Mass Spectrometry Methods for Analysis of Extracellular Matrix Components in Neurological Diseases

Abstract

The brain extracellular matrix (ECM) is a highly glycosylated environment and plays important roles in many processes including cell communication, growth factor binding, and scaffolding. The formation of structures such as perineuronal nets (PNNs) is critical in neuroprotection and neural plasticity, and the formation of molecular networks is dependent in part on glycans. The ECM is also implicated in the neuropathophysiology of disorders such as Alzheimer’s disease (AD), Parkinson’s disease (PD) and Schizophrenia (SZ). As such, it is of interest to understand both the proteomic and glycomic makeup of healthy and diseased brain ECM. Further, there is a growing need for site-specific glycoproteomic information. Over the past decade, sample preparation, mass spectrometry, and bioinformatic methods have been developed and refined to provide comprehensive information about the glycoproteome. Core ECM molecules including versican, hyaluronan and proteoglycan link proteins (HAPLNs), and tenascin are dysregulated in AD, PD, and SZ. Glycomic changes such as differential sialylation, sulfation, and branching are also associated with neurodegeneration. A more thorough understanding of the ECM and its proteomic, glycomic, and glycoproteomic changes in brain diseases may provide pathways to new therapeutic options.

1. Introduction

1.1. Importance of ECM Proteomics and Glycoproteomics

Application of high-throughput technologies such as genomics, epigenetics, transcriptomics, metabolomics, and proteomics have advanced the understanding of the mechanisms of brain diseases (Amaro et al., 2015). Recently, the importance of glycomics and glycoproteomics has been recognized for analysis of the vital roles of the brain extracellular matrix (ECM) in the progression and pathogenesis of brain diseases including Alzheimer’s disease (AD), Parkinson’s disease (PD), Schizophrenia (SZ). In this review, we summarize mass spectrometry-based proteomics, glycomics, and glycoproteomics studies on neurological disorders published over the last decade and highlight approaches that reveal altered expression of brain ECM components associated with neurodegeneration.

The brain ECM constitutes approximately 20% of the adult brain volume (Nicholson & Syková, 1998). It consists of highly glycosylated proteins, including N- and O-glycoproteins, collagens, organized into condensed networks that include perineuronal nets (PNNs) (Figure 1A). Brain ECM networks act as scaffolds that mediate cell-to-cell communication, bind secreted proteins such as growth factors, and regulate the activities of protein complexes. PNNs consist of chondroitin sulfate proteoglycan (CSPG)-containing protein domains that bind partner molecules to create an organized network that surrounds neurons and plays critical roles in neuroprotection (Suttkus et al., 2014). Further, proteoglycans modulate neural plasticity (Rowlands et al., 2018; Yamaguchi, 2000).

Figure 1:

Summary of major ECM components. (A) Perineuronal nets (PNNs) are highly glycosylated ECM structures that surround neurons and play crucial roles in neuroplasticity. (B) Heparan sulfate proteoglycans (HSPGs) found on the cell surface and pericellular matrix bind growth factors and their receptors and contribute to cell signaling. (C) The cell surface of neurons is populated with highly glycosylated molecules, whose function is modulated by glycans. (D) Chondroitin sulfate proteoglycans (CSPGs) include the hyalectan family (aggrecan, versican, neurocan, and brevican), each of which contains an N-terminal domain that binds hyaluronan, an extracellular GAG polysaccharide, and contributes to the structure of the ECM network.

The exact glycan makeup of mature matrix glycoproteins and proteoglycans remains ill-defined. About 50% of human proteins are glycosylated (An et al., 2009), and glycosylation participates in a variety of neuronal processes, including maintenance of resting membrane potential, axon firing, and synaptic vesicle release (Conroy et al., 2021). Interestingly, many ECM proteins have lectin domains that recognize glycan structures, allowing the formation of robust molecular networks. The ECM also contains sialidases, sulfatases, and heparanases, which modify glycosylation patterns on proteins (Hynes & Naba, 2012), Antibody-based and histological staining studies have shown spatial and temporal regulation of glycan epitopes in various organs (Kuppevelt et al., 1998; Pantazopoulos et al., 2013), but do not explicitly define the underlying structure. Conventional mass spectrometry-based proteomics methods often do not describe ECM protein glycosylation due to low sequence coverage and a large number of possible proteoforms (Raghunathan, Sethi, Klein, et al., 2019). Nonetheless it is clear that defining spatial and temporal distributions of ECM proteins is a crucial aspect for understanding the roles of ECM molecules in neurological disorders (Dauth et al., 2016).

Heparan sulfate proteoglycans (HSPGs) and CSPGs are integral parts of the brain ECM. Both contain glycosaminoglycan (GAG) units (heparan sulfate (HS) and chondroitin sulfate (CS) attached to the core protein via a linker tetrasaccharide (Iozzo & Schaefer, 2015). HSPGs are associated with both cell surface and ECM in a wide range of vertebrate and invertebrate tissues (Lindahl et al., 1998). The HS GAGs are synthesized as nascent disaccharide repeating units of N-acetyl glucosamine (GlcNAc) and glucuronic acid (GlcA) that undergo extensive modifications by a series of enzymes in Golgi apparatus, including N-deacetylase/N-sulfotransferase and epimerization by uronic acid epimerase to convert GlcA to iduronic acid (IdoA), and a series of O-sulfotransferases, sulfating 2O-position of IdoA, the NO- position of GlcNAc, and 6O- and 3O-position of GlcNS (Moon et al., 2012; Sasisekharan & Venkataraman, 2000; Turnbull et al., 1999) (Figure 1B). HSPGs, including glypicans and syndecans, participate in synaptic development and neural plasticity. They serve as essential components of synapse-organizing protein complexes and as ligands for leucine-rich repeat transmembrane neuronal proteins (LRRTMs), leukocyte common antigen-related (LAR) family receptor protein tyrosine phosphatases (RPTPs), and G-protein-receptor 158 (GPR158), regulating synapse formation (Kamimura & Maeda, 2021) (Figure 1C).

By contrast, CSPGs are comprised of disaccharide repeating units made up of N-acetyl galactosamine (GalNAc)-and GlcA that may be sulfated at the C2 position of GlcA, C4, and/or C6 position of GalNAc. Large CSPGs, namely aggrecan, neurocan, brevican, and versican, known collectively as hyalectans, are major components of the brain ECM (Figure 1D) (Iozzo & Schaefer, 2015). CSPGs interact with numerous growth factors, including the fibroblast growth factor family (FGF-2, FGF-7), and play critical roles in regulating cell proliferation and signaling (Mizumoto et al., 2015; Sami et al., 2020). Moreover, they are involved in the development and maturation of the nervous system as well as the pathological mechanisms associated with neurodegenerative disorders (Lin et al., 2021).

1.2. Altered ECM in neurodegeneration

Neurodegenerative diseases, including AD, PD, and SZ result from gradual and progressive losses of neural cells and changes in ECM components that lead to brain dysfunction (Bonneh-Barkay & Wiley, 2008; Dauth et al., 2016). Thus, there is growing interest in documenting the roles of GAGs, proteoglycans, and their binding partners and underpinning biosynthetic and degrading enzymes in neurodegeneration and aging (Smith et al., 2015).

AD is characterized in part by extracellular plaques made of misfolded amyloid-β protein (Aβ), which has been demonstrated to interact with key ECM molecules (Srivastava et al., 2021). Several studies showed the roles of HSPGs and CSPGs in AD pathology (DeWitt et al., 1993; Patey et al., 2008). HSPGs colocalize with Aβ deposits in human AD brain and animal models (O’Callaghan et al., 2008; Snow & Kisilevsky, 1985; Snow & Wight, 1989), alter the processing of the amyloid precursor protein (APP) and increase Aβ aggregation (Castillo et al., 1997; Cotman et al., 2000). They also interact with β-site APP-cleaving enzyme 1 (BACE1), which cleaves APP to produce the amyloidogenic β-amyloid peptide (Cole & Vassar, 2007). The accumulation and clearance of Aβ plaques depend on the sulfation patterns of the GAG chains of HSPGs (C.-C. Liu et al., 2016; G. Zhang et al., 2014). Additionally, CSPGs are associated with amyloid plaques and are involved in reduced synaptic plasticity in AD (Howell et al., 2015; W. Yang et al., 2017). Specific changes in 4-O-sulfated CSPG have been reported with aging, and similar processes may be occurring in AD (Foscarin et al., 2017). Aberrant extracellular glycoprotein expression and glycosylation have been found in AD, and several meta-studies have suggested that similar trends are common among other neurodegenerative disorders (Freitas et al., 2021; Ruffini et al., 2020; Q. Zhang et al., 2020)

PD is characterized primarily by α-synuclein pathology, which has been shown to involve key ECM molecules. HSPGs have been reported to participate in uptake and seeding of α-synuclein depositions (Holmes et al., 2013; Ihse et al., 2017). In addition, GAGs may bind proteases and delay the degradation of α-synuclein of α-synuclein (Lehri-Boufala et al., 2015). Matrix metalloproteinases (MMPs) have been identified as key players in the processing of α-synuclein (Sung et al., 2005). Importantly, HSPGs and CSPGs are involved in the uptake of α-synuclein aggregates and the spread of amyloid aggregates in AD and PD brains (Ihse et al., 2017).

Accumulating evidence points towards the critical role of the brain ECM in the pathophysiology of SZ, including a decrease in PNNs and altered expression of CSPGs in glial cells identified in several brain regions, regulating key functions relevant to SZ pathophysiology, including synaptic transmission, synaptic plasticity, and protection from oxidative stress (Pantazopoulos et al., 2021). Mature oligodendrocytes originate predominately from oligodendrocyte progenitor cells (OPCs) in the adult brain (Rivers et al., 2008). CSPGs interact with OPCs and participate in axon guidance, fasciculation, myelination, and impulse conduction (Dours-Zimmermann et al., 2009; Hunanyan et al., 2010; Ichihara-Tanaka et al., 2006; Kucharova & Stallcup, 2010; Lau et al., 2012; Pendleton et al., 2013; Sami et al., 2020; Siebert et al., 2014; Siebert & Osterhout, 2011). In SZ, lower numbers of oligodendrocytes have been reported in several brain regions, including the prefrontal cortex (Vostrikov et al., 2008) and two thalamic nuclei (Byne et al., 2008), indicating the possibility that OPCs fail to differentiate into mature myelinating oligodendrocytes.

Taken together, a growing body of work highlights compellingly the critical roles brain ECM plays in the neuropathophysiology of brain disorders. Mass spectrometry omics studies have focussed on brain ECM and the roles of its components in brain pathology. We summarize current sample preparation and mass spectrometry methods used for brain samples and scrutinize current proteomics, glycomics, and glycoproteomics studies in various brain disorders, including AD, PD, and SZ, that indicate altered ECM structures associated with neuropathophysiology.

2. Methods

2.1. Sample Preparation Methods

Sample preparation methods are compared and contrasted in Table 1, and schematics of the methods are shown in Figure 2.

Table 1:

Advantages and disadvantages of sample preparation methods for brain tissue and CSF/exosomes

Technique	Advantages	Disadvantages	References
Tissue sections
On-slide digestion	-Can target specific regions or structures of interest -Good for analysis of sulfated GAGs -Enzyme application and biomolecule extraction may be automated	-Does not permit enrichment of glycopeptides due to low sample volume -Lower spatial resolution than MALDI	(Raghunathan et al., 2020; Raghunathan, Sethi, & Zaia, 2019; Raghunathan, Sethi, Klein, et al., 2019)
Tissue microarray	-Allows high-throughput screening of tissues -Provides spatial information when combined with MALDI	-Time-consuming preparation -Necessity of experienced pathologist for preparation	(Behling & Schittenhelm, 2018; Groseclose et al., 2008; Jawhar, 2009)
Mass spectrometry imaging (MSI)	-High spatial resolution -Rapid analysis	-Not recommended for analysis of glycosaminoglycans, due to dissociation of fragile glycan substituents	(Drake, Powers, et al., 2018; Drake, West, et al., 2018; Škrášková et al., 2016)
Laser microdissection	-Allows cell-type-specific proteomics	-Requires specialized equipment	(E. Drummond et al., 2017; Hondius et al., 2016; Longuespée et al., 2016)
Solubilized tissue
In-gel digestion	-High degree of robustness and reproducibility	-2D-DIGE associated with difficulty in quantitation -Sample loss can occur due to re-solubilization of protein aggregates	(Choksawangkarn et al., 2012; Korovesi et al., 2020; Ping et al., 2020)
In-solution digestion or filter-aided sample preparation	-Allows single-pot analysis of GAGs and proteins -Can be combined with enrichment methods -May be used on tissue or biofluids, allowing potential biomarker discovery	-Sample loss can occur with repeated centrifugation steps	(Sethi et al., 2020; Wiśniewski et al., 2009, 2011)
Glycoblotting	-Can be combined with specific exoglycosidase treatment to deduce specific structures and linkages -Distinguishes different isobaric glycan isomers	-Very complex samples may present challenges for interpretation -Incompatible with proteomics analysis	(Jensen et al., 2012)

Figure 2:

Sample preparation and enrichment methods. (A) Schematic of on-slide digestion of proteins and GAGs from brain tissue slides adapted from (Raghunathan, Sethi, Klein, et al., 2019). (B) Schematic of in-solution digestion of proteins and GAGs from brain tissue or biofluids, adapted from (Sethi et al., 2020). (C) Summary of commonly used enrichment methods for glycopeptides adapted from (J. A. Klein et al., 2018; Riley et al., 2021; Xiao et al., 2016).

2.1.1. Tissue sections

Mass spectrometry analysis from tissue slides allows direct comparison of molecular patterns among regions of interest. The Zaia lab has introduced and optimized a method for on-slide tissue digestion that allows the extraction of glycan classes and peptides from small (~1-2 mm) tissue targets (Figure 2A) (Raghunathan, Sethi, & Zaia, 2019; C. Shao et al., 2013; Turiák et al., 2014). It can be used on fresh-frozen or fixed tissue and allows the quantification of 14 HS and 11 CS disaccharides, including the differentiation of positional isomers, 50 N-glycan compositions, and approximately 1200 proteins from a 1.8-mm area on a 10-μm tissue section. The enzymes are applied in serial order to the target area, incubated overnight, and digestion products are extracted and cleaned before LC-MS/MS analysis. The processes of enzyme application and biomolecule extraction may be automated with similar results (Raghunathan, Sethi, & Zaia, 2019), and may also be applied to tissue microarrays (TMAs) (Sethi et al., 2022).

Laser microdissection targets a specific brain tissue section area or single neuronal cells, relying on microscope-coupled lasers to cut defined areas of a tissue section (Espina et al., 2006; von Eggeling & Hoffmann, 2020). A number of different laser microdissection instruments exist, but the term 'laser-capture microdissection (LCM)' is often used as a catchall term. The original laser capture microdissection (LCM) had inverted microscopy and an IR laser used to melt thermoplastic film to adhere it to underlying tissue regions. The user then lifts microdissected cells as a polymer-cell composite from tissue sections on a cap of a microcentrifuge tube. It is a groundbreaking technology that isolates pure cell populations from a heterogeneous tissue sample to capture target-specific tissue or cells for downstream analysis. However, the film may become sticky and adhere to the cell and might interfere with downstream LC-MS/MS analysis even after removal of the film from the cells. There are two general classes of LCM: infrared (IR) capture systems and ultraviolet (UV) cutting systems (Espina et al., 2006). The LCM techniques have been extensively modified since first developed. Laser microdissection and pressure catapulting (LMPC), primarily pioneered by Zeiss, is a non-contact technique, first cutting the desired area and using a laser pulse to eject the tissue upward into a microfuge cap. Laser microdissection (LMD) is similar to LMPC, but the collection vessel is placed underneath the specimen, allowing the microdissected tissue to fall into the collection tube with gravity (von Eggeling & Hoffmann, 2020). The collection efficiency is higher for LMD compared to LMPC as tissue faces down and, thus, is easier to collect. LMD has been used to precisely target hippocampal sub-areas and microdissect neurons and plaques from tissue slides to gain a more specific understanding of the changes associated with AD (E. Drummond et al., 2017; E. S. Drummond et al., 2015; Hondius et al., 2016).

Another widely used method applied to brain tissue sections is mass spectrometry imaging (MSI) using Matrix-assisted laser desorption/ionization (MALDI), which has been used to study glycans and gangliosides (Drake, Powers, et al., 2018; Drake, West, et al., 2018; Škrášková et al., 2016). These MALDI methods employ a derivatization step to stabilize sialic acid residues and have been used to image N-glycan distributions at impressive resolution. Because sulfate groups cannot be stabilized by chemical derivatization, the use of MALDI imaging is not recommended for analysis of glycosaminoglycans (GAGs) (Juhasz & Biemann, 1995). TMAs are often used in conjunction with MALDI imaging to allow high-throughput screening of tissues, providing in-depth spatial information about glycans and proteins (Aichler & Walch, 2015; Groseclose et al., 2008). They may be used with LC-MS as well and are useful in instances of low sample availability as they require very little material (Meurs et al., 2020, 2021).

2.1.2. Solubilized tissue

Several proteomic methods have been developed for solubilized or homogenized brain tissue followed by in-solution or in-gel (SDS-PAGE) enzymatic digestion prior to LC-MS/MS analysis (Korovesi et al., 2020; Ping et al., 2020). The in-gel digestion technique is a robust and reproducible but laborious technique that is associated with relatively high sample losses due to the inefficient solubilization of protein aggregates (Choksawangkarn et al., 2012). However, an LMD-proteomic study compared various in-solution digestion methods with and without detergent versus in-gel digestion methods using a total area of 10 mm² of the temporal cortex, equivalent to approximately 80,000 neurons from AD brain tissues, and observed similar numbers of proteins identified between the methods. Investigators used an in-solution method without detergent for neurons microdissected using a Leica LMD system and detected 202 proteins from an area of 0.5 mm², with 159 being neuron-specific proteins (E. S. Drummond et al., 2015). Another gel-based method that has been used for solubilized brain tissue is known as 2D fluorescence difference gel electrophoresis (2D-DIGE) (Prabakaran et al., 2004), but this technique has not been extensively used in neuroproteomics due to lack of reproducibility and quantitation (Diez et al., 2010).

Filter-aided sample preparation (FASP) is another widely used approach that was initially described in 2009 and has since been modified to optimize the analysis of biomolecules of interest (Wiśniewski et al., 2009, 2011). It has been described as a universal method but is associated with limited reproducibility and loss of protein on the filter (Liebler & Ham, 2009). A single-pot FASP-type protocol has been described recently to extract GAGs and proteins from wet tissue or cultured cells. This approach can identify approximately 1000 proteins and 1 HA, 6 CS, and 8 HS unsaturated disaccharides, as well as 2 CS and 1 HS, saturated disaccharides, using 100 μg of starting material (Figure 2B) (Sethi et al., 2020). N- and O-Glycans may be released from solubilized tissue using chemical or enzymatic methods. N-Linked glycans may be released for analysis using protein N-glycosidases or by alkaline hydrolysis (Kameyama et al., 2018). While there is no single enzyme that will release all mucin-type O-glycans, they may be released by β-elimination. Specific exoglycosidases may be used to trim and further elucidate the underlying O-glycan structures (Jensen et al., 2012; Wilkinson & Saldova, 2020). Glycoblotting allows the release and analysis of both N- and O-linked glycans from a solubilized glycoprotein by blotting to a PVDF membrane, increasing the accessibility of the glycans to enzymatic or chemical release (Jensen et al., 2012). Particularly for MALDI-MS analysis, glycans may be permethylated to improve their ionization efficiency (Wada et al., 2007). Other methods have been developed that solubilize brain tissue using sequential buffer extractions to extract CSPGs present in the central nervous system (CNS) matrix and in the condensed matrix of PNNs (Deepa et al., 2006).

2.1.3. Other sample types

Especially for biomarker discovery, other sample types may be used in lieu of brain tissue. Cultured cells may be used in conjunction with isotopic labeling to study protein lifetimes and dynamics, and the increasing use of induced pluripotent stem cells provides an avenue to define protein expression levels associated with neurodevelopment (Djuric et al., 2017; Dörrbaum et al., 2020). Cerebrospinal fluid (CSF) is the only body fluid in direct contact with the brain and, as such, is a valuable source of biomarkers for brain diseases (Macron et al., 2018). Compared to blood or plasma, CSF has been less studied historically due to the more invasive method of sample collection and its nature as a complex mixture of proteins with a high dynamic range. The dynamic range can be corrected somewhat using depletion strategies, including immunoaffinity depletion and strong cation exchange, allowing the detection of low-abundance species (Guldbrandsen et al., 2014; Macron et al., 2018). Several studies have attempted discovery of biomarkers from blood or serum as signatures of brain diseases such as AD, but there has been little success in identifying a reproducible biomarker (Geyer et al., 2017). CSF, blood, and serum may be prepared for MS analysis using FASP or in-solution-based methods (Z. Chen, Wang, et al., 2021; Z. Chen, Yu, et al., 2021; Dislich et al., 2015).

Extracellular vesicles (EVs) include exosomes (50 to 150 nm in size), ectosomes/microvesicles (150 to 1000 nm), and apoptotic bodies (1000 to 5000 nm) and are released from neural cells into the extracellular space. They may be transferred from the central nervous system through the systemic circulation and, as such, may be isolated from brain tissue, CSF, blood, or plasma (You & Ikezu, 2019). EVs isolated from plasma have been found to contain brain-specific markers, suggesting that this is a potential source of biomarkers for brain diseases (Chiasserini et al., 2014; Goetzl et al., 2016). Additionally, EVs are a potential mechanism for the spread of pathology in AD (DeLeo & Ikezu, 2018; Rajendran et al., 2006); thus, studying them may provide insight into the mechanisms of neurodegeneration. EVs are isolated for analysis using a sucrose gradient and ultracentrifugation (Perez-Gonzalez et al., 2012; Théry et al., 2006), but as the knowledge of EVs grows, it becomes difficult to classify them neatly based on size or protein markers alone, creating a need for reproducible purification and characterization methods.

2.1.4. Glycoprotein, proteoglycan, and ECM enrichment and fractionation methods

Fractionation and enrichment are common practices in sample preparation for mass spectrometry. Subcellular fractionation allows for organelle-specific experiments, which may permit detection of less-abundant species (Masuda et al., 2020). Purification of synaptic vesicles from brain tissue increases the number of synaptic vesicle proteins identified by three-fold relative to unpurified tissue (Taoufiq et al., 2020). Fractionation or depletion of abundant proteins is often necessary when studying biofluids such as serum. Immunodepletion of the six most abundant proteins from serum and plasma, followed by anion-exchange chromatography and reversed-phase chromatography increased the number of protein identifications by 38% (Faca et al., 2007). In the brain, fractionation permits the study of signaling peptides, which are low abundance (Li & Sweedler, 2008). Post-translational modifications provide a convenient means for enrichment on the basis of chemistry. Phosphopeptides may be enriched using immobilized metal affinity chromatography, reversible covalent binding, and metal oxide affinity chromatography (Dunn et al., 2010).

Major glycopeptide enrichment strategies are summarized in Figure 2C. An advantage of using solubilized tissue for glycoproteomics analysis is that it permits the enrichment of glycosylated molecules, which is often necessary due to their low abundances and heterogeneities. Enrichment has been applied to other PTMs, such as phosphorylation, to elucidate their role in brain diseases (Low et al., 2021). Hydrophilic interaction chromatography (HILIC) has long been used to fractionate glycopeptides on the basis of their chemistry (Qing et al., 2020). Lectin affinity enrichment is also commonly used, but lectin affinities vary widely; one must choose the lectin reagents carefully (Merkle & Cummings, 1987; Y. Wang et al., 2006). Boronic acid enrichment is regarded as an unbiased approach for glycopeptide enrichment, but due to the relative weakness of the glycan-boronic acid interaction, it may not be suitable for the analysis of low-abundance glycopeptides (Riley et al., 2021). High-pH fractionation is widely used for phosphopeptide enrichment, but its use has recently been described for glycopeptides (Fang et al., 2016). Researchers may choose to use one or multiple of these strategies for enrichment; Chen et al. recently found that HILIC followed by boronic acid enrichment provided the greatest degree of glycopeptide coverage compared to other strategies (Z. Chen, Yu, et al., 2021). For proteoglycans, the strategies differ somewhat due to the size and acidity of the GAG chains. GAG-linked peptides may be fractionated from unmodified peptides using molecular weight cutoff filters due to their size differences relative to unmodified peptides (J. A. Klein et al., 2018). DEAE-Sepharose chromatography has been described as a method to isolate both HSPGs and CSPGs from brain tissue (Klinger et al., 1985; H. Yamada et al., 1994, 1997). Strong anion exchange (SAX) is another commonly-used enrichment technique, commonly used for GAG-linked peptides and employed on the basis of the high negative charge of GAGs (Cao et al., 2013; Riley et al., 2021; W. Yang et al., 2017). SAX and other resins are available both as spin columns and as magnetic beads, the latter of which may allow higher-throughput proteomics and glycoproteomics studies (Batth et al., 2019).

Protein-level fractionation and cell type-specific methods have been shown to improve the detection of ECM components (McCabe et al., 2021). A method for high-performance metabolically labeled secretome protein enrichment with click sugars (hiSPECS) mapped the cellular origin of CSF proteins, including 476 glycoproteins (Tüshaus et al., 2020). Primary cell cultures have also been used to obtain cell-type-specific proteomes of neurons and glia and brain region-specific proteomics experiments to allow increased detection of cell surface and extracellular proteins (Sharma et al., 2015). Subcellular fractionation by differential centrifugation maximizes protein detection in specific organelles of interest and is useful in describing the role of specific organelles in the pathophysiology of disease (Reis-de-Oliveira et al., 2019). This has been applied to the nuclear proteome in SZ and resulted in the identification of several dysregulated nuclear factors (Saia-Cereda et al., 2017). Similar methods apply for ECM enrichment and isolation of structures such as PNNs. Decellularization and selective solubilization of ECM components aid in the identification of ECM components that are insoluble under many typical lysis conditions (McCabe et al., 2021). Notably, the use of chaotropic agents, such as urea and guanidine hydrochloride alone or in combination with CHAPS detergent, increases the solubilization of ECM proteins (Ashraf Kharaz et al., 2017). PNNs may be isolated by serial extraction of brain tissue with tris-buffered saline, detergent, sodium chloride, and 6 M urea, with the PNNs in the urea fraction (Deepa et al., 2006).

2.2. Mass spectrometry and quantification of ECM components

Typical mass spectrometry methods for acquisition and quantification of the brain proteome, including data-dependent acquisition (DDA), data-independent acquisition (DIA), stable isotope labeling (SILAC), tandem mass tagging (TMT)-labeling, ITRAQ, and label-free quantification. These methods have been extensively discussed in recent reviews (Pappireddi et al., 2019; Rozanova et al., 2021; Vidova & Spacil, 2017). ECM proteins may be quantified using these methods, but as ECM proteins are highly glycosylated, the number of possible proteoforms increases exponentially (Aebersold et al., 2018) Electron transfer/higher energy dissociation (EThcD), which allows peptide and glycan fragmentation, comes at the cost of a slower duty cycle (Q. Yu et al., 2017). Dissociation of glycopeptides using HCD can be obtained with fast analyzer speeds and produces abundant diagnostic oxonium ions, but favors dissociation of the glycan especially at low collision energies. A commonly used approach is that of stepped collision energy, which produces more abundant peptide backbone fragments at high collision energies (Riley et al., 2020; H. Yang et al., 2018). ETD favors peptide backbone dissociation at the expense of lower duty cycle (Thaysen-Andersen et al., 2016). As glycoproteins are inherently heterogeneous, it is difficult to fully sample the glycoproteome, and data may contain missing values due to slower analyzer speeds. As such, data-independent acquisition methods are of interest (Hackett & Zaia, 2021). The changes to individual glycoproteins may be examined using molecular similarity approaches, which identity significantly altered glycopeptides between two sample groups (D. Chang et al., 2020; Hackett & Zaia, 2021; Sethi et al., 2022). Ion mobility is an emerging technology, which can provide an extra dimension to separate ions based on collision cross section. This increases the capacity to distinguish gas phase ions and is of interest for glycoproteomics applications. The resulting data files are large and analyzing the resulting data is challenging (Mookherjee & Guttman, 2018).

2.3. Bioinformatics tools for quantifying changes in ECM structure associated with neurological diseases

Most of the brain MS omics studies utilizes commercial database searching and quantitation tools and software, including Proteome Discoverer, MaxQuant, OpenMS, and PEAKS Studio (Han et al., 2011; Ma et al., 2003; Orsburn, 2021; Röst et al., 2016; Tyanova et al., 2016). Details about these may be obtained from recent reviews (C. Chen et al., 2020; Dupree et al., 2020). ECM proteomes can be identified using these commercial software packages. Gene set enrichment analysis (GSEA) and network topology analysis (NTA), can be utilized to assess the changes in ECM molecules. For example, WebGesTalt (Liao et al., 2019) has been used for GSEA and NTA analysis in recent brain studies (Downs et al., 2022; Sethi et al., 2022). Other publicly available gene ontology and pathway analysis platforms include DAVID, Cytoscape, Reactome, KEGG, and Ingenuity pathway analysis (Du et al., 2016; Good et al., 2021; Huang et al., 2007; Shannon et al., 2003; J. Yu et al., 2016). While these platforms include ECM genes and proteins information, they rely on computational predictions and manual curation of gene sets from experimental data from the literature. Thus, the systems and pathways extensively studied are more broadly annotated (Naba & Ricard-Blum, 2020). Towards this end, tremendous effort has been put forth in the ECM field by Naba et al. to overcome the underrepresented and misannotated ECM genes and proteins. Their initial work found that about 4% of the total human genome encodes the ECM (Naba et al., 2012). Further, the Naba group has curated an ECM database known as MatrisomeDB (http://matrisomedb.pepchem.org/), which collects and integrates experimental proteomics data on the ECM composition of healthy and diseased tissues (X. Shao et al., 2020). Other ECM databases include adhesome (Winograd-Katz et al., 2014), GO resource extracellular matrix (The Gene Ontology Consortium, 2019), laminin database (Golbert et al., 2014), and matrixDB (Clerc et al., 2019). Protein-protein interaction analysis has also been used in a recent meta-study to examine the ECM-related processes underlying neurodegeneration using BioVenn, EvoPPI, and ClueGO plugins from Cytoscape software (v.3.8.0) (Freitas et al., 2021). Machine-learning methods have also been developed to segregate disease states (Hua et al., 2019; Muraoka et al., 2020).

In addition to protein expression levels, it is of interest to quantify the changes in PTMs in ECM components. Proline hydroxylation is a common PTM of collagens, which may affect fibril formation (Kirchner et al., 2021). The degree of occupancy of hydroxyproline has been found to be altered in glioblastoma and in PD (Downs et al., 2022; Sethi et al., 2022). For glycopeptides, tools including Byonic, pGlyco, and GlycReSoft have been developed (Bern et al., 2012; J. Klein et al., 2018; Maxwell et al., 2012; Zeng et al., 2021). GlycReSoft was developed in the Zaia lab and allows for the identification and quantification of glycopeptides, assembling a list of theoretical glycopeptides from peptide and glycan hypotheses, then scoring the tandem mass spectra in the LC-MS data files (J. Klein et al., 2018; J. A. Klein & Zaia, 2020; Maxwell et al., 2012; M. Wang et al., 2021). It has been used for the identification of glycopeptides from purified glycoproteins as well as in human brain samples (J. A. Klein et al., 2018; Sethi et al., 2022). Using the resulting glycoproteomics data, molecular similarity calculations have been applied to quantify the degree of similarity between glioblastoma subtypes (Sethi et al., 2022) using a glycosimilarity tool developed in the Zaia lab (D. Chang et al., 2020; Hackett & Zaia, 2021). This tool uses a modified Tanimoto coefficient to compare chemical structures using the presence/absence of data to evaluate co-occurrences (Bajusz et al., 2015; Chung et al., 2019). In summary, existing proteomics and glycoproteomics tools may be applied to gain a deeper understanding of the ECM.

3. Mass Spectrometry-based studies to assess ECM alterations in neurological diseases

The ECM alterations discussed below are summarized in Table 2 and Figure 3. We focused our search on mass spectrometry studies within the last ten years, except a few SZ glycomics studies from 2010, as none were observed within the last decade. We used the following terms for the search including: “proteomics OR glycoproteomics OR glycomics Alzheimer’s disease,” “proteomics OR glycoproteomics OR glycomics Parkinson’s disease, “proteomics OR glycoproteomics OR glycomics Schizophrenia,” “extracellular matrix proteomics OR glycoproteomics OR glycomics Alzheimer’s disease,” “extracellular matrix proteomics OR glycoproteomics OR glycomics Parkinson’s disease,” extracellular matrix proteomics OR glycoproteomics OR glycomics Schizophrenia,” “mass spectrometry Alzheimer’s disease or Parkinson’s disease or Schizophrenia.” These were searched in various platforms, including Google (and Google Scholar), PubMed, and Web of Science.

Table 2a:

Summary of ECM proteomic (P), glycomic (G), and glycoproteomic (GP) studies in Alzheimer’s disease using human samples over the last decade.

Study Type	Brain Region/Sample Type	Method	Major Findings	Dataset Availability	Reference
Brain tissue
P	Hippocampus, parietal cortex, cerebellum; AD n=4, controls n=4 for each region	RP LC-HCD-MS/MS, quantified by iTRAQ	-950 proteins identified; 31 differentially expressed -Tenascin-R decreased in abundance in AD	Not publicly available	(Manavalan et al., 2013)
P	Synapses from hippocampus and primary motor cortex; AD n=6 and control n=6	RP LC-HCD-MS/MS (DIA)	-2077 proteins identified; 68 differentially expressed -Versican decreased in abundance in AD	Not publicly available	(R. Y. K. Chang et al., 2015)
P	Hippocampal subareas CA1 and subiculum, isolated using laser capture microdissection; n=40 with five to seven cases per Braak stage	RP LC-HCD-MS/MS validated using immuno-blotting and immuno-histochemistry	-3216 proteins identified; 372 differentially expressed -Increased abundance of tenascin-C, versican, laminin subunit beta-2, vimentin, galectin-1, annexin A5, and cathepsin D in AD.	Not publicly available	(Hondius et al., 2016)
P	Amygdala, caudate nucleus, cerebellum, entorhinal cortex, inferior parietal lobule, middle frontal gyrus, superior temporal gyrus, thalamus, visual cortex from three individual brains (one with no tangles, one with intermediate tangles, one with severe tangles)	RP LC-HCD-MS/MS	-9735 proteins identified in at least one sample; 6256 quantified strong divergence -HAPLN4, Tenascin-A are increased in abundance in a region-specific manner	Available in Proteome-Xchange, PRIDE partner repository; identifier PXD010603	(McKetney et al., 2019)
G	Brain tissue (Frontal, parietal, occipital, and temporal cortices brain tissue), serum, and CSF AD n=6 and Control n=6	Glycoblotting and MALDI-TOF MS/MS analysis of enzymatically released N-glycans	-52 N-glycan structures identified -Bisecting-type and multiply branched glycoforms were increased significantly in AD CSF and serum compared to controls.	Not publicly available	(Gizaw et al., 2016)
Cerebrospinal fluid
P	Dorsolateral prefrontal cortex and CSF, n=453	RP LC-HCD-MS/MS + TMT, protein co-expression analyzed using WGCNA	-5,668 proteins quantified, 13 protein coexpression modules identified -ECM module found to be correlated with pathological, cognitive, and functional measures	Not publicly available	(Johnson et al., 2020)
G	CSF; AD n=24, MCI n=11, healthy control n=21	Permethylated N-glycan MALDI TOF/TOF MS/MS analysis	-90 N-glycan structures identified -Decrease in sialylated N-glycans and increase in bisecting type N-glycans	Not publicly available	(Palmigiano et al., 2016)
G	CSF; AD n=8 (4 male, 4 female), control n=8 (4 male, 4 female)	Reduced permethylated N-glycans RP LC-MS/MS	-Fucosylated and bisecting GlcNAc structures were higher in abundances in females with AD, while both females and males exhibited lower abundances of high-mannose structure	Not publicly available	(Cho et al., 2019)
G	CSF; control n=31, MCI n=25, AD n=27	MALDI TOF/TOF, 2AB nano-LC-MS/MS analysis	-30 N-glycans identified in each sample pool -Increase in N-glycans carrying bisecting N-acetylglucosamine in AD	Not publicly available	(Schedin-Weiss et al., 2020)
GP	CSF; AD n=16, control n=16	RP LC-EThcD MS/MS	-285 N-glycoproteins identified -Altered glycosylation of alpha-1-antichymotrypsin and carnosinase CN1 -ECM structural constituent proteins and cell adhesion proteins enriched	Available in Proteome-Xchange Consortium, MassIVE partner repository; identifier PXD02274	(Z. Chen, Yu, et al., 2021)
GP	CSF; AD n=16, MCI n=16, control n=16	RP LC-EThcD MS/MS	-308 O-glycopeptides from 110 glycoproteins identified -Decreased fucosylation and increasing endogenous O-glycosylation in AD	Available in Mass Spectrometry Interactive Virtual Environment; deposit ID MSV000087160	(Z. Chen, Wang, et al., 2021)
Other samples
P	Extracellular vesicles; AD n=20, control n=18	RP LC-HCD-MS/MS	-949 proteins quantified; 18 differentially expressed proteins -Annexin A5 increased in abundance in AD relative to controls	Not publicly available	(Muraoka et al., 2020)
GP	Blood serum; AD n=136 and healthy controls without cognitive impairment n=183	Serum glycoproteomics RP LC-MS/MS	-871 glycoproteins were identified, including 266 and 259 unique proteins in control and AD groups. 49 and 297 differentially abundant glycoproteins in AD -Increased abundance ECM glycoproteins-ADAM-TS8, coagulation factor X Decreased abundance-von Willebrand factor C domain-containing protein 2-like, -4.25-fold increase in ECM glycoproteins and ECM dysfunction as one of the altered pathways associated with AD	Not publicly available	(Kerdsaeng et al., 2021)

Figure 3:

Major core ECM changes in brain disorders. Arrows indicate the direction of abundance change (increased or decreased). References:

a. (Hondius et al., 2016; Saia-Cereda et al., 2015; van Dijk et al., 2012)

b. (Föcking et al., 2015; Muraoka et al., 2020; Raghunathan et al., 2020; Raiszadeh et al., 2012; L. Wang et al., 2010)

c. (Föcking et al., 2015)

d. (Y. Liu et al., 2015)

e. (Hondius et al., 2016; Raghunathan et al., 2020)

f. (Manavalan et al., 2013)

g. (McKetney et al., 2019)

h. (Licker et al., 2014)

i.(Y. Liu et al., 2015; Saia-Cereda et al., 2015)

j. (McKetney et al., 2019)

k. (Jaros et al., 2012)

l. (Cho et al., 2019; Gizaw et al., 2016; Schedin-Weiss et al., 2020)

m. (Stanta et al., 2010)

n. (Raghunathan et al., 2020; Wilkinson et al., 2021)

o. (Raghunathan et al., 2020)

p. (Wilkinson et al., 2021)

3.1. Alzheimer’s Disease (AD)

Alzheimer’s disease (AD) is the primary cause of dementia in individuals over the age of 60. It is characterized by the presence of Aβ plaques and the formation of intracellular aggregates of the tau protein (Srivastava et al., 2021). Due to the role of the ECM in ligand binding and cell-cell communication, it has long been studied in the development of AD. Heparan sulfate proteoglycans, which are major ECM components, have been shown to interact with Aβ plaques and impair effective clearance (Bruinsma et al., 2010; C.-C. Liu et al., 2016; Snow et al., 1990). The prefrontal cortex is often studied due to its early involvement in Aβ deposition (Bereczki et al., 2018; Johnson et al., 2020; Q. Zhang et al., 2018). The hippocampus, the involvement of which is indicative of severe disease, is also a common region for analysis (Hondius et al., 2016; Manavalan et al., 2013). AD is classified into Braak & Braak (B&B) stages based on tau pathology (Braak & Braak, 1991). Studies often compare individuals across B&B stages to individuals with no cognitive impairment or mild cognitive impairment. Several mass spectrometry proteomics, glycomics, and glycoproteomics studies have been published from the last decade, some of which indicate the involvement of the ECM in neurodegeneration. Major ECM changes associated with AD are summarized in Table 2a and Figure 3.

Proteomics

In proteomics studies, aberrant expression of structural ECM proteins has been identified. Several isoforms of type VI collagen, as well as type XVIII collagen, were identified by weighted gene co-expression networks as being over-represented as hub proteins, which is indicative of the correlation of protein expression with the disease process (Pei et al., 2017; Q. Zhang et al., 2018). These findings are supported by proteomics studies of cerebrospinal fluid (CSF) and blood serum, in which ECM structural constituent gene sets are enriched in AD (Z. Chen, Yu, et al., 2021; Kerdsaeng et al., 2021). In addition to their structural roles, collagens also affect signaling. Endotrophin is the C-terminal cleavage product of COL6A3 and is known to induce tissue fibrosis and metabolic dysfunction (Sun et al., 2014). Type XVIII collagen may also be cleaved to generate a fragment referred to as endostatin, which affects endothelial proliferation and angiogenesis (van Horssen et al., 2002). Proteomic and glycoproteomic alterations in CSF and exosomes have also been studied in the context of AD. The major ECM finding in a recent study of exosomes was an increase in abundance of annexin A5 (Muraoka et al., 2020). Annexin A5 is a key ECM-affiliated protein that has been implicated in GAG binding as well as interaction with amyloid-β (Bartolome et al., 2020; Ishitsuka et al., 1998).

Members of the laminin family in the brain ECM, specifically laminin alpha-5 and laminin beta-2, are increased in abundance in the prefrontal cortex and hippocampus, respectively (Hondius et al., 2016; Q. Zhang et al., 2018). Laminins are ECM glycoproteins which are the most abundant non-collagen proteins in the basement membrane, shown to thicken during AD. This thickening may affect the pathophysiology of AD and affect the delivery of therapeutics, further complicating the treatment of AD (Thomsen et al., 2017). Members of the tenascin family, an ECM glycoprotein and member of the core ECM, are also dysregulated in AD. Tenascin-C, which may promote fibrosis and disruption of the blood-brain barrier, is increased in abundance in the AD hippocampus relative to controls (Hondius et al., 2016; Okada & Suzuki, 2021). The abundance of tenascin-R, by contrast, is decreased in the hippocampus (Manavalan et al., 2013). Tenascin-R is an essential component of PNNs, and its loss is associated with neurological abnormalities (Chiquet-Ehrismann, 2004; Morawski et al., 2014; Suttkus et al., 2014). Versican, which is a member of the hyalectan family of CSPGs that binds to hyaluronan, is elevated in the AD hippocampus (Hondius et al., 2016). Versican has also been implicated in mediating immune-regulated inflammation, which can be dysregulated in neurodegeneration (Wight et al., 2020; Z. Zhang et al., 2012). Taken together, these findings indicate that ECM components are altered in AD and play critical roles in AD pathology that requires further focused investigation.

Glycomics and Glycoproteomics

Modifications to N-linked glycans have been described in studies of AD brain tissue, serum, and CSF. The most commonly observed alteration is the increase in bisecting-type N-glycans; Cho et al. suggest that the difference may be sex-specific (Cho et al., 2019; Gizaw et al., 2016; Palmigiano et al., 2016; Schedin-Weiss et al., 2020). This may impact AD pathology directly, as a modification with bisecting GlcNAc has been shown to stabilize BACE1, which increases the generation of Aβ (Kizuka et al., 2016). A decrease in the level of sialylated N-glycans has also been observed in AD CSF (Palmigiano et al., 2016), as well as a decrease in the abundance of high-mannose structures (Schedin-Weiss et al., 2020). Sialic acids have been implicated in neuronal plasticity, myelin stability, and remodeling of neuronal connections, and sialyltransferase-deficient mice have displayed neurological symptoms such as impaired coordination and cognitive deficits (Rawal & Zhao, 2021; Yoo et al., 2015). Both sialylation and the presence of immature high mannose-containing glycans have been linked to aging, with possible implications for decreased brain function (Simon et al., 2019).

Glycoproteomics studies of CSF reveal alterations in both N- and O-glycoproteins(Z. Chen, Wang, et al., 2021; Z. Chen, Yu, et al., 2021). AD is associated with aberrant glycosylation of alpha-1-antichymotrypsin and carnosinase CN1 (Z. Chen, Yu, et al., 2021). Additionally, decreased fucosylation of O-glycans has been observed in AD (Z. Chen, Wang, et al., 2021). Validation of these findings in AD brain tissue, with further investigation of site-specific glycoproteomic changes, is of interest.

3.2. Parkinson’s Disease (PD)

Parkinson’s disease (PD) affects nearly 3% of the population over age 80 and is associated with the death of dopaminergic neurons in the brain, which leads to the characteristic features of tremor, rigidity, postural instability, and bradykinesia, and more variable non-motor symptoms (Henchcliffe & Beal, 2008; Poewe et al., 2017). Additionally, the diagnosis of PD depends on the presence of intraneuronal Lewy bodies and Lewy neurites, which are made up of aggregated α-synuclein (Del Tredici et al., 2002). Recent mass spectrometry-based studies have explored the roles of ECM glycans, proteins, and glycoproteins in the pathogenesis of PD. In contrast to senile or Aβ plaques, Lewy bodies do not contain HSPGs (van Horssen et al., 2004). PD studies usually include age-matched controls and may compare PD to Parkinson’s disease with dementia (PDD) or frontotemporal lobe dementia (FTD/FTLD). Major ECM changes associated with PD are summarized in Table 2b and Figure 3.

Table 2b:

Summary of ECM proteomic (P), glycomic (G), and glycoproteomic (GP) studies in Parkinson’s disease using human samples over the last decade.

Study Type	Brain Region/Sample Type	Method	Major Findings	Dataset Availability	Reference
Brain tissue
P	Locus coeruleus; PD n=6, control n=6	RP LC-MS/MS	-2495 proteins identified; 87 differentially expressed -Versican abundance elevated in PD	Not publicly available	(van Dijk et al., 2012)
P	Substantia nigra; PD n = 5 and neurologically intact controls n = 8	RP LC-MS/MS + TMT	-1795 proteins identified; 204 differentially expressed -Neurocan abundance decreased in PD	Available in Proteome-Xchange, identifier PXD000427	(Licker et al., 2014)
P	Substantia nigra; PD n=3, control n=3	RP LC-MS/MS; quantitative proteomic analysis using isotopically labeled cultured cells	-3934 proteins identified; 229 in all samples and 11 differentially abundant proteins -HAPLN2 increased in abundance in PD, annexin A1 decreased in abundance in PD	Not publicly available	(Y. Liu et al., 2015)
P, G	Prefrontal cortex Cohort 1: PD n=12, aged n=12 Cohort 2: PD n=16, aged n=13, young n=12	RP LC-MS/MS	-Increase in abundance of ECM structural components, including collagen types I, II, III, and IV and PNN-associated glycoproteins and proteoglycans -Enrichment of GAG-binding proteins, including annexins -Increased abundance of N-acetylated domains, accompanied by a decrease in abundance of N-sulfated domains	Proteomics data available in PRIDE, identifier PXD018736; glycomics data available in GlycoPOST, identifier GPST0000033	(Raghunathan et al., 2020)
G	Striatum and substantia nigra; n=33	HILIC-MS	-70 O-glycans identified -O-glycan sialylation is increased and sulfation is decreased in PD	Available through GlycoPOST, identifier GPST000185 and GPST000195	(Wilkinson et al., 2021)
GP	CSF; PD n=24, PDD n=21, AD n=9, FTLD n=6, control n=24	-LC-MS/MS, validated by immunoblotting	-Differential sialylation of Serpin A1 in PD vs. PDD	Not publicly available	(Jesse et al., 2012)
Cerebrospinal fluid
P	CSF; 3 cohorts Cohort 1a: PD n=34, control n=35 Cohort 1b: PD n=10, control n=8 Cohort 2: PD n=24, control n=12	GC-MS	-Mannose and fructose levels higher in PD -ERGIC-53 (mannose-specific lectin, marker for ER-Golgi intermediate compartment) decreased in abundance in PD	Not publicly available	(Trezzi et al., 2017)
P	CSF; PD n=81, control n=115	RP LC-MS/MS (DIA)	-341 protein groups identified; 53 differentially expressed -Abundance of SerpinC1 increased in PD	Available in PRIDE repository; accession PXD011216	(Rotunno et al., 2020)

Because PD is associated with the loss of dopaminergic neurons in the nigrostriatal pathway, proteomic, glycomic, and glycoproteomic changes to the substantia nigra and striatum are often examined (Licker et al., 2014; Raghunathan et al., 2020; Wilkinson et al., 2021). Aberrant dopamine signaling in the prefrontal cortex is associated with cognitive dysfunction observed in PD; thus, Brodmann area 9 is another common region of study (Bereczki et al., 2018; Narayanan et al., 2013; Raghunathan et al., 2020).

Proteomics

In proteomics studies, structural ECM molecules have been shown to be enriched in PD relative to controls. Abundance of collagen types I, II, III, and IV is elevated in PD, along with PNN-associated glycoproteins and proteoglycans, including tenascin-C and aggrecan (Raghunathan et al., 2020). These changes are associated with dysfunction in fibrillar collagen as well as PNNs, which form key structures in the brain. Neurocan has been identified as part of an enriched gene set in a proteomics study, but in another proteomics study, its abundance was found to be decreased (Licker et al., 2014). The two studies examined different brain regions, thus, the role of neurocan in neuronal adhesion may be affected in a region-specific manner. Versican, the role of which in inflammation is discussed in section 3.1, was also found to be increased in PD, consistent with the possible connection between inflammation and neurodegeneration (van Dijk et al., 2012; Wight et al., 2020).. Hyaluronan-binding proteins are another important group of proteins that are dysregulated in PD. Hyaluronan and proteoglycan link proteins (HAPLNs) mediate interactions between hyaluronan, a free-floating GAG, and other ECM structural components (Lindahl et al., 2015). HAPLN2, which along with HAPLN4, is highly expressed in the brain, is increased in abundance in the substantia nigra in PD relative to controls (Y. Liu et al., 2015; Q. Wang et al., 2019, p. 2). Members of the annexin family are also dysregulated in PD; annexin A5 is increased in abundance while annexin A1 is decreased in abundance (Y. Liu et al., 2015; Raghunathan et al., 2020). As described in section 3.1, annexins have been shown to bind GAGs, and their differential abundance may be indicative of signaling abnormalities (Gerke & Moss, 2002; Ishitsuka et al., 1998). Overall, ECM plays an intricate role in PD pathophysiology that warrants further investigations.

Glycomics and Glycoproteomics

Studies of brain tissue and CSF have indicated aberrant glycosylation in PD. A study of O-linked glycans identifies an increase in O-glycan sialylation and a decrease in sulfation in PD (Wilkinson et al., 2021). GAG glycomics also identifies a decrease in the abundance of N-sulfated domains, accompanied by an increase in N-acetylated domains (Raghunathan et al., 2020). Little information is available about site-specific glycoproteomic changes in PD; an early study of CSF indicates differential sialylation of Serpin A1 in PD compared to PDD (Jesse et al., 2012). Serpin A1 is a serine protease, the differential modification of which has been identified in recent years as a potential biomarker for PDD (Halbgebauer et al., 2016). Further investigation is needed to elucidate further glycoproteomic changes which are associated with PD.

3.3. Schizophrenia

Schizophrenia is a severe mental illness that affects approximately 24 million people worldwide, affecting about 1 in 222 people (0.45%) of the population over the age of 18 (Charlson et al., 2018). It is a low prevalence disorder, but it has a significant burden associated with it. While there is no cure for this disease, antipsychotic medicines employed for its treatment have shown some success. Due to individual variations of the disease, however, the treatments are only beneficial to a subset of symptoms (Nascimento & Martins-de-Souza, 2015). Thus, it is of interest to understand potential biomarkers and disease targets, of which some have been linked to the ECM.

The involvement of various brain regions in SZ has led to an interest in defining their proteomes and glycomes. SZ has been shown to be associated with morphological abnormalities of the corpus callosum, which plays an important role in hemispheric communication and lateralized brain function (Ahmadvand et al., 2017). Hypoactivity in the anterior cingulate cortex has also been shown in SZ (Nelson et al., 2015). Biofluids such as eccrine sweat and saliva have also been studied in the context of SZ, and they present a possible avenue for biomarker detection (Iavarone et al., 2014; Raiszadeh et al., 2012). Major ECM findings are summarized in Table 2c and Figure 3.

Table 2c:

Summary of ECM proteomic (P), glycomic (G), and glycoproteomic (GP) studies in schizophrenia using human samples.

Study Type	Brain Region/Sample Type	Method	Major Findings	Dataset Availability	Reference
Brain tissue
P	Corpus callosum (CC) brain tissue, 9 SZ and 7 control patients	nano LC-MS/MS, cytosolic proteins, Label free spectral counting	5678 unique peptides, 1636 proteins 65 differentially expressed (28 up, 37 down Decrease in abundance of versican and HAPLN2, increase in abundance of vimentin	Not publicly available	(Saia-Cereda et al., 2015)
P	Human post-mortem brain tissue of the supragenual (BA24) anterior cingulate cortex (ACC) N=20, 10 SZ 10 controls	RP-LC-MS/MS analysis	734 proteins identified, 135 differentially expressed proteins Decrease in abundance of Annexin A7 and A5.	Not publicly available	(Föcking et al., 2015)
Cerebrospinal fluid
G	38 serum samples, (19 SZ and 19 controls), 32 CSF samples (14 SZ and 18 controls)	2-AB labeled N-glycan NP-HPLC analysis, glycan analysis was performed by using GU values	Bisecting and sialylated N-glycans were decreased for SZ in CSF. Gender specific N-glycan signatures were observed.	Not publicly available	(Stanta et al., 2010)
Other sample types
P	Fibroblasts from 11 SZ and 11 controls	LC-MSE analysis	580 identified proteins, 16 differentially expressed, increase in Annexin A5 abundance in SZ	Not publicly available	(L. Wang et al., 2010)
P	Eccrine sweat; 78 subjects (55 controls and 23 patients)	LC-MS/MS analysis	-220 proteins identified -Increase in Annexin A5, Cystatin A and S100A7 abundance in SZ	Not publicly available	(Raiszadeh et al., 2012)
P	blood sera from 20 antipsychotic-naïve SZ patients and 20 matched healthy controls	LC-MSE analysis	-694 unique proteins were detected by LC-MSE and 312 of these were identified by at least 3 unique peptides, 35 were differentially expressed. - Increased abundance of ficolin-3, Insulin like growth factor binding protein3, and decreased abundance of fibrinogen alpha-2 in SZ patient blood sera compared to the controls	Not publicly available	(Jaros et al., 2012)
P	Whole saliva- 32 SZ, 17 bipolar disorder (BD), and 31 healthy subjects	RP-HPLC-ESI-MS top-down platform	-S100A2 (ECM-associated secreted factor) increased in abundance for SZ and BD.	Not publicly available	(Iavarone et al., 2014)

Proteomics

Among proteomics studies, several ECM-associated secreted factors have been reported to be dysregulated in SZ biofluids. S100A2 is increased in saliva(Iavarone et al., 2014), and annexin A5, cystatin A, and S100A7 are enriched in eccrine sweat in SZ (Raiszadeh et al., 2012). An increase in annexin A5 abundance is also found in skin fibroblasts (Raiszadeh et al., 2012). Extracellular S100-family proteins have roles in regulating inflammatory cells, neurons, microglia, and other neural cells, and increased abundance may be indicative of inflammatory processes (Donato, 2003). However, in a study of brain tissue, annexin A5 was found to be decreased in abundance, along with annexin A7 (Föcking et al., 2015). This difference may be attributed to the fact that the samples were from different sources, and the collective findings demonstrate that there are signaling abnormalities in the ECM of SZ. Annexins have GAG-binding properties and may function as recognition elements for proteoglycans in the extracellular space. Dysregulated expression of annexin family members may indicate ECM signaling abnormalities (Gerke & Moss, 2002; Ishitsuka et al., 1998). Vimentin, which is present as a non-filamentous form within the cells and the ECM and has been identified as an axonal growth factor, has been shown to be increased in abundance in the corpus callosum of SZ (Saia-Cereda et al., 2017; Shigyo & Tohda, 2016). Interestingly, vimentin has been observed in astrocytes in AD and PD exclusively related to senile plaques (T. Yamada et al., 1992). Additionally, decreases in abundance of the core ECM components versican and HAPLN2 has been observed (Saia-Cereda et al., 2015). Versican and HAPLN2 are both core structural molecules in the ECM, and their decreased abundance is indicative of decreased ECM structure formation in SZ (Deepa et al., 2006; Dours-Zimmermann et al., 2009; Q. Wang et al., 2019). Several ECM proteins were observed to be dysregulated in SZ patient blood sera compared to controls; abundances of ficolin-3 and insulin-like growth factor binding protein-3 were increased, and abundance of fibrinogen alpha-2 was decreased (Jaros et al., 2012). Increased fibrinogen contact has been previously reported linked to hyperfibrinogenemia, which is associated with inflammation related to vascular disease in various organs, including the brain (Brackenridge & Jones, 1968; Kerlin et al., 2004). In addition, stress has been linked to raised fibrinogen levels (Fessel, 1962), and an increase in fibrin degradation linked to inflammatory processes has been observed in SZ (Körschenhausen et al., 1996).

Glycomics and Glycoproteomics

Only a few mass spectrometry-based glycomics and glycoproteomics studies have been reported in SZ. On the other hand, various glycosylation enzymes, including FUT9, MAN2A1, TMTC1, GALNT10, and B3GAT1, have been implicated in SZ via GWAS analysis emphasizing the fundamental role of under-explored glycosylation in the SZ pathogenesis (Mealer et al., 2020). An early study of serum and CSF identified glycomic changes in SZ; specifically, bisecting and sialylated N-linked glycans were decreased in CSF (Stanta et al., 2010). Sialylation has been associated with both aging and neurodegeneration, and sialyltransferase-deficient mice display neurological differences relative to controls (Rawal & Zhao, 2021; Yoo et al., 2015). Bisecting N-glycans are most highly expressed in the nervous system and have been implicated in neurite outgrowth as well as regulation of the synthesis of other glycan epitopes (Q. Chen et al., 2020; Kizuka & Taniguchi, 2018). The decreased abundance of bisecting and sialylated N-glycans implies structural and signaling dysfunction in SZ.

3.4. Profiling ECM in the healthy brain

In recent years, numerous studies have been undertaken to profile the ECM of various regions and structures in the healthy brain. ECM components have been found to be differentially expressed in different neural stem cell niches (Lam et al., 2019). An immunohistochemistry and mass spectrometry-based study of the isocortex, hippocampus, amygdala, cerebellum, and brainstem found that expression of key extracellular matrix proteins was region-specific (Dauth et al., 2016). A multi-organ proteomics study of the ECM identified 170 out of 274 theoretical matrisome proteins and found that the brain contained less fibrillar collagen relative to other organs; additionally, brain-specific ECM facilitates the formation of neuronal networks (Kjell et al., 2020; McCabe et al., 2022). A growing area of interest is cerebrovasculature and its changes during aging, neurodegeneration, and brain injury. A study of mouse and human cerebrovasculature identified a total of 103 and 114 ECM components, respectively (Pokhilko et al., 2021). These brain studies are creating a pool of collective information to understand the ECM composition of the healthy brain, which is essential to understand prior to recognizing the role of ECM in brain disorders.

4. Conclusions & Perspectives

Neurodegenerative disorders are associated with changes in the expression of extracellular matrix (ECM) molecules. Core ECM components and ECM regulatory proteins are implicated in neurodegenerative disease processes, resulting in both structural and signaling aberrations. There has been extensive development in techniques for thorough characterization of ECM composition and bioinformatics tools at the genomic and proteomics level (Naba et al., 2016),but the role of altered glycosylation of ECM components requires specialized application of proteomics and glycoproteomics approaches. Notably, site-specific glycoproteomics studies have gained momentum in recent years with new MS technology and data processing tools. Thus, further studies will elucidate the precise roles of many ECM components in neurodegeneration. Dedicated multi-omics efforts to utilize MS-based proteomics, glycomics, and glycoproteomics will continue to shed light on the pathophysiology of neurodegenerative disorders. Importantly, LC-MS/MS analysis requires a streamlined quality control workflow to overcome biases and reproducibility issues linked to sample heterogeneity, sampling handling, data acquisition, and data processing. One approach is to utilize data-independent acquisition (DIA) rather than more commonly used data-dependent acquisition (DDA). DIA is ideal for tissue cohort studies in that it produces relatively unbiased and reproducible data (Rosenberger et al., 2017), although it presents challenges for data processing. It has been shown to produce more reproducible coverage for ECM proteins in mouse lung and liver tissue (Krasny et al., 2018). Another approach is the validation of MS results using orthogonal techniques such as immunohistochemistry, Western blotting, ELISA, or transcriptomics.

Numerous research avenues are being pursued to target the ECM therapeutically. Extracellular vesicles have been investigated for their potential as drug carriers, and two drugs targeting brain disorders are undergoing clinical trials (Herrmann et al., 2021). Drug delivery focused on tissue-specific ECM components has the potential to allow targeted drug release. Additionally, the activity of ECM-modifying enzymes such as matrix metalloproteinases, which is affected in a number of diseases, may be modified to reverse or regulate ECM changes (Ahmad, 2021). Altered ECM mechanics, particularly increased stiffness and fibrosis, are associated with a variety of diseases. Growth factor therapies and targeting of advanced glycation end products are being investigated for their potential to remodel the ECM into a healthier state (Lampi & Reinhart-King, 2018). As data acquisition and processing methods become more sophisticated, the understanding of the complex environment of the brain ECM will continue to grow and will bring us closer to understanding the role of ECM in neuropathophysiology and uncovering therapeutic options.

Abbreviations

GAGs

Glycosaminoglycans

HSPGs

heparan sulfate proteoglycans

CSPGs

chondroitin sulfate proteoglycans

LC-MS/MS

liquid chromatography-tandem mass spectrometry

ECM

extracellular matrix

AD

Alzheimer's disease

PD

Parkinson's disease

SZ

Schizophrenia

PNNs

perineuronal nets

CNS

central nervous system

CSF

cerebrospinal fluid