Quantitative analysis of proteostasis networks: are there applications for dementia?

1. Introduction

1.1. Dementia

Dementia is defined as the progressive decline of two or more cognitive functions that results in a loss of quality of life. There is currently no treatment that can significantly alleviate the symptoms of dementia or its progression. According to WHO, 55 million people worldwide are afflicted with dementia and 10 million new cases are expected each year[1]. The three most common dementias are Alzheimer’s Disease (AD), Dementia with Lewy Body disease (DLB), and Frontotemporal Dementia (FTD). All dementias are characterized by neurodegeneration, with age being the most significant risk factor. In the United States, 6.7 million people suffer from AD, making it the most common dementia and the fifth leading cause of death for people over 65 years old[2].

1.2. Proteostasis Network

The brain proteome is dynamic but achieves a stable equilibrium by balancing the protein synthesis with protein degradation through a process of protein homeostasis or proteostasis. The proteostasis network (PN) maintains this equilibrium while also ensuring that proteins possess the correct structure, subcellular localization, post-translational modifications (i.e., phosphorylation, acetylation, and glycosylation) and protein-protein interactions (PPI) to be fully operational. The PN includes two principal mechanisms of cellular degradation, autophagy (lysosomes) and the ubiquitin proteasomal system (UPS) (proteasomes). All cells in the body depend on proteostasis, but the brain is especially vulnerable to dysfunction. Since neurons do not divide, minute perturbations in proteostasis can accumulate over a lifetime resulting in impairment, whereas other cells dilute toxic substances with each cell division. In addition, neurons are compartmentalized with a cell body, long axons with presynaptic boutons and spines on intricate dendritic arbors, which demand precise transport to maintain proteostasis.

2. Proteostasis dysfunction in Dementia

Proteostasis dysfunction in dementia is evidenced by the aggregation of misfolded proteins. AD is characterized by extracellular deposits of amyloid-beta (Abeta) peptides and intracellular aggregates of hyperphosphorylated tau protein, whereas alpha-synuclein aggregates characterize DLB and tau or TDP-43 proteins aggregate in FTD. Protein aggregation is hypothesized to protect cells, at least initially, from the toxicity of misfolded proteins. It is not known why these proteins accumulate abnormally in dementia, but evidence points to a dysfunctional PN (Figure 1). The PN components heavily implicated in dementia pathogenesis are chaperones and the endolysosomal network (ELN). Chaperones are ATP-dependent enzymes that refold proteins to their optimal structure by interacting with exposed hydrophobic regions. If the correct structure is not achieved, chaperones play an important role in directing proteins to degradation. Evidence suggests that chaperones become impaired with age. The ELN is a continuum of intracellular vesicles and a hub for protein signaling, trafficking, localization, recycling, and delivery of proteins to the lysosomes for degradation. Defects in the ELN are observed in AD brains reproducibly and early in pathogenesis[3]. GWAS studies have revealed an enrichment of ELN proteins in AD risk factors[4]. In addition, rare genetic mutations that cause AD disrupt the ELN[5]. How ELN abnormalities contribute to pathogenesis is unclear and may be multifaceted. Importantly, the PN has been manipulated by drugs to improve the dementia phenotype in animal models[6].

Figure 1:

Hypothesis of proteostasis network (PN) dysfunction in dementia.

3. Mass spectrometry (MS) analysis of Dementia

A recent review of the AD literature[7] reveals the use of the most current proteome technologies to analyze mouse models, human fluids and post-mortem tissues and different cell types generated from induced pluripotent stem cell (IPSC) lines derived from patients. Comparisons of patients who are Asymptomatic AD(AsymAD) (possess Abeta and tau aggregates but are cognitively normal) with patients who are Mildly Cognitively Impaired (MCI) (possess a phenotype that lies between normal and AD) has helped distinguish proteome changes that are responding to protein aggregates from those that are responsible for cognitive deficits. In the largest AD proteomic study, Johnson et al. used label-free quantification to study >2000 brains from different cohorts including AD, AsymAD, Non-AD[8]. Other studies obtained deeper quantitative proteome coverage using extensive peptide fractionation techniques to analyze smaller cohorts[9]. The massive amount of quantitative global proteomic data has confirmed that PN dysfunction is an early event in pathogenesis. However, there have been few proteomics studies that directly quantitate changes in the PN in response to dementia. Inda et al. demonstrated that the PPI chaperone network was altered in AD human brains and a mouse model using an affinity probe for stressed chaperones and label-free quantitation[10]. They found that the spatial memory of the model could be improved by inhibiting the formation of the stressed PPI network. Bourdenx et al. used isobaric tags to perform quantitative proteomic analysis on the insoluble fractions of brains from transgenic mice where chaperone-mediated autophagy (CMA) was impaired in neurons. When the CMA deficient mouse was crossed with an AD mouse, quantitative proteomics revealed an enhancement of insoluble proteins and exacerbation of protein aggregation whereas CMA activation decreased protein aggregation[11]. Lee et al. focused on the protein SORLA, which is an endocytic receptor that mediates trafficking through the ELN[12]. SORLA is a risk factor for AD and regulates Abeta. Isobaric tag quantitation was performed on differentiated neurons from >50 IPSC lines of AD patients. Strikingly, expression of SORLA significantly correlated with other AD risk factors localized to the ELN.

4. Expert opinion

Only a few drugs are currently approved for AD with modest cognitive improvement and potentially lethal side-effects. They target Abeta, so they are not applicable for dementias besides AD. These disappointing AD drugs and the abundance of AsymAD cases suggest that misfolded protein aggregates are essential but not sufficient to cause dementia. Proteostasis is clearly involved in pathogenesis, but our knowledge is fragmented. Drugging the PN has improved or “cured” dementia in animal models but has not been translated to humans. This may stem from the inherent difficulties of clinical trials, and the caveats of modeling a human specific disease in rodents. Rodent models are excellent for studying discrete timepoints to identify disease triggers, but none recapitulates all aspects of human dementia. Most models overexpress human genes so cognitive dysfunction is observed in young mice. Overexpression causes artifacts and ignores the strong biological link between dementia and age. Newer knock-in models without overexpression will provide a better analysis of the influence of age. Post-mortem human brain analysis also has limitations. After neurodegeneration has ravaged the brain, it often contains comorbidities, so it is unclear if the initial trigger is identifiable. A quantitative proteomic comparison of MCI, AsymAD, and AD brains will help deconvolute the dementia proteome. Human brain studies can be complemented with patient IPSC and organoids, which can combine drug screening and quantitative proteomics. Both human samples and animal models will be needed to discover potential drug targets in the PN.

To find PN drug potential for dementia, quantitative proteomics needs to directly investigate the PN to understand its roles in the brain, aging, and dementia. This is no small task as the PN encompasses thousands of genes and hundreds of signaling pathways. The quantitative strategy will require the use of labels (i.e., heavy stable isotope or isobaric tags) since label-free strategies are less accurate. PPI are promising drug targets[13], but we are still unsure how they are affected by dementia in the PN. Large PPI MS studies have been conducted in cultured cells, but this translates poorly to the brain proteome, especially neurons. PPI MS analysis of most AD risk factors has never been performed in human brain tissue, which requires immunoprecipitation (IP) of endogenous proteins using antibodies that may not be available. Alternatively, strategies can be used that IP tagged proteins or proximity labeling (i.e., BioID and APEX) that enriches proteins that transiently bind the protein of interest in mouse models and IPS-neurons. The brain proteome is a circuit of cells, but it is unclear how the PN function and impairment differs between cell-types. Patient IPSC should be differentiated into multiple types of neurons and glia to understand the PN in different cell types. For animal models, new techniques to label cell-specific proteomes can quantitate the cellular heterogeneity of the PN[14,15]. Although still in its infancy, single cell proteomics could localize PN disruptions to individual cells, since it is poorly understood why certain cells are more vulnerable early in pathogenesis. Intracellularly, the PN encompasses multiple subcellular compartments. Although subcellular compartments have been enriched by biological fractionation techniques and quantitated by MS, only a few studies have simultaneously quantitated multiple fractions which will be able to determine how PN disruptions are spatially distributed. This quantitative spatial proteomic strategy should be combined with PPI and PTM analyses to identify spatial PN protein isoforms. This editorial has highlighted AD research due to its abundance compared to FTD and DLB, but quantitative proteomic PN research is needed for all dementias. The strategy of targeting aggregates has not been successful in AD, so it should not be expected to help other dementias. It remains unknown whether there is a common underlying PN impairment prior to aggregation. Thus, quantitative proteomic comparison studies of AD, FTD, and DLB are required to determine if future therapeutic strategies can be universal or dementia specific.