Roquin1 Suppresses Breast Cancer Growth by Inducing Cell Cycle Arrest via Selectively Destabilizing Cell Cycle–Promoting Gene mRNAs


 Background: Dysregulation of cell cycle progression is one of the common features of human cancer cells, however, its mechanism remains unclear. This study aims to clarify the role and the underlying mechanisms of Roquin1 in cell cycle arrest induction in breast cancer.Methods: Public cancer databases were analyzed to identify the expression pattern of Roquin1 in human breast cancers and the significant association with patient survival. Quantitative real-time PCR and western blots were performed to detect the expression of Roquin1 in breast cancer samples and cell lines. Cell counting, MTT assay, flow cytometry, and in vivo study were conducted to investigate the effects of Roquin1 on cell proliferation, cell cycle progression and tumor progression. RNA-sequencing was applied to identify the differential genes and pathways regulated by Roquin1. RNA immunoprecipitation assay, luciferase reporter assay, mRNA half-life detection, RNA affinity binding assay, and RIP-ChIP were used to explore the molecular mechanisms of Roquin1.Results: We showed that Roquin1 expression in breast cancer tissues and cell lines was inhibited, and the reduction in Roquin1 expression was associated with poor overall survival and relapse free survival of patients with breast cancer. Roquin1 overexpression inhibited breast cancer cell proliferation and induced G1/S cell cycle arrest without causing significant apoptosis. In contrast, knockdown of Roquin1 promoted breast cancer cell growth and cycle progression. Moreover, in vivo induction of Roquin1 by adenovirus significantly suppressed breast tumor growth and metastasis. Mechanistically, Roquin1 selectively destabilizing cell cycle–promoting genes, including Cyclin D1, Cyclin E1, cyclin dependent kinase 6 (CDK6) and minichromosome maintenance 2 (MCM2) through targeting the stem–loop structure in the 3’untranslated region (3’UTR) of mRNAs via its ROQ domain, leading to the downregulation of cell cycle–promoting mRNAs.Conclusions: Our findings demonstrated that Roquin1 was a novel breast tumor suppressor and could induce G1/S cell cycle arrest by selectively downregulating the expression of cell cycle–promoting genes, which might as a potential molecular target for breast cancer treatment.

autoimmunity by destabilizing the inducible T cell costimulator (ICOS) mRNA through its ROQ domain [7]. Therefore, Roquin1 is regarded as a regulator of immune system and orchestrate the differentiation of various immune cells, including Follicular helper T cell (TfH), Natural killer T (NkT), and Regulatory T (Treg) cells [8][9][10]. Roquin1 de ciency induces the death of mouse embryos and severe autoimmune reaction and enteritis [11,12]. Evidence showed that Roquin1 could decay mRNA by binding the stemloop structure in the 3'UTR of target genes [13][14][15]. Besides, Roquin1 serves as a regulator of multiple signaling pathways, such as AMP-activated protein kinase (AMPK) [16], NF-κB [17], and PI3K-mTOR [18], to regulate immune responses. Roquin1 is additionally able to regulate microRNA homeostasis [19]. In cancerous TfH cells, Roquin1 expression level is similar to that in normal TfH cells [20]. However, it remains unknown whether Roquin1 plays a role in cancer progression.
In this study, we identi ed that Roquin1 is a potent breast tumor suppressor by inducing tumor cell cycle arrest through selectively suppressing the expression of cell cycle-promoting genes, including CCND1, CCNE1, CDK6, and MCM2. Roquin1 expression was reduced in breast cancer tissues and cells, which might contribute to their absence from cell cycle regulation. Ectopic Roquin1 expression induces G1/S cell cycle arrest in breast tumor cells. In contrast, further suppression of Roquin1 expression by shRNAs facilitates tumor cell proliferation and cell cycle progression. Consistent with these in vitro observations, Roquin1 expression in vivo signi cantly inhibits tumor growth and metastasis. By analyzing the database of human breast tumors [21], low Roquin1 levels in tumor samples were found to be strongly associated with poor survival of luminal A, luminal B, and Basal breast cancer patients. Moreover, Roquin1 expression is negatively correlated with CCNE1 and MCM2 in human breast tumors. These results suggested that Roquin1 was a potential tumor suppressor, which was involved in regulating cell cycle progression by suppressing the cell cycle-promoting genes expression.

Cell lines and plasmids
The human breast cancer cell line (MDA-MB-231, MDA-MB-468, MCF7 and T47D), human normal mammary epithelial cell lines (MCF-10A and MCF-12A), human lung cancer cell line A549 cell, and human liver cancer cell line (HepG2) cells were obtained from the American Type Culture Collection (ATCC) and cultured in DMEM or RPMI-1640 with 10% FBS plus 1% Peni/Stro, respectively. HEK293 and HEK293T cells were obtained from National Infrastructure of Cell Line Resource (Beijing, China). The human full-length Roquin1 coding sequence (NM_172071) was synthesized, sequenced and inserted into pEGFP-N1 vector at EcoR I and Age I sites. Roquin1 serial deletion plasmids were generated by inserting the PCR-ampli ed fragments into pEGFP-N1 vector at EcoR I and Age I sites. A set of luciferase reporters were constructed by inserting the full-length 3'UTRs of human MCM2, Cyclin D1, and Cyclin E1 or part 3'UTR (1-1440 bp) of CDK6 into the pGL3 control vector (Promega) between Xba I and Fse I sites, respectively. For stem-loop deletion reporters, point mutated and truncated CCNE1-3'UTR (1-256 bp) (∆stem-loop) and MCM2-3'UTR (1-360 bp) (∆stem-loop) were ampli ed, sequenced, and inserted into pGL3 control vector using Phusion Site-Directed Mutagenesis Kit (Thermo Scienti c). For stem-loop insertion constructs, the stem-loop sequences of CCNE1 3'UTR (257-276) and MCM2 3'UTR (361-377) were inserted into pGL3-β-actin 3'UTR reporter at 555 base pair.

Cell cycle analysis
Cells were trypsinized and washed twice with PBS, and then xed in cold ethanol (70%). The cells were then stained with propidium iodide (20 µg/mL) and RNase A (0.2 mg/mL) for 30 min. The stained cells were analyzed by ow cytometry and the data were analyzed with FlowJo software.
RNA-sequencing analysis MCF7/Roquin1-GFP, MDA-MB-468/Roquin1-GFP, A549/Roquin1-GFP, HepG2/Roquin1-GFP, and their control cells (expressing GFP) were cultured for 36 h and total RNA was extracted using the TRIzol method. RNA sequencing (RNA-seq) was completed by Allwegene Technology Inc., Beijing. The cDNA library was then constructed using polymerase chain reaction (PCR) ampli cation. RNA-seq was performed with the PE150 sequencing strategy using an Illumina second-generation high-throughput sequencing platform. RNA-seq reads with inferior quality or adapters were ltered. Clean read data were processed using Tophat2 and Cu inks software to complete the alignment of transcriptomes. Genes not expressed in any sample were excluded from further analysis. Differentially expressed genes and transcripts were then ltered for false discovery rate (FDR)-adjusted P values less than or equal to 0.05. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis RNA-seq data were deposited (PRJNA637876). The common up-and -downregulated mRNAs by Roquin1 in two breast cancer cells were classi ed using the Venn diagram. Gene Ontology (GO) (biological process) and KEGG pathway analyses of commonly downregulated genes were done using DAVID Bioinformatics Tools and Ingenuity Pathway Analysis.

RNA immunoprecipitation
Roquin1/GFP fusion protein was expressed in MDA-MB-468/Roquin1 cells, and then whole cell lysates were pre-cleared with isotype IgG, followed by incubation with anti-GFP antibody at 4℃ for 4 hrs. The protein-RNA complexes were then pulled down by protein G agarose beads (sc-2002, SantaCruz) and total RNA extracted with TRIzol, followed by detection of cell cycle-promoting genes with RT-PCR.
Luciferase reporter assays Luciferase assay was performed as described previously [22]. pGL3 luciferase reporter constructs containing full-length or segment of 3'UTR of different genes were transfected into HEK293 cells along with Roquin1/GFP, aa 1-441 (contain RING, ROQ, zinc nger domains), aa 441-1133 (contain PRD), aa 174-326 (only contain ROQ domain), and GFP-control constructs, respectively. All transfections were conducted in triplicate and repeated at least three times. The luciferase activity was measured 36 h after transfection using a Dual-Luciferase Reporter Assay System (Promega). mRNA stability Roquin1/GFP was expressed in MDA-MB-468 cells, and then actinomycin D (ActD, 5 µg/mL) and 5, 6dichlorobenzimidazole riboside (DRB, 5 µg/mL) were added to block de novo RNA synthesis. Total RNA was collected at indicated time points, and the relative mRNA level was analyzed by qPCR. The half-life of mRNA was determined by comparing with the levels of mRNA before adding ActD and DRB. The halflife of different genes in Roquin1 knockdown cells and their corresponding scrambled control cells was also tested as described earlier.

Hematoxylin and Eosin (H&E staining)
Mouse lung tissues were immersed in 10% formalin for at least two weeks, and then stained with H&E staining. Zeiss Imaging System is used for visualizing of H&E sections.

Statistical analysis
Data in bar graphs represent mean ± SD of at least three biological repeats. Statistical analysis was performed using Student's t-test by comparing treatment versus vehicle control or otherwise as indicated. P-value < 0.05 was considered to be statistically signi cant.

Results
Roquin1 expression was reduced in breast cancer patients and was associated with poor survival Roquin1 expression was rst analyzed in the breast tumor database (www.oncomine.org). Low levels of Roquin1 were found in various breast cancers, including breast carcinoma, invasive ductal breast carcinoma, and invasive mixed breast carcinoma, although a slightly high level in invasive lobular breast cancer tissues ( Fig. 1a; Additional le 1: Figure S1A). Experimentally, Roquin1 mRNA expression was signi cantly reduced in breast tumors compared with normal tissues (Fig. 1b). Meanwhile, Roquin1 protein expression was also lower in randomly selected four pairs of breast cancer tissues than in their surrounding normal tissues (Fig. 1c). Moreover, Roquin1 expression was signi cantly repressed at both protein ( Fig. 1d) and mRNA (Fig. 1e) levels in several human breast cancer cell lines compared with normal mammary gland epithelial cells. Notably, by surveying Roquin1 expression across a gene array dataset [21], we found that low Roquin1 expression in human breast tumor samples was strongly associated with poor overall survival and relapse-free survival of patients ( Fig. 1f and 1 g). Furthermore, the levels of Roquin1 in breast tumors were also associated with patient survival in the Luminal A, Luminal B, and Basal-like subsets ( Fig. 1h-j). Although no signi cant correlation was found between Roquin1 expression and patient survival in HER2 + subsets, a similar trend like another three subsets was found (Fig. 1k). These results suggested that Roquin1 was of great signi cance for the prognosis of patients with breast cancer. Besides, we also found that Roquin1 was suppressed in other types of human cancers, including lung cancer, ovarian cancer, gastric cancer, and bladder carcinoma (Additional le 1: Figure S1B-1E). Striking, Roquin1 expression levels also signi cantly correlated with the prognosis of patients with these above cancers and liver cancer (Additional le 1: Figure S1F-1J), indicating Roquin1 might be clinical predictive for multiple cancers.

Roquin1 inhibits cell growth by inducing G1-S phase cell cycle arrest in breast cancer cells
To determine Roquin1 function in breast cancer progression, Roquin1/GFP fusion protein was expressed in breast cancer cells MCF7 and MDA-MB-468, respectively, and identi ed by Western blot (Fig. 2a). When Roquin1 overexpression, we noticed that the proliferation (Fig. 2b, c) and activity (Fig. 2d, e) of breast cancer cells were considerably reduced. Similar results were also found in human lung cancer cell A549 and human liver cancer cell HepG2 with Roquin1 overexpression (Additional le 1: Figure S2A-2D). To determine whether Roquin1 inhibited cell proliferation by affected tumor cell cycle progression, we evaluated the effect of Roquin1 overexpression on cell cycle by ow cytometry (FCM). The G1 phase percentage of breast tumor cells was signi cantly increased in Roquin1-overexpressing cancer cells compared with the controls. Meanwhile, a signi cant decrease in S phase percentage was detected after Roquin1 overexpression. Similar results were also found in A549 and HepG2 cells with Roquin1 overexpression (Additional le 1: Figure S2G-2J). However, the G2 phase percentage of cells did not change consistently among tumor cells which might be due to different cell types (Fig. 2f, g; Additional le 1: Figure S2I-2J). These ndings suggested that Roquin1 was able to induce the G1/S cell cycle arrest in breast tumor cells. Indeed, the protein levels of p21, a typical cell cycle inhibitor, were induced by Roquin1 in breast tumor cells (Fig. 2h, i), and A549, HepG2 cells (Additional le 1: Figure S2K-2L). To determine whether Roquin1 induced apoptosis in breast tumor cells, cleaved caspase3 and PARP1, two key apoptosis indicators, were detected by Western blotting. Roquin1 was not able to cause signi cant cleavages of pro-caspase3 and pro-PARP1 in breast tumor cells, although a slightly cleaved PARP1 was detected in MDA-MB-468 cells 72 h after Roquin1 overexpression (Fig. 2j, k), indicating Roquin1 did not cause cell apoptosis in breast tumor cells. This detection was consistent with the observation that no obvious cell death occurred after Roquin1 overexpression during cell culture. Collectively, these data clearly demonstrate that Roquin1 induces G1/S cell cycle arrest of breast tumor cells.
Roquin1 selectively inhibited the mRNA expression of cell cycle-promoting genes through targeting 3'UTRs Next, we identi ed the genes affected by Roquin1 using RNA-seq in Roquin1-overexpressing MCF7 and MDA-MB-468 cells. Venn diagrams showed that 6556 genes were common downregulated and 7067 genes were common upregulated in two breast tumor cells (Additional le 1: Figure S3A). We further focused on the expression of cell cycle-related genes. Interestingly, the genes that promote cell cycle progression, including G1/S transition, G2/M transition, S phase transition, and M phase transition, were suppressed, whereas the genes inhibiting cell cycle (p21 and Rb1) were enhanced by Roquin1 in MCF7 (Fig. 3a) and MDA-MB-468 cells (Additional le 1: Figure S3B). Similar trends were also found in A549 and HepG2 cells (Additional le 1: Figure S3C-3D), indicating that Roquin1 could regulate the expression of cell cycle-related genes in tumor cells. Detailed RNA-seq data were summarized in Additional le 3: Table S1. Moreover, the 'cell cycle' pathway was the rst of the top ten signaling pathways signi cantly enriched in the KEGG pathway analysis of downregulated genes (Fig. 3b). And, cell cycle-related terms 'cell division' and 'mitotic nuclear division' were enriched in the Gene Ontology (GO) analysis of downregulated genes (Fig. 3c). These computational analyses further supported our experimental ndings. To validate the RNA-seq data, four downregulated cell cycle-promoting genes (i.e., CCND1, CCNE1, CDK6, and MCM2) and three upregulated cell cycle-inhibiting genes (i.e., p21, p27, and Rb1) were measured by real-time PCR. The mRNA expression of four cell cycle-promoting genes was reduced in a time-dependent manner by Roquin1 in breast cancer cells (Fig. 3d, e). Also, the protein levels of CCNE1 and MCM2 were downregulated by Roquin1 over time (Fig. 3f, g). However, the upregulated cell cycleinhibiting genes without time-dependent changes (Additional le 1: Figure S3E). Notably, no timedependent changes in the protein levels of p21 were observed in breast tumor cells (Fig. 2h, i). These results con rmed our RNA-seq data. In agreement with the overexpression results, the cell cyclepromoting genes were upregulated in Roquin1 San/San MEF cells (Additional le 1: Figure S3F) [23], which further strengthened our ndings. Taken together, these results indicate that Roquin1 regulates cell cycle pathway by inhibiting the mRNA expression of cell cycle-promoting genes.
As an RBP, we next examined whether Roquin1 binding to these cell cycle-promoting gene mRNAs. An RNA pull-down assay was performed with an anti-GFP antibody in Roquin1/GFP-expressing MDA-MB-468 cells, followed by detecting bound mRNAs by RT-PCR. Four cell cycle-promoting genes were ampli ed by PCR, whereas the GAPDH and cell cycle-inhibiting mRNAs were not ampli ed (Fig. 3h, i). TNFα was used as a positive control. These results indicated that Roquin1 selectively bound to the cell cycle-promoting genes but not cell cycle-inhibiting genes. To determine if the mRNA binding was mediated through 3'UTR, we cloned the 3'UTRs of CCNE1, CCND1, CDK6 (part), and MCM2 downstream of the luciferase gene as previously described [22], and then co-transfected these reporters with Roquin1 expression vector and its empty vector into HEK293 cells, followed by the measurement of luciferase activity. As shown in Fig. 3J, Roquin1 signi cantly inhibited luciferase activities of all four 3'UTR reporters compared with cells transfected with control vector. The β-actin 3'UTR was used as a negative control. Collectively, these results suggested that Roquin1 speci cally suppressed the mRNA expression of cell cycle-promoting genes by targeting their 3'UTRs.
Roquin1 destabilized the mRNAs of cell cycle-promoting genes via the ROQ domain We speculated that Roquin1 might reduce cell cycle-promoting gene mRNAs through destabilizing the mRNAs. To con rm that, Roquin1/GFP was expressed in MDA-MB-468 cells and then de novo mRNA synthesis blocked using ActD (5 µg/mL) and DRB (5 µg/mL), followed by the measurement of the remaining mRNAs at different time points. The half-lives of indicated cell cycle-promoting mRNAs were shortened about 2-fold in Roquin1-overexpressing cells compared with cells expressing empty vector ( Fig. 4a-d), while the half-lives of cell cycle-inhibiting mRNAs (including p21, Rb1, and p27) were barely affected by Roquin1 (Additional le 1: Figure S4A-4C), demonstrating that Roquin1 indeed inhibits cell cycle-promoting genes expression through the mRNA stability.
Roquin1 protein contains a RING nger, a ROQ domain, a zinc nger (ZF), and a proline-rich domain (PRD), of which the ROQ domain is involved in the destabilization of mRNAs [8]. To determine whether the ROQ domain is also responsible for the cell cycle-promoting mRNAs decay, a series of truncated Roquin1 mutants, including aa 1-441 containing RING, ROQ, and ZF domains, aa 441-1133 containing PRD domain, and aa 174-326 containing ROQ domain, were generated (Fig. 4e) and identi ed by Western blot analysis (Fig. 4f). Then, the mutants were co-transfected with wild-type (WT) Roquin1 as well as different 3'UTR reporters (Additional le 1: Figure S4D) into HEK293 cells. As shown in Fig. 4g and  4 h, WT, mutants aa 1-441 and aa174-326 suppressed the mRNA expression of four cell cycle-promoting genes and the luciferase activities of their 3'UTR reporters, but not mutant aa 441-1133, which was also consistent with the previous report [24]. In addition, aa 174-326 signi cantly inhibited the proliferation ( Fig. 4i) and cell cycle progression ( Fig. 4j; Additional le 1: Figure S4E) of MDA-MB-468 cells, which also signi ed that the ROQ domain in Roquin1 is essential for the induction of breast tumor cell cycle arrest.

Roquin1 knockdown stabilizes cell cycle-promoting gene transcripts and promotes tumor cell cycle progression
To further con rm the inductive effects of Roquin1 on tumor cell cycle arrest, we suppressed Roquin1 expression with two shRNAs in MDA-MB-231 cells, another triple-negative breast cancer cell. Roquin1 was reduced approximately 65% and 74% by #1shRNA and #2shRNA, respectively (Fig. 5a). Although Roquin1 is lowly expressed in breast tumors, the knockdown of Roquin1 considerably promoted the proliferation and activities of breast tumor cells (Fig. 5b, c), and increased the mRNA expression of cell cyclepromoting genes (Fig. 5d). These results were also consistent with previous data in Roquin1 san/san MEF cells [23]. However, depletion of Roquin1 had no effect on the mRNA levels of p21, Rb1, and p27 (Additional le 1: Figure S5A), again suggesting that Roquin1 directly suppressed the mRNA expression of cell cycle-promoting genes. Next, we examined the effect of Roquin1 knockdown on the half-life of cell cycle-promoting gene transcripts. As expected, reduced Roquin1 signi cantly prolonged the halflives of the indicated cell cycle-promoting mRNAs (Fig. 5e-h). Furthermore, we also found reduced percentage of G1 phase cells and increased S phase cell percentage in MDA-MB-231 cells after knocking down Roquin1 (Fig. 5i; Additional le 1: Figure S5B). To con rm whether the cell cycle-promoting genes were involved in Roquin1-induced cell cycle arrest, CCNE1 and MCM2 were knocked down by shRNA lentivirus in Roquin1 knockdown MDA-MB-231 cells. Figure 5J showed that these shRNAs effectively knocked down CCNE1 and MCM2 expression. Upon co-knockdown of Roquin1 and CCNE1/MCM2, cell proliferation was closed to the scramble control compared to Roquin1 knockdown alone (Fig. 5k). Additionally, the percentage of G1 phase cells was signi cantly increased compared with the group of Roquin1 knockdown alone, and the percentage of S phase cells signi cantly decreased (Fig. 5l). Collectively, these results con rmed that Roquin1 repression indeed promotes breast tumor cell cycle progression through stabilizing cell cycle-promoting genes.
Roquin1 binds to the stem-loop structure of cell cyclepromoting genes for degradation Roquin1 is known to degrade target mRNAs by binding to the stem-loop structure [14]. The 3'UTR sequences of four cell cycle-promoting genes were analyzed and a conserved sequence was identi ed across species, respectively, which could form a similar stem-loop structure (Additional le 1: Figure   S6A-6D) using the RNAfold WebServer [25]. To investigate the role of the stem-loop structure in Roquin1mediated degradation of cell cycle-promoting mRNAs, we generated deletion constructs by deleting the sequences containing the stem-loop in the 3'UTRs of CCNE1 and MCM2 (Fig. 6a). Then, full-length and deletion reporters with Roquin1 were co-transfected into HEK293 cells, followed by luciferase activity measurement. Roquin1 could signi cantly inhibit the luciferase activity of full-length CCNE1 and MCM2 3'UTR, but not the deletion mutant reporters (Fig. 6b). In addition, Roquin1 reduced the activities of the reporters containing human β-actin 3'UTR with CCNE1 or MCM2 stem-loop structures, compared with that in the control group (Fig. 6c, d). These ndings indicated that the stem-loop structure was pivotal for Roquin1-mediated cell cycle-promoting mRNAs decay.
To determine the necessity of stem-loop secondary conformation for mRNA degradation, two 3'UTR mutant reporters of CCNE1 and MCM2 were generated, of which the stem-loop structure of mutant1 was deleted by replacing two or four nucleotides and mutant2 retained the stem-loop structure after replacing four nucleotides (Fig. 6e). Subsequently, the two reporters were co-transfected with Roquin1 for measuring luciferase activity. Deletion of the stem-loop structure in the 3'UTRs of CCNE1 and MCM2 (mutant1) allowed them completely to resistant to Roquin1 inhibition, while the mutant2 that kept the stem-loop structure remained sensitive to Roquin1 suppression (Fig. 6f), indicating that the stem-loop structure in 3'UTRs was critical for cell cycle-promoting mRNAs decay. To further determine if Roquin1 physically bound to the stem-loop in the 3'UTRs of CCNE1 and MCM2, we performed RNA a nity binding assay with biotin-labeled RNA probes. Wild-type RNA probes and mutant probes with the stemloop structure either disrupted (mutant1) or retained (mutant2) were incubated with lysates of MDA-MB-468 cells expressing Roquin1/GFP fusion protein. Then, streptdavidin-coated magnetic beads were used for the pull-down assay, followed by Western blot detection with anti-GFP antibody. Roquin1/GFP fusion protein was pulled down by wild-type and mutant2 probes, but cannot by the stem-loop structurede cient mutant1 probe (Fig. 6g), indicating that Roquin1 indeed interacted with the stem-loop structure of CCNE1 and MCM2 in vitro. Furthermore, a modi ed RNA immunoprecipitation-chromatin immunoprecipitation (RIP-ChIP) assay was performed to verify that Roquin1 could bind the stem-loop structure in vivo. Roquin1/GFP fusion protein was expressed in MDA-MB-468 cells, the protein-RNA complex was pulled down by GFP antibody-coated beads, after the bound mRNAs sonicated, followed by ampli cation of the stem-loop sequences by RT-PCR. As expected, the stem-loop sequences in the 3'UTRs of CCNE1 and MCM2 could be ampli ed in the GFP antibody pull-down group, but not in the group using isotype IgG (Fig. 6h), indicating that the binding of Roquin1 to the 3'UTRs of cell cyclepromoting mRNAs inside breast tumor cells. Overall, these data demonstrate that Roquin1 recognized and bound to the stem-loop structure in the 3′UTRs of cell cycle-promoting genes for degradation.

Roquin1 suppressed breast tumor growth and metastasis
To determine the inhibitory effect of Roquin1 on breast cancer progression in vivo, MDA-MB-468/Roquin1-GFP cells (expressing the Roquin1/GFP fusion protein) and MDA-MB-468/GFP cells were inoculated into mammary gland fat pad of female nude mice (BALB/c). The growth and sizes of tumors expressing Roquin1/GFP fusion protein signi cantly reduced compared with the control tumors ( Fig. 7a; Additional le 1: Figure S7A). Roquin1/GFP fusion protein expression in tumors was con rmed by the Western blot analysis (Fig. 7b). Meanwhile, a signi cant decrease in the number of metastatic foci (Fig. 7c) and metastatic white nodules (Additional le 1: Figure S7B) were observed in the lung tissues from Roquin1/GFP tumor-bearing mice. To avoid the impacts of the manual manipulation of gene expression and simulate the clinical treatment of breast cancer, adenoviruses expressing Roquin1/GFP fusion gene and its control virus (expresses GFP) were prepared to treat the established MDA-MB-231 breast tumors in nude mice. When tumor mass reaching approximately 5 mm in diameter, 10 10 pfu of Roquin1/GFP adenovirus in 100 µL of PBS and the control adenovirus were injected every other day for ve injections totally (Fig. 7d). Two days after injection, tumors began to shrink and grew slowly, while the tumors treated with control adenovirus continued growing well (Fig. 7e). At the end of the experiment, the sizes of tumors treated with Roquin1/GFP adenovirus were signi cantly smaller than those in the control group (Additional le 1: Figure S7C). Tumor metastasis was also signi cantly suppressed by Roquin1 adenovirus treatment ( Fig. 7f; Additional le 1: Figure S7D). In line with the in vitro results, the protein levels of CCNE1 and MCM2 were also reduced in Roquin1 adenovirus-treated tumors (Fig. 7g), further con rming that Roquin1 suppressed the expression of cell cycle-promoting genes in vivo. Interestingly, the expression of CCNE1 and MCM2 was also signi cantly inhibited as Roquin1 increased in 1,006 human breast cancer samples (Fig. 7h, i) (Additional le 4: Table S2) (Oncolnc.org/). Notably, higher levels of CCNE1 and MCM2 negatively correlated with poor survival of patients with breast cancer (Fig. 7j, k). Conclusively, these ndings strongly suggested that Roquin1 was a promising breast tumor suppressor, and the Roquin1-cell cycle-promoting genes axis might be considered as a new therapeutic target for breast tumor treatment in future.

Discussion
Currently, the role of Roquin1 in tumor progression is less reported. The analysis of the The Cancer Genome Atlas (TCGA) breast cancer databases showed that Roquin1 was expressed at low levels in human breast cancers. This low-level expression pattern was further con rmed in human breast cancer tissues and cell lines. And, low Roquin1 expression in tumors from various subtypes of breast cancer patients was strongly associated with poor survival signi ed the clinical signi cance of Roquin1. Despite no signi cant association in HER2 + breast cancer, the trend was similar to that of other subtypes. Besides, the levels of Roquin1 in several types of human cancers were also associated with patient survival, indicating that Roquin1 might be important for the prognosis of multiple types of human cancers.
We initially focused on the changes of cytokines secreted by tumor cells following Roquin1 overexpression in breast cancer cells. Unexpected, we found that Roquin1 signi cantly suppressed breast cancer cell proliferation and induced cell cycle arrest in vitro. To our knowledge, the relationship between Roquin1 and cell cycle progression has not been reported to date, especially in breast tumor progression. Using FCM, we found that Roquin1 could signi cantly stop G1/S transition in breast cancer cells as well as lung and liver cancer cells. However, we did not observe obviously cell apoptosis occurs after Roquin1 overexpression in breast tumor cells, which was different from the apoptosis induction role of another RBP, monocyte chemotactic protein-induced protein 1 (MCPIP1) in breast cancer progression [22]. Cell cycle arrest is triggered upon the balance between cell cycle-promoting and -inhibiting gene being broken. Our results clearly demonstrated that most of the genes that promoted cell cycle progression were downregulated by Roquin1, while those that inhibited cell cycle progression were upregulated. Indeed, four cell cycle-promoting genes were con rmed suppressed in a time-dependent manner but not the cell cycle-inhibiting genes. A selective targeting of Roquin1 to cell cycle-promoting mRNAs but not to cycle-inhibiting mRNAs might trigger an imbalance between cell cycle-promoting and inhibiting genes in tumor cells. Roquin1 also inhibits cell cycle progression in human liver cancer cells HepG2 and lung cancer cells A549, indicating that Roquin1 might be implicated in eliciting an antitumor response in a wide range of human cancers.
Among the Roquin1 target genes, CCNE1, CCND1, and CDK6 are involved mainly in G1/S transition [26][27][28], while MCM2 is important for DNA replication and S phase transition of cell cycle [29]. The broad targets of Roquin1 in the cell cycle pathway suggest that Roquin1 induces cell cycle arrest through targeting multiple molecules in breast tumor cells. We showed that Roquin1 suppresses the expression of these genes at the post-transcriptional level through enhancing mRNA degradation. The 3′UTR is important for the post-transcriptional regulation of genes. Many RBPs, including human antigen R (HuR) [30], tristetraprolin (TTP) [31], and MCPIP1, have been reported to regulate mRNA stability by the 3'UTR.
Indeed, Roquin1 inhibits luciferase activity through the 3'UTRs of cell cycle-promoting genes. It has been reported that the ROQ domain is essential for Roquin1-mediated mRNA degradation and immune regulatory effects [32]. We demonstrate that the ROQ domain is also required for the expression of cell cycle-promoting genes and cell cycle arrest induction by mutating the domains of Roquin1.
Many elements responsible for mRNA degradation are primarily localized in the 3'UTR, such as ARE, GUrich element (GRE), and stem-loop structure [33][34][35]. Roquin is known to recruit other deadenylases to the 3'UTR for RNA decay by recognizing and binding the stem-loop structure [36]. Our results demonstrate that it is the stem-loop structure but not the ARE in the 3'UTRs is required for Roquin1mediated cell cycle-promoting mRNAs decay. Roquin1 has been shown to bind a constitutive decay element (CDE) in the 3'UTR of TNFα mRNA and this CDE can fold into a stem-loop structure [37]. Indeed, we found a conserved consensus sequence was shared among different speci es in four cell cyclepromoting genes, respectively, and these sequences could fold into a stem-loop structure. Strikingly, no common stem-loop sequences were identi ed among four cell cycle-promoting genes, which support our hypothesis that Roquin1 mainly recognizes the secondary structure, instead of the linear sequence in the 3′UTR, which was also consistent with previous ndings [13].
Finally, we proposed a model to elucidate the potential role of Roquin1 in the suppression of breast cancer cell cycle progression (Additional le 1: Figure S7E). Roquin1 selectively targeted cell cyclepromoting gene mRNAs for degradation via its ROQ domain by binding to the stem-loop structures in the    Roquin1/GFP and its truncated mutations were con rmed by immunoblotting with an anti-GFP antibody. g MDA-MB-468 cells were transiently transfected with Roquin1 and its mutants. After 36 hours, total RNA was extracted to measure the mRNA expression of indicated genes by qPCR. h HEK293 cells were cotransfected with indicated reporters and Roquin1 as well as its truncations. After 36 hours, luciferase activity was measured in cell lysates and compared with that cells transfected with empty vector. i Cell counting was conducted in MDA-MB-468 cells after overexpressing Roquin1 and its mutants (n = 3). ***P < 0.0001. j Cell cycle was examined by FCM in MDA-MB-468 cells after overexpressing Roquin1 and its mutants (n = 3). *P < 0.05