MS-grade acetonitrile (ACN) was purchased from Thermo Fisher Scientific, United States, and ultrapure water was filtered through the Milli-Q system (Millipore, Billerica, MA). Formic acid (FA) was obtained from DIMKA, and ammonium bicarbonate was purchased from Fluka (Honeywell Fluka, United States). Sodium dodecyl sulfate (SDS), SDS-free protein lysate, dithiothreitol (DTT), iodoacetamide (IAM), Coomassie brilliant blue G-250, trypsin and acetone were obtained from Aladdin, China.

This investigation was carried out according to the World Medical Association Declaration of Helsinki and approved by the ethics committee of Maoming People’s Hospital. All patients had completed written informed consents before cardiac surgery, and six patients were enrolled in the urinary proteomics study after the diagnosis of AKI which was defined in accordance with the KDIGO Clinical Practice Guidelines based on sCr criteria: the increase in sCr ≥ 50% within 7 days, or the increase in sCr ≥ 0.3 mg/dl (≥ 26.5 μmol/L) within 48 h, or oliguria (Khwaja, 2012). Patients were excluded if they were age under 18 years, had preexisting chronic renal disease or undergoing renal replacement therapy, a history of renal transplantation or nephrectomy. In order to explore the effect of CSA-AKI on kidney, spot urine samples were prospectively collected at two time points, preoperative (Before_AKI) and postoperative-the first day of diagnosed CSA-AKI (AKI-Day1), which was used to compare the changes in urinary proteomic profiles. Serum creatinine was measured before cardiac surgery, at least twice a day for the first 3 days after cardiac surgery, and then daily thereafter.

Approximate 5 ml of urine sample from Biological Resource Center of Maoming People’s Hospital, was transferred into a 50 ml centrifuge tube, and 25 ml (5 times the volume of urine sample) of precooled acetone was added into the tube. After placing at -20°C overnight, the urine samples were centrifuged at 25,000 g and 4°C for 15 min, and the supernatant of each sample were discarded. Then, an appropriate amount of SDS-free protein lysate was added into the drying urine sample, and an automatic grinder was used to promote protein dissolution. The supernatants were collected after centrifuging at 25,000 g at 4°C for another 15 min. Final concentration (10 mM) of DTT was added into the supernatant and bathed in water at 37°C. After 30 min water bath, final concentration (55 mM) of IAM was added into the reduced samples and kept for another 45 min in the dark at room temperature. Protein solution could be obtained from the supernatant of alkylated samples by centrifuging at 25,000 g at 4°C for 15 min. At last, the Bradford assay kit (Bio-Rad, Hercules, CA, United States) was used to determine the concentration of the protein solution.

As for protein enzymatic digestion, 100 μg of protein solution per sample was transferred and diluted with 4 times volumes of 50 mM NH₄HCO₃ solution. According to the ratio of 40: 1 (protein to enzyme), 2.5 μg of trypsin was added into the diluted protein solution for enzymatic digestion at 37°C. After 4 h, enzymatic peptides were desalted using a Strata X column and vacuumed to dryness.

In Shimadzu LC-20AB high performance liquid chromatography (HPLC) system, the Gemini C18 00G-4435-E0 column (5 μm, 4.6 mm internal diameter×250 mm length) was used for peptide separation with mobile phase A (5% ACN, pH 9.8) and mobile phase B (95% ACN). Equal amount of peptides were mixed from all urinary proteome samples, which was further diluted with mobile phase A and eluted at a flow rate of 1 ml/min by the following gradient: 0–10 min 95% A, 5% B; 10–50 min, 95%–65% A, 5%–35% B; 50–51 min 65%–5% A, 35%–95% B; 51–54 min, 5%–0% A, 95%–100% B; finally 95% A and 5% B eluting 10 min to re-equilibrate the column. The elution peaks were monitored at 214 nm and the fractions were collected every minute. At last, the peptide components were combined into a total of 10 fractions for lyophilization and Data Dependent Acquisition (DDA) analysis.

The morphology of Gemini C18 silica microspheres was characterized by a JSM-6700F scanning electron microscope (SEM, JEOL, Japan) at 3.0 kV, and element analysis was performed on an energy dispersive spectrometer (EDS) of the SEM. The particle size distribution was measured by ImageJ software through SEM image, and the zeta potential of the particles (dispersed in ethanol) was measured by a nanosize and zeta potential analyzer (Malvern, United Kingdom). The water contact angles (CAs) on the coating composed of the silica microspheres were measured on a CA system (OCA20, Dataphysics, Germany) at ambient temperature, and the average CA value was obtained by measuring three different positions of the particles.

An Orbitrap Exploris 480 (Thermo Fisher Scientific, San Jose, CA) system equipped was used to analyze urinary proteome for peptide sequencing and protein quantification. A Thermo Ultimate 3000 UHPLC liquid chromatography with the trap column and a tandem self-packed C18 column (150 μm internal diameter, 1.8 μm of silica microspheres diameter, 35 cm column length), were coupled online to the mass spectrometer through a nanoESI ion source.

DDA Library Construction: The segmented dried peptide samples were reconstituted with mobile phase A (2% ACN, 0.1% FA aqueous solution), and the supernatants were collected for injection after centrifuging at 20,000 g for 10 min. Enriched peptides were separated at a flow rate of 500 nL/min by the following effective gradient for DDA mode detection: 0–5 min, 5% mobile phase B (98% ACN, 0.1% FA); 5–120 min, 5%–25% B; 120–160 min, 25%–35% B; 160–170 min, 35%–80% B; 170–175 min, 80% B; 175–175.5 min, 80%–5% B; 175.5–180 min, 5% B. The main parameters of Orbitrap MS were set as follows: ion source voltage 1.9kV; MS1 scanning range m/z 350∼1,650; MS1 resolution 120,000; maximal injection time (MIT) 90 ms; MS/MS collision type HCD; collision energy NCE 30; MS/MS resolution 30,000; auto mode for MIT; dynamic exclusion duration 120 s. The start m/z for MS/MS was fixed to auto mode, and AGC was set as MS 300% and MS/MS 100%. Precursor peptide ions for MS/MS scan were satisfied: charge range from 2 + to 6+, top 30 precursors with intensity over 2E4.

Data Independent Acquisition (DIA) Protein Quantification: The dried enzymatic urinary proteome samples were reconstituted with mobile phase A (2% ACN, 0.1% FA aqueous solution) and centrifuged at 20,000 g for 10 min. The supernatants were injected into the nanoLC-MS system and separated at a flow rate of 500 nL/min by the following gradient program: 0–5 min, 5% mobile phase B (98% ACN, 0.1% FA); 5–90 min, 5%–25% B; 90–100 min, 25%–35% B; 100–108 min, 35%–80% B; 108–113 min, 80% B; 113–113.5 min, 80%–5% B; 113.5–120 min, 5% B. The main parameters of MS in DIA mode were set as follows: ion source voltage 1.9 kV, MS1 scanning range m/z 400∼1,250; MS1 resolution 120,000; MIT 90 ms. For MS/MS scan, the scanning range m/z 400∼1,250 was equally divided to 50 continuous windows, and fragment ions were scanned in Orbitrap with resolution of 30,000. MS/MS collision type was also selected as HCD, collision energy NCE was set to 30, MIT was auto mode, and AGCs were separately set as MS 300% and MS/MS 1000%.

In the DDA library construction, the software MaxQuant (version 1.5.3.30) was used for peptide identification, and the retrieved database was uniprot_homosapiens_irt.fasta (20,303 sequences). Trypsin was used for proteins cutting and up to two missed cuts were allowed. Methyl carbamate was used for the fixed modification at cysteine (C) site, while oxidized of methionine (M), acetyl groups (N-terminal of protein), glutamine (Q) to pyro-glutamate (N-terminal of Q) and deamidated (NQ) were used for variable modification. The error rate for peptide mapping matching with protein was set as 1%, and the shortest peptide length was seven amino acids. The DIA MS data were analyzed by the software Spectronaut, and the retention time (RT) was corrected by iRT peptides. Based on the target-decoy model suitable for sequential window acquisition of theoretical fragment ion mass spectrometry (SWATH-MS), the false positive control was set as 1%. Finally, the significant differences of urinary proteome were statistically evaluated by the MSstats software package. After error correction and normalization were performed on each sample, the differential expressed proteins (DEPs) were screened out according to the absolute value of fold change (|FC|) ≥ 2 and p value < 0.05 as the judgment criteria for significant differences.

Gene ontology (GO) annotation of urinary proteome was analyzed using the online Database for Annotation, Visualization and Integrated Discovery (DAVID) software. The software Cluster 3.0 was used for hierarchical clustering and visualization on the urinary DEPs. Kyoto Encyclopedia of Genes and Genomes (KEGG) is a database collection referring to genomes, diseases, biological pathways, drugs and chemical materials, which was used for pathway analysis of the identified urinary proteins using the online software. The possible functions of urinary proteins were predicted by Eukaryoyic Orthologous Groups (KOG) software, and the online WoLF PSORT software was used to predict the possible subcellular locations of proteins. Finally, Search Tool for the Retrieval of Interacting Genes (STRING, 11.0) software was used to analyze the protein-protein interaction (PPI) network, which is important to understand cell physiology under normal and disease conditions.

The basic information of six patients with CSA-AKI, such as age, body mass index (BMI), sCr levels and urinary protein amounts, is displayed in Supplementary Table S1. The urine samples were collected from patients with CSA-AKI and divided into two groups: Before_AKI (uninjured kidney) and AKI_Day1 (injured kidney). The sCr levels and urinary protein amounts of the AKI_Day1 group were significantly increased compared with those of the Before_AKI group. According to the 2012 KDIGO clinical practice guideline, AKI can be divided into three stages based on sCr elevation: stage 1A as an absolute increase in sCr of 0.3 mg/dl (or 26.5 μmol/L) within 48 h, stage 1B as a 50% relative increase in sCr within 7 days, stage 2 as a 100% relative increase in sCr over a 7-days window of observation, and stage 3 as a 200% relative sCr increase within 7 days. In total, the severity of six AKI patients separately belonged to three patients in stage 1A, two patients in stage 1B, and one patient in stage 2.

The high-magnification SEM image in Figure 1A showed the uniform particle size distribution of the silica microspheres. EDS mapping of the particles showed that carbon (C), oxygen (O) and silicon (Si) were only detected in the exterior region, suggesting the hydrophobicity of the surface of silica microspheres due to the grafted layer of alkane chains. The silica particles exhibited uniform size distribution with an average diameter of 5.4 ± 0.70 µm (mean ± SD, Figure 1B). The average zeta potential of these silica particles in ethanol was 13.70 ± 0.97 mV (three test values: 13.83, 12.73 and 14.53 mV), indicating that these particles could easily aggregate. Furthermore, the contact angle (CA, θ) of the three phases of the silica microspheres was measured to study their surface wettability (Figure 1C). The value of 136.1° ± 0.6° (mean ± SD, n = 3) was obtained, further proving the hydrophobic properties of the silica microsphere coating.

DDA MS was used to identify and construct the peptide library of urinary proteomes from the Gemini column enriched samples (Supplementary Table S2). A total of 10,869 peptides and 1,942 proteins were identified from the preoperative and postoperative urine samples using DIA MS as listed in Supplementary Table S3 and Supplementary Table S4, respectively. The cysteine sites in the samples can be reduced and alkylated through the classical carbamidomethylated reaction with DTT and IAM prior to tandem MS analysis. A total of 1,781 peptide segments contained cysteines with carbamidomethylation. With the development of tandem MS, the number of posttranslational modifications (PTMs) identified on proteins has increased rapidly, providing valuable information on the signal pathways and cellular processes regulated by PTMs. In the study, the PTMs on the peptides included 961 oxidation of M, 105 acetylation of N-terminal, 218 of pyroglutamalytion on Q and 856 deamidation. The annotated data of the urinary proteomes of patients with AKI are listed in Supplementary Table S5.

The basic statistical results of the urinary proteins are shown in Supplementary Figure S1 and comprise unique peptide distribution, protein mass distribution and protein coverage distribution. According to the unique peptide distribution in Supplementary Figure S1A, 1,004 proteins had the number of unique peptides ≥ 3, accounting for 51.70% of all identified urinary proteins. Bar charts in Supplementary Figure S1B showed that the molecular masses of urinary proteins were mostly distributed in the range of 10–60 KDa. Therefore, most urinary proteins were medium proteins. For protein coverage distribution, the length of the identified sequence was divided by the total length of the protein sequence. Supplementary Figure S1C shows that the identification rate was mainly concentrated within the range of 40%.

The urinary proteome differences between groups were compared using univariate analysis to investigate the significant DEPs before and after AKI. The results were visualised through the volcano plot in Figure 2A. Changes in urinary proteins were determined by the combination of FC and p value; FC ≥ 2 and p value < 0.05 were expressed as up-regulation, and FC ≤ −2 and p value < 0.05 were expressed as down-regulation. Compared with the Before-AKI group, 324 proteins were differentially expressed in the AKI-Day1 group, with 96 up-regulated DEPs and 228 down-regulated DEPs. Furthermore, Figure 2B shows the cluster analysis chart of DEPs to visually reflect the expressed differences between the two groups.

GO is an international standard gene function classification system with three categories: biological process (BP), cellular component (CC) and molecular function (MF). It can provide a timely updated standard vocabulary to comprehensively describe the characteristics of genes and gene products in organisms. Blast2GO software was used for GO annotation analysis to evaluate the functional significance of all identified proteins. As shown in Supplementary Figure S2A, the most enriched BP (out of 29 GO terms) were “cellular process”, “biological regulation”, “regulation of biological process”, “metabolic process” and “response to stimulus”. Meanwhile, the most enriched CC (out of 19 GO terms) were “cell”, “cell part”, “organelle”, “extracellular region” and “extracellular region part”. The most enriched MF (out of 13 GO items) were “binding”, “catalytic activity”, “molecular function regulator”, “molecular transducer activity” and “signal transducer activity”.

GO enrichment analysis was also carried out for the identified DEPs (Supplementary Table S6). Supplementary Figure S2B shows the GO functional classification maps of all DEPs, and Figure 3 displays the GO classification of up-regulated and down-regulated differential proteins. The up-regulated and down-regulated DEPs can interfere with common structural or functional processes. However, unique enrichments were detected for some GO terms, including the up-regulated proteins (e.g., detoxification, presynaptic processes involved in chemical synaptic transmission, protein tag and transcription regulator activity) and down-regulated proteins (e.g., cell aggregation, pigmentation, nucleoid, virion and virion part).

The biological functions of the identified DEPs were further characterised via KEGG enrichment analysis (Supplementary Table S7). The results showed that the up-regulated proteins were annotated as 29 major pathways and the most enriched pathways were “protein processing in endoplasmic reticulum” and “pathway in cancer”. Meanwhile, the down-regulated proteins were annotated to 27 major pathways, and the most enriched ones were “cell adhesion molecules”, “pathways in cancer” and “phagosome” (Figure 4A). The 19 highest ranked biological functions for the DEPs set are shown in Figure 4B. Among these pathways, “cell adhesion molecules” was enriched in ICOS ligand, poliovirus receptor, platelet endothelial cell adhesion molecule, cadherin-2, tumour necrosis factor receptor superfamily member 5, receptor-type tyrosine-protein phosphatase mu, cadherin-5, contactin-1, mucosal addressin cell adhesion molecule 1, CD276 antigen, vasorin, neuronal growth regulator 1, neogenin, golgi apparatus protein 1, junctional adhesion molecule C, cell adhesion molecule one and neurexin-3 (Supplementary Table S8).

The potential functions of the identified DEPs were predicted using KOG, a database for the classification of protein orthologs (Supplementary Figure S3). The most representative KOG category was ‘cell processes and signalling’, which indicated that the DEPs were closely associated with signal transduction mechanisms, posttranslational modification, protein turnover and chaperones. Meanwhile, WoLF PSORT software was used to predict the subcellular localisation of the identified DEPs (Supplementary Figure S4). The most represented structures were located in extracellular, plasma membrane compartments, intracellular and mitochondria.

Proteins often carry out a specific function after combining with a complex through PPI. For the top 100 confidence intervals, the STRING database was used for network interaction analysis of the DEPs to construct protein–protein relationships (Figure 5 and Supplementary Table S9). The hub eight central nodes were identified as vitamin K-dependent protein S (PROS), α-1-antitrypsin (A1AT), aggrecan core protein (PGCA), β-2-microglobulin (B2MG), AGT, matrix metalloproteinase-9 (MMP-9), growth arrest-specific protein 6 (GAS6) and urokinase-type plasminogen activator (PLAU). Among these central nodes, A1AT, B2MG, AGT and PROS belonged to the up-regulated DEPs, and the down-regulated DEPs included GAS6, MMP-9, PLAU and PGCA.