Screening of LncRNAs and characterization of LINC00908 in LUAD.
To explore new effective LncRNAs regulating glycolysis in LUAD, we analyzed the data related to LUAD according to the screening strategy (Fig. 1A). The differentially expressed LncRNAs in normal and cancer tissues were achieved using data from TCGA database. As shown in Fig. 1B, there were 1843 LncRNAs exhibited differentially expression, including 992 upregulated and 851 downregulated LncRNAs in paired samples. In addition, 1906 differentially expressed LncRNAs were screened in unpaired samples involving 883 upregulated and 1023 downregulated LncRNAs (Figure S1). To sum up, there were 1491 overlapped LncRNAs comprising 790 upregulated and 701 downregulated LncRNAs (Fig. 1C). Next, we analyzed these differentially expressed LncRNAs related to survivable prognosis by independent prognostic analysis. The overall prognosis was correlated with 8 upregulated, including LINC00908, and 5 downregulated LncRNAs (Fig. 1D). Subsequently, we decided to explore the correlation between the prognosis-associated LncRNAs and a nine-gene risk score associated with glycolysis. Interestingly, LINC00908 had the greatest correlation coefficient with risk score among these 13 LncRNAs (Fig. 1E), indicating its possible important role in LUAD. Moreover, the expression of LINC00908 in normal tissues was higher than that in tumor tissues (Figs. 1F, G). Further analysis found that LINC00908 showed a good clinical outcome in LUAD (Fig. 1H). Additionally, we analyzed LUAD RNA-seq data for gene set enrichment analysis (GSEA). Consistently, LINC00908 was negatively correlated with poor survival of lung cancer (Fig. 1I). To further explore specific functions of LINC00908 in LUAD, RNA levels were determined in 10 pairs of LUAD tissues and their matched adjacent normal tissues by qRT-PCR. The results similarly demonstrated that LINC00908 expressions in cancer tissues were significantly lower than that in normal tissues (Fig. 1J). Notably, LINC00908 expressions in tumor cells were lower than that in normal lung epithelial cells, remarkably in A549 and H1299 cells (Fig. 1K).
DDX54, a negative prognostic factor for LUAD, is regulated by LINC00908.
The mechanisms by which LINC00908 regulated glycolysis in LUAD were further explored. First, LUAD samples were divided into high and low risk score sets from TCGA employing the above-mentioned nine-gene risk signature model. There were 1 upregulated and 1837 downregulated differential expressed genes in our screen (Fig. 2A). Second, 25 target genes that might be regulated by LINC00908 were predicted based on the Starbase database. Third, an intersection of both the 25 target genes and 1838 differential expressed genes were conducted. Finally, we obtained 3 potential genes, DDX54, IGF2BP3 and TAF15, triggered by LINC00908. (Fig. 2B). Excitingly, the highest correlation was observed between DDX54 and LINC00908 (Fig. 2C, Figure S2). We thus gave priority to the possible regulatory relationship between LINC00908 and DDX54. In 10 LUAD and adjacent tissues, LUAD tissues contained higher levels of DDX54 mRNA and protein than adjacent tissues (Fig. 2D). Similarly, DDX54 was significantly increased in both paired and unpaired analysis based on TCGA database (Fig. 2E). Furthermore, high DDX54 expression levels resulted in a poor OS in LUAD (Fig. 2F). In addition, the pivotal results were that knockdown of LINC00908 significantly fueled DDX54 expressions both in A549 and H1299 cells (Fig. 2G). In total, the intrinsic mechanisms of DDX54 and LINC00908 in LUAD deserved to be estimated.
LINC00908 inhibition facilitated LUAD tumorigenesis through mediating DDX54 in vitro.
Next, we detected the roles of LINC00908 and DDX54 on the growth, apoptosis, migration and invasion in independent LUAD cells. LINC00908 overexpression decreased the protein levels of DDX54 in both A549 and H1299 cells. Remarkably, this impairing effect can be reversed by DDX54 overexpression in the cells transfected LINC00908 (Fig. 3A). Next, CCK8, apoptosis, wound healing, colony assay and transwell assays were conducted. As shown, LINC00908 overexpression decreased the proliferation, migration, invasion in both LUAD cells. Moreover, apoptosis assays indicated that overexpression of LINC00908 could reduce cell apoptosis. Further, cotransfected DDX54 and LINC00908 abolished these effects generated by LINC00908 overexpression (Figs. 3B-E).
On the other hand, we explored whether the effects of LINC00908 knockdown depend on DDX54 in both LUAD cells. As expected, LINC00908 knockdown demonstrably upregulated the ability of proliferation, migration, invasion and downregulated apoptosis. DDX54 consistently restrained these effects of LINC00908 (Figs. 4A-E). Collectively, these data demonstrated that LINC00908 dampens the growth, migration and invasion through DDX54 in LUAD cells. Overexpression of LINC00908 could restrain LUAD progression, and this role of tumor suppressor mainly depended on DDX54 expression in LUAD cells.
DDX54 knockdown downregulates glycolysis in LUAD cells.
We also conducted GSVA analysis to detect the relationship between DDX54 and glycolysis-related pathways. There were 6 pathways implicated in glycolysis, involving KEGG_GLYCOLYSIS_GLUCONEOGENESIS, REACTOME_GLYCOLYSIS, WINTER_HYPOXIA_UP, MOOTHA_GLYCOLYSIS, WINTER_HYPOXIA_DN, GROSS_HYPOXIA_VIA_HIF1A_UP, QI_HYPOXIA_TARGETS_OF_HIF1A_AND_FOXA2 (Figs. 5A-C, Figure S3).
Additionally, we interrogated the possible correlations between DDX54 and 12 glycolysis-related genes according to the database from TCGA. Strong positive correlations were detected between DDX54 and 11 genes (SLC2A1, HK2, GPI, PFKL, ALDOA, GAPDH, PGK1, PGAM1, ENO1, PKM, and LDHA) (Figs. 5D, E). We further found that DDX54 knockdown downregulated the mRNA and protein expressions of SLC2A1, GPI, PFKL, ALDOA, PGK1, PGAM1, ENO1, PKM in both LUAD cells. Excitingly, these outcomes were reversed via reexpression of DDX54 in DDX54 knockdown cells (Figs. 5F, G), which provided further evidence of DDX54 in glycolysis. Moreover, there was a decrease in ECAR and an increase in OCR with DDX54 knockdown, indicating lower glycolytic flux and higher mitochondrial respiration. Once more, DDX54 reexpression reversed these effects (Figs. 5H, I). Besides, glucose uptake, lactate production, ATP generation, and pyruvate were tested to evaluate the glycolysis level. Corresponding with expected, DDX54 knockdown decreased glucose uptake, lactate production, ATP generation, and pyruvate, while DDX54 reexpression reversed these effects (Figs. 5J, K). To sum up, these data suggested that DDX54 knockdown downregulates glycolysis in both A549 and H1299 cells.
LINC00908 downregulates glycolysis through DDX54 in vitro.
Since we proved the regulatory relationship between LINC00908 and DDX54, we next explored whether LINC00908 plays an important role in regulating glycolysis. GSVA analysis revealed that LINC00908 was negative correlated with KEGG_GLYCOLYSIS_GLUCONEOGENESIS, REACTOME_GLYCOLYSIS, WINTER_HYPOXIA_UP, GROSS_HYPOXIA_VIA_HIF1A_UP, REACTOME_REGULATION_OF_GLYCOLYSIS_BY_FRUCTOSE_2_6_BISPHOSPHATE_METABOLISM (Figs. 6A-C, Figure S4). In vitro, LINC00908 knockdown upregulated the mRNA and protein levels of DDX54, SLC2A1, GPI, PFKL, ALDOA, PGK1, PGAM1, ENO1, PKM. DDX54 knockdown abolished these effects in both cells (Figs. 6D, E, I, K). Moreover, the results of glucose uptake, lactate production, ATP generation, pyruvate, ECAR and OCR also confirmed the effects of LINC00908 inhibiting glycolysis via DDX54 (Figs. 6F, G, H, J, L, M).
RFX2 was screened as one of the transcript factors for LINC00908.
We predicted 383 transcription factors which may bind to the promoter region of LINC00908 using the bioinformatics database. These predicted transcription factors were then overlapped with the results of DEGs from the analysis results based on TCGA database. Finally, we obtained 40 candidate transcription factors (Fig. 7A), of which 28 were associated to positive prognostic impact in LUAD. One of these, RFX2, is of particular interest. Because further analyses showed that there were significant correlations between RFX2 and LINC00908 (Figs. 7B-D). Immunoblot analysis and qPCR showed that Based on the important role of RFX2 in LUAD, we then detected the mRNA and protein expressions of RFX2 in 10 pairs of tumor and adjacent tissues. The results demonstrated that RFX2 expression was significantly lower in tumor tissues than that in adjacent tissues (Fig. 7E). Besides, similar results were confirmed in paired or unpaired LUAD samples from TCGA database (Fig. 7F). Thus, it is of importance to further investigate the specific association between RFX2 and LINC00908. We predicted the expected binding sites by a bioinformatics method (http://jaspar.genereg.net/). We then found that RFX2 overexpression increased the LINC00908 promoter reporter activity containing the putative RFX2-binding site, whereas the mutated binding site had little effect on RFX2 activation of the LINC00908 promoter (Fig. 7G). Consistent with this, chromatin immunoprecipitation (ChIP) assay showed that RFX2 was proved to bind to the LINC00908 promoter, not upstream (Fig. 7H). These results suggested that RFX2 might be one of the critical transcription factors for LINC00908.
RFX2 suppresses proliferation, apoptosis, migration, invasion and glycolysis by regulating LINC00908 in LUAD cells.
Subsequently, we validated the possible association between RFX2 and LINC00908 in regulation of LUAD progressions. The results showed that RFX2 overexpression inhibited LUAD cell growth, while LINC00908 knockdown lead to RFX2 inactivation in regulating cell proliferation (Figs. 8A-B). We further explored whether RFX2 inhibits cell apoptosis, migration, and invasion via LINC00908. As expected, RFX2 overexpression inhibited LUAD cell apoptosis, migration, and invasion. LINC00908 knockdown greatly dampened these effects of RFX2 overexpression (Figs. 8C-G).
We also detected the functions of RFX2 on glycolysis in LUAD. RFX2 overexpression suppressed glycolytic genes, including SLC2A1, GPI, PFKL, ALDOA, GAPDH, PGAM1, ENO1, and PKM. had reduced mRNA and protein expression levels of DDX54 and Moreover, these functions of RFX2 were abolished by LINC00908 knockdown (Figs. 9A-D). Further studies also revealed that RFX2 dampened glucose uptake, lactate production, ATP generation, and pyruvate ratio. This reduction could also be reversed by LINC00908 knockdown (Figs. 9E, F). Corresponding with the above results, RFX2 displayed decreased ECAR and increased OCR, which were abolished by LINC00908 knockdown (Figs. 9G-J). Taken together, these data showed that RFX2 suppresses LUAD cell growth, apoptosis, migration, invasion, and glycolysis largely dependent on LINC00908.
RFX2/LINC00908/DDX54 axis has a tumor-promoting effect in vivo.
We next aimed to investigate the phenotype of the RFX2/LINC00908/DDX54 pathway in vivo. First, Injecting BALB/c mice with A549 cells carrying the indicated constructs was performed subcutaneously. We detected that LINC00908 knockdown significantly elevated LUAD tumor growth, whereas knockdown of DDX54 blocked tumor growth. Furthermore, DDX54 knockdown evidently elucidated the regulation by LINC00908 in the tumor growth of xenografts (Figs. 10A-D). Next, we examined the important roles of RFX2 and LINC00908 in the regulation of LUAD tumor growth. As shown in Figs. 10E-G, RFX2 overexpression significantly decreased LUAD tumor growth, while this effect was dramatically attenuated when LINC00908 was knocked down. These data suggesting the key role of RFX2/LINC00908/DDX54 axis in vivo.
Clinical Relevance of LINC00908, DDX54 and RFX2 in the LUAD patients.
We performed immunohistochemical (IHC) staining to assess DDX54 expression and FISH assays to examine LINC00908 expression in 130 human LUAD samples. Positive correlation was found between LINC00908 expression and RFX2 expression, but negative correlation with DDX54 expression (Figs. 11A, B). The potential regulatory pathway is shown in Fig. 8C. Overall, these results indicated that RFX2/LINC00908/DDX54 axis is likely to play important pathological roles in LUAD.