Long noncoding RNA LINC01578 drives colon cancer metastasis through a positive feedback loop with the NF‐κB/YY1 axis

LINC01578 was upregulated and correlated with metastasis and prognosis in colon cancer. Nuclear factor kappa B (NF‐κB) and Yin Yang 1 (YY1) bound to LINC01578 promoter to activate LINC01578 expression. LINC01578 interacted with and recruited EZH2 to NFKBIB promoter to repress NFKBIB expression, thereby activating NF‐κB signaling. LINC01578 upregulated YY1 via activating NF‐κB. LINC01578 enhanced colon cancer liver metastasis through forming a positive feedback loop with NF‐κB/YY1.

Metastasis accounts for poor prognosis of cancers and related deaths. Accumulating evidence has shown that long noncoding RNAs (lncRNAs) play critical roles in several types of cancer. However, which lncRNAs contribute to metastasis of colon cancer is still largely unknown. In this study, we found that lncRNA LINC01578 was correlated with metastasis and poor prognosis of colon cancer. LINC01578 was upregulated in colon cancer, associated with metastasis, advanced clinical stages, poor overall survival, disease-specific survival, and disease-free survival. Gain-of-function and loss-of-function assays revealed that LINC01578 enhanced colon cancer cell viability and mobility in vitro and colon cancer liver metastasis in vivo. Mechanistically, nuclear factor kappa B (NF-jB) and Yin Yang 1 (YY1) directly bound to the LINC01578 promoter, enhanced its activity, and activated LINC01578 expression. LINC01578 was shown to be a chromatin-bound lncRNA, which directly bound NFKBIB promoter. Furthermore, LINC01578 interacted with and recruited EZH2 to NFKBIB promoter and further repressed NFKBIB expression, thereby activating NF-jB signaling. Through activation of NF-jB, LINC01578 further upregulated YY1 expression. Through activation of the NF-jB/YY1 axis, LINC01578 in turn enhanced its own promoter activity, suggesting that LINC01578 and NF-jB/YY1 formed a positive feedback loop. Blocking NF-jB signaling abolished the oncogenic roles of LINC01578 in colon cancer. Furthermore, the expression levels of LINC01578, NFKBIB, and YY1 were correlated in clinical tissues. Collectively, this study demonstrated that LINC01578 promoted colon cancer metastasis via forming a positive feedback loop with NF-jB/YY1 and suggested that LINC01578 represents a potential prognostic biomarker and therapeutic target for colon cancer metastasis.

Introduction
Colon cancer is the fourth most common malignancies and the fifth leading cause of cancer-related death worldwide [1]. A total of 1 096 601 newly diagnosed colon cancer cases and 551 269 deaths were estimated to occur in 2018 globally [1]. Colon cancer is the fourth most common malignancies and the second leading cause of cancer-related death in the United States with 104 610 newly diagnosed colon cancer cases and 53 200 deaths estimated to occur in 2020 [2]. Therefore, colon cancer is still an important healthy challenge for humans. Due to the ineffectiveness of early diagnosis,~15-25% of CRC patients are diagnosed with metastasis [3]. Metastasis is the leading cause of cancer-related death [4]. Although primary colon cancer could be effectively resected, currently therapeutic methods for colon cancer metastasis are still far from satisfactory [5]. Therefore, better understanding of the molecular mechanisms underlying colon cancer metastasis is imperative for the development of more effective therapeutic strategies against colon cancer metastasis.
Given that metastasis is the leading cause of colon cancer-related death, we hypothesized that the lncRNAs involved in colon cancer metastasis should also be associated with colon cancer patients' survival. Therefore, we first searched the lncRNAs associated with colon cancer patients' survival using The Cancer Genome Atlas (TCGA) colon adenocarcinoma (COAD) data. Through the analysis of the TCGA COAD data, we found lncRNA LINC01578 was associated with poor overall survival of colon cancer patients. LINC01578, named as CHASERR in a previous report, is a conserved lncRNA locating at chromatin 15q26.1 and upstream of CHD2 [32]. Mouse Linc01578 is reported to be essential for cell viability in neurological disease [32]. Here, we further studied the expression pattern, biological roles, and molecular mechanisms of LINC01578 in human colon cancer metastasis.

Clinical specimens
Seventy pairs of colon cancer and matched normal tissues were acquired at Tianjin Medical University Cancer Institute and Hospital from colon cancer patients who received surgical resection. The inclusion criteria included (a) histopathologically confirmed colon cancer; and (b) curative surgical resection. The exclusion criteria included (a) presurgical chemotherapy or radiotherapy; and (b) history of other cancers. The clinicopathological features of these 70 colon cancer patients are presented in Table 1. Among these 70 colon cancer patients, 10 have liver metastases and the liver metastatic lesions were resected and collected at Tianjin Medical University Cancer Institute and Hospital. All of the patients were followed up on a regular basis and lasted for a median of 57 months. Disease-specific survival (DSS) time was defined as the date of surgery to the date of colon cancer-caused death or the date of the last follow-up. Disease-free survival (DFS) time was defined as the date of surgery to the date of colon cancer recurrence or the date of the last follow-up. This study conformed to the standards set by the Declaration of Helsinki, and was reviewed and approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital. Written informed consents were acquired from all participants.

Generation of stable cell lines
To construct colon cancer cells with LINC01578 stable overexpression, LINC01578 overexpression plasmid pcDNA3.1-LINC01578 was transfected into HCT116 and DLD-1 cells. Forty-eight hours after transfection, HCT116 and DLD-1 cells were treated with neomycin (1000 µgÁmL À1 ) for 4 weeks to select LINC01578-overexpressed cells. To construct LINC01578 stably silenced colon cancer cells, LoVo and HT-29 cells were infected with LINC01578 shRNAs lentivirus. Ninetysix hours after infection, the cells were treated with puromycin (2 µgÁmL À1 ) for 4 weeks to select LINC01578-depleted cells. The overexpression and depletion efficiencies were verified using qRT-PCR as below described.

Luciferase reporter assay
To measure LINC01578 promoter activity, LINC01578 promoter firefly luciferase reporter plasmid pGL3-LINC01578 was cotransfected with pRL-TK into indicated colon cancer cells. pRL-TK, which encodes Renilla luciferase, was used as endogenous control. Forty-eight hours after transfection, the firefly and Renilla luciferase activities were detected by the Dual-Luciferase Reporter Assay System (Promega).
LINC01578 promoter activity was calculated as the ratio of firefly luciferase activity to Renilla luciferase activity. To measure NF-jB transcriptional activity, firefly luciferase reporter plasmid pNFjB-luc was cotransfected with pRL-TK into indicated colon cancer cells. Forty-eight hours after transfection, the firefly and Renilla luciferase activities were detected by the Dual-Luciferase Reporter Assay System (Promega). NF-jB transcriptional activity was calculated as the ratio of firefly luciferase activity to Renilla luciferase activity.

NF-jB activation measurement
The nuclear extracts from indicated cells were used to measure NF-jB transcription factor DNA binding activities using the NF-jB p50 Transcription Factor Assay Kit (Cat. ab207217; Abcam) and NF-jB p65 Transcription Factor Assay Kit (Cat. ab133112; Abcam) following the manufacturer's protocols.

Subcellular fractionation and polyadenylated RNA enrichment
Subcellular fractionation was conducted as previously described [34]. The RNA in different subcellular components was extracted and detected using qRT-PCR as above described. Polyadenylated RNA was captured and purified using oligo(dT) polystyrene beads (GenElute mRNA Miniprep Kit; Sigma-Aldrich, St Louis, MO, USA). The enriched polyadenylated RNA was detected using qRT-PCR as above described.

RNA pull-down
RNA pull-down assays were conducted as previously described [34]. Briefly, LINC01578 and antisense LINC01578 were in vitro-transcribed and biotin-labeled from pSPT19-LINC01578 by the Biotin RNA Labeling Mix (Roche) and T7 and Sp6 RNA polymerase (Roche), respectively. After being purified, 3 µg of in vitro-transcribed LINC01578 was incubated with 1 mg of whole-cell lysate of DLD-1 cells at 25°C for 1 h. Next, the complexes were enriched by streptavidin agarose beads (Invitrogen). The proteins present in the pull-down material were measured using western blot.

RNA immunoprecipitation (RIP)
RIP assay was conducted by the Magna RIP TM RNA-Binding Protein Immunoprecipitation Kit (Millipore) and primary antibody against EZH2 (Cat. ab195409; Abcam). The enriched RNA was measured using qRT-PCR.

Cell viability and proliferation assays
Glo cell viability assay was conducted to measure cell viability. 5-Ethynyl-2'-deoxyuridine (EdU) assay was conducted to measure cell proliferation. Glo cell viability and EdU assays were carried out as previously described [35]. Briefly, Glo cell viability assay was conducted with the CellTiter-GloÒ Luminescent Cell Viability Assay (Cat. G7570; Promega). EdU assay was conducted with the EdU Kit (Cat. C00031; RiboBio, Guangzhou, China). The percentage of EdUpositive cell was calculated using a fluorescence microscope based on at least five random fields.

Cell migration and invasion assays
Cell migration and invasion abilities were detected by transwell migration and invasion assays as previously described [27]. Briefly, 50 000 indicated cells were seeded into the upper chamber of 24-well transwell insert precoated with (invasion) or without (migration) Matrigel. After culture for 48 h, the cells remaining in the upper chamber were removed. The migrated or invasive cells were fixed, stained, and counted using a microscope based on at least five random fields.

Mice studies
Five-week-old male BALB/c athymic nude mice were used for animal studies. For liver metastasis experiments, 3 9 10 6 indicated cells were injected into the spleen of nude mice. The mice were allowed to grow for another 6 weeks. Next, the livers were resected and liver metastases were detected by hematoxylin and eosin (H&E) staining. The liver metastatic lesions were further used to perform immunohistochemistry (IHC) staining as previously described [36] with primary antibodies against Ki67 (Cat. ab16667; Abcam), p50 (Cat. ab32360; Abcam), or YY1 (Cat. ab109228; Abcam). In addition, the liver metastatic lesions were also used to perform TdT-mediated dUTP Nick-End Labeling (TUNEL) staining using the One Step TUNEL Apoptosis Assay Kit (Cat. C1086; Beyotime). Mice studies conformed to the animal care guidelines and were approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital.

Statistical analysis
All statistical analyses were conducted by the GRAPH-PAD PRISM 6.0 Software, (GraphPad, La Jolla, CA, USA). The detailed statistical methods were described in the figure and table legends. P < 0.05 was recognized as statistically significant.

LINC01578 was correlated with metastasis and poor prognosis of colon cancer
To search the lncRNAs associated with outcome of colon cancer patients, we analyzed TCGA COAD data using the GEPIA. The most differentially survival genes are shown in Table S1. Among the top differentially survival genes, we noted lncRNA LINC01578 whose high expression was correlated with remarkably worse overall survival of colon cancer patients with a hazard ratio (HR) value of 2.9 (Fig. S1A). In addition, the GSE17538 dataset that contains 226 colon cancer tissues with prognostic information showed that increased expression of LINC01578 was correlated with worse DSS (Fig. 1A-C). To further elucidate the expression pattern and clinical relevance of LINC01578 in colon cancer, we collected 70 pairs of colon cancer and matched normal tissues. The expression of LINC01578 in these tissues was detected by qRT-PCR, and the results revealed that LINC01578 was upregulated in colon cancer tissues with respect to normal tissues (Fig. 1D). Correlation analysis between LINC01578 expression and clinicopathological characteristics revealed that high expression of LINC01578 was positively correlated with lymph node metastasis, distant metastasis, and advanced TNM stages, but not correlated with gender, age, grade, and microsatellite  expression in 60 colon cancer tissues without metastasis, 10 colon cancer tissues with metastasis, and 10 matched liver metastatic tissues was measured by qRT-PCR. **P < 0.01, ***P < 0.001, ns, not significant, by the Kruskal-Wallis test followed by Dunn's multiple comparisons test. instability (MSI) status (Table 1). The Kaplan-Meier analysis revealed that high expression of LINC01578 was correlated with worse DSS (Fig. 1E,F) and DFS (recurrence) (Fig. 1G,H). Furthermore, univariate and multivariate survival analyses revealed that high LINC01578 expression was independently correlated with poor DSS and DFS (Table S2). To elucidate the expression pattern of LINC01578 in colon cancer metastatic processes, we detected the expression of LINC01578 in colon cancer tissues without metastasis, colon cancer tissues with metastasis, and liver metastatic tissues using qRT-PCR. The results revealed that LINC01578 was remarkably increased in colon cancer tissues with metastasis with respect to colon cancer tissues without metastasis (Fig. 1I). The expression of LINC01578 in liver metastatic tissues was not significantly different from that in matched primary colon cancer tissues (Fig. 1I). GSE37892 dataset that contains 93 colon cancer without metastasis and 37 colon cancer with metastasis presented that LINC01578 was also remarkably increased in colon cancer tissues with metastasis with respect to colon cancer tissues without metastasis (Fig. S1B). In addition, the GSE37892 dataset also showed that LINC01578 was correlated with advanced TNM stages, but not correlated with gender, age, and tumor localization (Fig. S1C-F). The expression of LINC01578 in colon cancer cell lines was measured by qRT-PCR, and the results revealed that LINC01578 was increased in all colon cancer cell lines compared with a normal colon mucosa epithelial cell line NCM460 (Fig. S1G). Collectively, these results demonstrated that increased expression of LINC01578 was associated with metastasis and poor prognosis of colon cancer patients.
LINC01578 has three different isoforms annotated by the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/) (Fig. S1H). qRT-PCR results showed that the isoform NR_037601 was the major transcript in colon cancer cell lines and tissues (Fig. S1I,J). Therefore, in this study we focused on NR_037601. NR_037601 has six exons and 1787 nucleotides in length. Three online in silico tools including the CPAT, the CPC, and PhyloCSF all indicate that LINC01578 was a noncoding RNA, with CPAT score of 0.0067, CPC score of À0.9136, and PhyloCSF score of below zero. LINC01578 had a poly (A) tail (Fig. S1K).

LINC01578 enhanced colon cancer cell viability and mobility in vitro
Considering the correlation between LINC01578 and metastasis and poor prognosis in colon cancer, we further studied whether LINC01578 contributes to colon cancer progression. LINC01578 overexpression plasmid pcDNA3.1-LINC01578 was transfected into colon cancer cells DLD-1 and HCT116 to sufficiently increase LINC01578 expression (Fig. 3A,B). Glo cell viability assays showed that cell viabilities of DLD-1 and HCT116 cells increased significantly after   LINC01578 overexpression (Fig. 3C,D). EdU assays showed that LINC01578 overexpression led to significant acceleration in cell proliferation (Fig. 3E). Transwell migration assays presented that cell migration increased remarkably after LINC01578 overexpression (Fig. 3F). Transwell invasion assays presented that cell invasion increased remarkably after LINC01578 overexpression (Fig. 3G). To preclude the possible clonal effect, another LINC01578 overexpressing clone of DLD-1 cells was constructed (Fig. S3A). Glo cell viability and EdU assays showed that cell viability and proliferation consistently increased after LINC01578 overexpression (Fig. S3B,C). Transwell migration and invasion assays showed that cell migration and invasion consistently increased after LINC01578 overexpression (Fig. S3D,E).  In addition, two independent short hairpin RNAs (shRNAs) targeting to LINC01578 were infected into colon cancer cells LoVo via lentivirus to sufficiently inhibit LINC01578 expression (Fig. 4A). Glo cell viability assays showed that cell viabilities of LoVo cells reduced significantly after LINC01578 inhibition (Fig. 4B). EdU assays showed that LINC01578 inhibition led to significant repression in cell proliferation (Fig. 4C). Transwell migration assays presented that cell migration reduced remarkably after LINC01578 inhibition (Fig. 4D). Transwell invasion assays presented that cell invasion reduced remarkably after LINC01578 inhibition (Fig. 4E). LINC01578-silenced another colon cancer cell HT-29 was constructed via shRNA lentivirus (Fig. S4A). Glo cell viability and EdU assays showed that cell viability and proliferation of HT-29 cells were reduced after LINC01578 inhibition (Fig. S4B,C). Transwell migration and invasion assays showed that cell migration and invasion of HT-29 cells were reduced after LINC01578 inhibition (Fig. S4D,E). Collectively, these findings demonstrated that overexpression of LINC01578 enhanced, but depletion of LINC01578 repressed, colon cancer cell viability and mobility in vitro.

LINC01578 drove colon cancer metastasis in vivo
Given the correlation between LINC01578 and metastasis and poor prognosis in colon cancer, we investigated the roles of LINC01578 in colon cancer metastasis in vivo. DLD-1 cells with LINC01578 overexpression or control were injected into spleen of nude mice to construct liver metastasis model. Six weeks after injection, liver metastases were measured by H&E staining. As shown in Fig. 5A-C, DLD-1 cells formed more and larger liver metastases after LINC01578 overexpression. LINC01578-depleted and control LoVo cells were also injected into spleen of nude mice. Six weeks after injection, liver metastases were evaluated by H&E staining. As shown in Fig. 5D-F, LoVo cells formed less and smaller liver metastases after LINC01578 depletion. Proliferation marker Ki67 IHC staining presented that the liver metastatic tumors formed by DLD-1 cells with LINC01578 overexpression had more Ki67-positive cells with respect to the liver metastatic tumors formed by control DLD-1 cells (Fig. 5G). Conversely, the liver metastatic tumors formed by LINC01578-depleted LoVo cells had less Ki67-positive cells with respect to the liver metastatic tumors formed by control LoVo cells (Fig. 5H). Apoptosis marker TUNEL staining showed that the liver metastatic tumors formed by DLD-1 cells with LINC01578 overexpression had less TUNEL-positive cells with respect to the liver metastatic tumors formed by control DLD-1 cells (Fig. 5I). Conversely, the liver metastatic tumors formed by LINC01578-depleted LoVo cells had more TUNELpositive cells with respect to the liver metastatic tumors formed by control LoVo cells (Fig. 5J). Taken together, these findings demonstrated that LINC01578 drove colon cancer liver metastasis, promoted colon cancer cell proliferation, and repressed colon cancer cell apoptosis in vivo.

LINC01578 activated NF-jB
Considering NF-jB could activate LINC01578, we next studied whether LINC01578 was implicated in NF-jB signaling in colon cancer. p50 nuclear expression was measured to determine NF-jB activation in liver metastatic tumors formed by LINC01578-overexpressed DLD-1 and LINC01578-depleted LoVo cells. As presented in Fig. 6A,B, the liver metastatic tumors formed by LINC01578-overexpressed DLD-1 cells had higher p50 nuclear expression with respect to the liver metastatic tumors formed by control DLD-1 cells (Fig. 6A). Conversely, the liver metastatic tumors formed by LINC01578-depleted LoVo cells had lower p50 nuclear expression with respect to the liver metastatic tumors formed by control LoVo cells (Fig. 6B). Next, NF-jB nuclear expression in LINC01578-overexpressed DLD-1 and LINC01578-depleted LoVo cells was measured by western blot. LINC01578-overexpressed DLD-1 cells had significantly higher p50 and p65 nuclear expression with respect to control DLD-1 cells (Fig. 6C). Conversely, LINC01578-depleted LoVo cells had significantly lower p50 and p65 nuclear expression with respect to control LoVo cells (Fig. 6D). To further investigate whether LINC01578 modulated NF-jB, NF-jB transcriptional activity was measured using luciferase reporter assays. As shown in Fig. 6E,F, NF-jB transcriptional activity increased after LINC01578 overexpression and reduced after LINC01578 depletion. NF-jB activation in nuclear extracts was measured by NF-jB Transcription Factor Assay. As shown in Fig. 6G,H, p50 and p65 activation in nuclear extracts increased after LINC01578 overexpression and reduced after LINC01578 depletion. Therefore, these findings demonstrated that LINC01578 activated NF-jB in colon cancer.
To elucidate whether the activation of NF-jB mediated the oncogenic roles of LINC01578 in colon cancer, DLD-1 cells with LINC01578 overexpression or control were treated with NF-jB inhibitor JSH-23. Glo cell viability assays presented that JSH-23 abolished the roles of LINC01578 overexpression in increasing cell viability (Fig. 7A). EdU assays showed that JSH-23 abolished the roles of LINC01578 overexpression in accelerating cell proliferation (Fig. 7B). Transwell migration assays presented that JSH-23 abolished the roles of LINC01578 overexpression in promoting cell migration (Fig. 7C). Transwell invasion assays showed that JSH-23 abolished the roles of LINC01578 overexpression in promoting cell invasion (Fig. 7D). Collectively, the data demonstrated that LINC01578 enhanced colon cancer cell viability and mobility via activating NF-jB.

LINC01578 epigenetically repressed IjBb via recruiting EZH2
To investigate the mechanisms mediating the activation of NF-jB by LINC01578, LINC01578-overexpressed and control DLD-1 cells were treated with different NF-jB inhibitors, and then, NF-jB activation in nuclear extracts was measured by NF-kB Transcription Factor Assay. NF-jB nuclear translocation inhibitor JSH-23 abolished the increased NF-jB activation caused by LINC01578 (Fig. S5A). IKK inhibitor Bay 11-7085 did not influence the increased NF-jB activation caused by LINC01578 (Fig. S5B). Therefore, the targets of LINC01578 were locating at the downstream of IKK and upstream of p50/p65. Next, we investigated the effects of LINC01578 on IjBa and IjBb. As shown in Fig. 8A-D, IjBb expression decreased after LINC01578 overexpression and increased after LINC01578 depletion. Furthermore, the liver metastatic tumors formed by LINC01578overexpressed DLD-1 cells had lower IjBb expression with respect to the liver metastatic tumors formed by control DLD-1 cells (Fig. S5C). The liver metastatic tumors formed by LINC01578-depleted LoVo cells had increased IjBb expression with respect to the liver metastatic tumors formed by control LoVo cells (Fig. S5D). The data further supported the repressive roles of LINC01578 on IjBb.
To study the mechanisms mediating the repressive roles of LINC01578 on IjBb, we determined the subcellular localization of LINC01578. As shown in Fig. 8E, LINC01578 mainly localizes at chromatin. Next, we predicted a putative LINC01578 complementary binding locus in NFKBIB (IjBb encoding gene) promoter (Fig. 8F). ChIRP assay presented the specific binding of LINC01578 to NFKBIB promoter, but not a distant NC region (Fig. 8G). To elucidate whether the binding between LINC01578 and NFKBIB promoter modulated chromatin structure, chromatin marks at NFKBIB promoter region were investigated using ChIP assay. As shown in Fig. 8H, repressive chromatin mark H3K27me3 significantly increased after LINC01578 overexpression,  which was abolished by the mutation of the binding region. Furthermore, H3K27me3 significantly reduced after LINC01578 depletion (Fig. 8I). H3K27me3 was mainly catalyzed by histone methyltransferase EZH2. Intriguingly, the online in silico tool RPISeq (http:// pridb.gdcb.iastate.edu/RPISeq/) predicted a potential binding between LINC01578 and EZH2 with a score of 0.96. RNA pull-down assay presented specific enrichment of EZH2 in the LINC01578 group (Figs 8J  and S5E). RIP assays showed specific enrichment of LINC01578 in the EZH2 antibody group (Fig. 8K). EZH2-interacted lncRNA HEIH was used as positive control [40]. Thus, the RNA pull-down and RIP assays suggested the specific binding between  LINC01578 and EZH2. ChIP assays further demonstrated that the binding of EZH2 to NFKBIB promoter increased after LINC01578 overexpression and reduced after LINC01578 depletion, while EZH2 levels were not changed by LINC01578 (Fig. 8L,M). Collectively, these results demonstrated that LINC01578 binds and recruits EZH2 to NFKBIB promoter, and therefore epigenetically represses LINC01578 expression.

LINC01578 formed a positive feedback loop with NF-jB/YY1
Yin Yang 1 was frequently reported as downstream target of NF-jB in several different cell models [39,41,42]. In the above-established liver metastasis models, we found that the liver metastatic tumors formed by LINC01578-overexpressed DLD-1 cells had higher YY1 expression with respect to the liver metastatic tumors formed by control DLD-1 cells (Fig. S5F). The liver metastatic tumors formed by LINC01578-depleted LoVo cells had lower YY1 expression with respect to the liver metastatic tumors formed by control LoVo cells (Fig. S5G). To further confirm whether LINC01578 regulates YY1 through NF-jB in colon cancer, YY1 expression in LINC01578-overexpressed DLD-1 and LINC01578-depleted LoVo cells was detected by qRT-PCR and western blot. As presented in Fig. 9A,B, YY1 expression increased after LINC01578 overexpression and reduced after LINC01578 depletion. Furthermore, both the increase in YY1 caused by LINC01578 overexpression and reduction in YY1 caused by LINC01578 depletion were abolished by JSH-23 (Fig. 9C,D). These results suggested that LINC01578 upregulated YY1 through activation of NF-jB. Given that NF-jB and YY1 directly upregulated LINC01578, LINC01578 and NF-jB/YY1 may form positive feedback loop in colon cancer (Fig. 9E). To verify whether this feedback loop existed, pGL3 luciferase reporter plasmid assay was conducted to detect the effects of LINC01578 on LINC01578 promoter activity. As presented in Fig. 9F,G, LINC01578 promoter activity increased after LINC01578 overexpression and reduced after LINC01578 depletion. Furthermore, the binding of NF-jB on LINC01578 promoter increased after LINC01578 overexpression and reduced after LINC01578 depletion (Fig. 9H,I). Consistently, the binding of YY1 on LINC01578 promoter increased after LINC01578 overexpression and reduced after LINC01578 depletion (Fig. 9J,K). Therefore, these findings demonstrated a positive feedback loop between LINC01578 and NF-jB/YY1.
To investigate whether this positive feedback loop also exists in colon cancer metastasis in vivo, we measured the expression and correlation between LINC01578, NFKBIB, and YY1 in clinical tissues. In contrast to LINC01578, NFKBIB was remarkably downregulated in colon cancer tissues with metastasis with respect to colon cancer tissues without metastasis (Fig. 10A). The expression of NFKBIB was remarkably inversely correlated with that of LINC01578 in colon cancer tissues (Fig. 10B). The inverse correlation between NFKBIB and LINC01578 was also found in GSE17538 dataset (Fig. 10C). Consistent with LINC01578, YY1 was also increased in colon cancer tissues with metastasis with respect to colon cancer tissues without metastasis (Fig. 10D). The expression of YY1 was positively correlated with that of LINC01578 in colon cancer tissues (Fig. 10E). The positive correlation between YY1 and LINC01578 was also found in GSE17538 dataset (Fig. 10F). These findings supported the positive feedback loop between LINC01578 and NF-jB/YY1 in human colon cancer metastasis.

Discussion
Metastasis accounts for the majority of cancer-related death [43,44]. Identifying the aberrantly expressed molecules and critical molecular events during metastasis would provide prognostic biomarker and therapeutic target for metastasis [45]. In this study, using publicly available dataset including TCGA and GEO data, and our own cohort collected in our hospital, we identified lncRNA LINC01578 as a metastasis and prognosis correlated with lncRNA in colon cancer. Our findings identified that increased expression of LINC01578 was correlated with worse overall survival, DSS, and DFS. Further investigation on multiple central cohorts is needed to establish the optimal cutoff value of LINC01578 for potential clinical application to predict the outcomes of colon cancer patients. Furthermore, our findings revealed that LINC01578 was increased in colon cancer tissues with respect to normal tissues. Increased expression of LINC01578 was positively associated with lymph node metastasis, distant metastasis, and advanced clinical stage. The expression of LINC01578 was increased in colon cancer tissues with metastasis with respect to colon cancer tissues without metastasis. However, LINC01578 was not differentially expressed between primary colon cancer tissues and matched liver metastatic tissues. Therefore, our findings suggest that different LINC01578 expressions in primary colon cancer tissues may determine different metastatic potential of corresponding colon cancer.
Gain-of-function and loss-of-function assays demonstrated that ectopic expression of LINC01578 enhanced colon cancer cell viability and mobility in vitro and colon cancer liver metastasis in vivo. Depletion of LINC01578 inhibited colon cancer cell viability and mobility in vitro and colon cancer liver metastasis in vivo. Further analysis of the liver metastatic lesions revealed that LINC01578 promoted colon cancer cell proliferation and repressed colon cancer cell apoptosis in vivo. Therefore, our findings suggested that LINC01578 drove colon cancer liver metastasis via enhancing colon cancer cell viability and mobility. Our findings also suggested LINC01578 as a potential therapeutic target for colon cancer liver metastasis. IjBb protein level in LoVo cells infected with shRNAs targeted to LINC01578 was measured by western blot. (E) Subcellular localization of LINC01578 was detected by qRT-PCR in biochemically fractionated DLD-1 cells. GAPDH, RNU1-1, and NEAT1 were used as cytoplasmic, nucleoplasmic, and chromatinic controls, respectively. (F) The predicted interaction between LINC01578 and NFKBIB promoter. (G) ChIRP assays using LINC01578 capture probes were carried out in DLD-1 cells. The enrichment of NFKBIB promoter and a distant NC region was determined by qPCR. (H) ChIP assays using H3K27me3 and H3K27ac antibodies were carried out in DLD-1 cells overexpressing wide-type or binding region-mutated LINC01578. The enrichment of NFKBIB promoter was determined by qPCR. (I) ChIP assays using H3K27me3 and H3K27ac antibodies were carried out in LINC01578-depleted and control LoVo cells. The enrichment of NFKBIB promoter was determined by qPCR. (J) RNA pull-down assays were performed using in vitro-transcribed biotinylated LINC01578. The enriched proteins were detected by western blot. GAPDH was used as NC. (K) RIP assays using EZH2 antibody were carried out in DLD-1 cells. The enrichment of LINC01578 was determined by qRT-PCR. GAPDH and HEIH were used as negative and positive controls, respectively. (L) ChIP assays using EZH2 antibody were carried out in LINC01578-overexpressed and control DLD-1 cells. The enrichment of NFKBIB promoter and a distant NC region was determined by qPCR. EZH2 protein level in LINC01578overexpressed and control DLD-1 cells was determined by western blot. (M) ChIP assays using EZH2 antibody were carried out in LINC01578-depleted and control LoVo cells. The enrichment of NFKBIB promoter and a distant NC region was determined by qPCR. EZH2 protein level in LINC01578-depleted and control LoVo cells was determined by western blot. Data are shown as mean AE SD based on three independent experiments. **P < 0.01, ***P < 0.001, ****P < 0.0001, ns, not significant, by Student's t-test (A,K,L), or one-way ANOVA followed by Dunnett's multiple comparisons test (B,G,H,I,M). The mechanisms exerted by lncRNAs are complicated [46][47][48]. Different subcellular locations of lncRNAs influence their mechanisms of action.
Cytoplasmic lncRNAs interact and sponge micro-RNAs, and relieve the repressive roles of microRNAs on their targets [49]. Cytoplasmic lncRNAs also bind   Luciferase reporter assays for LINC01578-depleted and control LoVo cells transfected with luciferase reporter plasmids containing LINC01578 promoter. (H) ChIP assays using p50 and p65 antibodies were carried out in LINC01578-overexpressed and control DLD-1 cells. The enrichment of LINC01578 promoter was determined by qPCR. (I) ChIP assays using p50 and p65 antibodies were carried out in LINC01578depleted and control LoVo cells. The enrichment of LINC01578 promoter was determined by qPCR. (J) ChIP assays using YY1 antibody were carried out in LINC01578-overexpressed and control DLD-1 cells. The enrichment of LINC01578 promoter was determined by qPCR. (K) ChIP assays using YY1 antibody were carried out in LINC01578-depleted and control LoVo cells. The enrichment of LINC01578 promoter was determined by qPCR. Data are shown as mean AE SD based on three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant, by Student's t-test (A,C,F,H,J), or one-way ANOVA followed by Dunnett's multiple comparisons test (B,D,G,I,K). mRNAs and change the stability and/or translation of interacted mRNAs [50]. Nuclear lncRNAs may regulate the transcription of target genes via binding epigenetic modification enzymes or transcription factors [51]. Chromatin-bound lncRNAs may directly change the epigenetic modification and chromatin structure, and further modulate target gene expression [52]. Another important common mechanism of lncRNAs is binding proteins [53,54]. The interaction between lncRNAs and proteins may regulate the post-translational modification, location, expression, and/or role of the bound proteins.
Here, we identified LINC01578 as a chromatinbound lncRNA, and further found that LINC01578 specifically binds NFKBIB promoter. NFKBIB encodes IjBb, an important inhibitor of NF-jB. Combining bioinformatic analyses and experimental investigation, we further revealed that LINC01578 bound and recruited EZH2 to NFKBIB promoter, upregulated H3K27me3 level in NFKBIB promoter, and therefore repressed NFKBIB expression. Via repressing NFKBIB expression, LINC01578 activated NF-jB signaling. NF-jB signaling was frequently reported to play critical roles in cancer cell viability and tumor growth and metastasis [55,56]. Here, we further found that blocking NF-jB signaling abolished the oncogenic roles of LINC01578 in colon cancer, suggesting that NF-jB signaling activation was an important mediator of the oncogenic roles of LINC01578 in colon cancer. NF-jB family contains five different transcription factors, including p65, p50/p105, p52/p100, RelB, and c-Rel. Here, we focused on the canonical p65 and p50. The effects of LINC01578 on noncanonical NF-jB signaling need further study. The identification of LINC01578 as a chromatin-bound lncRNA provided novel evidence for the functional variety of lncRNAs. lncRNAs NEAT1 and MALAT1 were revealed to bind active chromatin sites [52]. But the effects of NEAT1 and MALAT1 on the binding sites are not clear [52]. lncRNA DILC was shown to bind IL-6 promoter and inhibited IL-6 transcription [57]. lncRNA RBAT1 bound to E2F3 promoter and activated E2F3 transcription [58]. These reports demonstrated that the binding to chromatin by lncRNAs may further cause different effects on their bound targets.
Yin Yang 1 is a multifunctional transcription factor, which could positively or negatively regulate target gene expression [59,60]. YY1 was also reported to be a downstream target of NF-jB in myoblasts and pancreatic cancer cells [39,41]. Here, we further found that via activating NF-jB, LINC01578 upregulated YY1 expression, which was abolished by NF-jB inhibitor. Intriguingly, LINC01578 was also identified as a direct transcriptional target of NF-jB and YY1. Our findings revealed that both NF-jB and YY1 activated LINC01578 promoter activity and expression. Therefore, LINC01578 formed positive feedback loop with NF-jB/YY1. Through this positive feedback loop, LINC01578 activated its own promoter activity. The correlation between LINC01578 and NF-jB/YY1 was also verified in clinical colon cancer tissues and colon cancer liver metastasis. Several feedback loops have been reported in cancers [39,61]. The feedback loop amplifies the effects of the participants and drives the progressively oncogenic and metastatic processes.

Conclusion
Here, we found lncRNA LINC01578 was correlated with colon cancer metastasis and poor prognosis. LINC01578 enhances colon cancer cell viability, mobility, and in vivo metastasis through forming a positive feedback loop with NF-jB/YY1. LINC01578 represents a valuable prognostic biomarker and therapeutic target for colon cancer metastasis.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. The expression and characteristics of LINC01578 in colon cancer. Fig. S2. NF-jB and YY1 activated LINC01578 expression in HT-29 cells. Fig. S3. Overexpression of LINC01578 enhanced DLD-1 cell viability and mobility. Fig. S4. Depletion of LINC01578 repressed HT-29 cell viability and mobility. Fig. S5. LINC01578 repressed IjBb and activated YY1. Table S1. The list of genes correlated with overall survival of colon cancer. Table S2. Univariate and multivariate survival analysis of patients with LINC01578 low and high expression.