**2. Proteomics studies of dementia and AD**

**1. Introduction**

110 Update on Dementia

Dementia is progressively more common disease in aging population. Worldwide, in 2015 about 46.8 million people were affected by dementia and projected to increase to about 74.7 million by 2030 [1]. The increase in dementia patients is in part due to the aging society, lack of effective prevention strategies, and curative treatments. Due to this exponential increase in dementia population, the social and economic cost of this disorder is surpassing those attributed to cancer and heart diseases [1, 2]. High global prevalence, impact of this disorder on families, caregiv‐ ers, and communities have posed significant public health challenge [3] forcing the global health community to recognize the need for action and to place dementia on the public health agenda. Recently, the World Health Organization (WHO) has identified dementia as a major public health priority [3]. Unfortunately, dementia research has not been given priority as well as funding share, which could be another reason for significant increase in dementia population. For example, in the UK, only 11% of research funding has been allocated for dementia re‐

The most common forms of dementia are Alzheimer's disease (AD) and vascular dementia (VaD), with respective frequencies of 70 and 15% of all dementias [4]. However, the boundaries between the subtypes are sometimes not clear and mixed forms often coexist [5]. In past decades, research in different subtypes of dementia has failed to improve our understanding of dementia pathogenesis and to develop effective treatments or interventions for this disorder [6, 7]. The major mystery is the lack of information on the main causes of the disorder. This remains the main obstacle in developing a cure for the disorder. Therefore, an urgent inter‐ vention is needed to identify the key molecular mechanism that promotes dementia patho‐ genesis. Several theories have been put forward and only few have survived the test of time. Induction of dementia by ischemic cerebral vascular diseases or stroke was first described in clinics a century ago. However, the later discovery of aggregated β-amyloid and *tau* proteins in the brain tissues of dementia patients diverted the majority of subsequent research toward the study of these two molecules. Accordingly, it was hypothesized that this disorder is triggered by the toxicity of oligomerized protein that forms senile plaque including amyloidbeta (Aβ) and *tau* proteins [8]. However, this hypothesis failed to answer several questions regarding pathogenesis and further development in therapeutics. Although Aβ-deposition has been considered as the main cause of AD, the degree of its deposition in the brain does not correlate with dementia severity [9]. According to Arriagada et al. [10], patients without dementia have the same density of senile plaques as patients with AD. Amyloidal hypothesis could not answer questions such as why healthy elderly people have abundant senile plaques

The burden of senile plaques does not correlate with cognitive dysfunction in dementia indicating that protein aggregation alone is not sufficient to explain the pathology of these disorders. Accumulation of degenerative protein modifications (DPMs) triggered by nonen‐ zymatic spontaneous posttranslational modifications, loss of protein function, protein misfolding, protein aggregation, and their depositions in brain tissues could be key features of multiple neurodegenerative diseases since protein dysfunction is likely to extend beyond

search while 64% was spent on cancer research in 2012.

in their brains but no signs of AD [11].

Dementia is caused by damage to brain cells, which further interfere with the ability of brain cells to communicate with each other. The broad range of symptoms includes a decline in memory, thinking skills, and decision making. This potentially affects a person's ability to perform everyday activities. According to Alzheimer's association, dementia have several types such as AD, VaD, mixed dementia, Parkinson's Disease (PD), frontotemporal dementia, mild cognitive impairment, posterior cortical atrophy, traumatic brain injury, Down syn‐ drome, Creutzfeldt-Jakob disease, and normal pressure hydrocephalus [23]. These subtypes are associated with damage to specific types of brain cell in particular regions of the brain. For example, hippocampus is the center of learning and memory in the brain, and damage to hippocampus cells results in memory loss, which is one of the earliest symptoms of AD. The presence of aggregated protein plaque is a common clinical manifestation of the diseases, but the specific molecular mechanisms in each type of dementia that trigger neurodegeneration remain a mystery. The main reasons are the lack of well-characterized clinical samples of brain from particular region, suitable technology to isolate plaque and aggregated proteins, the technique that profiles quantitative composition of both soluble and aggregated proteins, and the technique that accurately identifies DPMs. Proteomic technique enables the comprehensive analysis of the protein and its work flow involves the identification of proteins following their separation, digestion by trypsin, determination of the molecular weight of the resulting peptides, and database searching to make the identification and quantification of the proteins as well as the characterization of the DPMs. In addition to label-free proteomic methods, isobaric tags for relative and absolute quantitation (iTRAQ) and tandem mass tag (TMT) protein labeling are widely accepted approaches for quantitative profiling of cell lines and clinical brain tissue samples [24–26]. Proteomics has also been used for the accurate identifi‐ cation of protein modifications [26–31].

#### **2.1. Novel amyloidal protein-enrichment techniques and DPMs**

The alteration in protein function and aggregation is the key feature of neurodegenerative diseases. However, what initiates the protein aggregation, and their deposition and formation of insoluble plaque are poorly defined. Due to poor solubility and self-association of these amyloidal plaque proteins, their accurate identification and quantitation in brain tissue extracts are technically challenging. Researchers [32, 33] have attempted to isolate amyloid proteins using detergents or detergent-free buffers. They adopted sequential extraction and quantification by enzyme-linked immunosorbent assay (ELISA), immunoblotting, or immu‐ nocytochemistry. But these approaches were unable to determine the aggregation state of the amyloids and complete composition of amyloidal proteins. Recently, Adav et al. [34] success‐ fully developed ultracentrifugation-electrostatic repulsion hydrophilic interaction chroma‐ tography (UC-ERLIC)-coupled mass spectrometry-based proteomics technologies to characterize aggregated proteins in human brain tissues affected by dementia. Using a detergent buffer, they extracted soluble proteins, amyloidal proteins, and insoluble aggregated proteins to identify dementia-associated changes in amyloid protein composition, relative abundances, and the extent of DPMs such as deamidation. These authors profiled both soluble and aggregated amyloidal plaque by LC-MS/MS and found significant enrichment of proteins such as S100A9, ferritin, hemoglobin subunits, creatine kinase, and collagen among the aggregated brain proteins. According to their findings, amyloid plaque was enriched in the deamidated variant of protein S100A9. Yet, the following modified protocol (**Figure 1**) could further improve the detection and identification of amyloidal protein profile in clinical samples.

Most DPMs cause small shift in mass and also involve the addition of small chemical motifs to protein side-chain functional group. This causes alterations in charge and hydrophobicity of the peptide/protein. The detection of the DPM-modified peptide/protein is challenging because the DPMs containing peptides in the trypsin-digested protein sample usually exhibit very low stoichiometry; hence, it is very difficult to identify these from high abundant unmodified peptides during LC-MS/MS analysis. However, these DPMs containing peptides with different charges and hydrophilicities can be separated from unmodified peptides by using ion exchange column running in hydrophilic interaction liquid chromatography (HILIC) mode that facilitates the detection and identification by LC-MS/MS [35]. Moreover, the unmodified and modified peptides elute from ion exchange column in a predictable order based on their charge densities in LC-MS/MS mobile phase. Accordingly, the modified and unmodified peptides can be separated by electrostatic-interaction-modified HILIC (emHILIC) methods using weak anion exchange (WAX)/strong anion exchange (SAX) columns in ERLIC mode for online ERLIC-MS/MS analysis, or using weak cation exchange (WCX) columns in electrostatic attraction hydrophilic interaction chromatographic mode (EALIC) for online ERLIC-MS/MS analysis.

from particular region, suitable technology to isolate plaque and aggregated proteins, the technique that profiles quantitative composition of both soluble and aggregated proteins, and the technique that accurately identifies DPMs. Proteomic technique enables the comprehensive analysis of the protein and its work flow involves the identification of proteins following their separation, digestion by trypsin, determination of the molecular weight of the resulting peptides, and database searching to make the identification and quantification of the proteins as well as the characterization of the DPMs. In addition to label-free proteomic methods, isobaric tags for relative and absolute quantitation (iTRAQ) and tandem mass tag (TMT) protein labeling are widely accepted approaches for quantitative profiling of cell lines and clinical brain tissue samples [24–26]. Proteomics has also been used for the accurate identifi‐

The alteration in protein function and aggregation is the key feature of neurodegenerative diseases. However, what initiates the protein aggregation, and their deposition and formation of insoluble plaque are poorly defined. Due to poor solubility and self-association of these amyloidal plaque proteins, their accurate identification and quantitation in brain tissue extracts are technically challenging. Researchers [32, 33] have attempted to isolate amyloid proteins using detergents or detergent-free buffers. They adopted sequential extraction and quantification by enzyme-linked immunosorbent assay (ELISA), immunoblotting, or immu‐ nocytochemistry. But these approaches were unable to determine the aggregation state of the amyloids and complete composition of amyloidal proteins. Recently, Adav et al. [34] success‐ fully developed ultracentrifugation-electrostatic repulsion hydrophilic interaction chroma‐ tography (UC-ERLIC)-coupled mass spectrometry-based proteomics technologies to characterize aggregated proteins in human brain tissues affected by dementia. Using a detergent buffer, they extracted soluble proteins, amyloidal proteins, and insoluble aggregated proteins to identify dementia-associated changes in amyloid protein composition, relative abundances, and the extent of DPMs such as deamidation. These authors profiled both soluble and aggregated amyloidal plaque by LC-MS/MS and found significant enrichment of proteins such as S100A9, ferritin, hemoglobin subunits, creatine kinase, and collagen among the aggregated brain proteins. According to their findings, amyloid plaque was enriched in the deamidated variant of protein S100A9. Yet, the following modified protocol (**Figure 1**) could further improve the detection and identification of amyloidal protein profile in clinical

Most DPMs cause small shift in mass and also involve the addition of small chemical motifs to protein side-chain functional group. This causes alterations in charge and hydrophobicity of the peptide/protein. The detection of the DPM-modified peptide/protein is challenging because the DPMs containing peptides in the trypsin-digested protein sample usually exhibit very low stoichiometry; hence, it is very difficult to identify these from high abundant unmodified peptides during LC-MS/MS analysis. However, these DPMs containing peptides with different charges and hydrophilicities can be separated from unmodified peptides by using ion exchange column running in hydrophilic interaction liquid chromatography (HILIC)

cation of protein modifications [26–31].

112 Update on Dementia

samples.

**2.1. Novel amyloidal protein-enrichment techniques and DPMs**

**Figure 1.** Isolation and identification of both soluble and insoluble amyloid proteins.

#### **2.2. Quantitative clinical proteomics of brain tissue**

Protein quantification through the incorporation of stable isotopes has become a vital tech‐ nology in modern proteomics research. Applying two-dimensional (2D) liquid chromatogra‐ phy coupled with tandem mass spectrometry-based iTRAQ (2D-LC-MS/MS-iTRAQ) technique, Brodmann area 21 of pathologically confirmed cases of VaD and matched non‐ neurological controls were studied [25]. In the study, 144 differentially expressed proteins including superoxide dismutase, neural cell adhesion molecule, and ATP synthase subunit alpha were characterized to be significantly up-regulated in VaD patients, suggesting a state of hypometabolism and vascular insufficiency along with an inflammatory condition during vascular dementia. iTRAQ quantitative proteomics of brain tissue samples from VaD subjects discovered down-regulation of ion channel proteins including proteins such as V-type proton ATPase subunit D (VATD), ATP synthase, H+ transporting, mitochondrial F0 complex, subunit b-isoform (ATP5F1), Obg-like ATPase 1(OLA1), and V-type proton ATPase subunit F (VATF) [24]. The ion channel protein Na+ −K<sup>+</sup> −ATPase exhibits multiple functions including the maintenance of differential membrane potential in neurons, which is an essential feature of the signal transduction. Using proteomics and structural modeling of Na+ −K<sup>+</sup> −ATPase, Sze and coworkers [24] showed that the impaired regulation and compromised activity of Na+ −K + −ATPase contribute to the pathophysiology of VaD. Dysregulated Na+ −K<sup>+</sup> −ATPase expression or function have been reported in both animal models and brain tissues in AD, PD, and Huntington's disease (HD) [36].

Synaptic failure is the most common feature observed in both VaD and AD. The loss of synapses and synaptic contacts is also most significant contributor to the cognitive impairment in VaD and other neurodegenerative disease [30, 37]. Similarly, a decline in synapse number in the hippocampal dentate gyrus in AD has been correlated with impairment on a variety of cognitive tests [38]. This suggests that hippocampal degeneration is central to memory loss in AD. Mitochondrial dysfunction is a vital feature of AD, but the fundamental mechanism is still unclear. Mitochondrial dysfunction in neurodegenerative disorders remains a key to the development of oxidative stress. According to Caspersen et al. [39], mitochondrial Aβaccumulation impairs neuronal function contributing to cellular dysfunction in transgenic (Tg) mice expressing human-mutant amyloid precursor protein (mAPP). During the early stages of AD, a reduced number of mitochondria in neurons [40] and decreased brain glucose metabolism [41] have been reported. As reviewed by Butterfield et al. [42], autopsied AD brain tissue revealed a decreased pyruvate dehydrogenase activity in the parietal, temporal, and frontal cortex. Activities of cytochrome c oxidase and mitochondrial complex IV were signif‐ icantly low in AD brain.

Dementia risk in women is higher than that in men. Recently, our group [43] applied discoverybased proteomics approach to evaluate gender differences in AD with cerebrovascular disease (CVD) subjects. Quantitative proteomics revealed gender-specific-altered mitochondriome. Proteomic analysis of AD-CVD brain tissues suggested hypercitrullination of arginine and deamidation of glutamine (Gln) in myelin basic protein (MBP) from female patients. It has been revealed that an increased citrullination of MBP is due to the down-regulation of cathepsin D and other enzymes that degrade the damaged proteins, leading to axonal dysfunction and progressive loss of neuron function [44].

#### **2.3. Insights from hypoxia/ischemia-induced neuropathy**

**2.2. Quantitative clinical proteomics of brain tissue**

(VATF) [24]. The ion channel protein Na+

Huntington's disease (HD) [36].

icantly low in AD brain.

+

114 Update on Dementia

Protein quantification through the incorporation of stable isotopes has become a vital tech‐ nology in modern proteomics research. Applying two-dimensional (2D) liquid chromatogra‐ phy coupled with tandem mass spectrometry-based iTRAQ (2D-LC-MS/MS-iTRAQ) technique, Brodmann area 21 of pathologically confirmed cases of VaD and matched non‐ neurological controls were studied [25]. In the study, 144 differentially expressed proteins including superoxide dismutase, neural cell adhesion molecule, and ATP synthase subunit alpha were characterized to be significantly up-regulated in VaD patients, suggesting a state of hypometabolism and vascular insufficiency along with an inflammatory condition during vascular dementia. iTRAQ quantitative proteomics of brain tissue samples from VaD subjects discovered down-regulation of ion channel proteins including proteins such as V-type proton ATPase subunit D (VATD), ATP synthase, H+ transporting, mitochondrial F0 complex, subunit b-isoform (ATP5F1), Obg-like ATPase 1(OLA1), and V-type proton ATPase subunit F

−K<sup>+</sup>

of the signal transduction. Using proteomics and structural modeling of Na+

−ATPase contribute to the pathophysiology of VaD. Dysregulated Na+

the maintenance of differential membrane potential in neurons, which is an essential feature

or function have been reported in both animal models and brain tissues in AD, PD, and

Synaptic failure is the most common feature observed in both VaD and AD. The loss of synapses and synaptic contacts is also most significant contributor to the cognitive impairment in VaD and other neurodegenerative disease [30, 37]. Similarly, a decline in synapse number in the hippocampal dentate gyrus in AD has been correlated with impairment on a variety of cognitive tests [38]. This suggests that hippocampal degeneration is central to memory loss in AD. Mitochondrial dysfunction is a vital feature of AD, but the fundamental mechanism is still unclear. Mitochondrial dysfunction in neurodegenerative disorders remains a key to the development of oxidative stress. According to Caspersen et al. [39], mitochondrial Aβaccumulation impairs neuronal function contributing to cellular dysfunction in transgenic (Tg) mice expressing human-mutant amyloid precursor protein (mAPP). During the early stages of AD, a reduced number of mitochondria in neurons [40] and decreased brain glucose metabolism [41] have been reported. As reviewed by Butterfield et al. [42], autopsied AD brain tissue revealed a decreased pyruvate dehydrogenase activity in the parietal, temporal, and frontal cortex. Activities of cytochrome c oxidase and mitochondrial complex IV were signif‐

Dementia risk in women is higher than that in men. Recently, our group [43] applied discoverybased proteomics approach to evaluate gender differences in AD with cerebrovascular disease (CVD) subjects. Quantitative proteomics revealed gender-specific-altered mitochondriome. Proteomic analysis of AD-CVD brain tissues suggested hypercitrullination of arginine and deamidation of glutamine (Gln) in myelin basic protein (MBP) from female patients. It has been revealed that an increased citrullination of MBP is due to the down-regulation of

and coworkers [24] showed that the impaired regulation and compromised activity of Na+

−ATPase exhibits multiple functions including

−K<sup>+</sup>

−K<sup>+</sup>

−ATPase, Sze

−ATPase expression

−K

In mild cognitive impairment (MCI) and early phase of AD, a decrease in the cerebral blood flow has been noted and correlated with the symptoms of dementia [45]. At cellular level, a decrease in the blood flow triggers hypoxia. The conditions such as hypoxia/ischemia have been linked to the pathogenesis of AD [46]. Unbiased proteomic analysis of hypoxia-ischemia pathology in numerous disease models and clinical setting including neuronal cell lines [47], a rat model of ischemic middle cerebral artery occlusion [48], a mouse model of cardiovascular disease [49], blood or tissues samples from patients with dementia [24, 26, 30] has provided novel insight into molecular pathology of hypoxia-ischemic injury and confirmed that hypoxia induced mitochondrial dysfunction and oxidative stress, induced epigenetic changes, and dysregulated proteostasis. Thus, oxygen availability is a crucial regulator of cellular metabo‐ lism and homeostasis. Proteomic study using ischemic neuronal injury model also identified the dysregulation of proteins such as Park7 and VAP-A implicated in the chronic neurological disorders such as AD and PD [47]. When neuronal cell response to hypoxia and glucose depletion stress was studied by iTRAQ proteomics in hypoxia-ischemic penumbra model, dysregulation of housekeeping proteins, antioxidative defense, chaperone response, and protein metabolism were observed [47]. Proteomic of pathological progression from hypoxiaischemia brain injury to clinical dementia revealed the dysregulation of energy metabolism, mitochondrial dysfunction, neuro-inflammation, synaptic failure, etc. [24–26, 50]. Further, the activity of α-ketoglutarate dehydrogenase appears to be inhibited in the cerebral cortex of AD patients, and there are substantial evidences indicating that the function of the Krebs cycle is impaired in AD brains [51, 52]. The impact of hypoxia and the γ-aminobutyric acid (GABA) shunt activation in the pathogenesis of AD has been reviewed by Salminen et al. [51]. Restated, neurodegeneration is caused by a progressive cycle of hypoxic-ischemic brain injury that induces DPMs, protein misfolding, and aggregation, leading to cognitive decline and demen‐ tia. Hypoxia-inducible transcription factor (HIF) is the key inducer of hypoxia-responsive genes that functions during general development and pathological processes in association with decreased oxygen availability. In hypoxic condition, HIF is accumulated while it is rapidly degraded in normoxic cells. HIF prolyl 4-hydroxylases (HIF-P4Hs, commonly known as PHDs and EglNs) act as oxygen sensing.

Recent studies suggest that neurodegeneration is caused by progressive cycles of hypoxiaischemic brain injury that induces DPMs, protein misfolding, and aggregation. These processes result in cognitive decline and dementia. The molecular events that drive this proteinopathy preceding dementia symptoms have not yet been well identified. However, unbiased, global, discovery-driven approaches such as proteomics have the potential to uncover the complex molecular pathology of human proteinopathies including dementia. Our groups adopted systematic proteomic studies to investigate hypoxia effects on neuronal cell lines, animal models of ischemic brain injury, human blood plasma samples, and postmor‐ tem brain tissues from patients affected by stroke or dementia [47–50, 53–57]. We and other

investigators [53–57] have yielded a good progress in understanding how protein DPMs, and protein aggregation induced by hypoxic-ischemic brain injury can promote neurodegenera‐ tion in dementia. This "vicious cycle" of brain tissue damage is summarized in **Figure 2**.

**Figure 2.** Vicious cycle of hypoxia-ischemic brain injury, degenerative protein modifications, and amyloidosis.
