyloids are thought to be abnormal and are associated with a range of clinical pathologies including systemic amyloidoses and some neurodegenerative disorders.

• Alzheimer's disease, the most common neurodegenerative disorder, is currently the focus of some of the most exciting and rapidly progressing research on amyloid therapeutics. Two main approaches are being pursued to deplete cerebral amyloid β (Aβ) levels, the primary component of senile plaques associated with Alzheimer's disease. The first approach involves inhibition of the secretases responsible for Aβ production. The second approach involves mobilization of the immune system to promote Aβ clearance.

• Aβ is produced from the sequential proteolysis of amyloid precursor protein (APP) by β- and γ-secretase. The development of small-molecule inhibitors for these enzymes or complexes as a therapeutic method of depleting monomeric Aβ from the brain has met with considerable success, as well as several challenges. Most notably, the secretases have other non-APP substrates (for example, γ-secretase cleavage of Notch and β-secretase cleavage of Neuregulin 1) that are important for normal physiology.

• Enhanced clearance of monomeric Aβ and Aβ aggregates by Aβ immunotherapy has been used successfully to lower cerebral Aβ levels and promote cognitive improvement in murine amyloid models. Initial attempts to actively immunize against Aβ in humans were suspended owing to the development of adverse immune responses in some patients. However, passive immunotherapeutic approaches have progressed to clinical trials, and administration of intravenous immunglobulins may also be effective for lowering Aβ levels in the brain.

• Increasing evidence implicates specific forms of Aβ (for example, soluble Aβ oligomers and intraneuronal Aβ) in Alzheimer's disease-associated neurotoxicity, raising the intriguing possibility that therapies targeted at these pools of Aβ (for example, conformation-specific antibodies) may be effective in ameliorating cognitive deficits associated with Alzheimer's disease by blocking cell death pathways rather than altering Aβ homeostasis.

• Prion diseases are the only known amyloidoses that act as genuine infectious diseases, with the possible exception of AA amyloidosis. The unusual properties of prion amyloid have presented significant challenges for therapeutics, as well as some opportunities to explore unique therapeutic modalities.

• In many cases of acquired prion disease, prions first colonize and replicate in extraneural secondary lymphoid organs before being transmitted to the central nervous system (CNS). Blockade of prion replication at these extraneural sites (for example, by inhibition of lymphotoxin β receptor (LTβR) signalling) is an effective method to prevent the spread of prions from the periphery to the CNS and may be useful for post-exposure prophylaxis against prion infections.

• Attempts to generate active immune responses against PrP^Sc (PrP with abnormal conformation) have had little success owing to immune tolerance of ubiquitously expressed endogenous PrP^C(PrP with normal conformation). However, passive immunization with large doses of PrP antibodies isolated from Prnp^−/− mice can prevent the spread of prions from the periphery to the CNS upon intraperitoneal inoculation. It has not yet been demonstrated that PrP immunotherapy is effective in slowing the rate of prion disease once it has reached the CNS. However, the development of PrP antibodies that effectively cross the blood–brain barrier may dramatically enhance the efficacy of PrP immunotherapy for the treatment of CNS prion infections.

• Numerous compounds have been investigated or are being developed for their anti-aggregation properties including various amyloid-binding dyes, anti-malarial compounds, protein X mimetics, β-sheet breakers and scyllo-inositol. These compounds are still in development for the treatment of prion diseases and other neurodegenerative disorders; however, some of these compounds are currently in clinical trials for the treatment of systemic amyloidoses.

• Several lines of evidence indicate that protein aggregates trigger specific cellular toxicity pathways in the brain, including the resistance of non-neuronal cell types and Prnp^−/− neurons to PrP^Sc-induced toxicity. Intriguingly, PrP^C was recently identified as a receptor for oligomeric Aβ, suggesting that diverse protein aggregates may activate common neurotoxicity pathways, which may have important therapeutic implications for amyloid diseases.

Abstract

A growing number of diseases seem to be associated with inappropriate deposition of protein aggregates. Some of these diseases — such as Alzheimer's disease and systemic amyloidoses — have been recognized for a long time. However, it is now clear that ordered aggregation of pathogenic proteins does not only occur in the extracellular space, but in the cytoplasm and nucleus as well, indicating that many other diseases may also qualify as amyloidoses. The common structural and pathogenic features of these diverse protein aggregation diseases is only now being fully understood, and may provide novel opportunities for overarching therapeutic approaches such as depleting the monomeric precursor protein, inhibiting aggregation, enhancing aggregate clearance or blocking common aggregation-induced cellular toxicity pathways.

Main

Inappropriate aggregation of proteins is normally prevented by complex cellular quality control mechanisms. However, under certain circumstances, an unusual subset of proteins is able to aggregate within or around cells. Although the amino-acid sequences of this class of proteins are diverse, they all seem to adopt a similar, insoluble, highly ordered structure when aggregated known as the cross-β spine¹ (Fig. 1a). The histology of the resulting protein deposits was appreciated more than 150 years ago and was termed amyloid² (Fig. 1b). Although it is possible to predict the approximate tendency of proteins to form amyloid on the basis of their sequence³, amyloid formation is far from a simple function of protein sequences and secondary structures. It is becoming increasingly apparent that amyloid-forming proteins exist in a complex dynamic equilibrium between soluble monomeric or oligomeric states and various insoluble states of higher-order aggregation. The formation of these aggregates depends on the protein concentration, complex interactions with other proteins and the specific cellular environment. A more thorough understanding of the factors that influence this equilibrium is crucial for determining how protein aggregation disorders arise and for developing effective therapies against them.

Figure 1: Common features of protein aggregation and amyloids.

a | Ribbon diagrams of the three-dimensional structure of amyloid-β (Aβ42) (residues 17–40). Amyloid-forming proteins such as Aβ are all thought to form similar tertiary structures when aggregated, known as a cross-β spine or amyloid¹⁶². The cross-β spine consists of an ordered arrangement of β-sheets (thick coloured arrows). b | Congo red-stained sections of human kidney affected by amyloidosis. The cross-β spine structure is able to intercalate with molecules of the azo-dye Congo red and cause them to emit a characteristic apple-green birefringence upon exposure to polarized light¹⁶³. This unique feature of amyloids has historically been used to identify them histologically.

From protein aggregation to pathology

Aggregation and accumulation of amyloid-forming proteins can lead to a wide range of protein aggregation diseases known as amyloidoses. These diseases include systemic amyloidoses in which deposits may occur in any part of the body, such as AL amyloidosis due to the accumulation of immunoglobulin light chain amyloid fibrils; amyloid A (AA) or reactive systemic amyloidosis resulting from deposits of catabolic products of the serum amyloid A protein (SAA); and ATTR due to transthyretin (TTR) accumulation. There are also amyloidoses in which a single organ is affected — possibly including pancreatic accumulation of islet amyloid polypeptide in type 2 diabetes⁴.

Many protein aggregation diseases affect the nervous system. Interestingly, certain proteins specifically aggregate and are toxic in the central nervous system (CNS) despite the fact that they are ubiquitously expressed. These neurodegenerative diseases include disorders in which the pathological proteins may accumulate within the nucleus, as is the case with polyglutamine expansion diseases (such as Huntington's disease and spinocerebellar ataxias), disorders characterized by cytoplasmic inclusions (such as α-synuclein in Parkinson's disease), as well as disorders in which pathological proteins accumulate extracellularly (for example in prion diseases) or both intracellularly and extracellularly (for example, tau and amyloid-β (Aβ) in Alzheimer's disease).

It is becoming clear that amyloid formation is not always pathological, but rather might have a normal physiological function⁵. However, not all pathological protein aggregates necessarily form amyloids. Nevertheless, the definition of amyloid is broadening to encompass a wider range of diseases⁶. Originally, the term was used exclusively to describe extracellular amyloid deposits that stain with histological dyes such as Congo red. Now, it is recognized that many cytoplasmic⁷ and even intranuclear inclusions⁸, which do not necessarily stain with these dyes, are composed of ordered fibrillar structures similar to those of classical amyloids. Furthermore, some of these dyes seem to bind to aggregated intermediates of amyloid-forming proteins in addition to mature amyloid fibrils (for example, thioflavin appears to bind to Aβ oligomers)^9,10,11. Interestingly, staining with Congo red and thioflavin S has been reported intracellularly in some cases^12,13. Thus, the current classification of amyloids is far from clearly defined.

In this Review, we summarize various therapeutic approaches aimed at preventing or reversing pathological protein aggregation (Table 1). Our discussion is focused around two neurodegenerative diseases: prion disease and Alzheimer's disease. However, a number of extracerebral amyloidoses are also benefiting from therapeutic advances. Several systemic amyloidoses are becoming manageable chronic diseases rather than definite death sentences.

Table 1 Summary of therapeutic strategies for protein aggregation diseases

Alzheimer's disease

Alzheimer's disease, the most common neurodegenerative disorder, is currently the focus of some of the most exciting and rapidly progressing research on amyloid therapeutics. Alzheimer's disease pathology is characterized by the formation of two types of protein aggregates in the brain: amyloid plaques (Fig. 2a) — which form an extracellular lesion composed of the Aβ peptide; and intracellular neurofibrillary tangles (Fig. 2b) — which are composed of hyperphosphorylated filaments of the microtubule-associated protein tau. Genetic evidence implicates deregulated Aβ homeostasis as an early event in Alzheimer's disease pathology¹⁴. Indeed as all familial Alzheimer's disease mutations lead to increased production of this peptide or preferential production of a more fibrillogenic Aβ isoform (Aβ42)¹⁵. For this reason, most Alzheimer's disease therapeutics have targeted the Aβ peptide, although tau-targeted therapies are also being pursued^16,17.

Figure 2: Characteristics of Alzheimer's disease.

a | A human cortical section from a patient affected by Alzheimer's disease, stained with an amyloid-β (Aβ)-specific antibody. One of the classical hallmarks of Alzheimer's disease histopathology is the appearance of extracellular lesions known as senile or amyloid plaques, which are primarily composed of the Aβ peptide. b | A human cortical section from a patient affected by Alzheimer's disease, stained with a phospho-tau-specific antibody. The second histopathological hallmark of Alzheimer's disease is the presence of intraneuronal lesions known as neurofibrillary tangles (indicated by an arrow), which are composed of abnormal, hyperphosphorylated filaments of the microtubule-associated protein tau.

Targeting Aβ production: the secretase inhibitors. One widely pursued method of combating protein aggregation diseases is to inhibit the production of the monomeric form of the protein with the aim of reducing the amount of protein available to aggregate. In the case of Alzheimer's disease, the amyloidogenic Aβ fragment associated with amyloid plaques is derived from proteolytic processing of a longer, non-aggregating precursor protein — amyloid precursor protein (APP) (Fig. 3). Therefore, pharmacological inhibition of the enzymes responsible for Aβ formation (γ-secretase and β-secretase) is a prime strategy for blocking Aβ production. The γ-secretase complex consists of four proteins¹⁸, the catalytic activity of which is thought to be mediated by the presenilin 1 (PS1) and PS2 proteins. The functional assembly of these proteins has the unusual ability to target a stretch of APP that is entirely buried within the lipid bilayer of the plasma membrane. The γ-secretase complex is responsible for the carboxy-terminal cleavage of APP to produce Aβ40 or Aβ42.

Figure 3: Amyloid-β formation.

Amyloid precursor protein (APP) undergoes a series of proteolytic cleavages in neurons to form the amyloid-β (Aβ) peptide that is associated with senile plaques in Alzheimer's disease. In the amyloidogenic pathway (a), internalized APP is initially cleaved at its amino terminus by endocytic β-secretase (β) to form secreted APPsβ and C99. C99 then becomes a substrate for intramembraneous cleavage by the γ-secretase complex, leading to the release of Aβ40 (or Aβ42 at low frequency) and the APP intracellular domain (AICD), which may regulate gene expression. In a competing, non-amyloidogenic pathway (b), α-secretase cleaves cell surface APP to liberate secreted APPsα and C83. C83 is then cleaved by γ-secretase to form the soluble p3 peptide and the AICD. The majority of Aβ and N-terminal APP cleavage fragments are eliminated from the neuron by the secretory pathway.

Potent small-molecule inhibitors of γ-secretase can dramatically reduce Aβ40 and Aβ42 production^19,20,21,22. The primary caveat to targeting γ-secretase for Alzheimer's disease therapeutics is that APP is not the only substrate of γ-secretase. The most notable alternative cleavage substrate is the Notch receptor^23,24. Cleavage of Notch by γ-secretase is crucial for normal development, as shown by the fact that Ps1^−/− mice suffer from embryonic lethality owing to deficient Notch cleavage^25,26. Although this does not necessarily disqualify γ-secretase as a drug target, Notch seems to mediate differentiation of several cell types throughout adulthood (for example, gut epithelium and T cells). The unfortunate consequence of this is that potent γ-secretase inhibitors have serious gastrointestinal and immunological side effects²⁷.

As a result of this drawback, the field has shifted towards the development of γ-secretase modulators. These compounds either selectively inhibit γ-secretase cleavage of APP, leaving Notch cleavage unaffected, or alter γ-secretase cleavage of APP to favour Aβ40 production rather than Aβ42. The longer, less abundant 42 amino-acid isoform of Aβ, Aβ42, seems to be more closely associated with the development of amyloid pathology than its counterpart, Aβ40. Studies have shown that dominantly-inherited PS1 mutations, which cause Alzheimer's disease with 100% penetrance, lead to a shift in γ-secretase cleavage of APP in favour of the Aβ42 isoform^{28,29,30,31,32,33}. Drugs that modulate γ-secretase activity in this manner include non-steroidal anti-inflammatory drugs (NSAIDs) such as ibuprofen³⁴ and compounds that interact with the ATP-binding motif of PS1 near the γ-secretase active site³⁵.

One such γ-secretase-modulating compound, the NSAID (R)-flurbiprofen (also known as tarenflurbil), effectively reduced amyloid plaque formation³⁶ and rescued memory deficits³⁷ in APP-transgenic mice. It also yielded encouraging results in early human trials^38,39. However, (R)-flurbiprofen failed to significantly enhance cognitive performance of patients with Alzheimer's disease in Phase III clinical trials and has recently been abandoned as a potential therapy⁴⁰. The compound-screening approach may identify additional promising small-molecule γ-secretase modulators, and other compounds may be more successful than (R)-flurbiprofen in human trials (for example, semagacestat from Lilly and Elan is currently in Phase III trials⁴¹), researchers may also decide to capitalize on recent results demonstrating the existence of endogenous γ-secretase activity modifiers that are APP-specific⁴² and the observation that some components of the γ-secretase complex have tissue-specific isoforms, which may provide a novel means of achieving brain-specific γ-secretase inhibition⁴³.

The amino-terminal cleavage of APP to form both Aβ40 and Aβ42 results from β-secretase activity. After the discovery that β-secretase cleavage of APP seemed to be due to the activity of a single aspartic protease, β-secretase 1 (BACE1; also known as memapsin 2 and ASP2), there was much interest in the possibility of targeting β-secretase for the treatment of Alzheimer's disease^{44,45,46,47,48}. Deletion of this enzyme does not have an overt effect on phenotype in mice, suggesting that BACE1 inhibitors, unlike γ-secretase inhibitors, might lack serious target-related side effects⁴⁹. Inhibition of BACE1 activity can block the production of Aβ, prevent the development of amyloid pathology in the brain and rescue Alzheimer's disease-related memory deficits in mice^50,51,52,53.

However, nearly a decade has passed since the initial identification of BACE1 as the β-secretase, and researchers continue to struggle with the development of effective BACE1 inhibitors that are active in the CNS. The large BACE1 active site requires the identification of large compounds for potent BACE1 inhibition that also readily penetrate the blood–brain barrier (BBB)⁵⁴ and are reasonably stable. Unfortunately, the slow progress of the BACE1 inhibitor field is a testament to the fact that such molecules are relatively rare. Furthermore, it is becoming increasingly apparent that BACE1 has alternative cleavage substrates and may have subtle physiological roles in humans (for example, in peripheral nerve myelination and synaptic transmission)^{55,56,57,58,59,60,61,62,63,64}. Nevertheless, some BACE1 inhibitors have progressed to early clinical trials⁶⁵. As with γ-secretase inhibition, problems with developing small-molecule BACE1 inhibitiors may ultimately be circumvented by indirect approaches to BACE1 inhibition — for example, by modulating regulatory mechanisms that control BACE1 expression^{66,67,68,69,70,71,72} or by taking advantage of the fact that BACE1 activity is optimal in acidic cellular compartments⁷³. Despite the various drawbacks to the approach of secretase inhibition, γ-secretase inhibitors, γ-secretase modulators and β-secretase inhibitors continue to be actively pursued as drug targets for Alzheimer's disease therapy in the hope that the eventual benefits might outweigh the risks.

Mobilizing the immune system: Aβ immunotherapy. An alternative approach to protein aggregation therapeutics is to enhance the degradation of the aggregating protein or the aggregates themselves. Manipulating the immune system for the purpose of enhancing Aβ clearance has been pursued as a therapeutic approach for Alzheimer's disease since the turn of the millennium. This was when several studies reported dramatically reduced Aβ levels and plaque pathology and/or cognitive improvements upon active immunization of APP-transgenic mice with full-length Aβ peptide^74,75,76, Aβ peptide fragments⁷⁷ and passive transfer of Aβ-specific antibodies^78,79,80,81. Based on these studies and encouraging results from Phase I trials, active Aβ immunotherapy in humans subsequently progressed to a widely publicized Phase II clinical trial in 2001. Unfortunately, this trial was halted in January 2002 owing to the development of sterile meningoencephalitis in some patients⁸².

Nonetheless, a follow-up study on a small subset of patients from the Phase II immunotherapy trial suggested that individuals that had generated high Aβ antibody titres exhibited a decreased rate of cognitive decline compared with placebo-treated individuals⁸³. However, these results are controversial and follow-up analyses have questioned the validity of this positive interpretation.

As a complementary strategy, second-generation Aβ immunotherapies are attempting passive transfer approaches. A monoclonal Aβ antibody from Pfizer is currently in Phase II clinical trials. Additionally, bapineuzumab from Elan–Wyeth caused a significant delay in cognitive decline, at least in a subset of individuals¹⁷. Therefore, the outlook on Aβ immunotherapy remains hopeful.

Interestingly, preliminary studies have shown that intravenous treatment with human immunoglobulins, an approved treatment for immune deficiencies and autoimmune disorders, also seems to be effective in reducing Aβ levels and cognitive decline in patients with Alzheimer's disease^84,85. Although the mechanism of action remains unclear, these studies raise the interesting question of whether immunotherapy for Alzheimer's disease needs to be directed against the Aβ peptide, or whether general immunomodulation (possibly through Fcγ-mediated engulfment of amyloid deposits by microglia, or through poorly understood modulation of cerebral cytokine signalling)⁸⁶ might be sufficient to enhance the clearance and degradation of pathological peptides in the brain. If so, this finding would have important implications for the treatment of other protein aggregation diseases.

Curiously, despite the initial success of Aβ immunotherapeutic methods, the mechanisms by which these antibodies function to clear amyloid and/or to elicit cognitive improvement remains an area of intensive debate. Three scenarios have been envisaged⁸⁷. First, immunotherapy might enhance microglial phagocytic activity, which could be directed against the soluble monomeric and/or oligomeric forms of the amyloid-forming protein. This would reduce the amount of protein that is available to aggregate. Alternatively, microglia might directly disassemble the insoluble deposits, as microscopic analysis suggests that Aβ antibody administration promotes microglial uptake of insoluble deposits⁷⁸. However, the stability of deposits in the brain argues against this hypothesis^81,88. Second, the 'peripheral sink hypothesis' postulates that antibodies mediate their effects by sequestering amyloid-forming proteins in the periphery (based on the theory that peripheral plasma proteins are in equilibrium with CNS amyloid-forming proteins )^79,80. In this scenario, antibodies may bind to amyloid-forming proteins and act as a physical barrier, preventing amyloid-forming protein from crossing the BBB, or antibodies may stimulate immune-mediated uptake of amyloid-forming proteins. In either case, the proposed net result is efflux of protein out of the CNS into the periphery, where the protein is presumably degraded by circulating macrophages. Third, antibodies might inhibit the aggregation process by preventing the recruitment of monomeric species to aggregated moieties, either by binding to the monomers or the aggregates themselves. Evidence that Aβ antibodies, which bind amyloid plaques, are the same antibodies that effectively ameliorate plaque pathology⁷⁸ supports this hypothesis⁸⁹.

Regardless of the exact immunotherapeutic mechanism of action, the net result is likely to be enhanced clearance of amyloid-forming proteins, as non-aggregated and aggregated states are in equilibrium and soluble forms of the protein are more accessible to clearance and degradation mechanisms than insoluble forms. Several of these mechanisms probably act simultaneously, with the relative contribution of each to net reduction in aggregation depending upon the specific properties of the amyloid-forming protein being targeted, the stage of disease progression and the antibody being used.

All aggregates are not created equal: toxic oligomers and neurodegeneration. An important assumption underlying therapeutic strategies aimed at reducing protein aggregation is that protein aggregates are toxic to cells. But is this the case? It has long been known that the number of amyloid plaques in Alzheimer's diseased brains correlates poorly with the degree of dementia⁹⁰. More refined measurements that take into account levels of total Aβ aggregates (or amyloid load, of which plaques are only a subset) seem to be better predictors of cognitive performance. One theory that has emerged to explain this observation is that the large, insoluble lesions observed upon histological examination are benign 'tombstones' and that the pathological mediators are highly toxic oligomeric intermediates, which exist in a kinetic state that is between the monomeric and the insoluble, aggregated state. Oligomers derived from cultured cells consist primarily of Aβ dimers and trimers^91,92, whereas oligomers isolated from brain tissue consist of a mixture of dimers, trimers and higher-order aggregates^93,94. Interestingly, a reduction in higher-order Aβ aggregates (and not Aβ trimers) was co-incident with cognitive improvement in APP-transgenic mice receiving scyllo-inositol (ELND005) treatment⁹⁵. A large body of evidence has accumulated over the past decade establishing a role for Aβ oligomers in synaptic dysfunction and neurotoxicity (reviewed in Refs 96, 97). In addition, more recent studies have further implicated this Aβ species in other aspects of Alzheimer's disease pathogenesis, including tau phosphorylation⁹⁸, long-term depression⁹⁹, retrograde trafficking of brain-derived neurotrophic factor¹⁰⁰ and insulin signalling^101,102. With regards to therapeutics, it is hoped that traditional approaches will clear oligomeric species with as much (or more) efficacy as they clear mature amyloid fibrils.

However, the theory of toxic oligomers has other interesting implications for therapeutics. For example, it implies that therapies do not have to clear protein aggregates to be beneficial. Indeed, studies in APP-transgenic mice indicate that certain Aβ immunotherapies lead to cognitive improvement without any detectable clearance of Aβ⁸¹. In the future, conformation-specific antibodies that recognize and neutralize the action of these toxic species in addition to promoting Aβ clearance could be the most effective type of antibodies in eliciting cognitive improvement in Alzheimer's disease. Interestingly, some of these antibodies recognize a common motif in diverse protein aggregates¹⁰³, and so could be broadly useful for the treatment of many aggregation disorders.

Inside or out? The role of intracellular Aβ in amyloid pathology. As with any new discovery, the idea of toxic oligomers has ultimately raised as many questions as it has answered. For example, how and where do Aβ oligomers induce neurotoxicity? Amyloid plaques seem to be extracellular lesions; however, if soluble oligomers are the primary toxic species, they could mediate their toxic effects inside the neuron. Although the majority of Aβ is eliminated from the cell through the secretory pathway (Fig. 3), supporting the idea that Aβ exerts its effects extracellularly, there is mounting evidence that intracellular pools of Aβ could also have an important role in the disease process. The disease-associated isoform of Aβ, Aβ42, seems to be more prone to intracellular accumulation than Aβ40. Also, intracellular Aβ occurs most frequently in the hippocampus and entorhinal cortex (ERC), which are the brain regions to be affected first in Alzheimer's disease¹⁰⁴. Expression of the ε4 allele of apolipoprotein E (APOE4), a major genetic risk factor for Alzheimer's disease, also increases intracellular Aβ¹⁰⁵. More recent studies have found that increased synaptic activity reduces intracellular Aβ¹⁰⁶, and overexpression of neprilysin, which significantly ameliorates plaque pathology in APP-transgenic mice, also decreases intracellular Aβ¹⁰⁷. Therapies aimed at reducing intracellular Aβ levels therefore warrant further investigation. The ability of current methods to influence intracellular Aβ levels might be indicative of their efficacy. Indeed, Aβ immunotherapy can reduce intracellular Aβ in addition to its other effects, even when the antibody is not internalized¹⁰⁸. Interestingly, Aβ immunotherapy has recently been used to reduce intracellular Aβ and attenuate motor impairment in a mouse model of inclusion body myositis — a disease in which intracellular Aβ accumulation is a prominent feature¹⁰⁹.

The idea that soluble oligomeric intermediates and intraneuronal pools of Aβ have key roles in pathogenesis is steadily gaining support, indicating that traditional ideas about amyloid are rapidly evolving. Nevertheless, the idea that amyloid deposits are entirely non-pathological seems unlikely. In the case of Alzheimer's disease, it is well established that neurons surrounding neuritic plaques have an abnormal appearance. It is likely that many forms of Aβ can contribute to neuronal dysfunction, as proposed by one recent study investigating neuronal Arc expression in the vicinity of extracellular and intracellular pools of Aβ¹¹⁰. The extent to which each Aβ pool contributes to the disease process at different stages will undoubtedly be the subject of future studies.

Infectious amyloid: the prion diseases

Prion diseases, also known as transmissible spongiform encephalopathies, include Creutzfeldt–Jakob disease, kuru, fatal familial insomnia and Gerstmann–Sträussler–Scheinker syndrome in humans; scrapie in sheep; bovine spongiform encephalopathy in cattle; and chronic wasting disease in cervids¹¹¹. The prion protein (PrP) is normally present in its native conformation (PrP^C). However, in all prion diseases, the protein is present in an abnormal conformation (PrP^Sc) that accumulates and forms deposits around neurons (Fig. 4a,b). A unique feature of prion diseases is that they are transmissible among humans and across species. Although a seeding mechanism with subsequent propagation of amyloid fibrils is thought to occur generally in amyloidoses, prion diseases seem to be the only amyloidoses that are genuinely infectious diseases (with the possible exception of AA amyloidosis)^112,164. Remarkably, the infectious agent in all prion diseases appears to be composed exclusively of PrP^Sc aggregates, although the conversion process from PrP^C to PrP^Sc may require other cellular cofactors. Upon association with the PrP^Sc conformer, PrP^C undergoes a dramatic shift in secondary structure from a protein composed of ∼45% α-helices and relatively few β-sheets to a protein composed of ∼30% α-helices and ∼45% β-sheets. PrP^Sc aggregates are unusually resistant to degradation, and proteinase K digestion (which entirely degrades PrP^C, but only partially degrades PrP^Sc) is the most common biochemical method of distinguishing PrP^C from PrP^Sc in tissues and cells (Fig. 4c,d). The unusual properties of prion amyloid propagation have presented considerable challenges for the development of effective therapeutics to treat prion diseases, as well as opportunities to explore unique therapeutic modalities.

Figure 4: Characteristics of prion disease.

a | A human cortical section from a patient affected by prion disease, stained with prion protein (PrP)-specific antibodies. All prion diseases are marked by this widespread accumulation of insoluble deposits in the brain consisting of the human prion protein (PrP^C) folded in an abnormal conformation (PrP^Sc). b | A haematoxylin- and eosin-stained human cortical section from a patient affected by prion disease. As seen here, another pathological hallmark of prion diseases is the formation of large vacuoles, or 'spongiosis', in the affected brain tissue. It is because of this lesion that prion diseases are also known as transmissible spongiform encephalopathies. c | Proteolytic degradation followed by PrP^Sc-specific immunostaining (PrP^Sc histoblot) of coronal mouse brain sections from C57BL/6 (prion-susceptible) and Prnp^−/− (prion-resistant) mice infected with a mouse-adapted prion strain, Rocky Mountain Laboratory passage 5 (RML5). d | A PrP-specific immunoblot of brain homogenates digested by proteinase K (PK) and separated by SDS–PAGE, taken from RML5-infected C57BL/6 (PrP^Sc) or uninfected (PrP^C) mice. PrP^Sc and PrP^C have identical amino-acid sequences and are thought to differ only in secondary structure. Although the crystal structure of PrP^Sc has not been solved, its conformation can be distinguished from that of PrP^C using biochemical techniques. For example, PrP^Sc is partially resistant to protease digestion, whereas PrP^C is fully sensitive to proteolytic degradation. MW, molecular weight. Panel c is reproduced, with permission, from Ref. 163 © (2007) Elsevier Science.

The immune system revisited: prion immunotherapy. The immune system has a complex role in the course of many amyloidoses, and prion diseases are no exception. The peripheral immune system is crucial for extraneural prion replication and the spread of prions to the CNS. Mice that lack various components of the immune system — for example, B cells and follicular dendritic cells (FDCs) — are much less susceptible to prion disease upon peripheral infection^{113,114,115,116}. Conversely, pro-inflammatory conditions increase susceptibility to prion infection^117,118, promote peripheral prion deposition^119,120 and accelerate prion pathogenesis^{113,114,115,121}.

However, the role of the immune system in prion replication in the CNS is less clear. Although extensive astrogliosis is a hallmark of prion diseases, transgenic mice in which various components of inflammatory pathways had been deleted were not more susceptible to prion infection when inoculated intracerebrally, indicating that many key immune pathways do not contribute to prion replication in the CNS. However, depletion of microglia led to a nearly 15-fold increase in prion infectivity in a slice culture model of prion disease¹²². This suggests that immune-mediated phagocytic activity in the CNS might be involved in degrading prion aggregates and prolonging neuronal survival. In either case, as with Alzheimer's disease, targeting the immune system (both peripherally and in the CNS) has become a prime therapeutic strategy against prion disease pathogenesis.

Targeting peripheral immune cells. Unlike other neurodegenerative diseases, prion diseases can spread from the periphery to the CNS (and possibly vice versa). The immune system has a pivotal role in this process. FDCs in secondary lymphoid organs accumulate PrP^Sc in both acquired and inherited forms of the disease. In the case of most acquired prion diseases, it is from these sites that PrP^Sc is subsequently transmitted to the CNS¹²³. Consequently, targeting FDCs has proved to be an effective method of preventing the spread of PrP^Sc from the periphery to the CNS in animal models. For example, lymphotoxin-β receptor (LTβR) signalling is required for FDC maintenance and might even be directly involved in enabling prion replication at the cellular level^{116,124,125,126}. Blocking peripheral LTβR signalling in mice by administering soluble LTβR–immunoglobulin fusion protein transiently depletes FDCs and prevents PrP^Sc accumulation in peripheral lymphoid organs (for example, the spleen) and subsequent transmission of PrP^Sc to the CNS^124,165. The drawback to this approach is that it requires the time of peripheral prion infection to be known, which is not possible in many cases of acquired prion disease. However, this approach may prove useful to medical personnel for cases of accidental prion infection in the clinic.

From Aβ to PrP: PrP immunization. The initial success of Aβ immunotherapy for the treatment of Ab pathology in APP-transgenic mice inevitably led to the investigation of whether the immune system could also be mobilized to clear PrP^Sc aggregates. Both active and passive PrP immunotherapy have been tested for efficacy in the treatment of prion disease. These approaches have revealed additional challenges facing the immunotherapeutic approach to amyloid diseases. In the case of active PrP immunization, the immune systems of wild-type mice were largely tolerant to PrP as it is ubiquitously expressed. The result was the generation of low PrP-specific antibody titres and minimal or non-significant increases in survival time when used to treat intraperitoneal prion inoculation^{127,128,129,130,131}. Transgenic expression in mice of a PrP epitope raised in Prnp^−/− mice prevented disease following intraperitoneal prion inoculation and showed that PrP immunotherapy is theoretically an effective strategy for combating prion infection¹³². Despite these encouraging results, subsequent attempts of passive PrP immunization have required unrealistically high levels of PrP antibodies to elicit any meaningful effect on the survival of mice following intraperitoneal inoculation with prions. Furthermore, passive immunization has so far been ineffective in slowing the rate of disease progression in mice following intracerebral prion inoculation or in mice showing clinical signs of disease¹³³.

Another potential problem with the PrP immunotherapy approach was highlighted by the observation that certain PrP-specific antibodies are capable of triggering a pro-apoptotic signalling cascade in neurons that is reminiscent of PrP^Sc infection or deletion of the central domain of PrP^C (Ref. 134). Fortunately, not all PrP-specific antibodies seem to have this capability. Nevertheless, this study urges caution to those investigating PrP immunotherapeutics. In its current state, PrP immunotherapy is still far from entering the clinic. However, new innovations, such as antibodies that are able to penetrate the BBB more readily or PrP–antibody fusion proteins designed to specifically target the immune system to PrP^Sc deposits, may dramatically increase the efficacy of PrP immunotherapy.

Preventing amyloid formation: aggregation inhibitors. In recent years, there has been tremendous progress in developing therapeutic strategies aimed at reducing the production or enhancing the degradation of amyloid-forming proteins. However, in the case of prion diseases, in which the amyloid-forming protein is not produced from enzymatic activity and the pathological aggregates are unusually resistant to degradation, it is likely that alternative strategies will have to be pursued. One viable approach might be to physically interfere with the aggregation process. As the process of aggregation itself is not entirely understood, this is a difficult approach at present. Nonetheless, numerous compounds, peptides and nucleic acids have been investigated for their ability to prevent PrP^Sc aggregation and slow the progression of prion disease. These types of compounds and their efficacy in different model systems have been extensively reviewed elsewhere¹³⁵. The exact mechanism of action of this type of therapy is not known; however, it is thought that the three-dimensional structure of this class of molecules might enable them to intercalate within growing amyloid structures, preventing further recruitment of monomers and promoting the dissolution of aggregates. Alternatively, these compounds may mimic the amyloid-binding properties of and/or compete with endogenous amyloid-stabilizing proteins. Certain endogenous proteins (and metal ions) associate with and promote the stability of amyloid in vivo, including heparin sulphate proteoglycans, APOE, APOJ (also known as clusterin), α1-antichymotrypsin, complement factors, serum amyloid P (SAP), copper and zinc.

A potential concern with this class of compounds is that mimicking the properties of amyloid-stabilizing proteins could also promote aggregation of amyloid-forming proteins. This may partially explain the conflicting results on the efficacy of these compounds on PrP^Sc aggregation across studies and in different model systems. An interesting alternative approach may be to deplete endogenous levels of amyloid-stabilizing proteins. For example, successful depletion of SAP from serum and cerebrospinal fluid of patients with Alzheimer's disease has been achieved using subcutaneous injections of the compound CPHPC (R-1-[6-[R-2-carboxy-pyrrolidin-1-yl]-6-oxo-hexanoyl] pyrrolidine-2-carboxylic acid), which binds SAP and targets this peptide for hepatic degradation¹³⁶.

Another class of drugs that may prevent the aggregation of PrP^Sc are the antimalarial compounds quinacrine and chloroquine^137,138,139. These drugs localize to lysosomes, the cellular compartment in which PrP^Sc accumulates, and presumably interfere with the interaction of PrP^C with PrP^Sc and the formation of PrP^Sc aggregates. However, antimalarial compounds also interfere with lysosomal function. As lysosomes might be involved in degrading endogenous PrP^C, these drugs have the potential to prolong PrP^C half-life, thereby promoting PrP^C–PrP^Sc interaction and the conversion of PrP^C to PrP^Sc. This might explain why some studies have reported an accelerated disease course and enhanced PrP^Sc deposition following administration of antimalarial compounds^140,141. Despite these conflicting data about the efficacy of antimalarial drugs in preventing PrP aggregation, quinacrine was recently tested in a clinical trial of patients with Creutzfeldt–Jakob disease. The drug was reasonably well tolerated, but it did not have a significant effect on the course of prion disease¹⁴².

A viable alternative strategy could be to develop agents that compete with endogenous amyloid-stabilizing proteins using a compound or peptide that cannot be incorporated into aggregates and has no known amyloid-stabilizing properties. The effectiveness of this approach has been demonstrated by transgenic overexpression of a soluble form of PrP (PrP-Fc), which significantly delayed incubation time following prion inoculation¹³². A similar mechanism might underlie the action of several other anti-aggregation compounds. For example, protein X mimetics of PrP are compounds designed to interact with and block PrP amino acids that are thought to comprise the binding site of an unknown endogenous protein required for the conversion of PrP^C to PrP^Sc (Refs 143,144). Compounds known as β-sheet breakers contain sequences from target amyloid-forming proteins plus additional proline residues designed to interfere with amyloid formation¹⁴⁵. Similar compounds have been developed to interfere with the interaction of Aβ and APOE or Aβ and SAP^146,147,148. Additionally, the ability of Congo red to interfere with amyloid formation in some studies raises the intriguing possibility that commonly used amyloid diagnostic or biochemical tools may also have therapeutic applications. An example of one such tool is luminescent conjugated polymers (LCPs), which bind amyloids in a manner similar to Congo red or thioflavin but are sterically more flexible than these compounds, perhaps making them less likely to stabilize amyloid conformations^149,166. Interestingly, compounds that block amyloid aggregation may ultimately reach the clinic for the treatment of systemic amyloidoses (which do not have the additional requirement for compounds to cross the BBB) long before neurodegenerative disorders. For example, two such compounds for the treatment of ATTR amyloidosis (tafamidis from FoldRx¹⁵⁰ and scyllo-inositol from Elan) are currently in Phase II and III clinical trials.

The prion protein as an aggregate receptor: the normal function of PrP^C revealed? Another fascinating implication of the toxic-oligomer theory is that specific cell surface receptors could trigger neuronal dysfunction and/or neuronal apoptotic pathways following binding of oligomeric species. This concept is not new to the prion field: it has long been known that PrP^Sc-mediated neuronal toxicity is completely abrogated in the absence of Prnp expression, even in experimental paradigms in which PrP^Sc can still accumulate^151,152,153.

Intriguingly, a recent study has claimed that endogenous PrP^C may be one of the cell surface receptors mediating synaptic dysfunction associated with oligomeric Aβ¹⁵⁴. In this study, PrP^C was identified in an expression cloning screen for proteins that could enhance oligomeric Aβ binding to the surface of COS-7 cells. Furthermore, it was shown that the absence of PrP expression partially rescued Aβ oligomer-mediated deficits in long-term potentiation. Conformation-specific antibodies that recognize oligomeric forms of Aβ also seem to be capable of recognizing the oligomeric forms of other amyloid-forming proteins, indicating that oligomers of diverse proteins have common structural motifs, similar to mature amyloid fibrils. This observation suggests the interesting possibility that PrP^C might be capable of binding other oligomeric species and function physiologically as a general 'aggregation receptor' for a wide range of amyloid-forming proteins.

Several seemingly unrelated observations support this hypothesis. For example, PrP^C is a glycosylphosphatidylinositol-anchored protein and therefore a permanent resident of lipid rafts^155,156. If PrP^C is an aggregation detector, this could explain why oligomerization increases the affinity of proteins for lipid rafts¹⁵⁷. Furthermore, it has been shown that endosomes containing lipid rafts are preferentially targeted to the lysosomal pathway rather than recycling to the cell surface¹⁵⁸, which implies that PrP^C binding of protein aggregates could target these aggregates for degradation. So, at what point does protein aggregation become toxic? One possibility is that, at sufficient concentrations, extracellular protein aggregates may cause clustering of PrP^C, leading to a conformational change in PrP secondary structure that activates apoptotic pathways. This hypothesis is supported by several lines of evidence indicating that aggregation and/or conformational change of PrP^C triggers a pro-apoptotic signalling cascade^{134,159,160,161,167}. Alternatively, internalization and lysosomal accumulation of insoluble aggregates might also be toxic (Fig. 5). Regardless of the precise mechanism, it is an intriguing idea that PrP^C may serve as a general cell surface aggregation detector, and it could have important therapeutic consequences. First, this theory implies that oligomeric species from diverse aggregation disorders may all trigger cellular toxicity through common pathways. Second, it raises the possibility that PrP^C-targeted therapies might be broadly applicable to many amyloidoses.

Figure 5: Possible roles of PrP^C in oligomer-mediated toxicity.

a | Normally folded prion protein (PrP^C) is a glycophosphatidylinositol -anchored cell surface protein and a permanent resident of lipid rafts. The physiological function of PrP^C is unknown; however, an intriguing possibility is that cell surface PrP^C may bind potentially toxic extracellular oligomeric species and target them to the lysosome for degradation. b | At high, pathological concentrations, toxic oligomers may induce clustering of and/or a conformational change in cell surface PrP^C. This may lead to the direct induction of cell death pathways. Alternatively (or in combination), PrP^C-mediated lysosomal accumulation of oligomers may induce cellular toxicity.

The future of amyloid therapeutics

Although many challenges remain for amyloid therapeutics, the field is steadily moving forward thanks to vigorous and innovative research efforts. Encouraging results from animal models and several promising immunotherapies, secretase inhibitors and anti-aggregation compounds that are currently in clinical trials provide hope that the first disease-modifying treatments for amyloidoses may soon become available.