Comparative sera proteomics analysis of differentially expressed proteins in oral squamous cell carcinoma

Samples collection

The serum and formalin-fixed paraffin-embedded (FFPE) tissue samples of OSCC were obtained from the Malaysian Oral Cancer Database and Tissue Bank System (MOCDTBS) coordinated by the Oral Cancer Research & Coordinating Centre, University of Malaya (OCRCC-UM) {Medical Ethics Approval Number: DFOP1504/0084(L)} following approval by the Ethics Committee of the Faculty of Dentistry, University of Malaya {Medical Ethics Approval Number: DFOC1803/0026(P)}. These samples were acquired from patients with a histologically confirmed diagnosis of OSCC. No patients received radio- or chemotherapy prior to surgery. OPMD samples were also included in this study as OPMDs have been associated with a risk of malignant transformation to OSCC (Warnakulasuriya, Johnson & Waal, 2007). For the control group, serum samples were obtained from healthy individuals, while FFPE tissue samples were obtained from subjects who are visiting the clinic for removal of impacted wisdom teeth with informed consent. These control subjects did not have a previous history of OSCC or other cancers.

Serum samples consisted of early OSCC (n = 20), advanced OSCC (n = 20), OPMD (n = 10), and normal control (n = 10) were used for the proteomics analysis. For validation purposes, antibody-based methods were conducted in a larger cohort. A total of 120 serum samples consisted of OSCC patients from early stage (n = 34) and advanced stage (n = 39), OPMD patients (n = 12), and healthy individuals (n = 35) were used for ELISA analysis. Out of the 120 serum samples, 60 samples are from the proteomics study set. Briefly, the blood samples (3–4 ml) were collected into BD Vacutainer Plus plastic serum tubes that containing increased silica act clot activator and silicone-coated interior (Becton Dickinson, Heidelberg, Germany) by trained phlebotomists. The blood samples were collected and processed in accordance with the established protocol for sample collection as outlined by the International Society for Biological and Environmental Repositories (ISBER) Best Practices for Repositories, Collection, Storage and Retrieval of Human Biological Materials for Research (Klont et al., 2019; Vaught & Henderson, 2011). Upon processing of the blood samples, the serum samples were aliquoted into 300 µl portions and stored at −80 °C until further use. Whereas 70 FFPE tissue samples consisted of early stage OSCC (n = 18), advanced stage OSCC (n = 21), OPMD (n = 11), and normal control (n = 10) were included for IHC analysis. The serum and FFPE tissue samples are independent of each other. All analysis of these samples was performed individually, and detailed clinical information of the samples is shown in Tables S1 and S2.

Two-dimensional gel electrophoresis (2-DE) and Mass Spectrometry (MS) analyses

The 2-DE and MS analyses were performed as previously described using three µl of unfractionated serum samples with an estimated protein concentration of 150 µg (Chen et al., 2014). The first-dimensional of the 2-DE analysis was carried out on 11 cm, pH 4–7 L Immobiline DryStrip gels (GE Healthcare Biosciences, Uppsala, Sweden). This was followed by the second-dimensional separation with 12.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel using the SE 600 Ruby system (GE Healthcare Biosciences, Uppsala, Sweden). The 2-DE gels were then visualized by silver staining and the images of 2-DE gels were scanned with ImageQuant LAS 500 (GE Healthcare Biosciences, Uppsala, Sweden). After that, the gel images were analyzed using Progenesis SameSpots v4.0 software (Totallab, Newcastle, UK). The protein spots of interest were manually excised from the 2-DE gels and proceeded to in-gel trypsin digestion. The tryptic peptides were then reconstituted in 0.1% formic acid and desalted using ZipTip C18 pipette tips (Millipore, Massachusetts, USA). Subsequently, the eluted peptides were mixed with an equal volume of α-cyano-4-hydroxy-cinnamic acid (CHCA) matrix solution (six mg/ml) and spotted onto a MALDI target plate. After air dried, the peptide mixtures were analyzed with a 4,800 Plus MALDI-TOF/TOF Analyzer (AB Sciex, Foster City, CA, USA). Calibration of the instrument was performed using trypsin-digested beta-galactosidase (Mass Standards kit, AB Sciex, Foster City, CA. USA). In-house prepared silver-stained bovine serum albumin (five µg/ml) gel plugs was used as internal control. The MS results were acquired automatically with a trypsin auto-digest exclusion list and 20 most intense precursor ions were selected for tandem mass spectrometry (MS/MS) analysis with a minimum signal-to-noise (S/N) of at least 10. The protein identification was conducted using the MASCOT search engine (Matrix Science, London, UK) against all human proteins and sequence information in the UniProtKB database (last updated: December 2016). Search parameters for the MASCOT peptide mass fingerprint were set as a fixed modification on carbamidomethylation of cysteines, variable modification of methionine oxidation, up to one missed tryptic cleavage is allowed, MS precursor ion mass tolerance at 100 ppm, and MS/MS fragment ion mass tolerance of ±0.2 Da. A MASCOT score greater than 50 were considered significant (p < 0.05).

Functional classification, annotation and pathway analyses

The identified proteins were further determined using Protein ANnotation Through Evolutionary Relationships (PANTHER) v16.0 (http://www.pantherdb.org/) and Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8 (https://david.ncifcrf.gov/home.jsp) for functional classification and annotation analysis. Both PANTHER and DAVID are bioinformatics tools that enable functional protein grouping and annotation using gene ontology terms to understand the biological importance associated with a list of genes or proteins of interest. The proteins were also subjected to pathway analysis using Ingenuity Pathway Analysis (IPA) v7.1 (Qiagen Ingenuity System, California, USA) software to explore the pathway annotations and interaction networks. The interaction networks and predominant canonical pathways of the proteins were generated algorithmically using the Ingenuity pathways knowledge base.

Enzyme-linked Immunosorbent Assay (ELISA)

A quantitative sandwich ELISA technique was used to determine the serum concentrations of clusterin (CLU) and haptoglobin (HP) in this study. The serum samples were analyzed in duplicate using commercially available CLU and HP ELISA kits (Cusabio Biotech, Wuhan, China) according to the manufacturer’s protocol. Once the chromogenic reaction was stopped with stop solution, the optical density (OD) value was measured at absorbance 450 nm. After that, the concentration of CLU and HP in the serum samples was determined by assessing the OD value of the samples against the respective standard curve.

Immunohistochemistry (IHC)

For immunohistochemical analysis, the FFPE tissue sections were dewaxed with xylene and rehydrated with gradient ethanol prior to heat-induced antigen retrieval. The peroxidase-blocking solution (Dako, Agilent, Santa Clara, CA, USA) and background sniper (Biocare Medical, Concord, CA, USA) were employed to quench the endogenous peroxidase activity and non-specific background, respectively. Subsequently, the sections were incubated with the diluted primary antibodies: CLU (1:400 dilution, ab92458, Abcam, Cambridge, MA, USA), and HP (1:160 dilution, ab23100, Abcam, Cambridge, MA, USA) for an hour at room temperature. Secondary antibody and 3,3′-diaminobenzidine solution (DAB) chromogen substrate were then applied according to the manufacturer’s protocol (Dako Real EnVision Detection System and Peroxidase/DAB+, Agilent, Santa Clara, CA, USA). After counterstained with hematoxylin, the sections were dehydrated and mounted.

Statistical analysis

Statistical analysis was performed using SPSS 20.0 statistical software package (SPSS Inc, Chicago, IL, USA). The data of ELISA are exhibited as mean ± standard error mean (SEM). Whereas IHC data are presented as immunoreactive scores (IRS). The immunoreactive scores (IRS) were calculated based on the intensity scores (negative = 0, mild = 1, moderate = 2, and strong = 3) and the percentage of immunopositive staining (0 = ≤10%, 1 = 11–25%, 2 = 26–50%, 3 = 51–75%, and 4 = >75%), whereby the scores were determined by multiplying the intensity and percentage of immunopositive staining scores. Receiver operator characteristic (ROC) curve and logistic regression analyses were used to evaluate the ELISA and IHC data. Statistical comparisons with p-values less than 0.05 were considered statistically significant.

Identification of differentially expressed proteins by 2-DE and MS analysis

High-resolution serum protein profiles of control, OPMD, early OSCC, and advanced OSCC were obtained from the 2-DE separation and silver staining of the unfractionated serum samples. Thirty-two protein spots that corresponded to 20 differentially expressed proteins showed a statistically significant difference with fold change cut-off of 1.2 between the study groups in the comparative analysis of the gel images (p < 0.05, FC ≥ 1.2) (Fig. 1, Table 1). For OPMD, five proteins (AAT, APOA1, IGKC, SAMP, and VDBP) were found to have an increased expression, and the expression levels of 7 proteins (AINX, AMBP, CLU, HP, PRDX2, RAD50, and RBP4) were decreased when compared with controls. In early OSCC, the expression levels of 4 proteins (IGKC, IGHA2, IGHG2, and TF) were increased, and 8 proteins (ALB, AMBP, ARF, CLU, HP, LRG1, RAD50, and VCL) were decreased in expression. Whereas in advanced OSCC, five proteins (AAT, APOA1, C3, IGHG2, and VDBP) were increased in expression, and 2 proteins (ARF and PRDX2) were found to have a decrease in expression when compared with controls.

Functional classification, annotation and pathway analyses

The identified proteins were classified according to their cellular component, biological process, molecular function, and protein class in PANTHER analysis (Fig. 2). In the aspect of cellular component, most of the identified proteins contained cellular anatomical entity (59%). The predominant biological processes of the identified proteins were biological regulation (15%) and metabolic process (15%). These proteins were also involved in binding (59%) and catalytic activity (31%) as most of them are classified as transfer/carrier protein (42%).

The functional annotation clustering analysis revealed that the identified proteins are related to the regulation of cellular processes (Table 2). Most of the identified proteins were enriched within the extracellular space or region. Among these proteins, five proteins (C3, CLU, IGHA2, IGHG2, and IGKC) were implicated in the activation of classical complement pathway. The major molecular function of these proteins was related to catalytic and binding activity, of which 4 proteins (C3, IGHG2, IGKC, and HP) were associated with the serine-type endopeptidase activity.

The IPA showed the pathway annotations and interaction networks between the identified proteins (Fig. 3). For the category of molecular and cellular function, these proteins are involved in lipid metabolism, small molecule biochemistry, cellular compromise, cell cycle, and cell-to-cell signaling and interaction. Whereas for the category of associated network function, the most prominent network of the identified proteins is related to inflammatory response, and organismal injury and abnormalities. The top five canonical pathways in IPA were liver X receptor/retinoid X receptor (LXR/RXR) activation, farnesoid X receptor/retinoid X receptor (FXR/RXR) activation, acute phase response signaling, clathrin-mediated endocytosis signaling, and atherosclerosis signaling. Of these pathways, the LXR/RXR activation and acute phase response signaling pathways are actively associated with OSCC. Nine proteins (AAT, ALB, AMBP, APOA1, C3, CLU, RBP4, TF, and VDBP) from our dataset were found to be involved in the LXR/RXR activation pathway. There were also nine proteins (AAT, ALB, AMBP, APOA1, C3, HP, RBP4, SAMP, and TF) that are associated with the acute phase response signaling pathway.

ELISA analysis

Among the identified proteins, CLU and HP were selected for further validation, and the results are shown in Fig. 4. The mean serum levels of CLU in OPMD, early OSCC, and advanced OSCC were significantly lower than those in the control (p < 0.05) (Fig. 4A). Although only the mean serum level of HP in early OSCC was significantly lower compared to the control (p < 0.05), the concentration of HP also showed a lower level in advanced OSCC (Fig. 4B). Furthermore, we had performed receiver operator characteristic (ROC) analysis to evaluate the potential utility of serum CLU and HP in the detection of OSCC. The area under the (AUC) for CLU and HP was 0.945 (p < 0.001, 95% CI [0.91–0.99]) and 0.617 (p = 0.049, 95% CI [0.51–0.73]), respectively. The corresponding optimal cutoff values of CLU and HP to discriminate OSCC from control were 111.29 µg/ml (94.5% sensitivity, 89.0% specificity) and 1360.29 µg/ml (34.3% sensitivity, 90.4% specificity), respectively.

IHC analysis

Immunohistochemical analysis revealed that the epithelial cells in control, OPMD, and OSCC tissues showed membranous and granular cytoplasmic staining of CLU and HP. Some OPMD and OSCC tissue samples displayed weak or no staining of CLU and HP (Figs. 5A–5B). We observed that the expressions of CLU and HP were significantly lower in early and advanced OSCC when compared with control (p < 0.05) (Figs. 5C–5D). In ROC analysis, the AUC for CLU and HP was 0.833 (p = 0.001, 95% CI [0.66–1.00]) and 0.804 (p = 0.003, 95% CI [0.66–0.95]), respectively. The optimal IRS cutoff for CLU and HP was 1.50 (70.0% sensitivity, 95.9% specificity) and 10.50 (60.0% sensitivity, 89.8% specificity), respectively. There were 54.5% OPMD and 95.9% OSCC patients had low CLU expression level. Similarly, 81.8% OPMD and 89.8% OSCC patients exhibited low HP expression levels as well. However, the expression levels of CLU and HP were not associated with the clinicopathological parameters of the disease.