Our study included two independent experimental cohorts. The first group included 35 GC cases with fresh frozen tumors and adjacent tissues selected from January to March 2018 in the Second Hospital of Jilin University, these samples were analyzed by quantitative real-time PCR (qRT-PCR) and western blot analysis. The second group included 164 GC tissue samples from patients who received radical gastrectomy at the Gastrointestinal Cancer Surgery Department of the First Affiliated Hospital of China Medical University from March 2009 to June 2012; these samples were analyzed by immunohistochemistry for SKA3 and DUSP2. All included patients had a single lesion as confirmed by pathological verification and complete postoperative follow-up data; the patients did not receive chemoradiotherapy before surgery. The study protocol was approved by the hospital ethics committee, and all patients signed informed consent.

After dewaxing and hydration, paraffin-embedded sections were immersed in 3% H₂O₂ solution for 20 min at 37°C to eliminate endogenous peroxidase followed by incubation in sodium citrate buffer (pH 6.0) at 121°C for 2 min. Sections were then incubated with primary antibodies against SKA3 (1:200, Abcam, Cambridge, MA, United States), or DUSP2 (1:200, Abcam) at 4°C overnight. Sections were then incubated with secondary antibodies at 37°C for 30 min and then stained with 3, 3′-diaminobenzidine, followed by counter-staining with hematoxylin for 1 min. Sections were dehydrated in a gradient ethanol series followed by xylene. Samples were scored according to the staining intensity and proportion of positive cells. Staining intensity was scored as follows: 0 for colorless, 1 for yellow, 2 for brown, and 3 for dark brown. The proportion of positive cells was scored as follows: < 5% was 0, 5–25% was 1, 25–50% was 2, 50–75% was 3, and >75% was 4. The final score was the product of the staining intensity and proportion of positive cell scores; samples with scores ≥5 were considered positive and samples with scores ≤4 were negative. Two authors evaluated and scored, and were blinded to the patient’s pathological information. The controversial cases were re-scored and finally reached an agreement.

The immortalized gastric epithelium cell line GES-1 and GC cell lines SGC-7901, MGC-803, HGC-27, and MKN-45 were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). Cells were cultured in DMEM or RPMI 1640 medium containing 10% serum. The human peritoneal mesothelial cell (HPMC) HMrSV5 cell line was provided by Youming Peng (Second Hospital of Zhong nan University, Changsha, China) and cultured in RPMI 1640 medium containing 10% serum. Cells were cultured in an incubator at 37°C in 5% CO₂.

SKA3 silencing lentiviruses (sh-SKA3), DUSP2 silencing lentiviruses (sh-DUSP2), DUSP2 overexpressing lentiviruses (oe-DUSP2) and empty vector virus (control) were produced by Genechem (Shanghai, China). GC cells were seeded in 12-well plates and when cells achieved 60–70% confluence, an appropriate amount of virus was added according to the cell line MOI value. Stably infected cell lines were selected in 5 μg/ml puromycin or neomycin. The sequences used to contrast sh-SKA3 are follows: sh-1, CTT​ATG​AGA​ATC​TGC​TCA​GAA and sh-2, GCA​TAG​CTT​TGG​TAT​CCA​CAA. The sequence used for sh-DUSP2 is CAT​CTG​TCT​GGC​ATA​CCT​CAT and the control sequence is TTC​TCC​GAA​CGT​GTC​ACG​T.

Total RNA was extracted from GC tissues and cells by Trizol reagent (Invitrogen, Carlsbad, CA, United States). cDNA was generated by reverse transcription reaction using a reverse transcription kit (Takara, Tokyo, Japan), and the SYBR Premix Ex Taq ™ kit (Takara, Japan) was used to conduct qRT-PCR. The experiment was repeated three times. GAPDH mRNA was used as the internal reference gene. Data expression was determined by the −ΔΔCt method (Livak and Schmittgen, 2001).

The primer sequences for qRT-PCR are as follows.

SKA3 forward: 5′-GCC​ATA​GAT​ACA​GAA​TCC​AGG​CT-3′

SKA3 reverse: 5′-CCA​AAG​GAG​AGT​TGG​TAT​ATT​CGG-3′

DUSP2 forward: 5′-TGT​GGA​GGA​CAA​CCA​GAT​GGT​G-3′

DUSP2 reverse: 5′-GAG​GTA​TGC​CAG​ACA​GAT​GGT​G-3′

GAPDH forward: 5′-GTC​TCC​TCT​GAC​TTC​AAC​AGC​G-3′

GAPDH reverse: 5′-ACC​ACC​CTG​TTG​CTG​TAG​CCA​A-3′

HGC-27 cell lysates were incubated with primary antibody (or IgG as control) and A/G agarose beads at 4°C for 6 h. After three washes with IP buffer, samples were analyzed by western blot.

Cell viability was measured using the Cell Counting Kit-8 (CCK8, MedChemExpress, United States). GC cells were seeded at 2 × 10³ cells/well in 96-well plates. After culture for 24, 48, 72 and 96 h, 10 μl CCK8 solution was added to each well. After incubation for 2 h, the absorbance value was measured at 450 nm using a Biotek microplate reader.

For colony formation assays, cells were seeded in 6-well plates (1 × 10³ cells/well). After culture for 14 days, the cells were fixed with paraformaldehyde for 20 min and stained with 0.4% Trypan blue for 15 min for clone counting.

Cells collected by trypsinization were fixed overnight with 900 μl of pre-cooled 75% ethanol at 4°C. Cells were rinsed with PBS and incubated in 300 μl propidium iodide staining solution at 37°C for 30 min in the dark. Cell cycles distribution was then analyzed by flow cytometry.

For wound-healing experiments, cells were uniformly seeded in a 6-well plate. Once cells achieved 80% confluence, a 100 μl pipette tip was used to scratch the cell monolayer. The same field of vision was photographed under an inverted microscope at 0 and 48 h after the scratch, and the scratch width was measured.

For migration experiments, a Transwell chamber (8 μm, Corning, United States) was placed in a 24-well plate and cells were seeded into the upper chamber at a density of 2 × 10⁴/well. Serum-free medium (100 μl) was added to the top well, and the lower chamber was filled with 600 μl medium containing 10% serum. After 16 h of culture, the Transwell chamber was removed, fixed with paraformaldehyde, and stained with 0.4% trypan blue.

For invasion experiments, 90 μl of diluted Matrigel (BD Biosciences, NJ, United States) was added to the upper chamber and 2 × 10⁴/well cells were also seeded into the upper chamber. The subsequent steps were the same as for the migration experiment.

HMrSV5 cells were seeded in a 12-well plate 1 day in advance and GC cells were seeded in a 6 cm Petri dish. Next, 2 μl Calcein-AM (Sigma, United States) was added to GC cells and cells were incubated in the dark for 1 h. The stained GC cells (1 ×10⁴) were then added to each well of the plated HMrSV5 cells. The cells were incubated in the dark for 1 h and then washed. The number of GC cells that adhered to HMrSV5 cells was counted under a fluorescence microscope.

All animal experiments were approved by the Ethics Committee of China Medical University and all operations were in accordance with animal ethic guidelines.

A subcutaneous tumor model was generated in 12 BALB/c immunodeficient nude mice (SPF level; 4–5 weeks old, female, and approximately 20 g in weight). The MGC-803 cells infected with lentivirus for SKA3 silencing or controls were injected into 3 × 10⁶ cells respectively in the left and right axillae of nude mice. Tumor volume was measured every 3 days. After 4 weeks, the subcutaneous tumors were dissected and weighed.

An intraperitoneal metastasis model was established in 20 BALB/c nude mice. Mice were assigned to two groups (n = 10 per group), and 5 × 10⁶ MGC-803 cells infected with lentivirus for SKA3 silencing or controls were injected into the abdominal cavity. After 6 weeks, mice were killed and dissected. The number of tumors on the peritoneal tissue was observed.

RNA-seq analysisTrizol reagent was used to extract the total RNA of the MGC-803 silencing and control groups of cells, and each group of cells repeated three biological samples. The subsequent RNA-seq analysis was completed by professional sequencing company. The sequencing data is not uploaded to the GEO public database, but owned by ourselves.

Statistical analysis was performed using SPSS 22.0 and GraphPad Prism 7.0. Experiments were repeated three times. Data are expressed as mean ± SD. The unpaired t test was used to compare the groups. The chi-square test was used to analyze the relationship between SKA3 expression and clinicopathological factors. The Pearson test was used to analyze the correlation between SKA3 and DUSP2 expression. Survival analysis was performed using the Kaplan–Meier method. Log-rank test calculated the survival differences among different groups and fitted the univariate and multivariate Cox proportional hazards model. p < 0.05 was considered statistically significant.

We next established a subcutaneous tumor model and examined the effects of SKA3 knockdown on tumor growth (Figure 4B). We found that tumor growth was reduced in the SKA3 silenced group; the volume and weights of tumors derived from SKA3 silenced cells were smaller than those in the control group (Figures 4C,D).

We further studied the impact of SKA3 silencing on the metastatic ability of GC in vivo using an intraperitoneal metastasis model (Figure 4E). While extensive and diffuse cancer nodules in the abdominal cavity were observed in the control group, the weight of nodules in the SKA3 silenced group was significantly reduced (Figure 4F). Together, these results showed that reduced expression of SKA3 inhibited the proliferation, invasion and metastasis of GC in vivo.

DUSP2 inhibits the MAPK/ERK pathway activity through binding ERK1/2 and inhibits the proliferation, migration, invasion, adhesion and EMT of GC cells.

To study the mechanism by which SKA3 promotes the progression of GC, we performed RNA-seq using MGC-803 cells silenced for SKA3 compared with controls. The results revealed 997 genes that were upregulated and 236 genes that were downregulated in MGC-803 cells silenced for SKA3. We screened out the downstream gene DUSP2 (Figure 4G). Previous studies showed that DUSP2 functions as an upstream negative regulator of the MAPK/ERK pathway. The MAPK/ERK pathway plays an important role in the progression of GC (Wang et al., 2012; Hu J. et al., 2018). We screened DUSP2 as a key downstream gene through qRT-PCR and western blot (Figures 4H–I).

We established two stable cell lines overexpressing DUSP2 (MGC-803-oe-DUSP2 and HGC-27-oe-DUSP2), as well as a control (oeNC). Western blot was used to confirm overexpression efficiency in the DUSP2 overexpressing cell lines (Figure 6A). CCK-8 cell proliferation assays showed that the cell proliferation ability of the DUSP2 overexpression group was significantly lower than that of the control group (Figure 5A). Colony formation assays showed that the number of cell clones in the DUSP2 overexpression group was less than that in the control group (p < 0.05) (Figure 5B). Cell cycle analysis of the DUSP2 overexpression group showed a cell cycle arrest in G0/G1 and G2/M phase (Figure 5C). Wound-healing experiments showed that DUSP2 overexpression inhibited the migration of GC cells (Figure 5D), and Transwell migration and invasion experiments showed that DUSP2 overexpression also significantly inhibited the migration and invasion of GC cells (Figures 5E,F). In adhesion experiments, we found that the number of GC cells adhering to mesothelial cells in the DUSP2 overexpressed group was significantly lower than that in the control group, indicating that overexpression of DUSP2 inhibited the adhesion ability of GC cells (Figure 5G). Western blot showed that DUSP2 overexpression resulted in up-regulation of E-cadherin and down-regulation of ZEB1, N-cadherin, and Vimentin (Figure 6A). We performed co-immunoprecipitation experiments and found that DUSP2 binds to ERK1/2 (Figure 6B), thereby dephosphorylating ERK1/2 and inhibiting MAPK/ERK pathway activity.

Knockdown of DUSP2 partially rescued the effects on proliferation and invasion caused by SKA3 knockdown.

To verify the SKA3-DUSP2-ERK1/2 signal axis, we performed functional rescue experiments by co-infecting GC cells with shSKA3 and shDUSP2. First, we verified the DUSP2 silencing efficiency in co-infected cells (Figure 6C). The co-infected group showed partially increased cell proliferation (Figures 6D,E), G0/G1 phase population (Figure 6F), wound-healing ability (Figures 7A,E), migration (Figures 7B,F), invasion (Figures 7C,G) and adhesion (Figures 7D,H) compared with SKA3 knockdown cells. Western blot showed that SKA3, ZEB1, N-cadherin, Vimentin, and phosphorylated ERK1/2 expression were partially increased, E-cadherin expression was partially decreased, and total ERK1/2 was unchanged in the double knockdown cells compared with SKA3 knockdown cells (Figure 8A). We also used the ERK1/2 pathway inhibitor PD98059 in cells with overexpression of SKA3, and western blot showed that SKA3, ZEB1, N-cadherin, Vimentin, and phosphorylated ERK1/2 expression were partially decreased, E-cadherin expression was partially increased, and total ERK1/2 was unchanged compared with cells with SKA3 overexpression (Figure 8B). These results show that the effect of SAK3 on proliferation and metastasis is likely largely achieved through downregulation of DUSP2. Our findings suggest that SKA3 negatively regulates DUSP2, through binding ERK1/2 to activate the MAPK/ERK pathway to regulate the malignant phenotype of GC (Figure 8C).