3.1 Elevated expression of RCN2 correlates with poor prognosis in NPC

Aberrant gene expression is repeatedly observed in NPC, which correlates with cancer progression, chemotherapy failure, and poor clinical outcomes [33, 34]. Therefore, based on the analysis of NPC microarray datasets from the Gene Expression Omnibus (GEO) databases (GSE12452, GSE13597, and GSE53819) and the survival-related dataset (GSE102349_COX), 29 significantly upregulated genes were identified. Among these, RCN2 drew our attention because of its involvement in calcium homeostasis (P < 0.05) (Fig. 1A, B). Analysis of TCGA Database reveals elevation of RCN2 (Supplementary Fig. 1B) and the expression of RCN2 protein in HNSC was second only to thyroid cancer among the 20 different types of tumors (Supplementary Fig. 1A). All six NPC cell lines demonstrated significantly high levels of RCN2 protein and mRNA by Western blotting and qRT-PCR (Fig. 1C, D). Then we confirmed the clinical role of RCN2 using the in-situ hybridization (ISH) data with NPC tissue. RCN2 is highly expressed in NPC tissues and high RCN2 expression were positively correlated with terminal stage disease (Fig. 1E, F). As expected from ISH results, patients with higher RCN2 expression displayed worse overall survival and progression-free survival (Fig. 1G, H). Additionally, RCN2 play a similar function in cancer progression and present similar prognostic features for NPC from GEO (GSE102349) database (Fig. 1I), and head and neck cancer (HNSC) from The Cancer Genome Atlas (TCGA) database (Supplementary Fig. 1C). Collectively, these data suggest that the overexpression of RCN2 in NPC may have significant clinical relevance.

Fig. 1

Expression and clinical significance of RCN2. (A) Venn diagram of upregulated genes among GEO databases (GSE12452, GSE13597, GSE53819, and GSE102349_COX). (B) Expression of RCN2 in three GEO databases. (C) RCN2 protein levels were assessed in cell lines by western blotting (WB). (D) qRT-PCR of RCN2 mRNA levels in NP69 and NPC cell lines analyzed using one-way ANOVA. (E) Representative RCN2 ISH staining (scale bar: 100 μm) and RCN2 expression in different clinical stages. (F) TheRCN2 staining scores were divided into low expression (scores of 0–9) or high expression (scores of 10–16). Kaplan–Meier curves representing overall survival under the influence of lymph node metastasis. (G) Kaplan–Meier curves representing overall survival were stratified (log-rank test). (H) Kaplan–Meier curves representing progression-free survival were stratified (log-rank test). (I) Survival differences were analyzed using the log-rank test among GSE102349_COX. (J) Lentivirus efficiency of RCN2 was detected by Western blot in CNE2 and 5-8 F cells. (K) Effects of cas 9 on p53 and RCN2 protein expressions. Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

3.2 RCN2 regulates the NPC malignant phenotype

As the upregulation of RCN2 was significantly associated with worse survival, we investigated the direct effect of RCN2 in NPC cells. The relatively high-expression cell lines (CNE2, 5-8 F) were chosen for further study. To investigate the endogenous function of RCN2, we established RCN2 overexpression and knockout cell lines. We utilized CRISPR-Cas 9 gene editing and a lentiviral-mediated overexpression approach to develop stable RCN2 knockout and overexpression cell lines by pre-selecting single cells and transfection, respectively [35, 36]. We verified the efficiency of the RCN2 knockout via western blot. Cas 9 − 5, cas 9 − 6, and cas 9 − 7 showed high efficiency of the constructs (Supplementary Fig. 1D). The knockout efficiency of cas 9 and overexpression of CNE2 and 5-8 F were validated by qRT-PCR (Supplementary Fig. 1E). After selection, a stable knockout of the cas 9 − 7 clone was established. Knockout and overexpression efficiencies of CNE2 and 5-8 F were confirmed using Western blotting (Fig. 1J). It was observed that the negative control of cas 9 and the negative control of RCN2 had a similar effect. In addition, our findings showed that the expression of a selected group of cancer-related genes such as P53 was not significantly affected by CRISPR-cas 9. (Fig. 1K).

Next, we investigated the effects of RCN2 on the malignant biological behavior of tumor cells. CCK8, EDU, and colony formation assays were utilized to assess the cell proliferation potential, showing that the knockout of RCN2 inhibited the proliferative capacity, while cells with high RCN2 levels had a higher proliferative capacity in CNE2 and 5-8 F cells (Fig. 2A-E; Supplementary Fig. 1F). A three-dimensional sphere culture system has been shown to provide a more suitable tumor microenvironment in vivo than 2D monolayer cell culture models. Using this culture system, higher RCN2 expression was found to enhance the mammosphere-forming capacity (Fig. 2F, G). In addition, a similar tendency was observed in mobility and invasive capacity through Transwell (Fig. 2H-J) and wound-healing assays (Supplementary Fig. 1G, H). Overall, these findings suggest that RCN2 regulates the proliferation, diffusion, and spheroid formation of NPC cells in vitro.

Fig. 2

RCN2 regulates NPC malignant biological properties in NPC cells. EDU assay (scale bar: 100 μm) (A-C), and colony formation assay (scale bar: 5 mm) (D, E) in CNE2 and 5-8 F were performed to measure cell proliferation and quantifications. (F, G) Cell mammosphere-forming capacity was analyzed by performing three-dimensional spheroid formation assay (scale bar: 400 μm). (H-J) Transwell assays (scale bar: 200 μm) were performed to analyze cell mobility and invasive capacity. Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

3.3 RCN2 regulates mitochondrial-related apoptosis driven by cellular stress in vitro

Given that RCN2 is an EF-hand Ca2+-binding protein [37, 38], we suspected that the differential expression of RCN2 would disrupt calcium homeostasis and affect the apoptosis of NPC cells. Flow cytometry was performed to assess the extent of apoptosis. None of these differences between cells expressing different levels of RCN2 was statistically significant (Fig. 3A, B). Oxidative stress of the tumor microenvironment plays an important role in apoptotic signaling cascades. Therefore, we hypothesized that RCN2 regulates apoptosis under intracellular stress. To further validate this assumption, the above experiment was repeated in the presence of starvation-induced oxidative stress. RCN2 knockout cells comprised a significant proportion of apoptotic cells, indicating that programmed cell death was induced (Fig. 3A, B), whereas very little apoptosis developed in RCN2 overexpressed cells. Given that ROS generation is involved in stress-induced apoptosis [39, 40], we cultured cells in glucose-free, hypoxic, and oxygen-glucose deprivation medium to mediate stress. As a result, we observed that the knockout of RCN2 resulted in lower ROS levels (Fig. 3C; Supplementary Fig. 2A). In contrast, the overexpression of RCN2 significantly elevated cellular ROS levels. However, ROS did not exhibit this trend in the absence of oxidative stress (Fig. 3C; Supplementary Fig. 2A). ROS can trigger cell death via the mitochondrial apoptosis pathway [41].

Fig. 3

Pro-apoptotic activity driven by RCN2 under cellular stress. (A, B) Flow cytometry analysis of CNE2 and 5-8 F cells after transfection to induce different levels of RCN2 expression under starvation or not to measure the apoptosis rate and quantify cell apoptosis. (C) The ROS levels were quantified under conditions of glucose-free, hypoxia, or oxygen-glucose deprivation or not. (D, E) Apoptosis in CNE2 cells as defined by TFAR19 immunofluorescent staining (scale bar: 40 μm). (F, G) Western blot analysis of Cl-Cas 3, GRP78, and Cl-Cas 12 in CNE2 cells under starvation or not in CNE2 and 5-8 F cells. (H) WB analysis to assess activation of mitochondrial apoptotic pathway markers Apaf-1, cyto C, and Cl-Cas 9 under starvation or not in CNE2 and 5-8 F cells based on the separation of the cytoplasm and mitochondria. Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Additionally, nuclear translocation of programmed cell death protein 5 (TFAR19) is an early event of apoptosis, meanwhile, caspase-3, in particular cleaved caspase-3 (the active form), is considered an apoptotic marker. Notably, the level of apoptosis in stressed cells was upregulated with the concomitant downregulation of RCN2, as detected by nuclear translocation (Fig. 3D, E) and WB (Fig. 3F, G), while RCN2 overexpressed cells showed less apoptosis. Taken together, these results indicate that RCN2 is involved in the regulation of cancer cell apoptosis under intracellular stress.

Cell death pathways are highly regulated by multiple mechanisms, including mitochondria-mediated, ER-mediated, and autophagy-dependent cell death. RCN2 was described to localize predominantly to the ER. We then examined whether RCN2 regulates cell death through endoplasmic reticulum autophagy (ER-phagy). The ER and lysosome marker molecules were stained to quantify co-localization. Unfortunately, no evidence was obtained that RCN2 influences the fusion of the ER and lysosomes under starvation conditions or not (Supplementary Fig. 2B, C). These results indicate that RCN2 does not mediate cell death via ER-phagy. The GRP78 and caspase-12 activities were evaluated to determine the level of ER stress and ER pathway apoptosis. Increased ER stress occurred in the lower RCN2 level group under starvation, whereas no difference was observed without stress (Fig. 3F, G). No significant differences were observed in cleaved caspase-12 expression of the ER-apoptotic factor (Fig. 3F, G). In summary, while our results point to a significant activation of ER stress responses under stress when RCN2 is knocked out, the absence of RCN2 does not appear to cause increases in ER pathway apoptosis.

Thus, we detected the activation of mitochondria-related intrinsic apoptotic death. Moreover, based on the separation of the cytoplasm and mitochondria, the relative content of cytochrome c (cyto C), Apaf-1, and cleaved caspase-9 protein in the cytoplasm or mitochondria was detected. The study involved the isolation and purification of mitochondria from CNE2 and 5-8 F cancer cells while retaining the cytoplasm. Protein extraction was then performed separately. The results of the study demonstrated that cyto C, Apaf-1, and cleaved caspase-9 protein was significantly increased in the cell lysates, especially in mitochondria. There was no apparent difference in the cytoplasm under stress conditions (Fig. 3H). We concluded that cell death induced by the deletion of RCN2 arose specifically from the mitochondrial apoptosis pathway rather than other cell death pathways under stress.

Taken together, our findings suggest that the deletion of RCN2 induces apoptosis in stressed cells through the activation of intrinsic mitochondrial-dependent and Apaf-1-associated pathways. High levels of RCN2 expression in a subset of cancer cells may facilitate cell survival in adverse environments.

3.4 Confirmation of mitochondrial-related apoptosis in vivo

Based on the results obtained for RCN2 in the in vitro experiments, different expression levels of RCN2 CNE2 cells were subcutaneously inoculated into nude mice, and a subcutaneous xenograft BALB/c mouse model was established to assess cell proliferation ability in vivo (Fig. 4A). After three weeks, we resected the subcutaneous tumors and measured their weights and volumes. The subcutaneous tumors in the knockout RCN2 group had a smaller volume and weighed less than those in the control group, while RCN2 overexpression was associated with a significant increase in tumor size and weight (Fig. 4A-D). IHC testing demonstrated that xenografts with lower RCN2 expression had higher levels of mitochondrial apoptosis pathway markers, such as Apaf-1, cyto-C, caspase 9, caspase 3, and caspase 7, while there was no significant difference in ER apoptosis pathway markers, such as caspase 12 (Fig. 4E).

Fig. 4

Confirmation of mitochondrial-related apoptosis in vivo. (A) Representative NPC xenografts in the indicated group of mice (n = 5 per group). (B) Tumor volume was recorded at the time points indicated. (C) The volume of tumors on day 21. (D) Tumor weight was measured after the tumors were surgically dissected. (E) Tumors were analyzed for MKI67, RCN2, cyto-C, Apaf-1, cas 9, cas 3, cas 7, and cas 12 expression by IHC (scale bar: 80 μm). (F) CNE2-luc and 5-8 F-luc cells were transfected and injected into mice through the tail vein. The lung metastasis visualization is representative of five independent experiments. (G) The arrow indicates the nodules of metastatic lung tumors in experimental mice. (H) Average intensity of fluorescence. (I) Typical images of histological lung metastases were analyzed using HE or RCN2 staining, and the number of lung metastatic tumors was calculated (J). Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Next, a lung metastasis model was established via tail vein injection of mice with control, RCN2 knockout, or overexpressed CNE2 and 5-8 F cells. Tumor growth and metastasis were monitored by evaluating the resulting bioluminescence values, which along with examination of lungs revealed that the RCN2 knockout group formed fewer pulmonary metastatic foci than the control group, with many more lung metastatic foci of overexpressed RCN2 cells than those of the negative control (Fig. 4F-H). Lung slices from each group were stained with hematoxylin and eosin, consistent with the in vivo imaging results (Fig. 4I, J).

In summary, these findings provide evidence that RCN2 accelerates growth and metastasis in vivo and confirm that RCN2 decreases stress-mediated apoptosis, mainly by inhibiting the mitochondrial apoptosis pathway.

3.5 RCN2 regulates mitochondrial dysfunction and Ca2+ overload

In addition to confirming the influence of the mitochondrial apoptotic pathway, we investigated how RCN2 regulates mitochondria-mediated apoptosis. First, JC1 staining was used to evaluate the mitochondrial membrane potential. The knockout of RCN2 cells retained monomer-emitting green fluorescence and showed mitochondrial dysfunction under cellular stress due to a reduction in mitochondrial membrane potential, while overexpressed RCN2 cells exhibited normal mitochondrial function (Fig. 5A, B; Supplementary Fig. 3A, B). Mitochondrial permeability transition pores (MPTP) also play a role in the development of mitochondrial dysfunction. Oxidative stress conditions stimulate MPTP to open, whereby part of calcein is discharged into the cytoplasm and combines with cobalt ions to lose fluorescence [42]. In our study, RCN2 knockout enhanced the opening of MPTP under oxidative stress with weaker green fluorescence, whereas RCN2 overexpression resulted in stronger green fluorescence (Fig. 5C-E). Mitochondrial calcium overload can elevate mitochondrial apoptosis by inducing permeability transitions and MPTP openings [43]. To detect cells that exhibited spontaneous calcium transients, Fluo-4 and Rhod-2 were used to measure cytosolic and mitochondrial calcium levels, respectively. As shown in the results, the knockout of RCN2 revealed increased Rhod-2 fluorescence and no difference in Fluo-4 fluorescence, which reflected an augmentation of calcium levels in the mitochondria, while calcium was gradually released from the mitochondria in the overexpressed RCN2 group under stress (Fig. 5F-I). Overall, under cellular stress damage, RCN2 deficiency stimulates MPTP opening, calcium overload in the mitochondria, mitochondrial dysfunction, cytochrome c release, and ultimately the induction of apoptosis.

Fig. 5

RCN2 regulates mitochondrial dysfunction and Ca2+ overload. (A) Confocal microscopy of mitochondrial membrane potential was evaluated by JC1 staining. Red puncta-maintained mitochondria; green puncta, depolarized mitochondria (scale bar: 30 μm). (B) Quantification of red/green represent JC1 aggregates/monomers. (C) MPTP openings were evaluated using the calcein cobalt method (scale bar: 80 μm) in CNE2 and 5-8 F cells. (D, E) MPTP openings were evaluated using the calcein cobalt metho, and the normalized relative fluorescence units (NRFU) of calcein in the experimental groups. (F) Mitochondrial calcium visualized by Rhod-2 under confocal microscopy (scale bar: 80 μm). (G) Quantitation of peak Rhod-2 fluorescence. (H) Cytoplasmic calcium visualized by Fluo-4 under confocal microscopy (scale bar: 80 μm). (I) Quantitation of peak Fluo-4 fluorescence. (J) RCN2 protein interaction network was generated using the STRING database. (K) CALR expression was detected following different expression of RCN2 by western blot in CNE2 and 5-8 F cells. (L) CoIP of endogenous RCN2 and CALR was performed to validate protein–protein interaction in CNE2 and 5-8 F cells. Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

MPTP: mitochondrial permeability transition pores

RCN2 is highly connected with CALR (Calreticulin) by putative protein-protein interactions in the STRING database (Fig. 5J). CALR functions in calcium homeostasis and we found CALR expression was significantly decreased following knockout of RCN2 and increased by the overexpression of RCN2 in CNE2 and 5-8 F cell lines (Fig. 5K). In addition, co-immunoprecipitation (CoIP) assay was performed to confirm their binding (Fig. 5L). These observations demonstrate that CALR interacts with RCN2, thereby promoting calcium storage in the ER, decreasing the potential for strong signaling to mitochondria and inhibiting apoptosis.

3.6 YY1 and GSC are upstream regulators of RCN2 and correlate with poor prognosis in HNSCC

To further elucidate the molecular mechanisms involved in the upregulation of RCN2 in NPC cells, the transcription factor binding profiles were collected from JASPER (https://jaspar.genereg.net/) [44], and the top five were selected for further analysis (YY1, GSC, ETS1, TCF3, and USF1). Silencing of YY1 or GSC was found to downregulate the expression of RCN2 (Fig. 6A, B; Supplementary Fig. 3C, D), whereas the knockdown of the other three predicted transcription factors had no influence on the expression of RCN2 (Supplementary Fig. 3E-J). We performed luciferase reporter and chromatin immunoprecipitation (ChIP) analyses to determine the promoter sites and demonstrated that RCN2 is directly transcriptionally regulated by YY1 and GSC (Fig. 6C, D). In particular, both YY1 and GSC were upregulated in NPC cell lines compared to NP69 cells by PCR and western blotting (WB) (Fig. 6E, F). In addition, YY1 or GSC expression was positively correlated with RCN2 expression in the TCGA database (Fig. 6G, H). We then verified the regulatory effects of YY1 and GSC on RCN2 expression. We observed that silencing YY1 or GSC inhibited the expression of RCN2, whereas RCN2 overexpression reversed the silencing effect of YY1 or GSC on RCN2 expression (Fig. 6I-J; Supplementary Fig. 3K, L). Hence, YY1 and GSC are the upstream transcription factors of RCN2. Moreover, patients with higher expression of YY1 and GSC had a poorer prognosis and reduced survival time, according to TCGA database (Supplementary Fig. 3M, N).

Fig. 6

YY1 and GSC are upstream regulators of RCN2. (A, B) qRT-PCR analysis of RCN2 expression induced by YY1 or GSC interference transfection in CNE2 and 5-8 F cells. (C) Binding of YY1 or GSC to the RCN2 promoter was determined by dual-luciferase reporter assays in NPC cells. (D) ChIP assays in NPC cells were used to define the binding of both YY1 and GSC to the RCN2 promoter region. (E) qRT-PCR analysis of YY1 and GSC expression in cell lines. (F) Western blot (WB) analysis of YY1 and GSC in NP69 and NPC cells. (G, H) Analysis of the correlation between RCN2 expression level and YY1 or GSC levels in TCGA database using Spearman’s rank correlation analysis. (I, J) Rescue experiments were performed to verify the combination with WB of RCN2 levels. Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

ChIP: chromatin immunoprecipitation

3.7 YY1 and GSC synergistically regulate calcium flow-mediated mitochondrial apoptosis

The specific mechanism of YY1 and GSC transcriptional regulation of RCN2 is not yet fully understood. Thus, we first studied RCN2 expression after different transfections and found that the simultaneous knockdown of both YY1 and GSC resulted in a stronger downregulation of RCN2 than knockdown alone, concomitantly with a trend towards increase in cleaved caspase 3 levels, while RCN2 overexpression was unable to reverse the silencing role of YY1 and GSC simultaneously on RCN2 expression and cleaved caspase 3 levels (Fig. 7A-B; Supplementary Fig. 4A-D). These results illustrate that YY1 and GSC coordinately regulate the expression of RCN2 and apoptosis levels. Through a series of experiments, we verified that the knockdown of YY1 and/or GSC induced calcium overload in the mitochondria under stress without evident change in cytosolic calcium levels (Fig. 7C, D; Supplementary Fig. 4E, F), promoted apoptosis (Fig. 7E; Supplementary Fig. 5A), and dissipated the mitochondrial membrane potential (Fig. 7F; Supplementary Fig. 5B). Overexpressing RCN2 in the YY1 and/or GSC knockdown background reversed the mitochondrial dysfunction, and apoptosis (Fig. 7E, F; Supplementary Fig. 5A, B). Taken together, we can conclude that upstream transcription factors YY1 and GSC co-regulate mitochondrial calcium overload and mitochondrial apoptosis under stress.

Fig. 7

YY1 and GSC synergistically regulate calcium flow-mediated mitochondrial apoptosis. (A, B) Rescue experiments of knocked-down YY1 and/or GSC expression with overexpressed RCN2 or normal RCN2 to detect RCN2 and Cl-Cas3 expression by western blot (WB) analysis. (C) Quantitation of peak Rhod-2 fluorescence following rescue experiments to measure the levels of mitochondrial calcium. (D) Quantitation of peak Fluo-4 fluorescence. (E) Flow cytometry analysis were performed to measure the apoptosis rate and quantification of apoptosis ratio. (F) Rescue experiments followed by JC1 staining to measure the mitochondrial membrane potential and quantification of JC1 aggregates/monomers. CCK8 assays were carried out to measure cell proliferation upon the knockdown of YY1 and/or GSC combined with RCN2 overexpression or normal RCN2 expression in CNE2 (G) and 5-8 F (H) (two-way ANOVA). (I, J) Transwell assays were performed to assess migration ability in CNE2 and 5-8 F. (K) Western blot analysis of RCN2, YY1, GSC in CNE2 cells with anoikis resistance or not. (L) Flow cytometry analysis of CNE2 and 5-8 F cells after AR (Anoikis resistance) cell was induced and quantify cell apoptosis (M). Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

We then examined whether YY1 and GSCs had co-effects on the malignant phenotype. Silencing YY1 and/or GSC decreased malignant features via decreased cell proliferation (Fig. 7G, H; Supplementary Fig. 5C, D), migration, and invasiveness (Fig. 7I, J; Supplementary Fig. 5E, F). Based on the knockdown of YY1 and/or GSC, the overexpression of RCN2 reversed the malignant phenotype.

Then, the CNE2 and 5-8 F AR (Anoikis resistance) cell was induced, screened, and obtained, and we sought to determine whether AR cells were induced successfully. RCN2, YY1 and GSC expression were compared between bulk cells and CNE2 and 5-8 F AR. Finally, we found that their expression was increased in AR cells with high metastatic potential (Fig. 7K). Then, we confirmed successful CNE2 and 5-8 F AR induction by flow cytometry (Fig. 7L, M).

Collectively, these data suggest that YY1, in synergy with GSC, is an upstream regulator of RCN2 and plays a significant role in the development of the malignant biology of NPC.

3.8 RCN2, YY1 and GSC upregulation are associated with progression to more severe stages of NPC

We confirmed the clinical role of RCN2, GSC and YY1 using the ISH data with NPC tissue. High GSC, and YY1 expression were positively correlated with terminal stage disease (Fig. 8A-D). The expression of GSC correlated significantly with tumor node metastasis stages and the results showed that patients with high GSC expression had short overall or progression-free survival times (Fig. 8E-G). In addition, YY1 exhibited the same tendency (Fig. 8H-J). The results also revealed that RCN2, GSC, and YY1 were significantly correlated with NPC recurrence and distant metastasis (Fig. 8K-M). Of note, GSC or YY1 were both found to have positive protein levels correlation with RCN2 by analysis of three parallel tissue microarrays (Fig. 8N, O). We further evaluated the prognostic value and found that combined use of the three biomarkers were significantly associated with OS (Fig. 8P), and these findings suggest that high expression of RCN2 in combination with high expression of GSC and YY1 may serve as an important clinical biomarker of poor prognosis in patients with NPC.

Fig. 8

RCN2, YY1 and GSC upregulation are associated with advanced stage, recurrence, metastasis, lymph node metastasis, and poor clinical outcomes of NPC. (A) Representative GSC ISH staining (scale bar: 100 μm). GSC expression in different clinical stages (B), as the same YY1 (C, D), and lymph node metastasis (E). The GSC staining scores were divided into low expression (scores of 0–9) or high expression (scores of 10–16). Kaplan–Meier curves representing overall survival (F) and progression-free survival (G) were stratified (log-rank test). The lymph node metastasis (H), overall survival (I) and progression-free survival (J) of YY1 were stratified. Recurrence with or without distant metastasis in RCN2 (K), GSC (L) and YY1 (M) (one-way ANOVA). (N) Pearson correlation between RCN2 and GSC expression was analyzed. (O) Pearson correlation between RCN2 and YY1 expression was analyzed. (P) Overall survival with differential gene expression of the three biomarkers was assessed. Data show the mean ± SD of at least three independent experiments: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Finally, we concluded that overexpression of RCN2, GSC, and YY1 correlate with malignancy in NPC progression and potentially constitute as prognostic biomarkers.