Protein misfolding in neurodegenerative diseases: implications and strategies

Proteostasis

The term “proteostasis” refers to the integrated activity of cellular mechanisms that regulate protein production, folding, trafficking, degradation, and clearance. Cellular responses to proteotoxic stress, like the heat shock response and the unfolded protein response (UPR) involve large-scale rebalancing of the proteostatic network via transcriptional regulation of both chaperones (e.g., Hsp70, Hsp90) and non-chaperone proteins (including transcription factors, signaling proteins and receptors, and cell cycle regulators [17]. Post-translational modifications can also radically change the activity of some chaperones [18], and likely also play a key role in proteostasis, although this area remains largely unexplored. During the ageing process, or in diseases associated with misfolded proteins, cells may experience “proteostatic collapse.” Proteostatic collapse is associated with the accumulation of ubiquitinated inclusion bodies (IBs), which are seen in many neurodegenerative diseases [9]. It has been suggested that ubiquitinated aggregates can directly inhibit or clog proteasomes [19, 20]. However, in the case of ubiquitinated Huntingtin (Htt), this does not appear to be the case, nor is ubiquitination required for Htt to accumulate in IBs [21]. Rather, the accumulation of ubiquitinated species in misfolded protein diseases may reflect a global perturbation of proteostasis, in which chaperones and proteasomes are simply overwhelmed with client proteins.

Propagation

A key feature of misfolded protein diseases is the ability of the pathogenic protein species to propagate in a prion-like manner by recruiting normally folded counterparts to adopt pathogenic conformations. Pathogenic amyloids can also spread from neurons to other neurons and neighboring glia to initiate new pathology after injection into the brains of normal animals [22, 23]. Both in vivo and in vitro studies have shown that misfolding of one disease causing protein can induce misfolding of other aggregation-prone proteins [23], and aggregates of different disease proteins may be found in the same patient [24]. Moreover, the accumulation of one species of misfolded proteins can impair the entire proteostatic network, thereby triggering the misfolding of unrelated proteins that would otherwise fold normally [25, 26]. The mechanisms by which misfolded proteins spread from one neuron to another are currently an area of active investigation. New evidence suggests that inter-neuronal spread of misfolded proteins involves 1) activity-dependent secretion by exosomes ([27] and/or 2) chaperone-mediated pathways [28, 29].

Mechanisms of misfolded protein toxicity

In the long term, all neurodegenerative disease proteins produce synaptic dysfunction and loss and, ultimately, neuronal cell death. The precise upstream mechanisms by which different misfolded disease proteins cause neurotoxicity are still unclear, and appear to differ depending on the protein species involved. Misfolded disease proteins appear to act primarily by toxic gain-of-function and/or dominant-negative effects, although loss-of-function effects have also been observed. Direct, acute effects of misfolded proteins on neuronal function have been observed after treating neurons with purified oligomers or transfecting them with expression vectors. To give just a few examples, amyloid-beta, tau, and alpha-synuclein all interfere with synaptic signaling [30–32]; mutant tau disrupts microtubule function and neuronal transport mechanisms [32, 33]; and alpha-synuclein disrupts mitochondrial protein import [32, 34]. In addition, larger aggregates of misfolded proteins may exert toxic effects by binding to and sequestering other cytosolic proteins. For example, proteomic studies of artificial proteins designed to form amyloid-like fibrils showed that the toxicity of these proteins correlated with the ability of their aggregates to engage in aberrant protein interactions and disrupt the cytosolic stress response [35]. Notably, the endogenous cellular proteins sequestered by the amyloid aggregates tended to be relatively large in size and enriched in intrinsically unstructured regions, and many play key roles in essential cellular activities such as transcription, translation and protein quality control. Indeed, another emerging common feature among misfolded disease proteins is their ability to disrupt proteostasis (see more below). More recently, cytosolic aggregates of several different proteins, including artificial β-sheets, fragments of mutant huntingtin, and TAR DNA binding protein-43 (TDP-43) have also been shown to disrupt nucleocytoplasmic transport of both proteins and RNA [36].

In addition to synaptic dysfunction, other cellular changes common to the major neurodegenerative diseases include calcium signaling abnormalities, mitochondrial dysfunction, oxidative stress, and neuroinflammation. These symptoms of cellular distress often occur early in the disease process, and are believed to be a cause as well as a consequence of neurodegeneration. That is, the relationship between the accumulation of misfolded disease proteins and other signs of cellular distress is bidirectional, and in many cases mutually exacerbating. For example, amyloid-β, a-synuclein, and mHtt all cause acute oxidative stress in neurons and/or astrocytes, and impair astroglial anti-oxidant responses [37–40]. Conversely, oxidative stress promotes the aggregation of disease proteins, and contributes to age- and disease-related proteostatic collapse [41, 42]. Similarly, there appearaas a downward spiralling cycle of interactions between protein misfolding and neuroinflammation, which has been most extensively studied for AD. Soluble Aβ oligomers and insoluble Aβ aggregates have been shown to bind to and activate and microglia and astrocytes, stimulating a chronic low level state of neuroinflammation [43]. Several lines of evidence suggest that the pro-inflammatory effects of Aβ, while perhaps helpful in the short-term, ultimately impair the microglial and astroglial function, including their ability to dispose of Aβ and other misfolded proteins [38, 43–45]. The destructive consequences of the neuroinflammation provoked by misfolded disease proteins are likely exacerbated by ongoing, age-related senescence of the immune system senescence [46, 47].

Targeting production, misfolding and aggregation

The development of drugs targeting protein misfolding or aggregation has been challenging due to the lack of certainty about which form/s of a given disease protein is primarily responsible for the disease. In the case of amyloid-β (Aβ), it was originally thought that fibrils and plaques were the pathogenic species in Alzheimer’s disease, but more recent studies point to aggregation intermediates (oligomers and soluble protofibrils) as the primary culprits; similar findings have emerged with respect to different species of α-synuclein in Parkinson’s disease [4, 48]. The situation is further complicated by the existence of the variety of intermediate species that exist during the folding and oligomerization processes. Recent studies have demonstrated that aggregated fibrils of tau, α-synuclein, and Aβ exist in different conformational variants, or ‘strains’, that have different propagation properties and different levels of neurotoxicity [49–51].

Identifying key toxic species of misfolded proteins has been challenged the inability of conventional biochemical analytic methods to detect and characterize intermediate species. For example, denaturing gel electrophoresis (SDS-PAGE) has been shown to alter the oligomerization state of Aβ42 oligomers [52]. Recently, ion mobility spectrometry-mass spectrometry (IMS-MS) and nuclear magnetic resonance (NMR) spectroscopy have been used to analyze the folding and aggregation of amyloid proteins in solution and to identify inhibitors of these processes [53–55]. As an alternative to preventing the initial misfolding and aggregation of amyloids, another approach now being explored is to stabilize mature fibrils to prevent their prion-like propagation [48, 56].

In cases where the relative pathogenicity of various misfolded protein species is unknown, one strategy would be to intervene therapeutically as far upstream as possible in the protein synthesis pathway: i.e., at the level of protein translation, cleavage, or post-translational modification. In theory, targeting early steps in the processing pathway would provide the highest degree of therapeutic specificity and eliminate toxic gains or losses of function caused by misfolded or aggregated forms of the protein, while preventing the propagation of abnormal folding and aggregation. Protein cleavage and post-translational modification targets are being actively explored for Aβ using Beta Amyloid Cleaving Enzyme (BACE) inhibitors (some of which are now in clinical trials) and for tau using inhibitors of tau phosphorylation (e.g., glycogen synthase 3β inhibitors) and acetylation [57–60].

Targeting chaperones

Chaperones are another possible target class for therapeutic intervention in protein misfolding disease states. As chaperones are involved in all aspects of proteostasis, they offer potential therapeutic entry points to each step in the processing of a pathogenic protein. There are over 200 different chaperone proteins expressed in the mammalian brain [61], and different cell types express different sets of chaperones and co-chaperones [62, 63]. Cell type-specific expression of chaperone subsets may help to explain why some misfolded proteins are toxic in one cell type and not in others, and also presents opportunities to develop drugs targeting neuron- or glial cell-specific chaperones. However, the sheer number of chaperones and the diversity in their mechanisms of action also presents challenge to therapeutics development; we still have limited knowledge of which chaperones interact with which disease proteins and how. Some clues have been offered by links between mutations in specific chaperones and hereditary forms of certain neurodegenerative diseases. For example, mutations in Hsp70 and Hsp40 have been linked to Parkinson’s disease [64, 65], and mutations in the co-chaperone valosin-containing protein (VCP) have been found in ALS [66]. In addition, to date over 20 different chaperones have been found to confer neuroprotection when over-expressed in cell or animal models of various neurodegenerative diseases, and in many cases individual chaperones appear to protect against several different disease proteins [67].

One approach to therapeutically targeting the chaperone system has been to develop small molecule inhibitors or activators of specific chaperones. Among chaperones relevant to neurodegeneration, Hsp70 and Hsp90 have been the most intensely studied. Hsp70 and Hsp90 have opposing effects on client protein stability: Hsp70 promotes their degradation via the UPS system, whereas Hsp90 stabilizes client proteins and inhibits their ubiquitination. The activities of Hsp90 and Hsp70 are tightly linked via HSF1. Hsp90 inhibitors typically activate HSF1, which in turn induces Hsp70 [68]. A variety of small molecule drugs have been developed that inhibit Hsp90, activate Hsp70, or both, and have been shown to reduce the formation of disease protein aggregates, reduce cellular toxicity, and improve neurological phenotypes in cellular and animal models of SBMA, HD, PD, and AD [62, 67, 69–71]. None of these drugs have yet entered clinical trials for use in neurodegenerative diseases, due to issues of low brain penetration and/or peripheral cytotoxicity, but active effort in this area is continuing [70, 72–74].

A second approach to developing chaperone-based therapeutics has been through protein engineering. For example, it was discovered that the yeast disaggregate Hsp104 has the ability to dissolve in vitro fibrils formed from a variety of neurodegenerative disease proteins, including tau, polyglutamine, Aβ42, α-synuclein and prion protein [75, 76]. However, relatively high concentrations of Hsp104 are needed to dissolve these proteins. Making small changes in Hsp104′s sequence yielded proteins with much higher dissaggregase activity and lower toxicity, including variants that reduced neurodegeneration in a C. elegans model of PD [77]. Other chaperones to which this approach might be applied include the yeast chaperonin Tric, which has the unusual ability to cross cell membranes and has been shown to protect against Htt toxicity [16], and metazoan chaperones known to have disaggregase activity (e.g., Hsp110, Hsp70, and Hsp40) [78].

Targeting degradation

Defects in both the UPS and autophagy pathways of protein degradation are often seen in neurodegenerative diseases [79, 80]. For example, many of the gene mutations that cause familial PD encode proteins involved in the UPS and/or autophagy, including PINK-1, Parkin (a ubiquitin ligase), UCH-L1 (Ub carboxy terminal hydrolase L1), DJ-1 (PARK7), and LRRK2/PRAK8 [79, 81, 82]. As with the chaperone system, choosing promising drug targets from the UPS or autophagy pathways is challenging because of the number of proteins involved (there are ~ 500 to 1000 associated just with the UPS system). In addition, for most diseases it isn’t known which form/s of the disease protein are primarily responsible for cellular toxicity. This issue is critical when targeting protein degradation, because the pathway by which a given protein is degraded (e.g., UPS versus autophagy) can vary depending on whether the protein is in the soluble or fibrillar state, and on what specific post-translational modifications it bears [83, 84].

A number of small molecules have been identified that upregulate components of the UPS, promote the degradation of disease proteins and in vitro, and (in some cases) have neuroprotective effects on cultured cells [84–86], but few have yet been shown to be effective in vivo. One interesting exception is rolipram, an agent that stimulates the phosphorylation and activity of the 26S proteasome. Myeku and colleagues [21] showed that 26S proteasome function is impaired in a mouse tauopathy model, and that treating the mice with rolipram treatment improved 26S proteasome function, decreased tau aggregation, and improved cognition.

Inhibition of the mammalian target of rapamycin (mTOR) pathway has proven to be an exceptionally effective approach for stimulating the degradation of neurodegenerative disease proteins. mTOR, a serine/threonine kinase, is a signaling nexus that collects information about ambient levels of resources necessary for cell growth (e.g., nutrients, ATP, growth factors, and oxygen) and up- or downregulates protein synthesis and degradation accordingly. When growth conditions are favorable, mTOR inhibits autophagy by inhibiting the ULK1 complex, which is required for the biogenesis of autophagosomes. Thus, mTOR inhibitors typically have the net effect of stimulating autophagy [87, 88].

Rapamycin and other mTOR inhibitors increase the clearance of abnormal protein aggregates and slow neurodegeneration in both cell and animal models of a variety of neurodegenerative diseases, including AD, PD, spinocerebellar ataxia type 3, and frontotemporal dementia [87, 88]. In most of these cases, mTOR inhibitors have been shown to act at least in part via stimulation of autophagy. A recent study showed that in yeast, inhibiting mTOR also produces a rapid, coordinated upregulation of proteasomes and their 19S regulatory chaperones [89]. Thus, the efficacy of mTOR inhibitors in clearing a variety of neurodegenerative disease proteins may be due to the ability of these drugs to upregulate both the proteasomal and autophagic routes of protein degradation.

Several mTOR-dependent activators of autophagy, including the natural compounds curcurmin and resveratrol, are currently in clinical trials for treating neurodegenerative diseases. However, mTOR is a multifarious protein that regulates many cellular processes in addition to protein degradation, and in clinical trials to date wmTOR inhibitors have caused unpredictable and undesirable side effects. Autophagy can be stimulated mTOR-independent mechanisms [79, 87, 88], and a number of compounds, including FDA-approved drugs, have now been shown to stimulate clearance of abnormal proteins and confer protective effects in cell or animal models of neurodegenerative disease [87, 88, 90, 91]. It has been suggested that these compounds might be used in conjuction with mTOR inhibitors to maximize therapeutic benefit and minimize side effects [87, 90].

Targeting extracellular clearance

One of the best-explored examples of targeting clearance of misfolded proteins has been the use of antibodies to promote clearance of Aβ. These antibodies are thought to operate by either or both of two mechanisms: 1) by penetrating the blood brain barrier (BBB) to bind Aβ in the extracellular space, and 2) through a”peripheral sink” effect [92]. Almost all misfolded proteins show some extracellular leakage [93], and an advantage of targeting extracellular misfolded proteins is that it can theoretically be accomplished by highly selective antibodies. For example, monoclonal antibodies have been developed to Aβ and α-synuclein that show >1000-fold higher affinities for the oligomeric versus monomeric forms of the proteins [94–96].

Current challenges in the use of Aβ antibodies include low rates of BBB penetration [97], nonspecific engagement of Aβ and uncertainty as to which antibodies engage clinically relevant forms of Aβ [98–100]. Detailed structural studies of how different antibodies interact with specific epitopes in the Aβ molecule are now underway [101] and help inform antibody design and epitope-targeting in the future. An alternative method of improving the clearance of toxic misfolded species is to harness endogenous mechanisms of protein clearance into the extracellular space. For example, LXR beta receptor agonists, which promote clearance of Aβ into the extracellular space by the ABCA1 transporter, have shown therapeutic effects in AD mouse models [102, 103]. Yet another potential avenue for new therapies targeting extracellular clearance of disease proteins is the use of non-antibody scaffold drugs [104].

Rebalancing the proteostatic network

The ability of the cell’s proteostatic machinery to counter proteotoxic stressors deteriorates with age, and is further compromised by mutations and other disease conditions that lead to the accumulation of misfolded proteins [16]. Thus, another potential approach to the development of therapeutics would involve large-scale rebalancing of the proteostatic network. Indeed, the efficacy of mTOR inhibitors may reflect their ability to provoke large-scale rebalancing of protein synthesis and degradation pathways. Another attractive target in this regard is heat shock factor 1 (HSF1), a transcriptional activator that helps coordinate the heat shock response. The heat shock response (and other proteotoxic stress responses) diminish with age and in neurodegenerative disease [105]. In addition, it was recently shown that HSF1 degradation is abnormally elevated in mouse and human α-synucleinopathy [106]. Over-expression of human HSF1 has been shown to be neuroprotective in cell models of neurodegenerative diseases [107, 108], to reduce polyglutamine aggregate formation and prolong lifespan in a mouse model of HD [17] and to reduce pathogenic androgen receptor accumulation and neurotoxicity in a mouse model of spinobulbar muscular atrophy [109]. Small molecule activators of HSF1 have now been identified and shown to have neuroprotective effects in cell or animal models of neurodegenerative diseases [107, 108].

Agents targeting HSF1 or other master regulators of the proteostatic network have the advantages of being fast-acting and relatively agnostic to the identities of the misfolded proteins involved in a given neurodegenerative disease and to their mechanisms of aggregation and toxicity. The effects of such drugs are hard to predict, however. For example, the induction of the heat shock response actually exacerbates Htt IB formation in a cellular model of HD [110]. In addition, it has been pointed out that, under normal physiological conditions, the heat shock response and other proteotoxic stress response pathways are activated only transiently, and that multiple cellular mechanisms are in place to limit and down-regulate these responses [111]. Consistent with those facts, an Hsp90 inhibitor that induces the heat shock response in HD model mice was found to provide short-term beneficial effects, but those benefits proved transient [112]. Yet another potential issue with HSF1 activators is that HSF1 promotes tumorigenesis and is activated in a broad range of highly malignant human cancers [113, 114]. This issue is not necessarily unsurmountable, however, as it has also been shown that the HSF1 drives a different transcriptional program and stimulates different sets of cellular processes in cancer cells (including proliferation, invasion, and metastasis) than it does in normal cells [113]. The ability of HSF1 to activate distinct transcriptional programs in cancer cells versus normal cells is thought to result in part from differences in post-transcriptional modifications to HSF1 in the different cell types [113, 114], which in turn raises the possibility that the neuroprotective effects of HSF1 could be harnessed separately from its tumorigenic ones.