SNP rs4142441 and MYC co-modulated long non-coding RNA OSER1-AS1 suppresses non-small cell lung cancer by sequestering RNA-Binding Protein ELAVL1


 Background: Single nucleotide polymorphisms (SNPs) and long non-coding RNAs (lncRNAs) have been involved in the process of lung cancer. Following clues given by lung cancer risk-associated SNPs, we aimed to find novel functional lncRNAs as candidate targets in non-small cell lung cancer (NSCLC). Methods: Case-control analyses were performed in 626 cases and 736 controls matched up on sex and age. The lncRNA OSER1-AS1 was identified near a lung cancer risk-associated SNP rs4142441. Kaplan–Meier survival analysis was performed to investigate the association between OSER1-AS1 expression and overall survival. The influence of rs4142441 on the expression level of OSER1-AS1 was confirmed using Luciferase assays. Subsequently, the biological function of OSER1-AS1 was assessed in vitro by cell proliferation, migration, and invasion experiments through gain- and loss-of-function approaches, and in vivo by subcutaneous tumor model and tail vein injection lung metastasis model. ChIP and RIP experiments were carried out to investigate the interaction between transcription factors, RNA-binding proteins, and OSER1-AS1.Results: OSER1-AS1 was down-regulated in tumor tissue and its low expression was significantly associated with poor overall survival among non-smokers in NSCLC patients. Gain- and loss-of-function studies revealed that OSER1-AS1 acted as a tumor suppressor by inhibiting lung cancer cell growth, migration and invasion in vitro. Xenograft tumor assays and metastasis mouse model confirmed that OSER1-AS1 suppressed tumor growth and metastasis in vivo. The promoter of OSER1-AS1 was repressed by MYC, and the 3’-end of OSER1-AS1 was competitively targeted by microRNA hsa-miR-17-5p and RNA-binding protein ELAVL1. Conclusion: Our results indicated that OSER1-AS1 exerted tumor-suppressive functions by acting as an ELAVL1 decoy to keep it away from its target mRNAs. Our findings characterized OSER1-AS1 as a new tumor suppressive lncRNA in NSCLC, suggesting that OSER1-AS1 may be suitable as a potential biomarker for prognosis, and a potential target for treatment.

In this study, we rst performed case-control studies to identify SNPs associated with NSCLC risks. To increase the strength of evidence, we integrated information gathered from various bioinformatics platforms and examined the potential prognostic values of lncRNAs at these NSCLC-associated loci.
Next, we performed functional studies to investigate the underlying molecular mechanisms of these lncRNAs. LncRNAs have been known to function primarily through their interaction with microRNAs, DNA, RNA or RNA-binding proteins via the competitive endogenous RNA network (ceRNA network) [12][13][14].
Therefore, we particularly focused on investigating the potential ceRNA interactions in the posttranscriptional gene regulation process.
We now show that a lncRNA OSER1-AS1, near the NSCLC-associated SNP rs4142441, was differentially expressed between lung cancer and adjacent normal tissues. We examined the effect of OSER1-AS1 knockdown and overexpression on cell proliferation, migration and invasion in NSCLC cells. We found that OSER1-AS1 was transcriptionally repressed by MYC at the promoter, and down-regulated by microRNA hsa-miR-17-5p at the 3'-end. Interestingly, OSER1-AS1 may function as a decoy for RNAbinding protein ELAV-like protein 1 (ELAVL1, or human antigen R (HuR)), which was one of the most widely studied regulators of cytoplasmic mRNA stability.

Selection of SNPs and genotyping
We selected 16 SNPs located near potential lung cancer-related lncRNAs following the procedure described in Supplementary Fig. 1. Additional details about the selecting procedure are available in the Supplementary Methods. The selected SNPs (Supplementary Table S1) were used for case-control analysis.
Case-control analysis New cases diagnosed with NSCLC were collected from Xinqiao Hospital of the Army Medical University (Third Military Medical University) in Chongqing, China. Healthy controls were collected from the annual physical examination group in the same hospital. Additional details about the inclusion/exclusion criteria for case and controls are described in the Supplementary Methods. We matched up 645 patients on age and sex with 748 healthy controls. Demographic and other risk information was obtained from subjects via a combination of a structured subject interview and medical records. Blood samples were collected from controls and patients prior to any treatment. All subjects provided informed written consent. Research protocol was approved by the ethics committee of the Army Medical University.

Human tissue samples
Tissue specimens were collected from lung cancer patients prior to any radiation or chemotherapy during operation. The freshly frozen lung tumors and matched normal lung tissues were sectioned and reviewed by a pathologist to con rm the diagnosis of lung cancer, histological grade, tumor purity, and lack of tumor contamination in the normal lung. Tumor samples with ≥ 70% tumor-cell content and matched normal lung tissues were used in the study.

Cell culture and treatment
The lung cancer cell lines A549, SPCA1, and H1299 were obtained from the Cell Bank of the Chinese Academy of Science (Shanghai, China) and the American Type Culture Collection (ATCC, Manassas, VA, USA), cultured in RPMI-1640 (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (HyClone, Logan, UT, USA).

qRT-PCR analysis
Total RNA was extracted using the TRIzol reagent (TaKaRa, Dalian, China). The cDNAs were ampli ed using PrimeScript™ RT reagent Kit with gDNA Eraser (TaKaRa). Real-time RT-PCR assay was performed using a SYBR PrimeScript RT-PCR Kit according to the standard manufacturer's instructions (TaKaRa, Dalian, China). Results were normalized to the housekeeping gene β-actin. The nuclear and cytoplasmic RNA were isolated using PARIS™ Kit (Invitrogen, NY, USA). Primers sequences were provided in Supplementary Table S2.

RNA Fluorescence in situ hybridization
Cy3-labeled OSER1-AS1 probes were synthesized by GenePharma (Shanghai, China). RNA-FISH was performed as described by Zhou et al [15] following the manufacturer's instructions. U6 snRNA and 18S rRNA probes were purchased from GenePharma (Shanghai, China) and were used as nuclear and cytoplasmic localization control, respectively.
Cell Counting Kit-8 assay, colony formation assay, cell migration and invasion assays in vitro Cell Counting Kit-8 (CCK-8) assay, colony formation assay, cell migration and invasion assays were performed as described in Yuan et al., 2016. [17] Page 6/28

Animal experiments in vivo
For in vivo tumorigenicity, ten male BALB/c-nude mice (4-week-old) were randomly divided into two groups, with ve mice in each group. Stable transfected A549 cells (5.0 × 10 6 ) were injected subcutaneously into the left anks of the nude mice (100 ul per mouse). The tumor volume was calculated using the equation V = 0.5 × D × d 2 (V, volume; D, longitudinal diameter; d, latitudinal diameter). We observed the tumor growth for 5 weeks. For metastasis model, ten male BALB/c mice (4-week-old) were randomly divided into two groups, with ve mice in each group. Stable transfected A549 cells (1 × 10 6 ) were injected into their tail veins (100 ul per mouse). The mice were sacri ced 5 weeks after injection and the lungs were removed for further analysis. All experimental animal procedures were approved by the Institutional Animal Care and Use Committee of Third Military Medical University.

Luciferase reporter assay
To evaluate the miRNA/ELAVL1-lncRNA interaction by luciferase reporter assay, the hsa-miR-17-5p binding sites of OSER1-AS1 3'-end region were inserted into the Pischeck-2 vector (Promega), and ELAVL1 was inserted into pcDNA3.1 vector (Sangon, Shanghai, China). Fire y and Renilla luciferase activities were measured 48 hours after transfection using the Dual-Luciferase Assay System (Promega). The relative luciferase activity was calculated using renilla/ re y luciferase activity.
To evaluate the binding of MYC to the promoter of OSER1-AS1, the OSER1-AS1 promoter sequence (-1000 bp ~ + 1000 bp) was synthesized and subcloned into the luciferase reporter vector pGL3-basic (Promega), for which two versions were constructed: WT with allele A for rs4142441 and MUT with allele G for rs4142441. The plasmids were co-transfected with pcDNA3.1-MYC as well as pRL-SV40 Renilla luciferase plasmid (Promega) for internal control.

Chromatin immunoprecipitation assay
Chromatin immunoprecipitation (ChIP) assay was performed following the protocol of ChIP Assay Kit (Beyotime Institute of Biotechnology, Jiangsu, China). Brie y, crosslinked chromatin was prepared with 1% formaldehyde for 10 min at 37 °C and the DNA was shredded to an average length of 200-1000 bp by sonication. Immunoprecipitation were conducted using rabbit polyclonal to MYC (Abcam) or IgG control. Precipitated DNA was ampli ed by PCR. Primers sequences are provided in Supplementary Table   S2. RNA-binding protein immunoprecipitation assay RNA-binding protein immunoprecipitation assay was performed following the protocol of Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (MilliporeSigma, US). The nuclear and cytoplasmic RNA was separated by using the Membrane, Nuclear and Cytoplasmic Protein Extraction kit (Sangon) with added RNase Inhibitor (Beyotime Institute of Biotechnology, Jiangsu, China) to a nal concentration of 1 U/ul. After the lysate was prepared, 100 ul of whole/nuclear/cytoplasmic lysate were incubated at 4 o C overnight with 50 ul of magnetic beads pre-coated with 5ug ELAVL1 antibody or 5ug of AGO2 antibody. Another 100 ul aliquot of cell lysate was incubated with 50 ul of magnetic beads pre-coated with 5 ug of IgG antibody as a negative control.
Bioinformatics analysis for TCGA-LUAD data Additional descriptions of the bioinformatic analysis for TCGA-LUAD data are shown in the Supplementary Methods.

Statistical analysis
We performed a multivariate logistic regression analysis on the association of SNPs with lung cancer by adjusting for age, sex, and smoking. Cell growth and migration results were evaluated using the twotailed Student's t-test. Gene expression analyses were evaluated using the two-tailed Student's t-test or Mann-Whitney U-test. Pearson correlation was performed to analyze the correlation between miRNA and mRNA expressions. A two-sided P-value less than 0.05 was taken as statistically signi cant. Statistical analyses were performed using the software STATA version 12.0 (STATA Corp., Texas, USA). and software R (version 3.4.2).

Results
Rs4142441 associated with NSCLC risk in non-smokers is located in the promoter of OSER1-AS1 We performed case-control analysis for the 16 SNPs (Supplementary Table S1) selected through the bioinformatic pipeline as described in Supplementary Figure S1 in 626 cases and 736 controls. Baseline characteristics of cases and controls were shown in Supplementary Table S3. We found no signi cant associations between any of the SNPs and lung cancer risks (Table 1). However, after stratifying by smoking status, we identi ed one signi cant association in the non-smoker group at rs4142441 in the promoter region of lncRNA OSER1-AS1 and lung cancer risk (OR = 1.712;95% CI:1.137-2.579 P = 0.01) ( Table 1).

Rs4142441 is associated with the expression of OSER1-AS1 in vitro
To investigate whether rs4142441 was associated with the expression level of OSER1-AS1, we performed luciferase reporter assays in H1299 cells. We constructed a luciferase reporter vector containing the promoter region of OSER1-AS1 (1000bp on both sides of rs4142441). Two versions of the promoter sequences were constructed: rs4142441-WT (with allele A for rs4142441) or rs4142441-MUT (with allele G for rs4142441) (Fig. 1a). We observed that the expression of re y luciferase fused to the promoter carrying rs4142441-MUT(G) was signi cantly lower than that of re y luciferase fused to the promoter carrying rs4142441-WT(A) (Mann-Whitney U-test, P = 0.0005) (Fig. 1b). To further validate the results, we constructed a pcDNA3.1 vector containing the WT and MUT rs4142441 alleles of OSER1-AS1 promoter sequence and the full-length cDNA sequence of OSER1-AS1 (Transcripts ID: ENST00000442383.1) (Fig. 1c). We transiently transfected the plasmids into H1299 cells and measured the OSER1-AS1 expression levels using qRT-PCR. OSER1-AS1 expression levels were signi cantly lower with vectors carrying rs4142441-MUT(G) allele (Mann-Whitney U-test, P = 0.0041) (Fig. 1d). Taken together, these pieces of evidence indicated that the allele G of rs4142441 was associated with down-regulation of OSER1-AS1.

OSER1-AS1 down-regulation is associated with poor overall survival in TCGA LUAD patients
We measured the expression level of OSER1-AS1 in 129 paired NSCLC tumor and adjacent normal lung tissues using qRT-PCR. The clinical characteristic of the 129 NSCLC patients were summarized in Supplementary Table S4. We rst examined the correlation between the expression level of OSER1-AS1 and clinical characteristics of the NSCLC patients. OSER1-AS1 was signi cantly lower among smokers (P = 0.001, Supplementary Table S5 Table S6). Collectively, these results revealed that the low expression levels of OSER1-AS1, which was associated with the allele G of rs4142441, were also associated with poor overall survival in TCGA-LUAD patients among non-smokers.
We further used an RNA uorescence in situ hybridization (RNA-FISH) assay to investigate the subcellular localization of OSER1-AS1. OSER1-AS1 was predominantly observed in cell cytoplasm (Supplementary Figure S3a). The qRT-PCR quali cation of nuclear and cytoplasmic RNA fractions further validated that OSER1-AS1 was mainly located in the cytoplasm (cytoplasm:nucleus ratio, 7:1) (Supplementary Figure  S3b).

OSER1-AS1 inhibits proliferation, migration, and invasion of NSCLC cells in vitro
To evaluate the potential role of OSER1-AS1 in NSCLC, we established gain-of-function cell models by transfecting pcDNA3.1-OSER1-AS1 expressing vectors into the H1299 and SPCA1 cells (Supplementary Figure S4a). In addition, we transfected H1299 and SPCA1 cells with two siRNAs targeting independent regions of OSER1-AS1 to investigate the effect of down-regulation of OSER1-AS1 on cell proliferation, migration and invasion (Supplementary Figure S4b). We examined the effects of OSER1-AS1 overexpression on cell proliferation, migration and invasion. CCK-8 assays showed that overexpression of OSER1-AS1 signi cantly decreased the proliferation of H1299 and SPCA1 cells (Supplementary Figure  S5a). Moreover, the in vitro transwell assays showed that overexpression of OSER1-AS1 signi cantly decreased the migration and invasion of H1299 and SPCA1 cells compared with vector control (Supplementary Figure S5c, d). The down-regulation of OSER1-AS1 signi cantly increased the proliferation, migration and invasion of H1299 and SPCA1 cells compared with vector control (Supplementary Figure S5b, e, f).

OSER1-AS1 represses tumor growth and metastasis in vivo
To investigate whether OSER1-AS1 suppresses tumorigenesis in vivo, we implanted A549 cells stably transfected with OSER1-AS1 or control vector subcutaneously in immunode ciency mice. The tumor mass and volume in OSER1-AS1 overexpression group were signi cantly smaller and lighter than those in control group (Fig. 3a, b and c). To validate the anti-metastatic effects of OSER1-AS1 in vivo, A549 cells stably transfected with OSER1-AS1 or control vector were injected into the tail veins of nude mice. The number of metastatic nodules on the surface of the lung was signi cantly decreased in mice receiving OSER1-AS1 stable overexpressing A549 cells compared with the control group (Fig. 3d). Hematoxylin and eosin (H&E) staining of the mice lung tissue slice con rmed that less metastatic nodules were present in the OSER1-AS1 overexpression group than the control group. Furthermore, proliferation marker Ki-67 was evaluated through immunohistochemistry in tumor tissues. Ki-67 expression levels were also signi cantly lower in OSER1-AS1 overexpression group than those in control vector group (Fig. 3e).

OSER1-AS1 promoter is suppressed by MYC
To explore the potential transcription factor that regulates the expression of OSER1-AS1, we searched the Transcription Factor ChIP-seq from ENCODE (Txn Factor ChIP) track in the UCSC genome browser. We found a binding site for MYC within the OSER1-AS1 promoter region (-333 bp ~ + 116 bp TSS of OSER1-AS1), covering rs4142441 (Fig. 4a). Thus, we then examined the regulation effect of MYC on OSER1-AS1 expression. By using ChIP experiments, we con rmed the direct interaction of MYC to OSER1-AS1 promoter using two pairs of primers binding to different positions of the potential binding region of MYC (Fig. 4b & c). We next performed luciferase reporter assays to further investigate the effect of the WT and MUT alleles of rs4142441 on the regulatory e ciency of MYC. The sequences of OSER1-AS1 promoter region (-1000 bp ~ + 1000 bp) were inserted into the reporter plasmid. Co-transfection of the MYCoverexpression vectors with OSER1-AS1 promoter luciferase construct signi cantly repressed the luciferase activity. Moreover, the luciferase activity from the promoter carrying rs4142441-MUT(G) was signi cantly lower than that of re y luciferase fused to the promoter carrying rs4142441-MUT(A), both with or without MYC co-transfection (P < 0.0001) (Fig. 4d). In addition, we transiently transfected the MYC overexpression plasmids into H1299 cells and measured the OSER1-AS1 expression levels using qRT-PCR. The results from qRT-PCR showed that MYC-overexpression vectors signi cantly decreased the expression of OSER1-AS1 in comparison to the control vector (P < 0.05) (Fig. 4e). These lines of evidence demonstrated that MYC suppressed the transcription of OSER1-AS1, and the allele G of rs4142441 was associated with down-regulation of OSER1-AS1 regardless of the binding status of MYC.
In addition, up-and down-stream in close proximity of this hsa-miR-17-5p target site, 4 binding sites of RNA-binding protein ELAVL1 (ELAV-like protein 1 or ELAVL1) were reported by independent CLIP-seq experiments (Supplementary Figure S6b). ELAVL1 was an widely studied RNA-binding protein that played a role in promoting cell proliferation by stabilizing mRNAs of oncogenes involved in cell cycle regulation [22]. Therefore, we next explored the impacts of hsa-miR-17-5p and ELAVL1 binding on the expression levels of OSER1-AS1. We rst investigated the differential luciferase activity of Pischeck-2 vector containing the OSER1-AS1 3'-end region co-transfected with hsa-miR-17-5p mimic, NC mimic, inhibitor and NC-inhibitor. We found out that hsa-miR-17-5p mimics signi cantly reduced the luciferase activities of the Pischeck-2 reporter containing the OSER1-AS1 3'-end region, whereas inhibitors of hsa-miR-17-5p caused the opposite effect (Supplementary Figure S6c). Subsequently, we investigated the expression levels of OSER1-AS1 after transfection with hsa-miR-17-5p mimic, NC mimic, inhibitor and NC-inhibitor.

OSER1-AS1 sequesters ELAVL1 by direct lncRNA-protein interaction
We next investigated the biological mechanism by which OSER1-AS1 suppressing tumorigenesis. The activity of ELAVL1 has been known to be dependent on its subcellular distribution: it predominantly localizes within the nucleus of resting cells. Under certain biological circumstances, it transports to the cytoplasm and binds to its target mRNA's 3'UTR [23]. The ELAVL1-mRNA complex would prevent the mRNA from degradation and thereby increase its translational level [24]. Since OSER1-AS1 were distributed predominantly in cell cytoplasm (Supplementary Figure S3), we hypothesized that it might function as natural sponges to prevent ELAVL1 from binding its target mRNAs in the cytoplasm.
To further investigate the potential in uence of OSER1-AS1 on the subcellular localization of ELAVL1, we performed western blot in whole cell, nuclear and cytoplasmic cell lysate prepared from H1299 cells transfected with scramble siRNA (NC), two separate siRNAs targeting OSER1-AS1 (siRNA-323 and siRNA-373), stable empty vector and OSER1-AS1 overexpression vector. Consistent with the previous studies, our results showed that ELAVL1 was predominantly distributed in the nucleus, and AGO2 was in both nucleus and cytoplasm [23]. Although we found no signi cant alteration of ELAVL1 and AGO2 expressions in the whole cell, OSER1-AS1 changed the subcellular localization of ELAVL1. Knockdown of OSER1-AS1 attenuated the translocation of ELAVL1 from nucleus to cytoplasm in H1299 cells, whereas OSER1-AS1 overexpression increased the amount of cytoplasmic ELAVL1 proteins (Fig. 5b).
Our ndings in RIP and western blot experiments together suggested that OSER1-AS1 directly bound with ELAVL1 and the up-regulation of OSER1-AS1 caused the accumulation of ELAVL1 proteins in the cytoplasm.
OSER1-AS1 regulates the target genes of ELAVL1 and miR-17-5p MYC has been reported to stimulate the expression of microRNAs in the miR-17-92 cluster, including miR-17-5p [26][27][28], which in turn suppressed the expression of the pro-apoptotic protein BCL2L11 in lymphoma [29] and in NSCLC [30]. On the other hand, the RBP ELAVL1 was known to up-regulate the translational level of MYC [31][32][33]. Considering that the 3'-end of OSER1-AS1 was bound by both miR-17-5p and ELAVL1, we hypothesized that OSER1-AS1 could act as a miR-17-5p sponge to change the expression of its target gene BCL2L11. By the same theory, OSER1-AS1 may also sponge ELAVL1 away from its target MYC mRNAs, leading to decreased translational e ciency and decreased MYC protein levels. Hence, we performed western blot to assess MYC and BCL2L11 protein level after knockdown and overexpression of OSER1-AS1. The result of western blotting indicated that knockdown of OSER1-AS1 increased the MYC expression, whereas the overexpression of OSER1-AS1 decreased the MYC expression. The change of MYC expression was more profound in the nucleus, consistent with the knowledge that MYC was a transcription factor that mainly executed its functions in the nucleus [28]. The in uence of OSER1-AS1 on the BCL2L11 expression was in the opposite direction: knockdown of OSER1-AS1 decreased the BCL2L11 expression, whereas the overexpression of OSER1-AS1 increased the BCL2L11 expression. The pattern of BCL2L11 expression change was similar in the nuclear and cytoplasmic fraction, consistent with the knowledge that BCL2L11 was distributed both in the cytoplasm and in the nucleus (Fig. 6).

Discussion
In this study, we tested a list of SNPs located near lncRNAs to detect their associations with lung cancer risks. From one signi cantly associated SNP rs4142441, we narrowed down our focus to lncRNA OSER1-AS1. We demonstrated that the mutant allele (G) of rs4142441 was associated with higher lung cancer risk and lower expression level of OSER1-AS1. Rs4142441 was previously reported in a large-scale GWAS study as associated with monocyte count in whole blood. The allele G was associated with increased monocyte count (Beta coe cient = 0.033; [95% CI 0.023-0.043]; P = 4 10 −11 ) [34]. An elevated peripheral monocyte count has been reported to have a poor prognosis impact on lung adenocarcinoma [35].
The qRT-PCR analysis con rmed OSER1-AS1 was signi cantly down-regulated in lung adenocarcinoma tissues in comparison to adjacent normal tissues. Moreover, the differential expression was more profound among non-smokers. Therefore, we hypothesized that OSER1-AS1 played a tumor suppressive × role in NSCLC. We then investigated the prognostic value of OSER1-AS1 using clinical data from TCGA public database by Kaplan-Meier survival analysis. We found that high expression of OSER1-AS1 was associated with better overall survival in TCGA-LUAD non-smoker patients. In addition, OSER1-AS1 suppressed lung cancer cells proliferation, migration, and invasion in vitro and promoted tumorigenesis in vivo.
Furthermore, our results showed that transcription factor MYC suppressed OSER1-AS1 expression.
Transcription factor MYC was a commonly known regulator which activated growth-related genes and suppressed genes involved in cell cycle arrest, cell adhesion, and cell-cell communication [36]. In addition, MYC was up-regulated in human NSCLC cells [37,38] and MYC depletion has been reported to be able to reverses immune evasion and enables effective treatment of lung cancer [39]. MYC was known to bind to the core promoter of the genes it repressed directly [40], and itself has been known to be regulated by a number of lncRNAs [41]. The results from our ChIP experiment con rmed the directing binding of MYC at the promoter region of OSER1-AS1. The binding site of MYC covered the SNP rs4142441, but its allelic status did not interfere with MYC binding capacity. One possible explanation was that rs4142441 in uenced the expression level of OSER1-AS1 via other transcription factors such as RNA polymerase II. The exact biological mechanism that may cause the association between allele G of rs4142441 and the lower expression level of OSER1-AS1 remains to be explored.
One the other hand, the 3'-end of OSER1-AS1 was bound competitively by hsa-miR-17-5p and RNAbinding protein ELAVL1. The RNA sequence of OSER1-AS1 contained more than 9 ARE motifs required for ELAVL1 binding, and the nearest ARE motif was only 51 bp 5' upstream of the miR-17-5p binding site.
Some studies have implicated the roles of ELAVL1 in affecting the binding capacity of AGO2 and miRNAs [23], and the effects could be either a competition [42] or cooperation [43,44], probably depending on the conformational changes of the 3'UTR caused by the initial binding. In the study of Chang et al. (2013), ELAVL1 has been reported to antagonize the suppressive effect of miR-200b on VEGF-A expression by competitive binding [45]. However, unlike VEGF-A, OSER1-AS1 as a lncRNA cannot be stabilized by ELAVL1 and translated into proteins. Therefore, we hypothesized that OSER1-AS1 may execute its biological function as a decoy by forming a stable RNA-protein complex with ELAVL1 and sequestering it from its target mRNAs.
Several studies had shown that some lncRNAs could serve as a sponge to restrict RBPs' availability to its target mRNAs [46][47][48]. The post-transcriptional regulations by ELAVL1 mainly took place in the cytoplasm [49]. Our RNA FISH assays demonstrated that OSER1-AS1 located mainly in the cytoplasm (cytoplasm:nucleus ratio 7:1), so the subcellular localization of OSER1-AS1 was consistent with our "ELAVL1 sponge" hypothesis.
To test this hypothesis, we performed RIP assays in the whole cell lysate, nuclear lysate, and cytoplasmic lysate separately. The results from RIP experiments con rmed that physical interaction happened between OSER1-AS1 and ELAVL1. In the whole cell, OSER1-AS1 preferentially formed RBP-RNA complex with ELAVL1. Moreover, OSER1-AS1 was also bound by AGO2, which is consistent with our nding that OSER1-AS1 was targeted by hsa-miR-17-5p. In the nucleus, OSER1-AS1 was predominantly bound by ELAVL1. However, in the cytoplasm, OSER1-AS1 was mainly bound by AGO2, suggesting it was under the suppressive status in RISC complexes in the cytoplasm. This seems to be counter-intuitive to our "ELAVL1 sponge" hypothesis, because ELAVL1 was known to carry out its mRNA stabilizer function in the cytoplasm. One possible explanation could be that ELAVL1 mainly localized in the nucleus, the cytoplasmic level of ELAVL1 was extremely low so the presence ELAVL1-OSER1-AS1 complex would be di cult to be accurately quanti ed in RIP assays.
Then we interfered the expression level of OSER1-AS1 to explore further on whether OSER1-AS1 had any impact on the ELAVL1 protein level. Our results indicated that, although OSER1-AS1 had no signi cant in uence on the overall protein level of ELAVL1, it changed its subcellular localization. Overexpression of OSER1-AS1 resulted in the accumulation of ELAVL1 proteins in the cytoplasm, whereas the knockdown of OSER1-AS1 caused the opposite effect. However, this was in con ict with the ndings from previous studies which showed higher cytoplasmic level of ELAVL1 associated with increased tumorigenic activity and poor prognostic outcome in NSCLC [50,51]. One possible explanation might be that, even though OSER1-AS1 increased the cytoplasmic level of ELAVL1, these ELAVL1 were not functional because they were not be able to bind their target mRNAs, and they could not be transported back to the nucleus.
Subsequently, we investigated whether OSER1-AS1 changed the level of MYC, which was both a regulator of OSER1-AS1 and a target gene of ELAVL1. We found that the knockdown of OSER1-AS1 increased the MYC expression, whereas the overexpression of OSER1-AS1 decreased the MYC expression. Taking together with the nding that MYC suppressed the transcriptional level of OSER1-AS1, we concluded that MYC and OSER1-AS1 controlled each other in a negative feedback loop (Fig. 7). In addition, we found that OSER1-AS1 positively regulate the expression level of BCL2L11, which was targeted by miR-17-5p, further validating the hypothesis that OSER1-AS1 functioned as a miRNA sponge of miR-17-5p.
Taking together, we demonstrated that OSER1-AS1 exercised tumor suppressive functions in NSCLC by acting as an ELAVL1 decoy and sequestered ELAVL1 from its target mRNAs involved in cell proliferation, migration and tumorigenicity. In company with previous studies [22,23,52,53], we showed that ELAVL1 can bind to the 3'-end of lncRNA OSER1-AS1 in competition with a microRNA hsa-miR-17-5p. These ndings highlighted that, other than sponging microRNAs, another possible mechanism of lncRNAs to exercise its tumor-suppressing function might be sponging oncogenic RBPs away from its target mRNAs.

Conclusion
Our study demonstrates that OSER1-AS1 is down-regulated in tumor and acts as a tumor suppressor in NSCLC. OSER1-AS1 is co-regulated by SNP rs4142441 and MYC at the promoter, and competitively targeted by both microRNA hsa-miR-17-5p and RBP ELAVL1 at the 3'-end. It performs tumor suppressive function by forming an RNA-protein complex with ELAVL1 and sequester it from binding and stabilizing its target mRNAs, numerous of which are implicated in carcinogenesis. Our results suggest that OSER1-AS1 could play important roles in ELAVL1-regulated mRNA stability. These ndings could provide biological insight into the regulation of the ceRNA network in NSCLC, and give new clues for future development of new therapeutic targets and biomarkers.

Declarations
The case control datasets generated and/or analysed during the current study are not publicly available due to individual privacy but are available from the corresponding author on reasonable request and with

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.