Abstract
The long noncoding RNA nuclear-enriched abundant transcript 1 (NEAT1) has been shown to regulate multiple cancer-related cellular activities including cell proliferation, apoptosis, and migration. In this study, we confirm that repression of NEAT1 induces DNA damage, disturbs the cell cycle, and arrests the proliferation of prostate cancer cells. By taking advantage of the prostate cancer tumor transcriptome profiles from The Cancer Genome Atlas, our data-mining pipeline identified a series of transcription factors (TF) whose regulatory activities on target genes depended on the level of NEAT1. Among them was putative TF CDC5L, which bound directly to NEAT1. Silencing NEAT1 in prostate cancer cells repressed the transcriptional activity of CDC5L, and RNA-seq and ChIP-seq analyses further revealed a handful of potential targets of CDC5L regulated by NEAT1 expression. One target of CDC5L, ARGN, mediated the strong phenotypic consequences of NEAT1 reduction, including DNA damage, cell-cycle dysregulation, and proliferation arrest. In summary, we have established the requirement of the CDC5L–AGRN circuit for the essential oncogenic role of NEAT1 in prostate cancer cells.
Significance: An integrative methodology uncovers CDC5L–AGRN signaling as critical to the tumor-promoting function of long noncoding RNA NEAT1 in prostate cancer cells. Cancer Res; 78(15); 4138–49. ©2018 AACR.
Introduction
Nuclear enriched abundant transcript 1 (NEAT1), which is a long noncoding RNA (lncRNA), has been shown to play key roles in a variety of cancer-related cellular activities, including cell proliferation, apoptosis, DNA damage, invasion, and migration (1–4). However, the detailed mechanisms have not been fully elucidated. NEAT1 is an essential component of the paraspeckle (5, 6), which is a nuclear speckle near the cell nucleolus. Specifically, NEAT1 binds directly to the core proteins in the paraspeckle, such as SFPQ (7) and NONO/p54 (8), and the major function of this RNA-protein complex is believed to be the regulation of gene expression at the transcriptional (7, 9, 10) and post-transcriptional (11–13) levels. For example, NEAT1 and the paraspeckle were shown to be enriched around gene transcription start and termination sites in breast cancer cells (9), although their involvement in transcription remains unclear. In addition, previous high-throughput profiling studies also revealed more proteins that interact directly with NEAT1 but are not essential for the architecture of the paraspeckle (9, 14). The functions of these interactions remain largely unknown.
Although substantial abnormalities of NEAT1 have been observed frequently in various cancer-related contexts (1–3, 15–18), the exact role of NEAT1 in tumorigenesis is still debated, and the potential underlying machineries remain largely unclear. For example, NEAT1 has been identified as a direct transcriptional target of p53 in multiple studies; however, those studies derived seemingly contradictory functions of NEAT1 (4, 19). Being proposed as an oncogenic lncRNA, NEAT1 was shown to serve as a negative feedback signal that attenuates p53 activity by preventing DNA damage accumulation (4). In contrast, NEAT1 was also characterized as a major component of the p53-mediated suppression of transformation and cancer development (19). In the present study, we confirmed that NEAT1 is essential for the proliferation and tumorigenesis of the castration-resistant prostate cancer (CRPC) cell lines, including PC3, which bears a hemizygous p53 deletion and mutation, and DU145, which carries two nonsynonymous mutations of p53. We showed that knocking down NEAT1 resulted in significant DNA damage and suppression of cell proliferation and tumor growth.
Mechanistic investigations for many lncRNAs has been challenging and the studies remain limited, partly due to lack of prior knowledge and difficulties in generating plausible hypotheses to start with. Regulation of gene transcription has been one of the major known functions of multiple lncRNAs (20, 21). To elucidate the machinery responsible for the essential function of NEAT1 in prostate cancer cells, we applied an integrative data-mining strategy to identify potential transcription factors (TF) whose transcriptional activity depends on the level of NEAT1. This analysis based on transcriptome profiling data from about 500 prostate tumors in The Cancer Genome Atlas yielded a series of known TFs and other DNA-binding proteins, including CDC5L, a putative TF shown to bind directly to NEAT1 by the CHART-MS assay (9). CDC5L, cell division cycle 5-like protein, is a critical element for mitotic progression. Inhibition of CDC5L resulted in arrest of the mitotic and activated spindle assembly checkpoint (22). Moreover, CDC5L was involved in the regulation of ATR-associated cell-cycle progression, and silencing CDC5L affected not only the mitotic checkpoint but also the S phase checkpoint by interacting with ATR (23). Guided by our data-mining results, our experimental studies confirmed the regulation of CDC5L by NEAT1 and identified the target gene of CDC5L, AGRN, which was modulated by NEAT1. We finally proved that this transcriptional regulatory circuit, NEAT1–CDC5L–AGRN, is essential for proper tumor cell growth, and repression of the pathway causes DNA damage and potent arrest of the cell cycle and proliferation. Therefore, we are proposing a novel transcription regulation machinery that mediates the oncogenic role of NEAT1 in prostate cancer cells.
Materials and Methods
Cell culture
PC3 and DU145 cells were purchased from the ATCC. The ATCC has performed authentications for both cell lines by short tandem repeat DNA profiling, morphology, and karyotyping. PC3 cells were cultured in RPMI-1640, and DU145 cells were cultured in Eagle's Minimum Essential Medium, which were both supplemented with 10% FBS (HyClone). The cells were maintained in a humidified incubator with 5% CO2 at 37°C. Cells after the second passage and before the 10th passage were used for the experiments. All cells were routinely tested as Mycoplasma-free with the Mycoplasma Detection Kit (Bimake, B39032).
Transfection and quantitative PCR
All siRNAs in this study were purchased from GenePharma (Shanghai, China). The antisense and sense sequences of the siRNAs are the following. siNEAT1_2: UUAAGAUUGAGAUUUACCCag and GGGUAAAUCUCAAUCUUAAtt, siNEAT1: GGAAAGUUUCUAAGCAACUUCUCACUU and GUGAGAAGUUGCUUAGAAACUUUdCdC, siCDC5L-1: UCAGACAAAAGUGUACUGGaa and CCAGUACACUUUUGUCUGAtt, siCDC5L-2: AAACUUGACUGUAGCAUUCtt and GAAUGCUACAGUCAAGUUUtt, siAGRN-1: UCAGUUCAAAGUGGUUGCUct and AGCAACCACUUUGAACUGAtt, siAGRN-2: AUAAAUCGCACGUGCUCCUGC and GCAGGAGCACGUGCGAUUUAT, siB2M: UACAAGAGAUAGAAAGACCAG and CUGGUCUUUCUAUCUCUUGUA, siLMNA: AUCUCAUCCUGAAGUUGCUUC and GAAGCAACUUCAGGAUGAGAU, siNC: ACGUGACACGUUCGGAGAA and UUCUCCGAACGUGUCACGU. A total of 10 nmol/L siRNA was transfected with the Lipofectamine RNAiMAX Reagent (Invitrogen) according to the manufacturer's instructions. For plasmid transfection, plasmids were transfected with Lipofectamine 2000.
Total RNA was isolated using the RaPure Total RNA Micro Kit (MaGen). Then, 1 μg of total RNA was reverse transcribed using a High-Capacity cDNA Reverse Transcription Kit (Invitrogen). Quantitative PCR was carried out with SYBR Green Master Mix (Invitrogen). The forward and reverse primers are the following. NEAT1_2: TTCACCTGCTCTGGCTCTTG and GCCAGGCACCGTGTTATACT, NEAT1: GACCTCTCACCTACCCACCT and ATGCCCAAACTAGACCTGCC, CDC5L: TCCGTTTAGGGTTGTTGGGC and TCTCGTATGGCCTGCTTTCG, AGRN: ACACCGTCCTCAACCTGAAG and CCAGGTTGTAGCTCAGTTGC, LMNA: ATGAGGACCAGGTGGAGCAGTA and ACCAGGTTGCTGTTCCTCTCAG, B2M: AGTATGCCTGCCGTGTGAA and AGCAAGCAAGCAGAATTTGGA.
Cell proliferation assay
The cells transfected with control or gene knockdown siRNAs were cultured in appropriate plates. The IncuCyte live-cell imaging and analysis system (ESSEN Bioscience) was used to monitor long-term cell growth and morphology changes. Cell proliferation was quantified by measuring the occupied area (% confluence) in the cell images over time.
Western blotting and antibodies
The cells were lysed with RIPA lysis buffer (Solarbio) supplemented with proteinase inhibitor (Solarbio). Protein concentrations were quantified using the BCA protein assay (Pierce). The cell lysate containing 20 to 30 μg proteins was heat denatured and subjected to SDS-PAGE, followed by transfer to a negative control membrane. The membrane was incubated with the primary and secondary antibodies and visualized with the SuperSignal West Pico Chemiluminescent HRP substrate (Thermo). The primary CDC5L antibody and secondary antibodies Alexa Fluor 488 (Donkey anti-Mouse), Alexa Fluor 594 (Donkey anti-sheep), and Alexa Fluor 594 (Goat anti-Rabbit) were purchased from Abcam. The ACTB antibody and secondary antibodies Goat to Rabbit and Goat to Mouse were from Bioss. The Phos-H2AX (Ser139) antibody was from Ruiying Biological, AGRN (D2) antibody from Santa Cruz Biotechnology, secondary antibody Alexa Fluor 488 (Donkey anti-Rabbit) from Life Technology, and PSPC1 and SFPQ antibodies from Sigma.
DNA damage quantification
Cells transfected with siRNA were cultured with 5% CO2 at 37°C for 24 hours. Genomic DNA was isolated with the AxyPre Multisource Genomic DNA Miniprep Kit (AxyGEN). A total of 0.1 μg/μL of purified genomic DNA was mixed with 5 μL of ARP solution and incubated at 37°C for 1 hour to tag the DNA AP site. The AP-labeled DNA was isolated and measured according to the manufacturers' instructions (Biovision).
TUNEL analysis
The siRNA-transfected cells were washed twice with 1× PBS solution, fixed with 4% paraformaldehyde for 30 minutes, and then washed twice with 1× PBS. The fixed cells were permeabilized with 0.3% Triton X-100 in PBS solution for 5 minutes and then rinsed with PBS solution. After preparing the TUNEL measurement solution (Beyotime), the fixed cells were incubated with the appropriate TUNEL measurement solution at 37°C for 30 minutes and then photographed with a microscope.
Flow cytometry assay
Cells transfected with siRNAs were harvested and centrifuged at 1,200 rpm, 4°C for 5 minutes. The supernatant was removed, and the cells were re-suspended in 1× PBS solution. The cells were then re-centrifuged at 12,000 rpm and 4°C for 5 minutes and then re-suspended in 0.3 mL of 1× PBS solution. Then, the cell suspension was gently mixed with 0.7 mL of 100% ethanol and stored at −20°C overnight. On day 2, the cells were washed once with 1x PBS solution and then centrifuged. Then, the cell pellet was re-suspended in 0.25 mL of 1× PBS solution, and added 5 μL of 10 mg/mL RNAse A (the final concentration was 0.2–0.5 mg/mL) for incubation at 37°C for 30 minutes. After 30 minutes, 62.5 μL of a 50 μg/mL propidium iodide (PI) solution was added; the cells were then kept in the dark at 4°C until the time of analysis.
Assessment of apoptosis was done by flow cytometry as well, with the Alexa Fluor 488 Annexin V/Dead Cell Apoptosis Kit (Invitrogen). Specifically, cells transfected with siRNA for 48 hours were washed with 1× PBS solution. Cells were harvested and re-suspended with 1× Annexin-binding buffer (Invitrogen) to make final cell concentration of 1 × 106 cells per ml. 100 μL of cell suspension was transferred to a new tube, and 5 μL of Annexin V solution and 5 μL of PI solution were added to the cell suspension. Cells were incubated at room temperature for 15 minutes with protection from light, followed by addition of 400 μL 1× Annexin-binding buffer for a flow-cytometry assay.
Dual-luciferase reporter assay
The CDC5L- or CMV-motif-luciferase plasmid was constructed as described previously (24). The two plasmids, 50 ng of CMV- or CDC5L-motif-Luc (firefly) and 1 ng of pRL-TK (Renilla) were co-transfected into cells with 10 nmol/L of siRNA negative control (siNC) or siNEAT1_2 at 96-well format. After 24 hours of transfection, the cells were washed with 1× PBS solution and analyzed with a dual-luciferase reporter assay kit (Promega: E1910) as the manufacturer's instruction. The luciferase activity was measured using ELISA and normalized to Renilla luciferase.
RNA fluorescence in situ hybridization and immunofluorescence staining
A 0.5 kb fragment of NEAT1_2 was amplified using the relevant primer sets (GCCTTCATTTATCCTCAGATCAGGTGAG and GTGTCTTTCATTTCATGCCCGCACTGCAC). The amplified DNA clone was generated with pGEM-T (Promega). The NEAT1_2 probe was synthesized with DIG RNA labeling Mix (Roche) and SP6 polymerase (Thermo), according to the manufacturers' instructions. The RNA–FISH assay was performed as described previously (25).
For immunofluorescence staining, cells were fixed with 4% formaldehyde solution for 15 minutes and then washed thrice with 1× PBS solution at room temperature for 15 minutes. Subsequently, the slides were permeabilized with PBT (0.1% Triton X-100 in 1× PBS solution) and then washed with 1× PBS solution at room temperature for 15 minutes. The slides were incubated with blocking solution (1% BSA in PBT) at room temperature for 1 hour and then incubated with antibody diluted in blocking solution at 4°C overnight. On day 2, the slides were washed thrice with 1× PBS and then incubated with Alexa Fluor 488/594 2nd antibody diluted in blocking solution at room temperature for 1 hour. The slides were imaged with a Zeiss LSM710 confocal microscope.
Tet-on CRISPR/Cas9-mediated knockdown
The lentiviral expression vector for Tet-on Cas9 was purchased from Addgene (Addgene ID: 50661). The lentiviral expression vector for single guide RNA (sgRNA) was obtained from Dr. Wang Dong's laboratory. The sgRNA sequences are the following. sgNEAT1-1: GGTTACCATGCTCTCCTACA, sgNEAT1-2: GGTAGGAGGTGAGCCTGGGA, Scramble sgRNA-1: GAACGTTGGCACTACTTCAC, Scramble sgRNA-2: GCGCCTTAAGAGTACTCATC. Both vectors were individually mixed with the packaging plasmid psPAX2 and VSVG. The plasmids were transfected into HEK293T cells with the Lipofectamine 2000 reagent (Invitrogen). The media with virus was collected 72 hours after transfection and passed through a 0.45 μm filter. The cells were infected in media containing 8 μg/mL of polybrene and spin-fected by centrifugation at 2,200 rpm for 2 hours. After 24 hours of infection, the virus was removed, and the cells were cultured using media with appropriate antibiotics.
Xenograft assay
The animal model was carried out in male BALB/c nude mice (6–8-weeks-old) First, 5 million cells with inducible gene knockdown were gently mixed into 150 μL of PBS and used to infect the mice subcutaneously. The tumor volume was recorded after 7 days of treatment. The drinking water of the mice contained 2 mg/mL of doxycycline and 5% sucrose. The tumor volume was measured every 5 days. Tumor weight was measured at the end of experiment. All the studies with mice have been approved by the Institutional Animal Care and Use Committee at Tsinghua University.
RNA sequencing
Total RNA was isolated from cells transfected with siNC and siNEAT1_2. Before library construction, ribosomal RNA was removed using a Ribozero Kit (Epicentre). The RNA-seq library was generated using a NEB Next Ultra Directional RNA Library Prep Kit (NEB). RNA-seq was carried out with Illumina HiSeq X Ten.
Differential expression analysis and gene set function annotation
Truseq library 2 × 125 reads from total RNA sequencing (RNA-seq) were first pre-processed using Cutadapt to remove adaptors and trim low-quality bases from the 5′ and/or 3′ ends. After discarding reads shorter than 20 bp, paired-end reads were mapped to the hg38 genome using the splice-aware algorithm RSEM (v1.2.15), with GENCODE v23 reference annotation and the following parameters: “—bowtie2—paired-end.” Differential gene-expression analyses were performed with the R package DEseq and marked on a volcano plot. Significantly up- and downregulated (|log2-fold change| >1 and P value < 0.05. siNEAT1_2 vs. siNC) genes were used for gene set function annotation analysis with Metascope (http://metascape.org/gp/index.html#/main/step1) and illustrated on a heat map.
Modulator inference using the network dynamics algorithm
The Modulator inference using the network dynamics (MINDy) algorithm was used to infer the TFs whose activities on the target gene expressions depend on the level of NEAT1 (26). The transcriptome profiles obtained with RNA-seq (level 2) for 499 prostate cancer tumor samples were downloaded from TCGA and used for the MINDy algorithm. The P value cutoff of MINDy was set at 1E−8. The results of MINDy were then subjected to two levels of filtering. First, for the target genes of each TF inferred by MINDy, the ones that are weakly correlated with the TF (correlation coefficient below 0.4) were discarded. Next, the TFs with no more than 20 target genes that were inferred by MINDy to be modulated by NEAT1 were discarded. Finally, the analysis and filtering pipeline yielded 111 TFs, which has more than 20 target genes that are highly correlated with the TFs and inferred by MINDy to be dependent on NEAT1.
RNA-binding protein immunoprecipitation
PC3 cells were harvested into a fresh 1.5 mL tube, lysed with native lysis buffer (Solarbio, Lot: R0030) with proteinase inhibitor cocktail (Roche) and RNAsin (TianGen) for 30 minutes by shaking every 4 minutes, and then centrifuged at 14,000 × g for 10 minutes. The antibodies (IgG and CDC5L) were incubated with Dynabeads Protein-G at 4°C by rotating for 6 to 8 hours. Then, the antibody-bead slurry was washed five times with NT-2 buffer (50 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 1 mmol/L MgCl2, and 0.05% NP-40). The antibody-bead slurry was resuspended with 900 μL of NET-2 buffer (NT-2 buffer, 20 mmol/L EDTA (pH 8.0), 1 mmol/L DTT, and 200 U/mL RNasin). The supernatants of the cell lysate were mixed overnight with the antibody-bead slurry at 4°C by rotating. Afterwards, the antibody-bead slurry was washed five times with NT-2 buffer. The antibody-bead slurry was resuspended with proteinase k buffer (117 μL of NT-2 buffer, 15 μL of 10% SDS, and 18 μL of proteinase K; 10 mg/mL). Then, RNA was isolated using TRizol. For cross-linking IP, the cells were exposed to 150 mj/cm2 UV light (254 nmol/L), as described previously (25).
Chromatin immunoprecipitation sequencing
The PC3 cells were fixed with formaldehyde at a 1% final concentration by incubation at 37°C for 10 minutes. Then, glycine was added to a final concentration of 0.14 mol/L, and the cells were incubated at room temperature for 30 minutes. The cells were harvested into a fresh tube and lysed with 400 μL of lysis buffer with 8 μL of protease inhibitor, with ice vortexing every 2 minutes. The cell lysate was sonicated to shear cross-linked DNA and was then mixed with Dynal beads and washed 3 times with PBS/BSA. The beads were resuspended with 1 mL of PBS/BSA with the tube against a magnet. Then, 5 μg of antibody were added to the bead slurry, for a total volume of 1 mL. This mixture was then incubated for 6 hours on a rotating platform at 4°C. The antibody-bead slurry was washed 3 times with PBS/BSA. Sheared chromatin was added to the antibody-bead slurry and incubated at 4°C overnight on a rotating platform. Then, 30 μL of chromatin were saved for input. The antibody-bead slurry was washed 5 times with RIPA buffer (50 mmol/L Hepes pH 8.0, 1% NP-40, 0.7% DOC, 0.5 M LiCl, and 1× proteinase inhibitor cocktail), re-suspended with 100 μL of elution buffer (10 mmol/L Tris pH 8.0, 1 mmol/L EDTA, and 1% SDS), and incubated at 65°C for 10 minutes. Reverse crosslinking was performed at 65°C overnight. Then, a 0.4 mg/mL final concentration of proteinase k was added to the elution, and the sample was incubated at 55°C for 1 hour. Genomic DNA was isolated using a Tiangen DNA purification kit and then incubated with 1 μL of RNAse (20 mg/mL) at 37°C for 1 hour. Finally, the DNA was re-isolated using AMpure XP beads. The chromatin immunoprecipitation sequencing (ChIP-seq) library was generated with a TD503 kit (Vazyme). Sequencing was carried out using Illumina HiSeq X Ten. ChIP-seq peak finding was performed with the MACS algorithm (27).
Data availability
The gene expression datasets generated in this study are available in the GEO database repository (https://www.ncbi.nlm.nih.gov/geo/, GEO accession ID: GSE114959).
Results
NEAT1 knockdown arrested the proliferation and tumorigenesis of PC3 and DU145 prostate cancer cells
In androgen-independent prostate cancer cells, NEAT1 was shown to be upregulated by ERα signaling and to act as a transcriptional regulator for tumor-promoting ER-signature genes (1). The goal of the current study is to further elucidate the essential physiological role of NEAT1 in CRPC cells in absence of the ERα signaling activity. The NEAT1 gene has two transcripts: the short isoform NEAT1_1 and the long isoform NEAT1_2. NEAT1_2, but not NEAT1_1, has been recognized as the predominant isoform for the function of NEAT1 in the paraspeckle (28). Our qPCR assay also suggested that NEAT1_2 is the main transcript of NEAT1 in PC3 and DU145 cells (Supplementary Fig. S1A). The following studies therefore focused on NEAT1_2.
Silencing of NEAT1 via an siRNA targeting both NEAT1_1 and NEAT1_2 or an siRNA targeting NEAT1_2 only (Supplementary Fig. S1B) resulted in nearly identical arrest of PC3 cell proliferation (Fig. 1A). Another type of CRPC cell, DU145, showed similar growth inhibitory responses to silencing of NEAT1 (Fig. 1B; Supplementary Fig. S1C). The sgRNA-mediated inducible knockdown of NEAT1 by CRISPR (Supplementary Fig. S1D–S1F) also resulted in a significant arrest of PC3 cell proliferation (Fig. 1C) and the growth of xenograft tumors derived from PC3 cells (Fig. 1D; Supplementary Fig. S1G). Taken together, these results confirmed the essential role of NEAT1 in maintaining the regular cell proliferation and tumor growth of CRPC cells. However, the cellular response to NEAT1 loss that led to proliferation arrest and the underlying pathway for the essential function of NEAT1 in CRPC cells remain largely unknown.
Silencing NEAT1 resulted in DNA damage in PC3 and DU145 cells
To elucidate the role of NEAT1 in the prostate cancer cells, we measured the transcriptome profile shift that occurs upon the loss of NEAT1 via RNA-seq experiments in PC3 cells. Silencing NEAT1 caused more downregulation than upregulation of genes (Fig. 2A; Supplementary Fig. S2, and detailed data in Supplementary Table S1). Strikingly, out of 333 downregulated genes, 26 genes were annotated as being involved in chromatin assembly (Fig. 2B), which was the top enriched function repressed upon NEAT1 knockdown. The other significantly enriched processes included nucleosome assembly and epigenetic gene-expression regulation. There were far fewer upregulated genes (Fig. 2A), and those genes did not show any functional enrichment with P value smaller than 1E−4.
In fact, many of the downregulated genes that composed the repressed processes, such as chromatin and nucleosome assembly, were histone genes (Supplementary Table S2). Previous studies have shown that DNA damage induced downregulations of histones at both the mRNA and protein levels (29, 30). Therefore, we investigated DNA damage in response to NEAT1 repression; indeed, silencing NEAT1 caused a dramatic DNA damage response in PC3 and DU145 cells, as shown by the phosphorylation of histone 2A.X (Fig. 2C), which is a marker of the DNA damage response. The TUNEL assay confirmed breakdown of the DNA upon silencing of NEAT1 (Fig. 2D), and quantification of apurinic/apyrimidinic sites also supported the significant DNA damage upon silencing of NEAT1 (Fig. 2E). Taken together, these results indicated the essential role of NEAT1 in maintaining the genomic DNA integrity of PC3 and DU145 cancer cells. Finally, it is worth noting that our flow cytometry assays did not detect significant changes in the apoptosis rate in response to NEAT1 silencing in PC3 and DU145 cells (Supplementary Fig. S3A). Consistently, the caspase-3 activity and the expression levels of BCL2 and BAX all remained unaltered upon NEAT1 silencing (Supplementary Fig. S3B and S3C). Indeed, as both PC3 and DU145 bear loss of p53 function, which is the major mediator of apoptosis induction upon DNA damage, it is not surprising that NEAT1 silencing induced strong DNA damage without triggering apoptosis.
One of the major downstream effects of the DNA damage response is the arrest of cell-cycle progression at the G1–S, intra-S and G2–M checkpoints (31). Indeed, silencing of NEAT1 induced cell-cycle arrest at the G1 and G2–M phase and reduced the number of cells in the S phase in both PC3 and DU145 cells (Fig. 2F; Supplementary Fig. S4A and S4B; Supplementary Fig. S5A). Such senescence of DNA synthesis has been shown to repress the tumorigenesis and malignancy of cancer cells (32). Furthermore, NEAT1 silencing reduced the phosphorylation of RB at Thr821 (Supplementary Fig. S5B), indicating CDK2 activity inhibition and RB-E2F1 dissociation (33), which could contribute to the cell-cycle arrest that we observed.
NEAT1 regulates the transcriptional activity of CDC5L
Previous studies have proposed that NEAT1 functions as a transcriptional regulator by modulating the epigenetic landscape of tumor-promoting genes (1). NEAT1 was also shown to bind to transcriptional start and termination sites (9). These findings indicated that NEAT1 may be deeply involved in the process of transcriptional regulation, but it is unclear whether NEAT1 targets any specific TF during the process of transcriptional regulation. Here, we sought to identify specific TF(s) whose activity on target genes depends on the level of NEAT1.
We took advantage of transcriptomic profiling data from about 500 prostate tumors in the TCGA consortium and used the MINDy algorithm (26, 34) to screen for TFs that appear to be dependent on NEAT1 expression. Approximately 1,400 genes that were annotated as TFs or putative TFs were tested, and 111 TFs passed the filters, showing strong signatures of dependency on NEAT1 expression (top 14 shown in Fig. 3A, and more details in Supplementary Table S3). Previously, a CHART-MS assay had profiled the proteins pulled down by RNA probes targeting NEAT1 via mass spectrometry (9), generating a list of 30 NEAT1-interacting proteins. Interestingly, among the top 14 TFs that were inferred to be dependent on NEAT1, only CDC5L was among this list of 30 proteins that bind directly to NEAT1, as revealed by the CHART-MS data (9). Furthermore, our RIP-qPCR assay, with or without UV cross-linking, confirmed the direct interaction between CDC5L and NEAT1 (Fig. 3B). In addition, an RNA-FISH assay also demonstrated the partial colocalization of NEAT1 and CDC5L in the cell nucleus (Fig. 3C).
Next, we constructed a dual-luciferase reporter system that contains CDC5L-specific–binding motifs (24) upstream of a firefly luciferase gene. The luciferase reporter assay showed that silencing NEAT1 indeed repressed the transcriptional effect of CDC5L on firefly luciferase expression (Fig. 3D). Finally, although CDC5L was found to be generally colocalized with PSPC1, which is a core component and marker of paraspeckles, the repression of NEAT1 broke up the integrity of the paraspeckle and its association with CDC5L (Fig. 3E). Collectively, these results established the physical interaction between NEAT1 and CDC5L, and more importantly, both the data-mining and the experimental tests showed that the activity of CDC5L depends on NEAT1 expression.
Furthermore, the silencing of CDC5L in PC3 and DU145 cells (Supplementary Fig. S6A) also resulted in the arrest of cell proliferation (Fig. 4A) and DNA damage (Fig. 4B and C), similar to the phenotypic responses of the cells to NEAT1 repression (Figs. 1 and 2). Importantly, overexpressing CDC5L completely rescued the proliferation arrest caused by silencing NEAT1 and restored normal tumor cell growth (Fig. 4D). The DNA damage resulted from loss of NEAT1 was also reversed by overexpressing CDC5L in PC3 cells (Fig. 4E). Taken together, the above results suggest that the physiological indispensable function of NEAT1 in CRPC cells is mediated, at least in part, by its regulatory function on the activity of the putative TF CDC5L.
AGRN, a target of CDC5L, mediated the effect of NEAT1 in CRPC cells
Although recognized as a putative TF, CDC5L has not been assessed in a ChIP-seq experiment. In the present study, we performed ChIP-seq profiling for CDC5L, and the CDC5L-binding regions were identified with the peak-calling program MACS (Supplementary Table S4), which also identified the CDC5L target genes harboring the peak regions in their approximate promoters (Supplementary Table S5). Interestingly, this CDC5L target gene set exhibited a strong enrichment in a gene set enrichment analysis using the differential mRNA expression data upon NEAT1 silencing (Fig. 5A). This result again supports our finding that CDC5L activity depends on NEAT1 expression.
Among the putative CDC5L targets revealed by the ChIP-seq data, 10 genes were subjected to differential expression upon silencing of NEAT1 (adjusted P < 0.05 and fold change >1 or < −1). These genes include FAM43A, RECQL4, ACAP3, TIGD5, AGRN, EPS8L2, KLHL17, TONSL, CDC42BPA, and PLEKHB2, and some of them indeed showed up- or downregulation by CDC5L knockdown (Fig. 5B). Among these 10 genes, AGRN was the most sensitive gene in response to repression of CDC5L (Fig. 5B). Furthermore, the suppressed AGRN expression caused by NEAT1 knockdown was rescued by overexpressing CDC5L in the PC3 cells (Fig. 5C). Collectively, these results confirmed the CDC5L–AGRN circuit that was modulated by NEAT1.
Importantly, compared with repression of CDC5L or NEAT1, silencing of AGRN (Supplementary Fig. S6B) caused nearly identical phenotypic responses in the CRPC cells PC3 and DU145, including proliferation arrest, DNA damage responses, and the reduction of S-phase cells (Fig. 6A–D; Supplementary Fig. S7A and S7B). Finally, similar to CDC5L, overexpressing AGRN in PC3 cells rescued the suppressed proliferation rate observed after NEAT1 knockdown (Fig. 6E). Therefore, we propose that the tumor-promoting function of NEAT1 in the prostate cancer cells was mediated by the CDC5L–AGRN transcriptional activation circuit. Loss of NEAT1 attenuated the CDC5L activity on its target AGRN, resulting in DNA damage and cell-cycle arrest that hindered cell proliferation (working model summarized in Fig. 7).
Discussion
The regulation of gene transcription has been recognized as one of the major functions of lncRNAs (20, 21). Various lncRNAs have been shown to interact directly with histone modifying proteins, RNA polymerase II, and many other DNA binding proteins, including transcription factors (20). As an essential component of the paraspeckle, NEAT1 was shown to activate the expression of IL8 by titrating away the transcriptional repressor SFPQ from the IL8 promoter DNA (7). A systematic lncRNA–chromatin interaction profiling assay (CHART-seq) showed NEAT1 binding to active chromatin regions, especially around transcriptional start sites and transcriptional termination sites, suggesting a strong association between transcription and NEAT1 (9). Furthermore, a set of NEAT1-interacting proteins was identified in a CHART-MS experiment (9), suggesting both previously known and uncharacterized functions of NEAT1. Inspired by these findings related to NEAT1, we sought to further elucidate the involvement of NEAT1 in the transcriptional regulation that underlies the physiological functions of NEAT1 in tumor cells.
We applied a high-throughput computational screen strategy to search for transcription factors that may depend on NEAT1 expression. Specifically, the MINDy algorithm quantifies how much TF-target association depends on a candidate regulator or TF activity (26). Using prostate cancer tumor transcriptome profiling data from TCGA, our analysis identified a group of TFs whose activities appeared to depend on NEAT1 expression. This insight was then integrated with previous NEAT1-protein interaction data, leading to the hypothesis that CDC5L is a regulatory target of NEAT1. Although CDC5L has been recognized as a putative TF (24, 35), this factor has been better studied as an RNA binding protein involved in the regulation of RNA splicing (24, 35–38). Here, we focused on the activity of CDC5L as a TF and showed that its TF activity was reduced upon NEAT1 repression. Furthermore, we found that AGRN, as a target of CDC5L, was responsible, at least in part, for the indispensable function of NEAT1 in tumor cell growth.
A previous study has reported that AGRN was involved in the proliferation, migration, and invasion of liver cancer cells by regulating focal adhesion integrity (39). Here, we have confirmed the tumor-promoting function of AGRN in prostate cancer. The knockdown of AGRN via RNAi repressed the proliferation of prostate cancer cells and induced DNA damage in both PC3 and DU145 cells. Although AGRN has not been reported to be associated with DNA integrity, it was shown to bind to extracellular BMP2, BMP4, and TGFβ1 (40). Specifically, AGRN binding was shown to promote the activity of TGFβ (40), which is necessary for maintaining genomic stability and facilitating DNA damage repair (41–43). Although DNA damage has been acknowledged as a hallmark of cancer and a driver of tumorigenesis, cancer cells also need a certain level of genome stability to ensure their viability and proliferation potential (44). Therefore, a proper level of TGFβ signaling activity could be advantageous for tumor development (41). On the basis of our findings, we suspect that AGRN helps maintaining DNA integrity and promotes DNA repair by activating the TGFβ signaling activity.
In conclusion, we propose that NEAT1 plays an essential role in maintaining tumor cell growth and preventing DNA damage via direct binding with CDC5L. The repression of NEAT1 reduced the expression of AGRN, which is a target of CDC5L and a binding partner of TGFβ1, resulting in the accumulation of DNA damage and cell-cycle arrest and hindering the growth and tumorigenesis of PC3 and DU145 prostate cancer cells.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: X. Li, X. Yang
Development of methodology: X. Li, X. Yang
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Li, W. Song, H. Xu, R. Huang, Y. Wang, Z. Xiao
Writing, review, and/or revision of the manuscript: X. Li, X. Yang
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Wang, W. Zhao
Study supervision: X. Yang
Acknowledgments
The authors wish to thank Mathew P. Daniels and Michael Ferns for generously providing the AGRN overexpression plasmids. The authors wish to acknowledge the supports from the Platforms of Genome Sequencing, High-Performance Computing, shRNA Library, and Cell Imaging & Function of the National Protein Science Facility (Beijing), the Laboratory Animal Center at Tsinghua University, and the Flow Cytometry Core Facility of Center of Biomedical Analysis at Tsinghua University. This work was supported by the National Key Research and Development Program, Precision Medicine Project (2016YFC0906001 to X. Yang), the National Natural Science Foundation of China (91540109 and 81472855 to X. Yang), the Tsinghua University Initiative Scientific Research Program (2014z21046 to X. Yang), the Tsinghua–Peking Joint Center for Life Sciences, and the 1000 talent program (Youth Category; to X. Yang).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.