Complement C3 overexpression activates JAK2/STAT3 pathway and correlates with gastric cancer progression

Complement gene expression in GC by RNA sequencing were retrieved from the UCSC Cancer Genomics Browser, which were collected from The Cancer Genome Atlas (TCGA) data portal. 384 GC tissues and 37 tumor-adjacent normal tissues were harvested from the TCGA cohort. Additional 12 samples with paired tissues were extracted from the Oncomine database to assess gene and protein expression of the complement system for GC.

From August to December 2013, adult patients with confirmed diagnosis of GC were prospectively enrolled. Consecutive participants referred or admitted to our center for surgical treatment were screened for eligibility. All included patients were managed and followed up as our published procedures [2, 19]. Inclusion criteria were: (1) A primary GC tumor should be resectable on the basis of preoperative evaluation, with no suspicion of distant metastasis; (2) adult age between 18 and 75 years, with no limitation of gender; (3) A radical gastrectomy with adequate lymphadenectomy was suggested after a multidisciplinary team meeting, with adjuvant chemotherapy scheduled or not.

Exclusion criteria included: (1) another synchronized malignancy, severe concomitant basic disease (cardiopulmonary dysfunction, tuberculosis, Crohn’s disease or psychosis) and uncontrolled infection except for Hp infection; (2) requirement of emergency surgery due to tumor progression, history of major abdominal surgery within the last six months; (3) long-term use of corticosteroids, insulin, oral antidiabetic drugs, or other agents for obesity; (4) history of blood transfusion or purification therapy within the last three months.

This study was carried out in accordance with the recommendations of NCCN guidelines for gastric cancer, with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Institutional Review Board of First Affiliated Hospital of Sun Yat-sen University and registered at ClinicalTrials.​gov (NCT02425930). The last follow-up date was 7th July 2018.

Human SGC-7901 and MGC-803 cells, normal gastric epithelial cells (GES-1) were purchased from the Cell Bank of Chinese Academy of Medical Science (Shanghai, China). All cells were cultured in RPMI-1640 medium supplied with 10% fetal bovine serum (FBS, Thermo Fisher Scientific, Waltham, MA), penicillin (100 U/mL), and streptomycin (100 mg/mL) at 37 °C in a humidified atmosphere of 5% CO₂. Cells were routinely tested for mycoplasma contamination (MycoAlert™ PLUS Mycoplasma Detection Kit, Lonza). For cell culture, 50–60% confluent cells were transiently incubated with specific agents for 48 h until the extraction of RNA and protein lysates.

Purified recombinant human C3 protein (HuC3, 20 ng/ml, MBS230377, MyBioSource, San Diego, CA) was added to or left out of the culturing medium. Exogenous C3 was depleted with cobra venom factor (CVF; Heng Fei biological technology, Shanghai, China), as previously described [20]. Complement receptor 1 (CR1/CD35, MBS717740, MyBioSource) was used to block C3 activation as previously confirmed [21], with JAK2 blocker (AG490, 25 μM, InvivoGen, Hongkong) used to inhibit STAT3 signaling pathway [22].

Cell lysates were extracted from gastric tissues and cancer cell lines. The primary antibodies targeted C3, C3a, C5a, CD35, Factor B (fB), IL-6, JAK2, STAT3, pSTAT3 and GAPDH proteins (Abcam, USA). Total protein was obtained through a cell lysis buffer (KeyGene, Nanjing, China) and the protein concentration was quantified using the enhanced BCA protein assay kit (KeyGene, Nanjing, China). The PageRuler™ prestained protein ladder (#26616, Thermo Fisher Scientific, USA) was used to estimate protein sizes. Proteins were separated by 8–10% sodium dodecyl sulfate polyacrylamide gel electrophoresis and electrotransferred to polyvinylidene fluoride (PVDF) membranes (BioRad, Richmond, CA). After that, the membranes with deposited proteins were blocked for 1 h in tris-buffered saline Tween (TBST; T8060, Solarbio) and probed with various primary antibodies, overnight at 4 °C, followed by incubation with rabbit and mouse radish peroxidase-coupled secondary antibody (BA1055, 1:2500; Biosterbio, Wuhan, China) for 1 h. Protein bands were visualized by using an enhanced ECL™ detection kit (KGP1121, KeyGene, Nanjing, China) and captured with a camera (Canon Inc., Japan).

qRT-PCR experiments were performed to detect mRNA expression of C3, C3a and C5, as previously reported [23, 24]. Briefly, total RNA was extracted from GC cell lines using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Aliquots of total RNA were reversely transcribed into single-strand-cDNA by incubation with virus reverse transcriptase (6110A; Takara Biochemicals, Kusatsu, Japan). After that, the specific primers for C3, C3a and C5 mRNAs (Additional file 1: Table S1) were used to guide the amplification of cDNA products with 40 cycles at 95 °C for 20s and 60 °C for 1 min. The abundance of each target mRNA was normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA and presented as 2[(Ct/GAPDH - Ct/gene of interest)].

IHC method was employed to detect regional deposition of complement components, including C3 (ab200999, Abcam), C3a (neo-epitope, ab11873, Abcam), C5a (neo-epitope, ab11878, Abcam), fB (ab192577, Abcam), and CR1 (anti-CD35, ab25, Abcam). IHC results were analyzed by two experienced pathologists blinded to patient’s information, and were scored by a semi-quantitative method in which staining of more than 10% of tumor cells was considered positive. The staining intensity was scaled as 0 for negative, 1 for weak (10~40%), 2 for moderate (40~70%) and 3 for strong (> 70%). The average score of staining intensity was calculated with five independent high-power fields using IMAGE PLUS software (Version 6.0, Media Cybernetics, USA). Low and high C3 deposition were defined as ≤1 and > 1 point, respectively. Deparaffinized sections from harvested human tissues were pretreated with 10 mM sodium citrate buffer (pH 6.0, boiling temperature, 30 min), blocked in normal serum (Vectastain ABC Kit; Vector Lab., Inc., CA, USA), incubated with primary antibodies (solution with saline, 1:100) at 4 °C overnight, rinsed and incubated with secondary antibody (EliVision plus, DAB Kit, 9902).

Double-marker immunofluorescence staining on paraffin-embedded human tissues were performed as previously described [25]. The primary and secondary antibodies included rabbit anti-human C3 (1:2000, ab20099, Abcam), rabbit anti-human C5a (1:2000, ab11876, Abcam), rabbit anti-phosphorylated (p)-STAT3 (1:2000, Cell Signaling Technology, Danvers, MA, USA) and goat anti-human IL-6 (1:800, R&D systems, Minneapolis, MN). Bound antibodies were revealed by fluorochrome-conjugated antibodies: Alexa Fluor 594 goat anti-Rabbit IgG (H + L; 1:300, ZF-0513) and Alexa Fluor 488 goat anti-Rabbit IgG (H + L; 1:300, ZF-0512) both from Invitrogen Molecular Probes (Carlsbad, CA). All slides were counterstained with DAPI Nucleic Acid Stain (1:1000, Carlsbad, CA) for 60 min. Finally, confocal analysis was performed with a Nikon C2 confocal system (Nikon, Melville, NY) to capture separated and merged images from all sections.

Plasma levels of C3a, C5a and fB in GC patients at perioperative period were monitored using specific ELISA Kits (Thermo Scientific, Frederick, USA). Briefly, 100 μl per well of plasma with standard solution were added to antibody-coated 96-well plates and incubated for 2 h at room temperature, followed by incubation of polyclonal antibody for specific effector for 1 h. Then the plate was washed and incubated with avidin conjugated to horseradish peroxidase (Lifespan BioScience, USA) for 1 h, followed by optical density detection with an ELISA plate-reader at 450 nm. Assays detected only cleaved C3a and C5a peptides in plasma.

A transwell assay system (Corning co. Ltd., USA) was employed to assess GC cell invasion, with a wound-healing assay used to assess tumor migration, as previously published [24, 26].. Nearly 2.0 × 10⁴ cells in 100 μl of serum-free medium were added to each upper chamber for 24 h, with 5 replicate wells set up for each condition. Medium containing 10% FBS applied to the lower chamber as chemo-attractant. After 24 h of incubation, the cells that migrated and adhered to the surface of lower chamber were fixed with ethanol, stained with 0.5% crystal violet, photographed at 200x, and counted at 400x magnification (Olympus, Japan).

A wound-healing assay was performed to assess cancer cell migration. In brief, cells were plated (2 × 10⁵/well) in 6-well plates and grown to 90% confluency. Central bands were artificially scratched with a sterile pipette tip to make a 1 cm wide ribbon. Afterward, the dislodged cells were removed by two PBS washes and serum-free culture medium was added for 48 h. The wound width and cell density were observed at 12 h, 24 h and 48 h, respectively.

Flow cytometric analysis detected the apoptosis rate and cell cycle of GC cells, as previously described [27, 28]. Cancer cells were harvested after 48 h of culture via ethylene diamine tetra-acetic acid-free trypsinization. Human purified C3 protein (20 ng/ml) or CVF protein (40 ng/ml) was selectively added into the culture medium, with PBS added as normal control (NC). Early apoptosis rate was detected using an annexin V-fluorescein isothiocyanate apoptosis detection kit (Oncogene Research, Boston, MA). The cell cycle was investigated via PI/RNase staining methods by using FACScan and CellQuest software (Becton Dickinson, CA).

The relationship between regional expression of C3 and clinical characteristics was analyzed with the chi-square test. Continuous variables were compared between both groups with the t test. Correlation between C3 deposition and other factors was revealed with a linear regression. Survival analyses were performed using the Kaplan-Meier estimate. The prognostic value of selective parameters was determined with the receiver operating characteristic (ROC) curve analysis, with a value of area under the curve (AUC) approaching 1.0 showing predictive power. All data were analyzed with SPSS® (Version 23.0). Statistical significance was set at 0.05.

A total of 106 patients were analyzed, with 65(61.3%) males and 41(38.7%) females. The flow chart of study design is shown in Fig. 1. Briefly, 41(38.7%) and 65(61.3%) patients were assigned to low and high C3 deposition groups, respectively. The median follow-up period was 41 (range, 1–57) months, which was significantly shortened in the high C3 group compared with the low C3 group (29 months vs. 43 months, P = 0.006). The demographic and baseline characteristics (Table 1) were almost comparable between the two groups (P > 0.05), except plasma levels of C3 and C4, and tumor histology (P < 0.05). Open gastrectomy plus adequate lymphadenectomy was performed in 98 patients (92.5%), with laparoscopic approach applied in only eight patients (7.5%). The surgical parameters were similar between both groups (Additional file 2: Table S2).

In the bioinformatics TCGA cohort, the overall mRNA levels of C3 expressed in tumor tissues were markedly upregulated compared with normal gastric tissues (P = 0.007; Fig. 2a, left panel). The C3 upregulation was further validated in paired tumor and adjacent normal tissues (P = 0.002; Fig. 2a, middle panel); however, the C5 expression was not significantly different between paired samples (P = 0.546; Fig. 2a, right panel). In the Oncomine cohort, the C3 deposition was significantly enhanced in GC tissues compared with gastric mucosa or adjacent normal tissues (P < 0.001; Fig. 2b). Afterward, we analyzed the expression of C3 and other C3-related components in the paired GC and normal tissues from enrolled subjects. The protein levels of both C3 and C3a in GC tissues were highly increased compared with adjacent normal tissues (P < 0.001; Fig. 2c), with no significance observed for C5a, CR1 or fB levels. Moreover, regional depositions of C3 and C3a in GC tissues were markedly enhanced compared with C5a and other complement proteins (Fig. 2d&E).

We employed the average score of C3 with IHC method (Fig. 3a) and divided patients into low and high C3 deposition groups with a cutoff value of 1.0 (Fig. 3b, left panel). We found that all patients were distributed into three subsets as trisection of IHC C3 scores, with 1:2 ratio for low and high C3 groups (Fig. 3b, right panel).

We investigated the relationship between localized C3 deposition and plasma levels including complement C3 and effectors at baseline, during surgery and after surgery, respectively. The linear regression results showed that average IHC C3 score was negatively correlated with systemic C3 level at baseline (r² = 0.658, P < 0.001) and positively correlated with systemic C3a (r² = 0.944, P < 0.001; Fig. 3c) and fB (r² = 0.871, P < 0.001; Fig. 3d) levels during surgery. However, it was not associated with either C4 or C5a plasma level in the current cohort. The further external validation using TCGA cohort showed a non-correlation between localized C3 and plasma C5 expression in GC patients (P = 0.137; Fig. 3e).

First, we explored the correlation between C3 deposition and tumor stage (Fig. 4 a). The findings indicated that it was positively correlated with pathological T (r² = 0.459, P < 0.001) and TNM stages (r² = 0.2155, P < 0.001), but not related to pathological N stage (P = 0.287) or clinical TNM stage (P = 0.383).

Second, we reviewed the long-term outcomes of GC patients from the TCGA dataset. We found that patients with high C3 expression in GC tissues had poorer overall survival (OS; Fig. 4b, left upper quadrant) and recurrence-free survival (RFS; Fig. 4b, right upper quadrant) than those with low C3 expression, with a survival significance observed in OS (P = 0.028). Subsequently, we compared the 5-year survival outcomes with our data (Fig. 4bb, left and right lower quadrants) and confirmed that high C3 deposition was a predictive factor of poor OS (P = 0.008) and RFS (P = 0.036). The 5-year OS and RFS rates were 52.6 and 50.7% in the low C3 group and 29.7 and 28.2% in the high C3 group, respectively. By further subgroup survival analyses based on tumor stage (Fig. 4c), we detected a survival significance of C3 deposition in stage III patients (P = 0.034), with no significance observed in other stages (P > 0.05). The results for RFS were not significant in each stage (Fig. 5).

Third, we performed a ROC curve analysis including IHC C3 score, plasma C3 level at baseline, pathological stage and two tumor markers to determine the prognostic value for tumor-related death (Fig. 4d). The findings indicated that both IHC C3 score (AUC = 0.651) and serum CEA level (AUC = 0.646) were valuable in predicting oncologic outcome, but inferior to pathological tumor stage (AUC = 0.842). We combined both useful parameters and obtained a better value (AUC = 0.744) that was comparable to tumor stage. The ROC analyses revealed that optimal cut-off values were 1.4 for IHC C3 score and 4.2 ng/ml for CEA level. We also compared the incidence of postoperative morbidity between both groups (Fig. 4e), which suggested that the localized C3 deposition was not significantly associated with any morbidities after surgery (P > 0.05). Of note, the relative risk (RR) of surgical site infection (RR, 0.525; 95% confidential interval [CI], 0.187–1.476) and anastomotic leak (RR, 0.300; 95%CI, 0.034–2.665) was both reduced in the low C3 group compared with the high C3 group.

At last, we determined the prognostic value of localized C3 deposition by using univariate and multivariate Cox regression analyses against 5-year OS (Table 2). We verified that high C3 deposition in GC tissues (odds ratio [OR], 1.848; 95%CI, 1.015–3.363; P = 0.045), along with advanced tumor stages (stage III and IV; OR, 2.609; 95%, 1.725–4.194; P < 0.001), depleted plasma C3 level (< 0.75 mg/ml; OR, 1.801; 95%CI, 1.049–3.090; P = 0.033) and any morbidities after surgery (OR, 2.770; 95%, 1.446–5.305; P = 0.002), were independent factors for poor 5-year OS in GC patients.

We examined RNA and protein expression of C3 and complement effectors in GC (SGC-7901 and MGC-803) and gastric mucosa (GES-1) cell lines (Fig. 6a). We found that both C3 and C3a were highly expressed in SGC-7901 and MGC-803 compared with GES-1; whereas C5 was similarly expressed across those cell lines. Besides, we observed a significantly decreased cell migration in CVF-treated SGC-7901 after 48 h of culturing (Fig. 6b, left panel). Exogenous C3 treatment could enhance cell proliferation in both SGC-7901 and MGC-803, but quickly shut down such growth once CVF was added into the C3-contained culture medium (Fig. 6b, right panel). Additional invasion experiments indicated that exogenous C3 could promote invasion capacity, which could be markedly depressed by CVF (Fig. 6c).

Next, we performed flow cytometric analysis of cell cycle and apoptosis (Fig. 6d). Exogenous C3 caused a dramatic decrease of apoptosis in MGC-803 cells compared with NC (10.8% vs. 7.3%, P = 0.0462). The use of CVF in the CM resulted in a reverse increase of apoptosis compared with NC (22.5% vs. 7.3%, P < 0.001). Meanwhile, the cell cycle study in SGC-7901 also confirmed an increased percentage of cells in S phase from C3 treatment (32.6% vs. 19.7%, P = 0.013) and an enhanced population in apoptotic phase from CVF interference (15.3% vs. 6.4%, P = 0.003).

We detected the activation of JAK2/STAT3 axis in human GC tissues first. Expression of both STAT3 phosphorylation (p-STAT3) and IL-6 were significantly enhanced in GC tissues compared to adjacent normal tissues (Fig. 7xa). A similar result was observed in comparing SGC-7901 with GES-1 in vitro. Afterward, we treated SGC-7901 with exogenous C3 and detected increased expression of p-STAT3 and p-JAK2 (Fig. 7b). However, pre-incubation of cells with AG490 and exogenous C3 significantly blocked the C3-induced increases in JAK2/STAT3 phosphorylation, which indicated that C3 might work as an upstream regulation of JAK2/STAT3 activation. We employed CR1 to block exogenous activation of C3 and detected a weakened expression of p-STAT3 and IL-6 compared with AG490-treated cancer cells (Fig. 7c). These data indicated that localized activation and deposition of C3 may play a role in tumor growth and metastasis by potentiating JAK2/STAT3 activation (Fig. 7d).