IGF2BP3 functions as a potential oncogene and is a crucial target of miR-34a in gastric carcinogenesis

Human GC cell lines (MKN1, MKN7, MKN28, MKN45, SNU1, SNU16, AGS, KatoIII, NCI-N87, MGC-803, SGC-7901, TMK-1) and immortalized gastric epithelial cells (GES-1 and HFE-145) were described previously [18]. Cells were cultured at 37 °C in humidified air atmosphere containing 5% CO2 in RPMI 1640 (GIBCO, Grand Island, NY) medium supplemented with 10% fetal bovine serum (GIBCO). A cohort of 247 formalin-fixed paraffin embedded tissues of GCs diagnosed between 1999 and 2006 in the Prince of Wales Hospital, Hong Kong was retrieved. Ethical approval was obtained from the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee.

Cultured cells were harvested for extracting total RNA with TRIzol reagent (Invitrogen, Carlsbad, CA). cDNA synthesis was performed with a High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems, Carlsbad, CA). The variations of mRNA expression of related genes were quantified by qRT-PCR and primers were listed as following: IGF2BP3 (sense: AGT TGT TGT CCC TCG TGA CC; anti-sense: GTC CAC TTT GCA GAG CCT TC); B2M (sense: ACT CTC TCT TTC TGG CCT GG; anti-sense: ATG TCG GAT GGA TGA AAC CC). The protocol of qRT-PCR was described in a previous study [19]. For microRNA expression detection, miR-34a expression in GC was detected by Taqman miRNA assays, and they were used to quantify the levels of mature miR-34a (Assay ID: #4427975, Life Technologies). The relative expression level of miR-34a was normalized by RNU6B expression (Assay ID: #001093). 7500 Fast Real-Time System (Applied Biosystems) was used for the qPCR reaction. And the reaction system was incubated at 95 °C for 30 s, followed by 40 cycles of 95 °C for 8 s and 60 °C for 30 s.

The protein extraction and western blot analysis protocol were described in our previous study [20]. The primary antibodies detected IGF2BP3 (#07-104) was from Merck Millipore. Other primary antibodies were from Cell Signaling (Danvers, MA) including p21 (#2946), p27 (#2552), p-Rb (Ser807/811) (#9308), cleaved-PARP (Asp214) (#9541), Cyclin D3 (#2936), CDK4 (# 12790), CDK6 (#3136) and GAPDH (#2118). The secondary antibodies were anti-Mouse IgG-HRP (Dako, 00049039, 1:30000) and anti-Rabbit IgG-HRP (Dako, 00028856, 1:10000).

Immunohistochemistry was to conduct tissue microarray within a 4 μm-thick section of each clinical sample using Ventana NexES automated Stainer (Ventana Corporation). All sections were performed microwaving in EDTA antigen retrieval buffer after de-waxing in xylene and graded ethanol. The IGF2BP3 primary antibody (1:100, 07–104) was from Merck Millipore. The cytoplasmic expression of IGF2BP3 was evaluated according to the proportion of tumor cells with intensity of cytoplasmic staining [20].

The miRNA precursor miR-34a (PM11030) and scramble control (AM17110) were commercially available from Life Technologies. siIGF2BP3-1 (SI03230759) and siIGF2BP3-2 (SI04234167) were obtained from Qiagen (Valencia, CA). Lipofectamine 2000 Transfection Reagent (Invitrogen) was used for all transfection assays. The cell proliferation experiments, colony formation assays in monolayer, cell invasion assays, flow cytometry analysis for cell cycle distribution have been described in our previous work [21]. The experiments were repeated in triplicate to obtain standard deviations.

The putative miR-34a binding site in IGF2BP3 3'UTR of was subcloned into pMIR-REPORT Vector (Ambion). The oligonucleotides that encompass the miR-34a recognition site and the oligonucleotides which contain the mutated binding site were listed in Additional file 1: Table S1. Prior to digestion and subcloning, oligonucleotides were annealed in 30 mmol/L HEPES buffer that contains 100 nmol/L potassium acetate and 2 mmol/L magnesium acetate [22]. The firefly luciferase constructs were co-transfected with Renilla luciferase vector control into MGC-803 cells. Dual luciferase reporter assays (Promega, Madison, WI) were performed 24 h after transfection.

AGS, MKN1, NCI-N87 and MGC-803 were treated with demethyltransferase inhibitor (5-Aza, Sigma, St Louis, MO) and histone-deacetylase inhibitor trichostatin A (TSA, Sigma) [23]. Cells were incubated with 10 μM 5-Aza for 72 h in 5-Aza group, while in TSA group, cells were treated with 100 nM TSA for 24 h. As for the combination treatment group, cells were treated with 5-Aza for 96 h and 100 nM TSA was added into the culture medium in the last 24 h. Equal amount of vehicle DMSO (Sigma) was used in negative control groups.

MGC-803 cells (10⁷ cells suspended in 0.1 ml PBS) were transfected with miR-34a or scramble control then were injected subcutaneously into the left and right dorsal flank of 4-week old Balb/c nude mice respectively. Diameters of tumors were measured and documented each 5 days with a total of 25 days. Tumor volume (mm³) was accessed by measuring the longest and shortest diameter of the tumor and calculating as follows: volume = (shortest diameter) ² × (longest diameter) × 0.5. All animal handling and experimental procedures were approved by Department of Health in Hong Kong.

For the rescue experiments, AGS and MKN28 cells were first transfected with miR-34a precursor or negative control respectively. After 24 h incubation, IGF2BP3 expression plasmid (#19879, Addgene) together with empty vector (pcDNA3.1, Life Technologies) were transfected to the cells using FuGENE HD Transfection Reagent (Roche, Nutley, NJ). After another 24 h, cells were harvested for functional study (MTT proliferation, monolayer colony formation, and cell invasion assays). The in vivo rescue experiments were performed using MGC-803 cells [24].

The Student T test was used to compare the differences in functional differences between siIGF2BP3 and siScramble control transfected cells. It is also used to compare the biological behavior between miR-34a transfected cells and scramble miRNA transfectant counterparts. Nonparametric Pearson Chi-Square test was used to evaluate the correlation between IGF2BP3 expression and selected clinicopathologic parameters. Kaplan-Meier method was used to estimate the survival rate for each parameter and the equivalences of the survival curve were examined by log-rank statistics. For those parameters were found statistically significant in the univariate survival analysis (P < 0.05), the Cox proportional hazards model was employed to further evaluate them for multivariate survival analysis. All statistical analysis was performed by SPSS software (version 22.0; SPSS Inc). A two-tailed P-value of less than 0.05 was considered statistically significant and the P-value less than 0.001 was considered highly significant.

By gene expression microarray analysis in nine GC cell lines (Additional file 2: Table S2), IGF2BP3 was found to be the most up-regulated gene in the top ten list (Fig. 1a). Consistently, both mRNA and protein expression of IGF2BP3 were up-regulated in most of the GC cell lines comparing with immortalized gastric epithelium cell line (GES-1) (Fig. 1b). In primary samples, reports from GENT dataset manifested that IGF2BP3 showed a dramatically overexpression in 351 GC tissues when compared with the normal gastric tissues (P < 0.01, Fig. 1c) [25]. In addition, according to a published dataset NCBI/GEO/GSE63089 [26], IGF2BP3 was demonstrated to be highly expressed in GC samples compared with adjacent non-tumorous tissues (n = 45, P < 0.001, Fig. 1d). Moreover, based on TCGA cohort, IGF2BP3 showed significantly up-regulated in both intestinal- and diffuse- types of GC when compared with normal control (P < 0.001, Fig. 1e). In particular, when it came to the molecular characterization proposed by TCGA [5], high level of IGF2BP3 was observed frequently in the subtype of chromosomal instability (CIN) and EBV positive GC, while its level was relative lower in genomically stable (GS) subtype (Fig. 1f). For deep investigation of the regulation mechanisms of IGF2BP3 in GC, the genomic alteration of IGF2BP3 in TCGA cohort was analyzed by cBioPortal (www.​cbioportal.​org/​) [27, 28], including copy number changes, somatic mutation and mRNA upregulation. It was found that 18% cases (47/258) have at least one alteration of IGF2BP3. And the rate of mRNA upregulation counted for 14% cases (Fig. 1g). Although the copy number gain showed positive correction with IGFBP3 mRNA expression (P < 0.05, Additional file 3: Figure S1), its mRNA upregulated in all GC samples can not merely be explained by copy number change.

From Kaplan Meier plotter (www.​kmplot.​com), overexpression of IGF2BP3 was correlated with poor survival of both overall and first progression GC cases (P = 0.018, overall survival; P < 0.001, first progression survival; Fig. 2a and Additional file 4: Table S3) [29]. In TCGA cohort, although the P-value was not significant, there was still a trend that aberrant richness of IGF2BP3 was associated with unfavorable outcome (P = 0.141, overall survival; P = 0.078, recurrence free survival; Fig. 2b). Subsequently, we conducted immunohistochemistry to elucidate the protein expression of IGF2BP3 in GC tissue microarray. In normal gastric epitheliums, there was a negative expression of IGF2BP3. Significantly, a strong cytoplasmic staining was observed in both intestinal- and disuse-types of GC cells (left panel, Fig. 2c), and this strong staining predicted a poor disease specific survival if we classified the samples as two subgroups (negative/week and moderate/strong subgroups, P = 0.012, right panel, Fig. 2c). The correlations between IGF2BP3 and clinicopathologic parameters in 247 GC samples were summarized in Additional file 5: Table S4. Specifically, IGF2BP3 upregulation was only correlated with N stage (P = 0.025). In univariate analysis, age, diffuse type, high grade, advanced stage, lymph node metastasis and overexpression of IGF2BP3 had a statistical association with poor survival. However, by multivariate Cox regression analysis, only age and advanced stage were two independent factors (Additional file 6: Table S5).

From the results of GSEA (Gene Set Enrichment Analysis) in a published GC cohort NCBI/GEO/GSE62254 (n = 300) [30], IGF2BP3 upregulation was found to have a positive correlation with expression of a common cancer gene set which was defined by a group of Singaporean researchers (P < 0.001) [31]. Furthermore, IGF2BP3 was also significantly correlated with cell growth (P < 0.001) [32] and cell cycle progression (P = 0.004, Fig. 3a) [33, 34]. Since IGF2BP3 was expressed abundantly in MKN28 and SGC-7901 cell lines, we conducted siRNA-mediated knockdown experiments to reveal its function in these two cell lines. After treatment of two siIGF2BP3s, both mRNA and protein expressions of IGF2BP3 were declined in cells (Fig. 3b). Silencing IGF2BP3 slowed down the rate of cell growth, as indicated in a 4-day MTT assays (P < 0.001, Fig. 3c). Monolayer colony formation ability was inhibited after IGF2BP3 knockdown compared with scramble control (P < 0.001, Fig. 3d). Besides, cell invasive ability was markedly suppressed by siIGF2BP3 transfection (P < 0.001, Fig. 3e). As proliferation-inhibition phenotypes were observed in siIGF2BP3 groups, the associated cell cycle regulators and apoptosis markers were analyzed by Western blot. The phosphorylated Rb (p-Rb), CDK4, CDK6 expression were decreased but p21 and p27 were uniformly up-regulated in IGF2BP3 knockdown cells, supporting the G0/G1-phase cell cycle arrest. Together, knockdown of IGF2BP3 did not induce obvious late apoptosis in MKN28 and SGC-7901 cell (Fig. 3f). Then the transfectants for cell-cycle parameters were analyzed by flow cytometry to validate the Western blot results. As shown in Fig. 3g, MKN28 cells with siIGF2BP3 knockdown groups contained a higher percentage of G0/G1 phase cells (52.9 and 54.7% respectively) compared with siScramble control counterparts (46.7%). The cell population of G0/G1 phase in IGF2BP3-knockdown SGC-7901 cells were increased from siScramble control (56.8%) to 62.2 and 77.4% respectively (Fig. 3g).

In IGF2BP3 3'UTR, the binding site for miR-34a was predicted by miRNA.org (http://​www.​microrna.​org/​) with the mirSVR scores of −1.098 (Fig. 4a). miR-34a ranks the top miRNA list that potentially targets IGF2BP3. Both the mRNA and protein expression of IGF2BP3 were found decreased in AGS and MKN28 cells after ectopic expression of miR-34a (Fig. 4b and c). To test whether IGF2BP3 was a direct target of miR-34a, we conducted the luciferase experiment. As shown in Fig. 4d, miR-34a inhibited the luciferase activity of construct encompassing the wild type binding site in IGF2BP3 3'UTR, but it had no effect on the construct containing mutated sequence of the binding site (Fig. 4d). This result supported that miR-34a regulated IGF2BP3 expression by directly binding with its 3'UTR. miR-34a showed uniformly down-regulated in all 11 GC cell lines compared with immortalized gastric epithelium cell line HFE-145 (Fig. 4e). However, the expression of miR-34a was not restored after treating the GC cells with 5-Aza or TSA (Fig. 4f), suggesting that epigenetic modification, such as promoter methylation or histone deacetylation, may not play an important role in the regulation of miR-34a expression.

To investigate the biological role of miR-34a in GC, ectopic expression of miR-34a precursor in GC cells (MKN28 and SGC-7901) was performed. MTT proliferation assays indicated that overexpression of miR-34a reduced cell proliferation (Fig. 5a) and this inhibitory effect was subsequently validated via monolayer colony formation study (Fig. 5b), with less and smaller colonies found in miR-34a over-expressed group. Moreover, a descending cell invasion was found in the cells with ectopic miR-34a transfection compared with negative control (Fig. 5c). Due to the proliferation inhibition, cell cycle analysis by flow cytometry was conducted. As shown in Fig. 5d, miR-34a increased G0/G1 proportion from 71.9 to 77.6% in MKN28 cells. In SGC-7901, miR-34a overexpression resulted in 67.8% proportion of G0/G1 cells when compared to 49.5% in negative control group (Fig. 5d). The results were further verified by Western blot of cell cycle regulators and apoptosis markers. The expression of p-Rb and Cyclin D3 showed decreased but p21 and p27 displayed up-regulated after overexpression of miR-34a in AGS, MKN28 and SGC-7901 cell. In addition, miR-34a induced late apoptosis, represented by the activation of cleaved-PARP in all the three GC cell lines (Fig. 5e). The effect of miR-34a on in vivo tumor growth was also investigated. MGC-803 cells transfected with negative control and miR-34a were subcutaneously injected into 4-week-old nude mice. Tumors grew slower and showed smaller size in miR-34a group than those in the negative control group (P < 0.001, Fig. 5f).

Apart from the functional study, the expression of miR-34a in primary samples and its clinical correlation were analyzed. In two published cohorts, TCGA (n = 285) as well as a Japanese cohort (E-TABM-341, n = 184) [35], a considerably lower expression of IGF2BP3 was observed in diffuse type in contrast with intestinal type GC (P = 0.006 and P = 0.008 respectively, Fig. 6a). For molecular characterization of GC, miR-34a expression showed a lower expression in in GS subtype (P < 0.001, Fig. 6b), suggesting downregulation of miR-34a might strongly be associated with metastasis. Furthermore, downregulation of miR-34a in GC samples predicted poorer outcome in overall survival (P = 0.024) [36], but for recurrence free survival, decreased miR-34a only showed a trend to correlate with poor prognosis (P = 0.098, Fig. 6c and Additional file 7: Table S6). To further investigate if miR-34a is a crucial regulator of IGF2BP3 in GC, the expression correlation of miR-34a and IGF2BP3 was analyzed in TCGA cohort. As shown in Fig. 6d, there was a negative association between their expression in overall cases (r = −0.132, n = 277, P = 0.028). Remarkably, when the cohort was re-analyzed according to Lauren classification, a more stringent reverse correlation between miR-34a and IGF2BP3 was observed in intestinal type GC (r = −0.249, P < 0.001). These results implicated that IGF2BP3 was modulated by miR-34a in primary GC, especially in intestinal type GC.

Finally, we investigated if the IGF2BP3 was a real functional target of miR-34a in GC by rescue experiments. Re-expression efficiency of IGF2BP3 was first verified by Western blot analysis (Fig. 7a). Both MTT proliferation and colony formation assays explicated growth-inhibition effect of miR-34a was partly alleviated by IGF2BP3 re-overexpression (Fig. 7b and c). In addition, re-expression of IGF2BP3 partly restored the impaired cell invasion ability in miR-34a treated GC cells (Fig. 7d). Importantly, in vivo experiments demonstrated that re-expressed IGF2BP3 partly abolished the anti-carcinogenic function of miR-34a, further emphasizing that IGF2BP3 was the bona fide functional target of miR-34a in GC (Fig. 7e).