Background
Dementia originates from a range of neurodegenerative and proteinopathic diseases that are predicted to affect over 130 million people worldwide by 2050 [
1]. Alzheimer’s disease (AD) is the leading cause of this epidemic syndrome, and it accounts for about 60% of dementia cases [
2]. Cerebrovascular disease (CVD) accounts for about 20% of dementia cases [
3] and often coexists with AD in AD + CVD, a mixed form of dementia that remains as the most prevalent subtype among the elder population [
4].
Several protein posttranslational modifications (PTMs) are implicated in the neuropathology of dementia [
5] which are also known as degenerative protein modifications (DPMs) [
6,
7]. Understanding the role(s) of DPMs is thus expected to provide novel insight on the pathogenesis of these aging-associated diseases, and it may aid on the development of new diagnostic and therapeutic tools. Two of these crucial DPMs that are thought to contribute to the neuropathology of dementia are citrullination and carbamylation [
8‐
10]. Citrullination is an aging-associated DPM [
11,
12] mediated by peptidylarginine deiminases (PADs), a calcium-dependent family of enzymes that generate the non-coded amino acid citrulline (Cit) from arginine (Arg) residues by hydrolysis [
13]. Carbamylation of lysine (Lys) residues oppositely occurs spontaneously using the urea intermediate isocyanic acid as substrate to generate the non-coded amino acid homocitrulline (HCit) [
10]. Cit and HCit are almost identical residues and differ only by one carbon in their side chain, although HCit results from a non-enzymatic chemical reaction and Cit formation is enzymatic [
14]. Particularly, both non-coded amino acids are considered as ureido DPMs (uDPMs); hence, they share an ureido group in the side chain [
14]. Citrullination is known to cause loss of positive charge in Arg residues, which in turn disturbs the structural order of the protein and makes it more prone to proteolysis [
11]. Similarly, carbamylation neutralizes the positively charged Lys residues in an irreversible manner [
15] and can result in altered function and pathogenic conformation [
16,
17]. Cit and HCit, thus, favor protein denaturation and proteolysis and can also be recognized by the immune system as non-naturally coded amino acids [
6,
7,
11,
18]. This former fact leads to the apparition of autoimmunity and inflammation, which are at the very basis of several chronic human diseases [
19‐
21].
The neuropathology of dementia, including AD + CVD, is associated with dysfunctional proteins that form oligomeric structures in the brain parenchyma known as amyloids [
6,
22]. Amyloids are resistant to proteolytic degradation and consequently sequestered into inclusion bodies, which are further compartmentalized in the brain parenchyma to reduce toxicity [
23‐
25]. Although the role of protein inclusions appears mainly to be protective [
23], chronic accumulation of these structures in the brain is likely to threaten neuronal function and viability [
26]. Within the central nervous system (CNS), several DPMs have been associated with the formation and accumulation of amyloids [
5] and some studies have suggested an association between citrullination, proteinopathy [
27,
28], and brain vascular dysfunction [
29,
30]. However, despite the implication that uDPMs might be involved in the neuropathology of AD + CVD, the extent of uDPMs in affected brain cells is unknown and the role(s) of citrullination and carbamylation in the neuropathology of dementia remain far to be fully elucidated [
7]. Similarly, although uDPMs are associated with multiple sclerosis (MS) [
20] and other autoimmune diseases [
21], any apparent nexus between uDPMs and neuroinflammation in AD + CVD has yet to be defined.
The ability to identify and quantify citrullinated and carbamylated proteins in complex samples is essential to understand their role(s) in normal cellular processes and disease states. Though until now, very few uDPM-containing proteins have been identified and characterized from brain-diseased proteomes [
31]. This is mainly due to the chemical and structural similarities that Arg and Lys present with their respective modified counterparts, and between Cit and HCit, which by the use of classical biochemical methods result in significant cross-reactivity [
14,
21,
32]. Recent advances in liquid chromatography mass spectrometry (LC-MS/MS)-based proteomics have made possible the unbiased identification of proteins in complex samples and characterization of isoforms expression, turnover rate, subcellular localization, PTMs, and quantification of altered abundances in disease states [
33]. Previously, we have adopted such approaches to characterize neurodegenerative proteomes and other conditions [
6,
7,
34]. As part of these studies, we recently developed ultracentrifugation-electrostatic repulsion hydrophilic interaction chromatography (UC-ERLIC)-coupled mass spectrometry to isolate and characterize soluble and aggregated proteins from human brain tissues [
35]. Here, we have used this approach to undertake a comparative characterization of uDPMs in the soluble and aggregated postmortem brain proteomes of AD + CVD and age-matched controls. Our results show the novel identification of relevant dementia-specific citrullinated and carbamylated brain proteins. This study thus provides new insight into the implications of uDPMs in the neuropathology of AD + CVD and will pave the way for further investigations on uDPMs as potential targets for therapeutic intervention or disease-state biomarkers in this fatal dementia subtype.
Methods
Chemicals and reagents
All the reagents used in this study were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise specified. Protease inhibitor cocktail tablets were purchased from Roche (Basel, Switzerland), and sequencing-grade-modified trypsin was purchased from Promega (Madison, WI, USA).
Postmortem brain tissues
AD + CVD and healthy age-matched control (Ctrl) autopsied human brain tissues (middle temporal lobe, BA21 region) were generously donated by the Harvard Brain Tissue Resource Center (HBTRC, Boston, MA, USA). All patients met clinical diagnosis for dementia and histological criteria for AD + CVD (mixed dementias) at the time of autopsy. The available details of the samples, which include age, postmortem delay, and gender can be found in Additional file
1: Table S1. Three biological replicates were independently analyzed in this study for each experimental condition (soluble and particulate proteomes). Brain tissues were stored in liquid nitrogen from the time of autopsy and subsequently kept at −150 °C until use. BA21 brain region was dissected, meninges were removed, and ~ 100 mg of tissue from each subject were washed thrice during 10 min in 1X PBS.
Homogenization of postmortem brain tissues
Dissected brain tissues were homogenized as previously described [
36], and all the following described procedures on postmortem tissues were performed on ice. Brain tissues were suspended in 250 μL of homogenization buffer (0.5% (
w/
v) sodium deoxycholate (SDC) in 100 mM ammonium acetate (AA) (pH 6.0) or 1% (
w/
v)
N-lauroylsarcosine in 30 mM Tris-HCl (pH 7.4) and 150 mM NaCl and were supplemented with protease inhibitor cocktail. Approximately, 100 mg of previously washed metallic beads (0.9–2.0-mm particles) were added to each safe-lock tube that contained the tissues and homogenization buffer. Homogenization was then performed using the bullet blender homogenizer (Next Advance, NY, USA) at high intensity during 5 min at 4 °C. The tissue homogenates were subsequently centrifuged at 10,000 ×
g, 4 °C, for 10 min, and the supernatants were collected. Remaining pellets were subject to further rounds of homogenization till the pellet was not observable. The obtained supernatants from each round of homogenization were finally combined and briefly vortexed.
Isolation of the soluble and particulate brain proteomes by UC-ERLIC
Soluble and particulate brain proteomes were isolated from brain homogenates by ultracentrifugation-electrostatic repulsion hydrophilic interaction chromatography (UC-ERLIC)-coupled mass spectrometry-based proteomics as we recently reported [
35], although with minor modifications in this study. Soluble homogenates were centrifuged at 3000 ×
g, 4 °C, during 10 min to remove cell debris. Supernatants were then subjected to ultracentrifugation at 112,000 ×
g, 4 °C, during 60 min, transferred to a new tube, and subjected to two further rounds of ultracentrifugation under the same conditions. Supernatants obtained from each round of ultracentrifugation were combined and transferred to new tubes (soluble fraction). The generated particulate pellets from the last two ultracentrifugation rounds were re-solubilized in 150 μL of 5% SDC (
w/
v) in 100 mM AA or 3% sodium dodecyl sulfate (SDS) in 1X PBS when the sample was homogenized in sarcosyl and combined to obtain the final particulate proteome.
In-solution tryptic digestion of human brain proteins
Soluble and particulate proteomes were subject to in-solution tryptic digestion as previously described [
37]. Briefly, 0.5 mg of proteins from each sample and condition previously quantified by bicinchoninic acid protein assay (BCA) were reduced with 10 mM dithiothreitol (DTT) at 60 °C during 30 min. Proteins were alkylated with 20 mM iodoacetamide at room temperature protected from the light during 45 min. Subsequently, the samples were diluted 10-fold with 100 mM AA containing 10 mM DTT and incubated at 37 °C for 30 min. Digestion was performed at 30 °C overnight at 1:20 (
w/
w) enzyme-to-substrate ratio using sequencing-grade-modified trypsin. Enzymatic digestion was quenched by addition of 0.5% formic acid (FA). SDC under acidic conditions was pelleted by centrifugation at 12,000 ×
g, 4 °C, for 10 min as previously indicated [
37]. The supernatant containing tryptic-digested peptides was then collected, and the SDC pellet was re-suspended in 0.5% ammonium hydroxide for further peptide recovery. SDC was again precipitated by the addition of 0.5% FA and pelleted by centrifugation at 12,000 ×
g, 4 °C, for 10 min. After centrifugation, the supernatant was collected. Recovery of peptides from the SDC pellet was performed thrice. All the supernatants were combined together. Tryptic-digested peptides were desalted using a C-18 Sep-pack 1 g cartridge (Waters, Milford, MA, USA). Elution was performed with 1 mL of 75% acetonitrile (ACN), 0.1% FA buffer. The eluted peptides were then dried using a vacuum concentrator (Eppendorf, Hamburg, Germany) and reconstituted for HPLC fractionation with 200 μL of 80% ACN, 0.1% acetic acid.
In-gel digestion of human brain proteins
SDS-solubilized proteins from the sarcosyl particulate fractions were resolved in a 12% SDS-PAGE gel. PAGE gel was stained using coomassie brilliant blue, and protein lanes were divided in 12 main areas according to the observed protein intensities. Gel lanes were cut into approximately 1-mm2 cubes and washed with 25 mM ammonium bicarbonate (ABB) in 50% ACN. Peptides contained in distained gel pieces were reduced in 10 mM DTT in 25 mM ABB at 60 °C for 1 h and subsequently alkylated in 55 mM IAA in 25 mM ABB at room temperature for 45 min in the dark. The gel cubes were then dehydrated in 100% ACN in two washes for 1 min each and dried in the speedvac for 5 min at 30 °C. Tryptic digestion of proteins was performed overnight at 37 °C by the addition of 10 ng/μl of sequencing-grade-modified trypsin (prepared in 25 mM ABB buffer). Tryptic-digested peptides were extracted from the gel cubes with 50% ACN and 5% acetic acid by vigorous vortexing during 30 min. The extraction step was repeated four times, and the obtained supernatants were combined and dried in the speedvac. The dried peptides were then re-suspended in 3% ACN, 0.1% FA, for subsequent LC-MS/MS analysis.
High-pressure liquid chromatography fractionation
Reconstituted tryptic peptides after in-solution digestion were fractionated by a PolyWAX column (4.6 × 200 mm, 3 μm, PolyLC, Columbia, MD, USA) using a Shimadzu Prominence UFLC system (Kyoto, Japan). Peptide intensities were monitored at 280 nm. Peptides were separated in a 72-min gradient using 80% ACN, 0.1% acetic acid, as mobile phase A, and 10% ACN, 0.1% FA, as mobile phase B. Separation of peptides was carried out in a 60-min gradient at a flow rate of 1 ml/min as follows: 0% for 5 min, 0–20% B for 25 min, 20–33% B for 10 min, 33–60% B for 10 min, and 60–100% B for 5 min, followed by 10 min at 100% B. Fractions were collected every minute, partially dried in a vacuum concentrator to reduce the volume, and pooled according to peak intensities. Combined fractions were dried completely and reconstituted in 3% ACN, 0.1% FA.
Liquid chromatography mass spectrometry analysis
Analysis of peptides was performed using an Orbitrap Elite mass spectrometer coupled with a Dionex UltiMate 3000 UHPLC system from Thermo Fisher Scientific Inc. (Bremen, Germany). Approximately, 2 μg of peptides were injected into a reverse phase Acclaim PepMap RSL column (75 μm ID × 15 cm, 2-μm particle size, Thermo Fisher Scientific Inc.) maintained at 35 °C and using a flow rate of 0.3 μL/min. Peptides were separated in a 60-min gradient using 0.1% FA as mobile phase A and 90% ACN, 0.1% FA, as mobile phase B. The gradient used for the separation of peptides was as follows: starting at 3% mobile phase B for 1 min, 3–35% mobile phase B over 47 min, 35–50% mobile phase B over 4 min, and 50–80% mobile phase B in 6 s and maintained isocratic for 78 s. Then, reverted to initial conditions over 6 s and maintained isocratic for 6.5 min. Tryptic-digested peptides were sprayed using a Bruker-Michrom Inc. Michrom’s Thermo Captive Spray nanoelectrospray ion source (Bruker-Michrom Inc., Auburn, CA, USA) with 1.5 kV spray voltage. Orbitrap Elite mass spectrometer was set to positive mode for data acquisition using Xcalibur 2.2 SP1.48 software (Thermo Fisher Scientific Inc., Bremen, Germany). Data acquisition was performed by alternating between full Fourier transform-mass spectrometry (FT-MS; 350–2000 m/z, resolution 60,000, 1 μscan per spectrum) and FT-MS/MS (150–2000 m/z, resolution 30,000, 1 μscan per spectrum) for the ten most intense ions above 500 count threshold with charge > + 2. Fragmentation was performed using high-energy collisional dissociation (HCD) mode with 32% normalized collision energy. Automatic gain control values for FT-MS and FT-MS/MS were set at 1 × 106.
LC-MS/MS data was searched in PEAKS Studio version 7.5 [
38] (Bioinformatics Solutions, Waterloo, Canada). Precursor ion tolerance and fragment ion tolerance were kept during the database search at 10 ppm and 0.05 Da, respectively. Trypsin with non-specific cleavage on both ends of the peptide was allowed. Carbamidomethylation of cysteine residues was considered as fixed modification. The PEAKS PTM algorithm [
39] was used to identify the presence of uDPMs in the brain proteomes. False discovery rate was kept as default by the software, and decoy-target identification of the peptides was performed. To extract the most abundant peptides from the particulate brain proteomes, only those peptides with relative signal intensities higher or equal to 10
8 were considered. This restrictive criterion allowed for the unique consideration of the most confidently identified peptides. The mean signal intensity was calculated individually for every subject and experimental condition (soluble and particulate brain proteomes) and contributed to the calculation of the grand mean signal intensity for each group, which has been reported here.
Enrichment distribution of the AD + CVD particulate brain proteomes was analyzed using the enrichment analysis software FunRich [
40] as described [
41]. The absolute number of unique proteins with identified citrullinated (Arg) and carbamylated (Lys) residues was included in this analysis. To analyze the aggregation pattern of the identified neuroinflammatory markers in AD + CVD only the proteins that were consistently identified in at least two individuals of the same group were considered. Student’s
t test was performed to establish statistical significance between experimental groups (
p < 0.05). All statistical analyses and graphs were performed and plotted in GraphPad Prism v6.01 (La Jolla, CA, USA). The bioinformatics study of the biochemical composition of hypercitrullinated proteins in the brain proteome of AD + CVD was performed using an in-house-created macro for Microsoft Excel. All the possible citrulline triplets were screened in this analysis resulting in a total of 400 combinations as shown in Additional file
2: Dataset 1. The total signal intensity for every group of peptides triplet was then ranked from the highest to lowest. Only those top five triplets that were consistently identified as hypercitrullinated in all subjects and conditions are reported. Data is reported as mean ± standard deviation (SD).
Discussion
Unbiased profiling of the soluble and particulate brain proteomes confirmed that formation of amyloids in the brain parenchyma is an aging-associated process that takes place with independence of the apparition of dementia symptoms. Higher accumulation of proteins was though observed in the brains of dementia subjects. These findings were consistent with previous reports showing that there is basal amyloid formation in the brain parenchyma of aged subjects [
44] and higher accumulation of dysfunctional proteins in brain tissues undergoing neurodegeneration [
35,
45]. Detailed analysis of the particulate proteomes has shown immune response-related proteins as a protein category that becomes preferentially aggregated in the temporal lobe of dementia subjects. Although neuroinflammation is considered a factor that plays a causal role in the development of dementia [
46,
47], the existing links between proteinopathy and neuroinflammation in this progressive and fatal disease remain poorly understood [
48].
Increase in uDPMs, as identified in dementia brains, suggests activation of a defective immune and pro-inflammatory response [
48]. uDPMs are associated with apparition of inflammation and proteinopathy; however, only very few proteins that might mediate these pathogenic outcomes in dementia have been until now identified [
27,
28]. We previously reported that an increase in citrullination in certain MBP residues did not apparently influence the aggregation ability of the protein [
49]. Here, we globally confirm this finding as citrullinated proteins tended to remain soluble in the temporal lobe of AD + CVD subjects. In further detail, we found that the purinergic receptor P2X7 was modified by a dementia-specific Cit residue in the soluble brain proteome of AD + CVD. To the best of our knowledge, this is the first time that this crucial immune receptor appears modified by PADs in a dementia-relevant manner. Activation of the immune response in the brain is an ATP-dependent process initiated by injury and death of the surrounding cells [
50,
51] and mediated by P2X7 in microglia [
52‐
54], a receptor that also induces increase in citrullination and release of Cit residues out of the cell [
55]. Similarly, several dementia-specific Cit sites were identified in histones H2 and H4 in the soluble brain fractions, which suggests ongoing formation of neurotrophil extracellular traps (NETs). NETs have been proven as molecular culprits in the pathogenesis of rheumatoid arthritis (RA) and other autoimmune diseases [
56‐
59]. Formation of NETs is mediated by citrullination of histones, which neutralizes the charge of these molecules promoting chromatin decondensation [
60]. Besides, the following proteins CNP, GFAP, Alpha-internexin, Hexokinase-1, Neurofilament Light Polypeptide, Adduccin-1, Reticulon-3, Adenylate kinase isoenzyme 1, Transgelin-3 Synaptotagmin-7, Stathmin, P2X purinoceptor 7, Heat shock protein HSP 90-alpha, Neuromodulin, MBP, and Cortactin were also identified containing dementia-specific Cit residues in the soluble brain proteome of AD + CVD. Considerably, all these modified proteins have been previously implicated in autoimmunity and inflammation in several human diseases [
61‐
75], though many of them were not previously identified in the sphere of dementia. An increase in citrulline in soluble brain fractions was associated with the increase of the MAC protein complement factor-5 in AD + CVD, the unique neuroinflammatory marker found upregulated in the soluble fraction of dementia subjects compared to age-matched controls. We also observed that an increase in citrullination in AD + CVD brains was favored by the small amino acid Ala and the aromatic amino acids Pro and His at the flank of Arg residues. The biochemical patterns that influence aberrant citrullination in dementia hold great promise in the search for novel pharmacological targets to fight chronic neuroinflammation in this disease.
Our main aim for this study was also to understand how uDPMs can influence the aggregation ability of temporal lobe proteome in AD + CVD. We observed that HCit, a spontaneous non-enzymatic aging-associated uDPM that becomes hardly distinguishable from Cit by conventional biochemical methods [
14,
32,
76], seems to clearly contribute to brain proteinopathy in this dementia subtype. The observed enrichment of HCit dementia-specific sites into the particulate fraction of AD + CVD brains may indicate involvement of HCit in early brain inflammatory processes. Mydel and colleagues demonstrated that HCit triggered early activation of proinflammatory cells, which was the primary step to recognition of Cit residues in autoimmune inflammation [
77]. Besides, Currais and colleagues [
78] recently suggested that certain proteins from the particulate brain fraction may contribute to the progression of neuroinflammation during dementia. In line with this hypothesis, we highlight the proinflammatory proteins GFAP and BASP1 which appear as good candidates for further research in this direction. These proteins were identified in this study as containing dementia-specific HCit residues in the particulate brain fraction of dementia subjects. Other proteins identified with exclusive AD + CVD particulate HCit sites are alpha-spectrin, sodium/potassium ATPase, actin 1, V-type proton ATPase 3, and calcium-transporting ATPase 4. Gorisse and colleagues suggested that although HCit-containing proteins should be degraded by common proteolytic mechanisms, some of these proteins seem to show surprising slow turnover rates even when the protein quality control systems in the cells are functional [
10]. Similarly, HCit is known to promote endothelial dysfunction and vascular disease [
79]; however, uncovering its specific contributions to the pathology of the neurovascular unit in mixed dementias requires of further research.