HOXA3-Induced Exosomal lncRNA RNF144A-AS1 Promotes Liver Metastasis of Gastric Cancer via Interacting with PUF60 Protein and Sponging miR-361-3p


 Background: Long non-coding RNA (lncRNA) plays an important regulatory role in the development and progression of gastric cancer(GC). However, the biological role and potential molecular mechanism of lncRNA RNF144A-AS1 in liver metastasis of gastric cancer(GCLM) remain unclear. Methods: LncRNAs associated with GCLM were identified by microarray. Gain or loss of function approaches was used to investigate the biological functions of RNF144A-AS1. The expression of RNF144A-AS1 was detected by real-time quantitative PCR, and the molecular mechanism of RNF144A-AS1 was investigated by RNA pull-down assay, Western blot, luciferase activity, RNA immunoprecipitation. Finally, nude mouse models and bioluminescence imaging were used to verify the role of RNF144A-AS1 in GCLM in vivo. Results:We identified that RNF144A-AS1 was highly expressed in GC by microarray.RNF144A-AS1 was related to GCLM,and poor DFS and OS. The expression of RNF144A-AS1 in serum exosomes of GC patients was significantly increased, while the level of RNF144A-AS1 was significantly decreased after GC resection. Overexpression of RNF144A-AS1 in exosomes can promote the growth and invasion of co-cultured GC cells in vitro. The down-regulation of RNF144A-AS1 induced the proliferation, migration, and invasion of GC cells in vitro and in vivo. Mechanistically, the transcription factor HOXA3 bound to the promoter region of RNF144A-AS1 could activate RNF144A-AS1, and RNF144A-AS1 promotes GCLM via interacting with PUF60 protein and sponging miR-361-3p.Conclusions: The present study indicated that exosome RNF144A-AS1 overexpression contributes to GCLM and would be a promising biomarker for the diagnosis and prognosis of GCLM.


Background
Gastric cancer(GC) is a serious threat to human health and is one of the ve major cancers in the world (1). Early GC is di cult to diagnose, and most of the patients are in the middle and late stage when they see a doctor. Invasion and metastasis of GC often occur in advanced GC, which seriously affects the treatment of GC patients, and the ve-year survival rate is less than 30% (2,3). Tumor metastasis is one of the main causes of death. Liver is the main organ of GC metastasis, and the incidence of liver metastasis of GC (GCLM) is as high as 44%. The ve-year survival rate of patients with GCLM is only about 10% (4,5). At present, early diagnosis of GCLM is still a di cult problem. Some tumor molecular markers have been used in clinical early detection and auxiliary diagnosis, such as CEA, CA125, CA72-4, CA199, etc., but their speci city is low (6). In order to further improve the prognosis of GC patients, it is necessary to discover new genes related to GCLM.
Long non-coding RNA (lncRNA) is a type of RNA with a length of more than 200bp with limited proteincoding potential. Many lncRNAs have been reported to function in both physiological and pathological processes (7,8). LncRNAs play crucial roles in multiple steps of gene regulation by serving as guides of chromatin-modifying complexes (9, 10) and transcription factors (11,12), scaffolds of protein-protein interactions (12,13), decoys of proteins (14,15), sponges for miRNAs (16-18), etc. Aberrant expression of lncRNAs is common in cancer (19), and lncRNAs have been found to involve in various aspects of cancer development such as cell growth, survival, invasion and metastasis (9,15,20,21). Although lncRNA has been widely acknowledged as a new contributor to human cancer, only a small number of lncRNAs have been functionally characterized in GC. The molecular mechanism of lncRNA involved in regulating GCLM is not fully understood. Exosomes are small vesicles that are rich in biologically active molecules such as protein, mRNA, miRNA, etc., which can be transported between cells and mediate the transfer of materials and information between cells, thereby affecting the physiological functions of cells (22). Exosome mainly affects tumor-related pathways by mediating substance transport, such as hypoxia-mediated EMT, angiogenesis, and tumor microenvironment involved in regulating tumor metastasis (22,23,24). Tumors will continue to release exosomes into the surrounding environment during the growth process. Exosomes are rich in ncRNAs. Therefore, the detection of exosomes and exosomal ncRNAs in body uids is helpful for tumor diagnosis and prognosis evaluation (25)(26)(27). Some scholars have found that malignant solid tumors such as breast cancer and GC can secrete lncRNAs to promote the occurrence and development of malignant tumors, which is expected to be used for the diagnosis and monitoring of malignant tumors (28-31).
Poly-U binding splicing factor 60KDa (PUF60) is a splicing factor related to U2, which plays an important role in the recognition of the 3'splicing site in the early stage of spliceosome assembly (32). Studies have found that PUF60 protein is highly expressed in colorectal cancer (33), GC (34), liver cancer (35), ovarian cancer (36) and other tumors, and is closely related to the occurrence and development of tumors. Some scholars have found that PUF60 can be detected in the serum of patients with early colon cancer, and the detection rate before surgery is signi cantly higher than that after surgery (37). This shows that PUF60 can be used as a combined or independent detection indicator for the diagnosis of colorectal cancer and other tumors, providing a new target for gene therapy.
The insulin-like growth factor-2 mRNA-binding protein 1(IGF2BP1) is a member of the conserved single-stranded RNA binding protein family (IGF2BP1-3).It is expressed only in a few normal adult tissues, but is highly expressed in fetal tissues and many cancer tissues, and plays an important role in embryogenesis, carcinogenesis, and drug resistance (38, 39). Studies have found that IGF2BP1 plays an important role in the proliferation, adhesion, migration and invasion of some cancer cells (40,41). In addition, IGF2BP1 is highly expressed in GC tissues and is associated with short overall survival(OS) and poor prognosis (42,43). Therefore, IGF2BP1 is considered to be one of the promising therapeutic targets for the treatment of cancer.
POU domain class 2 transcription factor 2(POU2) is a B-cell-speci c octamer transcription factor (44). Previous studies have shown that POU2F2 is usually expressed in B cells and B cell line tumor cells to regulate B cell proliferation and cell differentiation (44). Recent studies have found that POU2F2 is overexpressed in pancreatic cancer (45), GC (46,47), cervical cancer (48), and other epithelial malignancies. Wang et al. (47) found that the expression of POU2F2 in GC cells with high metastatic potential and GC tissues with lymph node or distant metastasis was signi cantly increased, and the survival time of GC patients with positive POU2F2 expression was shorter than that of GC patients with negative POU2F2 expression. However, the potential function and exact mechanism of POU2F2 in GC are still unclear.
In this study, we found that RNF144A-AS1 was overexpressed in the tumors and serum exosomes of GC patients, and was related to the pathological staging, poor prognosis, and GCLM. RNF144A-AS1 promotes GCLM in vivo and in vitro by interacting with PUF60 protein and sponging miR-361-3p. It was further found that the transcription factor HOXA3 can activate and promote the overexpression of RNF144A-AS1. This study provides a new perspective for exploring the pathogenesis of GCLM.

Patient samples
The 6 cases of GC and its adjacent normal tissues used for LncRNA high-throughput microarray detection were from patients with GC complicated by liver metastasis in general surgery who underwent palliative surgery due to obstruction or gastric bleeding. 120 cases of GC patients used for large sample veri cation were primary GC patients in the general surgery department of our hospital, all of which were con rmed by surgery and underwent D2 radical resection with R0 resection. All GC patients did not receive chemotherapy, radiotherapy, surgery, and other treatments before surgery. Peripheral blood was collected from 20 patients with GC and 20 healthy controls (not diagnosed with cancer). Preparation of exosomes from peripheral blood. All specimens were collected according to institutional protocol. In addition, the clinicopathological features of patients were collected, including age, gender, tumor location, tumor size, degree of differentiation, Lauren grade, TNM stage, etc.

Microarray analysis
Total RNAs were isolated from the paired tissue samples of six liver metastasis and their adjacent normal tissues and puri ed using TRIzol reagent (Invitrogen, Carlsbad, CA) and RNeasy mini kit (Qiagen Inc, Valencia, CA) according to the manufacturer's protocol. Following RNA isolation and cDNA synthesis, biotin-labeled cRNA was labeled and hybridized to the 8 × 60 K LncRNA Expression Microarray (ArrayStar). After having washed the slides, the arrays were scanned by the Agilent Scanner G2505C. Agilent Feature Extraction software (version 11.0.1.1) was used to analyze acquired array images. Quantile normalization and subsequent data processing were performed using the GeneSpring GX v11.5.1 software package (Agilent Technologies).

Cell culture
The AGS cell line was purchased from the American Type Culture Bank (ATCC,USA), and the HEK293T, HGC27, MKN45, and GES1 cell lines were purchased from the Shanghai Type Culture Bank of the Chinese Academy of Sciences. HEK293T, HGC27, MKN45, and GES1 cells were cultured in RPMI1640 medium (Gibco, Carlsad, CA, USA). AGS cells were cultured in F12K medium (Wisent, Canada). All cell lines were added with 100 µg/ml streptomycin, 100 U/ml penicillin, and 10% fetal bovine serum in a humidi ed atmosphere of 5% CO 2 at 37°C.

Isolation and identi cation of exosomes
Collect exosomes from 20 ml culture medium (1×10 7 cells). Collect the medium on ice, centrifuge at 800×g for 10 min to pellet the cells, and centrifuge at 12000×g for 30 min to remove cell debris. Centrifuge with an SW32 rotor (Beckman Coulter) at 100,000×g for 2 h to separate exosomes from the supernatant. The exosomal pellets were washed once in a large volume of phosphate buffer, and resuspended in 100 µl of phosphate buffer.
Human serum exosomes were obtained using ExoQuick Exosome precipitate (SBI, CA, USA) according to the instruction manual. In short, the serum was collected and centrifuged at 3000×g for 15 min. The 63 µL Exoquick Exosome precipitate was added to 250 µL supernatant, the serum was refrigerated at 4°C for 30 min, and centrifuged at 1500 ×g for 30 min. Then the Exosome pellets were suspended in 100 µL with 1×PBS.The particle size and shape of exosomes were then identi ed by transmission electron microscopy (TEM)(Philips Tecnai 20, Netherlands).Western blot method was used to identify exosomal protein markers.Nanoparticle tracking analysis (NTA) was used to measure the total amount of exosomes.

RNA preparation and quantitative real-time PCR (qRT-PCR)
Total RNA was extracted from cells or tissues using TRIzol reagent (Invitgen, MA, USA). The nucleus and cytoplasm were extracted with the Paris™ kit (Thermo Fisher, Massachusetts, USA). The extracted RNA was subjected to quantitative polymerase chain reaction using HiScript Q RT SuperMix (Vazyme, Jiangsu,China). ABI Prism7900 sequence detection system (Applied Biosystems Canada) was used to perform quantitative RTPCR with SYBR Green PCR Master Mix (Vazyme), GAPDH was used as an internal control, and the results of each sample were normalized to GAPDH expression. For RNase R treatment, incubate 2 µg of total RNA at 37°C and add or not add 3 U/µg of RNase R (Epicentre Technologies, WI, USA) in 1× reaction buffer. (The primers are listed in Supplementary Table S1.)

Plasmid and siRNA transfection and lentiviral transduction
The siRNAs of RNF144A-AS1 and HOXA3,and miRNA mimics or inhibitors were designed and synthesized(RiboBio,Guangzhou,China).Plasmids pcDNA3.1-RNF144A-AS1 and pcDNA3.1-PUF60 were designed and synthesized by Jiangsu Genssee Biological Technology Co. Plasmids and miRNA mimics or inhibitors were transfected with liposome 3000 (Invitgen). The cells were transfected with siRNAs with DharmaFECT4 (DharmaCon, IL, USA). An shRNA lentiviral vector (pGLV3/GFP/Puro) targeting RNF144A-AS1 was constructed by GenePharma (Shanghai, China) and transfected into HGC27 cells. A stable cell line was obtained by screening with puromycin. (RNF144A-AS1 and HOXA3 gene siRNA sequences are listed in Supplementary Table S1.) RNA pull-down assay and mass spectrometry According to the manufacturer's protocol, the magnetic RNA protein pull-down kit (TERMO) was used for lncRNA pull-down assay. 10 7 HGC27 cells were transfected with RNF144A-AS1 overexpressing vector or control vector. After 48 h, the total RNA of the two groups of cells was extracted and incubated with 100 nmol probes at 70°C for 5 min. Then the ribonucleic acid was slowly cooled to room temperature, 50 µl streptavidin magnetic beads were added, and the mixture was incubated at room temperature for 30 min with stirring. The unbound ribonucleic acid was eluted with 20 mMTris, and 100 µl 1×RNA-protein binding buffer with a total protein of 100 µg was added to the test tube containing streptavidin magnetic beads. After the streptavidin magnetic beads were incubated for 1.5 h with rotation at 4°C, they were washed 3 times with washing buffer, and then incubated with 50 µl of elution buffer at 37°C for 15 min with stirring. The supernatant was collected for silver staining and KeyGEN mass spectrometry analysis.

Wound healing analysis
Cell migration ability was detected by wound healing assay. A total of 2-4 × 10 5 cells were seeded in 6well plates, cultured for 12-24 h, and transfected with siRNA or control siRNA and PC-DNA3.1-RNF144A-AS1 or control vector. Once the culture reached an 85% fusion rate, the cell layer was scratched with sterile plastic tips, washed with culture medium, and then cultured for 24 and 48 h. At different time points, images of the steel plate were obtained using a microscope (Olympus Tokyo, Japan), and the relative area of the wound was obtained using ImageJ software to quantify and calculate the signi cance of the observed event.

Transwell analysis
The Transwell invasion test was performed in a 24-well plate (Corning, Massachusetts, USA). A Transwell cell with a diameter of 6.5 mm and an 8 µm pore polycarbonate membrane insert (Corning) was used.
The bottom of the upper cavity is coated with bronectin (Merck Millipol, Darmstadt, Germany). After 48 h of transfection, HGC27 cells (2×10 4 ) were seeded in 50 µl Matrigel (Corning) serum-free medium coated or uncoated in the upper cavity. RPMI1640 containing 10% FBS was added to the lower chamber as a chemical attractant. After incubating for 12 h at 37°C, the cells were xed with 4% paraformaldehyde, stained with crystal violet, and counted under a microscope ×200 times. The determination was repeated three times in duplicate. Take the average of the cell counts in 5 random areas.
Immunoprecipitates were collected by centrifugation and analyzed by SDS-PAGE.

RNA-protein immunoprecipitation (RIP) analysis
The MagnaRIP RNA Binding Protein Immunoprecipitation Kit (Merck Microwells) was used according to the manufacturer's instructions. The cell lysate was incubated with microspheres coated with 5 µg anti-Argonaute-2 antibody (AGO2) (Abcam, MA, USA) and anti-PUF60 (Abcam), and the control IgG was rotated overnight at 4°C. Then total RNA was extracted, and the expression of RNF144A-AS1 was detected by qRT-PCR.
Dual-luciferase reporter analysis HEK-293T cells were seeded in a 24-well plate at a density of 6×10 4 cells per well, 24 h before transfection. The luciferase reporter vector (PmirGLO) containing RNF144A-AS1-miR-361-3p/miR-619-5p binding sequence or mutation sequence was co-transfected with miRNA mimic (20 Nm) to detect the binding ability of miRNA. After 24 h, the luciferase activity was measured using the dual-luciferase reporter analysis system (Promega, Madison, WI, USA) according to the manufacturer's protocol.

RNA sequencing(RNA-seq) analysis
Total RNA was extracted with Trizol reagent, and the amount and purity of RNA were veri ed by Nanodrop2000 (Thermo Fisher). RNA integrity and gDNAs contamination were detected by denatured agarose gel electrophoresis.Before the RNA-seq library was constructed, the RNA of each sample was removed by the Ribominus Eukaryotic Cell Kit (QIGEN) for ribosomal RNA.The sequencing library was determined by the Agilent 2100 bioanalyzer using the Agilent DNA 1000 chip kit (Agilent, CA, USA).The library was adjusted to 10 NM before clustering. The cDNA was then sequenced using the HiSeq2000 system (Illumina, Sandiego, CA, USA) and a 100 bp paired-end run.

Animal research
All animal experiments were conducted following the procedures approved by the Institutional Animal Care and Utilization Committee of Jiangsu University. HGC27 cells of RNF144A-AS1 stable silence LV3sh-RNF144A-AS1 and control group LV3-sh-NC cells were collected and suspended in frozen PBS. Ten 5week-old male BALB/cnu/nu mice were randomly divided into 2 groups, and liver metastasis models (1.5 × 10 6 cells in 150 µl of PBS) were prepared by infrasplenic injection. After 3 weeks, bioluminescence imaging was performed to check liver metastases every week for 4 consecutive weeks. Prepare D-luciferin sodium salt stock solution with 15 mg/ml PBS. In order to produce bioluminescence, mice received intraperitoneal injection of uorescein stock solution (150mg/kg). All mice were anesthetized with 2% iso urane immediately, and imaging was performed 10 minutes later. These images were captured using the IVIS Spectrum Xenogen imaging system (Caliper Life Sciences). The mice were sacri ced 7 weeks later, and the liver samples were taken for H&E staining, Western blot, and IHC detection. Blood samples were collected for exosome analysis.

Statistical Analysis
All statistical analyses were performed using SPSS software version 20.0 (SPSS Inc., USA) and GraphPad Prism version 7.00 (GraphPad Software, USA). The differences between the two groups were assessed using Student's t-test. Overall survival curves were estimated by the Kaplan-Meier method, and the difference in survival was evaluated using the log-rank test.p Values < 0.05 were considered statistically signi cant.

Results
High-throughput microarray screening for lncRNAs related to GCLM To identify lncRNAs associated with GCLM and explore their functions and mechanisms, we collected 6 pairs of GC and adjacent normal tissues from patients with GCLM. LncRNA microarray was used for detection, and gene reannotation analysis was performed. In this study, the standard of Fold change (FC) greater than 10 times and p-value < 0.005 was used to screen lncRNA (Fig. 1A, 1B, 1C; Supplementary Table S2). One of the most up-regulated lncRNAs is RNF144A-AS1. RNF144A-AS1 is located on human chromosome 2 and has a 2348bp transcript. It has been found that RNF144A-AS1 is overexpressed in bladder cancer tissues, and promotes cancer cell migration and invasion (49), but there is no report on its expression and signi cance in GC.
Expression and clinical signi cance of RNF144A-AS1 in GC In this study, the expression level of RNF144A-AS1 mRNA in GC tissues was detected by qRT-PCR with expanded samples. The results showed that compared with normal tissues adjacent to cancer, RNF144A-AS1 was highly expressed in GC tissues, and the difference was statistically signi cant ( Fig. 2A) (P < 0.0001). We further evaluated RNF144A-AS1 as a potential molecular marker for GC Taking the normal tissues adjacent to cancer as a control, the ROC curve was made, and the results showed that the truncated value for distinguishing normal tissues from cancer and adjacent normal tissues was 3.493 (relative uorescence intensity), and the sensitivity and speci city were 0.8167 and 0.8917, respectively (Fig. 2B).
In this study, 120 GC tissue specimens were further divided into RNF144A-AS1 high expression group and low expression group according to the expression level of RNF144A-AS1, with 60 pairs in each group. The qRT-PCR method was used to analyze the expression of RNF144A-AS1 in GC tissues and clinical parameters such as age, gender, degree of tissue differentiation, depth of tumor invasion, lymph node  Table S3). Survival analysis of the two groups found that the disease-free survival (DFS) and overall survival (OS) of patients in the RNF144A-AS1 high expression group were lower than those in the RNF144A-AS1 low expression group (DFS, p = 0.002; OS, p < 0.001) (Fig. 2C, 2D). Multivariate analysis found that RNF144A-AS1 expression is one of the independent prognostic factors for GC (Supplementary Table S4). The above results indicate that the expression of RNF144A-AS1 is signi cantly up-regulated in GC, and may be used as one of the indicators for the diagnosis of GCLM.
To determine whether RNF144A-AS1 can be detected in serum exosomes, blood samples from 20 GC patients and 20 healthy controls were collected, and sera-derived exosomes were identi ed by transmission electron microscopy and Western blot analysis ( Fig. 2E, 2F). As expected, RNF144A-AS1 from serum exosomes was more abundant in GC patients than in healthy controls ( Fig. 2G). Furthermore, the level of RNF144A-AS1 in serum exosomes was linearly correlated with the level of RNF144A-AS1 in GC tissue tumors (P = 0.0050), making it possible to detect the expression of RNF144A-AS1 in blood samples. We also detected the expression of exosome RNF144A-AS1 in serum before and after gastrectomy (R0 resection), and found that the expression of exosome RNF144A-AS1 was signi cantly decreased after tumor resection (n = 20) ( Fig. 2H), indicating that GC tissue was the source of exosome RNF144A-AS1.
These results indicate that RNF144A-AS1 is an up-regulated lncRNA derived from GC tissue and can be effectively transported into the circulation by exosomes. In addition, the high expression of RNF144A-AS1 is related to GCLM, high TNM staging, and poor prognosis. It is a potential lncRNA biomarker for GCLM.
In addition, it was found that after transfection of siRNA or overexpression plasmid, the qRT-PCR analysis showed that the expression level of RNF144A-AS1 in exosomes was consistent with the expression change of RNF144A-AS1 in cells (Fig. 3I). We extracted exosomes from HGC27 cell culture medium transfected with RNF144A-AS1 plasmid and co-cultured them with untreated GC HGC27 cells at different concentrations. As expected, the overexpression of exosomal RNF144A-AS1 also enhanced the proliferation, migration, and invasion of GC HGC27 cells, and promoted the phenotype of malignant cells (Fig. 3J, 3K, 3L).
These results indicate that the overexpression of exosomal RNF144A-AS1 plays an important role in promoting the proliferation, migration and invasion of GC cells.

RNF144A-AS1 directly binds to PUF60
The above research suggested that RNF144A-AS1 can act as a miR-361-3p sponge and participate in the invasion and metastasis of GC cells. Are there other mechanisms? We performed RNA pull-down experiments in HGC27 cells. Silver staining results showed that several protein bands were enriched in GC cells in the RNF144A-AS1 group compared with the control group (Fig. 4A). Protein mass spectrometry was used to identify differentially expressed proteins. In the list of recognized proteins, the top one was PUF60. RIP analysis showed that compared with IgG (Fig. 4B), anti-PUF60 antibody down-regulated abundant RNF144A-AS1, con rming the direct interaction between RNF144A-AS1 and PUF60( Figure 4C). The qRT-PCR method detected the expression of PUF60 mRNA in the tissues, and the results showed that PUF60 mRNA was up-regulated in GC tissues compared with normal tissues (Fig. 4D). Considering the interaction between RNF144A-AS1 and protein, we analyzed the relationship between PUF60 mRNA expression and RNF144A-AS1 and found that they are positively correlated (Fig. 4E). We detected the transcriptional expression level of PUF60 in the OE-NC, OE-RNF144A-AS1, sh-NC, and sh-RNF144A-AS1 groups by qRT-PCR. It was found that after overexpression of RNF144A-AS1, the transcription and protein levels of PUF60 increased signi cantly, and after interference with RNF144A-AS1, the transcription and protein levels of PUF60 decreased signi cantly (Fig. 4F,4G). Transwell analysis showed that overexpression of PUF60 can reverse the inhibitory effect of down-regulation of RNF144A-AS1 on cancer cell migration and invasion (Fig. 4H,4I). These results suggest that RNF144A-AS1 directly binds to PUF60, thereby promoting the occurrence and development of GCLM.

RNF144A-AS1 acted as a sponge for miR-361-3p
How does RNF144A-AS1 regulate GCLM? In order to study the mechanism of RNF144A-AS1 in GC cells, we rst used cytoplasmic, nuclear RNA isolation test and FISH to detect the subcellular localization of RNF144A-AS1, and found that RNF144A-AS1 is mainly distributed in the cytoplasm (Fig. 5A, 5B). This shows that the way it regulates downstream target genes is mainly through post-transcriptional level regulation. Given that lncRNAs as miRNA sponges have been extensively studied, and RNF144A-AS1 is abundant in the cytoplasm, we next studied the miRNA binding ability of RNF144A-AS1. We performed RIP on AGO2 in HGC27 cells. The speci c enrichment of dropdown endogenous RNFA114-AS1 was found with qRT-PCR experiments, suggesting that RNF144A-AS1 can play a biological function as a miRNA sponge (Fig. 5C,5D). Three potential miRNAs (miR-361-3p, miR-619-5p and miR-4729) were predicted using online software (Starbase). We detected the expression of these miRNAs in HGC cells overexpressing RNF144A-AS1 and found that miR-361-3p and miR-619-5p were down-regulated (Fig. 5E).
To verify the direct interaction between miR-361-3p, miR-619-5p, and RNF144A-AS1, we constructed an RNF144A-AS1 fragment containing the predicted binding sites (wild type and mutant) of the identi ed miRNA,and insert it into the downstream of the dual-luciferase reporter gene (Fig. 5F). It was found that, compared with miR-NC mimics, miR-361-3p mimics resulted in down-regulation of relative luciferase activity in the RNF144A-AS1-miR-361-3p group, but there was no change in luciferase activity in the RNF144A-AS1-miR-361-3p mutant group. However, miR-619-5p mimics did not signi cantly affect the luciferase activity of RNF144A-AS1-miR-619-5p (Fig. 5G,5H). This study found that the expression level of miR-361-3p in GC tissues was lower than that of matched normal tissues (Fig. 5I), and the expression of miR-361-3p and RNF144A-AS1 was negatively correlated (Fig. 5J). These results indicate that RNF144A-AS2 can act as a sponge for miR-361-3p and reduce the expression of miR-361-3p.
To nd the target genes of miR-361-3p in the GC, we rst used Targetscan, miRDB, Tarbase and miRWalk databases to screen and cross (Fig. 5K), and found 47 mRNAs that can bind miR-361-3p. Further RNA Seq was used to detect the cells in the RNF144A-AS1 overexpression group, and it was found that IGF2BP1 and POU2F2 are important downstream targets of RNF144A-AS1 (Fig. 5L). Next, we detected the expression levels of IGF2BP1 and POU2F2 mRNA in GC tissues and found that the expression of IGF2BP1 and POU2F2 was negatively correlated with miR-361-3p (Fig. 5M, 5N), but positively correlated with RNF144A-AS1 (Fig. 5O,5P) The Western blot and qRT-PCR showed that down-regulation of miRNA-361-3p in HGC27 cells could signi cantly increase the mRNA and protein expression levels of IGF2BP1 and POU2 (Fig. 5Q,5R).
However, up-regulation of miR-361-3p can signi cantly inhibit the mRNA and protein expression levels of IGF2BP1 and POU2F2 ( Fig. 5S,5T), indicating that IGF2BP1 and POU2F2 are the direct target genes of miR-361-3p. It was further found that Western blot showed that down-regulation of RNF144A-AS1 can signi cantly reduce the protein levels of IGF2BP1 and POU2F2 (Fig. 5U), while up-regulation of RNF144A-AS1 can signi cantly increase the protein levels of IGF2BP1 and POU2F2 (Fig. 5V).
To con rm the biological function of miR-361-3p and whether RNF144A-AS1 affects the function of miR-361-3p, we carried out rescue experiments with miR-361-3p mimic alone or with RNF144A-AS1 plasmid.
CCK8 and Transwell analysis found that miR-361-3p mimic reversed the promotion of RNF144A-AS1 overexpression on the migration and invasion of GC cells (Fig. 5W, 5X). The above results suggest that RNF144A-AS1 can act as a sponge for miR-361-3p, and promote the migration and invasion of GC cells by up-regulating the expression of IGF2BP1 and POU2F2.
Regulation of transcription factor HOXA3 on RNF144A-AS1 Transcription factors play an important role in regulating gene expression. Is the abnormal up-regulation of RNF144A-AS1 in GC related to the activation of related transcription factors? To explore whether the activation of related transcription factors can lead to the up-regulation of RNF144A-AS1 expression in GC, we used online software (http://genome.ucsc.edu/) to analyze the transcription factors that might be bound by the RNF144A-AS1 promoter region, and the analysis found that the transcription factor HOXA3 and the gene were bound to the RNF144A-AS1 promoter region ( Fig. 6A). The transcription factor HOXA3 was found to bind with RNF144A-AS1 by luciferase assay (Fig. 6B).
This study used qRT-PCR to detect the level of HOXA3 mRNA and found that the HOXA3 gene was highly expressed in GC tissues, and it was signi cantly positively correlated with the expression level of RNF144A-AS1 in GC tissues (Fig. 6C, 6D). After knocking down the expression of HOXA3 in HGC27 cells with siRNA, RNF144A-AS1 was down-regulated. Further using CCK8, wound healing, and transwell detection, it was found that knocking down HOXA3 expression can reverse the up-regulation of RNF144A-AS1 overexpression on the enhancement of cancer cell proliferation, migration, and invasion ( Fig. 6E-H). This indicates that the high expression of RNF144A-AS1 in GC is likely to be regulated by HOXA3 transcription.
The study of RNF144A-AS1 in regulating GCLM in vivo In order to study the liver metastasis potential of RNF144A-AS1 in vivo, the HGC27 cells of stable silencing LV3-sh-RNF144A-AS1 and LV3-sh-NC cells of the control group were prepared with uorescent enzyme plasmids, and then the above cells were respectively injected into the liver metastasis animal model by subsplenic injection.Liver metastases were detected by in vivo bioluminescence imaging 3 weeks after injection.The results showed that RNF144A-AS1 gene knockout signi cantly reduced the number and size of liver metastases (Fig. 7A,7B). We extracted total RNA from serum exosomes and liver tumor tissues, and detected RNF144A-AS1 by qRT-PCR. The results showed that the expression of exosomal RNF144A-AS1 was linearly related to the expression of RNF144A-AS1 in nude mouse cancer tissues (Fig. 7C). qRT-PCR was used to detect the level of miR-361-3p in tumor tissues, and it was found that it was negatively correlated with the level of RNF144A-AS1 (Fig. 7D). H&E staining and IHC results showed that the staining of IGF2BP1, POU2F2 and HOXA3 in liver metastases in the RNF144A-AS1 gene knockout group was signi cantly reduced (Fig. 7E,7F). This study con rmed that RNF144A-AS1 can be delivered to distal sites through exosomes and detected in the peripheral circulation, suggesting that RNF144A-AS1 may become a promising biomarker for predicting GCLM.

Discussion
With the advent of next-generation sequencing, a large number of lncRNAs have been identi ed from the human genome. Many studies have shown that lncRNA is differentially expressed in tumor tissues and normal tissues, and may play an important role in the occurrence and development of tumors (9,15,20,21).However, the biological function of lncRNAs in GCLM is not very clear. Using high-throughput chip technology, we found that RNF144A-AS1 was signi cantly up-regulated in GC tissues with liver metastases. Compared with healthy people, RNF144A-AS1 was abundant and highly expressed in serum exosomes of patients with GC After gastrectomy, RNF144A-AS1 in serum exosomes decreased, suggesting that GC tissue is the source of RNF144A-AS1 in serum exosomes. In addition, it was found that RNF144A-AS1 overexpression is related to GCLM, TNM staging, and poor prognosis of GC cells. We found that overexpression of RNF144A-AS1 in exosomes can promote the growth and invasion of cocultured GC cells in vitro. The down-regulation of RNF144A-AS1 induced the proliferation, migration, and invasion of GC cells in vitro and inhibit GCLM cells in nude mice,. These results indicate that RNF144A-AS1 is an ideal biomarker for the diagnosis and prognosis of liver metastases from GC.
LncRNAs have been reported to function in multiple steps of gene regulation by sponges for miRNAs (16-18), scaffolds of protein-protein interactions (12,13), decoys of proteins (14,15),and serving as guides of chromatin-modifying complexes (9, 10), etc. Recent advances in studying the ncRNAs, such as lncRNAs and miRNAs, reveal the existence of complex RBP-ncRNA interactions that function in multiple biological processes such as epigenetic, transcriptional and post-transcriptional events (50. In this study,it We found that RNF144A-AS1 interaction with PUF60 protein by RNA pull-down, protein mass spectrometry,western blot,and rescue experiments in GC cells. In addition, RNF144A-AS1 binds to miR-361-3p and reduces the expression of the latter. miR-361-2p has low expression in malignant solid tumors such as prostate cancer (51), ovarian cancer (52), GC (53), non-small cell lung cancer (54,55) and has been con rmed as a tumor suppressor. In this study, it was found that miR-361-3p was low expressed in GC tissues and negatively correlated with RNF144A-AS1 expression. The downstream target genes IGF2BP1 and POU2F2 of miR-361-3p are expressed in the opposite way to miR-361-3p and are highly expressed in GC.
Transcription factors play an important role in regulating gene expression. Is the abnormal up-regulation of RNF144A-AS1 in GC related to the activation of related transcription factors? Online database analysis found that transcription factor HOXA3 and genes are bound to the promoter region of RNF144A-AS1. HOXA3 is a member of the homeobox genes (HOX) family. HOX genes are a class of genes that are highly conserved in evolution and can regulate the normal development of the body. Such as participating in the formation and remodeling of the aortic arch (56) and controlling the early embryonic development of the pharyngeal organs (57). Recent studies have found that HOXA3 is highly expressed in breast cancer (58), renal clear cell carcinoma (59) and T-cell lymphoblastic lymphoma (60), and is related to cancer cell proliferation and invasion. This study shows that HOXA3 is highly expressed in GC tissues and cells, and is positively correlated with RNF144A-AS1. It was also found that knocking down the expression of HOXA3 can reverse the up-regulation of RNF144A-AS1 to promote the proliferation, migration, and invasion of GC cells. This indicates that the high expression of RNF144A-AS1 in GC is likely to be regulated by HOXA3 transcription.

Conclusions
In summary, we proved that RNF144A-AS1 is up-regulated in GC patients and is related to GCLM, late TNM staging, and poor prognosis. RNF144A-AS1 can up-regulate IGF2BP1 and POU2F2 through sponge miR-361-3p, and promote the occurrence of liver development and metastasis of GC by interacting with PUF60. The overexpression of RNF144A-AS1 is related to the activation of the cancer-promoting transcription factor HOXA3 (Fig. 8). Serum exosomal liquid biopsy for RNF144A-AS1 is helpful for the diagnosis and prognosis prediction of GCLM. Therefore, RNF144A-AS1 may be a promising biomarker for the diagnosis and prognosis of GCLM, and one of the potential targets for the treatment of GCLM.

Consent for publication
The authors declare that they agree to submit the article for publication.

Availability of data and materials
All data and materials supporting the ndings of this work are available from its supplementary information les and from the corresponding author upon reasonable request.