Knockdown of CENPF inhibits the progression of lung adenocarcinoma mediated by ERβ2/5 pathway

Many studies have reported that estrogen (E2) promotes lung cancer by binding to nuclear estrogen receptors (ER), and altering ER related nuclear protein expressions. With the GEO database analysis, Human centromere protein F (CENPF) is highly expressed in lung adenocarcinoma (LUAD), and the co-expression of CENPF and ERβ was found in the nucleus of LUAD cells through immunofluorescence. We identified the nuclear protein CENPF and explored its relationship with the ER pathway. CENPF and ERβ2/5 were related with T stage and poor prognosis (P<0.05). CENPF knockout significantly inhibited LUAD cell growth, the tumor growth of mice and the expression of ERβ2/5 (P<0.05). The protein expression of CENPF and ERβ2/5 in the CENPF-Knockdown+Fulvestrant group was lower than CENPF- Negative Control +Fulvestrant group (P=0.002, 0.004, 0.001) in A549 cells. The tumor size and weight of the CENPF-Knockdown+Fulvestrant group were significantly lower than CENPF- Negative Control +Fulvestrant group (P=0.001, 0.039) in nude mice. All the results indicated that both CENPF and ERβ2/5 play important roles in the progression of LUAD, and knockdown CENPF can inhibit the progression of LUAD by inhibiting the expression of ER2/5. Thus, the development of inhibitors against ERβ2/5 and CENPF remained more effective in improving the therapeutic effect of LUAD.


INTRODUCTION
Lung cancer is one of the most commonly diagnosed cancers and a leading cause of cancer related mortality [1]. The global incidence and mortality of lung cancer is increasing significantly [2]. In addition to environmental exposures such as smoking, the growth factor pathway and hormonal regulation also play critical roles in the carcinogenesis of lung cancer [3]. AGING highly expressed in lung adenocarcinoma (LUAD). Furthermore, the expression of CENPF and ERβ2/5 in LUAD patients have been shown to be correlated to TNM staging, providing a basis for exploring the interactions between CENPF and ERβ2/5. Rattner et al. demonstrated that CENPF is involved in mitosis and tumor proliferation [7]. The full-length molecular weight of CENPF is 367 KDa and contains 3,210 amino acids [8]. In prostate cancer, CENPF has been shown to predict survival and tumor metastasis [9]. CENPF is directly associated with disease outcomes after undergoing gene amplification [10]. However, the role of CENPF in the progression of LUAD still remained unclear.
A large number of studies have shown that estrogen (E2) promotes the progression of lung cancer by binding to nuclear ERs [11]. Our previous studies have shown that among the five types of ERs, lung cancer tissues express found the ERβ1/2/5 [12,13]. As a fulllength fragment of ERβ subtype, ERβ1 is responsible for the action of other subtypes [14], and so the role of ERβ2/5 in the progression of LUAD remained the main focus of our current research.
Based on estrogen gene signaling pathway, a high coexpression of CENPF and ERβ2/5 showed association with the clinicopathological features and prognosis of LUAD patients. CENPF is hypothesized as one of the key nuclear proteins in estrogen gene signaling pathway. Therefore, this study mainly explored the relationship between CENPF and ERβ2/5, and also explored their role alone in the development of LUAD. This study will provide a better understanding of ERs gene signaling pathway and improve the prognosis of LUAD patients.

Bioinformatics analysis of lung cancer datasets and the determination of CENPF
Differential genes with similar expression pattern in each dataset (GSE19804, GSE30219, GSE32863, GSE63459) were used ( Figure 1A-1C, Supplementary Figure 1A-1I, β=5, 6, 5, 7) [15]. The module genes obtained by the four datasets showed significant association with TNM staging of lung cancer, which included the brown module of GSE19804 (n=185), the turquoise module of GSE30219 (n=413), the yellow module of GSE32863 (n=63) and the yellow module of GSE63459 (n=160) ( Figure 1B Table 1). By overlapping the module genes, five key genes including CENPF, CDC20, TOP2A, CCNB2 and BIRC5 were obtained ( Figure 1D-1G). Finally, based on the central degree of the five key genes in different datasets and relevant literature [9,16], the hub gene was identified as CENPF.

CENPF is highly expressed in LUAD and negatively correlated with the prognosis of LUAD patients
CENPF was highly expressed in LUAD when compared to normal lung tissues (Figure 2A-2H). The expression of CENPF was positively correlated to the TNM staging of LUAD (P<0.01, Figure 2I-2M). Additionally, high expression of CENPF was negatively correlated with overall survival and disease-free survival in LUAD patients (P=0.01, 0.027, 0.0267, Figure 2N-2P). The result of RNA-Seq included 1515 high expressed genes and 1370 low expressed genes in LUAD patients ( Figure 2Q). CENPF was included in high expressed genes ( Figure 2R, P<0.05) and was associated with poor prognosis in LUAD patients ( Figure 2S, 2T).

Expression of CENPF, ERβ, ERβ1, ERβ2 and ERβ5, and its relationship with TNM staging and prognosis of LUAD patients
The expressions of CENPF, ERβ, ERβ1, ERβ2 and ERβ5 in different TNM stages of LUAD and benign primary lesions (BPL) were examined ( Figure 3A, Supplementary Figure 2A-2D). The results revealed that CENPF, ERβ, ERβ2 and ERβ5 were highly expressed in LUAD and showed a positive correlation with TNM staging and T grade of LUAD patients (P<0.001, Figure 3B, 3C), but showed no association with nodal involvements ( Figure 3D). Moreover, analysis of high expression of CENPF in LUAD patients showed significant association of CENPF with shorter survival rate (Table 1).

CENPF knockdown inhibits biological effects of LUAD cells
Compared with BEAS-2B and other LUAD cells, CENPF was highly expressed in A549 and H1299 cells (P<0.05, Figure 4A; The corresponding gray value are shown in Supplementary Figure 3A, 3B). RT-PCR and cellular immunofluorescence confirmed that CENPF knockdown (KD) was seen in 70% in A549 and H1299 cells ( Figure 4B; Supplementary Figure 3C). Cell proliferation of CENPF-KD group was significantly weaker than the control (NC) group from day 3 in stable CENPF-deficient cell lines A549 and H1299 (P=0.007, 0.000, Figure 4C, 4D). At the same time, cells in CENPF-KD group demonstrated less Ki67 staining than NC group (Supplementary Figure 3D). In addition, the invasion and migration of A549 cells in the CENPF-KD group were significantly decreased (P=0.000, 0.000, Figure 4E; Supplementary Figure 3E, 3F) while AGING     Figure 4F; Scratch experiment also demonstrated similar results (P=0.000, 0.000, Figure 4H; Supplementary Figure  3I). In A549 cells, the cell percentage and DNA content were significantly increased in the G1 phase in the CENPF-KD group when compared with NC group (P=0.011, Figure 4I; Supplementary Figure 3J). At the same time, the expression of CCND1, CDK2 and CDK4 was significantly lowered in CENPF-KD group (P=0.022, 0.001, 0.002, Figure 4J; The corresponding gray value are shown in Supplementary Figure 3M). Similar results were obtained in the stable CENPFdeficient cell line H1299 (P<0.05, Supplementary Figure 3K-3M). The CENPF-KD group also showed a significant increase in apoptosis when compared with NC group (P=0.001, 0.001, Figure 4K; Supplementary Figure 3N).
The effect of CENPF knockdown on the expression of ERβ2/5 was examined in vitro in CENPF-NC and CENPF-KD groups. The protein expression of ERβ2/5 in CENPF-KD group was significantly lower than the CENPF-NC group (PA549=0.013, 0.000; PH1299=0.006, 0.002, Figure 5J; The corresponding gray value are shown in Supplementary Figure 5A). We further explored the effect of CENPF knockdown on the expression ERβ2/5 under the action of E2 and Ful in vitro. The results revealed that the protein expression of CENPF and ERβ2/5 in CENPF-KD+Ful group was significantly lower than CENPF-NC+Ful group (P=0.002, 0.004, 0.001, Figure 5K; The corresponding gray value are shown in Supplementary Figure 5C) in A549 cells. Similar results were obtained in stable CENPF-deficient cell line H1299 (P < 0.01, Figure 5J, 5K; Supplementary Figure 5C).

Knockdown of CENPF can inhibit ERβ2/5 pathwaymediated tumor growth in vivo
The tumor weight and size in the CENPF-KD group were lower than NC group (P<0.001, 0.001, Figure 6A-6C). The results of immunohistochemistry also showed that the expression of CENPF and ERβ2/5 was significantly decreased in CENPF-KD group when compared to the NC group (P=0.000, 0.000, 0.000, Supplementary Figure 5D, 5E).
Similar to the in vitro experiments, the effect of CENPF knockdown on the expression of ERβ2/5 was examined in CENPF-NC and CENPF-KD groups, and examined under the action of E2 and Ful in vivo. The protein expression of CENPF and ERβ2/5 in CENPF-KD group was lower than CENPF-NC group in vivo (P=0.024, 0.020, 0.003, Figure 6I; The corresponding gray value are shown in Supplementary Figure 5B). The protein expression of CENPF and ERβ2/5 was significantly decreased in CENPF-KD+Ful group when compared with CENPF-NC+Ful group in mice tumor tissues (P=0.020, 0.004, 0.002, Figure 6J; The corresponding gray value are shown in Supplementary Figure 5F).

DISCUSSION
Lung cancer is one of the most common malignancies and poses as a major health crisis globally. Targeted therapy for cell proliferation-related pathways and hormone therapy for lung cancer are considered important treatment modalities [17]. ER signaling pathways mainly included gene signaling and non-gene signaling pathways [18]. Among them, the gene signaling pathway involves the integration of estrogen with ER and the association with estrogen response element (ERE) to promote the recruitment of RNA polymerase II and regulate gene transcription. During this process, different combinations of synergistic activators determine the specificity of ER for activating the target genes [5]. Therefore, the key genes involved in TNM staging of LUAD in the four lung cancer datasets (GSE19804, GSE30219, GSE32863, GSE63459) were analyzed by the "WGCNA" R package ( Table 1). Ultimately, CENPF was found by overlapping the key genes ( Figure 1D-1G).
High expression of CENPF was shown to correlate with the malignant progression and poor prognosis in patients with LUAD. Studies have shown that CENPF has been up-regulated in a variety of malignancies, including nasopharyngeal carcinoma, esophageal squamous cell carcinoma and prostate cancer [19][20][21]. The expression of CENPF in LUAD was detected by the Oncomine database. Similarly, RNA-Seq data showed high expression of CENPF in LUAD ( Figure 6G, 6H), indicating that high expression of CENPF in LUAD might be related with its occurrence. At the same time, studies have shown that CENPF mediates mitosis and cell proliferation [22]. The expression of CENPF and DNA content were significantly reduced ( Figure 4I), and the cell proliferation was significantly decreased in stably CENPF-deficient LUAD cells ( Figure 4C, 4D). These results indicated that low expression of CENPF inhibited proliferation of tumor cells in LUAD. Other studies showed that forkhead box M1 (FOXM1) and CENPF synergistically promoted malignant progression AGING   Figure 5F). *P < 0.05. P values were calculated with two-tailed unpaired Student's t-test, or one-way analysis of variance. AGING and poor prognosis of prostate cancer [9]. It has been reported that chicken ovalbumin upstream promotertranscription factor 2 (COUP-TFII) promoted metastasis of prostate cancer through signal transduction of FOXM1 and CENPF [19]. We speculated that the dysregulation of miRNA-COUP-TFII-FOXM1-CENPF axis can be associated with malignant progression, poor prognosis and metastasis in LUAD. Our study revealed that LUAD cells showed a significant reduction in invasion and metastasis after CENPF knockdown ( Figure 4E-4H). The expression of CENPF was significantly related to TNM staging of LUAD ( Figure 2I-2L; Figure 3B, 3C). In addition, based on the clinical data of LUAD patients obtained from GEPIA and TCGA, LUAD patients with high expression of CENPF was correlated to a poor prognosis ( Figure 2N-2P). These results indicated that abnormal expression of CENPF was significantly associated with TNM staging, poor prognosis, and malignant metastasis of LUAD.
In addition, our previous study reported that E2 promoted the progression of lung cancer by binding to ERβ [23]. The biological effects of ERβ in E2 varies based on the targeted organ tumors, including breast, cervical, and prostate cancers [13,24]. Our previous findings indicated that ERβ2/5 are expressed in lung cancer [12]. Our study results revealed that ERβ2/5 is also highly expressed in LUAD patients (Supplementary Figure 2B, 2D). In addition, ERβ2/5 showed high positive association with the TNM staging of LUAD ( Figure 3B, 3C).
CENPF knockdown inhibited the progression of LUAD mediated by the ERβ2/5 pathway. High expression of CENPF and ERβ2/5 is associated with the development of LUAD. Invasion, migration and proliferation of LUAD cells in the CENPF-KD+E2 group showed significant reduction in vitro when compared to the CENPF-NC+E2 group ( Figure 5C-5H, Supplementary Figure 4A-4F). In the in vitro system, the protein expression of ERβ2/5 in CENPF-KD+E2 group was significantly lower than CENPF-NC+E2 group ( Figure  5K; Supplementary Figure 5C). In order to eliminate the effects of endogenous estrogen due to gender differences, in vivo experiments were only conducted in male mice. Consistent with the in vitro experiment results, the expression of ERβ2/5 protein in CENPF-KD+E2 group was significantly lower than that in CENPF-NC+E2 group ( Figure 6J; Supplementary Figure 5F). From these results, we confirmed that the knockdown of CENPF inhibited the progression of LUAD mediated by ERβ2/5 pathway both in vitro and in vivo. This is another important point regarding the mechanism of CENPF, except for FOXM1 [9] and COUP-TFII [19].
Taken together, these findings indicated that both CENPF and ERβ2/5 are highly expressed in LUAD cells and their expression is associated with TNM staging and prognosis in LUAD patients. CENPF knockdown inhibited proliferation, invasion and metastasis of LUAD cells mediated by ERβ2/5 pathway. Thus, the development of inhibitors against the ERβ2/5 subtypes and CENPF can have great therapeutic impact in LUAD. However, there are some shortcomings in this study that should be acknowledged. First, due to the large molecular weight of CENPF, plasmid construction can easily be created off target, thus LUAD cell lines expressing CENPF can be difficult to construct. Second, the number of specimen used for RNA-Seq is limited. These issues are key points that should be targeted for future studies.

Tissue specimens of patient and cell culture
This study was approved by the Ethics Review Committee of Wuhan University. Tissue specimens from 56 LUAD cases and 10 benign pulmonary lesions (BPL) cases who underwent surgery from April 2014 to July 2017 were collected for tissue chip. One pair of LUAD and peri-cancerous tissues were collected for RNA-Seq and three pairs of tissues for profiling protein were obtained from the Department of Thoracic and Cardiovascular Surgery, Zhongnan Hospital of Wuhan University. The tissue chip was customized by Shanghai Core Biotech Co., Ltd. [11]. The isolated tissue samples were immediately stored in liquid nitrogen and sent to Huada Gene (Beijing) for RNA-Seq analysis [4].

Lung cancer patient data set
Training data sets (GSE19804, GSE30219, GSE32863, GSE63459) based on the Affymetrix platform (Affymetrix HG-U133 Plus 2.0 array and HG-U133A array) and corresponding clinical information were retrieved from the Gene Expression Omnibus. Two nonsmall cell lung cancer (NSCLC) genome-wide expression profiles were extracted from GSE19804 (including 60 paired tumors and normal tissues) and GSE30219 (including 293 tumors and 14 non-tumor tissues). Two LUAD genome-wide expression profiles were extracted from GSE32863 (including 58 paired tumors and normal tissues) and GSE63459 (including 65 tumor tissues). AGING

Analysis and verification of hub gene
The data sets of GSE19804, GSE30219, GSE32863 and GSE63459 were used to construct co-expression networks and clinical functioning related modules. The genes were screened according to the false discovery rate (FDR) <0.05 and | log2 fold change (FC) |> 1.5. Next, a weighted gene co-expression network analysis (WGCNA) package was used to construct a coexpression network [25,26]. Finally, the hub gene was selected based on the degree of centrality using the Venn diagram to obtain key genes.
The raw data of RNA-Seq was subjected to quality control, and then mapped with STAR [27] to obtain differential genes. The screening criteria for differential genes were abs(log2FC) > 1 and p value < 0.05.
The Oncomine, Gene Expression Profiling Interactive Analysis (GEPIA) and clinical data from The Cancer Genome Atlas (TCGA) database were used to verify the expression, progression and prognosis of hub gene.

Immunohistochemistry
The detailed steps for conducting immunohistochemistry was described previously [11]. CENPF (ab5) and ERβ (ab3576) were purchased from Abcam. ERβ1 (MCA1974ST), ERβ2 (MCA2279GT) and ERβ5 (MCA4676T) were purchased from AbDSerotec. The specificity of the above antibodies was confirmed by numerous laboratories including ours [11,28]. Immunohistochemical method to analyze the optical density was calculated by Image-Pro Plus software.

Cell culture experiment
The sense sequence of CENPF knockdown (KD) and negative control (NC) were integrated into the pWSLV-sh08-GFP vector for transfection of A549 and H1299 cells. The stable CENPF-deficient A549 and H1299 cells were immunofluorescently labeled with anti-CENPF and anti-ERβ [31]. Cell apoptosis assay was conducted using TdT-mediated dUTP-biotin nick end labeling test (TUNEL, Roche Applied Science, Germany) according to the manufacturer's instructions [32]. MTT [33], invasion, migration, scratch and cell cycle experiments were used to evaluate the effects of LV-CENPF sh and LV-NC on the biological function of A549 and H1299 cells [34,35].

Xenograft mouse model
Male nude mice were obtained from Beijing HFK Bioscience Co., Ltd., China. Mice were housed in specific-pathogen free environment for one week, and then were subcutaneously injected with 100 μL of 8x10 6 LV-CENPF sh or LV-NC cells. Tumor size was measured every three days (tumor size = length × width 2 × 0.5 mm 3 ). When the tumor size has reached 80-120 mm 3 , the mice were injected with E2 (0.036 mg/ml, purity 98%, Sigma) or fulvestrant (Ful, 0.800 mg/ml, Sigma) subcutaneously twice a week (6 weeks) [11]. All mice were then sacrificed on day 45, wherein the xenograft tumors were harvested and the tumor weight and size were measured [36]. Tumor tissues were fixed in 4% paraformaldehyde or frozen with liquid nitrogen and stored at -80° C.