Worldwide, Alzheimer’s disease (AD) is the most common form of dementia and an important neurodegenerative disease affecting health and quality of life. With an unprecedented rapidly aging human population, dementia is increasing in prevalence. In 2019, 55 million people globally were reported to be living with dementia, and the World Health Organization predicts that this number will increase to 139 million by 2050 [1]. AD is responsible for approximately 60%–80% of all dementia cases [2]. Clinically, AD is characterized by gradual memory decline, dementia, personality changes, and a reduced ability for independent daily living. Although AD is a devastating disease, its underlying mechanisms are poorly understood, and no current cure exists [3]. The neuropathological hallmarks of AD include extracellular amyloid plaques composed of amyloid β (Aβ) and intracellular neurofibrillary tangles (NFTs) containing hyperphosphorylated tau [4]. Under normal physiological conditions, these proteins are correctly folded and functional; however, in AD, they misfold and aggregate irreversibly, leading to neuronal dysfunction and cell death [5]. Consequently, protein misfolding is recognized as a fundamental mechanism in AD development and advancement of AD [6].

This review aimed to provide a comprehensive analysis of protein misfolding in patients with AD. First, we discuss the cellular aspects of protein quality control and then delve into the role of the unfolded protein response (UPR) and endoplasmic reticulum (ER) stress in proteostasis. Finally, a brief account of the theoretical modeling of protein folding is outlined. Next, we shall examine molecular mechanisms of the misfolding and aggregation of Aβ and tau proteins, associating these with inflammation, oxidative damage, and neuronal stress.

Cytosolic proteins must fold into their native three-dimensional structures with high efficiency shortly after synthesis by ribosomes based on their linear amino acid sequence, sometimes assisted by molecular chaperones. Misfolding during this event renders the protein nonfunctional and exposes hydrophobic surfaces, which can generate abnormal interactions between molecules, such as aggregation. These aggregates can be cytotoxic; therefore, cells should develop a fine quality control system to more efficiently recognize and eliminate misfolded polypeptides [7]. Major protein quality control systems include the chaperone system, ubiquitin-proteasome system (UPS), ER-associated degradation (ERAD), and autophagy pathways. Chaperone proteins, e.g., heat shock protein 70 (HSP70) and HSP90, bind to the misfolded proteins to induce refolding or undergo catabolism [8].

Cellular protein quality control machinery plays a role in the equilibrium of proteins to fold, refold, and degrade misfolded proteins. Molecular chaperones are essential proteins that protect exposed hydrophobic regions, thereby avoiding deleterious protein–protein interactions. They help refold proteins or target them for degradation, as in UPS. Co-chaperones such as DNAJB8 are proposed therapeutic targets that hamper the early aggregation of tau [9]. UPS tags toxic proteins with a chain of enzymes (E1, E2, and E3), and sends them to the 26S proteasome for disposal. This pathway is related to ERAD, which is a mechanism for the retrotranslocation and ubiquitination of misfolded proteins, followed by cytoplasmic degradation. Central to this process are several important ubiquitin ligases such as Hrd1 and Doa10, whose expression and activity are downregulated in AD [10].

Autophagy is involved in the disposal of large-scale aggregates or dysfunctional organelles. Proteins such as microtubule-associated protein 1 light chain 3, sequestosome 1, histone deacetylase 6, and Bcl-2-associated athanogen 3 (BAG3) work together to package complexes by assembling them within a caging-like structure and directing them to lysosomes for degradation. Chaperone-assisted selective autophagy (CASA) is an example of selective autophagy that degrades misfolded or damaged proteins. First identified in muscle cells, CASA is a heteromeric complex composed of HSPA/HSP70, HSPB8, co-chaperone BAG3, and E3 ubiquitin ligase STUB1/CHIP (although STUB1/CHIP can be replaced by other ligases) [11]. Although HSPA/HSP70 proteins are ubiquitously expressed, CASA-specific components such as BAG3 and HSPB8 are highly expressed in striated muscles and neurons, highlighting their importance in maintaining cellular proteostasis [12,13]. While functioning synergistically to ensure proteostasis, defects in these three systems have also been observed in the brains of patients with AD and other neurodegenerative diseases [14]. This leads to abnormal protein deposition and generation of toxic aggregates [15].

ER plays a key role in protein folding, maintaining calcium balance, and cellular proteostasis. ER stress occurs when unfolded or misfolded proteins accumulate in ER, which disrupts both ER function and cellular homeostasis. In AD, growing evidence suggests that disruption of UPR—the adaptive signaling pathway for ER stress—not only leads to protein accumulation but also contributes to the development of neurodegeneration. Three important UPR activator proteins that regulate cellular stress responses are protein kinase RNA-like ER kinase (PERK), inositol-requiring enzyme 1α (IRE1α), and activating transcription factor 6 (ATF6). Each protein features a luminal domain to detect misfolded proteins and a cytoplasmic domain to initiate downstream signaling.

PERK is turned on when ER is stressed, either by breaking apart the 78 kDa glucose-regulated protein, forming dimers, and autophosphorylating, or by directly binding to misfolded proteins. This causes eukaryotic translation initiation factor 2 alpha (eIF2α) to be phosphorylated, which slows down global translation but speeds up the translation of stress-related proteins like ATF4. ATF4 regulates the expression of genes involved in amino acid metabolism, antioxidant defense, autophagy, and apoptosis. The initial activation of ATF4 may bolster cellular resistance; however, after extended ER stress, ATF4 activates the pro-apoptotic transcription factor C/EBP homologous protein (CHOP), leading to neuronal death [16]. Chronic stimulation of the PERK–eIF2α–ATF4 pathway increases β-site APP cleaving enzyme 1 (BACE1, also known as β-secretase 1) expression, which leads to more Aβ being made and worsens amyloidogenic disease [17]. In human AD brains, increased levels of phosphorylated PERK and eIF2α have been directly correlated with disease severity, indicating persistent activation of UPR [18]. Moreover, hyperactivated PERK signaling exacerbates neuroinflammation by activating nuclear factor kappa-light-chain-enhancer of activated B cells [19], while sustained eIF2α phosphorylation can suppress cAMP response element-binding protein, a major transcription factor for proteins essential for plasticity [20]. IRE1 functions as a transmembrane ER stress sensor that dimerizes and autophosphorylates into an active ribonuclease. Upon activation, it splices X-box binding protein 1 (XBP1) mRNA, generating its active spliced form (XBP1s), which is a potent transcription factor that increases the expression of chaperones and ERAD components, and promotes cell survival. This constitutes adaptive UPR during acute ER stress [21]. In contrast, sustained activation of XBP1 contributes to pathological conditions, including inflammation, metabolic disorders, tumor development, neuronal damage, and apoptosis [22]. In AD, sustained IRE1 activity may contribute to and exacerbate the pathology, cause Aβ accumulation and oligomerization, and cognitive and synaptic deficits; genetic ablation of the IRE1 ribonuclease domain resulted in reduced Aβ levels and improved memory in mouse models of AD, highlighting the maladaptive role of chronic IRE1 signaling in AD [23]. ATF6 seems to control Aβ production with a more direct mechanism. APP/PS1 transgenic mice have reduced ATF6 expression; however, activation of ATF6 inhibits BACE1 promoter activity and reduces Aβ42 levels [24]. In addition to its role within the ATF6 arm of UPR, ATF6 promotes ER chaperone expression and ERAD, and alleviates proteotoxic stress [25]. However, its specific role in human diseases is not as well characterized as those of the PERK and IRE1 pathways.

Currently, pharmacological approaches targeting ER stress are being investigated. Therapeutic interventions that inhibit the ATF4–CHOP arm of the integrated stress response, primarily through PERK blockade, or small molecules such as the integrated stress response inhibitor that alleviate eIF2α-phosphorylation–mediated translational repression, have been shown to ameliorate synaptic dysfunction and cognitive/behavioral deficits in AD-related and tauopathy models [26]. Additionally, pharmacological strategies aimed at activating ATF6 are actively investigated.

Protein folding refers to the spontaneous conversion of a linear unfolded polypeptide into a specific three-dimensional conformation that is essential for biological functions. This complex process has long been a central subject of biochemical, biophysical, and molecular biology research, and several theoretical models have been proposed to explain different aspects of protein folding. Anfinsen’s dogma [27] states that under physiological conditions, the native structure of a protein is determined solely by its amino acid sequence. Based on classical kinetic studies of ribonuclease A, Anfinsen [27] demonstrated that a denatured enzyme can spontaneously refold into its active state once the denaturant is removed, thereby establishing that proteins adopt their native structure by folding toward the most thermodynamically stable minimum free-energy state. Although groundbreaking, this view is limited because it does not account for the role of molecular chaperones in the cellular environment or the fact that misfolding and aggregation frequently occur in vivo, contributing to neurodegenerative diseases such as Alzheimer’s and Parkinson’s disease. A more dynamic view of protein folding is provided by the energy landscape theory [28], in which folding is the descent of a funnel-shaped free-energy surface. This model does not imagine protein folding along only one pathway, but rather sampling a broad conformational space and forming several intermediates prior to reaching the native state. This model illustrates several important aspects of modern folding science such as kinetic heterogeneity, multiple pathways, and the existence of folding intermediates. The other models emphasize alternative structural considerations. The nucleation–condensation model [29] claims that folding starts by segmenting a small native-like nucleus to which the rest of the molecule rapidly and cooperatively condenses. This model explains hierarchical and cooperative folding of small proteins, but is less appropriate for larger or modular polypeptides. In contrast, the modular folding model [30] proposes that multidomain proteins fold via partially or fully folded semi-independent modules or domains that interact later to produce the final structure. This viewpoint is especially pertinent to multidomain proteins or proteins with repeat domains, which may fold together or in a stepwise manner [31].

Aβ misfolding and aggregation play a critical role in the pathogenesis of AD. Aβ is generated by serial proteolytic processing of APP by BACE1 and γ-secretase, leading to the production of many Aβs of different lengths, including mainly Aβ40 and Aβ42 [32]. Among them, Aβ42 is more prone to misfold and aggregate because of its greater hydrophobicity and tendency to adopt β-sheet-rich structure [33].

Under physiological conditions, Aβ is produced and released from the brain. However, in pathological conditions, the fragile balance between production and clearance is lost and cerebral accumulation of Aβ monomers, which can also misfold and oligomerize downstream (such as soluble oligomers, protofibrils and finally insoluble amyloid fibrils, and senile plaques in brain parenchyma), is generated [34]. The misfolding process begins with a conformational change of Aβ from its native state to β-sheet structures, which are prone to aggregation [35].

Small Aβ oligomers are the most neurotoxic Aβ species, affecting synaptic activity, calcium homeostasis, oxidative stress, and neuroinflammation [36]. N-terminal oligomers can engage neuronal membrane receptors, mainly N-methyl-D-aspartate (NMDA) receptors, cellular prion protein (PrPC), and metabotropic glutamate receptor 5 (mGluR5), to stimulate synaptic and cellular signals for synaptic impairments and apoptosis [37].

Fibrillar Aβ aggregates, which are less soluble, are crucial in generating extracellular amyloid plaques, a signature feature of AD pathology. The seeding-nucleation model of aggregation describes the process as one where the level of misfolded Aβ surpasses a threshold and the first nucleation of fibrils occurs and then elongates rapidly as monomers quickly add on to the developing fibril [38]. Prion-like aspects of Aβ aggregation, including templated misfolding and cell-to-cell transmission, may explain the gradual regional spread of Aβ pathology [39]. Aβ aggregation kinetics are influenced by various factors including peptide concentration, pH, ionic strength, presence of metal ions (e.g., Zn²⁺, Cu²⁺), and post-translational modifications such as oxidation or pyroglutamylation [40,41]. In addition, neuronal membrane lipid content and isoforms of apolipoprotein E, especially ApoE4, influence Aβ aggregation and clearance [42]. Together, the misfolding and aggregation of Aβ is a multi-staged process which is influenced by both intrinsic properties of the peptide and extrinsic factors that the cell may control and one that is instrumental in AD.

The Aβ harbors multiple sequence motifs involved in its tendency to aggregate and in its toxicity. The seven residues 16–22 (¹⁶KLVFFAE²²) and overlapping shorter motifs e.g., ¹⁶KLVFFA²¹ have been widely studied for their amphipathicity, ability to form β-sheets, and self-recognition [43]. The ¹⁶KLVFF²⁰ segment is critical for the multimerization of Aβs. This hydrophobic motif is thought to allow π-π stacking and van der Waals force interactions toward oligomerization and fibrous aggregates, as observed in eukaryotic systems [44]. Structure and mutagenesis studies revealed that changes in this region cause a decrease in aggregation and cytotoxicity, revealing the importance of this region in Aβ self-assembly [45].

Beyond ¹⁶KLVFF²⁰, the ¹⁶KLVFFAE²² sequence contains a polar residue, E22 (glutamate), that alters the aggregation pattern and is the site of known pathogenic mutations, E22Q (Dutch) and E22G (Arctic), which are linked to familial AD [46]. These mutations increase the kinetics of aggregation and stability of protofibrils. A third major sequence motif, ²¹AEDVGSNKGA³⁰, that resides in the C-terminal half of Aβ40 and Aβ42, is less hydrophobic but is important for β-sheet stability and fibril architecture [47]. This region is involved in the growth of fibrils by elongation and lateral association, and connects to the central hydrophobic domain during secondary nucleation. In addition, molecular dynamics simulations indicated that this motif is involved in salt bridge formation (e.g., D23-K28) and loop stabilization in Aβ oligomers [48].

The interaction of these motifs is the molecular foundation of the nucleation-dependent polymerization of Aβ aggregation. The ¹⁶KLVFF²⁰ sequence specifically forms a self-recognition and nucleation center and the ²¹AEDVGSNKGA³⁰ region supports fibril elongation and stabilization, indicating that these regions could serve as potential targets for therapeutic intervention or imaging diagnostic probes [49].

The misfolding and aggregation of Aβ may be understood based on the major theories of protein folding. According to Anfinsen’s dogma, the amino acid sequence itself provides instructions for structural fate. In Aβ, the hydrophobic ¹⁶KLVFF²⁰ motif and C-terminal β-sheet–promoting region act almost like built-in triggers, predisposing the peptide to adopt β-sheet structures and slip into misfolding. Energy landscape theory offers a more dynamic perspective. Instead of following a smooth funnel toward a single, well-defined native structure, Aβ explores a rugged and frustrated energy landscape with multiple shallow minima, lacking a unique stable native state, which allows β-sheet–rich intermediates to be populated. The nucleation–condensation model describes the formation of small nuclei (especially around a ¹⁶KLVFF²⁰ nucleus) cooperatively seeding the addition of monomers, resulting in stepwise formation of oligomers, protofibrils, and fibrils at a faster rate for Aβ42, which shows greater aggregation propensity than for Aβ40. Finally, the modular folding model suggests that different stretches of the Aβ sequence behave almost like semi-independent modules. Each motif can fold or misfold locally, and over time, these pieces assemble hierarchically into larger toxic oligomers and, ultimately, the plaques seen in AD.

Tau is a neuronal microtubule-associated protein that plays an important role in stabilizing microtubules and regulating axonal transport [50]. Tau is a highly soluble, intrinsically disordered protein that undergoes a dynamic cycle of phosphorylation/dephosphorylation under physiological conditions and regulates its affinity for microtubules [51]. However, in a group of neurodegenerative diseases called tauopathies, including AD, frontotemporal dementia, and progressive supranuclear palsy, tau is abnormally hyperphosphorylated so as to form β-sheet-rich fibrous aggregates, leading to pathological paired helical filaments (PHFs) [52]. These pathological changes are the result of many types of post-translational modifications, particularly hyperphosphorylation at serine and threonine residues, which interfere with tau’s ability to bind microtubules and promote its aggregation. Hyperphosphorylated tau dissociates from microtubules, destabilizing the cytoskeleton and increasing the cytoplasmic pool of unbound tau. This unbound tau can misfold, oligomerize, and aggregate into β-sheet-rich fibrillar structures (Fig. 1) [53]. The major regions of tau that are prone to fibrillar aggregation are the microtubule-binding repeat domains, particularly the region including the ³⁰⁶VQIVYK³¹¹ (also known as PHF6) and ²⁷⁵VQIINK²⁸⁰ (PHF6*) motifs, which promote β-sheet formation and nucleation of filament assembly. Misfolded tau undergoes a series of transitions that give rise to filaments. Starting with monomers, it proceeds next to oligomers, and successively moves on with the further development of PHFs and straight structures, which become NFTs in neurons [54].

Notably, tau exists in two major isoforms: the 3-repeat (3R) form, which contains repeat R1, R3, and R4 domains, and the 4-repeat (4R) form, which includes an additional R2 domain encoded by exon 10; this extra repeat R2 harbors the PHF6* motif, whereas the PHF6 motif located in the R3 domain is shared by both 3R and 4R isoforms [55]. The PHF6 motif in the R3 domain is the most aggregation-prone region in tau. It acts as a seed that rapidly converts soluble tau monomers into β-sheet oligomers and fibrils. In vitro, short peptides derived from this motif can self-assemble into amyloid-like fibrils and nucleate the aggregation of the full-length tau proteins [56]. Structural studies such as solid-state high-resolution nuclear magnetic resonance and cryo-electron microscopy show that PHF6 forms a segment of the fibril core in AD tau filaments with a tight intermolecular β-sheet packing held together by hydrophobic and polar interactions [57].

Like the tau K18 fragment, the PHF6* motif at the beginning of the R2 domain, which is present only in tau isoforms containing exon 10 (encoding R2), is a β-sheet-prone sequence that is important in promoting tau aggregation. Moreover, experimental studies have shown that PHF6* cooperates with the PHF6 motif in the R3 repeat to nucleate fibril formation, thereby accelerating tau aggregation [58]. Because the R2 domain contains the PHF6* motif, its presence distinguishes the 4R tau isoforms from the 3R forms, and this difference is directly linked to the distinct aggregation patterns of tau isoforms observed in different tauopathies. For example, 4R tau is the major lesion in progressive supranuclear palsy, and 3R tau is the major lesion in Pick’s disease [59].

Mutating selected residues in PHF6 and PHF6* (e.g., isoleucine or valine to serine) markedly diminishes tau aggregation and fibril formation, confirming their importance in mediating pathological misfolding [60]. These have been used as potential targets for therapy, with peptide inhibitors (that target the repeat motifs), antibodies, and small molecules, which are all designed to interfere with the β-aggregation-prone interaction of the proteins and halt tau polymerization at the outset [61].

The misfolding of tau can also be understood in the context of folding theories. The amino acid sequence of Anfinsen’s dogma encodes the intrinsic structural propensities that drive tau toward misfolding. In particular, the β-sheet–prone motifs PHF6 (³⁰⁶VQIVYK³¹¹) and PHF6* (²⁷⁵VQIINK²⁸⁰) provide a sequence-based explanation for tau’s strong tendency to assemble into fibrils. The energy landscape theory offers a complementary perspective. Because tau is intrinsically disordered and highly flexible, it does not fold into a single stable native structure. Instead, extrinsic conditions such as hyperphosphorylation can reshape its conformational landscape. This framework suggests that tau is trapped in abnormal intermediates that favor β-sheet formation, thereby promoting its conversion into a pathological species. The nucleation–condensation model explains how tau molecules aggregate cooperatively. Small nuclei, often stabilized by PHF6 or PHF6*, serve as seeds that recruit additional monomers and drive the stepwise assembly of oligomers, PHFs, and ultimately, NFTs. The presence or absence of these critical motifs largely determines isoform-specific behaviors such as the differential aggregation of 3R versus 4R tau. Finally, the modular folding model suggests that tau-repeat domains behave as semi-independent modules. Misfolding in one region may destabilize adjacent domains and facilitate cross-seeding, leading to cooperative fibril growth. This modular perspective also highlights therapeutic opportunities. Compounds or antibodies specifically targeting these repeat motifs could block early misfolding events, thereby suppressing tau aggregation at its roots.

Aβ plaques and NFTs are pathologically and anatomically distinct. It has become evident that the interaction of Aβ with tau aggregates (itself potentially the result of Aβ-induced changes) is a dynamic and synergistic process. Additionally, it contributes significantly to accelerating the course of neurodegeneration and cognitive decline in AD.

Misfolded and aggregated proteins trigger intracellular stress and promote oxidative damage by increasing the accumulation of reactive oxygen species (ROS). Because mitochondria are the primary source of ROS, Aβ can accumulate within them and impair the electron transport chain, leading to reduced ATP production and a loss of mitochondrial membrane potential. Elevated ROS further cause DNA damage, protein oxidation, and lipid peroxidation, ultimately activating apoptotic pathways [78]. This oxidative stress is accompanied by the dysfunction of the Nrf2–ARE antioxidant defense system. In the brains of patients with AD, the suppression of Nrf2 results in impaired glutathione synthesis, reduced activity of antioxidant enzymes, and increased neuroinflammation. In parallel, activated microglia and astrocytes release pro-inflammatory cytokines such as interleukin (IL)-1β, tumor necrosis factor-α, and IL-6, which amplify neurotoxicity and contribute to neuronal death [79,80]. Chronic inflammation, together with defective autophagy and proteasomal clearance of protein aggregates, creates a vicious cycle that worsens protein deposition and neurodegeneration.

The interaction between neuroinflammation, oxidative stress, and protein aggregation is an important characteristic of AD pathogenesis. In this context, the UPR pathways—PERK, IRE1, and ATF6—are hyperactivated in the AD brain. Transient activation of these pathways may provide initial protection against protein aggregation; however, persistent activation induces inflammation and apoptosis. In particular, sustained activation of PERK-induced eIF2α phosphorylation increases BACE1 expression, leading to an elevation in Aβ levels in a positive feedback loop during AD progression [81]. Thus, although UPR can play a protective role, prolonged activation may be a key driver of disease progression.

Misfolding and aggregation of Aβ and tau are defining features of AD and other neurodegenerative proteinopathies. Clinically, extracellular Aβ plaques and intracellular NFTs composed of hyperphosphorylated tau correlate strongly with cognitive decline, synaptic dysfunction, and neuronal loss, and are therefore regarded as the classical hallmarks of AD [63,82]. Although Aβ deposition appears early in the disease, tau pathology shows a closer association with the degree of cognitive decline and overall disease severity, highlighting its pivotal role in AD progression [83]. Thus, Aβ and tau have become the main biomarkers for early diagnosis and disease monitoring: cerebrospinal fluid (CSF) Aβ42 decreases with plaque deposition, while phosphorylated tau (p-tau) and total tau increase, reflecting tauopathy and neurodegeneration [84]. More recently, blood-based assays for p-tau181, p-tau217, and Aβ42/40 ratios have shown promise as noninvasive diagnostic and prognostic tools [85].

Proteomic and multi-omics studies of brain, CSF, and blood have found molecular changes linked to Aβ, tau, proteostasis, autophagy, and endocytosis, many appearing before symptoms [86-89]. In both sporadic and inherited AD, CSF protein changes emerge years before onset, supporting established markers like p-tau and Aβ42/40 for better diagnosis, prognosis, and trial enrichment [86-89]. Single-cell and single-nucleus transcriptomic studies further show that excitatory neurons, oligodendrocytes, and microglia each display distinct stress responses, suggesting possible treatment targets and offering detailed tools to track drug effects [90-93].

The pathogenic accumulation of Aβ and tau has also been a therapeutic target for monoclonal antibodies (mAbs), small molecules, and anti-aggregation peptides. Anti-Aβ immunotherapies, including aducanumab and lecanemab, decrease amyloid load but with variable clinical efficacy; in particular, aducanumab was constrained by amyloid-related imaging abnormalities (ARIA) [94]. Tau-targeted strategies (such as antisense oligonucleotides, aggregate formation inhibitors, and tau immunotherapies) are under intense clinical investigation and hold promise for stopping the ongoing disease process [95,96]. However, AD is a multifactorial disease—Aβ, tau, neuroinflammation, and vascular dysfunction are all implicated. Therefore, mAbs that target a single mechanism of action may not have optimal utility. Therefore, growing emphasis is on dual-target therapeutics: for instance, as seen with the Dominantly Inherited Alzheimer Network Trials Unit, the Tau Next Generation platform of background anti-Aβ treatment (lecanemab) with add-on anti-tau immune therapy (E2814—microtubule-binding region-specific), reflecting a staged strategy of early amyloid clearance followed by suppression of tau propagation [97,98]. Additional advances are represented by bispecific antibodies [99], novel biologics that specifically target misfolded conformers of Aβ and tau (anti-β-sheet conformation mAb) [100], and brain-shuttle platforms that aim to improve blood–brain barrier permeability [101].

Recent phase 3 trials have clarified the clinical potential of these approaches. In the CLARITY AD trial (ClinicalTrials.gov number, NCT03887455), lecanemab slowed cognitive and functional decline by approximately 27% over 18 months, corresponding to a −0.45-point difference in the Clinical Dementia Rating–Sum of Boxes (CDR-SB) score compared to the placebo (1.21 vs. 1.66), accompanied by significant amyloid reduction; however, ARIA with edema (ARIA-E) occurred in 12.6% of participants. [102]. In TRAILBLAZER-ALZ 2, donanemab slowed decline on the Integrated Alzheimer’s Disease Rating Scale by 3.25 points and improved CDR-SB by −0.70, though with notable ARIA rates (ARIA-E 24.0%, ARIA-H 31.4%, symptomatic ARIA-E 6.1%), particularly in APOE ε4 homozygotes [103]. In contrast, the GRADUATE I/II trials of gantenerumab failed to meet their primary endpoints, underscoring the heterogeneity among anti-amyloid therapies and importance of drug-specific pharmacology, target engagement, and trial design [104].

In summary, scientists are working to understand how Aβ and tau proteins build up and spread in the brain, and they are combining this knowledge with detailed maps of single-cell proteostasis and multi-omics biomarker analyses. These studies provide important clues for improving diagnosis, developing more precise treatments, and identifying appropriate patients for each therapy. Growing evidence shows that using sensitive biomarker tests to guide treatments that target both Aβ and tau—especially when started early—may be the most promising way to slow or change the course of AD.

AD is a typical neurodegenerative disorder characterized by the misfolding and aggregation of Aβ and tau, which undergo conformational transitions from their native states into toxic, soluble oligomers and insoluble fibrils. These aggregates are mainly caused by defective protein quality control systems such as chaperones, UPS, ERAD, and autophagy, leading to synaptic dysfunction, neuroinflammation, oxidative stress, and neuronal death. Sustained activation of UPR, especially through the PERK–eIF2α pathway, further exacerbates pathology by enhancing Aβ production, promoting inflammation, and driving cell death. The amount of Aβ and tau aggregates that accumulate in the brain is closely related to the progression and severity of the disease, so biomarkers in CSF and blood can be used to indirectly monitor the pathological process. In particular, p-tau species (p-tau181, p-tau217) and the Aβ42/40 ratio are promising early diagnostic markers for AD diagnosis.

Therapeutic development until now has focused on mAbs and small molecule compounds targeting Aβ and tau, which show partial efficacy but are most useful when started before extensive aggregation and neurodegeneration occurs. Advancing the diagnosis and treatment of AD requires a more profound understanding of how protein-folding errors propagate between cells and how they interact with stress responses throughout the organism.

This study emphasized precision approaches. Single-cell proteostasis analyses have revealed how specific neuronal and glial populations activate distinct stress-response pathways, offering opportunities for targeted therapies and real-time monitoring of treatment effects. In parallel, multi-omics strategies, including proteomics, transcriptomics, and metabolomics of brain tissue, CSF, and plasma, identify biomarker panels that enable early diagnosis, patient stratification, and personalized therapeutic strategies. Importantly, mounting evidence suggests that combined targeting of Aβ and tau represents the most promising avenue for modifying disease trajectory, highlighting the need for integrative, mechanism-based interventions in AD.