JP3, an antiangiogenic peptide, inhibits growth and metastasis of gastric cancer through TRIM25/SP1/MMP2 axis

Human metastatic GC cell line BGC823 was gifted from Professor Q Guo, China Pharmaceutical University. The SGC7901 and MGC803 cell lines were obtained from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). Human umbilical vein endothelial cells (HUVECs) were purchased from the Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences (Shanghai, China). HUVECs, BGC823 and SGC7901 cells were cultured in RPMI-1640 (Hyclone, Thermo Scientific, Waltham, MA, USA), while MGC803 cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% fetal bovine serum (FBS) (Hyclone), 100 U/mL streptomycin and 100 μg/mL penicillin (Invitrogen, Carlsbad, CA, USA) in a humidified incubator at 37 °C with 5% CO₂. Cycloheximide (CHX) (Sigma-Aldrich, St Louis, MO, USA) and MG132 (Selleck Chemicals, USA) were used at the indicated concentrations.

The wild-type Flag-SP1(WT) or its mutants [Flag-SP1 (K610R), Flag-SP1 (K624R), Flag-SP1 (K685R), Flag-SP1 (K693R) and Flag-SP1(ALL)] were inserted into the GV141 eukaryotic expression vector (Genechem, Shanghai, China) using XhoI/KpnI sites. Small interfering RNA used for TRIM25 gene knockdown was produced by Genomeditech (Shanghai, China). The sequences of siRNAs were as follows: si-TRIM25 1 (si-RNA 1), 5′-GAGUGAGAUCCAGACCUUGAA-3′ (forward) and 5′-UUCAAGGUCUGGAUCUCAACUC-3′ (reverse); si-TRIM25 2 (si-RNA 2), 5′-GAACUGAACCACAAGCUGAUA-3′ (forward) and 5′-UAUCAGCUUGUGGUUCAGUUC-3′ (reverse); si-TRIM25 3 (si-RNA 3), 5′-GUGCCCGAUUCCUCUUAGAGA-3′ (forward) and 5′-UCUCUAAGAGGAAUCGGGCAC-3′ (reverse). A scrambled siRNA was used as the negative control. Plasmids and siRNAs were transfected into cells using Lipofectamine 3000 (Invitrogen).

All animal experiments were in accordance with the Jiangsu Provincial Guidelines for the use of experimental animals, and it was approved by the Animal Care and Use Committee of Nanjing Medical University. Briefly, BALB/c nude mice (female, 6 weeks) were obtained from NLARSH China (Shanghai, China). Animals were housed in plastic cages and kept in a temperature-, humidity-maintained, and light-controlled room (23 ± 1 °C; 50 ± 5% humidity; 12 h light/dark cycle starting at 7:00 am). They had ad libitum accesses to food and water.

For establishing xenograft mouse models, 5 × 10⁶/200 μL of BGC823 cells were subcutaneously injected into flank of BALB/c nude mice. When the tumor volume (V = 0.5 × length × width²) reached about 100 mm³, the mice were randomly divided into different groups and received the indicated treatments. General situations and activities of each mouse were daily recorded. Tumor sizes were recorded by caliper every 3 days for calculating tumor volumes. Body weight of the experimental mouse was also recorded every 3 days. The endpoint of the experimental therapy was determined until enough growth (up to 2000 mm³) of the xenograft tumor or intolerable side effects.

In the present study, the strategy to screen JWA protein based functional tumor inhibitory peptides included four steps, (1) all the candidate fragments were selected from non-transmembrane regions and contain amino acid can be phosphorylated in its sequence; (2) the intra-tumor injection in xenograft GC tumor in mouse model was used for first round screening for all candidate fragments; (3) the identified functional peptides in (2) were modified with HWGF as MMP2 target and used for secondary screening by intraperitoneal injection in xenograft GC mouse model; (4) the functional peptides selected in (3) were further used to inhibit metastasis of GC (Fig. 1a).

For establishment of metastatic GC mouse model, five-week-old BALB/c nude mice were randomly divided into two groups, 3 mice each. BGC823 cells were suspended in PBS. The mice were injected intravenously with 2 × 10⁶ cells in 0.15 mL of PBS through tail vein. On next day, the mice were treated by intra-peritoneal (ip) injection of Ctrl-H-P and JP3 and followed by luciferase substrate (GM-040611, Genomeditech, China); and photographed using the IVIS Spectrum in vivo imaging system (PerkinElmer, USA) on day 50^th to observe the cancer cell metastasis.

According to the predicted phosphorylation sites in 188 amino acids of JWA, six fragments (Fig. 1b) were selected and finally ten peptides were synthesized since only a single phosphorus modification to one amino acid residue was designed for each peptide. Both PJP3 and PJP6 fragments were split in three modified ones from each since three modifiable amino acid residues were contained in its sequences, respectively. A routine acetylation modification to each peptide was conducted to generate N-acetylation and C-amide, thereby increase its in vivo stability. According to the full length of JWA amino acid sequence and spatial structure, each synthesized peptide with a phosphorylated serine (S), threonine (T) or tyrosine (Y) in the center, and extension of 3–7 amino acid sequences in both sides. The MMP2-targeting motif HWGF was added after a four glycine (G) linker at C- termini of peptide (named as JP1-JP3).

All the peptides were synthesized by GL Biochem Ltd. (Shanghai, China), with a purity of more than 98%. These solid-phase synthesized peptides have good water solubility. The synthesized peptides were packed as 10 mg in sterile Eppendorf tube and always kept at − 20 °C. All peptides were dissolved in pre-cold PBS to the indicated doses/concentrations before use.

GC cells were treated with different concentrations of FITC-JP3 for different time points, followed by washing with phosphate buffer saline (PBS) for 3 times. Cells were then fixed with methanol for 15 min at room temperature and washed in PBST (PBS supplemented with 0.5% Tween-20). Cell nuclei were counterstained with DAPI (Beyotime, Jiangsu, China) for 20 min. The confocal images of stained cells were captured using the Zeiss AIM software on a Zeiss LSM 700 confocal microscope system (Carl Zeiss Jena, Oberkochen, Germany).

BGC823 and SGC7901 GC cells were grown to confluence and processed for coimmunoprecipitation by standard procedures as previously described [16]. The pre-cleared lysate was incubated with anti-SP1 or anti-TRIM25 polyclonal antibodies for 1 h, and then incubated with protein A/G agarose beads overnight. Collected beads by centrifugation were washed for 5 times with washing buffer, and resuspended in 1× sodium lauryl sulfate loading buffer. Immunoprecipitation was eluted by incubation at 95 °C for 5 min.The eluted proteins were separated by sodium dodecyl sulfate - polyacrylamide gel electrophoresis, and then Western blotting with anti-SP1 (CST, Danvers MA, USA), anti-TRIM25 and anti-ubiquitin antibodies, respectively.

GC cells were treated with or without JP3, followed by incubation with MG132 (10 μM) for another 6 h. Cells were harvested, prepared for protein samples and divided for Western blotting and co-IP assay, respectively. For co-IP assay, the protein (500 μg) was cultured with anti-SP1 or anti-TRIM25 antibody (dilution: 1:250) at 4 °C for 1 h, and Protein A/G Plus-Agarose overnight. After three cycles of washing in pre-cold IP buffer and centrifugation at 4 °C, 1000×g for 5 min, the immunoprecipitate was collected by centrifugation, eluted and examined by Western blotting.

Western blotting analysis was carried out as previously described. The antibodies against MMP2, Tubulin, Ub, HA-tag, Flag-tag and Ki-67 were purchased from Cell Signal Technology (Danvers MA, USA); anti-CD31 and PCNA were provided by Epitomics (MA, USA); anti-GAPDH and Actin were purchased from Beyotime (Jiangsu, China); anti-TRIM25 and SP1 were provided by Proteintech (Wuhan, China); anti-His was purchased from MBL (Japan). All above antibodies were used at a dilution of 1:1000.

According to the manufacturer’s protocol, total RNAs were isolated using TRIzol (Invitrogen, CA, USA) and reversely transcribed into cDNAs using reverse transcription kit and the SYBR Green Master Mix kit (Takara, Otsu, Japan). The complementary DNA (cDNA) was amplified with the following primers: 5′-GTCTCTACCCAGAACAGTTTCC-3′ (forward) and 5′-ATCCAACACAGGCTGATTCC-3′ (reverse) for TRIM25; 5′-GCCGGTGCTGAGTATGTC-3′ (forward) and 5′-CTTCTGGGTGGCAGTGAT-3′ (reverse) for GAPDH; 5′-GGTGCCTTTTCACAGGCTC’-3′ (forward) and 5′-GCTGTTCTCATTGGGTGACTC-3′ (reverse) for SP1. Quantitative RT-PCR was carried out an ABI Prism 7900 Sequence detection system (Applied Biosystems, Canada). Relative levels were normalized to that of GAPDH.

After 6 hours of peptide treatment of GC cells, the medium was replaced and cultured in a 60 mm petri dish containing 1% serum for 24 h. On the next day, 2 mL of conditional medium was collected. For tube formation assay, a 96-well plate was coated with 50 μL of Matrigel™ (BD Biosciences) and kept at 37 °C for 2 h. Then, 1.2 × 10⁴ HUVECs suspended in 100 μL of conditional medium were applied to each well of the precoated 96-well plate. After incubation at 37 °C for 4 h, tubular structures formed in the Matrigel were captured in five randomly selected fields per sample under a microscope.

I-TASSER (Iterative Threading Assembly Refinement) was designed for protein structure modelling by iterative threading assembly simulations. The molecular structures of JP3, MMP2, and TRIM25 were constructed by I-TASSER. The largest possible binding pocket of MMP2 and TRIM25 was then predicted by Discovery Studio 3.0. These predicted pockets were utilized to construct an initial coarse model of the JP3-MMP2, JP3-MEK1/2 and JP3-TRIM25 complex. The coarse model of a complex was then refined using RosettaDock. Since a peptide was considered as a rigid body in RosettaDock, the peptide–protein complex with the lowest energy was then refined by utilizing the FelxPepDock module of Rosetta. The output of 2000 models was then ranked based on their energy scores. The model with the lowest energy was chosen for further research, i.e., binding sites and electrostatic properties. The binding sites between proteins were obtained based on space coordinates and types of amino acid (hydrogen bond: ≤3.5 Å; ionic: ≤3.5 Å; disulphide: ≤2.5 Å; van-der-waals: ≤0.5 Å; pi-pi interaction: ≤6.5 Å). Electrostatic interactions were found in almost all biomolecular systems and processes. To predict potential phosphorylation sites of TRIM25, a visual inspection of the electrostatic potential at a solvent-accessible surface was used. The electrostatic properties of the structures were calculated with the Adaptive Poisson–Boltzmann Solver (APBS) and PDB2PQR. High-quality 3D images of proteins were drawn with PyMOL/VMD.

The construction of the GC TMAs was performed with standard procedures, and the assessment of the IHC employed a semiquantitative immunoreactivity score (IRS) as reported elsewhere [17]. The concordance for the IRS of the TRIM25 and SP1 staining scores between the two pathologists was 81 (90%) in 90 of the TMA cohort. The optimum value of cutoff points of the TRIM25 or SP1 IRS in this study were both 6 because the predictive value of this cutoff point for death was the best in the GC cohorts. Under these conditions, the samples with IRS scores 0–6 and IRS 8–12 were classified as low and high expressions of TRIM25 and SP1, respectively.

All analyses and depicted graphs were performed with GraphPad Prism 6.0 Software (GraphPad Inc., San Diego, CA, USA). Data were expressed as mean ± SEM. The Student’s t-test was used to analyze differences between groups, while one-way ANOVA was used when more than two groups. In all statistical comparisons, *P value < 0.05 were considered as statistically significant (*P value < 0.05, **P < 0.01 and ***P < 0.001).

To identify tumor inhibitory functional fragment, we completed series screenings as described in Fig. 1a for all designed fragments (PJP1-PJP6) from JWA protein non-transmembrane regions (Fig. 1b); in fact, ten polypeptides were initially synthesized and used for screening since both PJP3 and PJP6 contained three residues could be phosphorylated, respectively. The first round screening of intra-tumor injection mouse models (10 mg/kg/d for 5 days/week× 2 weeks) showed PJP3–1, PJP3–3 and PJP5 (Additional files 1: Table S1) had better effects than the others (Fig. 1c); JP1 was previously ruled out because of its weaker tumor-suppressing effect on gastric cancer; as a positive control, although cisplatin showed better effects than all candidate peptides (Additional files 2: Figure. S1a), it was also indicated severe toxicity on mouse and with significant body weight lost since the 3^rd day of treatment (Additional files 2: Figure. S1b). The three candidates were further modified with four glycine and linked with HWGF for next screening by intraperitoneal injection (50 mg/kg/d for 12 days) in the GC cell xenograft mouse models. Gramm-x and Rosetta Dock simulation showed that 19His, 20Trp and 22Phe of the HWGF in JP3 could bind with MMP2 (Additional files 2: Figure. S1c). As shown in Fig. 1d, Additional files 2: Figure. S1d, JP3 indicated the best tumor inhibitory rate among candidates. We further confirmed whether the acetylation modification was necessary for JP3; as shown in Additional files 2: Fig. S1e-f, the inhibitory effect on GC xenograft was slightly better in acetylated JP3 group than in non-acetylated JP3 (Non-Ac-JP3) group. Western blotting showed JP3 treatment down-regulated expressions of MMP2, CD31 and PCNA significantly in GC tumor tissues (Fig. 1e); IHC data showed that JP3 suppressed expressions of CD31 (microvessel density (MVD)) and Ki-67 in GC tumor tissues (Fig. 1f, g). Colony formation assay showed no obvious effects of JP3 on GC cell proliferation in vitro (Additional files 3: Figure. S2). To determine whether JP3 inhibits GC metastasis, we completed a BGC823 GC cell metastatic mouse model. As shown by bioluminescence imaging, the fluorescence intensity of GC cell metastatic foci was obviously reduced in JP3 treated mice compared with the non-sense-HWGF peptide (Ctrl-P-H) group (Fig. 1h, i). HE staining showed that the relative area of metastatic tumor in lung from JP3-treated group was significantly decreased (Fig. 1j, k); IHC data showed that JP3 significantly suppressed expressions of CD31 in GC metastatic tissues (Fig. 1l, m). Collectively, these results suggested that JP3 suppressed GC tumor angiogenesis, proliferation and metastasis in vivo.

To determine how JP3 inhibits angiogenesis in GC, we first observed the pharmacokinetics of JP3 in GC cells. The FITC labelled JP3 (100 μM) was co-cultured with BGC823 (Fig. 2a) and MGC803 (Fig. 2b) GC cells for the indicated time; and JP3 peptide was rapidly entered cells and localized in cytoplasm; the intensity of fluorescence was reached a peak at about 45 min in both cells (Fig. 2c). JP3 was also showed a dose-dependent increase in GC cells and reached a peak at 50 μM (Fig. 2d, e); further observation showed that the intensity of JP3 in cells was decayed over time (Fig. 2f) and the intracellular half-life was about 2 h (Fig. 2g).

We then completed tubular formation assay in HUVECs using the conditional supernatent of JP3 treated GC cells. As shown in Fig. 2h, i, the supernatent from JP3 treated BGC823 and SGC7901 GC cells dose-dependently inhibited tubular formation. To determine whether JP3 also directly inhibited tubular formation, we treated HUVECs by JP3 at similar dose for 24 h, unfortunately no obvious effects were observed (Additional files 4: Figure. S3). Western blotting showed that JP3 treatment down-regulated expressions of SP1 and MMP2 but no obvious effect on JWA in GC cells (Fig. 2j, k). These results suggested that the effect of JP3 on tubular formation may be related to its inhibition of SP1 and MMP2 in GC cells therefore reduced MMP2 content in supernatants of culture medium.

Our previous data have shown that JWA down-regulates SP1 expression by ubiquitination associated mechanism [12]. To confirm whether JP3 down-regulates SP1 through similar mechanism, we determined effects of JP3 on the stability of SP1 in GC cells. As shown in Fig. 3a-b, JP3 accelerated degradation of SP1 when BGC823 cells were treated with 50 μM JP3 for 24 h and then with CHX for indicated time. Ubiquitin modification assay showed ubiquitinated SP1 was increased in BGC23 (Additional files 5: Figure. S4a-b) and SGC7901 (Fig. 3c) cells after JP3 exposure; and MG132 treatment inhibited the downregulation of SP1 and MMP2 (Fig. 3c-d). However, the mRNA level of SP1 was not affected by JP3 treatment in both GC cells (Additional files 6: Figure. S5).

To determine the potential amino acids in SP1 protein that was ubiquitinated by JP3 in GC cells, we first conducted prediction assay via a web tool at https://​www.​phosphosite.​org/​. Data showed there were four potential lysine sites (K610, K624, K685 and K693) in the SP1 involved in JP3 induced ubiquitination. We then constructed Flag-tag SP1 mutants for these candidate amino acids and replaced lysine (K) with arginine (R), respectively (Fig. 3e). In addition, SP1 wild-type and four-K (ALL) mutant together with the four single K mutated plasmids were transfected into BGC823 cells separately for 48 h; and cells were treated with or without CHX for 6 h, respectively. Western blotting results showed that Flag-SP1 protein degradation did not occur in the cells transfected with SP1 (K610R) and SP1 (ALL mutant) plasmids compared with the other three mutants (K624R, K685R and K693R) transfected cells (Fig. 3f). The half-life of SP1 with K610R mutation was arrested compared with the wild-type SP1 (Fig. 3g, h) in cells. This was further confirmed in BGC823 and SGC7901 cells (Fig. 3i-j). Next, the ubiquitination of Flag-SP1 was examined, which was attenuated for the SP1 (K610R) mutant compared with that for SP1 (WT) (Fig. 3k). These results suggested that the K610 of SP1 was responsible for JP3 increased ubiquitination and degradation.

To investigate how JP3 triggers ubiquitination at K610 of SP1, we completed prediction by online tools (http://​ubibrowser.​ncpsb.​org/​;http://​genemania.​org/​). Among the top six of ubiquitin enzymes with the confidence scores higher than 0.7 (Additional files 7: Table S2), TRIM25, WWP1 and MDM2 may more reliable to interact with SP1 (Fig. 4a). We then determined their expressions in transcriptional and translational levels, respectively. As shown in Fig. 4b, only TRIM25 indicated a dose-dependent increase with JP3 treatment in BGC823 cells. The mRNA level of TRIM25 was not affected by JP3 treatment (Additional files 8: Figure. S6). Co-IP assay was performed and a physical interaction between TRIM25 and SP1 was identified in endogenous settings of BGC823 and SGC7901 cells (Fig. 4c-d).

To elucidate the functional roles of TRIM25 in the degradation of SP1 in GC, BGC823 cells were pre-transfected TRIM25 small interfering RNA for 48 h, and subsequently treated with or without JP3. Conditional supernatent was collected to culture HUVECs for additional 4 h. Data showed si-TRIM25 treated supernatent increased the number of tubes; in line with this, the expression of SP1 was increased by si-TRIM25, and JP3 could partly reverse the relationships between SP1 and TRIM25 (Fig. 4e-f). To confirm the regulations of JP3 on TRIM25 triggered SP1 degradation, we completed the ubiquitin modification assays in BGC823 cells. As shown in Fig. 4g, JP3 treatment enhanced ubiquitination of SP1 which was partially inhibited by si-TRIM25; in contrast, HA-TRIM25 increased ubiquitination in SP1 (Fig. 4h). The tubular formation assay further showed HA-TRIM25 suppressed tube numbers in HUVECs (Fig. 4i). These data suggested that JP3 promoted the degradation of SP1 by upregulation of TRIM25 in GC cells, further resulted in the reduction of tubular formation of HUVECs induced by supernatent of GC cells.

To investigate how JP3 modulates TRIM25, we determined effects of JP3 on the stability of TRIM25 in GC cells. As shown in Fig. 5a-b, JP3 slowed down the degradation of TRIM25 when BGC823 cells were treated with 50 μM JP3 for 12 h and then with CHX for indicated time. Ubiquitin modification assay showed ubiquitinated TRIM25 was decreased in BGC23 cells after JP3 exposure (Fig. 5c).

TRIM25 has been reported to auto-ubiquitylate itself. MAP3K13, a part of a non-canonical MAPK signaling pathway (PubMed: 28111074), phosphorylates TRIM25 at S12 and stabilizes it through decreasing its polyubiquitination [18]. As shown in Additional files 9: Figure. S7a, non-phosphorylated JP3 did not show obvious effects on SP1 and MMP2. Compared with JP3, the supernatant of Non(p)-JP3 treated GC cells was also no inhibitory role on tubular formation in HUVECs (Fig. 5d, Additional files 9: Figure. S7b). These results showed that the phosphorylation of JP3 was required for the inhibition of angiogenesis through increasing stability of TRIM25 by reducing its ubiquitination.

Our previous studies have proven that JWA affects tumor cells migration by activating the MAPK signaling pathway [19]. Given that TRIM25 was a specific E3 ubiquitin ligase on SP1 and was stabilized by JP3, we speculated that the activation of TRIM25 by JP3 may have two ways, one was MEK/MAPK, and the another one was a direct activation. JP3 binding capacity with MEK1/2 and TRIM25 were then analyzed based on predicted complex structures (Additional files 10: Figure. S8a, b). As shown in Fig. 5e, the main interaction types between amino acids of JP3 and TRIM25 were Hbond and PiPistack (Additional files 11: Table S3). However, VDW (van der Waals force) is the main interaction type between JP3 and MEK1/2. These results indicate that JP3-TRIM25 complex probably more stable than JP3-MEK1/2 complex. To determine whether JP3 directly interacts and activates TRIM25, we completed protein docking analysis using the PDB2PQR server. As shown in Fig. 5f, both the T9 site of JP3 and the S12 site of TRIM25 were distributed in the protein interaction region to form an ATP binding pocket, in which JP3 may transfer its phosphate group to and catalytic phosphorylation of S12 in TRIM25. This interaction was reversely confirmed by a non-phosphorylated JP3 (Non(p)-JP3). Data showed the non-phosphorylated T9 in JP3 had more positive potential and can’t bind with the S12 in TRIM25 (Additional files 10: Figure. S8c, Additional files 12: Table S4). We then constructed mutant and replaced S12 by alanine in TRIM25 to verify whether JP3 stabilizes TRIM25 protein expression by phosphorylating S12 sites. The TRIM25 wild-type and mutant plasmid were transfected into BGC823 cells for 48 h, followed by JP3 treatment for 12 h, with CHX for 0, 4, 8 or 12 h, respectively. The half-life of TRIM25 with S12A mutation was shortened compared with the wild-type TRIM25 (Fig. 5g). Finally, western blotting showed JP3 treatment up-regulated expression of TRIM25 and down-regulated expression of SP1 significantly in GC tumor tissues compared to the Ctrl-P-H treatment (Fig. 5h). Therefore, it was believed that JP3 could facilitate the interaction with TRIM25 and increased stability of TRIM25 by reducing polyubiquitination mediated degradation of itself. A stabilized and activated TRIM25 was necessary to down-regulate SP1 mediated expression of MMP2 in GC cells, and further inhibited tubular formation in HUVECs.

To determine the significance of expressions both TRIM25 and SP1 in GC, using the online bioinformatics tool Kaplan-Meier Plotter (http://​www.​oncolnc.​org/​), we found that patients with a low level of TRIM25 in GC tissues suffer a poor overall survival (OS) than a high level of TRIM25 ones (Fig. 6a). When combining TRIM25 and SP1 or MMP2 together as a new variable, patients with GC were divided into three subgroups: high level of TRIM25 and low level of SP1 or MMP2; low level of TRIM25 and high level of SP1 or MMP2; and both low/high. Kaplan-Meier curves demonstrated that the subgroup of high level of TRIM25 and low level of SP1 or MMP2 was the most favorable to the overall survival (log-rank test, P < 0.0001; Fig. 6b-c). To confirm this, we completed assay on GC tissue microarray. Data showed that expression of TRIM25 was lower in GC tissues than in adjacent normal tissues (P = 0.0029, Fig. 6d); in contrast, expression of SP1 was higher in GC tissues than that in adjacent normal tissues (P < 0.0001, Fig. 6e); and the SP1 protein level was inversely correlated with the TRIM25 in primary tumors from GC patients (r = − 0.255, P = 0.0153, Fig. 6f, Additional files 13: Table S5). Further, the OS data showed the patients with low TRIM25 or high SP1 expression alone in GC tumor tissues had poor OS than high TRIM25 or low SP1 ones (Fig. 6g-h). Taken together, these data suggest that activation of the TRIM25-SP1 signaling could be benefit on OS in GC patients.