Study cohorts
Blood samples were collected from individuals enrolled in a primary care-based cohort (pcb-cohort), which comprises volunteer individuals from the Baixo Vouga region of Aveiro. The inclusion and exclusion criteria were defined, and volunteers were submitted to a battery of cognitive tests as previously described [23,24]. The cognitive and functional performance of volunteers was categorized based on the score obtained in 2 cognitive tests, the Mini-Mental State Examination (MMSE) and the Clinical Dementia Rating (CDR). MMSE scale cut-offs were set according to Portuguese population: 0–2 years of literacy, cutoff = 22; 3–6 years of literacy, cutoff = 24; and ≥7 years of literacy, cutoff = 27. Scores below cutoff indicate possible cognitive impairment (MMSE+) and scores equal or above were classified as normal (MMSE-). CDR scale applied scores between 0 and 3 where 0 accounts for normal, 1 for mild dementia, 2 for moderate and 3 for severe dementia stages.
The pcb-cohort, herein designated as the UA-Cohort, included a subgroup of 32 individuals that scored CDR≥1 and MMSE+ (mean age 77.38±9.17); and 9 clinically reported AD cases (1 AD scored CDR=1, (mean age 78.67±5.07). Sex- and age-matched Controls (MMSE- and CDR=0) were randomly selected from the same cohort (n=32, mean age 76.69±8.07 and n=9, mean age 77.56±4.83).
Another independent cohort, established at the Department of Psychiatry and Psychotherapy at the University Medical Center Goettingen (UMG-Cohort) was also used for biomarker candidate’s validation. The UMG-cohort comprises 12 age-matched Controls (mean age 67.58±7.74) and 12 demented individuals clinically diagnosed as ADs (mean age 73.17±10.66) according to the 2011 McKhann criteria, as previously described [25,26]. The UMG-cohort is characterized by neuropsychological testing (CERAD battery testing), and AD diagnosis of these patients is supported by CSF biomarkers (CSF-NDD) and/or PET analysis (amyloid PET and/or FDG-PET). The CSF molecular biomarkers (Total-Tau, Phospho-Tau 181, Aβ42 and Aβ42/Aβ40 ratio) were monitored and cerebral imaging tests were also carried out.
EVs isolation and characterization
EVs, with exosome-like characteristics, were isolated from serum samples as previously described [10,27,28]. Two distinct exosomes isolation methods were used: the precipitation-based ExoQuick Serum Exosome Precipitation Solution (System Biosciences) (ExoQ) and the column-based Exo-spin Blood Exosome Purification Kit (Cell guidance systems) (ExoS). In brief, serum samples were centrifuged to remove cell debris and then incubated with the respective isolation reagent, followed by a centrifugation step to pellet the nanovesicles. For ExoQ, two exosome isolations were performed: one where the EVs were eluted in PBS for Transmission Electron Microscopy (TEM) and Nanoparticle Tracking Analysis (NTA) and the other where the EVs were resuspended in RIPA for MS and Western blot (WB) analysis. For ExoS, the pellet was ressuspended in PBS, passed through a purification column and eluted with PBS. Part of the resulting EVs was used to perform TEM and NTA; while the remaining EVs suspension was mixed with RIPA buffer with protease inhibitors to lyse the vesicles, allowing subsequent analysis. All exosome-enriched suspensions were aliquoted and stored at -20 °C prior to analyses. Controls and ADs samples were subjected to the same procedure for each EVs isolation method.
Exosome’s concentration and size distribution curves were assessed by NTA, using Nanosight NS300TM instrument and NTA 3.2 software (Malvern Instruments, UK), as previously described [27]. NTA analysis was carried out in duplicate for each sample and the particle concentration was corrected by the dilution factor (1:1000).
Exosome-enriched suspensions from both cohorts were randomly selected for TEM analysis [27]. Paraformaldehyde (2%) was added to the exosome suspensions in PBS and then, exosomes were allowed to adsorb in 75 mesh Formvar/carbon grids. A 3% phosphotungstic acid solution was added to perform the negative staining. TEM images were obtained using a Hitachi H-9000 transmission electron microscope at 300 kV and images were captured using a slow-scan CCD camera.
The protein concentration of exosomal preparations was determined by BCA protein assay and 50 μg of total protein were loaded from each sample, for sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE). Following gel transfer into a nitrocellulose membrane, immunodetection was carried out for the exosomal markers Hsp70, CD63 and RAB11 and for the negative exosomal marker Calnexin. In brief, membranes were blocked in 5% non-fat dry milk and incubated with the primary antibodies: anti-HSP70 (1:500) (SPA-812), anti-CD63 (1:500) (sc-5275), anti-RAB11 (1:500) (610657; BD transduction laboratories), and anti-Calnexin (1:200) (ADI-SPA-860-J). The secondary antibodies used were the anti-mouse (7076S) or anti-rabbit IgG, HRP-linked antibody (7074S) (Cell Signaling Technology) and protein bands were detected using the chemiluminescence reagent ECL Select (GE Healthcare Life SciencesTM). Images were acquired with ChemidocTM gel imaging system (Bio-Rad).
EVs mass spectrometry analysis
For MS analysis, exosomes-like EVs were isolated using ExoQ and ExoS. For each method, serum-derived exosomes were isolated from 5 sex- and age-matched Controls (mean age 77.4±5.41) and 5 clinically diagnosed AD patients (mean age 77.8±5.59) from the UA-cohort. Subsequent biomarker validation was carried out in a higher number of samples from the UA-cohort and from the UMG-cohort.
For MS analyses, EVs preparations in RIPA buffer (ExoQ) or PBS plus RIPA buffer (ExoS) were sonicated and protein was quantified through BCA assay, using Pierce™ BCA Protein Assay kit. Loading buffer (4x) containing β-mercaptoetanol was added to exosomal samples, normalized for protein content (25 µg per sample), and separated in a 5-20% gradient SDS-PAGE. The resulting gels were stained with Coomassie Blue and each individual gel lane was excised and divided into smaller fragments, to facilitate sample digestion. The fragment corresponding to the albumin molecular weight (around 66 kDa) was excluded and thus not analyzed by mass spectrometry. The purpose was to reduce biological sample complexity, containing high levels of albumin, which may interfere with other proteins detection by MS detection. Gel fragments were washed with ammonium bicarbonate and acetonitrile and the proteins were reduced with 10 mM DTT (45 min at 56 ºC) and alkylated with 55 mM iodo-acetamide (30 min at RT). Then, gel pieces were washed again, allowed to dry and rehydrated in digestion buffer containing 12.5 µg/mL-1 of sequencing grade modified trypsin in ammonium bicarbonate. Tryptic digestion was performed as previously described [29], with minor modifications. Trypsin was added at an enzyme-to-substrate ratio of 1:30 (w/w) followed by an overnight incubation with 50 mM ammonium bicarbonate at 37ºC. The peptides were extracted by the addition of 5% formic acid (FA, Fluka), 5% FA/50% ACN (20 min each wash, 2x), lyophilized in SpeedVac (Thermo Savant) and peptides were reconstituted in 40 μL 1% FA solution.
Samples were analyzed with a QExactive Orbitrap Mass Spectrometer (Thermo Fisher Scientific, Bremen) through the EASY-spray nano ESI source (Thermo Fisher Scientific, Bremen) that was coupled to an Ultimate 3000 (Dionex, Sunnyvale, CA) HPLC (high-pressure liquid chromatography) system. The trap (5 mm × 300 µm I.d.) and the EASY-spray analytical (150 mm × 75 µm) columns used were C18 Pepmap100 (Dionex, LC Packings) having a particle size of 3 µm. Peptides were trapped at 30 μl/min in 96% solvent A (0.1 % FA). Elution was achieved with the solvent B (0.1 % formic acid/80% acetonitrile v/v) at 300 nl/min. The 92 min gradient used was as follows: 0–3 min, 96% solvent A; 3–70 min, 4–25% solvent B; 70–90 min, 25–40% solvent B; 90–92 min, 90% solvent B; 90–100 min, 90% solvent B; 101-120 min, 96% solvent A. The mass spectrometer was operated at 1.8 kV in the data dependent acquisition mode. A MS2 method was used with a FT survey scan from 400 to 1600 m/z (resolution 70,000; AGC target 1E6). The 10 most intense peaks were subjected to HCD fragmentation (resolution 17,500; AGC target 5E4, NCE 28%, max. injection time 100 ms, dynamic exclusion 35 s).
MS data analyses
Spectra were processed and analyzed using Proteome Discoverer (version 2.2, Thermo), with the MS Amanda (version 2.0, University of Applied Sciences Upper Austria, Research Institute of Molecular Pathology) and Sequest HT search engines. Uniprot (TrEMBL and Swiss-Prot) protein sequence database (version of October 2017) was used for all searches under Homo sapiens. Database search parameters were as follows: carbamidomethylation of cysteine, oxidation of methionine, and the allowance for up to two missed tryptic cleavages. The peptide mass tolerance was 10 ppm and fragment ion mass tolerance was 0.02 Da. To achieve a 1% false discovery rate, the Percolator (version 2.0, Thermo) node was implemented for a decoy database search strategy and peptides were filtered for high confidence and a minimum length of 6 amino acids, and proteins were filtered for a minimum number of peptide sequences of 1. The obtained results were further filtered, applying a cut-off at 1.5-fold increase and another at 0.5-fold decrease. Also, abundances found in less than 2 out of 5 samples were not regarded as being present in the respective condition (AD or Control) and, when no abundance was measured for one of the groups of samples, a 100-fold increase/0.01-fold decrease was considered for the ratio.
Bioinformatic analysis of EVs
Proteomes obtained by MS (gene names from the proteins identified) were initially overlapped through Venn diagrams with a serum exosomal list (Exo Serum list), obtained from databases and from literature search as described in [28], using the Bioinformatics and Evolutionary Genomics website (http://bioinformatics.psb.ugent.be/webtools/Venn/; accessed on 3rd February 2020) to determine the percentage of exosomal proteins present in EVs samples analysed. Only MS proteins for which gene name was available were included in the overlap. From the exosomal proteomes obtained by MS, seven immunoglobulin chains were not included in this analysis since no gene name was available. Additionally, the set of proteins identified for Controls and ADs for each kit by MS (ExoQ and ExoS) was categorized according to their Gene Ontology (GO) annotation, obtained from the UniProt-SwissProt database. The GO terms were filtered according to the Generic GO Slim and categorization was carried out for “Molecular Function” and “Biological Process”.
The protein lists identified by MS were then analyzed through the use of a dedicated software framework (SysBioTK) as previously reported [2], with the exception of the Partial Least Square (PLS) Analysis.
Data was prepared independently for each analysis (AD vs Control for ExoQ; AD vs Control for ExoS). In a first step, the protein abundances obtained from MS were normalized by the median of the protein abundances of the sample. The abundances were then independently transformed for each protein in each sample through the use of the binary logarithm. For some proteins, in some samples, there was no abundance data from MS, regardless that these proteins were not removed. The prepared data was then converted into tabular format and exported into a text file for later use with Metaboanalyst 4.0 [30] for the Partial Least Square Analysis (PLS) analysis (performed at 11th February 2020), in order to maintain a consistent data set. PLS analysis was performed to evaluate which kit had the highest discriminatory capacity between Controls and AD cases.
To identify proteins with statistically significant differences in abundance, a Welch’s t-test with a significance level of 5% (α=0.05) was applied to the mean “normalized and transformed abundance” of the protein in each condition (i.e. the mean across samples for each condition). Volcano plots were created by plotting, for each protein, the p-Value of the Welch’s t-test against the fold increase of the mean “normalized and transformed abundance” of the protein. The fold change threshold was set to 2 and a line representing the 5% significance level was drawn.
Heatmaps were created by taking the “normalized abundance” for each protein in each sample. The dendrograms were calculated using the Ward method to cluster similar samples and proteins together. A Euclidian distance metric was used to calculate the distances for the Ward method. The color scale represents the “normalized abundance”.
EVs biomarker candidate analyses
Following MS and bioinformatic analysis, biomarker validation was then carried out in a higher number of samples from the UA-cohort and from the UMG-cohort. WB analyses were performed to assess the patterns of two biomarker candidates identified by mass spectrometry: alpha-1-antichymotrypsin (AACT) and C4b-binding protein alpha chain (C4BPα). The protein concentration of exosome samples isolated with ExoQ were determined, and 50 µg of protein were loaded, per sample, in a 5-20% SDS-PAGE followed by proteins transfer to nitrocellulose membranes. Membranes were then blocked with 5% non-fat dry milk and incubated with the primary antibodies anti-AACT (1:500) (sc-59430; Santa Cruz Biotechnology) and anti-C4BPα Antibody (1:500) (sc-398720; Santa Cruz Biotechnology). Subsequently the membranes were incubated with the anti-mouse IgG, HRP-linked antibody (1:2000) (Cell Signaling Technology). Protein bands were detected, as described above, using the chemiluminescence reagent ECL Select (GE Healthcare Life SciencesTM) and images were acquired with ChemidocTM gel imaging system (Bio-Rad). AACT or C4BPα densitometry values for each individual sample were normalized to an exosomal pool loaded in every membrane. Graphs presented express the relative density ratios. Further, AACT and C4BPα levels were also evaluated by enzyme-linked immunosorbent assay (ELISA), in serum-derived exosomes of the same individuals, using the commercial Human AACT ELISA Kit (ab217779; Abcam) or the Human C4 binding protein A ELISA Kit (NBP2-60550; Novus Biologicals), according to manufacturer’s instructions. EVs samples were diluted and equal amounts of protein were used for AACT or C4BPα quantification.
Statistical analysis
Statistical analysis were carried out using SPSS version 27 (IBM) or GraphPad Prism 7 (GraphPad Software, La Jolla, California, USA). Data distribution was assessed by Shapiro-Wilk test. Exosomes concentration, mode and the levels of biomarker candidates were compared using unpaired t-tests. Kolmogorov-Smirnov was used to compare the particle size distribution. p-values ≤0.05 were considered significant.