Patients and isolation of enriched exosome fractions
A total of 52 ESKD patients in PD treatment followed up at the Nephrology Department of the Gaslini Children’s Hospital were included in the study. Written informed parental consent was obtained before enrolment. The main demographic and clinical features are summarized in Table 2. The inclusion criteria were defined as follows: patients up to 18 years old, on stable PD for more than one month, without peritonitis in the three months preceding the study, and who had not received a previous kidney transplant. Twelve randomly selected patients were included in the proteomic analysis: six patients had primary focal segmental glomerulosclerosis as baseline nephropathy (FSGS group) and 6 have been affected by other diseases (No FSGS group) (Table 2). A group of 40 patients (20 FSGS and 20 No FSGS) were included in the validation part of the study.
PM function was evaluated through a 4-h peritoneal equilibration test (PET) conducted by using 1000 ml per m2 of patient’s body surface area of a 2.27% glucose PD solution56. Urea and creatinine dialysate-to-plasma concentration ratio (D/P) and glucose dialysate to initial dialysate concentration ratio (D/D0) were calculated.
Patients have been treated with automated peritoneal dialysis (APD) using biocompatible glucose-based solutions, with different glucose concentrations (1.36%, 2.27% and 3.86%) according to the required fluid removal, and with bicarbonate/lactate buffer.
The study was carried out in accordance to Italian and international ethical guidelines and approved by the Comitato Etico Regione Liguria (number: 408REG2014).
Sample collection was standardized by performing the analyses on the PDE obtained at the end of a 4-h PET (see above). Exosomes from mesothelial peritoneal cells were isolated by centrifugation plus immuno-magnetic beads affinity capture. Aliquots (100 ml) of PDE at PET were centrifuged at 22,000×g for 120 min at 16 °C to remove cells, debris, microvesicles and organelles such as mitochondria. Supernatants were then centrifuged at 100,000×g for 120 min at 16 °C to pellet the exosomes. The exosomal pellet was resuspended in 1 ml 0.25 M sucrose, loaded onto 1 ml 30% sucrose cushion and centrifuged at 100,000×g for 120 min at 16 °C. The pellet was rinsed in PBS and centrifuged again at 100,000×g for 120 min at 16 °C.The final pellet was stored at − 80 °C until use.
Immuno-magnetic beads capture
The method is based on the capture of a specific subset of exosomes from peritoneal dialysis effluent using a biotinylated antibody and streptavidin magnetic beads.
Enriched exosomes fractions were mixed with polyclonal biotin-conjugated anti-human mesothelin (MSLN) antibody (LifeSpan BioSciences, Seattle, WA, USA) and incubated 4 h at RT with gentle rotation. Then, streptavidin Dynabeads (ThermoFisher) were added according to the procedure of the manufacturer. Briefly, exosome-antibody-dynabeads complexes were incubated for 30 min at RT with gentle rotation, placed on the magnet and rinsed five times to remove unspecific exosomes and unbound antibodies.Then, sample was removed from the magnet and bound exosomes collected by adding 250 µl of elution buffer. Finally, supernatant was centrifuged at 100,000×g for 120 min at 16 °C to pellet the exosomes pelleted. Such rinse/centrifugation cycle was carried out five times to obtain a clean anti-human mesothelin-positive exosome fraction. Size and purity of the isolated exosomes were assessed by DLS.
Dynamic light scattering
Exosome size was determined by DLS using a Zetasizer nano ZS90 particle sizer at a 90° fixed angle (Malvern Instruments, Worcestershire, UK). The particle diameter was calculated using the Stokes–Einstein equation. For particle sizing in solution, exosome aliquots were diluted in 10% PBS and analyzed at a constant 25 °C.
Western blotting
Expression of exosomal or human peritoneal mesothelial cells (HPMC) markers were detected by western blot. Aliquots of exosome fractions or whole lysate of HPMC were solubilized in 2% w/v SDS, 10% glycerol and 62.5 mM Tris–HCl pH 6.8 and separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred to a nitrocellulose membrane. Full length membrane was blocked, rinsed and cut perpendicular to the electrophoresis migration front to obtain a full length strips of whole samples and to allow the individually labeled and detection with one of the following primary human antibodies diluted in 3% (w/v) bovine serum albumin (BSA) in PBS containing 0.05% v/v Tween-20 (PBS-T): monoclonal anti-CD63 (Novus Biological, Littleton, CA, USA, 1:1000 clone H5C6), monoclonal anti-CD81 (Novus Biological, 1:1000 clone 1D6), monoclonal anti-CD45 (LifeSpan BioSciences, Seattle, WA, USA, 1:1000 clone 3G4), monoclonal anti-CD4 (Abcam. 1:1000 clone EPR19514), monoclonal anti-CD8 (Abcam, 1:1000, clone BLR044F), monoclonal anti-CD3 (Abcan, 1:1000, clone SP7), monoclonal anti-CD68 (Abcam, 1:1000, clone KP1), monoclonal anti-CD79a (Abcam, 1:1000, clone EPR3619), anti-mesothelin biotin-conjugated (LifeSpan BioSciences, Seattle, WA, USA, 1:1000), anti-E-cadherin (Santa Cruz Biotechnologies, CA, USA, 1:1000), anti-α-SMA (kindly provided by Professor G Gabbiani, 1:750), anti-vimentin (Novocastra, 1:1000), anti-PTP4A1 (ThermoFisher Scientific, 1:1000), anti-TIMP1 (Abcam, 1:1000) and anti-GAPDH (Sigma-Aldrich, 1:1000). After rinsing in PBS-T, the membrane was incubated with HRP-conjugated secondary antibodies (diluted 1:10,000 in 1% w/v BSA in PBS-T). Chemiluminescence signal was acquired and quantified using respectively the ChemiDoc and Quantity One software (Bio-Rad, Hercules, CA, USA). Gel electrophoresis was digitized by GS-800 Densitometer (Bio-Rad, Hercules, CA, USA).
Mass spectrometry (MS) analysis
Samples were lysed, reduced and alkylated in 50 ul of iST-LYSE buffer (PreOmics) for 10 min at 95 °C and then digested over night at 37 °C with 0.7 ug Trypsin and 0.3 ugLysC. Digested samples were processed by iST protocol57.
Elution of the digested samples was performed with a 200 cm uPAC C18 column (PharmaFluidics) maintained at 40 °C in the thermostatic column compartment of an Ultimate 3000 RSLC. The peptides were separated with increasing organic solvent at a flow rate of 350 nl/min using a non-linear gradient of 5–45% solution B (80% CAN and 20% H2O, 5% DMSO, 0.1% FA) in 155 min.
MS data were acquired on an Orbitrap Fusion Tribrid mass spectrometer (ThermoScientific). MS1 was performed with Orbitrap detection at a resolving power of 120 K, while MS2 was performed with Ion Trap detection with Rapid Ion Trap Scan Rate. Top speed mode with a 2 s. cycle time was performed for data dependent MS/MS analysis, during which precursors detected within the range of m/z 375 − 1500 were selected for activation in order of abundance. Quadrupole isolation with a 1.6 m/z isolation window was used, and dynamic exclusion was enabled for 30 s. Automatic gain control targets was set at 4 × 105 for MS1 and at 1 × 104 for MS2 with 50 and 45 ms maximum injection times respectively. The signal intensity threshold for MS2 was 1 × 104. HCD was performed using 28% normalized collision energy. One microscan was used for both MS1 and MS2 events.
Raw data were processed with MaxQuant58 software version 1.6.10.0. A false discovery rate (FDR) of 0.01 was set for the identification of proteins, peptides and PSM (peptide-spectrum match). For peptide identification a minimum length of 6 amino acids was required. Andromeda engine, incorporated into MaxQuant software, was used to search MS/MS spectra against Uniprot human database (release UP000005640_9606 April 2019). In the processing the Acetyl (Protein N-Term), Oxidation (M) and Deamidation (NQ) were selected as variable modifications and the fixed modification was Carbamidomethyl (C).
Whole Mass spectrometry data are friendly available at ProteomeXchange Consortium59 via the PRIDE58 partner repository with the dataset identifier PXD024556 (Reviewer account details: Username: reviewer_pxd024556@ebi.ac.uk, Password: 1BfsyHkc).
ELISA assay
To quantify ANXA13 in undiluted serum of an independent group of 20 patients with FSGS and 20 No FSGS patients, commercial ELISA kit was used (Abnova, KA6081). Assay was performed following the manufacturer’s instructions. The Reference standards were run in triplicate and test samples were run in duplicate. Box plot was used to visualize the protein concentration. In box plot each point indicates the mean obtained from duplicate measurement. The lower detection limit was determined as the lowest protein concentration that could be differentiated from blank.
Cell culture
Human peritoneal mesothelial cells (HPMC) were purchased from Creative Bioarray (Shirley, New York, USA). They were grown in DMEM/F12 medium supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml), and maintained at 37 °C in a humidified incubator supplied with 5% CO2. After cells reached 70% confluence, they were cultured in serum-free medium in the presence or absence of 10 ng/ml TGFβ1 (R&D Systems, Minneapolis, MN) to 96 h and then processed for western blots. Six biological replicates have been done.
Statistical analysis
After normalization, whole mass spectrometry data were analyzed by unsupervised hierarchical clustering using multidimensional scaling (MDS) with k-means and Spearman’s correlation, in order to identify outliers and dissimilarity between samples. Then, the normalized whole dataset was used to construct a co-expression network using the weight gene co-expression network analysis (WGCNA) package in R20. A weighted adjacency matrix was constructed using the power function. After choosing the appropriate β parameter of power (with the value of independence scale set to 0.8) the adjacency matrix was transformed into a topological overlap matrix (TOM), which measures the network connectivity of all proteins.
To classify proteins that display co-expression profiles into protein modules, hierarchical clustering analysis was conducted according to the TOM dissimilarity, with a minimum size of 20 proteins per module. To identify the relationship between each module and each clinical trait, we used module eigengenes (MEs) and calculated the Spearman’s correlation between MEs and the clinical traits, namely: dialysis vintage; PET values; D/D0 glucose and D/P creatinine; FSGS and No FSGS patients. A heatmap was then used to visualize each degree of relationship. Same analysis was done for each protein. Proteins were considered in correlation with at least one clinical traits with a significant (two sides p values ≤ 0.05 after Benjamini–Hochberg correction for multiple interactions) Spearman’s correlation coefficient values > 0.7.
To identify the hub proteins of modules that maximize the discrimination between FSGS and No FSGS samples, we applied T-test, machine learning methods such as non-linear support vector machine (SVM) learning, and partial least squares discriminant analysis (PLS-DA). For the T-test, proteins were considered to be significantly differentially expressed between two conditions with power of 80% and an adjusted p value ≤ 0.05 after correction for multiple interactions (Benjamini-Hochberg) and a fold change of ≥ 2. In addition, the proteins needed to show at least 70% identity in the samples in one of two conditions and area under the curve (AUC) in the received operating characteristic (ROC) analysis > 0.8. Volcano plots were used to visualize the expression fold change differences between FSGS and No FSGS samples.
In SVM learning, a fourfold cross-validation approach was applied to estimate the prediction and classification accuracy. Besides, the whole matrix was randomly divided into two parts: one for learning (65%) and the other (35%) to determine the prediction accuracy.
Finally, gene set enrichment analysis60 was done to build a functional proteins network based on their Gene Ontology (GO) annotations extracted from the Gene Ontology Consortium (http://www.geneontology.org/). The protein profile expression data were loaded in the dataset and a ranked list was assigned to each GO annotation/pathway. These ranks take into account the number of proteins associated with each gene signature with respect to all proteins, their mean of fold change and the p value after False Discovery Rate (FDR) correction for multiple interactions. These ranks are confined between − 1 and 1, corresponding to minimal and maximal enrichment in each group. In the two-dimensional scatter plot utilized to visualize this analysis, the points located on the straight line passing through the coordinates (1x, 1y) and (− 1x, − 1y) represent the equally enriched signatures. The distance from this line is proportional to the increase in signature enrichment in one of the two groups (over or under the straight line are the GO annotation/pathway positively enriched in FSGS or No FSGS samples respectively).
For the ELISA assay, Mann–Whitney U-test for unpaired samples was used to assess the difference in the concentration of the potential biomarker of FSGS samples. Results were expressed as medians and interquartile range (IQr). Receiver operating characteristic (ROC) curves were generated to assess the diagnostic efficiency of assay. AUC value was classified as: 0.5, not discriminant; 0.5–0.6, fail; 0.6–0.7, poor; 0.7–0.8, fair; 0.8–0.9, good and 0.9–1, excellent. Youden’s index and Likelihood ratio were used to identify the cutoff and the diagnostic performance of each assay, respectively. Two sides p values ≤ 0.05 were considered as significant. All statistical tests were performed using Origin Lab V9 and the latest version of software package R available at the time of the experiments61.
