RPRD1B is overexpressed in GC and metastatic lymph nodes
High-throughput RNA-seq was performed in 4 patients in the Chinese gastric cancer cohort with primary tumors, metastatic lymph nodes and corresponding nontumor epithelial tissue (Supplementary RNA-seq Data 1). Overexpressed genes in lymph node metastasis compared with matched normal/primary tumors were analysed (fold changes of RSEM counts > 1), and 2236 target genes were selected for subsequent analysis. Compared with genes that were overexpressed and associated with shorter survival in the TCGA STAD cohort (450 patients, 415 for tumor tissues and 35 for normal tissues; limma-eBayes: log2(RSEM counts + 1) fold changes > 0.5 and P < 0.05), 8 target genes (RPRD1B, MAP4K4, MCM2, TOPBP1, FRMD8, KBTBD2, ADAM10 and CXCR4) were screened from these 2236 genes (Fig. 1A). As a transcriptional cofactor, RPRD1B was upregulated in lymph node metastatic lesions and further characterized. The expression status of RPRD1B was then confirmed using qRT–PCR in 36 patients with GC. RPRD1B was upregulated in 20/36 (55.6%) GC primary tumor tissues (P = 0.0061, paired Student’s t test) and 9/12 (75%) metastatic lymph nodes (P < 0.0001, paired Student’s t test) compared with the corresponding nontumor tissues (Fig. 1B).
For the subsequent survival analysis, IHC staining was performed on a tissue microarray (Fig. 1C). Overexpression of RPRD1B (IHC score ≥4) was detected in 78/191 (40.8%) informative GC primary tumor tissues (P < 0.0001, paired Student’s t test) and 44/93 (47.3%) metastatic lymph nodes (P < 0.0001, paired Student’s t test) (Fig. 1D). Furthermore, RPRD1B overexpression was significantly associated with the tumor invasion depth (P = 0.019), lymph node invasion (P = 0.014), and vascular invasion (P = 0.000) (Fig. 1E) (Supplementary Table 6). The Kaplan–Meier analysis revealed that RPRD1B overexpression in tumors (P = 0.037) and lymph nodes (P = 0.001) was significantly associated with shorter overall survival (Fig. 1F).
The Mettl3-induced m6A modification is involved in the upregulation of RPRD1B
The mechanism underlying the aberrant expression of RPRD1B is unknown. Previous studies have reported that the m6A modification modulates mRNA stability and plays a critical role in GC metastasis[21]. m6A RNA immunoprecipitation (RIP) revealed that the m6A modification of RPRD1B was significantly more enriched in GC cells than in normal GES1 cells (Fig. 1G). As Mettl3 is the key m6A methyltransferase, we evaluated Mettl3 expression and the correlation between Mettl3 and RPRD1B expression in TCGA database. Notably, Mettl3 was significantly upregulated in GC tissues, and Mettl3 expression was positively correlated with RPRD1B expression (R=0.49, P < 0.05) (Fig. 1H). Additionally, Mettl3 depletion apparently reduced RPRD1B mRNA and protein levels. In contrast, Mettl3 overexpression increased the RPRD1B mRNA and protein levels (Fig. 1I, J). MeRIP assays also revealed that Mettl3 silencing reduced the m6A modification of the RPRD1B mRNA in AGS cells (Fig. 1K). Furthermore, we examined the stability of the RPRD1B mRNA upon Mettl3 overexpression and silencing in AGS cells. Actinomycin D was used to block de novo mRNA synthesis in cells, and the stability of the RPRD1B mRNA decreased upon Mettl3 depletion but increased with Mettl3 overexpression (Fig. 1L). Thus, the m6A writer Mettl3 upregulates RPRD1B mRNA expression by promoting the m6A modification in GC cells.
RPRD1B promotes metastasis of GC cells in vivo and in vitro
Since the clinicopathological analysis revealed that RPRD1B overexpression was associated with vascular invasion and lymph node metastasis, we next investigated the effect of RPRD1B on tumor cell migration and invasion. Overexpression of RPRD1B was detected in 7 GC cell lines compared with the immortalized gastric cell line GES1 using western blotting and qRT–PCR (Fig. 2A). RPRD1B was stably overexpressed in HGC27 and SGC7901 cells and ablated with two short hairpin RNAs (shRNAs) in AGS and BGC823 cells (Fig. 2B). In vitro, RPRD1B overexpression potentiated the ability of HGC27 and SGC7901 cells to migrate and invade (Fig. 2C). Conversely, RPRD1B depletion in AGS and BGC823 cells resulted in the opposite effects (Fig. 2D). The wound-healing assay showed faster closure of the scratched “wound” by RPRD1B-overexpressing cells but slower closure by RPRD1B-silenced cells than in control cells (Fig. 2E, F).
An animal model of lymph node metastasis was established to test the metastasis-promoting effect of RPRD1B in vivo. Swollen popliteal lymph nodes were observed in all seven mice injected with RPRD1B-overexpressing HGC27 cells, and lymph node metastasis was confirmed by H&E staining in 5/7 mice (Fig. 2G). Only 1/7 of mice injected with control cells displayed tumor metastasis. In contrast, after RPRD1B was silenced in AGS cells, none of the mice exhibited popliteal lymph node metastasis compared with 3/7 mice exhibiting metastasis after the injection of control cells (Fig. 2H).
RPRD1B facilitates fatty acid metabolism by activating the c-Jun/c-Fos/SREBP1 axis
RNA sequencing analysis was applied between RPRD1B- and LacZ-transfected HGC27 cells to identify the differentially expressed genes regulated by RPRD1B. Ninety differentially expressed genes were identified (qval < 0.05, Supplementary RNA-seq Data 2). KEGG pathway analysis showed enrichment in the Wnt/PCP and mitogen-activated protein kinase (MAPK) signaling pathways. Functional annotation revealed that RPRD1B was involved in fatty acid (FA) metabolism and NAFLD, serving as a transcription cofactor binding to DNA fragments (Fig. 3A). GSEA indicated the significant enrichment of AP1 and SREBP1 transcriptional activity in RPRD1B-overexpressing cells (Fig. 3B).
Quantitative RT–PCR analysis was conducted to confirm the differentially expressed genes in RNA sequencing analysis. As the intersection of Wnt and MAPK pathway, the transcription factors c-Jun and c-Fos were most significantly upregulated. In addition to the transcription factors c-Jun and c-Fos, the fatty acid metabolism-related genes SREBP1, FASN, ACSS2 and FABP3 were upregulated in RPRD1B-overexpressing cells. An opposite expression pattern of these genes was observed in RPRD1B-silenced cells (Fig. 3C). As SREBP1 has a known role in regulating FA synthesis[22, 23] and was reported to be the target of AP1 members such as c-Jun and JNK2[24, 25], we hypothesized that RPRD1B promotes fatty acid metabolism and lymph node metastasis by activating c-Jun/c-Fos/SREBP1 axis and increasing FA uptake and synthesis (Fig. 3D). Cholesterol and triglyceride levels were increased in RPRD1B-overexpressing cells but decreased in RPRD1B knockdown cells (Fig. 3E, F). Oil Red O staining was performed to verify the presence of more droplets in RPRD1B-overexpressing cells and fewer droplets in RPRD1B-knockdown cells (Fig. 3G). Positive correlations between c-Jun, c-Fos, SREBP1, FASN, ACSS2, FABP3 and RPRD1B expression at the protein level were examined using western blot analysis (Fig. 3H).
RPRD1B occupies the promoter of c-Jun and c-Fos
We further examined the mechanism by which RPRD1B upregulates the expression of c-Jun and c-Fos using bioinformatics software (UCSC, PROMO and TFSEARCH) to analyse a 2 kb region upstream of the transcription start sites of c-Jun and c-Fos. All potential binding sites in the c-Jun and c-Fos promoters were predicted. A luciferase activity assay was then performed with four fragments of c-Jun or c-Fos promoter regions that encompass the predicted RPRD1B binding sites (F1: -2200 bp to +100 bp; F2: -1700 bp to +100 bp; F3: -1200 bp to +100 bp; F4: -700 bp to +100 bp). RPRD1B could significantly increase the transcriptional activity of both c-Jun and c-Fos in cells transfected with the F1, F2 and F3 fragments (P < 0.01) (Fig. 4A). A ChIP-PCR assay was then completed using four pairs of primers according to the predicted binding sites within the c-Jun promoter (P1: -2,157 bp to -2,002 bp, P2: -1,626 bp to -1,471 bp, P3: -1194 bp to -1049 bp, and P4: -690 bp to -535 bp) and c-Fos promoter (P1: -2,135 bp to -2,010 bp, P2: -1,622 bp to -1,517 bp, P3: -1160 bp to -1025 bp and P4: -633-bp to 508 bp). The P2 and P3 sequences in the c-Jun promoter and the P1 and P3 sequences in the c-Fos promoter were enriched in RPRD1B-ChIPed DNA fragments but not in IgG-ChIPed controls (P < 0.001) (Fig. 4B). Electrophoretic mobility shift assays (EMSAs) were performed to further confirm that more probes of c-Jun and c-Fos were pulled down in RPRD1B-overexpressing cells (Fig. 4C).
We inhibited the expression of c-Jun/c-Fos by applying SR11302 (AP1 inhibitor) to confirm that RPRD1B promotes tumor metastasis by activating the c-Jun/c-Fos/SREBP1 pathway. SR11302 effectively inhibited the expression of SREBP1, FASN, ACSS2 and FABP3 in a dose-dependent manner (Fig. 4D). Cell migration assays indicated that SR11302 significantly decreased cell motility in RPRD1B-overexpressing HGC27 cells (Fig. 4E).
The lncRNA NEAT1 is transcriptionally upregulated by c-Jun/c-Fos/AP1
According to previous studies, lncRNAs are key molecules that regulate transcriptional and posttranscriptional processes[26, 27]. In the present study, we noticed that the lncRNA NEAT1 was significantly upregulated in the RNA-seq data conducting between LacZ- and RPRD1B-overexpressing cells (Fig. 5A). In clinical samples, NEAT1 was frequently upregulated and significantly correlated with shorter survival of patients with GC in our cohort and TCGA database (Fig. 5B). However, the mechanism of NEAT1 upregulation in RPRD1B-overexpressing GC cells is unknown. The potential transcription factor and promoter region involved in NEAT1 transcription were analyzed by ALGGEN and Promoter 2.0 to further identify the mechanism of NEAT1 upregulation. Interestingly, c-Jun/c-Fos/AP1 were predicted to be transcription factors for NEAT1 (Figure 5C). We validated this prediction by performing a luciferase activity assay with the promoter DNA fragments of NEAT1 containing binding sites for c-Jun and c-Fos. NEAT1 transcriptional activity was significantly upregulated following c-Jun and c-Fos transfection. The effect was significantly diminished by SR11302 (Fig. 5D). RNA-FISH and qRT-PCR confirmed the upregulation of NEAT1 in RPRD1B-upregulated GC cells, which was decreased significantly by SR11302 (Fig. 5E, F). The opposite effect was detected in RPRD1B-silenced GC cells (Fig. 5G, H). These data suggested that RPRD1B transcriptionally upregulates NEAT1 in a c-Jun/c-Fos/AP1-dependent manner.
NEAT1 increases RPRD1B mRNA stability via the hnRNPA2B1-mediated m6A modification
Nuclear speckles and paraspeckles have been reported to be important locations involved in posttranscriptional processes, such as splicing of pre-mRNA and m6A methylation of mRNA[28]. As an essential component of nuclear paraspeckles, NEAT1 was reported to contribute to a cancer-favorable transcriptional program[29]. We explored whether NEAT1 is a transcriptional or posttranscriptional modifier of RPRD1B by identifying the proteins interacting with NEAT1 through capture hybridization analysis of RNA targets and mass spectrometry (CHART/MS) performed with a probe against NEAT1, as described in a recent study[30]. Interestingly, RNA-binding proteins involved in m6A modification, such as hnRNPA2B1, YTHDF1 and hnRNPC, were precipitated in protein complexes pulled down by NEAT1 (Fig. 6A). Gene Ontology (GO) analysis of NEAT1-interacting genes showed enrichment in several biological processes, such as mRNA splicing, translation initiation and mRNA stability (Fig. 6B). As one of the predicted proteins, the interaction of NEAT1 and hnRNPA2B1 was confirmed by RNA pull-down and RIP experiments (Fig. 6C). Notably, hnRNPA2B1, a well-known m6A reader, was reported to affect the posttranscriptional modification of m6A-methylated RNA in the nucleus, such as splicing, stability and translation. However, researchers had not determined whether NEAT1/hnRNPA2B1 participates in the posttranscriptional regulation of m6A-methylated RPRD1B. Positive correlations between RPRD1B and hnRNPA2B1 expression at the mRNA level were confirmed in TCGA database (R = 0.54, P = 0.00) (Fig. 6D). RIP assays confirmed the direct interaction between hnRNPA2B1 and the RPRD1B mRNA in GC cells (Fig. 6E). Silencing of hnRNPA2B1 in RPRD1B-overexpressing HGC27 cells significantly decreased the RPRD1B mRNA and protein levels (Fig. 6F). MeRIP assays revealed that silencing hnRNPA2B1 reduced m6A modification of RPRD1B mRNA in HGC27 cells (Fig. 6G). After actinomycin D treatment, the half-life of the RPRD1B mRNA was significantly reduced due to hnRNPA2B1 depletion (Fig. 6H). In addition, the interaction between hnRNPA2B1 and the RPRD1B mRNA was impaired after NEAT1 suppression (Fig. 6I). Moreover, RPRD1B was downregulated after NEAT1 silencing in HGC27 and SGC7901 cells (Fig. 6J). Taken together, our data suggest that NEAT1 facilitates hnRNPA2B1 binding to the RPRD1B mRNA and increases its stability in an m6A-dependent manner.
NEAT1 reduces degradation of RPRD1B by inhibiting TRIM25-mediated ubiquitination
Emerging studies reported that RNA binding protein (RBP) involved R-loops play important roles in genome stability and gene expression[31, 32]. RPRD1B was predicted to function as a new potential RBP for involving in R-loops formation[33]. Interestingly, we found that RPRD1B was present in the protein complex precipitated by the NEAT1 probe, suggesting a direct interaction between NEAT1 and the RPRD1B protein. Immunofluorescence (IF) staining for RPRD1B and RNA-FISH of NEAT1 were performed to verify their colocalization in RPRD1B-transfected HGC27 and SGC7901 cells (Fig. 7A). Their interaction was then confirmed by performing RNA pull-down and RIP experiments (Fig. 7B, C). Ubiquitin assays were performed to explore the mechanism between NEAT1 and RPRD1B protein. Cells with NEAT1 silencing were then treated with the protein synthesis inhibitor cycloheximide, and a noticeably shorter half-life of RPRD1B was detected in HGC27 and SGC7901 cells than in control cells. In contrast, NEAT1 overexpression extended the half-life of RPRD1B in AGS and BGC823 cells (Fig. 7D). Based on these results, the NEAT1-mediated ubiquitin–proteasome pathway may be involved in the stabilization of the RPRD1B protein. The proteasome inhibitor MG132 was used to verify whether NEAT1 inhibited the degradation of RPRD1B. Increased levels of polyubiquitinated RPRD1B were observed in NEAT1-silenced cells compared with control cells. In contrast, reduced accumulation of polyubiquitinated RPRD1B was detected in NEAT1-overexpressing cells compared with control cells (Fig. 7E). We surveyed a protein–protein interaction database, Biogrid (www.biogrid.org), to identify the mechanism of RPRD1B ubiquitination and identified 113 RPRD1B interactors, including the E3 ubiquitin ligase TRIM25. As an RNA-binding protein, TRIM25 also interacted with NEAT1 in CHART-MS (Fig. 7F). IP assays confirmed the interaction among RPRD1B, NEAT1 and TRIM25 (Fig. 7G), verifying our hypothesis. NEAT1-silencing mediated RPRD1B ubiquitination was rescued by TRIM25 knockdown in GC cells. (Fig. 7H). RPRD1B degradation was rescued by TRIM25 knockdown in NEAT1-silenced GC cells (Fig. 7I). According to these results, we concluded that NEAT1 decreases RPRD1B ubiquitination by blocking its interaction with TRIM25.
The RPRD1B/c-Jun/c-Fos/SREBP1 axis correlates with the lymph node metastasis of GC
Cells were treated with siNEAT1 and/or the AP1 inhibitor (SR11302) to further identify the role of the c-Jun/c-Fos/SREBP1 axis and NEAT1-mediated feedback loop in lymph node metastasis promoted by RPRD1B. Western blot analysis showed that siNEAT1 and/or SR11302 significantly decreased the levels of c-Jun, c-Fos, and fatty acid uptake and synthesis markers (SREBP1, FASN, ACSS2 and FABP3) in RPRD1B-overexpressing HGC27 and SGC7901 cells (Fig. 8A). As expected, migration and invasion assays revealed that both SR11302 and siNEAT1 significantly inhibited the migration and invasion of RPRD1B-overexpressing cells compared with control cells (Fig. 8B). Positive correlations between the expression of c-Jun, c-Fos, SREBP1, FASN, ASCC2, FABP3 and RPRD1B at the mRNA level were further observed in 36 pairs of clinical GC samples (Fig. 8C) and then confirmed in the public TCGA database (Fig. 8D). Finally, c-Jun/c-Fos/SREBP1 axis-mediated fatty acid metabolism was verified in RPRD1B-driven metastatic lymph nodes using IHC and Oil Red O staining (Fig. 8E).