CENPF/CDK1 signaling pathway enhances the progression of adrenocortical carcinoma by regulating the G2/M-phase cell cycle

All Formalin-fixed and paraffin-embedded (PPFE) samples were collected from patients diagnosed as ACC and adrenocortical adenoma and stored at 20–25 °C, from Jan 2011 to Jan 2021 at Taihe Hospital of Hubei University of Medicine, China. According to the pathological characteristics of two pathologists, all ACC patients were diagnosed and graded. Finally, 6 cases of ACC, 12 cases of adrenocortical adenoma (benign tumor), and 12 cases of normal adrenal cortex samples (adjacent to adrenocortical adenoma) were involved in this study. Details of all enrolled patients were listed in Additional file 1: Table S1.

Hematoxylin–eosin (HE) and immunohistochemistry (IHC) staining of PPFE samples were conducted following the steps in the manufacturer’s instructions. Specifically, 3 μm of ACC tissues, adrenocortical adenoma tissues, and normal adrenal cortex tissues were sliced from the PPFE. For HE, tissue sections were stained with a hematoxylin–eosin staining kit (E607318-0200, Sangon, Shanghai) for morphological observations. For IHC, all slices were dewaxed with xylene and rehydrated with graded ethanol. The 3% hydrogen peroxide was used to block endogenous peroxidase activity in methanol for 10 min. The sections were incubated with CENPF primary antibody (Additional file 2: Table S2) for 1 h at 37 °C, and incubated with HRP labeled the second antibody at 37 °C for 0.5 h, hematoxylin staining was performed at 37 °C for 30 s. Lastly, CENPF positive cells were tallied in five random high-power fields (400×), and the average positive cell ratio was calculated.

The Gene Expression Omnibus (GEO) of NCBI, an open-access genomics database, was widely used for the systematic analysis of cancer genes. Only two GEO datasets were retrieved for analysis of adrenocortical carcinoma, including GSE90713 and GSE19750. Details of those datasets were listed in Additional file 3: Table S3. The ACC data of TCGA were derived from Genomic Data Commons Data Portal, containing 79 ACC samples.

As an interactive online service platform, Gene Expression Profiling Interactive Analysis (GEPIA) is widely conducted to analyze gene expression of tumors and control samples from the TCGA and GTEx datasets [30]. Thus, the GEPIA dataset was utilized to measure the mRNA expression of CENPF between ACC and normal tissues, and different pathological stages.

The correlation between CENPF and clinicopathological features was studied based on the TCGA-ACC dataset. The79 ACC tissues in the expression matrix were divided into two groups, including 39 ACC tissues with CENPF low expression (CENPF^low) and 40 ACC samples with CENPF high expression (CENPF^high) by the median cutoffs. The expression groups were grouped by the 'ggrisk' package of R software (version 4.0.0) [31, 32]. Sanguini diagram was drawn via the ‘ggalluval’ package [33] for displaying the distribution of the gene expression in survival status, ages, genders, stages, and other clinical characteristics for ACC. Then, the univariate (uni-cox) and multivariate cox (multi-cox) regression analyses were analyzed and applied to develop the Nomogram. Then, ‘forestplot’ R packages were applied to present several parameters, including the p-values, HR, and 95% confidence interval (CI) by forest [34]. Based on the results of multi-cox analysis, The Nomogram was constructed to provide a graphical representation of the risk factors and calculate the 1-year, 2-year, and 3-year overall recurrence for an ACC patient via 'rms' R package [35].

The prognosis analysis, including overall survival (OS) and progression-free survival (PFS), was performed in ACC with CENPF^high and CENPF^low by ‘survival’ and ‘survminer’ packages [31]. The timeROC analysis was applied to predict the accuracy of CENPF and risk score by the 'timeROC' package [32], and the AUC threshold is set to 0.80.

The GSEA software 4.0.3 (Broad Institute, USA) was applied to probe and uncovered biological mechanisms in TCGA-ACC patients, including 39 CENPF^low and 40 CENPF^high ACC samples [36, 37]. The four predefined gene-sets, including 'c2.cp.kegg.v7.2.symbols.gmt', 'h.all.v7.2.symbols.gmt', 'c2.cp.go.v7.2.symbols.gmt', and 'c2.cp.biocarta.v7.2.symbols.gmt' were analyzed. Normalized enrichment scores (NES) were reckoned as the main GSEA statistic results. Statistical significance threshold were set as |NES|> 1, normalized p-values (NOM p-values) < 0.05 and FDR < 0.25. Additionally, analysis of protein–protein interaction (PPI) network was performed to explore the interaction between CENPF and related proteins based on The Search Tool for the Retrieval of Interacting Genes (STRING) database (version 11.5).

As one of the crucial indicators to speculate the effect of immunotherapy, the immune cell infiltration in tumors has become a research hotspot [38]. CIBERSORTx, the online service platform that estimated the abundance of multiple cell types in the mixed cell population by inputting a standardized gene expression matrix [39], was carried out to analyze immune cell infiltration based on the TCGA-ACC dataset. The results showed the abundance of 22 types of immune cells, including T cells, B cells, natural killer (NK) cells, monocytes, macrophages (Mφ), dendritic cells (DC), and granulocytes (mast cells, eosinophils, and neutrophils). 'Limma' package of R software was performed to normalize the expression matrix from the TCGA dataset. Then, the expression of LAG3, CTLA4, PD-1, PD-L1, and HAVCR2, closely related to immunotherapy, were explored in ACC [40]. To predict the anti-tumor effect of immunosuppressive agents, the 'ggstatsplot' package of R software was further conducted to analyze the correlation between the expression of CENPF and microsatellite instability (MSI) or tumor mutation burden (TMB) [41].

The gene-drug interaction network of CENPF was constructed via Comparative Toxicogenomics Database (CTD) [42] for chemotherapeutic drugs that could reduce or elevate the mRNA or protein expression levels of CENPF, and visualized by the OmicShare tools.

Significant differences were found in morphology and histology among normal adrenal cortex, adrenocortical adenoma (benign tumor), and ACC (malignant tumor). Firstly, gross specimen photography showed that for adrenocortical adenoma, the nodule section was golden and hard; For ACC, the capsule was intact or defective, some areas were gray red, and necrotic (Fig. 1a). Results of IHC-P showed that CENPF was mostly upregulated in ACC compared with adrenocortical adenoma samples or normal adrenal cortex tissues, but no significant difference was discovered between adrenocortical adenoma and normal tissues (p < 0.001, Fig. 1b, c). Furthermore, the results were validated in public databases, including GEO and TCGA datasets, for ACC patients. The mRNA expression of CENPF was dramatically overexpressed in GSE90713 (p < 0.01, Fig. 1d). But interestingly, there was no significant difference in the expression of this gene in GSE19750 (p = 0.11, Fig. 1e). This might be in case of too small sample size in normal adrenocortical tissues (n = 4). Meanwhile, the controversy attracted us to further explore. Moreover, correlation analysis by the GEPIA database showed that the mRNA expression of CENPF was evidently overexpressed in ACC tumors compared with that in normal ones (p < 0.001, Fig. 1f), and was positively correlated with the tumor stage for ACC (p < 0.001, Fig. 1g).

Additionally, Ki67, as a cell proliferation index, was investigated in ACC. According to the proportion of Ki67 positive cells in ACC, the relationship between the expression of CENPF and cell proliferation was further discussed. As result, the expression of CENPF, namely the proportion of CENPF positive cells, was positively correlated with the proportion of Ki67 positive cells (p = 0.008, R = 0.85, Fig. 2a, b). Moreover, the expression of CENPF was positively correlated with that of MKI67 expression based on the TCGA-ACC dataset (p = 0, R = 0.91, Fig. 2c). It indicated that the expression of CENPF was closely related to the cell proliferation in patients with ACC.

For figuring out the correlation between CENPF expression and clinicopathological parameters in ACC, all ACC patients were divided into two subgroups, including 39 CENPF^low and 40 CENPF^high samples according to the median cutoffs (Fig. 3a). As presented in Table 1, the expression of CENPF was not correlated with age, gender, laterality, and mitotane therapy (p > 0.05). Compared with CENPF^low, ACC patients with CENPF^high suffered manifestly advanced pathological stages (especially stage III and IV, p < 0.001). In tumor status, new events, or residual tumor, more ACC cases with CENPF^high were prone to lose the chance of surgery, relapse, or form residual tumor than CENPF^low (p < 0.001). Additionally, the CENPF^high group showed a higher Weiss score (especially Weiss score 6–9, p = 0.002). Moreover, the sanguini diagram described the distribution of CENPF in age, gender, pTNM stage, and status (Fig. 3b). Further, the correlation between CENPF and relevant clinical parameters, including age, gender, pTNM stage, etc. on the OS of ACC patients were identified via the uni-cox and multi-cox regression analysis. As result, the CENPF expression and pTNM stage were closely related to the OS of ACC patients in uni-cox analysis (Fig. 3c, all p < 0.05), and the CENPF expression and pTNM stage could be served as independent prognostic factors for ACC patients in multi-cox analysis (Fig. 3d, all p < 0.05). Finally, the survival rate of 1-year, 2- year, or 3- year in one ACC patient related to high CENPF expression was evaluated by Nomogram (Fig. 3e).

According to the survival rate analysis, upregulation of CENPF expression predicts poor OS (HR = 8.66, log-rank p = 1.8e−05; Fig. 4a) and PFS (HR = 4.11, log-rank p = 6.68e−05; Fig. 4c) in ACC. ROC curve analyses demonstrated that the AUC of CENPF at 1-year, 3-year, and 5-year for OS (Fig. 3b) and PFS (Fig. 3d) pointed out that the expression of CENPF has a good predictive effect on the prognosis of ACC (all AUC value > 0.75). Additionally, subgroup analyses indicated that overexpression of CENPF in ACC patients acted as a risk factor for 30-month, 60-month, and 120-month OS (all log-rank p ≤ 0.001, Fig. 4e–g).

To uncover the potential roles of CENPF in cancer-related signaling pathways, GSEA was carried out to explain the gene expression profiles of ACC samples with CENPF^low and CENPF^high. According to GSEA analysis of the KEGG pathway, CENPF^high patients were mainly enriched in cell cycle, DNA replication, nucleotide excision repair, and p53 signaling pathway, etc. (Fig. 5a, Additional file 5: Table S5). Then, GSEA enrichment analyses of Biocarta pathway and Hallmark description suggested that cell cycle (NES = 2.03, p = 0; especially G2-phase of mitosis, NES = 1.96, p = 0), G2/M checkpoint (NES = 2.01, p = 0), and E2F targets (NES = 2.04, p = 0) were mostly enriched in ACC patients with CENPF^high, which implied that CENPF might regulate cell cycle by interacting with E2Fs proteins (Fig. 5b, Additional file 6: Table S6, Additional file 7: Table S7, and Additional file 8: Table S8). Additionally, relevant studies have proclaimed that E2F1 played the role as a transcription factor of CENPF in the NCI-60 cell line [43] and regulated the cell cycle of the G2/M-phase transition [44, 45]. Thus, we supposed that CENPF may regulate the cell cycle by interacting with E2F1 in ACC. Venn graph suggested that 20 genes (CDK1, CDC25A, CDKN1A, E2F1, CCNB1, CDKN2D, ATR, TP53, ATM, CDK2, CCNE1, CDKN2A, TFDP1, CCND1, CDK6, CCNA1, CDK4, CDKN1B, CDKN2B and RB1) were co-expressed in at least two groups (Fig. 5c). It has been reported that CENPF mainly regulated the spindle separation in mitosis [46]. As listed in Fig. 5d and Additional file 7: Table S7, “spindle midzone” was the top 1 term in GESA analysis of GO terms. By screening the correlation, expression, and prognosis of these 20 genes, we picked out that CDK1, CCNB1and E2F1 were mostly overexpressed in ACC samples compared with normal ones (Fig. 5f, all p < 0.05), and positively relevant with the expression of CENPF (Fig. 5e, all p < 0.01). The overexpressed CDK1, CCNB1, and E2F1 also were related to the negative OS of ACC patients (Fig. 5g, all p < 0.01). In addition, the PPI network revealed that close interactions were found among CENPF, CDK1, and CCNB1 (Fig. 5h). Thus, it suggested that CENPF regulated the cell cycle by interacting with CDK1, E2F1, and CCNB1 in ACC. The overexpressed CENPF might play pivotal roles in the regulation of the G2/M-phase transition mediated cell cycle in the progression of ACC.

To further search the role of CENPF in ACC, CENPF siRNAs (siCENPF) were conducted via ACC cell line, human SW13 cells, as in vitro experiments. The expression of CENPF was overtly inhibited in siCENPF, compared with siNC Fig. 6a, b). In cell proliferation assay, CENPF interference evidently inhibited cell proliferation (Fig. 6c) and cell migration (Fig. 6d) in human SW13 cells after cells were transfected for 48 h. Inhibition of CENPF distinctly reduced adhesion between tumor cells and matrix (Fig. 6e) and cell invasion (Fig. 6f) in human SW13 cells. In addition, flow cytometry assay proclaimed that down-regulation of CENPF overtly intensified cell apoptosis, including early and late stages of cell apoptosis (18.46% vs 36.56%, Fig. 6g). Then, CENPF interference induced cell cycle arrest at the G2/M-phase transition in human SW13 cells (28.2% vs 6.4%, Fig. 6h). As analyzed above in Fig. 5, the expression of CENPF was closely related to CDK1 in ACC. Therefore, SiCDK1 was designed to probe the correlation between CENPF and CDK1 (Additional file 9: Fig. S1a, b). Down-regulation of CDK1 significantly suppressed cell proliferation (Additional file 9: Fig. S1c) and cell mobility (Fig. 6d) in human SW13 cells after cells were transfected 48 h. Down-regulation of CDK1 significantly restrained adhesion between tumor cells and matrix (Fig. 6e) and cell invasion (Fig. 6f) in human SW13 cells. Additionally, the flow cytometry experiment depicted that down-regulation of CDK1 overtly augmented cell apoptosis, including early and late stages of cell apoptosis (18.46% vs 28.5%, Fig. 6g). Lastly, CDK1 interference blocked the cell cycle at the G2/M-phase transition in human SW13 cells (28.2% vs 16.2%, Fig. 6h).

The molecular mechanism of CENPF was further explored in SW13 cells. As represented in Fig. 6, inhibition of the expression of CENPF resulted in the decrease of CDK1 expression Fig. 6i), while inhibition of CDK1 had no significant effect on the expression of CENPF (Fig. 6j). It suggested that CENPF might regulate the expression of CDK1. Furthermore, the GSEA enrichment analysis disclosed that the p53 signaling pathway might serve as a masked molecular mechanism of CENPF involved in the occurrence and development of human ACC (Fig. 5a). Therefore, the phosphorylation of p53 (P-p53), the expression levels of p53, p21, and Bax were detected by Western blotting in SW13 cells. As the interference of CENPF and CDK1 by siRNA, the expression levels of P-p53, p21, and Bax were significantly decreased compared with the siNC in SW13 cells (Fig. 6i, j). Therefore, inhibition of CENPF downregulated the expression of CDK1, suppressed the phosphorylation of p53, and altered downstream protein levels, including apoptosis and cell cycle-related proteins.

Next, the available chemical drugs and treatment strategies, including immunotherapy, that may be favorable for the treatment of ACC were further explored. Firstly, the immune cell infiltration in ACC with CENPF^high and CENPF^low were analyzed to investigate immunotherapy. All the 79 ACC tissues from the TCGA database were divided into two groups, including 40 CENPF^high and 39 CENPF^low tissues by the median value. CIBERSORTx was applied to analyze immune cell infiltration, including 22 cell types with low and high expression of CENPF. p < 0.05 was statistically significant. The normalized by 'Limma' packages and then filtered groups including 7 CENPF^low and 6 CENPF^high ACC tissues were identified for further analysis. The expression of 22 immune cell populations for each sample was presented as heatmaps (Fig. 7a). The immune enrichment score was dramatically various between 6 CENPF^high and 7 CENPF^low samples. As shown in Fig. 7b, follicular helper T cells, M0 macrophages, Eosinophils were mainly enriched in ACC with CENPF^high, compared with the CENPF^low group. However, the immune enrichment score of gamma delta T cells, monocytes, and M2 macrophages were much lower in CENPF^high, compared with the CENPF^low group. The principal component analysis (PCA) showed that manifest difference was probed in immune infiltration between CENPF^high and CENPF^low ACC samples (Fig. 7c). It might indicate that the immune heterogeneity of the CENPF was significant in ACC patients. Otherwise, the mRNA expression of LAG3, CTLA4, PD-1, PD-L1, and HAVCR2, tightly related to tumor immunosuppression or tolerance, were not significant differences in ACC with CENPF^high, compared with CENPF^low (Fig. 7d, all p > 0.05). Moreover, the correlation between the expression of CENPF and MSI or TMB was analyzed to unraveled that CENPF expression was significantly associated with MSI score (Fig. 7e, p = 0.034) and TMB score (Fig. 7f, p < 0.01) in ACC. It showed that overexpression of CENPF might augment the accumulation of abnormal gene mutation, namely TMB, in the process of DNA replication in ACC. Then, the gene-drug interaction network indicated that a variety of drugs could affect the expression levels of CENPF in mRNA or protein levels (Fig. 7g). For example, Cisplatin, Sunitinib, and Etoposide could restrain CENPF expression level while Paclitaxel and Genistein could induce CENPF expression level. Generally, all these CENPF inhibitors were deemed as potential targets for the treatment of ACC.