Introduction

Alzheimer’s disease (AD) is an incurable, progressive brain degeneration that results in a loss of higher cognitive function, memory impairment, language deterioration, depression, and other manifestations. This disease is named after a German neuropathologist, Alois Alzheimer, who first presented the case of his patient Auguste Deter in 1906 [1, 2]. An increase in human life expectancy has coincided with a rapid increase in neurodegenerative disease incidence in recent years. A recent statistical study projected that in 2050 approximately 13.8 million people in the USA will suffer from AD unless preventive measures are taken. AD has gained widespread attention because of its high economic costs, which have reached $400 billion per year in the USA alone, as well as the social costs, which are more difficult to quantify [3, 4].

AD-related pathological markers include a progressive death of neurons in specific areas with an accumulation of intracellular neurofibrillary tangles (NFTs) and extracellular depositions of amyloid plaques (APs). NFTs are composed of the misfolded hyperphosphorylated microtubule-associated protein Tau (MAPT or Tau), whereas APs are extracellular deposits of misfolded and aggregated amyloid-beta peptides (Aβ). Because both NFTs and APs are persistently found in areas with severe neuronal death, these proteins were considered to be the main cause of neuronal loss and the emergence of dementia, which is a crucial symptom of AD; however, numerous drug trials based on these proteins have failed to provide a useful AD therapy [5, 6]. A postmortem study demonstrated that the misfolded protein accumulation is a shared pattern in many neurodegenerative diseases, including AD [7, 8]. Therefore, it was concluded that an accumulation of misfolded proteins is the potential cause of neurodegeneration in AD.

Here, we review recent global gene expression and proteomic studies of AD brains versus normal brains and in AD-affected regions versus unaffected regions. The obtained data underline the crucial role of protein regulation systems, especially the ubiquitin-proteasome system (UPS), as well as molecular chaperones and co-chaperones. These interconnected systems reveal how AD pathology commences at the molecular level, long before the development of any symptoms. This review also discusses potential areas for AD research that focus on protein regulatory mechanisms as the basis for novel AD therapeutics.

AD Development in the Brain

The development of AD pathology is not isolated to one specific brain region; rather, it spreads throughout nearly the entire brain in a predictable pattern [9]. The earliest area affected by NFTs in AD is the medial temporal lobe, beginning at the transentorhinal region [9–12]. Subsequently, the limbic system, which is composed of the hippocampus and amygdala, followed by the parahippocampal area and frontal lobe exhibit NFT accumulation [12]. AD lesions are found in the most susceptible regions even at the presymptomatic stage. This fact was further confirmed by magnetic resonance imaging (MRI) scans that indicate the expansion of cerebrospinal fluid (CSF), which correlates with brain volume loss in the temporal lobe area [13]. Moreover, presymptomatic individuals and mildly affected patients show a significant rate of volume loss in the hippocampal area. As the disease progresses, the inferior and lateral temporal lobes begin to show accelerated atrophy rates. This observation suggests that atrophy begins from the medial temporal lobe area and spreads to the adjacent inferior and temporal lobes [14]. The mechanism of how misfolded proteins spread to other parts of the brain remains elusive; however, we address the hypothesis that this mechanism involves a prion-like transmission [15].

Genetics of AD

For a small proportion of cases (~5 %), AD is caused by a genetic mutation in the genes involved in processing AD hallmark proteins, including the Aβ precursor protein (APP) and presenilin-1 and presenilin-2 (PS1 and PS2, respectively) [16, 17]. The genetic form of AD is referred to as familial AD or early-onset AD (EOAD), whereas the sporadic form of AD is referred to as late-onset AD (LOAD) [18]. LOAD is the most prevalent type of AD and is likely caused by normal aging [19, 20] and its associated consequences, such as increased protein misfolding, oxidative stress, and perturbed cell regulation [21, 22]. Familial AD or EOAD is often diagnosed at ages younger than 65 years, whereas LOAD is commonly diagnosed in patients older than 65 years [2, 17, 20].

A possible link between APP and AD emerged in the late 1980s [23–26]. The evidence for a pivotal genetic role of APP in the progression of AD pathology stemmed from the shared APs with Down’s syndrome [27, 28]. Amino acid (aa) sequencing of Down’s syndrome-derived Aβ indicated results similar to AD, which suggests that genetics plays a role in the formation of APs [28, 29]. Down’s syndrome is characterized by trisomy of chromosome (chr) 21; as a consequence, the extra dose of the APP gene contributes to the lifelong overexpression of Aβ [30, 31]. This accumulation of Aβ in patients with Down’s syndrome often results in the early appearance of APs, which are an AD hallmark [32, 33]. These data underline the importance of AP formation in the development of AD-like pathology.

Mutant APP770 (V717I) was the first genetic mutation reported in AD [34, 35]. To date, various mutations in the APP gene have been identified in EOAD cases [36–45]. The exact mechanism by which APP mutations promote AD remains unknown. However, it has been demonstrated that mutations in close proximity to secretase cleavage sites can alter the cleavage and processing of APP, which may favor the production of the highly amyloidogenic Aβ42 fragment [46, 47].

Even though the APP mutation is the first AD-related mutation that was discovered, mutations in the PS1/2-encoding genes are the most prevalent genetic cause of EOAD, with the majority of mutations in PS1 [48]. This issue was raised after a reported poor correlation between EOAD and a specific genetic marker in the APP region on chr 21 [49, 50]. It was subsequently discovered that missense mutations in the PS-encoding genes shift the production of Aβ toward the AP-promoting toxic Aβ species [51, 52]. The role of the PS1/2 in the production of Aβ will be discussed later in this review. Extensive studies on PS mutants and gene knock-out models have provided a clear understanding of its role in the development of APs, and various mutations of the PS1/2 genes have been identified and reviewed elsewhere [53]. Another gene linked to AD is apolipoprotein E4 (APOE4). Mutations in the APOE4 gene, while not directly involved in the development of AD, are a genetic factor that increases the risk of AD up to 15-fold [54, 55]. The APOE4 mechanism that affects AD is not in the scope of this review; thus, readers are directed to other review papers [56, 57]. A frame-shifted mutant ubiquitin B (UBB) gene, UBB+, has also been documented to produce an AD phenotype [58, 59]. However, UBB is not directly involved in the current major hypothesis for AD pathogenesis, including the AP or NFT hypotheses. A more comprehensive assessment of UBB+ and its pathophysiological significance in AD is discussed later in this review.

APs

APs, one of the two major hallmarks of AD, predominantly consist of 40 or 42 aa Aβ (Aβ40–42). These Aβ species are commonly found in postmortem AD brains and are neurotoxic [28, 29, 60, 61]. The Aβ peptide is a product of regulated intramembrane proteolysis (RIP) of the APP protein, which is ubiquitously and constitutively expressed in many human cells [62]. APs are also often associated with dystrophic neurites, fibrillar paired helical filaments (PHFs) of hyperphosphorylated Tau, and activated glial cells [63]. The frequency of APs in AD brains led to the development of the “amyloid cascade hypothesis.” This idea has been the de facto foundation of the majority of AD research for the previous 20 years and has been previously reviewed by Selkoe [61, 64]. Diffuse plaques are a subtype of APs that are more granular and less compacted than the APs commonly found in a normal aging brain; furthermore, these plaques do not induce neuronal death or microglia activation [65, 66]. Accordingly, diffuse plaques are considered a presymptomatic sign of AD and can be observed in human brain biopsies and animal models long before the development of a full AD phenotype. To understand how Aβ is formed, we must first briefly discuss APP, the precursor protein of Aβ.

APP

APP is a type 1 transmembrane glycoprotein with hypothesized functions as a cell surface receptor [23, 67]. It has heterogeneous molecular weights between 110 and 135 kDa because of various posttranslational modifications (e.g., N and O glycosylation, phosphorylation, and sulfation) [2, 68]. APP is also a target for alternative splicing, which results in six isoforms (mainly 695, 751, and 770 aa residues) [2, 62, 69–71]. Neurons specifically express a high level of the APP695 isoform, which is not ubiquitously expressed in other cell types [72]. The difference between the APP695 and APP751 isoforms is the presence of a 56 aa Kunitz-type serine protease inhibitor (KPI) motif [70, 73, 74].

A structural analysis of APP has identified several domains; however, the fundamental physiological role of this protein remains unclear. As a transmembrane protein, APP has extracellular, transmembrane, and intracellular regions. Several domains have been identified within these regions, such as the heparin-binding domain (HBD), the Cu(II)-binding domain (CUBD), and the RERMS (Arg-Glu-Arg-Met-Ser) sequence [75–79]. These findings suggest that the physiological function of APP and its derived domains are related to neurotrophic cell signaling, synaptic plasticity, and oxidative defense [80–84]. For example, a recent study demonstrated that the HBD of sAPPα has neuroprotective properties in traumatic brain injury [85].

Regulated Intramembrane Proteolysis of APP

RIP is the processing of APP involving serial cleavages at different sites, which results in the production of several possible fragments, including non-amyloidogenic and soluble APP fragments (sAPPα and sAPPβ), an APP intracellular C-terminal domain (AICD), and amyloidogenic Aβ, as determined by the specific site of APP cleavage. This cleavage occurs at three different sites according to the activities of three different enzymes: α-, β-, and γ-secretases (Fig. 1). RIP primarily occurs intracellularly within the trans-Golgi network (TGN) and the endoplasmic reticulum (ER) compartment and partially on the cell surface [86–89].

Fig. 1
figure 1

RIP of APP. APP has α-, β-, and γ-secretase cleaving sites and a KPI domain (left). APP can be cleaved in a non-amyloidogenic pathway by α-secretase, thus producing sAPPα (top right). Amyloidogenic processing of APP by sequential cleaving with β- and γ-secretases generates sAPPβ, Aβ, and the nuclear signaling molecule AICD (bottom right)

Full size image

The first possible product of RIP is sAPPα [90], which is produced when APP is cleaved by α-secretase. The proposed biological functions of sAPPα are related to neuritogenesis/axonal outgrowth, cell proliferation, glial differentiation, and memory and learning-related processes, such as long-term potentiation (LTP), metal homeostasis, and neuroprotection [81, 91–97]. The detailed physiological role of sAPPα has been extensively reviewed by Chasseigneaux and Allinquant in 2012 [91].

α-Secretase is a zinc metalloproteinase from the A disintegrin and metalloproteinase (ADAM) family that cleaves native APP within the transmembrane region and mediates the shedding of sAPPα from the cell membrane. Because α-secretase cleaves APP within the Aβ region (i.e., K16 of Aβ or aa residue number 687 of APP770), the α-cleavage of APP does not produce an amyloidogenic product [98–100].

ADAM family proteins play an important role in the ectodomain shedding events of diverse biological molecules (e.g., cytokines, growth factors, receptors, and other signaling molecules) and have recently been suggested to be involved in defense mechanisms against cellular prion proteins (PrPs) [101–103]. A comprehensive list of ADAM substrates has been reviewed by Huovila et al. [104]. ADAM family proteins are crucial mediators during the development of the central nervous system (CNS), with effects that largely mediate Notch signaling [105, 106].

The knock-out of ADAM10 in transgenic mice results in multiple defects, including the disruption of Notch signaling, which ultimately induce prenatal death; however, fibroblast cells retain normal APP processing, which suggests ADAM activity is tissue-specific and ADAMs may compensate for one another [105–107]. The overexpression of ADAM10 in transgenic mice results in the competitive cleavage of APP and release of sAPPα, thus reducing the formation of Aβ and its downstream pathology [108]. Furthermore, the in vivo administration of an ADAM10 inhibitor peptide in mice successfully mimicked sporadic AD-like pathology [109].

ADAM17, also known as tumor necrosis factor-α converting enzyme (TACE), cleaves APP and Notch [110]. ADAM17 is also responsible for the shedding of cell adhesion proteins, cytokine and growth factor receptors, and many ligands of the epidermal growth factor receptor (EGFR) [111, 112]. In human primary neuron cultures, an ADAM17 inhibitor prevents the α-secretase activity of ADAM17 for APP shedding [113]. Another ADAM protein, ADAM9, has also been reported to have α-secretase activity for APP [114].

sAPPβ is the second possible product of APP proteolysis, which occurs inside the TGN and plasma membrane [89, 115]. It is produced when native APP is cleaved by β-secretase. The sAPPβ and sAPPα isoforms have identical sequences with the exception that sAPPβ is missing the last 16 aa residues at its C terminus. The functions of sAPPα and sAPPβ are generally similar because their common domains are identical. However, sAPPβ has significantly lower neuroprotective properties and is not involved in LTP because of the missing aa residues at the C terminus [116]. In neurons, the most significant β-secretase is the aspartyl protease beta-site APP cleaving enzyme 1 (BACE1). The cleavage site of BACE1 is in the N terminus of the resulting Aβ, between residues 671 and 672 of APP770. The β-secretase activity is regulated through binding to other proteins, such as the reticulon family proteins (RTNs). A co-localization study demonstrated that RTN3 is a BACE1-binding partner in neurons and negatively modulates BACE1 activity [117], and a gene expression study showed that RTN3 is under-expressed in AD brains [118]. Thus, the downregulation of RTN3 may result in increased Aβ production in neurons [119].

Amyloidogenic Aβ is produced when APP undergoes serial cleavages by β- and γ-secretases. γ-Secretase is a multi-subunit protease complex that consists of PS1 or PS2, nicastrin, anterior pharynx-defective 1 (APH1), and presenilin enhancer 2 (PEN2) [120, 121]. Upon β-/γ-cleavage, the γ-secretase cleaves APP within the transmembrane domain, thus producing an Aβ fragment that ranges from 38 to 43 aa residues in length. It has been suggested that the GxGD motif of PS1 is the active site for APP cleavage specificity [122]. A subsequent study indicated that a missense mutation in L383 (i.e., x residue of the GxGD motif) of PS1 shifted APP cleavage to produce Aβ42 [123]. This mechanism may explain why the modulation of PS1 activity or a mutation in the PS1 gene favors the production of Aβ42.

The majority of Aβs found in APs are in the Aβ40 form [124]; however, it is the Aβ42 fragment that has been highly studied in AD research because of its potency to induce aggregation and neurotoxicity [125]. Aβ42 is considered more hydrophobic than Aβ40 due to its two extra stretches of hydrophobic residues; thus, it more readily forms a β conformation [126, 127]. The elevated hydrophobicity increases the potential of Aβ42 to form insoluble aggregates and consequently promotes its toxicity [128–130]. Under normal physiological conditions, the Aβ42/Aβ40 ratio is maintained at approximately 1:9; however, the level of Aβ42 increases in AD pathology. This perturbed Aβ42/Aβ40 ratio is thought to contribute to the formation of neurotoxic aggregates, which eventually cause cell death [36, 51, 131, 132].

AICD

AICD is another product of γ-secretase cleavage of APP. AICD has properties similar to the Notch intracellular domain (NICD), which is also mediated by γ-secretase [105]. Therefore, it has been hypothesized that AICD may play a pivotal role in neuron signaling pathways. For example, AICD was reported to be transported into the nucleus by an FE65 adaptor protein (otherwise known as APBB1) and subsequently connected to TIP60 (official gene symbol KAT5) to form the AICD-FE65-TIP60 (AFT) complex, which regulates gene transcription [133–137]. The AFT complex is transcriptionally active via the histone acetylation activity of TIP60 [138] with vast downstream target genes, such as lipoprotein receptor-related protein 1 (LRP1) [139], APP [140], glycogen synthase kinase 3β (GSK3β) [141, 142], p53 [143], and EGFR (Fig. 2) [144].

Fig. 2
figure 2

AICD signaling from the membrane to the nucleus. AICD is released through RIP by γ-secretase. One known trigger for this process is TAG1 binding to APP. Released AICD binds to the adaptor protein FE65; it subsequently translocates to the nucleus, where these proteins bind to TIP60 and form the AFT complex to regulate gene expression. GSK3β, APP, p53, and DAB1 are overexpressed by aberrant AICD signaling, whereas the expression of LRP1, EGFR, and neurogenesis genes are repressed

Full size image

The activation of AICD signaling can produce changes in gene expression and cell proteome metabolism, such as the upregulation of DAB1, an AICD interacting protein, in AD frontal cortex brain samples [145]. Moreover, it has been suggested that AICD can mediate neuron-specific apoptosis [146]. AICD also binds to Numb, a Notch inhibitor, and competes with the Notch signaling pathway [147]. Ma et al. showed that AICD release increased after the binding of the transient axonal glycoprotein 1 (TAG1), an extracellular ligand to the extracellular part of APP [67]. This study also demonstrated that released AICD binds to FE65 and negatively modulates neurogenesis. TAG1 is a member of the F3 protein family and is a functional ligand for the Notch receptor [148]. The overexpression of AICD in mice has been reported to induce AD-like characteristics because of the overactivation of GSK3β, which induces the hyperphosphorylation of Tau [149]. This study also found that AICD levels are increased in the brains of AD patients. Thus, aberrant AICD signaling may contribute to the development of AD pathophysiology [150].

Physiological Function of Aβ

Despite its strong link to neurodegeneration in AD, Aβ has exhibited various neuroprotective and neurotrophic effects at suboligomer-promoting concentrations [91, 151, 152]. A study reported that the Aβ1–28 fragment has a neurotrophic effect on hippocampal pyramidal neurons in vitro [152]. Soluble monomeric Aβ40 can induce neurogenesis in neural progenitor cells, whereas Aβ42 favors astrogenesis [153]. Monomeric Aβ40 has been demonstrated to be involved in insulin/insulin-like growth factor (IGF-1) receptor signaling and displays neuroprotective properties [154]. Additional interesting results have been discussed by Calafiore et al. [155]. This group showed that the Aβ25–35 fragment and neurotoxic Aβ42 are not harmful to neural precursor cells (NPC). Furthermore, this study indicated that Aβ increased neurogenesis. A similar result was achieved in another study which suggested that the neurogenic effect of the Aβ42 oligomer is mediated by the mitogen-activated protein kinase (MAPK) pathway [156].

Aβ Pathophysiology

Many hypotheses speculate how APs and Aβ induce cell death, including mechanisms related to oxidative stress, calcium imbalance, mitochondrial damage, the disruption of energy production, the phosphorylation of Tau, and UPS impairment [157–163]. The potential role of oxidative damage as a downstream process in AD pathogenesis has garnered a great deal of interest [164, 165]. This interest is mainly because the brain is the most highly oxygen-metabolizing organ in the body; thus, it is highly susceptible to oxidative damage [166–170]. Moreover, the brain has a high lipid content that is prone to oxidation [165, 167]. Aβ-induced free radical production promotes cell-damaging species that cause oxidized proteins, oxidized DNA, and alterations in brain phospholipids. A relationship between Aβ and free radical-mediated oxidative stress has been proven in vitro and in vivo [158, 159, 171–173]. However, despite the plethora of data that support this hypothesis, some studies have reached a contrasting conclusion [174–177]. Moreover, the role of antioxidants in AD pathology remains unclear [178]. These data suggest that the direct and specific role of Aβ is rather marginal; thus, the correlation between Aβ and neurodegeneration is not as direct as previously suggested [179–181] and has been described with an “airbag” analogy [182]. A more convincing hypothesis suggests that Aβ induces Tau hyperphosphorylation, which enhances Tau aggregation and leads to neurodegeneration [163, 183–189].

AD and NFTs

Another significant AD pathophysiology hallmark is the NFTs, which are filamentous aggregates of the intracellular insoluble hyperphosphorylated Tau protein [190, 191]. Apart from AD, NFTs have been linked to other tauopathies, including frontotemporal dementia and Pick’s disease [192–194]. NFTs were first described in the 1980s and consist of smaller units referred to as PHFs [195]. PHF is a fibril-like aggregation of proteins, which was subsequently discovered to be a wound form of the Tau protein [196, 197].

Tau and Its Physiological Functions

Tau was first characterized in the 1970s from a porcine brain as a protein that binds to tubulin and is essential for microtubule formation [198–200]. In addition to its significance in AD, Tau mutations are responsible for the frontotemporal dementia related to chr 17 (FTDP-17; human Tau is located on chr17q21). Tau is crucial for the formation of a functional microtubule system and is almost exclusively expressed in neurons, particularly in axons [199, 201–203]. Tau is highly soluble and rich in phosphorylation-prone aa residues. Based on its native function, Tau is involved in the stabilization of the axonal microtubule architecture for neurite extension, the establishment of cell polarity, and neuronal axonal transport [204–209]. The role of axonal transport in neurons is critical for the maintenance of axons, synapses, and synaptic activities [210]. Because neuronal function is sensitive to any disturbance of the microtubule-based system, errors in the axonal transport system have been clearly linked to neurodegenerative conditions [211, 212].

The Tau protein has a few notable regions [213, 214]. One region is the tubulin-binding repeat domain, which is highly conserved and spans the C-terminal half of the protein. Some isoforms have three of this repeat, whereas other isoforms have four; thus, these isoforms are often referred to as 3R or 4R Tau. Adjacent to this domain are basic proline-rich repeats, which promote Tau protein binding using tubulin-binding repeats that bind to specific pockets of positively charged prolines attached to the negatively charged microtubule surface. This second domain is in the N-terminal region of Tau, which has either one or two highly acidic sets of 29 aa residues depending on the isoform. This domain is also presumed to protrude because its negatively charged region repels it from the microtubule surface [213]. These various domain combinations generate the six known Tau isoforms that are generated by alternative splicing during mRNA maturation [215]. These isoforms, which range from 352 to 451 aa in length, were characterized from PHFs [215], and all six isoforms have been demonstrated to be aberrantly phosphorylated [216].

AD and the Pathophysiology of Tau Misfolding

Another striking characteristic of the Tau protein is its hyperphosphorylation. As previously described, hyperphosphorylated misfolded Tau causes fibrillization and subsequent aggregation in AD pathophysiology. In the 441 aa residues of each Tau isoform, 79 aa residues of serine or threonine can be potentially phosphorylated [215]. Numerous kinases are linked to Tau phosphorylation, both from proline-directed kinases and non-proline-directed kinases. The major known Tau kinases are GSK3, cyclin-dependent kinase 5 (CDK5), MAPKs, protein kinase A (PKA), protein kinase C (PKC), calcium/calmodulin-dependent protein kinase II (CaMKII), casein kinase II (CKII), MAP/microtubule-affinity-regulating kinases (MARKs), dual-specificity tyrosine-regulated kinase 1A (Dyrk1A), and the AMP-activated kinase (AMPK) [186, 187, 217–222]. As a result of the diverse potential of phosphorylating enzymes, Tau phosphorylation can be a downstream of a myriad of events, including Ca2+-mediated toxicity [223, 224], impaired insulin signaling [225, 226], and brain metabolic deficiency [227]. The hyperphosphorylation of Tau alters its binding to microtubules and promotes its disassociation [162, 228, 229]. Phosphorylated Tau, however, is also subject to various protein phosphatases, such as PP1, PP2A, PP2B, and PP5 [230]. In vitro experiments have demonstrated that Tau dephosphorylation can restore its ability to bind to microtubules [162]. Other than phosphorylation, Tau is a known target for acetylation, deamidation, and ubiquitination [231–234].

Furthermore, the Tau protein is also classified as an intrinsically disordered protein (IDP). Similar to other IDPs, Tau is highly hydrophilic, which easily prevents appropriate folding and a lack of a well-defined three-dimensional structure under physiological conditions [235]. IDPs have been reported to play key roles in complex diseases, including cancer, cardiovascular disease, amyloidosis, neurodegenerative diseases, and diabetes [236]. Interestingly, IDPs are involved in most of the other major neurodegenerative diseases, for example, α-synuclein in Parkinson’s disease (PD), the PrP in transmissible spongiform encephalopathies, polyglutamine in Huntington’s disease (HD), mutant superoxide dismutase 1 (SOD1), and TAR DNA-binding protein of 43 kDa (TDP-43), as well as several others in amyotrophic lateral sclerosis (ALS) [237–241].

Transmission of Aβ and Tau

Studies suggest that misfolded proteins can be transmitted to adjacent cells and ultimately to other brain regions. However, the exact mechanism of this phenomenon is still unclear. One hypothesis is that the misfolded protein acts as a prion-like particle that can replicate and transmit itself to an adjacent cell, where it initiates the nucleation process of protein aggregation in the receiving cell [15, 242–248]. It has been proposed that a prion-like transmission of misfolded proteins (Aβ or Tau) from neuron to neuron could be responsible for the progressive neurodegeneration in AD, which also appears in the absence of Tau aggregation in the transentorhinal region and cortical Aβ pathology. This prion-like transmission may subsequently trigger misfolded Tau accumulation even during the asymptomatic phase of the disease [246, 249].

The misfolded protein transmission hypothesis is supported by data which indicate that the injection of brain extracts, which contain clumps of misfolded amyloidogenic Aβ into a normal mouse brain, triggers the spread of APs [250–252]. Similar experimental results have demonstrated that injected Tau protein triggered extracellular protein aggregates that were internalized by the cells, thereby inducing further Tau aggregation [253–255]. Firm evidence has indicated that mutant Tau can induce the aggregation of normal Tau in the human Tau-expressing mouse brain. More importantly, injected Tau was able to transmit the aggregation to different brain regions [253]. Furthermore, the newly aggregated human Tau can induce Tau aggregation when serially injected in healthy wild-type mouse brains [256].

The prion-like propagation of properties of misfolded Tau is also shared with IDPs in other neurodegenerative diseases, for example, α-synuclein in PD [257–259], polyglutamine in HD [260], and mutant SOD1 in ALS [261]. TDP-43, an RNA binding protein that is involved in ALS and frontotemporal lobar degeneration with ubiquitin (FTLD-U or FTLD-TDP), also exhibits prion-like transmission [240].

The propagation mechanism of misfolded proteins from one cell to another cell is unclear. Currently, two mechanisms of misfolded protein propagation have been proposed. The first mechanism involves the release and uptake of misfolded proteins [244, 262] that were potentially released through membrane damage [260], cell lysis [260], or exocytosis [263, 264]. These first two release options likely contribute to AD pathology via trauma and aging, whereas the third release option may be relevant in cases of neuronal synaptic transmission [264]. The evidence for the hypothesis of this mechanism has been demonstrated for Aβ [251] and Tau [253, 254]. The second mechanism uses a direct cell-to-cell propagation via tunneling nanotubes [244, 265], and it has been observed using a fluorescent-labeled PrP related to Creutzfeldt-Jakob disease (CJD); however, this second mechanism has not been shown in AD to date [246, 265].

Taken together, these data suggest that there is a strong correlation between proteinopathies, prion-like transmission, and AD; these findings imply that the impairment of the UPS, which is responsible for the removal of misfolded proteins, such as prion-like aggregates, occurs at an early stage of AD [15, 246, 266]. In addition, other pivotal proteins of the chaperone/proteasomal pathways, such as heat shock proteins (HSPs), FK506-binding proteins (FKBPs), and tripartite motif (TRIM) family proteins, were also altered in AD [118, 267].

The UPS

Protein degradation was initially thought to be an unregulated, non-specific, lysosome-dependent process. However, the discovery of the UPS has altered the paradigm of protein degradation because this alternative pathway provides a system with high specificity and involves a complex control mechanism. The UPS was first described in 1979 with the discovery of the 26S proteasome, an active proteolytic protein complex [268, 269].

The UPS is responsible for degrading the majority of cellular proteins and maintaining protein homeostasis [270]. It has been conserved during evolution because of its importance in cell regulation; thus, it is not surprising that it has approximately 1,000 components (which compose approximately 5 % of the genome), and the aberration of its function can lead to many diseases, including neurodegeneration [271]. The UPS is distinct from the other proteasome systems (e.g., lysosomal proteases) because it is ATP- and ubiquitin-dependent. The (poly)ubiquitination process initiates this cascade by covalently linking an ubiquitin molecule to a target protein. Subsequently, additional ubiquitin molecules are linked to form a polyubiquitin chain.

Ubiquitin is an 8.5 kDa protein composed of 76 aa. The ubiquitin protein is transcribed from UBB, ubiquitin C (UBC), RPS27a, and UBCEP2 genes. However, only the first two genes encode for a polyubiquitin precursor that is significant in the UPS signaling cascade. The other genes encode an ubiquitin that fuses to ribosomal proteins [272, 273]. The regulation of the attached protein depends on which ubiquitin lysine residues are linked. In the context of protein degradation, K-48 is the most significant residue because it signals for anchored protein degradation via the UPS. Other polyubiquitin functions may involve DNA repair, cell cycle regulation, lysosomal degradation, kinase modification, and endocytosis [273].

The 26S proteasome is the proteolytic complex of the UPS system, which is composed of a single 20S core catalytic protein and capped with 19S regulatory proteins. The ubiquitination of proteins occurs posttranslationally. The cascade is initiated with the activation of an ubiquitin molecule catalyzed by an E1 ligase, the activating enzyme that forms an ubiquitin-adenylate intermediate. This process is followed by a reaction of the intermediate bound to an E1 cysteine residue to form a high-energy thioester bond, allowing the ubiquitin molecule to be transferred to E2, a conjugating enzyme. The third enzyme, E3 ligase, acts as a substrate-recognition enzyme bound to the target protein. Finally, ubiquitin is transferred from the E2 enzyme to a lysine in the target protein (Fig. 3) [273, 274]. The E3 ligase is a pivotal enzyme in the UPS cascade and is crucial for substrate specificity. E3 plays a critical role in selecting the substrates, and the regulation of either its catalytic activity or substrate interaction properties is important for the UPS signaling pathway [273, 274].

Fig. 3
figure 3

The ubiquitin-proteasome system. The ubiquitin-proteasome degradation cascade begins when E1 activates an ubiquitin molecule at its lysine residue to create a high-energy thioester bond. The ubiquitin is then transferred to a reactive cysteine residue of the E2 enzyme. Meanwhile, the E3 enzyme (e.g., TRIM32/37) is bound to a target protein and moves it to the E2-ubiquitin, which transfers the ubiquitin molecule to the lysine residue of the target protein. This process is repeated for several rounds to obtain a polyubiquitinated target protein. This target protein is then recognized by the 26S proteasome and undergoes degradation

Full size image

The repetitive addition of ubiquitin moieties to the target protein results in polyubiquitination. As a result of its pivotal role in maintaining protein homeostasis, any impairment of the UPS can cause disease; thus, the UPS plays an important role in multiple diseases, including neurodegenerative pathologies in the CNS [275].

The TRIM family proteins are E3 ligases with a RING finger motif, which have ubiquitin ligase activity and are involved in various cellular processes, including apoptosis, tumorigenesis, and the viral response [276–278]. For example, the mutation of TRIM37 is linked to Mulibrey nanism [279], whereas TRIM32 is associated with skin carcinogenesis and is highly expressed in skin-derived tumors [280]. TRIM32 is also involved in mouse neural-progenitor cell signaling during neuronal differentiation [281] and in skeletal muscle remodeling and myogenic differentiation [282]. TRIM32 plays an important role in the proper formation of neurofilaments and axons. In addition, the knock-out of TRIM32 in mice resulted in reductions in the diameter of myelinated axons and low expression of type II (fast) myosin isoforms [283]. The knock-out of TRIM32 or TRIM37 using siRNA decreased survival in SHSY5Y neuroblastoma cells [118].

The UPS and Neurodegeneration

The accumulation of toxic protein aggregates is a common characteristic of neurodegenerative diseases, including AD, PD, HD, ALS, dementia with Lewy bodies (DLB), and prion diseases [7]. It has been demonstrated that Tau, α-synuclein, and TDP-43 co-occurred in AD and DLB [284, 285]. To protect cell function, several defense mechanisms against protein aggregate formation have been suggested [244, 270]. The UPS plays a crucial role in the management of general protein turnover and the degradation of misfolded proteins [272]; therefore, it is an important defensive mechanism against neurotoxic protein aggregation and prion-like protein propagation.

Evidence has indicated that a perturbed UPS can lead to the accumulation of toxic protein aggregates; thus, a perturbed UPS is likely to contribute to the development of many neurodegenerative disorders [266]. A well-known example of UPS impairment occurs in PD, which is a neurodegenerative disease characterized by the loss of dopaminergic neurons in the substantia nigra pars compacta. A loss-of-function mutation in Parkin, an E3 ligase, is a common familial cause of PD. The impairment of Parkin can lead to impaired aggregated α-synuclein degradation and the accumulation of unwanted protein aggregates [286]. In PD models, the inhibition of the UPS leads to the death of dopaminergic neurons in rat brains and primary neurons [287–289].

Proteasome inhibition has also been linked to ALS pathology. The proteasome inhibition in primary neurons (cortical, hippocampal, and motor neurons) resulted in the accumulation of TDP-43 aggregates [290]. An in vivo experiment of the conditional knock-out of a proteasome subunit gene resulted in the induction of ALS pathogenesis in mice; however, the knock-out of an autophagy-related gene failed to develop ALS pathology [291].

A recent study by Giannini et al. assessed the activity of the 20S and 26S proteasomes in young and aging rat brains and demonstrated that in aged rat brains, the proteasome’s substrate degradation activity was significantly decreased [292]. However, the ability of the 26S proteasome-specific polyubiquitin substrate degradation system was not impaired but rather slightly increased. This finding indicates that the 26S proteasome is not impaired during the normal aging process. Taken together, these results indicate that proteasome impairment is a key hallmark of neurodegenerative brains and is not observed in normal aging brains.

Increasing UPS activity in an HD patient’s fibroblast-derived neurons has been shown to improve cell survival [293]. In this study, UPS activity was rescued by overexpressing PA28γ, a 26S proteasome activator. Following exposure to various toxins, HD-derived neurons exhibited an increased susceptibility to cell death compared with neurons derived from donor healthy cells. However, when PA28γ was activated, HD-derived neurons exhibited an increased tolerance.

Autophagy is another major mechanism for the clearance of protein aggregates [21]. It depends on the lysosomal degradation of proteins by enveloping target proteins or organelles inside specialized compartments. A targeted protein for autophagy must be conjugated by an ubiquitin-like protein in a way that is similar to E1-E2-E3 signaling in the UPS [294]. Autophagy has been reported to be involved in various neurodegenerative diseases [295] and has recently been reviewed [294, 296–298].

Chaperones and Co-chaperones

Molecular chaperones, together with co-chaperones, are specialized machinery that assist protein folding and prevent toxic protein aggregation [299, 300]. The binding of chaperones to newly synthesized proteins aids in their correct folding. Many factors can promote protein misfolding, such as oxidative stress, aging, and misfolding-prone mutations. Chaperones also maintain general proteostasis, including protein trafficking and the targeting of severely misfolded proteins for proteolytic degradation [300]. As previously discussed, aggregating proteins may be cytotoxic; thus, the prevention of protein aggregation is critical for cell survival.

The HSPs belong to a major molecular chaperone family that assists protein folding by reversibly binding to unfolded and misfolded proteins to prevent their aggregation. In mammals, HSPs are classified according to their properties into several families, i.e., HSP90, HSP70, HSP60, HSP40, and small HSPs [301].

HSP90 (a 90 kDa HSP) is a major heat shock protein class involved in the stress response and interacts with a wide array of target proteins. It contains three domains, including an ATP-binding domain at its N terminus (~25 kDa) that displays low intrinsic ATPase activity, a middle segment for interaction with target proteins, and a C-terminal domain that is essential for protein dimerization [302]. Co-chaperone binding of HSP90 can alter enzyme activity; for example, the binding of Aha1 and Cpr6 can activate ATPase, whereas Hop/Sti1, Cdc37, and p23 suppress this activity [302].

Another important HSP class is HSP70. These proteins are one of the most ubiquitous classes of chaperones and are involved in diverse cellular functions [303, 304]. HSP70s have two domains, including a nucleotide binding domain for ATP-binding and a substrate binding domain used for target protein attachment [305]. Detailed aspects of the HSP70 machinery have been reviewed by Kampinga and Craig [306]. HSP70 is also required for the ubiquitin-dependent protein degradation of short-lived and aberrant proteins by the UPS. The interplay between HSP90 and HSP70 during the protein degradation cascade is intricate and is not completely understood. HSP90 stabilizes proteins by preventing an exposed hydrophobic cleft in its target protein and HSP70-dependent ubiquitination [307].

FKBP Co-chaperones

FKBP is a major member of the immunophilin protein family and a known co-chaperone for the HSP family [308]. The physiological function of FKBP is mediated by peptidyl-prolyl cis/trans isomerase activity (PPIase) in the FK506-binding domain (FKBD). The FKBD is the binding site for the immunosuppressant molecule FK506 (tacrolimus) [309]. Somarelli et al. classified 45 FKBPs across 23 species based on a tertiary structure prediction algorithm that resulted in six distinct groups of FKBPs [310]. In addition to chaperoning protein folding, FKBP plays a role in the regulation of cytokines, histone assembly, steroid receptor signaling, calcineurin signaling, and apoptosis [310, 311]. FKBPs are highly present in neuronal tissues and exhibit neuroprotective properties [312, 313]. There are 15 known FKBPs expressed in humans. A summary of a subset of FKBPs in the UniProt database is shown in Fig. 4.

Fig. 4
figure 4

The FKBP domains based on the UniProt database. All FKBPs have a characteristic FKBD that exhibits PPIase activity. FKBP14 has two additional EF-hand domains for Ca2+ signaling regulation. FKBP38 has three TPR domains for HSP90 binding and a transmembrane retention domain. FKBP51 and FKBP52 have two similar FKBDs and three TPRs

Full size image

FKBP12 (also known as FKBP1A) is the smallest and contains only a single FKBD [309]. FKBP38 (official gene symbol FKBP8) contains an unusual FKBD but functions similarly to FKBP12. It has been demonstrated that FKBP38 prevents apoptosis via the regulation of Bcl-2 to mitochondria; however, upon interaction with presenilins, FKBP38 promotes apoptosis by reducing mitochondrial Bcl-2 [314, 315]. FKBP51 and FKBP52 (also known as FKBP5 and FKBP4, respectively) have two FKBDs that are slightly different from each other and three tetratricopeptide (TPR) repeats [316, 317]. The TPR domain allows FKBP to mediate protein-protein interactions, and it has been identified as a binding site for HSP90. Based on the UniProt database, FKBP14 (other alias FKBP22) has a single FKBP domain and two EF-hand domains for Ca2+-mediated signaling. Whereas other FKBPs are commonly found in the cytoplasm, FKBP14 is located in the ER [318]. FKBP14 plays an important role in Notch signaling in Drosophila and may control γ-secretase activity via the regulation of presenilin, which thereby affects Notch cleavage [319].

Chaperones and Co-chaperones in AD

Resolving the role of chaperones and co-chaperones in AD can be an intricate task. A number of studies have identified links between chaperones and neurodegeneration. Dickey et al. demonstrated that phosphorylated Tau was selectively degraded via the HSP90-carboxyl terminus of the HSP70-interacting protein (CHIP) complex [234] (Fig. 5). CHIP is an ubiquitin ligase that selectively targets phosphorylated Tau and promotes Tau degradation. This same research group reported that the deletion of CHIP increases the phosphorylation of Tau in mouse brains [320]. Although the significance of the HSP and co-chaperone interplay in AD pathology remains unclear, increasing evidence suggests that HSP-UPS-mediated aggregate clearance is the pivotal mechanism of AD pathology and further emphasizes the role of HSP90 as a mediator of Tau protein regulation and degradation. However, the role of HSP90 in Tau degradation is paradoxical: evidence suggests that HSP90 both promotes and prevents the formation of abnormal Tau, depending on its interaction with different target proteins or chaperones [321] (Fig. 5). An insight regarding the interaction between HSP90 activity and Tau folding has been reported in a recent finding [322]. The authors used state-of-the-art nuclear magnetic resonance (NMR) technology to determine the specific binding sites of the HSP90-Tau interaction.

Fig. 5
figure 5

Predicted interactions between Tau protein, FKBP and CHIP in AD. FKBP12 catalyzes the hyperphosphorylated Tau assembly into PHF. The fate of HSP90-bound hyperphosphorylated Tau is determined by the binding of the HSP90 co-chaperone molecule. The binding of an FKBP51 co-chaperone will drive the accumulation of Tau, which is antagonized by FKBP52. In contrast, the binding of HSP90 to the CHIP co-chaperone will lead to hyperphosphorylated Tau proteasomal degradation

Full size image

The earliest study on the role of FKBPs in AD was conducted by Kraemer et al. who screened 46 genes that induced Tau pathology using Caenorhabditis elegans as a model [323]. A subset of the genes identified belonged to the chaperone or co-chaperone families. An in vivo knock-down experiment of FKBP51 by siRNA resulted in drastically reduced Tau and phosphorylated Tau levels, whereas the knock-down of FKBP52 slightly increased total Tau, which suggests that FKBP51 and FKBP52 have opposing actions [324]. Furthermore, the overexpression of FKBP51 increased Tau and phosphorylated Tau levels. To further confirm the involvement of FKBP51 in AD, a co-localization study in human AD brains indicated that FKBP51 was strongly co-localized with Tau. An F130A mutant of FKBP51, which abolished PPIase activity, did not prevent Tau degradation by chymotrypsin, indicating that FKBP51 prevents Tau degradation via its PPIase activity. A similar result was achieved by Blair et al. [325]. Furthermore, Blair et al. revealed that FKBP51 accumulates in AD brains and promotes the formation of Tau oligomers (Fig. 5).

Moreover, FKBP52 interacts with tubulin proteins and negatively modulates microtubule formation. The knock-down of FKBP52 results in neurite outgrowth and the induction of neuronal differentiation [326], whereas the overexpression of FKBP52 reduced Aβ-mediated toxicity in Drosophila; furthermore, the ablation of FKBP52 has opposing effects [327].

FKBP12 is observed in the area of neurodegeneration and is strongly associated with the protein aggregates in AD, PD, and DLB [328]. For example, an immunoblot analysis showed that FKBP12 expression is lower in AD total brain homogenates compared with controls and FKBP12 was identified in the intracellular NFTs of the hippocampus in AD brains [329], which suggest that FKBP12 is tangled with aberrant Tau protein. FKBP12 also has been reported to play a role in APP localization and processing. FKBP12 overexpression shifts the APP processing pathway towards the production of amyloidogenic products by changing the subcellular localization of APP [330]. It has also been reported that FKBP12 can bind AICD; however, the implication of this phenomenon in AD remains to be discovered [331].

FKBPs also play a role in other neurodegenerative diseases, including PD. The involvement of FKBP12 and FKBP52 in α-synuclein aggregation has been previously discussed by Gerard et al. [332, 333]. The overexpression of FKBP12 and FKBP52 in SHSY5Y cells induced α-synuclein aggregation, whereas inhibition, using FK506 and siRNA, reduced α-synuclein aggregation. Furthermore, an in vivo analysis indicated that the injection of FK506 (FKBP inhibitor) into PD mouse brains resulted in reduced α-synuclein aggregation.

Another distinct chaperone involved in AD pathogenesis is the peptidyl-prolyl cis/trans isomerase NIMA-interacting 1 (PIN1). PIN1 has been demonstrated to restore Tau binding to the microtubules via the promotion of Tau dephosphorylation [334, 335]. PIN1 has also been demonstrated to alter APP processing; its overexpression reduced Aβ production in vitro, whereas a PIN1 knock-out had the opposite outcome [336].

Clinical Studies Indicate UPS and Chaperone Deregulation in AD

As previously described, the pathophysiology of AD is initiated by an accumulation of aggregated Aβ and hyperphosphorylated Tau. The failure to clear these misfolded proteins may induce AD pathophysiology. The first evidence that underlines the importance of the UPS in AD is the cross-reactivity of ubiquitin in NFTs [337]. It has been observed that ubiquitination of the cerebral protein is impaired in human AD brains, which is accompanied by decreased proteolytic activity. Salon et al. reported that E1 and E2 ligase activities are lower in AD brains [338]. In agreement with this finding, an earlier postmortem study by Keller et al. identified a decrease in proteasome function in human AD brains compared with age- and gender-matched controls [339]. Moreover, the brain regions that demonstrated the most significant decrease in proteasome function are the regions most susceptible to AD, including the hippocampus, parahippocampal gyrus, superior and middle temporal gyri, and inferior parietal lobule. On the other hand, no significant decreases in proteasome function were observed in the occipital lobe or cerebellum, areas comparatively unaffected in AD pathology. Furthermore, this study showed that while the expression of the proteasome subunit(s) was uninterrupted, the function was inhibited. In agreement with previous studies, Keck et al.’s findings indicate that proteasome activity is decreased in AD brains [340]. Moreover, this study revealed that the proteasome activity was inhibited by PHFs and polymerized Tau, but not by soluble Tau. Isolated PHFs from AD brains and polymerized recombinant Tau protein significantly inhibited proteasome activity; however, unpolymerized recombinant Tau did not yield proteasome inhibition.

This theory is further supported by previous results that indicated proteasome activity impairment in AD-related neurodegeneration [266, 267, 341–344]. Furthermore, a human in vivo study that used the metabolic marker 13C6-leucine to trace the production and clearance rates of Aβ40 and Aβ42 in the CSF indicated that Aβ was produced at the same rate in both the AD (which ranged from very mild to mild AD) and control groups; however, the Aβ clearance in the AD group was significantly (~35 %) lower than the control group. [345]. The impaired proteolysis by the UPS may contribute to this observed phenomenon. The impeded proteolysis observed in very mild stages of AD suggests that proteolysis impairment precedes any symptoms from the development of AD pathology. Additional supporting evidence originated from a correlation study that identified a strong correlation between AD severity and Tau accumulation in AD brains [346]. This study also demonstrated a significant correlation between the level and activity of proteasome subunits and Tau accumulation in non-AD brains, which indicates that the UPS is crucial to prevent Tau aggregation.

Various groups have reported that the UPS can also be inhibited by Aβ [161, 347–351]. The use of scanning transmission electron microscopy has demonstrated that gold-labeled Aβ binds to the 20S proteasome, the catalytic core of the UPS, and inhibits its proteolytic activity [349]. This event, together with the activation of kinases, may explain why EOAD patients have observable symptoms at an accelerated rate. EOAD patients exhibit an increased production of Aβ that causes impaired UPS action, which is, alongside with contemporaneous activation of kinases that phosphorylate Tau, followed by the accumulation of misfolded Tau; these effects ultimately induce neuronal death accompanied by early-stage cognitive dysfunctions.

A clinical study by Yokota et al. indicated that Tau mRNA expression is not necessarily upregulated in AD brains; however, a related and more recent global proteome study by Manavalan et al. has shown that Tau protein is accumulated twofold in the temporal lobes of AD brains, which suggests that an impaired UPS is one of the pivotal causative factors of AD [118, 267]. This finding is in agreement with another previous report that indicated that impaired protein clearance, not increased protein production, causes the accumulation of neurotoxic aggregates [339, 340, 345]. Another important finding by Manavalan et al. was that most of the differentially regulated proteins from their study have a strong interaction with UBC. UBC is a polyubiquitin precursor involved in protein degradation by the UPS. These results underline that a perturbed UPS is almost certainly involved in the development of AD pathology [21, 271, 345].

In a global gene expression study of human AD subjects, Yokota et al. determined the differential gene expressions of TRIM32 and TRIM37, which are E3 ligase-encoding genes [118]. These data indicate that TRIM32/37 are expressed more than twofold in the occipital lobe (the brain region less susceptible to AD) compared with normal aging controls. Therefore, TRIM32/37 overexpression in the occipital lobe may prevent neurodegeneration, and areas affected by AD may have a perturbed UPS. Additional evidence for this phenomenon is that UBB+, a frame-shifted mutant ubiquitin, causes neuronal death and accumulates in the neurons of AD patients with this mutation [344]. UBB+ is a potent inhibitor of 26S proteasome [352]. Thus, this finding further emphasizes that the UPS can be inhibited in several ways that all lead to neuronal death in AD.

The upregulation of FKBP14 in the temporal lobes of AD brains has also been observed [118]. As previously described, immunophilins with PPIase activity may play a key role in AD pathogenesis by binding to hyperphosphorylated Tau and preventing ubiquitination and degradation. This notion is supported by reports which indicate that the inhibition of FKBPs by FK506 has a neuroprotective effect [353–355]. A study reported that FK506 also inhibits prion replication in a mouse model of CJD [356]. One explanation for these effects is that FK506 inhibits the formation of misfolded protein by inhibiting the PPIase activity of FKBP. Therefore, an increased level of small immunophilins, such as FKBP14, may represent another risk factor that prevents proper Tau degradation in neurons, which may thus contribute to AD pathogenesis.

Discussion

Despite more than 20 years as the de facto basis for AD research, the amyloid cascade hypothesis has failed to provide a clear mechanism of AD pathogenesis or halt disease progress in a myriad of drug trials [5, 6, 357]. In contrast, the Tau-based hypothesis has provided a more reasonable molecular explanation. First, other tauopathies exhibit neurodegeneration, even in the absence of APs. Thus, APs may induce neurodegeneration (as in EOAD) without being crucial (as in LOAD). A therapy based on a Tau aggregation inhibitor (TAI) has demonstrated more promising results compared with Aβ drugs. In a large phase 2 study of 321 subjects, methylthioninium chloride (methylene blue) was reported to stabilize the progression of both mild and moderate AD over 50 weeks [358, 359]. Therefore, a novel approach for the prevention of misfolded Tau accumulation is critically needed.

Here, we propose a hypothesis that AD is caused by collective insults to the ability of neurons to maintain proper proteostasis, especially in Tau protein regulation mediated by the UPS (Fig. 6). In neurons, Tau protein is exceptionally important because of its role in the maintenance of axonal transport and synaptic health. A failure to maintain Tau proteostasis results in the deposition of tangled Tau, which eventually manifests as PHFs and NFTs. Unfortunately, Tau protein is an IDP that is highly susceptible to aggregation without proper regulation and is subject to numerous modifications (i.e., phosphorylation and acetylation) that can induce Tau misfolding [217, 232, 233, 237]. Furthermore, in animal studies, Tau has exhibited prion-like characteristics that can be spread as aggregated units to nearby neurons [253, 254]. These intrinsic characteristics of Tau shed light on how AD is almost asymptomatic during its early stages and then exponentially progresses during the later stages.

Fig. 6
figure 6

The proposed AD model that involves Tau phosphorylation, prion-like misfolded protein propagation, and the deregulation of UPS, HSP90, and FKBP co-chaperones. In AD, various physiological conditions are altered, including kinase overexpression or Aβ-induction increases in cellular kinase activities. As a result, the phosphorylation-prone Tau protein becomes hyperphosphorylated and loses its ability to bind with microtubules. It is subsequently misfolded and acts as a seed for prion-like transmission to adjacent neurons where it aggregates into NFTs. In AD brains, the UPS system is impaired as a result of inhibition by Aβ or aberrant gene expression, which is possibly triggered by an aberrant AICD signaling cascade. TRIM32/37, important E3-ligases, are downregulated, which impairs the protein degradation pathway. In addition, the downregulation of HSP90 might also impair hyperphosphorylated Tau clearance. HSP90 and CHIP target hyperphosphorylated Tau for proteasomal degradation. The FKBP14 co-chaperone, which is upregulated in AD brains, prevents Tau degradation by binding to Tau and increasing its stability via interaction with the PPIase domain. Detached Tau loses its function and impairs the axonal transport pivotal for neuron survival. In this paradigm, Tau is involved in both the loss-of-function of axonal transport and gain-of-toxicity via prion-like transmission; in addition, impaired UPS and HSP90 function and FKBP14 overexpression promote pathophysiological conditions

Full size image

One source of alterations for Tau proteostasis is the downregulation of E3 ligase in AD-susceptible areas. In the AD brain, HSP90 is a binding partner for CHIP and has been reported to be downregulated. Previous reports have indicated that CHIP selectively targets phosphorylated Tau, and the deletion of CHIP leads to an accumulation of phosphorylated Tau protein [234, 320]. The downregulation of HSP90 in AD-affected areas may result in the impairment of CHIP function, which thus promotes Tau accumulation. We postulate that CHIP may be the canonical E3 ligase that targets Tau. In AD-affected neurons, CHIP is either impaired or overloaded. A clinical study by Yokota et al. demonstrated that TRIM32/37 upregulation in the occipital lobe may result in a gain-of-function of the UPS; therefore, this protective mechanism enables certain brain regions to become less susceptible to AD pathogenesis [118]. Interestingly, it has been reported that the overexpression of Parkin, an E3 ligase that is associated with familial PD, ameliorates Tau and Aβ aggregates in an AD mouse model [360]. This finding implies that the activity of canonical E3 ligases for Tau or Aβ can be compensated by other E3 ligases.

Another important finding from this clinical investigation is the involvement of small immunophilins. The overexpression of FKBP14 in the temporal lobe of AD brains suggests that it might promote Tau toxicity and inhibits Tau clearance via the stabilization of aggregated Tau and the prevention of its degradation. Another previous study also indicated that another small immunophilin, FKBP12, is bound to aberrant Tau in NFTs. This finding suggests that small immunophilins have a crucial role in the development of AD pathogenesis and could become important drug targets or biomarkers.

In light of these findings, further studies are needed to investigate the role of TRIM ligases in Tau ubiquitination and to shed more light on E3 ligases other than CHIP that are involved in the Tau degradation pathway. The rationale is that E3 ligases are one of the largest protein families in humans and are encoded by more than 600 genes [361]; thus, the identification of canonical E3 ligases involved in Tau ubiquitination will provide a better understanding of AD pathogenesis at the molecular level. The discovery of non-canonical E3 ligases that target Tau can aid in the development or identification of compensatory E3 ligases, which can be used as therapeutics for the treatment of AD (Fig. 7).

Fig. 7
figure 7

E3 ligases and FKBP14 are potential drug targets for AD therapy. Increasing the activity of non-canonical E3 ligases by drugs or gene therapy could prevent Tau aggregation. CHIP has been suggested to target phosphorylated Tau for ubiquitination and subsequent degradation; therefore, CHIP is hypothesized to be a canonical E3 ligase for phosphorylated Tau. We postulate that in the AD brain, CHIP is impaired or overloaded, which causes a failure to maintain Tau turnover. In the occipital lobe, a region less impacted by AD, TRIM32/37 act as compensatory E3 ligases and help to maintain Tau protein turnover. As previously described, FKBP14 catalyzes Tau aggregation. FK506 (an FKBP inhibitor) has been reported to have neuroprotective effects; however, it also acts as an immunosuppressant. Novel drug therapies for AD should also focus on a non-immunosuppressive FKBP inhibitor that can cross the blood-brain barrier or a compound that can activate E3 ligases

Full size image

The role of FKBPs has not been widely studied in AD. The inhibition of FKBPs by FK506 has been demonstrated to be neuroprotective in prion diseases. It is suggested that prions assert their toxicity via the activation of the calcineurin pathway, whereas the inhibition of calcineurin signaling by FK506 prevents neurodegeneration [362]. However, we also postulate that the neuroprotective effects of FK506 are achieved via the inhibition of misfolded protein formation, rather than calcineurin signaling. A study in neuroblastoma cells indicated that the addition of FK506 did not induce autophagy; however, it prevented PrP replication in cell culture [356]. The inhibition of FKBPs successfully attenuated protein aggregation in a mouse PD model [332]. Considering the shared properties between Tau and α-synuclein aggregates, non-immunosuppressive FKBP inhibitors could represent candidates for a therapy to prevent Tau aggregation (Fig. 7). A focus on FKBP inhibitors as drug candidates has additional advantages. FKBPs have been well studied for their pharmacological interest, several sequences have been identified, and three-dimensional structures from X-ray and NMR studies are available; therefore, screening strategies for the discovery of new drug candidates have been established [363].

Conclusions

Our review of the recent research data based on (pre)clinical studies elucidated a potential mechanism for AD pathogenesis as a result of the malfunction of the UPS and deregulation of molecular chaperones. AD neurons exhibit impaired proteostasis that ultimately leads to misfolded protein accumulation and manifests as characteristic AD lesions. We postulate that the impairment of E3 ligases, notably HSP90-CHIP and TRIM32/37, and the deregulation of molecular chaperones, notably FKBP immunophilins, are the real culprits that cause misfolded Tau protein accumulation. Thus, future AD research should not only focus on the amyloid cascade hypothesis but also on the involvement of the UPS and molecular chaperones to facilitate the understanding of the neuropathogenesis of AD and other tauopathies.