p23 expression is upregulated in lung cancer exhibiting metastatic features.
To evaluate the potential involvement of p23 in lymph node metastasis of lung cancer, we observed a significant upregulation of p23 expression in metastatic lung cancer tissues compared to primary counterparts by querying the TCGA database (Fig. 1A), and further confirmed in clinical samples (n = 133) by immunohistochemical staining (Fig. 1B), qPCR assays (Fig. 1C) and western blotting (Fig. 1D). We next evaluated p23 expression on the prognosis of 93 lung adenocarcinoma patients from a tissue microarray. High p23 expression facilitated the metastasis of lung cancer to lymph nodes in patients (Fig. 1E), and reduced the disease-free survival (DFS) of lung cancer metastasis patients (Fig. 1F). These results suggest that the elevated p23 expression is possibly associated with metastasis and predict the poor clinical outcomes of lung cancer metastasis patients.
Knocking down p23 inhibits the invasion and migration of NSCLC cells.
To further investigate the underlying mechanism of p23 in metastasis, we established stable A549 lung cancer cells with p23 knockdown (A549/shp23) (Supplementary Fig. S1A). Real-time inhibitory analysis using the xCELLigence provided compelling evidence that depletion of endogenous p23 impeded the migratory capacity of lung cancer cells (Fig. 2A). The similar results were also obtained from scratch assay (Fig. 2B), transwell invasion (Fig. 2C) and migration assay (Fig. 2D). Besides, several proteins involved in EMT were also evaluated by western blotting (Fig. 2E) and qPCR analysis (Fig. 2F). p23 knock-down (shp23) greatly decreased protein levels of N-Cadherin and Vimentin, and increased the expression of E-Cadherin. By contrast, p23 stable overexpression (H1299/OEp23) altered the metastasis in an opposite fashion in lung cancer cells (Supplementary Figs. S1B-1D). Furthermore, a mouse model of lung cancer metastasis was also established by intravenous injection of tumor cells via tail vein to evaluate the effect of p23 on metastasis in vivo. Consistent with in vitro observations, shp23 dramatically decreased the number of lung metastatic nodules (Figs. 2G-2I). These data indicate that p23 played an oncogenic role in promoting lung cancer metastasis.
The downstream target gene CXCL1 of p23 was found by RNA-seq sequencing.
In order to determine the molecular pathways involved in p23-driven lung cancer metastasis, we performed RNA-seq sequencing analysis to explore the potential target genes related to p23. A total of 92 differentially expressed genes (DEGs) were found to be commonly shared among the three p23 knock-down groups (Fig. 3A). Then, 11 candidate DEGs for p23 knock-down were selected based on the analysis of typical lung cancer EMT marker genes from the EMTome database (www.EMTome.org) (Fig. 3B), and subsequently confirmed through qPCR detection. The consistent qPCR results validated the reliability of the RNA seq analysis findings (Fig. 3C). According to the abundance of these genes in cell samples, we selected CXCL1 with the highest content as the follow-up research focus (Fig. 3D).
CXCL1 is a vital chemokine driver to promote tumor metastasis[14, 19, 36]. We then proposed that p23 is involved in lung cancer metastasis, at least partially, regulating CXCL1 expression. Importantly, both p23 and CXCL1 expressions were significantly increased in metastatic lung cancer tissues compared to non-metastatic counterparts (Fig. 3E and 3F). In addition, CXCL1 expression was significantly downregulated by p23 depletion, and upregulated by p23 over-expression at both mRNA and protein levels in lung cancer cells (Figs. 3G-3J), indicating that p23 regulates CXCL1 expression in lung cancer cells.
EMT induced by p23 is partially dependent on CXCL1.
To test whether p23 achieves its metastatic regulation functions in a CXCL1-depedent manner, we generated A549/shp23 lung cancer cells stably with over-expression of CXCL1 (A549/shp23 + OECXCL1) to detect the alterations in EMT. As shown in Figs. 4A-4C, CXCL1 significantly reversed the migration inhibition caused by p23 knockdown, as well as the changes in expression of EMT-relative proteins (N-Cadherin, Vimentin and E-Cadherin) (Fig. 4D). Moreover, the similar results were also observed in vivo, that is, shp23 significantly reduced the number of lung metastatic nodules in NYG mice, and this reduction was completely reversed by CXCL1 overexpression (Figs. 4E-4G). These results indicated that p23 promotes lung cancer metastasis through CXCL1 expression. To support this notion, p23 overexpression largely enhanced lung cancer cell in vitro migration, and further CXCL1 knockdown diminished the lung cancer metastasis enhancement induced by p23 overexpression (Supplementary Fig. S2A and 2B).
p23 regulates the transcriptional expression of CXCL1 rather than affect its mRNA stability.
To further assess the function of p23 on CXCL1 expression, we first examined the effect of p23 on the stability of CXCL1 mRNA. Although p23 depletion or overexpression could positively regulate the basal mRNA level of endogenous CXCL1, the half-life of CXCL1 mRNA was not affected by p23 alterations (Supplementary Fig. S3A and 3B), indicating that p23 upregulated CXCL1 expression at transcription level rather than affecting mRNA stability. Furthermore, we constructed luciferase reporter plasmids containing the optimal CXCL1 promoter region (-874 to + 38). Indeed, p23 expression dramatically induced the CXCL1 luciferase activity (Fig. 5A). Through generating truncated the variants of CXCL1 promoter, we observed that deletion of the − 744 to -549 regions fully abolished p23-induced CXCL1 promoter activity (Fig. 5B), indicating that this region containing the binding sites for p23. Therefore, a biotin-labeled probe corresponding to the optimal CXCL1 promoter (from − 744 to + 38) was employed to evaluate the binding capability of p23, with the CXCL1 promoter (from − 549 to + 38) as negative control. As expected, p23 protein was indeed visualized in the pulldown of CXCL1 promoter region − 744 to + 38 from the nuclear extract of A549 cells, whereas no signal was detected in negative control region (Fig. 5C).
To further identify the DNA binding motif of p23 in CXCL1 promoter, we truncated this region (-549 to -744 bp) into five fragments, including P1 (probe 1): -744 to -703, P2: -707 to -663, P3: -667 to -621, P4: -625 to -579, P5: -583 to -537. Nucleotide sequence of CXCL1-EMSA probe was listed in supplementary table 2. The interaction between p23 and each probe was analyzed using EMSA assay with HEK-293T cell lysate overexpressing p23. It was found that p23 had binding affinities with P1, P2 and P3 (Fig. 5D), indicating that three fragments contained p23 binding motifs. This observation was further validated through a super-shift assay using anti-p23 antibody (Fig. 5E). Further DMEME analysis revealed a 6 bp p23-binding motif from three fragments with conserved locations at nucleotides 1C, 2A, 3C, 4T, 5G/T and 6A (Fig. 5F). Mutation of the motifs totally abolished the p23-induced CXCL1 reporter activity (Fig. 5G). These results further confirmed the transcriptional regulatory function of p23 on CXCL1 expression. To further investigate the direct binding activity of p23 to the CXCL1 promoter, we conducted EMSA assay using purified p23 protein. Surprisingly, purified p23 exhibited no binding affinity towards the CXCL1 promoter (Fig. 5H), suggesting that p23's transcriptional regulation of CXCL1 may not be solely dependent on its direct interaction with the promoter, but rather necessitates the involvement of other protein factors.
RBM14 promotes the transcriptional regulation of p23 on CXCL1.
In order to investigate the participants involved in the regulation of CXCL1 by p23, a total of 34 proteins interacting with p23 were identified through co-immunoprecipitation (Co-IP) combined with mass spectrometry (MS), of which 4 proteins were selected due to their transcriptional function (Fig. 6A). Co-IP assay further verify their interaction with p23 (Fig. 6B and supplementary Fig. S4A), which also showed the accuracy of the identification results by mass spectrometry. However, only RBM14 could be pull downed by CXCL1 promoter probe (Fig. 6C and supplementary Fig. S4B), and it can significantly promote p23 reporter activity of the optimal CXCL1 promoter (Fig. 6D). A similar result was also obtained from western blot analysis (Fig. 6E). Furthermore, we conducted EMSA assay to investigate the binding activity of purified p23 and RBM14 protein to the CXCL1 promoter. As expected, p23 or RBM14 individually exhibited no direct binding to the CXCL1 promoter; however, their complexes demonstrated binding activity (Fig. 6F). These results implied that the interaction between p23 and RBM14 plays a crucial role in the regulation of CXCL1 expression.
We next explored the underlying mechanism of p23/RBM14 protein complex participating in the transcription regulation of CXCL1. Firstly, there was no reciprocal influence observed between p23 and RBM14 in terms of their protein and mRNA expression levels (Supplementary Figs. S4C-4F). Subsequently, we modulated intracellular complex production by conducting p23-knockdown and/or RBM14-overexpression as well as p23-overexpression and/or RBM14-knockdown (Fig. 6G and 6H), to elucidate the impact of the complex on CXCL1 transcription. The results demonstrated a significant reduction in CXCL1 promoter activity upon destruction of the p23/RBM14 complex via knockdown of either p23 or RBM14 (Fig. 6I and 6J). Similarly, the binding activity of p23 or RBM14 to CXCL1 promoter probe was significantly suppressed upon destruction of the p23/RBM14 complex (Figs. 6K and 6L). The above results show that the formation of the p23/RBM14 protein complex plays a pivotal role in mediating p23-induced CXCL1 expression.
The influence of p23 on EMT is partially dependent on RBM14.
Next, in order to explore whether the EMT process induced by p23 is RBM14-dependent, we constructed four stable cell lines by using lentivirus infection system: A549/NC, A549/OERBM14, A549/shp23, A549/shp23 + OERBM14. In vitro, the migration number of cells is monitored in real time by xCELLigence assay (Fig. 7A). The result implied RBM14 could promote the phenotype of EMT, but when p23 was knocked down at the same time, the phenotype of EMT could be reduced slightly. In addition, wound healing experiments (Fig. 7B) and transwell invasion (Fig. 7C) as well as migration (Fig. 7D) experiments were performed and brought out similar conclusion. The changes of E-Cadherin, N-Cadherin and Vimentin protein levels were evaluated by western blotting (Fig. 7E), which also proved that p23 promoted EMT phenotype in a RBM14-dependent manner. In vivo, four stably transfected cell lines, A549/NC, A549/shp23, A549/OERBM14 and A549/shp23 + OERBM14, were injected into the tail vein respectively, and the number of lung metastatic nodules (Fig. 7F and G) and H&E staining (Fig. 7H) of lung metastatic nodules were investigated. From above results, we can draw the conclusion that EMT induced by p23 is partially dependent on RBM14.