Plasma proteomic analysis of systemic lupus erythematosus patients using liquid chromatography/tandem mass spectrometry with label-free quantification

Collection of SLE samples and healthy control samples

A total of 19 SLE patients (female/male; 18/one; mean age 32.1 ranging from 22 to 54 years) and the same age and gender-matched 12 healthy controls (female/male; nine/three; mean age 32.9 ranging from 26 to 54 years of age) gave plasma samples that were obtained from the Cathay General Hospital (CGH), Taipei, Taiwan. The disease assessment of the collected SLE samples was performed according to the classification criteria of the American College of Rheumatology (Tan et al., 1982). The Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) scoring method was used to determine the level of SLE disease activity (Bombardier et al., 1992). Based on the disease activity index parameter results, all of the recruited SLE patients had active disease (SLEDAI more than eight) with high titers of anti-ds DNA antibodies, and the presence of proteinuria and low levels of complement. The demographic characteristics of patients and controls are presented in Table 1. Plasma was prepared from the whole blood samples, which were collected in heparin-containing tubes from both the SLE patients and the healthy controls. All of the patient samples were collected prior to the clinical treatment. The protein concentrations of the collected samples were determined by Bradford protein assay (Zor & Selinger, 1996) and the samples were stored at −80 °C until further analysis. This study obtained approval from the Institutional Review Board (IRB) (Approval code CT-09905) for Research Ethics at the Cathay General Hospital (CGH), Taipei, Taiwan. Informed consent was provided by all blood donors.

Depletion of high abundance serum albumin protein

In order to identify the various low and high abundance proteins by mass-based proteomic analysis, highly abundant serum albumin was depleted from the SLE and control samples by using a dye-based proteo-prep blue albumin removal kit (Thermo Fisher Scientific, Waltham, MA, USA). Using the manufacturer’s protocol, spin columns were suspended with slurry and centrifuged at 8,000 × g for 10 s. Next, 0.1 ml of serum samples were added to the spin columns and the columns were incubated for 10 min. Subsequently, the samples were centrifuged at 8,000 × g for 60 s; this was repeated twice to remove further serum albumin (Ahmed et al., 2003). The protein concentrations of the depleted serum samples were determined by Bradford assay and the samples were stored at −80 °C until further analysis.

Protein precipitation and in-solution digestion

The depleted SLE and healthy plasma samples were precipitated using 100% ice-cold acetone and then kept overnight at −20 °C. Then the samples were centrifuged at 14,000 × g for 10 min, and the pellets were dissolved in 100 μl of 25 mM NH₄HCO₃ and 6.5 M urea (0.1–1 μg/μl); this was followed by an in-solution digestion procedure used in previous studies (Ru et al., 2006). Next, the protein samples were reduced with 100 mM DTT at 37 °C for 30–40 min and alkylated with 200 mM IAA in the dark at room temperature for 25–35 min. After this, the protein samples were digested with sequencing grade trypsin (Promega, Madison, WI, USA) using a 50:1 ratio at 37 °C. The reaction was quenched by adding 2 μl of 50% formic acid (FA), the mixture vortexed and then centrifuged in order to collect the peptides. Finally, the solution was lyophilized and desalted using a C18 zip-tip procedure.

Nano UPLC and mass spectrometry conditions

ESI interface Q-TOF MS/MS was carried out at a resolution of about 10,000 full-width half maximum performance. In order to calibrate the instrument, an external standard of lock mass BSA was continuously infused at a constant flow rate of 0.25 μl/min using the Nano-ACQUITY auxiliary pump at an interval of 20 s (lock spray frequency). The precursor mass error was less than 2 ppm and for accuracy, the lock mass data were averaged. The acquired peptide spectra were eluted using the positive V mode with a scan mass range of 50–2,000 m/z and a scan time of 1 s. After reconstituting in 3% ACN and 0.1% FA, the digested 400 ng peptides were injected into an online nano-ACQUITY, UPLC coupled Q-TOF, Synapt-HDMS mass spectrometer (Waters Corporation, Milford, MA, USA). The peptides were separated using a C₁₈ reverse phase column (1.7 μm × 75 μm × 250 mm) (Waters Corporation, Milford, MA, USA). The binary solvent system used consisted of 99.9% water and 0.1% FA (mobile phase A) and 99.9% ACN and 0.1% FA (mobile phase B). The peptides were initially pre-concentrated and desalted online at a flow rate of 5 μl/min using a 5 μm symmetry C₁₈ trapping column (internal diameter 180 mm, length 20 mm) (Waters Corporation, Milford, MA, USA) with 0.1% FA. After injection, the peptides were eluted into the Nano-LockSpray ion source at a flow rate of 300 n/l and a gradient of 2–40% for 120 min. Then the column was washed and equilibrated. The digested plasma samples were run in triplicate and the data were analyzed by ProteinLynx Global Server 4.2 software (PLGS; Waters Corporation, Milford, MA, USA) (Aggarwal et al., 2017). Each sample was injected three times to obtain technical triplicates.

Label-free quantification

To quantify the proteins from the LC-MS/MS, a label-free quantification analysis was performed using PEAKS Studio 8.0 (Bioinformatics Solutions Inc., Waterloo, ON, USA) (Zhang et al., 2012). Independent samples from each triplicate analysis were studied and compared between patients and the controls. Total raw data files were imported and processed using the Peaks software program for the interpretation of spectra and the retention time was set from 600 to 10,500 s. An in-house constructed Uniprot’s reference database of Homo sapiens (release 03_2014) contained 20,272 entries was added and combined with a decoy database (the sequences were reversed). For label-free quantification the following parameters were specified: enzymatic digestion by trypsin, with two missed cleavages; precursor mass tolerance was 10 ppm; fragment mass tolerance: 0.7 Da, minimum charge: 2, maximum charge: 3. The specified fixed and variable modification consisted of carbamidomethylation (Cys), oxidation (M), and deamidated (N and Q). To determine the false-positive identification rate, the estimated spectra was used against decoy database. A false discovery rate (FDR) of ≤1%, with a peptide score of −10 log p ≥ 20 was considered adequate for confident protein identification. To determine the relative protein and peptide abundance in the tested samples, peptide feature based quantification was performed. The signal intensity of a peptide is directly proportional to the abundance of the peptide in the sample, therefore, the confidently identified peptide features were matched and the peptide intensity differences between two samples were able to be estimated. Likewise, the area under the curve of the extracted ion chromatograms (XICs) was measured and compared between two analyzed runs. To get the summed cumulative peak area of the protein, only unique peptides that are assigned to particular proteins were selected.

The FDR was calculated based on the target/decoy database, and the peptides with an FDR of ≤1% were chosen as true positive hits (considering the risk of having one false positive in 20 observations). By using this active feature based quantitative approach the detected peptides with p-values <0.05 and 0.01 which were identified in at least three observations from the SLE samples compared to control samples were considered. In order to identify the significant protein differential expressions an independent sample T-test was performed. The quantified datasets were normalized using their spectral abundance factor values (the average of the triplicate experiments) and this was used to generate a heat map showing the differentially expressed proteins between the two groups.

Quantification of altered proteins by emPAI

To acquire confident results from our proteomic study, the widely used exponentially modified protein abundance index (emPAI) was employed as an alternative method in order to quantify the differences in abundances of the identified proteins between the SLE patients and the control subjects. Numerous label-free relative quantification studies have successfully applied the formula derived for emPAI and developed by Ishihama et al. (2005). Specifically, the MASCOT database results were generated in an Excel file (composed of emPAI, protein description, peptide identifications, the charge states of the peptides, etc.), and this was uploaded to the freely accessible (http://empai.iab.keio.ac.jp) web-based system (http://empai.iab.keio.ac.jp) for emPAI evaluation. The following parameters were specified on the emPAI website: IPI_HUMAN database, trypsin digestion as the enzyme, carbamidomethyl (Cys) fixed modifications, mass range 500–3,000 Da, and no retention time filtering. The uploaded sample files were analyzed and the results were exported to a separate file. To estimate the protein abundances, the mean emPAI values of the triplicate MS analyses from patients and controls were calculated and Hochberg and Benjamini calculation of FDR (p < 0.05) was applied to correct for multiple testing errors. Protein level fold-difference was calculated from the relative means of normalized emPAI values and the fold change differences were estimated from all replicates, with the Student’s t-test (two-tailed, heteroscedastic) applied to the same values. A minimum of three unique peptide sequences in at least three replicates were required for quantification.

Protein identification

For the identification of proteins, the UniProtKB database (UniProt release 2015-10) and National Center for Biotechnology non-redundant (NCBInr) was used for the database search using the Mascot software (Matrix Science version 2.2, http://www.matrixscience.com) search engine. The following parameters were specified for searching altered proteins: enzymatic digestion by trypsin with two missed cleavages, followed by carbamidomethyl as fixed and oxidation (M) as variable modifications. The peptide mass tolerance was set at 50 ppm and the MS/MS tolerance was set at 0.1 Da with an FDR of ≤ 1%. Proteins identified in all three technical replicates or in at least two of the three analyses were considered to be identified and their theoretical molecular mass (MW) and isoelectric point (pI) were determined using Mascot database.

Bioinformatics analysis

To interpret the biological processes (BPs), molecular functions (MFs), and cellular components (CCs) of the identified proteins, the international standardized gene function classification system of gene ontology (GO) (http://www.geneontology.org/), and the DAVID (http://david.abcc.ncifcrf.gov/) (Database Annotation Visualization, and Integrated Discovery) database with for functional analysis were used (Huang, Sherman & Lempicki, 2009). For the protein–protein interaction (PPI) networks among these proteins, the STRING (Search Tool for the Retrieval of Interacting Genes/Proteins, Version 9.1) database at the website: http://string-db.org/ with default parameters were used.

Statistical analysis

The LC-ESI-MS/MS data was measured using peptide feature based and spectral counting based quantification in order to identify the differences in protein expression levels. Data are expressed as the mean ± standard deviation and differences were determined using two independent sample t-test and a paired t-test were used to determine the differentially regulated proteins between SLE and healthy controls. Statistical analysis were performed using the SPSS statistical package (SPSS16; SPSS Ltd., Working, and Surrey, UK) for Windows. A probability value of less than 0.05 with 95% confidence limit was considered as statistically significant and of p-values <0.01 were considered as highly significant.

Protein profile of the differentially expressed proteins

Based on the bioinformatics analysis, the identified differentially expressed proteins were found to be immunoglobulins, acute phase reactants, glycoproteins, transporters, antiproteases, and binding proteins. Mounting evidence reveals that most of the identified proteins play crucial roles in immune system regulation, inflammation, and acute inflammatory responses. SLE patients were found to have a 2.8- to more than 3.5-fold increase in expression of several immunoglobulin heavy and light chains (p < 0.01), a 2.6- to 3.5-fold increase in expression of acute phase proteins (APPs) (alpha-2 macroglobulin (A2M), serotransferrin (TF/TRFE), ceruloplasmin (CP), clusterin (CLU), serum albumin (ALB), and transthyretin (TTR)) (p < 0.01), a 2.7- to 4.0-fold increase in expression of various glycoproteins, including alpha-1-acid glycoprotein (ORM1/A1AG1), alpha-1-acid glycoprotein 2 (ORM2/A1AG2), and alpha-1-B glycoprotein (A1BG) (p < 0.04), a 3.2-fold increase in expression of alpha-1-antichymotrypsin (A1ACT/SERPINA3) (p < 0.01), alpha-1-antitrypsin (A1AT/SERPINA1) and a 1.4- to seven-fold increase in expression of hemoglobin subunit alpha-1 (HBA1), hemoglobin subunit beta (HBB), and haptoglobin (HP/HPT) (p < 0.01). By way of contrast, the expression of clusterin (CLU), apolipoprotein A-1 (APOA1), apolipoprotein A-2 (APOA2), and transthyretin (TTHY/TTR) were significantly decreased in SLE patients by 0.2- to 0.4-fold (p < 0.01).

Functional annotation of the identified proteins

To gain valuable insight of the identified differentially expressed proteins from our analysis the international standardized gene function classification system of GO and the DAVID database were employed. When analyzed for CC analysis by GO annotations, most of the identified differentially expressed proteins were found in the extracellular space (46%), while others are macromolecular in nature (27%) and yet others were found to be part of membrane complexes (27%) (Fig. 3A). As shown in Fig. 3B, the BP evaluations of the GO annotations revealed that the identified proteins are largely involved in biological regulation (15.5%), complement activation (9.5%), negative regulation of the immune response (12.1%), multicellular organismal processes (14%), immune system processes (8.6%), metabolic processes (12.1%), the acute inflammatory response (10.3%), and localization (14.7%). On the other hand, when analyzed in terms of MF, the identified proteins were mostly involved in binding (41.2%), catalytic activity (38.2%), receptor activity (14.7%), transporter activity (14.7%), and serine peptidase activity (7.2%) (Fig. 3C; Tables S3 and S4). Therefore, based on the functional analysis, the identified proteins seem to primarily play important roles in immune system regulation (p = 0.001), in complement system activation (p = 0.01), in the innate and adaptive immune system responses (p = 0.003) and, potentially, in the acute inflammatory response (p = 0.002).

According to KEGG pathway annotation, the identified protein cohort was classified into more than 25 pathways (Tables S5 and S6) with the top 18 of them being shown in Fig. 4. Among 19 identified differentially expressed proteins, nine proteins (A2M, A1AT, A1ACT, CLU, KV105, KV315, IGG4, SERPING1, TF) were involved in complement and coagulation pathway, 12 proteins (A1AT, A1BG, HBB, TRFE, APOA1, AACT KV315, CLUS, KV105, A2MG, APOB, A1AG2) were involved in homeostasis and immune system processes (p = 8.8 × 10⁻⁸). In addition to the above findings, the majority of them were involved in a number of potentially interesting pathways such as scavenging heme from plasma, platelet activation, acute-phase responses, degranulation, signaling, aggregation, and the creation of C4 and C2 activators, etc. These pathways are highly relevant to the present study since SLE is an autoimmune disease caused by abnormalities that affect the immune system and in turn cause an acute inflammatory response. Therefore, the identified proteins are likely to play a significant role in SLE disease development and/or pathogenesis.

Additionally, the STRING PPI network analysis demonstrated a tight PPI network when the medium PPI confidence score of 0.4 was applied in STRING database, whereas at the high confidence score of 0.7 a significant association of PPT networks was identified among the differentially expressed proteins as shown in Fig. 5. This PPI network suggests that the identified proteins have strong interactions with the proteins that contribute to numerous BP and have a wide range of MFs.