Upregulated expression of RNF144A-AS1 predicted dismal prognosis in GC
In recent years, using high-throughput sequencing technology, numerous gene expression data have been generated. To identify critical lncRNA implicated in gastric carcinogenesis, integrated analysis of gene expression profiles from the TCGA database was conducted, and the results suggested the expression of RNF144A-AS1 was significantly upregulated in GC tissues (n = 375; Fig. 1A). Furthermore, enhanced expression of RNF144A-AS1 was verified in 60 paired GC tissues, in which 78% of these GC tissues presented with increased expression of RNF144A-AS1 compared with adjacent normal tissues (Fig. 1B, C). Moreover, a higher expression level of RNF144A-AS1 was detected in tumors with distant metastasis than in localized tumors (n = 8) (Additional file 4: Figure S1A). In line with the above findings, GC cells also displayed a higher expression level of RNF144A-AS1 than normal epithelial cell line GES-1 (Fig. 1D). In addition, the clinical relevance of RNF144A-AS1 was also evaluated. From the analysis of the TCGA database, the higher expression level of RNF144A-AS1 was correlated with advanced tumor stages (T stages) and tumor distant metastasis (M stage) (Fig. 1E), as well as linked to poor overall survival rate (HR = 1.70, 95% CI 1.21–2.39) and recurrence-free survival rate (HR = 2.25, 95% CI 1.10–4.60) (Fig. 1F). Beyond this, through categorizing the expression of RNF144A-AS1 as high or low group by using the mean expression level as the cut-off value (n = 26 > mean; n = 34 ≤ mean) in above 60 paired GC tissues, remarkable associations of RNF144A-AS1 with later-stage diseases were discerned, especially with lymph node metastasis (LNM) (P = 0.015) (Additional file 3: Table S3). Next, we would like to determine the localization of RNF144A-AS1 in GC cells. Fluorescence in situ hybridization assay and subcellular fraction assay both indicated RNF144A-AS1 was largely cytoplasm-localized (Fig. 1G, H). Meanwhile, the coding potential of RNF144A-AS1 was also investigated. Using online tools like Coding Potential Assessment Tool and ORF finder, the noncoding feature of RNF144A-AS1 was confirmed (Additional file 4: Figure S1B, C).
RNF144A-AS1 promoted the metastasis, angiogenesis, and proliferation of GC
First, using gene expression data from the TCGA database and CCLE database, GSEA analysis indicated that RNF144A-AS1 was enriched in some key gene sets such as hypoxia, EMT, angiogenesis, apical junction (NES > 1.0, NOM P-value < 0.05) (Fig. 2A; Additional file 5: Figure S2A, B). Then, to validate the above predictions in GC cells, the expression of RNF144A-AS1 was downregulated by siRNAs in MKN45 and AGS cells and was overexpressed in HGC27 cells (Fig. 2B). Next, transwell assays showed the migrative and invasive ability of GC cells were seriously harmed when transfected with siRNAs targeting RNF144A-AS1 (Fig. 2C, D). Meanwhile, wound-healing assays further consolidated the inhibition of cellular migration by knockdown of RNF144A-AS1 in MKN45 and AGS cells (Additional file 5: Figure S2C, D). Additionally, conditioned medium obtained from RNF144A-AS1-silenced MKN45 (CM-MKN45) and AGS (CM-AGS) cells progressively decreased the tube-formation rate of HUVECs, suggesting attenuated tumor angiogenesis by the knockdown of RNF144A-AS1 (Fig. 2E, F). Of note, CCK8 assays also suggested that blockade of RNF144A-AS1 impaired the growth of MKN45 and AGS cells (Fig. 2G, H). In contrast, reintroduction of RNF144A-AS1 into HGC27 cells inversely promoted cellular migration, invasion (Fig. 2I), and proliferation (Additional file 5: Fig. S2E). Moreover, conditioned medium derived from HGC27 cells with overexpressed expression of RNF144A-AS1 stimulated the tube-formation rate of HUVECs, as compared with the control group (Fig. 2J). EMT represents one of the cardinal signs of cell invasion and cancer metastasis [25], thus, we then determined the influence of RNF144A-AS1 in this process. As depicted in Fig. 3A and B, silencing RNF144A-AS1 induced the expression of epithelial markers like E-cadherin, accompanied by a dampened expression of mesenchymal markers such as N-cadherin and Vimentin both in protein and RNA levels. Meanwhile, the expression of vascular endothelial growth factor A (VEGFA) was also hampered by the knockdown of RNF144A-AS1 in GC cells, indicating a VEGFA-dependent way for RNF144A-AS1-stimulated angiogenesis. In agreement, it is clear that RNF144A-AS1 contributes to the metastasis, angiogenesis, and proliferation of GC.
To delineate further the functional significances of RNF144A-AS1 in those malignant processes in vivo, we developed multiple animal models. Using RNF144A-AS1-specific short hairpin RNA, the expression of RNF144A-AS1 was diminished (Fig. 3C). First, the lung metastatic model was established by directly injecting treated MKN45 and AGS cells into the tail vein of nude mice. It was evident the number of lung metastatic nodules was sharply reduced in the RNF144A-AS1-silenced group than the control group (Fig. 3D, E). Next, in an attempt to evaluate the effect of RNF144A-AS1 on GC angiogenesis, a Matrigel plug assay was further conducted. Compared with control, downregulation of RNF144A-AS1 markedly decreased the rate of blood vessel formation of Matrigel plugs, as well as reduced skin vascularization adjoining the plug (Fig. 3F, H). In accordance, immunohistochemical analysis of plug sections for angiogenesis marker like platelet endothelial cell adhesion molecule-1 (CD31) also suggested lower blood vessel formation for the RNF144A-AS1-knockdown group (Fig. 3G, I). In addition, nude mice xenograft assays further indicated that depletion of RNF144A-AS1 in GC cells specifically delayed tumor formation and reduced the final tumor weight of subcutaneous tumors when compared with the control group (Fig. 3J, L). Immunohistochemical staining of Ki67, a cell proliferative marker, further substantiated our macroscopic observation (Fig. 3K, M). Hence, these results strongly supported the oncogenic role of RNF144A-AS1 in GC outgrowth, metastasis, and angiogenesis.
RNF144A-AS1 served as a miRNA decoy for miR-30c-2-3p
Having established that RNF144A-AS1 acts as an oncogene in GC, we next aimed to determine the underlying mechanism. Given the cytoplasmic localization of RNF144A-AS1 in GC cells, online tools such as RegRNA2.0 and RNA22-HAS were applied to predict the potential binding sites of miRNAs to RNF144A-AS1 (Fig. 4A). Because of the relatively low expression level in GC tissues from the TCGA database, miR-30c-2-3p and miR-139-3p were chosen for further research (Fig. 4B; Additional file 6: Figure S3A, B). However, GC cells transfected with RNF144A-AS1-specific siRNA only exhibited upregulated expression of miR-30c-2-3p, but not miR-139-3p (Fig. 4C; Additional file 6: Figure S3C). Consistent with above findings, HGC27 cells transfected with RNF144A-AS1 vector restricted the expression of miR-30c-2-3p (Fig. 4C). Thus, we then exclusively explored the relationship between miR-30c-2-3p and RNF144A-AS1. To validate the interaction, the binding sequences of miR-30c-2-3p to RNF144A-AS1 were mutated and engineered into a luciferase reporter vector (Fig. 4D). As showed in Fig. 4E, the luciferase activity was efficiently reduced following co-transfection with increasing concentration of miR-30c-2-3p and a reporter vector carrying the wild-type sequence of RNF144A-AS1 into HEK293T cells. Nevertheless, mutation of miR-30c-2-3p binding sites abolished the suppressive effect of miR-30c-2-3p mimics on RNF144A-AS1-driven luciferase activity (Fig. 4E). MiRNAs are involved in the formation of RNA-silencing complex with Argonaute 2 (Ago2) and can guide the complex to bind targeted genes, thereby reducing the expression of downstream targets [26]. In this regard, we then attempted to ensure the influence of RNF144A-AS1 on the miR-30c-2-3p-dependent RNA-silencing complex. As expected, using an anti-Ago2 antibody, transcripts of RNF144A-AS1 were efficiently immunoprecipitated from AGS cells (Fig. 4F). Moreover, the qRT-PCR analysis showed compared with the control group, both the expression level of RNF144A-AS1 and miR-30c-2-3p were drastically decreased in the immunoprecipitation purified from AGS cells transfected with miR-30c-2-3p inhibitor (Fig. 4G). Additionally, correlation analysis of 60 GC tissues indicated an inverse correlation between RNF144A-AS1 and miR-30c-2-3p (R2 = 0.2448, P < 0.001) (Additional file 6: Figure S3D). Based on these findings, we concluded that lncRNA RNF144A-AS1 could bind with miR-30c-2-3p and negatively regulate its expression in GC.
MiR-30c-2-3p is a well-defined tumor suppressor that can target a variety of cancer-associated genes, yet its function in GC is still poorly understood [27]. As suggested in Fig. 4H, significantly upregulated expression of miR-30c-2-3p was observed in GC cells transfected with miRNA mimics, while restricted expression of miR-30c-2-3p was discerned by miR-30c-2-3p-specific inhibitor (Additional file 6: Figure S3E). First, the tube-formation ability of HUVECs was severely inhibited after co-cultured with conditioned medium derived from AGS cells treated with miR-30c-2-3p mimics (Fig. 4I). Meanwhile, transwell assays with or without Matrigel coating showed overexpression of miR-30c-2-3p resulted in less migration and invasion than that observed in control cells (Fig. 4J, K). Remarkably, the inhibitory effect of miR-30c-2-3p on EMT was detected either, characterized by increased expression of E-cadherin and decreased expression of N-cadherin and Vimentin after overexpression of miR-30c-2-3p (Additional file 6: Figure S3F). Moreover, the expression level of VEGFA was potently abolished by miR-30c-2-3p either (Additional file 6: Figure S3F). Thus, these results indicated that a low expression level of miR-30c-2-3p may be indispensable for GC aggressiveness.
To determine whether miR-30c-2-3p mediates the promotive effect of RNF144A-AS1 on GC, we co-transfected miR-30c-2-3p mimics and RNF144A-AS1 vector into GC cells. First, transwell assays suggested RNF144A-AS1-mediated promotions on cellular migration and invasion were markedly reversed by miR-30c-2-3p mimics (Fig. 4L; Additional file 6: Figure S3G). Tube-formation assays also suggested miR-30c-2-3p could phenocopy RNF144A-AS-driven angiogenesis (Fig. 4M). In the same direction, ectopic miR-30c-2-3p rescued the effects of RNF144A-AS1 on cell proliferation (Additional file 6: Figure S3H, I). In summary, we found RNF144A-AS1 promotes tumor metastasis, angiogenesis, and proliferation by competitively binding miR-30c-2-3p.
LOX was determined as the target of miR-30c-2-3p
The competing endogenous RNA (ceRNA) model demands the interactions among lncRNA, miRNA, and downstream targets. Here, LOX was predicted as the target of miR-30c-2-3p through intersecting four datasets, including the gene set about hypoxia enriched by RNF144A-AS1 and target sets generated by miRDB, DIANA, and Targetscan (Fig. 5A). Notably, GC cells transfected with miR-30c-2-3p mimics exhibited lower expression of LOX both in RNA and protein level, compared with the control group (Fig. 5B, C). Meanwhile, downregulating of miR-30c-2-3p induced elevated expression of LOX either (Additional file 7: Figure S4A, B). Moreover, the 3′ UTR of LOX was fused into a luciferase reporter vector, and another reporter vector containing mutated miR-30c-2-3p binding sites was also constructed (Fig. 5D). Then these vectors together with miR-30c-2-3p mimics were co-transfected into HEK293T cells, respectively. As expected, mutating miR-30c-2-3p seed sequences in LOX 3′ UTR was sufficient to abolish miR-30c-2-3p-dependent regulation of luciferase activity (Fig. 5E). On the other hand, an inverse correlation between LOX and miR-30c-2-3p was detected in 60 GC tissues (R2 = 0.1293, P = 0.0048) (Fig. 5F). In this regard, it is evident that LOX is the target of miR-30c-2-3p. However, it is also important to ensure the regulation of RNF144A-AS1 to LOX. First, depletion of RNF144A-AS1 drastically suppressed the expression of LOX in MKN45 and AGS cells, while artificial expression of RNF144A-AS1 significantly boosted LOX expression in HGC27 cells (Fig. 5G, H). Using expression data of 375 GC tissues from the TCGA database, we detected a positive correlation between RNF144A-AS1 and LOX either (R2 = 0.2714, P < 0.001), which was consistent with the result from GC cells (R2 = 0.7135, P < 0.001) (Fig. 5I, J). Importantly, restoring the expression of miR-30c-2-3p progressively rescued RNF144A-AS1-induced upregulation of LOX in HGC27 cells (Fig. 5K). Meanwhile, co-transfection with miR-30c-2-3p inhibitor also rescued the expression of LOX in MKN45 cells transfected with siRNA against RNF144A-AS1 (Additional file 7: Figure S4C). For further confirmation, we constructed expression plasmids containing the sequence of RNF144A-AS1, which contains wild-type (Wild) or mutated miR-30c-2-3p binding sites (Mutated). We found overexpression of RNF144A-AS1 drastically enhanced the expression of LOX but not empty plasmid or mutant plasmid (Fig. 5L). Furthermore, in a similar manner, mutated LOX 3ʹUTR or empty plasmid could not vary the expression of RNF144A-AS1 except for the wild type with the binding sites of miR-30c-2-3p (Fig. 5M). Hence, the above findings confirmed the competition to miR-30c-2-3p between LOX and RNF144A-AS1. Previous researches have established the fundamental role of Dicer in miRNA biogenesis [28]. Therefore, we intend to further testify that miR-30c-2-3p acts as an intermediary between RNF144A-AS1 and LOX by knockdown of Dicer. First, silencing of Dicer successfully inhibited the expression of LOX, emphasizing the important role of miRNAs in the regulation of LOX (Fig. 5N). Then, as expected, when we co-transfected Dicer targeting-siRNA and RNF144A-AS1 vector into HGC27 cells, the expression of LOX did not make a difference compared with the Dicer knockdown group (Fig. 5O). Therefore, these results supported the ceRNA model among RNF144A-AS1, miR-30c-2-3p, and LOX.
LOX improved the aggressiveness of GC
HIF-1α-induced expression of LOX typically causes stiffening of the ECM in cancer, thereby allowing cancer cells to easily metastasize [29]. In our study, we delineate further the function of LOX in GC. Indeed, predominant overexpression of LOX was observed in GC tissues and cells (Fig. 6A–C). In addition, a high expression level of LOX was strongly associated with poor overall survival, particularly in patients diagnosed at advanced stages (Additional file 7: Figure S4D). Then, the expression of LOX was efficiently knocked down by siRNA in GC cells (Fig. 6D). As illustrated, depletion of LOX drastically inhibited the migrative and inhibitive ability of GC cells (Fig. 6E, F). Tube-formation assay also suggested dampened angiogenetic ability of HUVECs after co-cultured with LOX-silenced conditioned medium, as compared with control (Fig. 6G, H). Furthermore, Western blot analysis indicated knockdown of LOX inhibited EMT process on GC cells and the expression of VEGFA (Fig. 6I). Remarkably, in vivo observations revealed reduced blood vessel formation of plugs, as well as decreased skin vascularization in the LOX knockdown group (Fig. 6J, K). And IHC staining of CD31 further emphasized the attenuation of angiogenesis in Matrigel mixed with GC cells transfected by LOX siRNA (Fig. 6L). A similar phenomenon was also detected in cell proliferation, in which downregulating of LOX inhibited cell growth and tumor formation both in vivo and in vitro (Fig. 6M, N). Therefore, it is reasonable to believe that LOX could promote tumor metastasis, angiogenesis, and proliferation in GC.
LOX was required for the function of miR-30c-2-3p in GC
To evaluate the role of LOX in mediating the function of miR-30c-2-3p, we co-transfected LOX-specific siRNA and miR-30c-2-3p inhibitor into GC cells. First, Transwell assays indicated that induced promotion on cellular invasion and migration by miR-30c-2-3p inhibitor was mitigated through downregulating of LOX (Fig. 7A, B). And an analogous mode was detected in tube-formation assays (Fig. 7C). Notably, Western blot analysis showed the same influence on the expression of LOX, EMT-markers, and VEGFA, which miR-30c-2-3p inhibitor rescued the suppression of these genes in the LOX-knockdown group (Fig. 7D). Meanwhile, the CCK8 assays indicated that miR-30c-2-3p inhibitor could rescue the impaired cell growth in the LOX-silencing group (Fig. 7E, F). Therefore, the function of miR-30c-2-3p is orchestrated by its negative regulation of LOX.
RNF144A-AS1 was a TGF-β- and hypoxia-inducible gene in GC
Since RNF144A-AS1 tightly regulates the expression of LOX through targeting miR-30c-2-3p, we, therefore, conducted GSEA analysis on LOX to evaluate whether LOX and RNF144A-AS1 are jointly involved in certain cellular phenotypes or signaling pathways. Intriguingly, LOX and RNF144A-AS1 were enriched in the same gene sets with relatively similar enrichment scores using gene expression data from the CCLE database, which indicates RNF144A-AS1 functions synchronously with LOX in GC (Fig. 8A; Additional file 7: Figure S4E, F). Moreover, we then detected the expression of RNF144A-AS1, miR-30c-2-3p, and LOX under hypoxic conditions. It showed the expression of RNF144A-AS1 and LOX was significantly elevated under the hypoxic circumstance, while the expression of miR-30c-2-3p was greatly downregulated (Fig. 8B–D). Moreover, compelling evidence has shown the reciprocal interplay between LOX, HIF-1α, and hypoxia. Therefore, we decided to explore the potential interaction of RNF144A-AS1 and HIF-1α. However, through the correlation analysis among RNF144A-AS1, LOX, and HIF-1α, we did not find a relationship between RNF144A-AS1 and HIF-1α (Fig. 8E, F). Meanwhile, knockdown of HIF-1α did not influence the expression of RNF144A-AS1 but LOX (Fig. 8G; Additional file 8: Figure S5A). Together, these experiments revealed that hypoxia stimulated RNF144A-AS1 expression in a HIF-1α-independent manner.
Notably, a plethora of studies demonstrates that TGF-β is also a strong promoter of EMT and is significantly involved in tumor metastasis [30]. Thus, we wondering whether TGF-β signaling is one of the upstream activators of the RNF144A-AS1-miR-30c-2-3p-LOX axis. First of all, GSEA analysis indicated a positive correlation of LOX expression and TGF-β signaling, suggesting a potential interaction between them (Additional file 8: Figure S5B, C). Then, through exploring the TCGA database, we detected that the expression of LOX and RNF144A-AS1 were significantly associated with TGF-β1, the main member of the TGF-β family (R2 = 0.1892 for RNF144A-AS1; R2 = 0.2672 for LOX) (Fig. 8H, I). To further delineate the role of TGF-β1 on the regulation of RNF144A-AS1 and LOX, GC cells were treated with recombinant human TGF-β1. Compared with day 0, the expression of RNF144A-AS1 and LOX both increased in all GC cell lines after treatment for ten days, whereas the expression of miR-30c-2-3p was suppressed (Fig. 8J; Additional file 8: Figure S5D). Interestingly, a higher fold-change of LOX, rather than RNF144A-AS1, was discerned, indicating that TGF-β1 manipulates LOX expression in multiway. Thus, we concluded that RNF144A-AS1 was induced in a TGF-β1-dependent pathway.